File etc/gerrit.config

The optional file '$site_path'/etc/gerrit.config is a Git-style config file that controls many host specific settings for Gerrit.

Note
The contents of the etc/gerrit.config file are cached at startup by Gerrit. For most properties, if they are modified in this file, Gerrit needs to be restarted before it will use the new values. Some properties support being reloaded without restart.

Sample etc/gerrit.config:

[core]
  packedGitLimit = 200 m

[cache]
  directory = /var/cache/gerrit

Reload etc/gerrit.config

Some properties support being reloaded without restart when a reload config command is issued through SSH or the REST API. If a property supports this it is specified in the documentation for the property below.

Section accountPatchReviewDb

The AccountPatchReviewDb is a database used to store the user file reviewed flags.

accountPatchReviewDb.url

The url of accountPatchReviewDb. Supported types are CLOUDSPANNER, H2, POSTGRESQL, MARIADB, and MYSQL. Drop the driver jar in the lib folder of the site path if the Jdbc driver of the corresponding Database is not yet in the class path.

Default is to create H2 database in the db folder of the site path.

Changing this parameter requires to migrate database using the MigrateAccountPatchReviewDb program. Migration cannot be done while the server is running.

Also note that the db_name has to be a new db and not reusing an old ReviewDb database from a former 2.x site, otherwise gerrit’s init will remove the table.

[accountPatchReviewDb]
  url = jdbc:postgresql://<host>:<port>/<db_name>?user=<user>&password=<password>
accountPatchReviewDb.poolLimit

Maximum number of open database connections. If the server needs more than this number, request processing threads will wait up to poolMaxWait seconds for a connection to be released before they abort with an exception. This limit must be several units higher than the total number of httpd and sshd threads as some request processing code paths may need multiple connections.

Default is sshd.threads + httpd.maxThreads + 2.

database.poolMinIdle

Minimum number of connections to keep idle in the pool. Default is 4.

accountPatchReviewDb.poolMaxIdle

Maximum number of connections to keep idle in the pool. If there are more idle connections, connections will be closed instead of being returned back to the pool. Default is min(accountPatchReviewDb.poolLimit, 16).

accountPatchReviewDb.poolMaxWait

Maximum amount of time a request processing thread will wait to acquire a database connection from the pool. If no connection is released within this time period, the processing thread will abort its current operations and return an error to the client. Values should use common unit suffixes to express their setting:

  • ms, milliseconds

  • s, sec, second, seconds

  • m, min, minute, minutes

  • h, hr, hour, hours

If a unit suffix is not specified, milliseconds is assumed. Default is 30 seconds.

Section accounts

accounts.visibility

Controls visibility of other users' dashboard pages and completion suggestions to web users.

If ALL, all users are visible to all other users, even anonymous users.

If SAME_GROUP, only users who are also members of a group the current user is a member of are visible.

If VISIBLE_GROUP, only users who are members of at least one group that is visible to the current user are visible.

If NONE, no users other than the current user are visible.

Default is ALL.

accounts.defaultDisplayName

If a user account does not have a display name set, which is the normal case, then this configuration value chooses the strategy how to choose the display name. Note that this strategy is not applied by the backend. If the AccountInfo has the display name unset, then the client has to apply this strategy.

If FULL_NAME, then the (full) name of the user is chosen from AccountInfo.

If FIRST_NAME, then the first word (i.e. everything until first whitespace character) of the (full) name of the user is chosen from AccountInfo.

If USERNAME, then the username of the user is chosen from AccountInfo. If that is not set, then the (full) name will be used.

Default is FULL_NAME.

Section addreviewer

addreviewer.maxWithoutConfirmation

The maximum number of reviewers a user can add at once by adding a group as reviewer without being asked to confirm the operation.

If set to 0, the user will never be asked to confirm adding a group as reviewer.

Default is 10.

This setting only applies for adding reviewers in the Gerrit Web UI, but is ignored when adding reviewers with the set-reviewers command.

This value supports configuration reloads.

addreviewer.maxAllowed

The maximum number of reviewers a user can add at once by adding a group as reviewer.

If set to 0, there is no limit for the number of reviewers that can be added at once by adding a group as reviewer.

Default is 20.

This value supports configuration reloads.

addReviewer.baseWeight

The weight that will be applied in the default reviewer ranking algorithm. This can be increased or decreased to give more or less influence to plugins. If set to zero, the base ranking will not have any effect. Reviewers will then be ordered as ranked by the plugins (if there are any).

By default 1.

Section auth

See also SSO configuration.

auth.type

Type of user authentication employed by Gerrit. The supported values are:

  • OpenID

    The default setting. Gerrit uses any valid OpenID provider chosen by the end-user. For more information see openid.net.

  • OpenID_SSO

    Supports OpenID from a single provider. There is no registration link, and the "Sign In" link sends the user directly to the provider’s SSO entry point.

  • HTTP

    Gerrit relies upon data presented in the HTTP request. This includes HTTP basic authentication, or some types of commercial single-sign-on solutions. With this setting enabled the authentication must take place in the web server or servlet container, and not from within Gerrit.

  • HTTP_LDAP

    Exactly like HTTP (above), but additionally Gerrit pre-populates a user’s full name and email address based on information obtained from the user’s account object in LDAP. The user’s group membership is also pulled from LDAP, making any LDAP groups that a user is a member of available as groups in Gerrit. Hence the _LDAP suffix in the name of this authentication type. Gerrit does NOT authenticate the user via LDAP.

  • CLIENT_SSL_CERT_LDAP

    This authentication type is actually kind of SSO. Gerrit will configure Jetty’s SSL channel to request the client’s SSL certificate. For this authentication to work a Gerrit administrator has to import the root certificate of the trust chain used to issue the client’s certificate into the <review-site>/etc/keystore. After the authentication is done Gerrit will obtain basic user registration (name and email) from LDAP, and some group memberships. Hence the _LDAP suffix in the name of this authentication type. Gerrit does NOT authenticate the user via LDAP. This authentication type can only be used under hosted daemon mode, and the httpd.listenUrl must use https:// as the protocol. Optionally, certificate revocation list file can be used at <review-site>/etc/crl.pem. For details, see httpd.sslCrl.

  • LDAP

    Gerrit prompts the user to enter a username and a password, which it then verifies by performing a simple bind against the configured ldap.server. In this configuration the web server is not involved in the user authentication process.

    The actual username used in the LDAP simple bind request is the account’s full DN, which is discovered by first querying the directory using either an anonymous request, or the configured ldap.username identity. Gerrit can also use kerberos if ldap.authentication is set to GSSAPI.

    If auth.gitBasicAuthPolicy is set to HTTP, the randomly generated HTTP password is used for authentication. On the other hand, if auth.gitBasicAuthPolicy is set to HTTP_LDAP, the password in the request is first checked against the HTTP password and, if it does not match, it is then validated against the LDAP password. Service users that are internal-only are authenticated by their HTTP passwords.

  • LDAP_BIND

    Gerrit prompts the user to enter a username and a password, which it then verifies by performing a simple bind against the configured ldap.server. In this configuration the web server is not involved in the user authentication process.

    Unlike LDAP above, the username used to perform the LDAP simple bind request is the exact string supplied in the dialog by the user. The configured ldap.username identity is not used to obtain account information.

  • OAUTH

    OAuth is a protocol that lets external apps request authorization to private details in a user’s account without getting their password. This is preferred over Basic Authentication because tokens can be limited to specific types of data, and can be revoked by users at any time.

    Site owners have to register their application before getting started. Note that provider specific plugins must be used with this authentication scheme.

    Git clients may send OAuth 2 access tokens instead of passwords in the Basic authentication header. Note that provider specific plugins must be installed to facilitate this authentication scheme. If multiple OAuth 2 provider plugins are installed one of them must be selected as default with the auth.gitOAuthProvider option.

  • DEVELOPMENT_BECOME_ANY_ACCOUNT

    DO NOT USE. Only for use in a development environment.

    When this is the configured authentication method a hyperlink titled Become appears in the top right corner of the page, taking the user to a form where they can enter the username of any existing user account, and immediately login as that account, without any authentication taking place.

By default, OpenID.

auth.allowedOpenID

List of permitted OpenID providers. A user may only authenticate with an OpenID that matches this list. Only used if auth.type is set to OpenID (the default).

Patterns may be either a standard Java regular expression (java.util.regex) (start with ^ and end with $) or be a simple prefix (any other string).

By default, the list contains two values, http:// and https://, allowing users to authenticate with any OpenID provider.

auth.trustedOpenID

List of trusted OpenID providers. Only used if auth.type is set to OpenID (the default).

In order for a user to take advantage of permissions beyond those granted to the Anonymous Users and Registered Users groups, the user account must only have OpenIDs which match at least one pattern from this list.

Patterns may be either a standard Java regular expression (java.util.regex) (start with ^ and end with $) or be a simple prefix (any other string).

By default, the list contains two values, http:// and https://, allowing Gerrit to trust any OpenID it receives.

auth.openIdDomain

List of allowed OpenID email address domains. Only used if auth.type is set to OPENID or OPENID_SSO.

Domain is case insensitive and must be in the same form as it appears in the email address, for example, "example.com".

By default, any domain is accepted.

auth.maxOpenIdSessionAge

Time in seconds before an OpenID provider must force the user to authenticate themselves again before authentication to this Gerrit server. Currently this is only a polite request, and users coming from providers that don’t support the PAPE extension will be accepted anyway. In the future it may be enforced, rejecting users coming from providers that don’t honor the max session age.

If set to 0, the provider will always force the user to authenticate (e.g. supply their password). Values should use common unit suffixes to express their setting:

  • s, sec, second, seconds

  • m, min, minute, minutes

  • h, hr, hour, hours

  • d, day, days

  • w, week, weeks (1 week is treated as 7 days)

  • mon, month, months (1 month is treated as 30 days)

  • y, year, years (1 year is treated as 365 days)

Default is -1, permitting infinite time between authentications.

auth.registerEmailPrivateKey

Private key to use when generating an email verification token.

If not set, a random key is generated when running the site initialization.

auth.maxRegisterEmailTokenAge

Time in seconds before an email verification token sent to a user in order to validate their email address expires.

  • s, sec, second, seconds

  • m, min, minute, minutes

  • h, hr, hour, hours

  • d, day, days

  • w, week, weeks (1 week is treated as 7 days)

  • mon, month, months (1 month is treated as 30 days)

  • y, year, years (1 year is treated as 365 days)

Default is 12 hours.

auth.openIdSsoUrl

The SSO entry point URL. Only used if auth.type is set to OpenID_SSO.

The "Sign In" link will send users directly to this URL.

auth.httpHeader

HTTP header to trust the username from, or unset to select HTTP basic authentication. Only used if auth.type is set to HTTP.

auth.httpDisplaynameHeader

HTTP header to retrieve the user’s display name from. Only used if auth.type is set to HTTP.

If set, Gerrit trusts and enforces the user’s full name using the HTTP header and disables the ability to manually modify the user’s full name from the contact information page.

auth.httpEmailHeader

HTTP header to retrieve the user’s e-mail from. Only used if auth.type is set to HTTP.

If set, Gerrit trusts and enforces the user’s e-mail using the HTTP header and disables the ability to manually modify or register other e-mails from the contact information page.

auth.httpExternalIdHeader

HTTP header to retrieve the user’s external identification token. Only used if auth.type is set to HTTP.

If set, Gerrit adds the value contained in the HTTP header to the user’s identity. Typical use is with a federated identity token from an external system (e.g. GitHub OAuth 2.0 authentication) where the user’s auth token exchanged during authentication handshake needs to be used for authenticated communication to the external system later on.

Example: auth.httpExternalIdHeader: X-GitHub-OTP

auth.loginUrl

URL to redirect a browser to after the end-user has clicked on the login link in the upper right corner. Only used if auth.type is set to HTTP or HTTP_LDAP. Organizations using an enterprise single-sign-on solution may want to redirect the browser to the SSO product’s sign-in page for completing the login process and validate their credentials.

If set, Gerrit allows anonymous access until the end-user performs the login and provides a trusted identity through the HTTP header. If not set, Gerrit requires the HTTP header with a trusted identity and returns the error page 'LoginRedirect.html' if such a header is not present.

auth.loginText

Text displayed in the loginUrl link. Only used if auth.loginUrl is set.

If not set, the "Sign In" text is used.

auth.registerPageUrl

URL of the registration page to use when a new user logs in to Gerrit for the first time. Used only when auth.type is set to HTTP.

If not set, the standard Gerrit registration page /#/register/ is displayed.

auth.logoutUrl

URL to redirect a browser to after the end-user has clicked on the "Sign Out" link in the upper right corner. Organizations using an enterprise single-sign-on solution may want to redirect the browser to the SSO product’s sign-out page.

If not set, the redirect returns to the list of all open changes.

auth.registerUrl

Target for the "Register" link in the upper right corner. Used only when auth.type is LDAP, LDAP_BIND or CUSTOM_EXTENSION.

If not set, no "Register" link is displayed.

auth.registerText

Text for the "Register" link in the upper right corner. Used only when auth.type is LDAP, LDAP_BIND or CUSTOM_EXTENSION.

If not set, defaults to "Register".

auth.editFullNameUrl

Target for the "Edit" button when the user is allowed to edit their full name. Used only when auth.type is LDAP, LDAP_BIND or CUSTOM_EXTENSION.

auth.httpPasswordUrl

Target for the "Obtain Password" link. Used only when auth.type is CUSTOM_EXTENSION.

auth.switchAccountUrl

URL to switch user identities and login as a different account than the currently active account. This is disabled by default except when auth.type is OPENID and DEVELOPMENT_BECOME_ANY_ACCOUNT. If set the "Switch Account" link is displayed next to "Sign Out".

When auth.type does not normally enable this URL administrators may set this to login/, allowing users to begin a new web session. This value is used as an href in the Gerrit web app, so absolute URLs like https://someotherhost/login work as well.

If a ${path} parameter is included, then the Gerrit web app will substitute the currently viewed path in the link. Be aware that this path will include a leading slash, so a value like this might be appropriate: /login${path}.

auth.cookiePath

Sets "path" attribute of the authentication cookie.

If not set, HTTP request’s path is used.

auth.cookieDomain

Sets "domain" attribute of the authentication cookie.

If not set, HTTP request’s domain is used.

auth.cookieSecure

Sets "secure" flag of the authentication cookie. If true, cookies will be transmitted only over HTTPS protocol.

By default, false.

auth.emailFormat

Optional format string to construct user email addresses out of user login names. Only used if auth.type is HTTP, HTTP_LDAP or LDAP.

This value can be set to a format string, where {0} is replaced with the login name. E.g. "{0}+gerrit@example.com" with a user login name of "foo" will produce "foo+gerrit@example.com" during the first time user "foo" registers.

If the site is using HTTP_LDAP or LDAP, using this option is discouraged. Setting ldap.accountEmailAddress and importing the email address from the LDAP directory is generally preferred.

auth.contributorAgreements

Controls whether or not the contributor agreement features are enabled for the Gerrit site. If enabled a user must complete a contributor agreement before they can upload changes.

If enabled, the admin must also add one or more contributor-agreement sections in project.config and create agreement files under '$site_path'/static, so users can actually complete one or more agreements.

By default this is false (no agreements are used).

To enable the actual usage of contributor agreement the project specific config option in the project.config must be set: receive.requireContributorAgreement.

auth.trustContainerAuth

If true then it is the responsibility of the container hosting Gerrit to authenticate users. In this case Gerrit will blindly trust the container.

This parameter only affects git over http traffic. If set to false then Gerrit will do the authentication (using Basic authentication).

By default this is set to false.

auth.gitBasicAuthPolicy

When auth.type is LDAP, LDAP_BIND or OAUTH, it allows using either the generated HTTP password, the LDAP or OAUTH password, or a combination of HTTP and LDAP authentication, to authenticate Git over HTTP and REST API requests. The supported values are:

*HTTP

Only the HTTP password is accepted when doing Git over HTTP and REST API requests.

*LDAP

Only the LDAP password is allowed when doing Git over HTTP and REST API requests.

*OAUTH

Only the OAUTH authentication is allowed when doing Git over HTTP and REST API requests.

*HTTP_LDAP

The password in the request is first checked against the HTTP password and, if it does not match, it is then validated against the LDAP password.

By default this is set to LDAP when auth.type is LDAP and OAUTH when auth.type is OAUTH. Otherwise, the default value is HTTP.

When gitBasicAuthPolicy is set to LDAP or HTTP_LDAP and the user is authenticating with the LDAP username/password, the Git client config needs to have http.cookieFile set to a local file, otherwise every single call would trigger a full LDAP authentication and groups resolution which could introduce a noticeable latency on the overall execution and produce unwanted load to the LDAP server.

auth.gitOAuthProvider

Selects the OAuth 2 provider to authenticate git over HTTP traffic with.

In general there is no way to determine from an access token alone, which OAuth 2 provider to address to verify that token, and the BasicAuth scheme does not support amending such details. If multiple OAuth provider plugins in a system offer support for git over HTTP authentication site administrators must configure, which one to use as default provider. In case the provider cannot be determined from a request the access token will be sent to the default provider for verification.

The value of this parameter must be the identifier of an OAuth 2 provider in the form plugin-name:provider-name. Consult the respective plugin documentation for details.

auth.userNameToLowerCase

If set the username that is received to authenticate a git operation is converted to lower case for looking up the user account in Gerrit.

By setting this parameter a case insensitive authentication for the git operations can be achieved, if it is ensured that the usernames in Gerrit (scheme username) are stored in lower case (e.g. if the parameter ldap.accountSshUserName is set to ${sAMAccountName.toLowerCase}). It is important that for all existing accounts this username is already in lower case. It is not possible to convert the usernames of the existing accounts to lower case because this would break the access to existing per-user branches and Gerrit provides no tool to do such a conversion. Accounts created using the REST API or the create-account SSH command will be created with all lowercase characters, when this option is set.

Setting this parameter to true will prevent all users from login that have a non-lower-case username.

This parameter only affects git over http and git over SSH traffic.

By default this is set to false.

auth.userNameCaseInsensitive

If set the username will be handled case insensitively but case preserving, i.e. a user can login with johndoe or JohnDoe for the same account created for JohnDoe. The form of the username used during account creation will be used wherever the username is displayed. Sandbox branches created for a user can also only be created for this original form.

Note, that this does not work for all existing accounts, if they were not originally created with all lowercase, since the note keys of the external IDs will not match the new scheme. For more details refer to the External ID documentation.

Gerrit provides the offline and the online online tools to migrate existing accounts to match the new scheme.

Naturally, if there were two accounts only different in capitalization, e.g. johndoe and JohnDoe, the account JohnDoe will not be able to authenticate anymore after setting this option. If such duplicate accounts exist the migration tool will fail, since the newly computed note name would be identical and thus conflict. These duplicates thus have to be deleted manually by deleting the respective external ID.

For newly initialized sites this option defaults to true.

Default is false.

auth.userNameCaseInsensitiveMigrationMode

Setting migration mode to true allows to fallback to case sensitive behaviour if the migrated external ID cannot be found. This allows to trigger the migration while Gerrit process is running.

Default is false.

auth.enableRunAs

If true HTTP REST APIs will accept the X-Gerrit-RunAs HTTP request header from any users granted the Run As capability. The header and capability permit the authenticated user to impersonate another account.

If false the feature is disabled and cannot be re-enabled without editing gerrit.config and restarting the server.

Default is true.

auth.allowRegisterNewEmail

Whether users are allowed to register new email addresses.

In addition for the HTTP authentication type auth.httpemailheader must not be set to enable registration of new email addresses.

By default, true.

auth.autoUpdateAccountActiveStatus

Whether to allow automatic synchronization of an account’s inactive flag upon login.

If set to true, upon login, if the authentication back-end reports the account as active, the account’s inactive flag in NoteDb will be updated to be active.

If the authentication back-end reports the account as inactive, the account’s flag will be updated to be inactive and the login attempt will be blocked. Users enabling this feature should ensure that their authentication back-end is supported. Currently, only strict 'LDAP' authentication is supported.

In addition, if this parameter is not set, or false, the corresponding scheduled task to deactivate inactive Gerrit accounts will also be disabled. If this parameter is set to true, users should also consider configuring the accountDeactivation section appropriately.

By default, false.

auth.skipFullRefEvaluationIfAllRefsAreVisible

Whether to skip the full ref visibility checks as a performance shortcut when a user has READ permission for all refs.

The full ref filtering would filter out refs for pending edits, private changes and auto merge commits.

By default, true.

Section cache

cache.threads

Number of threads to use when running asynchronous cache tasks. The threads executor is delegated to when sending removal notifications to listeners, when asynchronous computations like refresh, refreshAfterWrite are performed, or when performing periodic maintenance.

NOTE: Setting it to 0 disables the dedicated thread pool and indexing will be done in the same thread as the operation. This may result in evictions taking longer because the listeners are executed in the caller’s thread.

By default, the JVM common ForkJoinPool is used.

cache.directory

Path to a local directory where Gerrit can write cached entities for future lookup. This local disk cache is used to retain potentially expensive to compute information across restarts. If the location does not exist, Gerrit will try to create it.

Technically, cached entities are persisted as a set of H2 databases inside this directory.

If not absolute, the path is resolved relative to $site_path.

Default is unset, no disk cache.

cache.enableDiskStatMetrics

Whether to enable the computation of disk statistics of persistent caches. This computation is expensive and requires a long time on larger installations.

By default, false.

cache.h2CacheSize

The size of the in-memory cache for each opened H2 cache database, in bytes.

Some caches of Gerrit are persistent and are backed by an H2 database. H2 uses memory to cache its database content. The parameter h2CacheSize allows to limit the memory used by H2 and thus prevent out-of-memory caused by the H2 database using too much memory.

Technically the H2 cache size is configured using the CACHE_SIZE parameter in the H2 JDBC connection URL, as described here

Default is unset, using up to half of the available memory.

H2 will persist this value in the database, so to unset explicitly specify 0.

Common unit suffixes of 'k', 'm', or 'g' are supported.

cache.h2AutoServer

If set to true, enable H2 autoserver mode for the H2-backed persistent cache databases.

See here for detail.

Default is false.

cache.openFiles

The number of file descriptors to add to the limit set by the Gerrit daemon.

Persistent caches are stored on the file system and as such participate in the file descriptors utilization. The number of file descriptors can vary depending on the cache configuration and the specific backend used.

The additional file descriptors required by the cache should be accounted for via this setting, so that the Gerrit daemon can adjust the ulimit accordingly.

If you increase this to a larger setting you may need to also adjust the ulimit on file descriptors for the host JVM, as Gerrit needs additional file descriptors available for network sockets and other repository data manipulation.

Default is 0.

cache.<name>.maxAge

Maximum age to keep an entry in the cache. Values should use common unit suffixes to express their setting:

  • s, sec, second, seconds

  • m, min, minute, minutes

  • h, hr, hour, hours

  • d, day, days

  • w, week, weeks (1 week is treated as 7 days)

  • mon, month, months (1 month is treated as 30 days)

  • y, year, years (1 year is treated as 365 days)

If a unit suffix is not specified, seconds is assumed. If 0 is supplied, the maximum age is infinite and items are never purged except when the cache is full.

Default is 0, meaning store forever with no expire, except:

  • "adv_bases": default is 10 minutes

  • "ldap_groups": default is 1 hour

  • "web_sessions": default is 12 hours

cache.<name>.memoryLimit

The total cost of entries to retain in memory. The cost computation varies by the cache. For most caches where the in-memory size of each entry is relatively the same, memoryLimit is currently defined to be the number of entries held by the cache (each entry costs 1).

For caches where the size of an entry can vary significantly between individual entries (notably "git_modified_files", "modified_files", "git_file_diff", "gerrit_file_diff", "diff_intraline"), memoryLimit is an approximation of the total number of bytes stored by the cache. Larger entries that represent bigger patch sets or longer source files will consume a bigger portion of the memoryLimit. For these caches the memoryLimit should be set to roughly the amount of RAM (in bytes) the administrator can dedicate to the cache.

Default is 1024 for most caches, except:

  • "adv_bases": default is 4096

  • "git_modified_files": default is 10m (10 MiB of memory)

  • "modified_files": default is 10m (10 MiB of memory)

  • "git_file_diff": default is 10m (10 MiB of memory)

  • "gerrit_file_diff": default is 10m (10 MiB of memory)

  • "diff_intraline": default is 10m (10 MiB of memory)

  • "diff_summary": default is 10m (10 MiB of memory)

  • "external_ids_map": default is 2 and should not be changed

  • "groups": default is unlimited

  • "groups_byname": default is unlimited

  • "groups_byuuid": default is unlimited

  • "groups_byuuid_persisted": default is 1g (1 GiB of disk space)

  • "plugin_resources": default is 2m (2 MiB of memory)

If set to 0 the cache is disabled; entries are loaded but not stored in-memory.

+ NOTE: When the cache is disabled, there is no locking when accessing the same key/value, and therefore multiple threads may load the same value concurrently with a higher memory footprint. To keep a minimum caching and avoid concurrent loading of the same key/value, set memoryLimit to 1 and maxAge to 1.

cache.<name>.expireFromMemoryAfterAccess

Time after last access to automatically expire entries from an in-memory cache. If 0 or not specified, entries are never expired in this manner. Values may use unit suffixes as in maxAge.

This option only applies to in-memory caches; persistent cache values are not expired in this manner, and are only pruned via diskLimit.

cache.<name>.diskLimit

Total size in bytes of the keys and values stored on disk. Caches that have grown bigger than this size are scanned daily at 1 AM local server time to trim the cache. Entries are removed in least recently accessed order until the cache fits within this limit. Caches may grow larger than this during the day, as the size check is only performed once every 24 hours.

Default is 128 MiB per cache, except:

  • "change_notes": disk storage is disabled by default

  • "diff_summary": default is 1g (1 GiB of disk space)

  • "external_ids_map": disk storage is disabled by default

  • "persisted_projects": default is 1g (1 GiB of disk space)

If 0 or negative, disk storage for the cache is disabled.

cache.<name>.refreshAfterWrite

Duration after which we asynchronously refresh the cached value.

Values should use common unit suffixes to express their setting:

  • ms, milliseconds

  • s, sec, second, seconds

  • m, min, minute, minutes

  • h, hr, hour, hours

    This applies only to these caches that support refreshing:

  • "projects": Caching project information in-memory. Defaults to 15 minutes.

cache.refreshThreadPoolSize

Number of threads that are available to refresh cached values that became out of date. This applies only to these caches that support refreshing:

  • "projects": Caching project information in-memory

    Refreshes will only be scheduled on this executor if the values are out of sync. The check if they are is cheap and always happens on the thread that inquires for a cached value.

    Defaults to 2.

Standard Caches

cache "accounts"

Cache entries contain important details of an active user, including their display name, preferences, and known email addresses. Entry information is obtained from NoteDb data in the All-Users repo.

If direct updates are made to All-Users, this cache should be flushed.

cache "adv_bases"

Used only for push over smart HTTP when branch level access controls are enabled. The cache entry contains all commits that are available for the client to use as potential delta bases. Push over smart HTTP requires two HTTP requests, and this cache tries to carry state from the first request into the second to ensure it can complete.

cache "default_preferences"

Caches the server’s default general, edit and diff preferences.

Default value is 1 to hold only the most current version in-memory.

cache "changes"

The size of memoryLimit determines the number of projects for which all changes will be cached. If the cache is set to 1024, this means all changes for up to 1024 projects can be held in the cache.

Default value is 0 (disabled). It is disabled by default due to the fact that change updates are not communicated between Gerrit servers. Hence this cache should be disabled in a cluster setup using multiple primary or multiple replica nodes.

The cache should be flushed whenever NoteDb change metadata in a repository is modified outside of Gerrit.

cache "changes_by_project"

Ideally, the memorylimit of this cache is large enough to cover all projects. This should significantly speed up change ref advertisements and git pushes, especially for projects with lots of changes, and particularly on replicas where there is no index.

cache "git_modified_files"

Each item caches the list of git modified files between two git trees corresponding to two different commits. This cache does not read the actual file contents nor does it include the edits (modified regions) of the files.

cache "modified_files"

Each item caches the list of modified files between two commits. This cache is similar to the git_modified_files cache but performs extra logic including filtering out files that are untouched by both commits because they were purely modified between the parent commits.

cache "git_file_diff"

Each item caches the pure git diff between two git trees for a specific file path. The diff includes all the file attributes (old/new paths, change/patch types) as well as the list of edits corresponding to the modified regions in the file.

cache "gerrit_file_diff"

Each item caches the diff between two git commits for a specific file path. This cache is similar to the git_file_diff cache but performs extra logic including identifying the edits that are due to rebase. The diff for the "commit message" and "merge list" can also be requested from this cache.

Entries in this cache are relatively large, so memoryLimit is an estimate in bytes of memory used. Administrators should try to target cache.diff.memoryLimit to fit all changes users will view in a 1 or 2 day span. The same applies for other diff caches: "git_modified_files", "modified_files" and "git_file_diff".

cache "diff_intraline"

Each item caches the intraline difference of one file, when compared between two commits. Gerrit uses this cache to accelerate display of intraline differences when viewing a file.

Entries in this cache are relatively large, so memoryLimit is an estimate in bytes of memory used. Administrators should try to target cache.diff.memoryLimit to fit all files users will view in a 1 or 2 day span.

cache "diff_summary"

Each item caches list of file paths which are different between two commits. Gerrit uses this cache to accelerate computing of the list of paths of changed files.

Ideally, disk limit of this cache is large enough to cover all changes. This should significantly speed up change reindexing, especially full offline reindexing.

cache "external_ids_map"

A singleton cache whose sole entry is a map of the parsed representation of all current external IDs. The cache may temporarily contain 2 entries, but the second one is promptly expired.

It is not recommended to change the in-memory attributes of this cache away from the defaults. The cache may be persisted by setting diskLimit, which is only recommended if cold start performance is problematic.

cache "git_tags"

If branch or reference level READ access controls are used, this cache tracks which tags are reachable from the branch tips of a repository. Gerrit uses this information to determine the set of tags that a client may access, derived from which tags are part of the history of a visible branch.

The cache is persisted to disk across server restarts as it can be expensive to compute (60 or more seconds for a large history like the Linux kernel repository).

cache "comment_context"

Caches the context lines of comments, which are the lines of the source file highlighted by the user when the comment was written.

cache "groups"

Caches the basic group information of internal groups by group ID, including the group owner, name, and description.

For this cache it is important to configure a size that is larger than the number of internal Gerrit groups, otherwise general Gerrit performance may be poor. This is why by default this cache is unlimited.

External group membership obtained from LDAP is cached under "ldap_groups".

cache "groups_byname"

Caches the basic group information of internal groups by group name, including the group owner, name, and description.

For this cache it is important to configure a size that is larger than the number of internal Gerrit groups, otherwise general Gerrit performance may be poor. This is why by default this cache is unlimited.

External group membership obtained from LDAP is cached under "ldap_groups".

cache "groups_byuuid"

Caches the basic group information of internal groups by group UUID, including the group owner, name, and description.

For this cache it is important to configure a size that is larger than the number of internal Gerrit groups, otherwise general Gerrit performance may be poor. This is why by default this cache is unlimited.

External group membership obtained from LDAP is cached under "ldap_groups".

cache "groups_byuuid_persisted"

Caches the basic group information of internal groups by group UUID, including the group owner, name, and description.

This is the persisted version of groups_byuuid cache. The intention of this cache is to have an in-memory size of 0.

External group membership obtained from LDAP is cached under "ldap_groups".

cache "groups_bymember"

Caches the groups which contain a specific member (account). If direct updates are made to the account_group_members table, this cache should be flushed.

cache "groups_bysubgroups"

Caches the parent groups of a subgroup. If direct updates are made to the account_group_includes table, this cache should be flushed.

cache "groups_external"

Caches all the external groups available to Gerrit. The cache holds a single entry which maps to the latest available of all external groups' UUIDs. This cache uses "groups_external_persisted" to load its value.

cache "groups_external_persisted"

Caches all external groups available to Gerrit at some point in history. The cache key is representation of a specific groups state in NoteDb and the value is the list of all external groups. The cache is persisted to enhance performance.

cache "ldap_groups"

Caches the LDAP groups that a user belongs to, if LDAP has been configured on this server. This cache should be configured with a low maxAge setting, to ensure LDAP modifications are picked up in a timely fashion.

cache "ldap_groups_byinclude"

Caches the hierarchical structure of LDAP groups.

cache "ldap_usernames"

Caches a mapping of LDAP username to Gerrit account identity. The cache automatically updates when a user first creates their account within Gerrit, so the cache expire time is largely irrelevant.

cache "permission_sort"

Caches the order in which access control sections must be applied to a reference. Sorting the sections can be expensive when regular expressions are used, so this cache remembers the ordering for each branch.

cache "plugin_resources"

Caches formatted plugin resources, such as plugin documentation that has been converted from Markdown to HTML. The memoryLimit refers to the bytes of memory dedicated to storing the documentation.

cache "persisted_projects"

Caches the project description records, from the refs/meta/config branch of each project. This is the persisted variant of the projects cache. The intention is for this cache to have an in-memory size of 0.

cache "projects"

Caches the project description records, from the refs/meta/config branch of each project. If a project record is updated or deleted, this cache should be flushed. Newly inserted projects do not require a cache flush, as they will be read upon first reference.

Note
This cache should be disabled or set with a low refreshAfterWrite in a cluster setup using multiple primary or multiple replica nodes.
cache "prolog_rules"

Caches parsed rules.pl contents for each project. This cache uses the same size as the projects cache when cache.prolog_rules.memoryLimit is not set.

cache "pure_revert"

Result of checking if one change or commit is a pure/clean revert of another.

cache "soy_sauce_compiled_templates"

Caches compiled soy templates. Stores at most only one key-value pair with a constant key value and the value is a compiled SoySauce templates. The value is reloaded automatically every few seconds if there are reads from the cache. If cache is not used for 1 minute, the item is removed (i.e. emails can be send with templates which are max 1 minute old).

cache "sshkeys"

Caches unpacked versions of user SSH keys, so the internal SSH daemon can match against them during authentication. The unit of storage is per-user, so 1024 items translates to 1024 unique user accounts. As each individual user account may configure multiple SSH keys, the total number of keys may be larger than the item count.

Note
This cache should be disabled or set with a low refreshAfterWrite in a cluster setup using multiple primary or multiple replica nodes.
cache "web_sessions"

Tracks the live user sessions coming in over HTTP. Flushing this cache would cause all users to be signed out immediately, forcing them to sign-in again. To avoid breaking active users, this cache is not flushed automatically by gerrit flush-caches --all, but instead must be explicitly requested.

If no disk cache is configured (or cache.web_sessions.diskLimit is set to 0) a server restart will force all users to sign-out, and need to sign-in again after the restart, as the cache was unable to persist the session information. Enabling a disk cache is strongly recommended.

Session storage is relatively inexpensive. The average entry in this cache is approximately 346 bytes.

The maxAge configuration is also used for as maximum lifetime of the HTTP servlet container session.

Cache Options

cache.git_file_diff.timeout

Maximum number of milliseconds to wait for git diff data before giving up and falling back on a simpler diff algorithm that will not be able to break down modified regions into smaller ones. This is a work around for an infinite loop bug in the default difference algorithm implementation.

Values should use common unit suffixes to express their setting:

  • ms, milliseconds

  • s, sec, second, seconds

  • m, min, minute, minutes

  • h, hr, hour, hours

If a unit suffix is not specified, milliseconds is assumed.

Default is 5 seconds.

cache.diff_intraline.timeout

Maximum number of milliseconds to wait for intraline difference data before giving up and disabling it for a particular file pair. This is a work around for an infinite loop bug in the intraline difference implementation.

If computation takes longer than the timeout, the worker thread is terminated, an error message is shown, and no intraline difference is displayed for the file pair.

Values should use common unit suffixes to express their setting:

  • ms, milliseconds

  • s, sec, second, seconds

  • m, min, minute, minutes

  • h, hr, hour, hours

If a unit suffix is not specified, milliseconds is assumed.

Default is 5 seconds.

cache.diff_intraline.enabled

Boolean to enable or disable the computation of intraline differences when populating a diff cache entry. This flag is provided primarily as a backdoor to disable the intraline difference feature if necessary. To maintain backwards compatibility with prior versions, this setting will fallback to cache.diff.intraline if not set in the configuration.

Default is true, enabled.

cache.projects.loadOnStartup

If the project cache should be loaded during server startup.

The cache is loaded concurrently. Admins should ensure that the cache size set under cache.projects.memoryLimit is not smaller than the number of repos.

Default is false, disabled.

cache.projects.loadThreads

Only relevant if cache.projects.loadOnStartup is true.

The number of threads to allocate for loading the cache at startup. These threads will die out after the cache is loaded.

Default is the number of CPUs.

cache.project_list.interval

The interval for running the project_list cache warmer.

By default, if cache.project_list.maxAge is set, interval will be set to half its value. If cache.project_list.maxAge is not set or interval is set to -1, it is disabled.

cache.project_list.startTime

The start time for running the project_list cache warmer.

Default is 00:00 if the project_list cache warmer is enabled.

Section capability

capability.administrateServer

Names of groups of users that are allowed to exercise the administrateServer capability, in addition to those listed in All-Projects. Configuring this option can be a useful fail-safe to recover a server in the event an administrator removed all groups from the administrateServer capability, or to ensure that specific groups always have administration capabilities.

[capability]
  administrateServer = group Fail Safe Admins

The configuration file uses group names, not UUIDs. If a group is renamed the gerrit.config file must be updated to reflect the new name. If a group cannot be found for the configured name a warning is logged and the server will continue normal startup.

If not specified (default), only the groups listed by All-Projects may use the administrateServer capability.

capability.makeFirstUserAdmin

Whether the first user that logs in to the Gerrit server should automatically be added to the administrator group and hence get the administrateServer capability assigned. This is useful to bootstrap the account data.

Default is true.

Section change

change.allowBlame

Allow blame on side by side diff. If set to false, blame cannot be used.

Default is true.

change.cacheAutomerge

When reviewing merge commits, the left-hand side shows the output of the result of JGit’s automatic merge algorithm. This option controls whether this output is cached in the change repository, or if only the diff is cached in the persistent diff caches ("git_modified_files", modified_files, "git_file_diff", "file_diff").

If true, automerge results are stored in the repository under refs/cache-automerge/*; the results of diffing the change against its automerge base are stored in the diff caches. If false, no extra data is stored in the repository, only the diff caches. This can result in slight performance improvements by reducing the number of refs in the repo.

Default is true.

change.commentSizeLimit

Maximum allowed size in characters of a regular (non-robot) comment. Comments which exceed this size will be rejected. Size computation is approximate and may be off by roughly 1%. Common unit suffixes of 'k', 'm', or 'g' are supported. The value must be positive.

The default limit is 16kiB.

change.cumulativeCommentSizeLimit

Maximum allowed size in characters of all comments (including robot comments) and change messages. Size computation is approximate and may be off by roughly 1%. Common unit suffixes of 'k', 'm', or 'g' are supported.

The default limit is 3MiB.

change.disablePrivateChanges

If set to true, users are not allowed to create private changes.

The default is false.

change.maxComments

Maximum number of comments (regular plus robot) allowed per change. Additional comments are rejected.

By default 5,000.

change.maxFiles

Maximum number of files allowed per change. Larger changes are rejected and must be split up.

By default 100,000.

change.maxPatchSets

Maximum number of patch sets allowed per change. If this is insufficient, recreate the change with a new Change-Id, then abandon the old change.

By default 1000.

change.maxUpdates

Maximum number of updates to a change. Counts only updates to the main NoteDb meta ref; draft comments, robot comments, stars, etc. do not count towards the total.

Many NoteDb operations require walking the entire change meta ref and loading its contents into memory, so changes with arbitrarily many updates may cause high CPU usage, memory pressure, persistent cache bloat, and other problems.

The following operations are allowed even when a change is at the limit:

  • Abandon

  • Submit

  • Submit by push with %submit

  • Auto-close by pushing directly to the branch

  • Fix with expect_merged_as

By default 1000.

change.mergeabilityComputationBehavior

This setting determines when Gerrit computes if a change is mergeable or not. This computation is expensive, especially when the repository is large or when there are many open changes.

This config can have the following states:

  • API_REF_UPDATED_AND_CHANGE_REINDEX: Gerrit indexes mergeability enabling the is:mergeable predicate in change search and allowing fast retrieval of this bit in query responses. Gerrit will always serve mergeable in ChangeInfo objects. Gerrit will reindex all open changes when the target ref advances (expensive).

  • REF_UPDATED_AND_CHANGE_REINDEX: Gerrit indexes mergeability enabling the is:mergeable predicate in change search and allowing fast retrieval of this bit in query responses. Gerrit will never serve mergeable in ChangeInfo objects. This state can be a final state for instances that only want to optimize the read path, but not the write path for speed or serve as an intermediary step for instances that want to optimize both and need to migrate callers of their API. Gerrit will reindex all open changes when the target ref advances (expensive).

  • NEVER: Gerrit does not index mergeable, so is:mergeable is disabled as query operator. Gerrit does not serve mergeable in ChangeInfo.

Note
Gerrit would only render conflict changes section on change screen if API_REF_UPDATED_AND_CHANGE_REINDEX value is set.

Default is NEVER.

change.conflictsPredicateEnabled

This setting determines when Gerrit renders conflict changes section on change screen and also supports conflicts predicate. This computation is expensive, computing ConflictsPredicate has a runtime complexity of O(nˆ2) with n number of open changes on a branch. When set to false GUI will silently ignore the error message and leave the conflict changes section on change screen empty. See also implications on rendering of conflict changes section in configuration section:change.mergeabilityComputationBehavior.

Default is true.

change.maxSubmittableAtOnce

Maximum number of changes that can be chained together in the same repository to be submitted at once.

Default is 32767.

change.move

Whether the Move Change REST endpoint is enabled.

The move change functionality has some corner cases with undesired side effects. Hence administrators may decide to disable this functionality. In particular, if a change that has dependencies on other changes is moved to a new branch, and the moved change gets submitted to the new branch, the changes on which the change depends are silently merged into the new branch, although these changes have not been moved to that branch (see details in issue 40009784).

By default true.

change.enableRobotComments

Are robot comments enabled in the Gerrit UI? This setting allows phasing out robot comments.

By default true.

change.robotCommentSizeLimit

Maximum allowed size in characters of a robot comment. Robot comments which exceed this size will be rejected on addition. Size computation is approximate and may be off by roughly 1%. Common unit suffixes of 'k', 'm', or 'g' are supported. Zero or negative values allow robot comments of unlimited size.

The default limit is 1MiB.

change.sendNewPatchsetEmails

When false, emails will not be sent to owners, reviewers, and cc for creating a new patchset unless they are project watchers or have starred the change.

Default is true.

change.showAssigneeInChangesTable

Show assignee field in changes table. If set to false, assignees will not be visible in changes table.

Default is false.

change.strictLabels

Reject invalid label votes: invalid labels or invalid values. This configuration option is provided for backwards compatibility and may be removed in future gerrit versions.

Default is false.

change.submitLabel

Label name for the submit button.

Default is "Submit".

change.submitLabelWithParents

Label name for the submit button if the change has parents which will be submitted together with this change.

Default is "Submit including parents".

change.submitTooltip

Tooltip for the submit button. Variables available for replacement include ${patchSet} for the current patch set number (1, 2, 3), ${branch} for the branch name ("master") and ${commit} for the abbreviated commit SHA-1 (c9c0edb).

Default is "Submit patch set ${patchSet} into ${branch}".

change.submitTooltipAncestors

Tooltip for the submit button if there are ancestors which would also be submitted by submitting the change. Additionally to the variables as in change.submitTooltip, there is the variable ${submitSize} indicating the number of changes which are submitted.

Default is "Submit all ${topicSize} changes of the same topic (${submitSize} changes including ancestors and other changes related by topic)".

change.submitTopicLabel

If change.submitWholeTopic is set and a change has a topic, the label name for the submit button is given here instead of the configuration change.submitLabel.

Defaults to "Submit whole topic"

change.submitTopicTooltip

If change.submitWholeTopic is configured to true and a change has a topic, this configuration determines the tooltip for the submit button instead of change.submitTooltip. The variable ${topicSize} is available for the number of changes in the same topic to be submitted. The number of all changes to be submitted is in the variable ${submitSize}.

Defaults to "Submit all ${topicSize} changes of the same topic (${submitSize} changes including ancestors and other changes related by topic)".

change.submitWholeTopic

Determines if the submit button submits the whole topic instead of just the current change.

Default is false.

change.updateDelay

How often in seconds the web interface should poll for updates to the currently open change. The poller relies on the client’s browser cache to use If-Modified-Since and respect 304 Not Modified HTTP responses. This allows for fast polls, often under 8 milliseconds.

With a configured 30 second delay a server with 4900 active users will typically need to dedicate 1 CPU to the update check. 4900 users divided by an average delay of 30 seconds is 163 requests arriving per second. If requests are served at \~6 ms response time, 1 CPU is necessary to keep up with the update request traffic. On a smaller user base of 500 active users, the default 30 second delay is only 17 requests per second and requires ~10% CPU.

If 0 the update polling is disabled.

Default is 5 minutes.

+ If false, Gerrit will always take latest project configuration to compute submit labels. This means that, closed changes (either merged or abandoned) will be evaluated against the latest configuration which may produce different results. Especially for merged changes, they may look like they didn’t meet the submit requirements.

+ When true, evaluation will be skipped and Gerrit will show the exact status of submit labels when change was submitted. Post-review votes will only be allowed on labels that were configured when change was closed.

+ Default it false.

Section changeCleanup

This section allows to configure change cleanups and schedules them to run periodically.

changeCleanup.abandonAfter

Period of inactivity after which open changes should be abandoned automatically.

By default 0, never abandon open changes.

[WARNING] Auto-Abandoning changes may confuse/annoy users. When enabling this, make sure to choose a reasonably large grace period and inform users in advance.

The following suffixes are supported to define the time unit:

  • d, day, days

  • w, week, weeks (1 week is treated as 7 days)

  • mon, month, months (1 month is treated as 30 days)

  • y, year, years (1 year is treated as 365 days)

changeCleanup.abandonIfMergeable

Whether changes which are mergeable should be auto-abandoned. When set to false, -is:mergeable is appended to the query used to find the changes to auto-abandon.

By default true, meaning mergeable changes are auto-abandoned.

If change.mergeabilityComputationBehavior is set to NEVER, setting this option to false has no effect and it behaves as though it were set to true.

changeCleanup.cleanupAccountPatchReview

Whether accountPatchReview data should be also removed when change gets auto-abandoned.

By default false.

changeCleanup.abandonMessage

Change message that should be posted when a change is abandoned.

'${URL}' can be used as a placeholder for the Gerrit web URL.

By default "Auto-Abandoned due to inactivity, see ${URL}Documentation/user-change-cleanup.html#auto-abandon\n\n If this change is still wanted it should be restored.".

changeCleanup.startTime

The start time for running change cleanups.

changeCleanup.interval

The interval for running change cleanups.

Schedule examples can be found in the Schedule Configuration section.

Comment links are find/replace strings applied to change descriptions, patch comments, in-line code comments and approval category value descriptions to turn set strings into hyperlinks. One common use is for linking to bug-tracking systems.

In the following example configuration the 'changeid' comment link will match typical Gerrit Change-Id values and create a hyperlink to changes which reference it. The second configuration 'bugzilla' will hyperlink terms such as 'bug 42' to an external bug tracker, supplying the argument record number '42' for display.

commentlinks supports configuration reloads. Though a flush-caches of "projects" is needed for the commentlinks to be immediately available in the UI.

[commentlink "changeid"]
  match = (I[0-9a-f]{8,40})
  link = "#/q/$1"

[commentlink "bugzilla"]
  match = "(^|\\s)(bug\\s+#?)(\\d+)($|\\s)"
  link = http://bugs.example.com/show_bug.cgi?id=$3
  prefix = $1
  suffix = $4
  text = $2$3

Comment links can also be specified in project.config and sections in children override those in parents.

commentlink.<name>.match

A JavaScript regular expression to match positions to be replaced with a hyperlink. Subexpressions of the matched string can be stored using groups and accessed with $'n' syntax, where 'n' is the group number, starting from 1.

The configuration file parser eats one level of backslashes, so the character class \s requires \\s in the configuration file. The parser also terminates the line at the first #, so a match expression containing # must be wrapped in double quotes.

To match case insensitive strings, a character class with both the upper and lower case character for each position must be used. For example, to match the string bug in a case insensitive way the match pattern [bB][uU][gG] needs to be used.

The commentlink.name.match regular expressions are applied to the raw, unformatted and unescaped text form. Regex matching against HTML is not supported. Comment link patterns that are written in this style should be updated to match text formats.

A common pattern to match is bug\\s+(\\d+).

In order to better control the visual presentation of the link prefix, suffix and text is used. With the generated link html looking like: prefix<a …​>text</a>suffix.

commentlink.<name>.link

The URL to direct the user to whenever the regular expression is matched. Groups in the match expression may be accessed as $'n'.

commentlink.<name>.prefix

The text inserted before the link. Groups in the match expression may be accessed as $'n'.

commentlink.<name>.suffix

The text inserted after the link. Groups in the match expression may be accessed as $'n'.

commentlink.<name>.text

The text content of the link. Groups in the match expression may be accessed as $'n'.

If not specified defaults to $& (the matched text).

commentlink.<name>.enabled

Whether the comment link is enabled. A child project may override a section in a parent or the site-wide config that is disabled by specifying enabled = true.

By default, true.

Note that the names and contents of disabled sections are visible even to anonymous users via the REST API.

Section container

These settings are applied only if Gerrit is started as the container process through Gerrit’s 'gerrit.sh' rc.d compatible wrapper script.

container.heapLimit

Maximum heap size of the Java process running Gerrit, in bytes. This property is translated into the '-Xmx' flag for the JVM.

Default is platform and JVM specific.

Common unit suffixes of 'k', 'm', or 'g' are supported.

container.javaHome

Path of the JRE/JDK installation to run Gerrit with. If not set, the Gerrit startup script will attempt to search your system and guess a suitable JRE. Overrides the environment variable 'JAVA_HOME'.

container.javaOptions

Additional options to pass along to the Java runtime. May be specified multiple times to configure multiple values. If multiple values are configured, they are passed in order on the command line, separated by spaces. These options are appended onto 'JAVA_OPTIONS'.

For example, it is possible to overwrite Gerrit’s default log4j configuration:

  javaOptions = -Dlog4j.configuration=file:///home/gerrit/site/etc/log4j.properties

Gerrit built-in loggers are then ignored: error logger (error_log file), httpd.requestLog and sshd.requestLog. The log.jsonLogging and log.textLogging options are also ignored.

container.daemonOpt

Additional options to pass to the daemon (e.g. '--enable-httpd'). If multiple values are configured, they are passed in that order to the command line, separated by spaces.

Execute java -jar gerrit.war daemon --help to see all possible options.

container.replica

Used on Gerrit replica installations. If set to true the Gerrit JVM is called with the '--replica' switch, enabling replica mode. If no value is set (or any other value), Gerrit defaults to primary mode enabling write operations.

container.slave

Backward compatibility for container.replica config setting.

container.startupTimeout

The maximum time (in seconds) to wait for a gerrit.sh start command to run a new Gerrit daemon successfully. If not set, defaults to 90 seconds.

container.shutdownTimeout

The maximum time (in seconds) to wait for a gerrit.sh stop command. This is added to the highest value between either 'sshd.gracefulStopTimeout' or 'httpd.gracefulStopTimeout'. If not set, defaults to 30 seconds

container.user

Login name (or UID) of the operating system user the Gerrit JVM will execute as. If not set, defaults to the user who launched the 'gerrit.sh' wrapper script.

container.war

Path of the JAR file to start daemon execution with. This should be the path of the local 'gerrit.war' archive. Overrides the environment variable 'GERRIT_WAR'.

If not set, defaults to '$site_path/bin/gerrit.war', or to '$HOME/gerrit.war'.

Section core

Note
The etc/jgit.config file supports configuration of all JGit options.
core.packedGitWindowSize

Number of bytes of a pack file to load into memory in a single read operation. This is the "page size" of the JGit buffer cache, used for all pack access operations. All disk IO occurs as single window reads. Setting this too large may cause the process to load more data than is required; setting this too small may increase the frequency of read() system calls.

Default on JGit is 8 KiB on all platforms.

Common unit suffixes of 'k', 'm', or 'g' are supported.

core.packedGitLimit

Maximum number of bytes to load and cache in memory from pack files. If JGit needs to access more than this many bytes it will unload less frequently used windows to reclaim memory space within the process. As this buffer must be shared with the rest of the JVM heap, it should be a fraction of the total memory available.

Default on JGit is 10 MiB on all platforms.

Common unit suffixes of 'k', 'm', or 'g' are supported.

core.packedGitUseStrongRefs

Set to true in order to use strong references to reference packfile pages cached in the WindowCache. Otherwise SoftReferences are used. If this option is set to false, the Java garbage collector will flush the WindowCache to free memory if the used heap comes close to the maximum heap size. This has the advantage that it can quickly reclaim memory which was used by the WindowCache but comes at the price that the previously cached pack file content needs to be again copied from the file system cache to the Gerrit process. Setting this option to true prevents flushing the WindowCache which provides more predictable performance.

Default is false.

core.deltaBaseCacheLimit

Maximum number of bytes to reserve for caching base objects that multiple deltafied objects reference. By storing the entire decompressed base object in a cache Git is able to avoid unpacking and decompressing frequently used base objects multiple times.

Default on JGit is 10 MiB on all platforms. You probably do not need to adjust this value.

Common unit suffixes of 'k', 'm', or 'g' are supported.

core.packedGitOpenFiles

Maximum number of pack files to have open at once. A pack file must be opened in order for any of its data to be available in a cached window.

If you increase this to a larger setting you may need to also adjust the ulimit on file descriptors for the host JVM, as Gerrit needs additional file descriptors available for network sockets and other repository data manipulation.

Default on JGit is 128 file descriptors on all platforms.

core.streamFileThreshold

Largest object size, in bytes, that JGit will allocate as a contiguous byte array. Any file revision larger than this threshold will have to be streamed, typically requiring the use of temporary files under '$GIT_DIR/objects' to implement pseudo-random access during delta decompression.

Servers with very high traffic should set this to be larger than the size of their common big files. For example a server managing the Android platform typically has to deal with ~10-12 MiB XML files, so 15 m would be a reasonable setting in that environment. Setting this too high may cause the JVM to run out of heap space when handling very big binary files, such as device firmware or CD-ROM ISO images.

Defaults to 25% of the available JVM heap, limited to 2g.

Common unit suffixes of 'k', 'm', or 'g' are supported.

core.packedGitMmap

When true, JGit will use mmap() rather than malloc()+read() to load data from pack files. The use of mmap can be problematic on some JVMs as the garbage collector must deduce that a memory mapped segment is no longer in use before a call to munmap() can be made by the JVM native code.

In server applications (such as Gerrit) that need to access many pack files, setting this to true risks artificially running out of virtual address space, as the garbage collector cannot reclaim unused mapped spaces fast enough.

Default on JGit is false. Although potentially slower, it yields much more predictable behavior.

core.asyncLoggingBufferSize

Size of the buffer to store logging events for asynchronous logging. Putting a larger value can protect threads from stalling when the AsyncAppender threads are not fast enough to consume the logging events from the buffer. It also protects from losing log entries in this case.

Default is 64 entries.

core.useRecursiveMerge

Use JGit’s recursive merger for three-way merges. This only affects projects that allow content merges.

As explained in this blog, the recursive merge produces better results if the two commits that are merged have more than one common predecessor.

Default is true.

core.repositoryCacheCleanupDelay

Delay between each periodic cleanup of expired repositories.

Values can be specified using standard time unit abbreviations (ms, sec, min, etc.).

Set it to 0 in order to switch off cache expiration. If cache expiration is switched off, the JVM can still evict cache entries when it is running low on available heap memory.

Set it to -1 to automatically derive cleanup delay from core.repositoryCacheExpireAfter (lowest value between 1/10 of core.repositoryCacheExpireAfter and 10 minutes).

Default is -1.

core.repositoryCacheExpireAfter

Time an unused repository should expire and be evicted from the repository cache.

Values can be specified using standard time unit abbreviations (ms, sec, min, etc.).

Default is 1 hour.

core.usePerRequestRefCache

Use a per request (currently per request thread) ref cache. The ref cache uses JGit’s SnapshottingRefDirectory to ensure that packed refs are checked and potentially read at least once per request (lazily) if needed. This helps reduce the overhead of checking if the packed-refs file is outdated.

Default is true.

Section dashboard

dashboard.submitRequirementColumns

The list of submit requirement names that should be displayed as separate columns in the dashboard.

Section download

[download]
  command = checkout
  command = cherry_pick
  command = pull
  command = format_patch
  scheme = ssh
  scheme = http
  scheme = anon_http
  scheme = anon_git
  scheme = repo
  hide = ssh

The download section configures the allowed download methods.

download.command

Commands that should be offered to download changes.

Multiple commands are supported:

  • checkout

    Command to fetch and checkout the patch set.

  • cherry_pick

    Command to fetch the patch set and to cherry-pick it onto the current commit.

  • pull

    Command to pull the patch set.

  • format_patch

    Command to fetch the patch set and to feed it into the format-patch command.

If download.command is not specified, all download commands are offered.

download.scheme

Schemes that should be used to download changes.

Multiple schemes are supported:

  • http

    Authenticated HTTP download is allowed.

  • ssh

    Authenticated SSH download is allowed.

  • anon_http

    Anonymous HTTP download is allowed.

  • anon_git

    Anonymous Git download is allowed. This is not default, it is also necessary to set gerrit.canonicalGitUrl variable.

  • repo

    Gerrit advertises patch set downloads with the repo download command, assuming that all projects managed by this instance are generally worked on with the repo multi-repository tool. This is not default, as not all instances will deploy repo.

If download.scheme is not specified, SSH, HTTP and Anonymous HTTP downloads are allowed.

download.hide

Schemes that can be used to download changes, but will not be advertised in the UI. This can be any scheme that can be configured in download.scheme.

This is mostly useful in a deprecation scenario during a time where using a scheme is discouraged, but has to be supported until all clients have migrated to use a different scheme.

By default, no scheme will be hidden in the UI.

download.checkForHiddenChangeRefs

Whether the download commands should be adapted when the change refs are hidden.

Git has a configuration option to hide refs from the initial advertisement (uploadpack.hideRefs). This option can be used to hide the change refs from the client. As consequence fetching changes by change ref does not work anymore. However by setting uploadpack.allowTipSha1InWant to true fetching changes by commit ID is possible. If download.checkForHiddenChangeRefs is set to true the git download commands use the commit ID instead of the change ref when a project is configured like this.

Example git configuration on a project:

[uploadpack]
  hideRefs = refs/changes/
  hideRefs = refs/cache-automerge/
  allowTipSha1InWant = true

By default false.

download.archive

Specifies which archive formats, if any, should be offered on the change screen and supported for git-upload-archive operation:

[download]
  archive = tar
  archive = tbz2
  archive = tgz
  archive = txz
  archive = zip

If download.archive is not specified defaults to all archive commands. Set to off or empty string to disable.

Zip is not supported because it may be interpreted by a Java plugin as a valid JAR file, whose code would have access to cookies on the domain. For this reason zip format is always excluded from formats offered through the Download drop down or accessible in the REST API.

Section gc

This section allows to configure the git garbage collection and schedules it to run periodically. It will be triggered and executed sequentially for all projects.

gc.aggressive

Determines if scheduled garbage collections and garbage collections triggered through Web-UI should run in aggressive mode or not. Aggressive garbage collections are more expensive but may lead to significantly smaller repositories.

Valid values are "true" and "false," default is "false".

gc.startTime

The start time for running the git garbage collection.

gc.interval

The interval for running the git garbage collection.

Schedule examples can be found in the Schedule Configuration section.

Section gerrit

gerrit.basePath

Local filesystem directory holding all Git repositories that Gerrit knows about and can process changes for. A project entity in Gerrit maps to a local Git repository by creating the path string "${basePath}/${project_name}.git".

If relative, the path is resolved relative to '$site_path'.

gerrit.allProjects

Name of the permissions-only project defining global server access controls and settings. These are inherited into every other project managed by the running server. The name is relative to gerrit.basePath.

The persisted_projects cache must be flushed after this setting is changed.

Defaults to All-Projects if not set.

gerrit.defaultBranch

Name of the default branch to use on the project creation, if no other branches were specified in the input.

Defaults to refs/heads/master if not set.

gerrit.allUsers

Name of the project in which meta data of all users is stored. The name is relative to gerrit.basePath.

Defaults to All-Users if not set.

gerrit.canonicalWebUrl

The default URL for Gerrit to be accessed through.

Typically this would be set to something like "http://review.example.com/" or "http://example.com:8080/gerrit/" so Gerrit can output links that point back to itself.

Setting this is highly recommended, as it is necessary for the upload code invoked by "git push" or "repo upload" to output hyperlinks to the newly uploaded changes.

gerrit.canonicalGitUrl

Optional base URL for repositories available over the anonymous git protocol. For example, set this to git://mirror.example.com/base/ to have Gerrit display patch set download URLs in the UI. Gerrit automatically appends the project name onto the end of the URL.

By default unset, as the git daemon must be configured externally by the system administrator, and might not even be running on the same host as Gerrit.

gerrit.docUrl

Optional base URL for documentation, under which one can find "index.html", "rest-api.html", etc. Used as the base for the fixed set of links in the "Documentation" tab. A slash is implicitly appended. (For finer control over the top menu, consider writing a plugin.)

If unset or empty, the documentation tab will only be shown if /Documentation/index.html can be reached by the browser at app load time.

gerrit.editGpgKeys

If enabled and server-side signed push validation is also enabled, enable the REST API endpoints and web UI for editing GPG keys. If disabled, GPG keys can only be added by administrators with direct git access to All-Users.

Defaults to true.

gerrit.installCommitMsgHookCommand

Optional command to install the commit-msg hook. Typically of the form:

fetch-cmd some://url/to/commit-msg .git/hooks/commit-msg ; chmod +x .git/hooks/commit-msg

By default unset; falls back to using scp from the canonical SSH host, or curl from the canonical HTTP URL for the server. Only necessary if a proxy or other server/network configuration prevents clients from fetching from the default location.

gerrit.gitHttpUrl

Optional base URL for repositories available over the HTTP protocol. For example, set this to http://mirror.example.com/base/ to have Gerrit display URLs from this server, rather than itself.

By default unset, as the HTTP daemon must be configured externally by the system administrator, and might not even be running on the same host as Gerrit.

gerrit.installBatchModule

Repeatable list of class name of additional Guice modules to load as override to the batchInjector’s modules during the init phases. Classes are resolved using the primary Gerrit class loader, hence the class needs to be either declared in Gerrit or an additional JAR located under the /lib directory.

By default unset.

gerrit.installDbModule

Repeatable list of class name of additional Guice modules to load at Gerrit startup as part of the dbInjector. Classes are resolved using the primary Gerrit class loader, hence the class needs to be either declared in Gerrit or an additional JAR located under the /lib directory.

By default unset.

gerrit.installIndexModule

Class name of the Guice modules to load as alternate implementation for the Gerrit indexes backend. Classes are resolved using the primary Gerrit class loader, hence the class needs to be either declared in Gerrit or an additional JAR located under the /lib directory.

Note
The gerrit.installIndexModule has precedence over the index.type.

By default unset.

Example:

[gerrit]
  installIndexModule = com.google.gerrit.elasticsearch.ElasticIndexModule

+ gerrit.installModule::

+ Repeatable list of class name of additional Guice modules to load at Gerrit startup as part of the sysInjector. Classes are resolved using the primary Gerrit class loader, hence the class needs to be either declared in Gerrit or an additional JAR located under the /lib directory.

+ By default unset.

+ Example:

[gerrit]
  installModule = com.googlesource.gerrit.libmodule.MyModule
  installModule = com.example.abc.OurSpecialSauceModule
  installDbModule = com.example.def.OurCustomProvider
  installBatchModule = com.example.ghi.CustomBatchInitModule
gerrit.listProjectsFromIndex

Enable rendering of project list from the secondary index instead of purely relying on the in-memory cache.

By default false.

Note
The in-memory cache (set to false) rendering provides an unlimited list as a result of the list project API, causing the full list of projects to be returned as a result of the /projects/ REST API or the gerrit ls-projects SSH command. When the rendering from the secondary index (set to true), the list is limited by the global capability queryLimit which is defaulted to 500 entries.
gerrit.primaryWeblinkName

Name of the Weblink that should be chosen in cases where only one Weblink can be used in the UI, for example in inline links.

By default unset, meaning that the UI is responsible for trying to identify a weblink to be used in these cases, most likely weblinks that links to code browsers with known integrations with Gerrit (like Gitiles and Gitweb).

Example:

[gerrit]
  primaryWeblinkName = gitiles
gerrit.reportBugUrl

URL to direct users to when they need to report a bug.

By default unset, meaning no bug report URL will be displayed. Administrators should set this to the URL of their issue tracker, if necessary.

gerrit.enablePeerIPInReflogRecord

Record actual peer IP address in ref log entry for identified user.

Defaults to false.

gerrit.secureStoreClass

Use the secure store implementation from a specified class.

If specified, must be the fully qualified class name of a class that implements the com.google.gerrit.server.securestore.SecureStore interface, and the jar file containing the class must be placed in the $site_path/lib folder.

If not specified, the default no-op implementation is used.

gerrit.canLoadInIFrame

For security reasons Gerrit will always jump out of iframe. Setting this option to true will prevent this behavior.

By default false.

gerrit.xframeOption

Add X-Frame-Options header to all HTTP responses. The X-Frame-Options HTTP response header can be used to indicate whether or not a browser should be allowed to render a page in a <frame>, <iframe>, <embed> or <object>.

Available values:

  1. ALLOW - The page can be displayed in a frame.

  2. SAMEORIGIN - The page can only be displayed in a frame on the same origin as the page itself.

    If link:#gerrit.canLoadInIFrame is set to false this option is ignored and the X-Frame-Options header is always set to DENY. Setting this option to ALLOW will cause the X-Frame-Options header to be omitted the the page can be displayed in a frame.

    By default SAMEORIGIN.

gerrit.cdnPath

Path prefix for Gerrit’s static resources if using a CDN.

gerrit.faviconPath

Path for Gerrit’s favicon after default URL, including icon name and extension (.ico should be used).

gerrit.instanceId

Optional identifier for this Gerrit instance. Used to identify a specific instance within a group of Gerrit instances with the same serverId (i.e.: a Gerrit cluster). Unlike instanceName this value is not available in the email templates.

The instance ID can also be configured by setting the Java system property gerrit.instanceId on startup. This will override the configuration in the gerrit.config.

gerrit.instanceName

Short identifier for this Gerrit instance. A good name should be short but precise enough so that users can identify the instance among others.

Defaults to the full hostname of the Gerrit server.

gerrit.experimentalRollingUpgrade

Enable Gerrit rolling upgrade to the next version. For example if Gerrit v3.2 is version N (All-Projects:refs/meta/version=183) then its next version N+1 is v3.3 (All-Projects:refs/meta/version=184). Allow Gerrit to start even if the underlying schema version has been bumped to the next Gerrit version.

Set to true if Gerrit is installed in [high-availability configuration](https://gerrit.googlesource.com/plugins/high-availability/+/refs/heads/master/README.md) during the rolling upgrade to the next version.

By default false.

The rolling upgrade process, at high level, assumes that Gerrit is installed on two or more nodes sharing the repositories over NFS. The upgrade is composed of the following steps:

  1. Set gerrit.experimentalRollingUpgrade to true on all Gerrit masters

  2. Set the first master unhealthy

  3. Shutdown the first master and [upgrade](install.html#init) to the next version

  4. Startup the first master, wait for the online reindex to complete (where applicable)

  5. Verify the the first master upgrade is successful and online reindex is complete

  6. Set the first master healthy

  7. Repeat steps 2. to 6. for all the other Gerrit nodes

    Warning
    Rolling upgrade may or may not be possible depending on the changes introduced by the target version of the upgrade. Refer to the release notes and check whether the rolling upgrade is possible or not and the associated constraints.
gerrit.importedServerId

ServerId of the repositories imported from other Gerrit servers. Changes coming associated with the imported serverIds are indexed and displayed in the UI but they are not searchable by changeNumber therefore the index.cacheQueryResultsByChangeNum must also be set to false. Imported changes are still discoverable in any other ways, for example:

project:someproject branch:main changeId:I78a7add1fe2597cad788c833d8f771f09b54cf33

Specify multiple gerrit.importedServerId for allowing the import from multiple Gerrit servers with different serverIds.

Note
The account-ids referenced in the imported changes are used for looking up the associated account-id locally, using the imported: external-id. Example: the account-id 1000 from the imported server-id 59a4964e-6376-4ed9-beef will be looked up in the local accounts using the imported:1000@59a4964e-6376-4ed9-beef external-id.

If this value is not set, all changes imported from other Gerrit servers will be ignored.

By default empty.

gerrit.serverId

Used by NoteDb to, amongst other things, identify author identities from per-server specific account IDs.

If this value is not set on startup it is automatically set to a random UUID.

Note
If this value doesn’t match the serverId used when creating an already existing NoteDb, Gerrit will not be able to use that instance of NoteDb. The serverId used to create the NoteDb will show in the resulting exception message in case the value differs.

Section gitweb

Gerrit can forward requests to either an internally managed gitweb (which allows Gerrit to enforce some access controls), or to an externally managed gitweb (where the web server manages access). See also Gitweb Integration.

gitweb.cgi

Path to the locally installed gitweb.cgi executable. This CGI will be called by Gerrit Code Review when the URL /gitweb is accessed. Project level access controls are enforced prior to calling the CGI.

Defaults to /usr/lib/cgi-bin/gitweb.cgi if gitweb.url is not set.

gitweb.url

Optional URL of an affiliated gitweb service. Defines the web location where a gitweb.cgi is installed to browse gerrit.basePath and the repositories it contains.

Gerrit appends any necessary query arguments onto the end of this URL. For example, ?p=$project.git;h=$commit.

gitweb.type

Optional type of affiliated gitweb service. This allows using alternatives to gitweb, such as cgit.

Valid values are gitweb, cgit, disabled or custom.

If not set, or set to disabled, there is no gitweb hyperlinking support.

gitweb.revision

Optional pattern to use for constructing the gitweb URL when pointing at a specific commit when gitweb.type is set to custom.

Valid replacements are ${project} for the project name in Gerrit and ${commit} for the SHA-1 hash for the commit.

gitweb.project

Optional pattern to use for constructing the gitweb URL when pointing at a specific project when gitweb.type is set to custom.

Valid replacements are ${project} for the project name in Gerrit.

gitweb.branch

Optional pattern to use for constructing the gitweb URL when pointing at a specific branch when gitweb.type is set to custom.

Valid replacements are ${project} for the project name in Gerrit and ${branch} for the name of the branch.

gitweb.tag

Optional pattern to use for constructing the gitweb URL when pointing at a specific tag when gitweb.type is set to custom.

Valid replacements are ${project} for the project name in Gerrit and ${tag} for the name of the tag.

gitweb.roottree

Optional pattern to use for constructing the gitweb URL when pointing at the contents of the root tree in a specific commit when gitweb.type is set to custom.

Valid replacements are ${project} for the project name in Gerrit and ${commit} for the SHA-1 hash for the commit.

gitweb.file

Optional pattern to use for constructing the gitweb URL when pointing at the contents of a file in a specific commit when gitweb.type is set to custom.

Valid replacements are ${project} for the project name in Gerrit, ${file} for the file name, ${hash} for the SHA-1 hash for the commit, and ${commit} for the change ref or SHA-1 of the commit if no base patch set.

gitweb.filehistory

Optional pattern to use for constructing the gitweb URL when pointing at the history of a file in a specific branch when when gitweb.type is set to custom.

Valid replacements are ${project} for the project name in Gerrit, ${file} for the file name and ${branch} for the name of the branch.

gitweb.linkname

Optional setting for modifying the link name presented to the user in the Gerrit web-UI.

The default linkname for custom type is gitweb.

gitweb.pathSeparator

Optional character to substitute the standard path separator (slash) in project names and branch names.

By default, Gerrit will use hexadecimal encoding for slashes in project and branch names. Some web servers, such as Tomcat, reject this hexadecimal encoding in the URL.

Some alternative gitweb services, such as Gitblit, allow using an alternative path separator character. In Gitblit, this can be configured through the property web.forwardSlashCharacter. In Gerrit, the alternative path separator can be configured correspondingly using the property gitweb.pathSeparator.

Valid values are the characters *, ( and ).

gitweb.urlEncode

Whether or not Gerrit should encode the generated viewer URL.

Gerrit composes the viewer URL using information about the project, branch, file or commit of the target object to be displayed. Typically viewers such as CGit and gitweb do need those parts to be encoded, including the / in project’s name, for being correctly parsed. However other viewers could instead require an unencoded URL (e.g. GitHub web based viewer).

Valid values are true and false. The default is true.

Section groups

groups.includeExternalUsersInRegisteredUsersGroup

Controls whether external users (these are users we have sufficient knowledge about but who don’t yet have a Gerrit account) are considered to be members of the REGISTERED_USERS group.

This setting only makes sense if you run custom code (e.g. from a plugin or a custom authentication backend). By default, Gerrit core always requires users to register and doesn’t use external users.

By default, true.

groups.newGroupsVisibleToAll

Controls whether newly created groups should be by default visible to all registered users.

By default, false.

groups.<uuid>.name

Display name for group with the given UUID.

This option is only supported for system groups (scheme 'global').

E.g. this parameter can be used to configure another name for the Anonymous Users group:

[groups "global:Anonymous-Users"]
  name = All Users

When setting this parameter it should be verified that there is no existing group with the same name (case-insensitive). Configuring an ambiguous name makes Gerrit fail on startup. Once set Gerrit ensures that it is not possible to create a group with this name. Gerrit also keeps the default name reserved so that it cannot be used for new groups either. This means there is no danger of ambiguous group names when this parameter is removed and the system group uses the default name again.

groups.relevantGroup

UUID of an external group that should always be considered as relevant when checking whether an account is visible.

This setting is only relevant for external group backends and only if the account visibility is set to SAME_GROUP or VISIBLE_GROUP.

If the account visibility is set to SAME_GROUP or VISIBLE_GROUP users should see all accounts that are a member of a group that contains themselves or that is visible to them. Checking this would require getting all groups of the current user and all groups of the accounts for which the visibility is being checked, but since getting all groups that a user is a member of is expensive for external group backends Gerrit doesn’t query these groups but instead guesses the relevant groups. Guessing relevant groups limits the inspected groups to all groups that are mentioned in the ACLs of the projects that are currently cached (i.e. all groups that are listed in the groups files of the cached projects). This is not very reliable since it depends on which groups are mentioned in the ACLs and which projects are currently cached. To make this more reliable this configuration parameter allows to configure external groups that should always be considered as relevant.

As said this setting is only relevant for external group backends. In Gerrit core this is only the LDAP backend, but it may apply to further group backends that are added by plugins.

This parameter may be added multiple times to specify multiple relevant groups.

Section has operand alias

'has' operand aliasing allows global aliases to be defined for query 'has' operands. Currently only change queries are supported. The alias name is the git config key name, and the 'has' operand being aliased is the git config value.

For example:

[has-operand-alias "change"]
  oldtopic = topic

This section is particularly useful to alias 'has' operands (which may be long and clunky as they include a plugin name in them) to shorter operands without the plugin name. Admins should take care to choose shorter operands that are unique and unlikely to conflict in the future.

Aliases are resolved dynamically at invocation time to the currently loaded version of plugins. If a referenced plugin is not loaded, or does not define the command, "unsupported operand" is returned to the user.

Aliases will override existing 'has' operands. In case of multiple aliases with same name, the last one defined will be used.

When the target of an alias does not exist, the 'has' operand with the name of the alias will be used (if present). This enables an admin to configure the system to override a core 'has' operand with an operand provided by a plugin when present and otherwise fall back to the 'has' operand provided by core.

Section http

http.proxy

URL of the proxy server when making outgoing HTTP connections for OpenID login transactions. Syntax should be http://’hostname':’port'.

http.proxyUsername

Optional username to authenticate to the HTTP proxy with. This property is honored only if the username does not appear in the http.proxy property above.

http.proxyPassword

Optional password to authenticate to the HTTP proxy with. This property is honored only if the password does not appear in the http.proxy property above.

http.addUserAsRequestAttribute

If true, 'User' attribute will be added to the request attributes so it can be accessed outside the request scope (will be set to username or id if username not configured).

This attribute can be used by the servlet container to log user in the http access log.

When running the embedded servlet container, this attribute is used to print user in the httpd_log.

  • %{User}r

    Pattern to print user in Tomcat AccessLog.

Default value is true.

http.addUserAsResponseHeader

If true, the header 'User' will be added to the list of response headers so it can be accessed from a reverse proxy for logging purposes.

Default value is false.

Section httpd

The httpd section configures the embedded servlet container.

httpd.listenUrl

Configuration for the listening sockets of the internal HTTP daemon. Each entry of listenUrl combines the following options for a listening socket: protocol, network address, port and context path.

Protocol can be either http://, https://, proxy-http:// or proxy-https://. The latter two are special forms of http:// with awareness of a reverse proxy (see below). Network address selects the interface and/or scope of the listening socket. For notes examples, see below. Port is the TCP port number and is optional (default value depends on the protocol). Context path is the optional "base URI" for the Gerrit Code Review as application to serve on.

Protocol schemes:

  • http://

    Plain-text HTTP protocol. If port is not supplied, defaults to 80, the standard HTTP port.

  • https://

    SSL encrypted HTTP protocol. If port is not supplied, defaults to 443, the standard HTTPS port.

    For configuration of the certificate and private key, see httpd.sslKeyStore.

    Note
    SSL/TLS configuration capabilities of Gerrit internal HTTP daemon are very limited. Externally facing production sites are strongly encouraged to use a reverse proxy configuration to handle SSL/TLS and use a proxy-https:// scheme here (below) for security and performance reasons.
  • proxy-http://

    Plain-text HTTP relayed from a reverse proxy. If port is not supplied, defaults to 8080.

    Like http://, but additional header parsing features are enabled to honor X-Forwarded-For, X-Forwarded-Host and X-Forwarded-Server. These headers are typically set by Apache’s mod_proxy.

    Note

    For secruity reasons, make sure to only allow connections from a trusted reverse proxy in your network, as clients could otherwise easily spoof these headers and thus spoof their originating IP address effectively. If the reverse proxy is running on the same machine as Gerrit daemon, the use of a loopback network address to bind to (see below) is strongly recommended to mitigate this.

    If not using Apache’s mod_proxy, validate that your reverse proxy sets these headers on all requests. If not, either configure it to sanitize them from the origin, or use the http:// scheme instead.

  • proxy-https://

    Plain-text HTTP relayed from a reverse proxy that has already handled the SSL encryption/decryption. If port is not supplied, defaults to 8080.

    Behaves exactly like proxy-http://, but also sets the scheme to assume https:// is the proper URL back to the server.

Network address forms:

  • Loopback (localhost): 127.0.0.1 (IPv4) or [::1] (IPv6).

  • All (unspecified): 0.0.0.0 (IPv4), [::] (IPv6) or * (IPv4 and IPv6)

  • Interface IP address, e.g. 1.2.3.4 (IPv4) or [2001:db8::a00:20ff:fea7:ccea] (IPv6)

  • Hostname, resolved at startup time to an address.

Context path is the local part of the URL to be used to access Gerrit on ('base URL'). E.g. /gerrit/ to serve Gerrit on that URI as base. If set, consider to align this with the gerrit.canonicalWebUrl setting. Correct settings may depend on the reverse proxy configuration as well. By default, this is / so that Gerrit serves requests on the root.

If multiple values are supplied, the daemon will listen on all of them.

Examples:

[httpd]
    listenUrl = proxy-https://127.0.0.1:9999/gerrit/
[gerrit]
    # Reverse proxy is configured to serve with SSL/TLS on
    # example.com and to relay requests on /gerrit/ onto
    # http://127.0.0.1:9999/gerrit/
    canonicalWebUrl = https://example.com/gerrit/
[httpd]
    # Listen on specific external interface with plaintext
    # HTTP on IPv6.
    listenUrl = http://[2001:db8::a00:20ff:fea7:ccea]

    # Also listen on specific internal interface for use with
    # reverse proxy run on another host.
    listenUrl = proxy-https://192.168.100.123

See also the page on reverse proxy configuration.

By default, http://*:8080.

httpd.reuseAddress

If true, permits the daemon to bind to the port even if the port is already in use. If false, the daemon ensures the port is not in use before starting. Busy sites may need to set this to true to permit fast restarts.

By default, true.

httpd.gracefulStopTimeout

Set a graceful stop time. If set, the daemon ensures that all incoming calls are preserved for a maximum period of time, before starting the graceful shutdown process. Sites behind a workload balancer such as HAProxy would need this to be set for avoiding serving errors during rolling restarts.

Values should use common unit suffixes to express their setting:

  • s, sec, second, seconds

  • m, min, minute, minutes

    By default, 0 seconds (immediate shutdown).

httpd.inheritChannel

If true, permits the daemon to inherit its server socket channel from fd0/1(stdin/stdout). When set to true, the server can be socket activated via systemd or xinetd.

By default, false.

httpd.requestHeaderSize

Size, in bytes, of the buffer used to parse the HTTP headers of an incoming HTTP request. The entire request headers, including any cookies sent by the browser, must fit within this buffer, otherwise the server aborts with the response '413 Request Entity Too Large'.

One buffer of this size is allocated per active connection. Allocating a buffer that is too large wastes memory that cannot be reclaimed, allocating a buffer that is too small may cause unexpected errors caused by very long Referer URLs or large cookie values.

By default, 16384 (16 K), which is sufficient for most OpenID and other web-based single-sign-on integrations.

httpd.sslCrl

Path of the certificate revocation list file in PEM format. This crl file is optional, and available for CLIENT_SSL_CERT_LDAP authentication.

To create and view a crl using openssl:

openssl ca -gencrl -out crl.pem
openssl crl -in crl.pem -text

If not absolute, the path is resolved relative to $site_path.

By default, $site_path/etc/crl.pem.

httpd.sslKeyStore

Path of the Java keystore containing the server’s SSL certificate and private key. This keystore is required for https:// in URL.

To create a self-signed certificate for simple internal usage:

keytool -keystore keystore -alias jetty -genkey -keyalg RSA
chmod 600 keystore

If not absolute, the path is resolved relative to $site_path.

By default, $site_path/etc/keystore.

httpd.sslKeyPassword

Password used to decrypt the private portion of the sslKeyStore. Java keystores require a password, even if the administrator doesn’t want to enable one.

If set to the empty string the embedded server will prompt for the password during startup.

By default, gerrit.

httpd.requestLog

Enable (or disable) the '$site_path'/logs/httpd_log request log. If enabled, an NCSA combined log format request log file is written out by the internal HTTP daemon. The httpd log format is documented here.

log4j.appender with the name httpd_log can be configured to overwrite programmatic configuration.

By default, true if httpd.listenUrl uses http:// or https://, and false if httpd.listenUrl uses proxy-http:// or proxy-https://.

httpd.acceptorThreads

Number of worker threads dedicated to accepting new incoming TCP connections and allocating them connection-specific resources.

By default, 2, which should be suitable for most high-traffic sites.

httpd.minThreads

Minimum number of spare threads to keep in the worker thread pool. This number must be at least 1 larger than httpd.acceptorThreads multiplied by the number of httpd.listenUrls configured.

By default, 5, suitable for most lower-volume traffic sites.

httpd.maxThreads

Maximum number of threads to permit in the worker thread pool.

By default 25, suitable for most lower-volume traffic sites.

Note
Unless SSH daemon is disabled, see sshd.listenAddress, the max number of concurrent Git requests over HTTP and SSH together is defined by the sshd.threads and sshd.batchThreads.
httpd.maxQueued

Maximum number of client connections which can enter the worker thread pool waiting for a worker thread to become available. 0 sets the queue size to the Integer.MAX_VALUE.

By default 200.

httpd.maxWait

Maximum amount of time a client will wait for an available thread to handle a project clone, fetch or push request over the smart HTTP transport.

Values should use common unit suffixes to express their setting:

  • s, sec, second, seconds

  • m, min, minute, minutes

  • h, hr, hour, hours

  • d, day, days

  • w, week, weeks (1 week is treated as 7 days)

  • mon, month, months (1 month is treated as 30 days)

  • y, year, years (1 year is treated as 365 days)

If a unit suffix is not specified, minutes is assumed. If 0 is supplied, the maximum age is infinite and connections will not abort until the client disconnects.

By default, 5 minutes.

httpd.filterClass

Class that implements the javax.servlet.Filter interface for filtering any HTTP related traffic going through the Gerrit HTTP protocol. Class is loaded and configured in the Gerrit Jetty container and run in front of all Gerrit URL handlers, allowing the filter to inspect, modify, allow or reject each request. It needs to be provided as JAR library under $GERRIT_SITE/lib as it is resolved using the default Gerrit class loader and cannot be dynamically loaded by a plugin.

Failing to load the Filter class would result in a Gerrit start-up failure, as this class is supposed to provide mandatory filtering in front of Gerrit HTTP protocol.

Typical usage is in conjunction with the auth.type=HTTP as replacement of an Apache HTTP proxy layer as security enforcement on top of Gerrit by returning a trusted username as HTTP Header.

Allow multiple values to install multiple servlet filters.

Example of using a security library secure.jar under $GERRIT_SITE/lib that provides a org.anyorg.MySecureHeaderFilter Servlet Filter that enforces a trusted username in the TRUSTED_USER HTTP Header and org.anyorg.MySecureIPFilter that performs source IP security filtering:

[auth]
	type = HTTP
	httpHeader = TRUSTED_USER

[httpd]
	filterClass = org.anyorg.MySecureHeaderFilter
	filterClass = org.anyorg.MySecureIPFilter
filterClass.<className>.initParam

Gerrit supports customized pluggable HTTP filters as filterClass. This option allows to pass extra initialization parameters to the filter. It allows for multiple key/value pairs to be passed in this pattern:

initParam = <key>=<value>

For a comprehensive example:

[httpd]
	filterClass = org.anyorg.AFilter
	filterClass = org.anyorg.BFilter
[filterClass "org.anyorg.AFilter"]
	key1 = value1
	key2 = value2
[filterClass "org.anyorg.BFilter"]
	key3 = value3
httpd.idleTimeout

Maximum idle time for a connection, which roughly translates to the TCP socket SO_TIMEOUT.

This value is interpreted as the maximum time between some progress being made on the connection. So if a single byte is read or written, then the timeout is reset.

The max idle time is applied:

  • When waiting for a new message to be received on a connection

  • When waiting for a new message to be sent on a connection

By default, 30 seconds.

httpd.robotsFile

Location of an external robots.txt file to be used instead of the one bundled with the .war of the application.

If not absolute, the path is resolved relative to $site_path.

If the file doesn’t exist or can’t be read the default robots.txt file bundled with the .war will be used instead.

httpd.registerMBeans

Enable (or disable) registration of Jetty MBeans for Java JMX.

By default, false.

Section index

The index section configures the secondary index.

Note that after enabling the secondary index, the index must be built using the reindex program before restarting the Gerrit server.

index.type

(DEPRECATED) The only supported value is LUCENE, which is the default, that means a Lucene index is used.

For using other indexing backends (e.g. ElasticSearch), refer to gerrit.installIndexModule setting.

index.threads

Number of threads to use for indexing in normal interactive operations. Setting it to 0 disables the dedicated thread pool and indexing will be done in the same thread as the operation.

If not set or set to a zero, defaults to the number of logical CPUs as returned by the JVM. If set to a negative value, defaults to a direct executor.

index.batchThreads

Number of threads to use for indexing in background operations, such as online schema upgrades, and also the default for offline reindexing.

If not set or set to a zero, defaults to the number of logical CPUs as returned by the JVM. If set to a negative value, defaults to a direct executor.

index.cacheQueryResultsByChangeNum

Allow to cache and reuse the change JSON elements by their Change number. This improves the performance of queries that are returning Changes duplicates. It needs to be turned off when having Changes imported from other servers because of the potential conflicts of change numbers.

Defaults to true.

index.onlineUpgrade

Whether to upgrade to new index schema versions while the server is running. This is recommended as it prevents additional downtime during Gerrit version upgrades (avoiding the need for an offline reindex step using Reindex), but can add additional server load during the upgrade.

If set to false, there is no way to upgrade the index schema to take advantage of new search features without restarting the server.

Defaults to true.

index.paginationType

The pagination type to use when index queries are repeated to obtain the next set of results. Supported values are:

  • OFFSET

    Index queries are repeated with a non-zero offset to obtain the next set of results. Note: Results may be inaccurate if the data-set is changing during the query execution.

  • SEARCH_AFTER

    Index queries are repeated using a search-after object. Index backends can provide their custom implementations for search-after. Note that, SEARCH_AFTER does not impact using offsets in Gerrit query APIs. Note: Depending on the index backend and its settings, results may be inaccurate if the data-set is changing during the query execution.

  • NONE

    Index queries are executed returning all results, without internal pagination. Note: Since the entire set of indexing results is kept in memory during the API execution, this option may lead to higher memory utilisation and overall reduced performance. Bear in mind that some indexing backends may not support unbounded queries; therefore, the NONE option is unavailable.

    Defaults to OFFSET.

index.maxLimit

Maximum limit to allow for search queries. Requesting results above this limit will truncate the list (but will still set _more_changes on result lists). Set to 0 for no limit.

When index.type is set to LUCENE, defaults to no limit.

index.maxPages

Maximum number of pages of search results to allow, as index implementations may have to scan through large numbers of skipped results when searching with an offset. Requesting results starting past this threshold times the requested limit will result in an error. Set to 0 for no limit.

Defaults to no limit.

index.pageSizeMultiplier

When index queries are repeated to obtain more results, this multiplier will be used to determine the limit for the next query. Using a page multiplier allows queries to start off small and thus provide good latency for queries which may end up only having very few results, and then scaling up to have better throughput to handle queries with larger result sets without incurring the overhead of making as many queries as would be required with a smaller limit. This strategy of using a multiplier attempts to create a balance between latency and throughput by dynamically adjusting the query size to the number of results being returned by each query in the pagination.

The larger the multiplier, the better the throughput on large queries, and it also improves latency on large queries by scaling up quickly. However, a larger multiplier can hurt latencies a bit by making the "last" query in a series longer than needed. The impact of this depends on how much the backend latency goes up when specifying a large limit and few results are returned. Setting index.maxPageSize that isn’t too large, can likely help reduce the impacts of this.

For example, if the limit of the previous query was 500 and pageSizeMultiplier is configured to 5, the next query will have a limit of 2500.

Note: ignored when paginationType is `NONE`

Defaults to 1 which effectively turns this feature off.

index.maxPageSize

Maximum size to allow when index queries are repeated to obtain more results. Note that, index.maxLimit will be used to limit page size if it is configured to a value lower than maxPageSize.

For example, if the limit of previous query was 500, pageSizeMultiplier is configured to 5 and maxPageSize to 2000, the next query will have a limit of 2000 (instead of 2500).

Note: ignored when paginationType is `NONE`

Defaults to no limit.

index.maxTerms

Maximum number of leaf terms to allow in a query. Too-large queries may perform poorly, so setting this option causes query parsing to fail fast before attempting to send them to the secondary index.

When the index type is LUCENE, also sets the maximum number of clauses permitted per BooleanQuery. This is so that all enforced query limits are the same.

Defaults to 1024.

index.autoReindexIfStale

Whether to automatically check if a document became stale in the index immediately after indexing it. If false, there is a race condition during two simultaneous writes that may cause one of the writes to not be reflected in the index. The check to avoid this does consume some resources.

Defaults to false.

Subsection index.scheduledIndexer

This section configures periodic indexing. Periodic indexing is intended to run only on replicas and only updates the group index. Replication to replicas happens on Git level so that Gerrit is not aware of incoming replication events. But replicas need an updated group index to resolve memberships of users for ACL validation. To keep the group index in replicas up-to-date the Gerrit replica periodically scans the group refs in the All-Users repository to reindex groups if they are stale.

The scheduled reindexer is not able to detect group deletions that happened while the replica was offline, but since group deletions are not supported this should never happen. If nevertheless groups refs were deleted while a replica was offline a full offline reindex must be performed.

This section is only used if Gerrit runs in replica mode, otherwise it is ignored.

index.scheduledIndexer.runOnStartup

Whether the scheduled indexer should run once immediately on startup. If set to true the replica startup is blocked until all stale groups were reindexed. Enabling this allows to prevent that replicas that were offline for a longer period of time run with outdated group information until the first scheduled indexing is done.

Defaults to true.

index.scheduledIndexer.enabled

Whether the scheduled indexer is enabled. If the scheduled indexer is disabled you must implement other means to keep the group index for the replica up-to-date.

Defaults to true.

index.scheduledIndexer.startTime

The start time for running the scheduled indexer.

Defaults to 00:00.

index.scheduledIndexer.interval

The interval for running the scheduled indexer.

Defaults to 5m.

Schedule examples can be found in the Schedule Configuration section.

Lucene configuration

Open and closed changes are indexed in separate indexes named 'open' and 'closed' respectively.

The following settings are only used when the index type is LUCENE.

index.name.ramBufferSize

Determines the amount of RAM that may be used for buffering added documents and deletions before they are flushed to the index. See the Lucene documentation for further details.

Defaults to 16M.

index.name.maxBufferedDocs

Determines the minimal number of documents required before the buffered in-memory documents are flushed to the index. Large values generally give faster indexing. See the Lucene documentation for further details.

Defaults to -1, meaning no maximum is set and the writer will flush according to RAM usage.

index.name.commitWithin

Determines the period at which changes are automatically committed to stable store on disk. This is a costly operation and may block additional index writes, so lower with caution.

If zero, changes are committed after every write. This is very costly but may be useful if offline reindexing is infeasible, or for development servers.

Values can be specified using standard time unit abbreviations (ms, sec, min, etc.).

If negative, commitWithin is disabled. Changes are flushed to disk when the in-memory buffer fills, but only committed and guaranteed to be synced to disk when the process finishes.

Defaults to 300000 ms (5 minutes).

index.name.maxMergeCount

Determines the max number of simultaneous merges that are allowed. If a merge is necessary yet we already have this many threads running, the incoming thread (that is calling add/updateDocument) will block until a merge thread has completed. Note that Lucene will only run the smallest maxThreadCount merges at a time. See the Lucene documentation for further details.

Defaults to -1 for (auto detection).

index.name.maxThreadCount

Determines the max number of simultaneous Lucene merge threads that should be running at once. This must be less than or equal to maxMergeCount. See the Lucene documentation for further details.

For further details on Lucene index configuration (auto detection) which affects maxThreadCount and maxMergeCount settings. See the Lucene documentation

Defaults to -1 for (auto detection).

index.name.enableAutoIOThrottle

Allows the control of whether automatic IO throttling is enabled and used by default in the lucene merge queue. Automatic dynamic IO throttling, which when on is used to adaptively rate limit writes bytes/sec to the minimal rate necessary so merges do not fall behind. See the Lucene documentation for further details.

Defaults to true (throttling enabled).

During offline reindexing, setting ramBufferSize greater than the size of index (size of specific index folder under <site_dir>/index) and maxBufferedDocs as -1 avoids unnecessary flushes and triggers only a single flush at the end of the process.

Sample Lucene index configuration:

[index]
  type = LUCENE

[index "changes_open"]
  ramBufferSize = 60 m
  maxBufferedDocs = 3000
  maxThreadCount = 5
  maxMergeCount = 50


[index "changes_closed"]
  ramBufferSize = 20 m
  maxBufferedDocs = 500
  maxThreadCount = 10
  maxMergeCount = 100
  enableIOThrottle = false

Section event

events.payload.listChangeOptions

List of options that Gerrit applies when rendering the payload of an internal event. This is the same set of options that are documented here.

Depending on the setup, these events might get serialized using stream events.

This can be set to the set of minimal options that consumers of Gerrit’s events need. A minimal set would be (SKIP_DIFFSTAT).

Every option that gets added here will have a performance impact. The general recommendation is therefore to set this to a minimal set of required options.

Defaults to all available options minus CHANGE_ACTIONS, CURRENT_ACTIONS and CHECK. This is a rich default to make sure the config is backwards compatible with what the default was before the config was added.

event.comment-added.publishPatchSetLevelComment

Add patch set level comment as event comment. Without this option, patch set level comment will not be included in the event comment attribute. Given that currently patch set level, file and robot comments are not exposed in the comment-added event type, those comments will be lost. One particular use case is to re-trigger CI build from the change screen by adding a comment with specific content, e.g.: recheck. Jenkins Gerrit Trigger plugin and Zuul CI depend on this feature to trigger change verification.

By default, true.

Section experiments

This section covers experimental new features. Gerrit uses experiments to research new behavior in frontend and core backend. Once the research is done, the experimental feature either stays and the experimentation flag gets removed, or the feature as a whole gets removed

experiments.enabled

List of experiments that are currently enabled. The release notes contain currently available experiments.

We will not remove experiments in stable patch releases. They are likely to be removed in the next stable version.

[experiments]
  enabled = ExperimentKey
experiments.disabled

List of experiments that are currently disabled. The release notes contain currently available experiments. This list disables experiments with the given key that are either enabled by default or explicitly in the config.

[experiments]
  disabled = ExperimentKey

Section ldap

LDAP integration is only enabled if auth.type is set to HTTP_LDAP, LDAP or CLIENT_SSL_CERT_LDAP. See above for a detailed description of the auth.type settings and their implications.

An example LDAP configuration follows, and then discussion of the parameters introduced here. Suitable defaults for most parameters are automatically guessed based on the type of server detected during startup. The guessed defaults support RFC 2307, Active Directory and FreeIPA.

[ldap]
  server = ldap://ldap.example.com

  accountBase = ou=people,dc=example,dc=com
  accountPattern = (&(objectClass=person)(uid=${username}))
  accountFullName = displayName
  accountEmailAddress = mail

  groupBase = ou=groups,dc=example,dc=com
  groupMemberPattern = (&(objectClass=group)(member=${dn}))
ldap.guessRelevantGroups

Filter the groups found in LDAP by guessing the ones relevant to Gerrit and removing the others from list completions and ACL evaluations. The guess is based on two elements: the projects most recently accessed in the cache and the list of LDAP groups included in their ACLs.

Please note that projects rarely used and thus not cached may be temporarily inaccessible by users even with LDAP membership and grants referenced in the ACLs.

By default, true.

ldap.server

URL of the organization’s LDAP server to query for user information and group membership from. Must be of the form ldap://host or ldaps://host to bind with either a plaintext or SSL connection.

If auth.type is LDAP this setting should use ldaps:// to ensure the end user’s plaintext password is transmitted only over an encrypted connection.

If you want to configure multiple ldap servers you can try to put multiple ldap urls separated by a space: server = ldaps://ldap1 ldaps://ldap2 See issue 40010644.

ldap.startTls

If true, Gerrit will perform StartTLS extended operation.

By default, false, StartTLS will not be enabled.

ldap.supportAnonymous

If false, Gerrit will provide credentials only at connection open, this is required for some LDAP implementations that do not allow anonymous bind for StartTLS or for reauthentication.

By default, true.

ldap.sslVerify

If false and ldap.server is an ldaps:// style URL or ldap.startTls is true, Gerrit will not verify the server certificate when it connects to perform a query.

By default, true, requiring the certificate to be verified.

ldap.groupsVisibleToAll

If true, LDAP groups are visible to all registered users.

By default, false, LDAP groups are visible only to administrators and group members.

ldap.username

(Optional) Username to bind to the LDAP server with. If not set, an anonymous connection to the LDAP server is attempted.

ldap.password

(Optional) Password for the user identified by ldap.username. If not set, an anonymous (or passwordless) connection to the LDAP server is attempted.

ldap.referral

(Optional) How an LDAP referral should be handled if it is encountered during directory traversal. Set to follow to automatically follow any referrals, or ignore to ignore the referrals.

By default, ignore.

ldap.readTimeout

(Optional) The read timeout for an LDAP operation. The value is in the usual time-unit format like "1 s", "100 ms", etc…​ A timeout can be used to avoid blocking all of the SSH command start threads in case the LDAP server becomes slow.

By default there is no timeout and Gerrit will wait for the LDAP server to respond until the TCP connection times out.

ldap.accountBase

Root of the tree containing all user accounts. This is typically of the form ou=people,dc=example,dc=com.

This setting may be added multiple times to specify more than one root.

ldap.accountScope

Scope of the search performed for accounts. Must be one of:

  • one: Search only one level below accountBase, but not recursive

  • sub or subtree: Search recursively below accountBase

  • base or object: Search exactly accountBase; probably not desired

Default is subtree as many directories have several levels.

ldap.accountPattern

Query pattern to use when searching for a user account. This may be any valid LDAP query expression, including the standard (&…​) and (|…​) operators. If auth.type is HTTP_LDAP then the variable ${username} is replaced with a parameter set to the username that was supplied by the HTTP server. If auth.type is LDAP then the variable ${username} is replaced by the string entered by the end user.

This pattern is used to search the objects contained directly under the ldap.accountBase tree. A typical setting for this parameter is (uid=${username}) or (cn=${username}), but the proper setting depends on the LDAP schema used by the directory server.

Default is (uid=${username}) for FreeIPA and RFC 2307 servers, and (&(objectClass=user)(sAMAccountName=${username})) for Active Directory.

ldap.accountFullName

(Optional) Name of an attribute on the user account object which contains the initial value for the user’s full name field in Gerrit. Typically this is the displayName property in LDAP, but could also be legalName or cn.

Attribute values may be concatenated with literal strings. For example to join given name and surname together, use the pattern ${givenName} ${SN}.

If set, users will be unable to modify their full name field, as Gerrit will populate it only from the LDAP data.

Default is displayName for FreeIPA and RFC 2307 servers, and ${givenName} ${sn} for Active Directory.

ldap.accountEmailAddress

(Optional) Name of an attribute on the user account object which contains the user’s Internet email address, as defined by this LDAP server.

Attribute values may be concatenated with literal strings, for example to set the email address to the lowercase form of sAMAccountName followed by a constant domain name, use ${sAMAccountName.toLowerCase}@example.com.

If set, the preferred email address will be prefilled from LDAP, but users may still be able to register additional email addresses, and select a different preferred email address.

Default is mail.

ldap.accountSshUserName

(Optional) Name of an attribute on the user account object which contains the initial value for the user’s SSH username field in Gerrit. Typically this is the uid property in LDAP, but could also be cn. Administrators should prefer to match the attribute corresponding to the user’s workstation username, as this is what SSH clients will default to.

Attribute values may also be forced to lowercase, or to uppercase in an expression. For example, ${sAMAccountName.toLowerCase} will force the value of sAMAccountName, if defined, to be all lowercase. The suffix .toUpperCase can be used for the other direction. The suffix .localPart can be used to split attribute values of the form 'user@example.com' and return only the left hand side, for example ${userPrincipalName.localPart} would provide only 'user'.

If set, users will be unable to modify their SSH username field, as Gerrit will populate it only from the LDAP data. Note that once the username has been set it cannot be changed, therefore it is recommended not to make changes to this setting that would cause the value to differ, as this will prevent users from logging in.

Default is uid for FreeIPA and RFC 2307 servers, and ${sAMAccountName.toLowerCase} for Active Directory.

ldap.accountMemberField

(Optional) Name of an attribute on the user account object which contains the groups the user is part of. Typically used for Active Directory and FreeIPA servers.

Default is unset for RFC 2307 servers (disabled) and memberOf for Active Directory and FreeIPA.

ldap.accountMemberExpandGroups

(Optional) Whether to expand nested groups recursively. This setting is used only if ldap.accountMemberField is set.

Default is unset for FreeIPA and true for RFC 2307 servers and Active Directory.

ldap.fetchMemberOfEagerly

(Optional) Whether to fetch the memberOf account attribute on login. Setups which use LDAP for user authentication but don’t make use of the LDAP groups may benefit from setting this option to false as this will result in a much faster LDAP login.

Default is unset for RFC 2307 servers (disabled) and true for Active Directory and FreeIPA.

ldap.groupBase

Root of the tree containing all group objects. This is typically of the form ou=groups,dc=example,dc=com.

This setting may be added multiple times to specify more than one root.

ldap.groupScope

Scope of the search performed for group objects. Must be one of:

  • one: Search only one level below groupBase, but not recursive

  • sub or subtree: Search recursively below groupBase

  • base or object: Search exactly groupBase; probably not desired

Default is subtree as many directories have several levels.

ldap.groupPattern

Query pattern used when searching for an LDAP group to connect to a Gerrit group. This may be any valid LDAP query expression, including the standard (&…​) and (|…​) operators. The variable ${groupname} is replaced with the search term supplied by the group owner.

Default is (cn=${groupname}) for FreeIPA and RFC 2307 servers, and (&(objectClass=group)(cn=${groupname})) for Active Directory.

ldap.groupMemberPattern

Query pattern to use when searching for the groups that a user account is currently a member of. This may be any valid LDAP query expression, including the standard (&…​) and (|…​) operators.

If auth.type is HTTP_LDAP then the variable ${username} is replaced with a parameter set to the username that was supplied by the HTTP server. Other variables appearing in the pattern, such as ${fooBarAttribute}, are replaced with the value of the corresponding attribute (in this case, fooBarAttribute) as read from the user’s account object matched under ldap.accountBase. Attributes such as ${dn} or ${uidNumber} may be useful.

Default is (|(memberUid=${username})(gidNumber=${gidNumber})) for RFC 2307, and unset (disabled) for Active Directory and FreeIPA.

ldap.groupName

(Optional) Name of the attribute on the group object which contains the value to use as the group name in Gerrit.

Typically the attribute name is cn for RFC 2307 and Active Directory servers. For other servers the attribute name may differ, for example apple-group-realname on Apple MacOS X Server.

It is also possible to specify a literal string containing a pattern of attribute values. For example to create a Gerrit group name consisting of LDAP group name and group ID, use the pattern ${cn} (${gidNumber}).

Default is cn.

ldap.mandatoryGroup

All users must be a member of this group to allow account creation or authentication.

Setting mandatoryGroup implies enabling of ldap.fetchMemberOfEagerly

By default, unset.

ldap.localUsernameToLowerCase

Converts the local username, that is used to login into the Gerrit Web UI, to lower case before doing the LDAP authentication. By setting this parameter to true, a case insensitive login to the Gerrit Web UI can be achieved.

If set, it must be ensured that the local usernames for all existing accounts are converted to lower case, otherwise a user that has a local username that contains upper case characters will not be able to login anymore. The local usernames for the existing accounts can be converted to lower case by running the server program LocalUsernamesToLowerCase. Please be aware that the conversion of the local usernames to lower case can’t be undone. For newly created accounts the local username will be directly stored in lower case.

By default, unset/false.

ldap.authentication

Defines how Gerrit authenticates with the server. When set to GSSAPI Gerrit will use Kerberos. To use kerberos the java.security.auth.login.config system property must point to a login to a JAAS configuration file and, if Java 6 is used, the system property java.security.krb5.conf must point to the appropriate krb5.ini file with references to the KDC.

Typical jaas.conf.

KerberosLogin {
    com.sun.security.auth.module.Krb5LoginModule
            required
            useTicketCache=true
            doNotPrompt=true
            renewTGT=true;
};

See Java documentation on how to create the krb5.ini file.

Note the renewTGT property to make sure the TGT does not expire, and useTicketCache to use the TGT supplied by the operating system. As the whole point of using GSSAPI is to have passwordless authentication to the LDAP service, this option does not acquire a new TGT on its own.

On Windows servers the registry key HKEY_LOCAL_MACHINE\System\CurrentControlSet\Control\Lsa\Kerberos\Parameters must have the DWORD value allowtgtsessionkey set to 1 and the account must not have local administrator privileges.

ldap.useConnectionPooling

(Optional) Enable the LDAP connection pooling or not.

If it is true, the LDAP service provider maintains a pool of (possibly) previously used connections and assigns them to a Context instance as needed. When a Context instance is done with a connection (closed or garbage collected), the connection is returned to the pool for future use.

By default, false.

ldap.connectTimeout

(Optional) Timeout period for establishment of an LDAP connection.

The value is in the usual time-unit format like "1 s", "100 ms", etc…​

By default there is no timeout and Gerrit will wait indefinitely.

LDAP Connection Pooling

Once LDAP connection pooling is enabled by setting the ldap.useConnectionPooling configuration property to true, the connection pool can be configured using JVM system properties as explained in the Java SE Documentation.

For standalone Gerrit (running with the embedded Jetty), JVM system properties are specified in the container section:

  javaOptions = -Dcom.sun.jndi.ldap.connect.pool.maxsize=20
  javaOptions = -Dcom.sun.jndi.ldap.connect.pool.prefsize=10
  javaOptions = -Dcom.sun.jndi.ldap.connect.pool.timeout=300000

Section lfs

lfs.plugin

The name of a plugin which serves the LFS protocol on the <project-name>/info/lfs/objects/batch endpoint. When not configured Gerrit will respond with 501 Not Implemented on LFS protocol requests.

By default unset.

Section log

log.jsonLogging

If set to true, enables error, ssh and http logging in JSON format (file names: logs/error_log.json, logs/sshd_log.json and logs/httpd_log.json).

The option only applies to Gerrit built-in loggers. It is ignored when a log4j configuration is specified via container.javaOptions, for example -Dlog4j.configuration=file://etc/log4j.properties.

Defaults to false.

log.textLogging

If set to true, enables error logging in regular plain text format. Can only be disabled if jsonLogging is enabled.

The option only applies to Gerrit built-in loggers. It is ignored when a log4j configuration is specified via container.javaOptions, for example -Dlog4j.configuration=file://etc/log4j.properties.

Defaults to true.

log.compress

If set to true, log files are compressed at server startup and then daily at 11pm (in the server’s local time zone).

Defaults to true.

log.rotate

If set to true, log files are rotated daily at midnight (GMT).

Defaults to true.

Section metrics

metrics.reservoir

The type of data reservoir used by the metrics system to calculate the percentile values for timers and histograms. It can be set to one of the following values:

  • ExponentiallyDecaying: An exponentially-decaying random reservoir based on Cormode et al’s forward-decaying priority reservoir sampling method to produce a statistically representative sampling reservoir, exponentially biased towards newer entries.

  • SlidingTimeWindowArray: A sliding window that stores only the measurements made in the last window using chunks of 512 samples.

  • SlidingTimeWindow: A sliding window that stores only the measurements made in the last window using a skip list.

  • SlidingWindow: A sliding window that stores only the last measurements.

  • Uniform: A random sampling reservoir that uses Vitter’s Algorithm R to produce a statistically representative sample.

    Defaults to ExponentiallyDecaying.

metrics.ExponentiallyDecaying.alpha

The exponential decay factor; the higher this is, the more biased the reservoir will be towards newer values.

metrics.<reservoirType>.size

The number of samples to keep in the reservoir. Applies to all reservoir types except the sliding time-based ones.

Defaults to 1028.

metrics.<reservoirType>.window

The window of time for keeping data in the reservoir. It only applies to sliding time-based reservoir types.

Section mimetype

mimetype.<name>.safe

If set to true, files with the MIME type <name> will be sent as direct downloads to the user’s browser, rather than being wrapped up inside of zipped archives. The type name may be a complete type name, e.g. image/gif, a generic media type, e.g. image/*, or the wildcard */* to match all types.

By default, false for all MIME types.

Common examples:

[mimetype "image/*"]
  safe = true

[mimetype "application/pdf"]
  safe = true

[mimetype "application/msword"]
  safe = true

[mimetype "application/vnd.ms-excel"]
  safe = true

Section noteDb

NoteDb is the Git-based database storage backend for Gerrit. For more information, including how to migrate data from an older Gerrit version, see the documentation.

notedb.accounts.sequenceBatchSize

The next available account sequence number is stored as UTF-8 text in a blob pointed to by the refs/sequences/accounts ref in the All-Users repository. Multiple processes share the same sequence by incrementing the counter using normal git ref updates. To amortize the cost of these ref updates, processes increment the counter by a larger number and hand out numbers from that range in memory until they run out. This configuration parameter controls the size of the account ID batch that each process retrieves at once.

By default, 1.

notedb.changes.sequenceBatchSize

The next available change sequence number is stored as UTF-8 text in a blob pointed to by the refs/sequences/changes ref in the All-Projects repository. Multiple processes share the same sequence by incrementing the counter using normal git ref updates. To amortize the cost of these ref updates, processes increment the counter by a larger number and hand out numbers from that range in memory until they run out. This configuration parameter controls the size of the change ID batch that each process retrieves at once.

By default, 20.

Section oauth

OAuth integration is only enabled if auth.type is set to OAUTH. See above for a detailed description of the auth.type settings and their implications.

By default, contact information, like the full name and email address, is retrieved from the selected OAuth provider when a user account is created, or when a user requests to reload that information in the settings UI. If that is not supported by the OAuth provider, users can be allowed to edit their contact information manually.

oauth.allowEditFullName

If true, the full name can be edited in the contact information.

Default is false.

oauth.allowRegisterNewEmail

If true, additional email addresses can be registered in the contact information.

Default is false.

Section operator alias

Operator aliasing allows global aliases to be defined for query operators. Currently only change queries are supported. The alias name is the git config key name, and the operator being aliased is the git config value.

For example:

[operator-alias "change"]
  oldage = age
  number = change

This section is particularly useful to alias operator names which may be long and clunky because they include a plugin name in them to a shorter name without the plugin name.

Aliases are resolved dynamically at invocation time to any currently loaded versions of plugins. If the alias points to an operator provided by a plugin which is not currently loaded, or the plugin does not define the operator, then "unsupported operator" is returned to the user.

Aliases will override existing operators. In the case of multiple aliases with the same name, the last one defined will be used.

When the target of an alias doesn’t exist, the operator with the name of the alias will be used (if present). This enables an admin to config the system to override a core operator with an operator provided by a plugin when present and otherwise fall back to the operator provided by core.

Section pack

Global settings controlling how Gerrit Code Review creates pack streams for Git clients running clone, fetch, or pull. Most of these variables are per-client request, and thus should be carefully set given the expected concurrent request load and available CPU and memory resources.

pack.deltacompression

If true, delta compression between objects is enabled. This may result in a smaller overall transfer for the client, but requires more server memory and CPU time.

False (off) by default, matching Gerrit Code Review 2.1.4.

pack.threads

Maximum number of threads to use for delta compression (if enabled). This is per-client request. If set to 0 then the number of CPUs is auto-detected and one thread per CPU is used, per client request.

By default, 1.

Section plugins

plugins.checkFrequency

How often plugins should be examined for new plugins to load, removed plugins to be unloaded, or updated plugins to be reloaded. Values can be specified using standard time unit abbreviations ('ms', 'sec', 'min', etc.).

If set to 0, automatic plugin reloading is disabled. Administrators may force reloading with gerrit plugin reload.

Default is 1 minute.

plugins.allowRemoteAdmin

Enable remote installation, enable and disable of plugins over HTTP and SSH. If set to true Administrators can install new plugins remotely, or disable existing plugins. Defaults to false.

plugins.mandatory

List of mandatory plugins. If a plugin from this list does not load, Gerrit will fail to start.

Disabling and restarting of a mandatory plugin is rejected, but reloading of a mandatory plugin is still possible.

plugins.jsLoadTimeout

Set the timeout value for loading JavaScript plugins in Gerrit UI. Values can be specified using standard time unit abbreviations ('ms', 'sec', 'min', etc.).

Default is 5 seconds. Negative values will be converted to 0.

plugins.transitionalPushOptions

Additional push options which should be accepted by gerrit as valid options even if they are not registered by any plugin(e.g. "myplugin~foo").

This config can be used when gerrit migrates from a deprecated plugin to the new one. The new plugin can (temporary) accept push options of the old plugin without registering such options.

Section receive

This section is used to configure behavior of the 'receive-pack' handler, which responds to 'git push' requests.

receive.allowGroup

Name of the groups of users that are allowed to execute 'receive-pack' on the server. One or more groups can be set.

If no groups are added, any user will be allowed to execute 'receive-pack' on the server.

receive.certNonceSeed

If set to a non-empty value and server-side signed push validation is enabled, use this value as the seed to the HMAC SHA-1 nonce generator. If unset, a 64-byte random seed will be generated at server startup.

As this is used as the seed of a cryptographic algorithm, it is recommended to be placed in secure.config.

Defaults to unset.

receive.certNonceSlop

When validating the nonce passed as part of the signed push protocol, accept valid nonces up to this many seconds old. This allows certificate verification to work over HTTP where there is a lag between the HTTP response providing the nonce to sign and the next request containing the signed nonce. This can be significant on large repositories, since the lag also includes the time to count objects on the client.

Default is 5 minutes.

receive.checkMagicRefs

If true, Gerrit will verify the destination repository has no references under the magic 'refs/for' branch namespace. Names under these locations confuse clients when trying to upload code reviews so Gerrit requires them to be empty.

If false Gerrit skips the sanity check and assumes administrators have ensured the repository does not contain any magic references. Setting to false to skip the check can decrease latency during push.

Default is true.

receive.allowProjectOwnersToChangeParent

If true, Gerrit will allow project owners to change the parent of a project.

By default only Gerrit administrators are allowed to change the parent of a project. By allowing project owners to change parents, it may allow the owner to circumvent certain enforced rules (like important BLOCK rules).

Default is false.

This value supports configuration reloads: reload-config

receive.checkReferencedObjectsAreReachable

If set to true, Gerrit will validate that all referenced objects that are not included in the received pack are reachable by the user.

Carrying out this check on gits with many refs and commits can be a very CPU-heavy operation. For non public Gerrit-servers this check may be overkill.

Only disable this check if you trust the clients not to forge SHA-1 references to access commits intended to be hidden from the user.

Default is true.

receive.enableInMemoryRefCache

If true, Gerrit will cache all refs advertised during push in memory and base later receive operations on that cache.

Turning this cache off is considered experimental.

This cache provides value when the ref database is slow and/or does not offer an inverse lookup of object ID to ref name. When RefTable is used, this cache can be turned off (experimental) to get speed improvements.

Default is true.

receive.enableSignedPush

If true, server-side signed push validation is enabled.

When a client pushes with git push --signed, this ensures that the push certificate is valid and signed with a valid public key stored in the refs/meta/gpg-keys branch of All-Users.

Defaults to false.

receive.maxBatchChanges

The maximum number of changes that Gerrit allows to be pushed in a batch for review. When this number is exceeded Gerrit rejects the push with an error message.

May be overridden for certain groups by specifying a limit in the 'Batch Changes Limit' global capability.

This setting can be used to prevent users from uploading large number of changes for review by mistake.

Default is zero, no limit.

receive.maxBatchCommits

The maximum number of commits that Gerrit allows to be pushed in a batch directly to a branch when bypassing review. This limit can be bypassed if a user skips validation.

Default is 10000.

receive.maxObjectSizeLimit

Maximum allowed Git object size that 'receive-pack' will accept. If an object is larger than the given size the pack-parsing will abort and the push operation will fail. If set to zero then there is no limit.

Gerrit administrators can use this setting to prevent developers from pushing objects which are too large to Gerrit.

This setting can also be set in the project.config (receive.maxObjectSizeLimit) in order to further reduce the global setting. The project specific setting is only honored when it further reduces the global limit.

Default is zero.

Common unit suffixes of 'k', 'm', or 'g' are supported.

receive.inheritProjectMaxObjectSizeLimit

Controls whether the project-level receive.maxObjectSizeLimit value is inherited from the parent project. When true, the value is inherited, otherwise it is not inherited.

Default is false, the value is not inherited.

receive.maxTrustDepth

If signed push validation is enabled, set to the maximum depth to search when checking if a key is trusted.

Default is 0, meaning only explicitly trusted keys are allowed.

receive.threadPoolSize

Maximum size of the thread pool in which the change data in received packs is processed.

Defaults to the number of available CPUs according to the Java runtime.

receive.timeout

Overall timeout on the time taken to process the change data in received packs. Only includes the time processing Gerrit changes and updating references, not the time to index the pack. Values can be specified using standard time unit abbreviations ('ms', 'sec', 'min', etc.).

After the timeout is exceeded the task processing the receive gets a cancellation signal that allows the tast to finish gracefully. receive.cancellationTimeout defines how much time the task has to react to the cancellation signal before it is focefully cancelled.

The receive timeout cannot be overriden by setting a higher deadline on the git push request.

Default is 4 minutes. If no unit is specified, milliseconds is assumed.

receive.cancellationTimeout

Defines the time that a receive task has to react to a cancellation signal and finish gracefully after receive.timeout is exceeded. If the receive task is still not terminated after the cancellation timeout is exceeded the task is forcefully cancelled. Values can be specified using standard time unit abbreviations ('ms', 'sec', 'min', etc.).

Default is 5 seconds. If no unit is specified, milliseconds is assumed.

receive.trustedKey

List of GPG key fingerprints that should be considered trust roots by the server when signed push validation is enabled. A key is trusted by the server if it is either in this list, or a path of trust signatures leads from the key to a configured trust root. The maximum length of the path is determined by receive.maxTrustDepth.

Key fingerprints can be displayed with gpg --list-keys --with-fingerprint.

Trust signatures can be added to a key using the tsign command to gpg --edit-key, after which the signed key should be re-uploaded.

If no keys are specified, web-of-trust checks are disabled. This is the default behavior.

Section repository

Repositories in this sense are the same as projects.

In the following example configuration Registered Users is set to be the default owner of new projects.

[repository "*"]
  ownerGroup = Registered Users

The only matching patterns supported are exact match or wildcard matching which can be specified by ending the name with a *. If a project matches more than one repository configuration, then the configuration from the more precise match will be used. In the following example, the default submit type for a project named project/plugins/a would be CHERRY_PICK.

[repository "project/*"]
  defaultSubmitType = MERGE_IF_NECESSARY
[repository "project/plugins/*"]
  defaultSubmitType = CHERRY_PICK
Note
All properties are used from the matching repository configuration. In the previous example, all properties will be used from project/plugins/* section and no properties will be inherited nor overridden from project/*.
repository.<name>.basePath

Alternate to gerrit.basePath. The repository will be created and used from this location instead: ${alternateBasePath}/${projectName}.git.

If configuring the basePath for an existing project in gerrit, make sure to stop gerrit, move the repository in the alternate basePath, configure basePath for this repository and then start Gerrit.

Path must be absolute.

repository.<name>.defaultSubmitType

The default submit type for newly created projects. Supported values are INHERIT, MERGE_IF_NECESSARY, FAST_FORWARD_ONLY, REBASE_IF_NECESSARY, REBASE_ALWAYS, MERGE_ALWAYS and CHERRY_PICK.

For more details see Submit Types.

Default is INHERIT.

This submit type is only applied at project creation time if a submit type is omitted from the ProjectInput. If the submit type is unset in the project config at runtime, for backwards compatibility purposes, it defaults to MERGE_IF_NECESSARY rather than INHERIT.

repository.<name>.ownerGroup

A name of a group which exists. Zero, one or many groups are allowed. Each on its own line. Groups which don’t exist are ignored.

Section retry

retry.maxWait

Maximum time to wait between attempts to retry an operations when one attempt fails (e.g. on NoteDb updates due to contention, aka lock failure, on the underlying ref storage). Operations are retried with exponential backoff, plus some random jitter, until the interval reaches this limit. After that, retries continue to occur after a fixed timeout (plus jitter), up to retry.timeout.

Defaults to 5 seconds; unit suffixes are supported, and assumes milliseconds if not specified.

retry.timeout

Total timeout for retrying operations when one attempt fails.

It is possible to overwrite this default timeout based on operation types by setting retry.<operationType>.timeout.

Defaults to 20 seconds; unit suffixes are supported, and assumes milliseconds if not specified.

retry.<operationType>.timeout

Total timeout for retrying operations of type <operationType> when one attempt fails. <operationType> can be ACCOUNT_UPDATE, CHANGE_UPDATE, GROUP_UPDATE and INDEX_QUERY.

Defaults to retry.timeout; unit suffixes are supported, and assumes milliseconds if not specified.

retry.retryWithTraceOnFailure

Whether Gerrit should automatically retry operations on failure with tracing enabled. The automatically generated traces can help with debugging.

By default this is set to false.

Section rules

rules.enable

If true, Gerrit will load and execute 'rules.pl' files in each project’s refs/meta/config branch, if present. When set to false, only the default internal rules will be used.

Default is true, to execute project specific rules.

rules.reductionLimit

Maximum number of Prolog reductions that can be performed when evaluating rules for a single change. Each function call made in user rule code, internal Gerrit Prolog code, or the Prolog interpreter counts against this limit.

Sites using very complex rules that need many reductions should compile Prolog to Java bytecode with rulec. This eliminates the dynamic Prolog interpreter from charging its own reductions against the limit, enabling more logic to execute within the same bounds.

A reductionLimit of 0 is nearly infinite, implemented by setting the internal limit to 2^31-1.

Default is 100,000 reductions (about 14 ms on Intel Core i7 CPU).

rules.compileReductionLimit

Maximum number of Prolog reductions that can be performed when compiling source code to internal Prolog machine code.

Default is 10x reductionLimit (1,000,000).

rules.maxSourceBytes

Maximum input size (in bytes) of a Prolog rules.pl file. Larger source files may need a larger rules.compileReductionLimit. Consider using rulec to precompile larger rule files.

A size of 0 bytes disables rules, same as rules.enable = false.

Common unit suffixes of 'k', 'm', or 'g' are supported.

Default is 128 KiB.

rules.maxPrologDatabaseSize

Number of predicate clauses allowed to be defined in the Prolog database by project rules. Very complex rules may need more than the default 256 limit, but cost more memory and may need more time to evaluate. Consider using rulec to precompile larger rule files.

Default is 256.

Section execution

execution.defaultThreadPoolSize

The default size of the background execution thread pool in which miscellaneous tasks are handled.

Default and minimum is 2 so that a single, potentially longer executing task (e.g. GC), is not blocking the entire execution.

execution.fanOutThreadPoolSize

Maximum size of thread pool to on which a serving thread can fan-out work to parallelize it.

When set to 0, a direct executor will be used.

By default, 25 which means that formatting happens in the caller thread.

Section receiveemail

receiveemail.protocol

Specifies the protocol used for receiving emails. Valid options are 'POP3', 'IMAP' and 'NONE'. Note that Gerrit will automatically switch between POP3 and POP3s as well as IMAP and IMAPS depending on the specified encryption.

Defaults to 'NONE' which means that receiving emails is disabled.

receiveemail.host

The hostname of the mailserver. Example: 'imap.gmail.com'.

Defaults to an empty string which means that receiving emails is disabled.

receiveemail.port

The port the email server exposes for receiving emails.

Defaults to the industry standard for a given protocol and encryption: POP3: 110; POP3S: 995; IMAP: 143; IMAPS: 993.

receiveemail.username

Username used for authenticating with the email server.

Defaults to an empty string.

receiveemail.password

Password used for authenticating with the email server.

Defaults to an empty string.

receiveemail.encryption

Encryption standard used for transport layer security between Gerrit and the email server. Possible values include 'NONE', 'SSL' and 'TLS'.

Defaults to 'NONE'.

receiveemail.fetchInterval

Time between two consecutive fetches from the email server. Communication with the email server is not kept alive. Examples: 60s, 10m, 1h.

Defaults to 60 seconds.

receiveemail.enableImapIdle

If the IMAP protocol is used for retrieving emails, IMAPv4 IDLE can be used to keep the connection with the email server alive and receive a push when a new email is delivered to the inbox. In this case, Gerrit will process the email immediately and will not have a fetch delay.

Defaults to false.

receiveemail.filter.mode

An allow and block filter to filter incoming emails.

If OFF, emails are not filtered by the list filter.

If ALLOW, only emails where a pattern from receiveemail.filter.patterns matches 'From' will be processed.

If BLOCK, only emails where no pattern from receiveemail.filter.patterns matches 'From' will be processed.

Defaults to OFF.

The previous filter-names 'BLACKLIST' and 'WHITELIST' have been deprecated since they may be considered disrespectful and there’s no technical or practical reason to use these exact terms for the filters. For backwards compatibility they are still supported but support for these deprecated terms will be removed in future releases.

receiveemail.filter.patterns

A list of regular expressions to match the email sender against. This can also be a list of addresses when regular expression characters are escaped.

Section sendemail

sendemail.enable

If false Gerrit will not send email messages, for any reason, and all other properties of section sendemail are ignored.

By default, true, allowing notifications to be sent.

sendemail.html

If false, Gerrit will only send plain-text emails. If true, Gerrit will send multi-part emails with an HTML and plain text part.

By default, true, allowing HTML in the emails Gerrit sends.

sendemail.connectTimeout

The connection timeout of opening a socket connected to a remote SMTP server.

Values can be specified using standard time unit abbreviations ('ms', 'sec', 'min', etc.). If no unit is specified, milliseconds is assumed.

Default is 0. A timeout of zero is interpreted as an infinite timeout. The connection will then block until established or an error occurs.

sendemail.threadPoolSize

Maximum size of thread pool in which the review comments notifications are sent out asynchronously.

By default, 1.

sendemail.from

Designates what name and address Gerrit will place in the From field of any generated email messages. The supported values are:

  • USER

    Gerrit will set the From header to use the current user’s Full Name and Preferred Email. This may cause messages to be classified as spam if the user’s domain has SPF or DKIM enabled and sendemail.smtpServer is not a trusted relay for that domain. You can specify sendemail.allowedDomain to instruct Gerrit to only send as USER if USER is from those domains.

  • MIXED

    Shorthand for ${user} (Code Review) <review@example.com> where review@example.com is the same as user.email. See below for a description of how the replacement is handled.

  • SERVER

    Gerrit will set the From header to the same name and address it records in any commits Gerrit creates. This is set by user.name and user.email, or guessed from the local operating system.

  • Code Review <review@example.com>

    If set to a name and email address in brackets, Gerrit will use this name and email address for any messages, overriding the name that may have been selected for commits by user.name and user.email. Optionally, the name portion may contain the placeholder ${user}, which is replaced by the Full Name of the current user.

By default, MIXED.

sendemail.allowedDomain

Only used when sendemail.from is set to USER. List of allowed domains. If user’s email matches one of the domains, emails will be sent as USER, otherwise as MIXED mode. Wildcards may be specified by including * to match any number of characters, for example *.example.com matches any subdomain of example.com.

By default, *.

sendemail.smtpServer

Hostname (or IP address) of a SMTP server that will relay messages generated by Gerrit to end users.

By default, 127.0.0.1 (aka localhost).

sendemail.smtpServerPort

Port number of the SMTP server in sendemail.smtpserver.

By default, 25, or 465 if smtpEncryption is 'ssl'.

sendemail.smtpEncryption

Specify the encryption to use, either 'ssl' or 'tls'.

By default, 'none', indicating no encryption is used.

sendemail.sslVerify

If false and sendemail.smtpEncryption is 'ssl' or 'tls', Gerrit will not verify the server certificate when it connects to send an email message.

By default, true, requiring the certificate to be verified.

sendemail.smtpUser

User name to authenticate with, if required for relay.

sendemail.smtpPass

Password for the account named by sendemail.smtpUser.

sendemail.allowrcpt

If present, each value adds one entry to the list of allowed email addresses that Gerrit can send emails to. If set to a complete email address, that one address is added to the list of allowed emails. If set to a domain name, any address at that domain can receive email from Gerrit.

If allowrcpt is configured, The set of allowed recipients is: allowrcpt - denyrcpt.

By default, unset, permitting delivery to any email address.

sendemail.denyrcpt

If present, each value adds one entry to the list of email addresses that Gerrit can’t send emails to. If set to a complete email address, that one address is added to the list of blocked emails. If set to a domain name, any address at that domain can not receive email from Gerrit.

By default, unset, permitting delivery to any email address.

sendemail.includeDiff

If true, new change emails and merged change emails from Gerrit will include the complete unified diff of the change. Variable maxmimumDiffSize places an upper limit on how large the email can get when this option is enabled.

By default, false.

sendemail.maximumDiffSize

Largest size of unified diff output to include in an email. When the diff exceeds this size the file paths will be listed instead. Standard byte unit suffixes are supported.

By default, 256 KiB.

sendemail.importance

If present, emails sent from Gerrit will have the given level of importance. Valid values include 'high' and 'low', which email clients will render in different ways.

By default, unset, so no Importance header is generated.

sendemail.expiryDays

If present, emails sent from Gerrit will expire after the given number of days. This will add the Expiry-Date header and email clients may expire or expunge mails whose Expiry-Date header is in the past. This should be a positive non-zero number indicating how many days in the future the mails should expire.

By default, unset, so no Expiry-Date header is generated.

sendemail.replyToAddress

A custom Reply-To address should only be provided if Gerrit is set up to receive emails and the inbound address differs from sendemail.from. It will be set as Reply-To header on all types of outgoing email where Gerrit can parse back a user’s reply.

Defaults to an empty string which adds sendemail.from as Reply-To if inbound email is enabled and the review’s author otherwise.

sendemail.allowTLD

List of custom TLDs to allow sending emails to in addition to those specified in the IANA list.

Defaults to an empty list, meaning no additional TLDs are allowed.

sendemail.addInstanceNameInSubject

When set to true, Gerrit will add its short name to the email subject, allowing recipients to quickly identify what Gerrit instance the email came from.

The short name can be customized via the gerrit.instanceName option.

Defaults to false.

Section site

site.allowOriginRegex

List of regular expressions matching origins that should be permitted to use the full Gerrit REST API. These should be trusted applications, as the sites may be able to use the user’s credentials. Applies to all requests, including state changing methods (PUT, DELETE, POST).

Expressions should not require trailing slash. For example a valid pattern might be .example[.]com.

By default, unset, denying all cross-origin requests.

site.refreshHeaderFooter

If true the server checks the site header, footer and CSS files for updated versions. If false, a server restart is required to change any of these resources. Default is true, allowing automatic reloads.

Section ssh-alias

Variables in section ssh-alias permit the site administrator to alias another command from Gerrit or a plugin into the gerrit command namespace. To alias replication start to gerrit replicate:

[ssh-alias]
  replicate = replication start

Section sshd

sshd.enableCompression

In the general case, we want to disable transparent compression, since the majority of our data transfer is highly compressed Git pack files and we cannot make them any smaller than they already are.

However, if there are CPU in abundance and the server is reachable through slow networks, gits with huge amount of refs can benefit from SSH-compression since git does not compress the ref announcement during handshake.

Compression can be especially useful when Gerrit replicas are being used for the larger clones and fetches and the primary server mostly takes small receive-packs.

By default, false.

sshd.backend

Starting from version 0.9.0 Apache SSHD project added support for NIO2 IoSession. To use the old MINA session the backend option must be set to MINA.

By default, NIO2.

sshd.listenAddress

Specifies the local addresses the internal SSHD should listen for connections on. The following forms may be used to specify an address. In any form, :'port' may be omitted to use the default of 29418.

  • 'hostname':'port' (for example review.example.com:29418)

  • 'IPv4':'port' (for example 10.0.0.1:29418)

  • ['IPv6']:'port' (for example [ff02::1]:29418)

  • *:'port' (for example *:29418)

If multiple values are supplied, the daemon will listen on all of them.

To disable the internal SSHD, set listenAddress to off.

By default, *:29418.

sshd.advertisedAddress

Specifies the addresses clients should be told to connect to. This may differ from sshd.listenAddress if a firewall based port redirector is being used, making Gerrit appear to answer on port 22. The following forms may be used to specify an address. In any form, :'port' may be omitted to use the default SSH port of 22.

  • 'hostname':'port' (for example review.example.com:22)

  • 'IPv4':'port' (for example 10.0.0.1:29418)

  • ['IPv6']:'port' (for example [ff02::1]:29418)

If multiple values are supplied, the daemon will advertise all of them.

By default uses the value of sshd.listenAddress.

sshd.tcpKeepAlive

If true, enables TCP keepalive messages to the other side, so the daemon can terminate connections if the peer disappears.

Only effective when sshd.backend is set to MINA.

By default, true.

sshd.threads

Number of threads to use when executing SSH command requests. If additional requests are received while all threads are busy they are queued and serviced in a first-come-first-served order.

By default, 2x the number of CPUs available to the JVM (but at least 4 threads).

Note
When SSH daemon is enabled then this setting also defines the max number of concurrent Git requests for interactive users over SSH and HTTP together.
sshd.batchThreads

Number of threads to allocate for SSH command requests from service users. If equals to 0, then all non-interactive requests are executed in the same queue as interactive requests.

Any other value will remove the number of threads from the queue allocated to interactive users, and create a separate thread pool of the requested size, which will be used to run commands from service users.

If the number of threads requested for service users is larger than the total number of threads allocated in sshd.threads, then the value of sshd.threads is increased to accommodate the requested value.

By default is 1 on single core node, 2 otherwise.

Note
When SSH daemon is enabled then this setting also defines the max number of concurrent Git requests for batch users over SSH and HTTP together.
sshd.streamThreads

Number of threads to use when formatting events to asynchronous streaming clients. Event formatting is multiplexed onto this thread pool by a simple FIFO scheduling system.

By default, 1 plus the number of CPUs available to the JVM.

sshd.commandStartThreads

Number of threads used to parse a command line submitted by a client over SSH for execution, create the internal data structures used by that command, and schedule it for execution on another thread.

By default, 2.

sshd.maxAuthTries

Maximum number of authentication attempts before the server disconnects the client. Each public key that a client has loaded into its local agent counts as one auth request. Users can work around the server’s limit by loading less keys into their agent, or selecting a specific key in their ~/.ssh/config file with the IdentityFile option.

By default, 6.

sshd.loginGraceTime

Time in seconds that a client has to authenticate before the server automatically terminates their connection. Values should use common unit suffixes to express their setting:

  • s, sec, second, seconds

  • m, min, minute, minutes

  • h, hr, hour, hours

  • d, day, days

By default, 2 minutes.

sshd.idleTimeout

Time in seconds after which the server automatically terminates idle connections (or 0 to disable closing of idle connections) not waiting for any server operation to complete. Values should use common unit suffixes to express their setting:

  • s, sec, second, seconds

  • m, min, minute, minutes

  • h, hr, hour, hours

  • d, day, days

By default, 0.

sshd.waitTimeout

Time in seconds after which the server automatically terminates connections waiting for a server operation to complete, like for instance cloning a very large repo with lots of refs. Values should use common unit suffixes to express their setting:

  • s, sec, second, seconds

  • m, min, minute, minutes

  • h, hr, hour, hours

  • d, day, days

By default, 30s.

sshd.gracefulStopTimeout

Set a graceful stop time. If set, Gerrit ensures that all open SSH sessions are preserved for a maximum period of time, before forcing the shutdown of the SSH daemon. During this period, no new requests will be accepted. This option is meant to be used in setups performing rolling restarts.

Values should use common unit suffixes to express their setting:

  • s, sec, second, seconds

  • m, min, minute, minutes

    By default, 0 seconds (immediate shutdown).

sshd.maxConnectionsPerUser

Maximum number of concurrent SSH sessions that a user account may open at one time. This is the number of distinct SSH logins that each user may have active at one time, and is not related to the number of commands a user may issue over a single connection. If set to 0, there is no limit.

By default, 64.

sshd.cipher

Available ciphers. To permit multiple ciphers, specify multiple sshd.cipher keys in the configuration file, one cipher name per key. Cipher names starting with + are enabled in addition to the default ciphers, cipher names starting with - are removed from the default cipher set.

Supported ciphers:

  • aes128-cbc

  • aes128-ctr

  • aes128-gcm@openssh.com

  • aes192-cbc

  • aes192-ctr

  • aes256-cbc

  • aes256-ctr

  • aes256-gcm@openssh.com

  • arcfour128

  • arcfour256

  • blowfish-cbc

  • chacha20-poly1305@openssh.com

  • 3des-cbc

  • none

    If your setup allows for it, it’s recommended to disable all ciphers except the AES-CTR modes.

    See also ciphers.

    By default, all supported ciphers except none are available.

sshd.mac

Available MAC (message authentication code) algorithms. To permit multiple algorithms, specify multiple sshd.mac keys in the configuration file, one MAC per key. MAC names starting with + are enabled in addition to the default MACs, MAC names starting with - are removed from the default MACs.

Supported MACs:

  • hmac-md5

  • hmac-md5-96

  • hmac-sha1

  • hmac-sha1-96

  • hmac-sha2-256

  • hmac-sha2-512

  • hmac-sha1-etm@openssh.com

  • hmac-sha2-256-etm@openssh.com

  • hmac-sha2-512-etm@openssh.com

    See also macs.

    By default, all supported MACs are available.

sshd.enableDeprecatedKexAlgorithms

Enable deprecated kex algorithms:

  • diffie-hellman-group1-sha1

  • diffie-hellman-group14-sha1

  • diffie-hellman-group-exchange-sha1

By default, the deprecated kex algorithms are disabled.

sshd.kex

Available key exchange algorithms. To permit multiple algorithms, specify multiple sshd.kex keys in the configuration file, one key exchange algorithm per key. Key exchange algorithm names starting with + are enabled in addition to the default key exchange algorithms, key exchange algorithm names starting with - are removed from the default key exchange algorithms.

Supported key exchange algorithms:

  • ecdh-sha2-nistp521

  • ecdh-sha2-nistp384

  • ecdh-sha2-nistp256

  • curve25519-sha256

  • curve25519-sha256@libssh.org

  • curve448-sha512

  • diffie-hellman-group-exchange-sha256

  • diffie-hellman-group18-sha512

  • diffie-hellman-group17-sha512

  • diffie-hellman-group16-sha512

  • diffie-hellman-group15-sha512

  • diffie-hellman-group14-sha256

See sshd.enableDeprecatedKexAlgorithms for deprecated key algorithms and how to enable them.

It is strongly recommended to disable at least diffie-hellman-group1-sha1 as it’s known to be vulnerable (logjam attack). Additionally, if your setup allows for it, it is recommended to disable the remaining two sha1 key exchange algorithms.

See also key exchange.

By default, all supported key exchange algorithms are available.

sshd.kerberosKeytab

Enable kerberos authentication for SSH connections. To permit kerberos authentication, the server must have a host principal (see sshd.kerberosPrincipal) which is acquired from a keytab. This must be provisioned by the kerberos administrators, and is typically installed into /etc/krb5.keytab on host machines.

The keytab must contain at least one host/ principal, typically using the host’s canonical name. If it does not use the canonical name, the sshd.kerberosPrincipal should be configured with the correct name.

By default, not set and so kerberos authentication is not enabled.

sshd.kerberosPrincipal

If kerberos authentication is enabled with sshd.kerberosKeytab, instead use the given principal name instead of the default. If the principal does not begin with host/ a warning message is printed and may prevent successful authentication.

This may be useful if the host is behind an IP load balancer or other SSH forwarding systems, since the principal name is constructed by the client and must match for kerberos authentication to work.

By default, host/canonical.host.name

sshd.requestLog

Enable (or disable) the '$site_path'/logs/sshd_log request log. If enabled, a request log file is written out by the SSH daemon. The sshd log format is documented here.

log4j.appender with the name sshd_log can be configured to overwrite programmatic configuration.

By default, true.

This value supports configuration reloads.

sshd.rekeyBytesLimit

The SSH daemon will issue a rekeying after a certain amount of data. This configuration option allows you to tweak that setting.

By default, 1073741824 (bytes, 1GB).

The rekeyBytesLimit cannot be set to lower than 32.

sshd.rekeyTimeLimit

The SSH daemon will issue a rekeying after a certain amount of time. This configuration option allows you to tweak that setting.

By default, 1h.

Set to 0 to disable this check.

Section suggest

suggest.maxSuggestedReviewers

The maximum numbers of reviewers suggested.

By default 10.

This value supports configuration reloads.

suggest.from

The number of characters that a user must have typed before suggestions are provided. If set to 0, suggestions are always provided. This is only used for suggesting accounts when adding members to a group.

By default 0.

suggest.relevantChanges

When suggesting reviewers, we go over recent changes of the user, and give priority to users that are present as reviewers in any of those changes. The number of changes we go over is sugggest.relevantChanges.

This nubmer is a tradeoff between speed and accuracy. A high number would be accurate but slow, and a low number would be fast but inaccurate.

By default 50.

suggest.skipServiceUsers

If service users should be skipped when suggesting reviewers.

By default true.

Section tracing

tracing.performanceLogging

Whether performance logging is enabled.

When performance logging is enabled, performance events for some operations are collected in memory while a request is running. At the end of the request the performance events are handed over to the PerformanceLogger plugins. This means if performance logging is enabled, the memory footprint of requests can be markedly increased. In one recorded case the impact was an overall heap increase of 40% (using the metrics-reporter-graphite plugin), in other instances the heap increase wasn’t nearly as dramatic and the impact is most likely dependent on which plugin is used.

By default, false.

tracing.exportPerformanceMetrics

Whether to export performance metrics.

Performace logged when performanceLogging is enabled, can be exported as metrics.

Note
Since the payload returned could be of tens of thousands metrics, assess the latency of the metrics endpoint before enabling this option.

By default, false.

Subsection tracing.<trace-id>

There can be multiple tracing.<trace-id> subsections to configure automatic tracing of requests. To be traced a request must match all conditions of one tracing.<trace-id> subsection. The subsection name is used as trace ID. Using this trace ID administrators can find matching log entries.

tracing.<trace-id>.requestType

Type of request for which request tracing should be always enabled (can be GIT_RECEIVE, GIT_UPLOAD, REST and SSH).

May be specified multiple times.

By default, unset (all request types are matched).

tracing.<trace-id>.requestUriPattern

Regular expression to match request URIs for which request tracing should be enabled except if they match excludedRequestUriPattern. Request URIs are only available for REST requests. Request URIs never include the '/a' prefix.

May be specified multiple times.

By default, unset (all request URIs are matched).

tracing.<trace-id>.excludedRequestUriPattern

Regular expression to match request URIs for which request tracing should not be enabled even if they match requestUriPattern. Request URIs are only available for REST requests. Request URIs never include the '/a' prefix.

May be specified multiple times.

By default, unset (no request URIs are excluded).

tracing.<trace-id>.account

Account ID of an account for which request tracing should be always enabled.

May be specified multiple times.

By default, unset (all accounts are matched).

tracing.<trace-id>.projectPattern

Regular expression to match project names for which request tracing should be always enabled.

May be specified multiple times.

By default, unset (all projects are matched).

Subsection deadline.<id>

There can be multiple deadline.<id> subsections to configure deadlines for request executions. For a deadline to apply all conditions of the deadline.<id> subsection must match. The subsection name is the ID of the deadline configuration and allows to track back an applied deadline to its configuration.

Clients can override the deadlines that are configured here by setting a deadline on the request.

Deadlines are only supported for REST, SSH and GIT_RECEIVE requests, but not for GIT_UPLOAD requests.

deadline.<id>.timeout

Timeout after which matching requests should be cancelled.

Values must be specified using standard time unit abbreviations ('ms', 'sec', 'min', etc.).

For some requests additional timeout configurations may apply, e.g. receive.timeout for git pushes.

By default, unset.

deadline.<id>.isAdvisory

Whether this deadline is an advisory deadline. Advisory deadlines do not cause requests to be aborted when they are exceeded. Instead, if an advisory deadline is exceeded, only the cancellation/advisory_deadline_count metrics is incremented and a log is written. This is useful to test how many requests would be affected by a new deadline configuration.

By default, false.

deadline.<id>.requestType

Type of request to which the deadline applies (can be GIT_RECEIVE, REST and SSH).

May be specified multiple times.

By default, unset (all request types are matched).

deadline.<id>.requestUriPattern

Regular expression to match request URIs to which the deadline applies except if they match excludedRequestUriPattern. Request URIs are only available for REST requests. Request URIs never include the '/a' prefix.

May be specified multiple times.

By default, unset (all request URIs are matched).

deadline.<id>.excludedRequestUriPattern

Regular expression to match request URIs to which the deadline should not be applied even if they match requestUriPattern. Request URIs are only available for REST requests. Request URIs never include the '/a' prefix.

May be specified multiple times.

By default, unset (no request URIs are excluded).

deadline.<id>.account

Account ID of an account to which the deadline applies.

May be specified multiple times.

By default, unset (all accounts are matched).

deadline.<id>.projectPattern

Regular expression to match project names to which the deadline applies.

May be specified multiple times.

By default, unset (all projects are matched).

Section trackingid

Tagged footer lines containing references to external tracking systems, parsed out of the commit message and saved in Gerrit’s secondary index.

After making changes to this section, existing changes must be reindexed with reindex.

The tracking ids are searchable using tr:<tracking id> or bug:<tracking id>.

[trackingid "jira-bug"]
  footer = Bugfix:
  footer = Bug:
  match = JRA\\d{2,8}
  system = JIRA

[trackingid "jira-feature"]
  footer = Feature
  match = JRA(\\d{2,8})
  system = JIRA
trackingid.<name>.footer

A prefix tag that identifies the footer line to parse for tracking ids.

Several trackingid entries can have the same footer tag, and a single trackingid entry can have multiple footer tags.

If multiple footer tags are specified, each tag will be parsed separately and duplicates will be ignored.

The trailing ":" is optional.

trackingid.<name>.match

A standard Java regular expression (java.util.regex) used to match the external tracking id part of the footer line. The match can result in several entries in the DB. If grouping is used in the regex the first group will be interpreted as the tracking id. Tracking ids longer than 32 characters will be ignored.

The configuration file parser eats one level of backslashes, so the character class \s requires \\s in the configuration file. The parser also terminates the line at the first #, so a match expression containing # must be wrapped in double quotes.

trackingid.<name>.system

The name of the external tracking system (maximum 20 characters). It is possible to have several trackingid entries for the same tracking system.

Section transfer

transfer.timeout

Number of seconds to wait for a single network read or write to complete before giving up and declaring the remote side is not responding. If 0, there is no timeout, and this server will wait indefinitely for a transfer to finish.

A timeout should be large enough to mostly transfer the objects to the other side. 1 second may be too small for larger projects, especially over a WAN link, while 10-30 seconds is a much more reasonable timeout value.

Defaults to 0 seconds, wait indefinitely.

Section upload

Options to control the behavior of upload-pack on the server side, which handles a user’s fetch, clone, or repo sync command.

[upload]
  allowGroup = GROUP_ALLOWED_TO_EXECUTE
  allowGroup = YET_ANOTHER_GROUP_ALLOWED_TO_EXECUTE
upload.allowGroup

Name of the groups of users that are allowed to execute 'upload-pack'. One or more groups can be set.

If no groups are added, any user will be allowed to execute 'upload-pack' on the server.

Section accountDeactivation

Configures the parameters for the scheduled task to sweep and deactivate Gerrit accounts according to their status reported by the auth backend. Currently only supported for LDAP backends.

accountDeactivation.startTime

The start time for running account deactivations.

accountDeactivation.interval

The interval for running account deactivations.

Note that the task will only be scheduled if the auth.autoUpdateAccountActiveStatus is set to true.

Schedule examples can be found in the Schedule Configuration section.

Section submodule

submodule.verboseSuperprojectUpdate

When using automatic superproject updates this option will determine how the submodule commit messages are included into the commit message of the superproject update.

If FALSE, will not include any commit messages for the gitlink update.

If SUBJECT_ONLY, will include only the commit subjects.

If TRUE, will include full commit messages.

By default this is TRUE.

submodule.enableSuperProjectSubscriptions

This allows to enable the superproject subscription mechanism.

By default this is true.

submodule.maxCombinedCommitMessageSize

This allows to limit the length of the commit message for a submodule.

By default this is 262144 (256 KiB).

Common unit suffixes of k, m, or g are supported.

submodule.maxCommitMessages

This allows to limit the number of commit messages that should be combined when creating a commit message for a submodule.

By default this is 1000.

Section user

user.name

Name that Gerrit calls itself in Git when it creates a new Git commit, such as a merge during change submission.

By default this is "Gerrit Code Review".

user.email

Email address that Gerrit refers to itself as when it creates a new Git commit, such as a merge commit during change submission.

If not set, Gerrit generates this as "gerrit@hostname", where hostname is the hostname of the system Gerrit is running on.

By default, not set, generating the value at startup.

user.anonymousCoward

Username that is displayed in the Gerrit Web UI and in e-mail notifications if the full name of the user is not set.

By default "Name of user not set" is used.

Schedule Configuration

Schedule configurations are used for running periodic background jobs.

A schedule configuration consists of two parameters:

  • interval: Interval for running the periodic background job. The interval must be larger than zero. The following suffixes are supported to define the time unit for the interval:

    • s, sec, second, seconds

    • m, min, minute, minutes

    • h, hr, hour, hours

    • d, day, days

    • w, week, weeks (1 week is treated as 7 days)

    • mon, month, months (1 month is treated as 30 days)

    • y, year, years (1 year is treated as 365 days)

  • startTime: The start time defines the first execution of the periodic background job. If the configured interval is shorter than startTime - now the start time will be preponed by the maximum integral multiple of interval so that the start time is still in the future. startTime must have one of the following formats:

    • <day of week> <hours>:<minutes>

    • <hours>:<minutes>

      The placeholders can have the following values:

      • <day of week>: Mon, Tue, Wed, Thu, Fri, Sat, Sun

      • <hours>: 00-23

      • <minutes>: 00-59

    The time zone cannot be specified but is always the system default time zone. Hours must be zero-padded, i.e. 06:00 rather than 6:00.

The section (and optionally the subsection) in which the interval and startTime keys must be set depends on the background job for which a schedule should be configured. E.g. for the change cleanup job the keys must be set in the changeCleanup section:

  [changeCleanup]
    startTime = Fri 10:30
    interval  = 2 days

Examples for a schedule configuration:

  • Example 1:

      startTime = Fri 10:30
      interval  = 2 days

    Assuming that the server is started on Mon 07:00 then startTime - now is 4 days 3:30 hours. This is larger than the interval hence the start time is preponed by the maximum integral multiple of the interval so that start time is still in the future, i.e. preponed by 4 days. This yields a start time of Mon 10:30, next executions are Wed 10:30, Fri 10:30. etc.

  • Example 2:

      startTime = 06:00
      interval = 1 day

    Assuming that the server is started on Mon 07:00 then this yields the first run on Tuesday at 06:00 and a repetition interval of 1 day.

File etc/All-Projects/project.config

The optional file '$site_path'/etc/All-Projects/project.config provides defaults for configuration read from project.config in the All-Projects repo. Unlike gerrit.config, this file contains project-type configuration rather than server-type configuration.

Most administrators will not need this file, and should instead make commits to All-Projects to modify global config. However, a separate file can be useful when managing multiple Gerrit servers, since pushing changes to defaults using Puppet or a similar tool can be easier than scripting git updates to All-Projects.

The contents of the file are loaded each time the All-Projects project is reloaded. Updating the file requires either evicting the project cache or restarting the server.

Caveats:

  • The path from which the file is read corresponds to the name of the repo, which is configurable.

  • Although the file lives in a directory that shares a name with a repository, this directory is not a Git repository.

  • Only the file project.config is read from this directory to provide defaults; any other files in this directory, such as rules.pl, are ignored. (This behavior may change in the future.)

  • Group names listed in the access config in this file are resolved to UUIDs using the groups file in the repository, not in the config directory. As a result, setting ACLs in this file is not recommended.

File etc/secure.config

The optional file '$site_path'/etc/secure.config overrides (or supplements) the settings supplied by '$site_path'/etc/gerrit.config. The file should be readable only by the daemon process and can be used to contain private configuration entries that wouldn’t normally be exposed to everyone.

Sample etc/secure.config:

[auth]
  registerEmailPrivateKey = 2zHNrXE2bsoylzUqDxZp0H1cqUmjgWb6

[ldap]
  password = l3tm3srch

[httpd]
  sslKeyPassword = g3rr1t

[sendemail]
  smtpPass = sp@m

[remote "bar"]
  password = s3kr3t

File etc/peer_keys

The optional file '$site_path'/etc/peer_keys controls who can login as the 'Gerrit Code Review' user, required for the suexec command.

The format is one Base-64 encoded public key per line with optional comment, e.g.:

# Comments allowed at start of line
AAAAC3...51R== john@example.net
# Another comment
AAAAB5...21S== jane@example.net

Configurable Parameters

site_path

Local filesystem directory holding the site customization assets. Placing this directory under version control and/or backup is a good idea.

Files in this directory provide additional configuration.

Other files support site customization.

File etc/jgit.config

Gerrit uses the $site_path/etc/jgit.config file instead of the system-wide and user-global Git configuration for its runtime JGit configuration.

Sample etc/jgit.config file:

[core]
  trustFolderStat = false

Section gc

Options in section gc are used when command gerrit gc is used or scheduled via options gc.startTime and gc.interval.

gc.auto

When there are approximately more than this many loose objects in the repository, auto gc will pack them. Some commands use this command to perform a light-weight garbage collection from time to time. The default value is 6700.

Setting this to 0 disables not only automatic packing based on the number of loose objects, but any other heuristic auto gc will otherwise use to determine if there’s work to do, such as gc.autoPackLimit.

gc.autodetach

Makes auto gc run in a background thread. Default is true.

gc.autopacklimit

When there are more than this many packs that are not marked with *.keep file in the repository, auto gc consolidates them into one larger pack. The default value is 50. Setting this to 0 disables it. Setting gc.auto to 0 will also disable this.

gc.packRefs

This variable determines whether gc runs git pack-refs. The default is true.

gc.reflogExpire

Removes reflog entries older than this time; defaults to 90 days. The value "now" expires all entries immediately, and "never" suppresses expiration altogether.

gc.reflogExpireUnreachable

Removes reflog entries older than this time and not reachable from the current tip; defaults to 30 days. The value "now" expires all entries immediately, and "never" suppresses expiration altogether.

Section protocol

protocol.version

If set, the server will accept requests from a client attempting to communicate using the specified protocol version. Default is 2. If set in file etc/jgit.config this option will be used for all repositories of the site. It can be overridden for a given repository by configuring a different value in the repository’s config file.

Supported versions:

0

the original wire protocol.

1

the original wire protocol with the addition of a version string in the initial response from the server.

2

wire protocol version 2. Speeds up fetches from repositories with many refs by allowing the client to specify which refs to list before the server lists them.

Section receive

receive.autogc

By default, up to Gerrit 3.2 git-receive-pack will run auto gc after receiving data from git-push and updating refs. You can stop it by setting this variable to false. This is recommended in gerrit to avoid the additional load this creates. Instead schedule gc using gc.startTime and gc.interval or e.g. in a cron job that runs gc in a separate process. Since Gerrit 3.3 the init command will auto-configure git-receive-pack = false in etc/jgit.config if it wasn’t set manually and show a warning if it was set to true manually.