Processing specific job classes

WARNING: These are advanced settings. While they are used on GitLab.com, most GitLab instances should add more processes that all listen to all queues. This is the same approach we take in our Reference Architectures.

GitLab has two options for creating Sidekiq processes that only handle specific job classes:

Routing rules are used on GitLab.com. They direct jobs inside the application to queue names configured by administrators. This lowers the load on Redis, which is important on very large-scale deployments.
Queue selectors perform the job selection outside the application, when starting the Sidekiq process. This was used on GitLab.com until September 2021, and is retained for compatibility reasons.

Both of these use the same worker matching query syntax. While they can technically be used together, most deployments should choose one or the other; there is no particular benefit in combining them.

Routing rules must be the same across all GitLab nodes as they are part of the application configuration. Queue selectors can be different across GitLab nodes because they only change the arguments to the launched Sidekiq process.

Routing rules

Introduced in GitLab 13.12.

Default routing rule value added in GitLab 15.4.

NOTE: Mailer jobs cannot be routed by routing rules, and always go to the mailers queue. When using routing rules, ensure that at least one process is listening to the mailers queue. Typically this can be placed alongside the default queue.

We recommend most GitLab instances using routing rules to manage their Sidekiq queues. This allows administrators to choose single queue names for groups of job classes based on their attributes. The syntax is an ordered array of pairs of [query, queue]:

The query is a worker matching query.
The queue name must be a valid Sidekiq queue name. If the queue name is nil, or an empty string, the worker is routed to the queue generated by the name of the worker instead. (See list of available job classes for more information). The queue name does not have to match any existing queue name in the list of available job classes.
The first query matching a worker is chosen for that worker; later rules are ignored.

Routing rules migration

After the Sidekiq routing rules are changed, administrators must take care with the migration to avoid losing jobs entirely, especially in a system with long queues of jobs. The migration can be done by following the migration steps mentioned in Sidekiq job migration.

Detailed example

This is a comprehensive example intended to show different possibilities. It is not a recommendation.

Edit /etc/gitlab/gitlab.rb:

sidekiq['routing_rules'] = [
  # Route all non-CPU-bound workers that are high urgency to `high-urgency` queue
  ['resource_boundary!=cpu&urgency=high', 'high-urgency'],
  # Route all database, gitaly and global search workers that are throttled to `throttled` queue
  ['feature_category=database,gitaly,global_search&urgency=throttled', 'throttled'],
  # Route all workers having contact with outside world to a `network-intenstive` queue
  ['has_external_dependencies=true|feature_category=hooks|tags=network', 'network-intensive'],
  # Route all import workers to the queues generated by the worker name, for
  # example, JiraImportWorker to `jira_import`, SVNWorker to `svn_worker`
  ['feature_category=import', 'import'],
  # Wildcard matching, route the rest to `default` queue
  ['*', 'default']
]

The queue_groups can then be set to match these generated queue names. For instance:

sidekiq['queue_selector'] = false
sidekiq['queue_groups'] = [
  # Run two high-urgency processes
  'high-urgency',
  'high-urgency',
  # Run one process for throttled, network-intensive, import
  'throttled,network-intensive,import',
  # Run one 'catchall' process on the default and mailers queues
  'default,mailers'
]

Save the file and reconfigure GitLab:
```
sudo gitlab-ctl reconfigure
```

Queue selectors

Introduced in GitLab 12.8.

Sidekiq cluster, including queue selector, moved to GitLab Free in 12.10.

Renamed from experimental_queue_selector to queue_selector in GitLab 13.6.

The queue_selector option allows queue groups to be selected in a more general way using a worker matching query. After queue_selector is set, all queue_groups must follow the aforementioned syntax.

Using queue selectors

Edit /etc/gitlab/gitlab.rb:

sidekiq['enable'] = true
sidekiq['routing_rules'] = [['*', nil]]
sidekiq['queue_selector'] = true
sidekiq['queue_groups'] = [
  # Run all non-CPU-bound queues that are high urgency
  'resource_boundary!=cpu&urgency=high',
  # Run all continuous integration and pages queues that are not high urgency
  'feature_category=continuous_integration,pages&urgency!=high',
  # Run all queues
  '*'
]

Save the file and reconfigure GitLab:
```
sudo gitlab-ctl reconfigure
```

Negate settings

This allows you to have the Sidekiq process work on every queue except the ones you list. This is generally only used when there are multiple Sidekiq nodes. In this example, we exclude all import-related jobs from a Sidekiq node.

Edit /etc/gitlab/gitlab.rb:

sidekiq['routing_rules'] = [['*', nil]]
sidekiq['negate'] = true
sidekiq['queue_selector'] = true
sidekiq['queue_groups'] = [
   "feature_category=importers"
]

Save the file and reconfigure GitLab:
```
sudo gitlab-ctl reconfigure
```

Migrating from queue selectors to routing rules

We recommend GitLab deployments add more Sidekiq processes listening to all queues, as in the Reference Architectures. For very large-scale deployments, we recommend routing rules instead of queue selectors. We use routing rules on GitLab.com as it helps to lower the load on Redis.

To migrate from queue selectors to routing rules:

Open /etc/gitlab/gitlab.rb.
Set sidekiq['queue_selector'] to false.
Take all queue selectors in the sidekiq['queue_groups'].
Give each selector a queue_name and put them in [selector, queue_name] format.
Replace sidekiq['routing_rules'] with an array of [selector, queue_name] entries.
Add a wildcard match of ['*', 'default'] as the last entry in sidekiq['routing_rules']. This "catchall" queue has to be named as default.
Replace sidekiq['queue_groups'] with queue_names.
Add at least one default queue and at least one mailers queue to the sidekiq['queue_groups'].
Save the file and reconfigure GitLab:
```
sudo gitlab-ctl reconfigure
```

Run the Rake task to migrate existing jobs:

sudo gitlab-rake gitlab:sidekiq:migrate_jobs:retry gitlab:sidekiq:migrate_jobs:schedule gitlab:sidekiq:migrate_jobs:queued

NOTE: It is important to run the Rake task immediately after reconfiguring GitLab. After reconfiguring GitLab, existing jobs are not processed until the Rake task starts to migrate the jobs.

The following example better illustrates the migration process above:

Check the following content of /etc/gitlab/gitlab.rb:

sidekiq['routing_rules'] = []
sidekiq['queue_selector'] = true
sidekiq['queue_groups'] = [
  'urgency=high',
  'urgency=low',
  'urgency=throttled',
  '*'
]

Update /etc/gitlab/gitlab.rb to use routing rules:

sidekiq['min_concurrency'] = 20
sidekiq['max_concurrency'] = 20

sidekiq['routing_rules'] = [
  ['urgency=high', 'high_urgency'],
  ['urgency=low', 'low_urgency'],
  ['urgency=throttled', 'throttled_urgency'],
  # Wildcard matching, route the rest to `default` queue
  ['*', 'default']
]

sidekiq['queue_selector'] = false
sidekiq['queue_groups'] = [
  'high_urgency',
  'low_urgency',
  'throttled_urgency',
  'default,mailers'
]

Save the file and reconfigure GitLab:
```
sudo gitlab-ctl reconfigure
```

Run the Rake task to migrate existing jobs:

sudo gitlab-rake gitlab:sidekiq:migrate_jobs:retry gitlab:sidekiq:migrate_jobs:schedule gitlab:sidekiq:migrate_jobs:queued

WARNING: As described in the concurrency section, we recommend setting min_concurrency and max_concurrency to the same value. For example, if the number of queues in a queue group entry is 1, while min_concurrency is set to 0, and max_concurrency is set to 20, the resulting concurrency will be set to 2 instead. A concurrency of 2 might be too low in most cases, except for very highly-CPU bound tasks.

Worker matching query

GitLab provides a query syntax to match a worker based on its attributes. This query syntax is employed by both routing rules and queue selectors. A query includes two components:

Attributes that can be selected.
Operators used to construct a query.

Available attributes

Introduced in GitLab 13.1 (tags).

Queue matching query works upon the worker attributes, described in Sidekiq style guide. We support querying based on a subset of worker attributes:

feature_category - the GitLab feature category the queue belongs to. For example, the merge queue belongs to the source_code_management category.
has_external_dependencies - whether or not the queue connects to external services. For example, all importers have this set to true.
urgency - how important it is that this queue's jobs run quickly. Can be high, low, or throttled. For example, the authorized_projects queue is used to refresh user permissions, and is high urgency.
worker_name - the worker name. Use this attribute to select a specific worker. Find all available names in the job classes lists below.
name - the queue name generated from the worker name. Use this attribute to select a specific queue. Because this is generated from the worker name, it does not change based on the result of other routing rules.
resource_boundary - if the queue is bound by cpu, memory, or unknown. For example, the ProjectExportWorker is memory bound as it has to load data in memory before saving it for export.
tags - short-lived annotations for queues. These are expected to frequently change from release to release, and may be removed entirely.

has_external_dependencies is a boolean attribute: only the exact string true is considered true, and everything else is considered false.

tags is a set, which means that = checks for intersecting sets, and != checks for disjoint sets. For example, tags=a,b selects queues that have tags a, b, or both. tags!=a,b selects queues that have neither of those tags.

Available operators

Routing rules and queue selectors support the following operators, listed from highest to lowest precedence:

| - the logical OR operator. For example, query_a|query_b (where query_a and query_b are queries made up of the other operators here) includes queues that match either query.
& - the logical AND operator. For example, query_a&query_b (where query_a and query_b are queries made up of the other operators here) will only include queues that match both queries.
!= - the NOT IN operator. For example, feature_category!=issue_tracking excludes all queues from the issue_tracking feature category.
= - the IN operator. For example, resource_boundary=cpu includes all queues that are CPU bound.
, - the concatenate set operator. For example, feature_category=continuous_integration,pages includes all queues from either the continuous_integration category or the pages category. This example is also possible using the OR operator, but allows greater brevity, as well as being lower precedence.

The operator precedence for this syntax is fixed: it's not possible to make AND have higher precedence than OR.

As with the standard queue group syntax above, a single * as the entire queue group selects all queues.

List of available job classes

For a list of the existing Sidekiq job classes and queues, check the following files: