Maintenance Rake tasks

DETAILS: Tier: Free, Premium, Ultimate Offering: Self-managed

GitLab provides Rake tasks for general maintenance.

Gather GitLab and system information

This command gathers information about your GitLab installation and the system it runs on. These may be useful when asking for help or reporting issues. In a multi-node environment, run this command on nodes running GitLab Rails to avoid PostgreSQL socket errors.

• Linux package installations:

  sudo gitlab-rake gitlab:env:info

• Self-compiled installations:

  bundle exec rake gitlab:env:info RAILS_ENV=production

Example output:

System information
System:         Ubuntu 20.04
Proxy:          no
Current User:   git
Using RVM:      no
Ruby Version:   2.7.6p219
Gem Version:    3.1.6
Bundler Version:2.3.15
Rake Version:   13.0.6
Redis Version:  6.2.7
Sidekiq Version:6.4.2
Go Version:     unknown

GitLab information
Version:        15.5.5-ee
Revision:       5f5109f142d
Directory:      /opt/gitlab/embedded/service/gitlab-rails
DB Adapter:     PostgreSQL
DB Version:     13.8
URL:            https://app.gitaly.gcp.gitlabsandbox.net
HTTP Clone URL: https://app.gitaly.gcp.gitlabsandbox.net/some-group/some-project.git
SSH Clone URL:  git@app.gitaly.gcp.gitlabsandbox.net:some-group/some-project.git
Elasticsearch:  no
Geo:            no
Using LDAP:     no
Using Omniauth: yes
Omniauth Providers:

GitLab Shell
Version:        14.12.0
Repository storage paths:
- default:      /var/opt/gitlab/git-data/repositories
- gitaly:       /var/opt/gitlab/git-data/repositories
GitLab Shell path:              /opt/gitlab/embedded/service/gitlab-shell


Gitaly
- default Address:      unix:/var/opt/gitlab/gitaly/gitaly.socket
- default Version:      15.5.5
- default Git Version:  2.37.1.gl1
- gitaly Address:       tcp://10.128.20.6:2305
- gitaly Version:       15.5.5
- gitaly Git Version:   2.37.1.gl1

Show GitLab license information

DETAILS: Tier: Premium, Ultimate Offering: Self-managed

This command shows information about your GitLab license and how many seats are used. It is only available on GitLab Enterprise installations: a license cannot be installed into GitLab Community Edition.

These may be useful when raising tickets with Support, or for programmatically checking your license parameters.

• Linux package installations:

  sudo gitlab-rake gitlab:license:info

• Self-compiled installations:

  bundle exec rake gitlab:license:info RAILS_ENV=production

Example output:

Today's Date: 2020-02-29
Current User Count: 30
Max Historical Count: 30
Max Users in License: 40
License valid from: 2019-11-29 to 2020-11-28
Email associated with license: user@example.com

Check GitLab configuration

The gitlab:check Rake task runs the following Rake tasks:

• gitlab:gitlab_shell:check
• gitlab:gitaly:check
• gitlab:sidekiq:check
• gitlab:incoming_email:check
• gitlab:ldap:check
• gitlab:app:check

It checks that each component was set up according to the installation guide and suggest fixes for issues found. This command must be run from your application server and doesn't work correctly on component servers like Gitaly. If you're running Geo, see also the Geo Health check Rake task.

You may also have a look at our troubleshooting guides for:

• GitLab.
• Linux package installations.

Additionally you should also verify database values can be decrypted using the current secrets.

To run gitlab:check, run:

• Linux package installations:

  sudo gitlab-rake gitlab:check

• Self-compiled installations:

  bundle exec rake gitlab:check RAILS_ENV=production

Use SANITIZE=true for gitlab:check if you want to omit project names from the output.

Example output:

Checking Environment ...

Git configured for git user? ... yes
Has python2? ... yes
python2 is supported version? ... yes

Checking Environment ... Finished

Checking GitLab Shell ...

GitLab Shell version? ... OK (1.2.0)
Repo base directory exists? ... yes
Repo base directory is a symlink? ... no
Repo base owned by git:git? ... yes
Repo base access is drwxrws---? ... yes
post-receive hook up-to-date? ... yes
post-receive hooks in repos are links: ... yes

Checking GitLab Shell ... Finished

Checking Sidekiq ...

Running? ... yes

Checking Sidekiq ... Finished

Checking GitLab App...

Database config exists? ... yes
Database is SQLite ... no
All migrations up? ... yes
GitLab config exists? ... yes
GitLab config up to date? ... no
Cable config exists? ... yes
Resque config exists? ... yes
Log directory writable? ... yes
Tmp directory writable? ... yes
Init script exists? ... yes
Init script up-to-date? ... yes
Redis version >= 2.0.0? ... yes

Checking GitLab ... Finished

Rebuild authorized_keys file

In some cases it is necessary to rebuild the authorized_keys file, for example, if after an upgrade you receive Permission denied (publickey) when pushing via SSH and find 404 Key Not Found errors in the gitlab-shell.log file. To rebuild authorized_keys, run:

• Linux package installations:

  sudo gitlab-rake gitlab:shell:setup

• Self-compiled installations:

  cd /home/git/gitlab
  sudo -u git -H bundle exec rake gitlab:shell:setup RAILS_ENV=production

Example output:

This will rebuild an authorized_keys file.
You will lose any data stored in authorized_keys file.
Do you want to continue (yes/no)? yes

Clear Redis cache

If for some reason the dashboard displays the wrong information, you might want to clear Redis' cache. To do this, run:

• Linux package installations:

  sudo gitlab-rake cache:clear

• Self-compiled installations:

  cd /home/git/gitlab
  sudo -u git -H bundle exec rake cache:clear RAILS_ENV=production

Precompile the assets

Sometimes during version upgrades you might end up with some wrong CSS or missing some icons. In that case, try to precompile the assets again.

This Rake task only applies to self-compiled installations. Read more about troubleshooting this problem when running the Linux package. The guidance for Linux package might be applicable for Kubernetes and Docker deployments of GitLab, though in general, container-based installations don't have issues with missing assets.

• Self-compiled installations:

  cd /home/git/gitlab
  sudo -u git -H bundle exec rake gitlab:assets:compile RAILS_ENV=production

For Linux package installations, the unoptimized assets (JavaScript, CSS) are frozen at the release of upstream GitLab. The Linux package installation includes optimized versions of those assets. Unless you are modifying the JavaScript / CSS code on your production machine after installing the package, there should be no reason to redo rake gitlab:assets:compile on the production machine. If you suspect that assets have been corrupted, you should reinstall the Linux package.

Check TCP connectivity to a remote site

Sometimes you need to know if your GitLab installation can connect to a TCP service on another machine (for example a PostgreSQL or web server) to troubleshoot proxy issues. A Rake task is included to help you with this.

• Linux package installations:

  sudo gitlab-rake gitlab:tcp_check[example.com,80]

• Self-compiled installations:

  cd /home/git/gitlab
  sudo -u git -H bundle exec rake gitlab:tcp_check[example.com,80] RAILS_ENV=production

Clear exclusive lease (DANGER)

GitLab uses a shared lock mechanism: ExclusiveLease to prevent simultaneous operations in a shared resource. An example is running periodic garbage collection on repositories.

In very specific situations, an operation locked by an Exclusive Lease can fail without releasing the lock. If you can't wait for it to expire, you can run this task to manually clear it.

To clear all exclusive leases:

WARNING: Don't run it while GitLab or Sidekiq is running

sudo gitlab-rake gitlab:exclusive_lease:clear

To specify a lease type or lease type + id, specify a scope:

# to clear all leases for repository garbage collection:
sudo gitlab-rake gitlab:exclusive_lease:clear[project_housekeeping:*]

# to clear a lease for repository garbage collection in a specific project: (id=4)
sudo gitlab-rake gitlab:exclusive_lease:clear[project_housekeeping:4]

Display status of database migrations

See the background migrations documentation for how to check that migrations are complete when upgrading GitLab.

To check the status of specific migrations, you can use the following Rake task:

sudo gitlab-rake db:migrate:status

To check the tracking database on a Geo secondary site, you can use the following Rake task:

sudo gitlab-rake db:migrate:status:geo

This outputs a table with a Status of up or down for each Migration ID.

database: gitlabhq_production

 Status   Migration ID    Migration Name
--------------------------------------------------
   up     migration_id    migration_name

Run incomplete database migrations

Database migrations can be stuck in an incomplete state, with a down status in the output of the sudo gitlab-rake db:migrate:status command.

1. To complete these migrations, use the following Rake task:

   sudo gitlab-rake db:migrate

2. After the command completes, run sudo gitlab-rake db:migrate:status to check if all migrations are completed (have an up status).

3. Hot reload puma and sidekiq services:

   sudo gitlab-ctl hup puma
   sudo gitlab-ctl restart sidekiq

Rebuild database indexes

DETAILS: Status: Experiment

WARNING: This feature is experimental, and isn't enabled by default. Use caution when running in a production environment, and run during off-peak times.

Database indexes can be rebuilt regularly to reclaim space and maintain healthy levels of index bloat over time. Reindexing can also be run as a regular cron job. A "healthy" level of bloat is highly dependent on the specific index, but generally should be below 30%.

Prerequisites:

• This feature requires PostgreSQL 12 or later.
• These index types are not supported: expression indexes, partitioned indexes, and indexes used for constraint exclusion.

To manually rebuild a database index:

1. Optional. To send annotations to a Grafana (4.6 or later) endpoint, enable annotations with these custom environment variables (see setting custom environment variables):

   1. GRAFANA_API_URL: The base URL for Grafana, such as http://some-host:3000.
   2. GRAFANA_API_KEY: A Grafana API key with at least Editor role.

2. Run the Rake task to rebuild the two indexes with the highest estimated bloat:

   sudo gitlab-rake gitlab:db:reindex

3. The reindexing task (gitlab:db:reindex) rebuilds only the two indexes in each database with the highest bloat. To rebuild more than two indexes, run the task again until all desired indexes have been rebuilt.

Notes

• Rebuilding database indexes is a disk-intensive task, so you should perform the task during off-peak hours. Running the task during peak hours can lead to increased bloat, and can also cause certain queries to perform slowly.
• The task requires free disk space for the index being restored. The created indexes are appended with _ccnew. If the reindexing task fails, re-running the task cleans up the temporary indexes.
• The time it takes for database index rebuilding to complete depends on the size of the target database. It can take between several hours and several days.

Dump the database schema

In rare circumstances, the database schema can differ from what the application code expects even if all database migrations are complete. If this does occur, it can lead to odd errors in GitLab.

To dump the database schema:

SCHEMA=/tmp/structure.sql gitlab-rake db:schema:dump

The Rake task creates a /tmp/structure.sql file that contains the database schema dump.

To determine if there are any differences:

1. Go to the db/structure.sql file in the gitlab project. Select the branch that matches your GitLab version. For example, the file for GitLab 16.2: https://gitlab.com/gitlab-org/gitlab/-/blob/16-2-stable-ee/db/structure.sql.
2. Compare /tmp/structure.sql with the db/structure.sql file for your version.

Check the database for schema inconsistencies

• Introduced in GitLab 15.11.

This Rake task checks the database schema for any inconsistencies and prints them in the terminal.

gitlab-rake gitlab:db:schema_checker:run

Import common metrics

Sometimes you may need to re-import the common metrics that power the Metrics dashboards.

This could be as a result of updating existing metrics.

To re-import the metrics you can run:

sudo gitlab-rake metrics:setup_common_metrics

Troubleshooting

Advisory lock connection information

After running the db:migrate Rake task, you may see output like the following:

main: == [advisory_lock_connection] object_id: 173580, pg_backend_pid: 5532
main: == [advisory_lock_connection] object_id: 173580, pg_backend_pid: 5532

The messages returned are informational and can be ignored.

PostgreSQL socket errors when executing the gitlab:env:info Rake task

After running sudo gitlab-rake gitlab:env:info on Gitaly or other non-Rails nodes , you might see the following error:

PG::ConnectionBad: could not connect to server: No such file or directory
Is the server running locally and accepting
connections on Unix domain socket "/var/opt/gitlab/postgresql/.s.PGSQL.5432"?

This is because, in a multi-node environment, the gitlab:env:info Rake task should only be executed on the nodes running GitLab Rails.