Troubleshooting Gitaly Cluster

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

Refer to the information below when troubleshooting Gitaly Cluster (Praefect). For information on troubleshooting Gitaly, see Troubleshooting Gitaly.

Check cluster health

The check Praefect sub-command runs a series of checks to determine the health of the Gitaly Cluster.

gitlab-ctl praefect check

If Praefect is deployed by using the Praefect chart, run the binary directly.

/usr/local/bin/praefect check

The following sections describe the checks that are run.

Praefect migrations

Because Database migrations must be up to date for Praefect to work correctly, checks if Praefect migrations are up to date.

If this check fails:

1. See the schema_migrations table in the database to see which migrations have run.
2. Run praefect sql-migrate to bring the migrations up to date.

Node connectivity and disk access

Checks if Praefect can reach all of its Gitaly nodes, and if each Gitaly node has read and write access to all of its storages.

If this check fails:

1. Confirm the network addresses and tokens are set up correctly:
   • In the Praefect configuration.
   • In each Gitaly node's configuration.
2. On the Gitaly nodes, check that the gitaly process being run as git. There might be a permissions issue that is preventing Gitaly from accessing its storage directories.
3. Confirm that there are no issues with the network that connects Praefect to Gitaly nodes.

Database read and write access

Checks if Praefect can read from and write to the database.

If this check fails:

1. See if the Praefect database is in recovery mode. In recovery mode, tables may be read only. To check, run:

   select pg_is_in_recovery()

2. Confirm that the user that Praefect uses to connect to PostgreSQL has read and write access to the database.

3. See if the database has been placed into read-only mode. To check, run:

   show default_transaction_read_only

Inaccessible repositories

Checks how many repositories are inaccessible because they are missing a primary assignment, or their primary is unavailable.

If this check fails:

1. See if any Gitaly nodes are down. Run praefect ping-nodes to check.
2. Check if there is a high load on the Praefect database. If the Praefect database is slow to respond, it can lead health checks failing to persist to the database, leading Praefect to think nodes are unhealthy.

Praefect errors in logs

If you receive an error, check /var/log/gitlab/gitlab-rails/production.log.

Here are common errors and potential causes:

• 500 response code
  • ActionView::Template::Error (7:permission denied)
    • praefect['configuration'][:auth][:token] and gitlab_rails['gitaly_token'] do not match on the GitLab server.
    • git_data_dirs storage configuration is missing on the Sidekiq server.
  • Unable to save project. Error: 7:permission denied
    • Secret token in praefect['configuration'][:virtual_storage] on GitLab server does not match the value in gitaly['auth_token'] on one or more Gitaly servers.
• 503 response code
  • GRPC::Unavailable (14:failed to connect to all addresses)
    • GitLab was unable to reach Praefect.
  • GRPC::Unavailable (14:all SubCons are in TransientFailure...)
    • Praefect cannot reach one or more of its child Gitaly nodes. Try running the Praefect connection checker to diagnose.

Praefect database experiencing high CPU load

Some common reasons for the Praefect database to experience elevated CPU usage include:

• Prometheus metrics scrapes running an expensive query. Set praefect['configuration'][:prometheus_exclude_database_from_default_metrics] = true in gitlab.rb.
• Read distribution caching is disabled, increasing the number of queries made to the database when user traffic is high. Ensure read distribution caching is enabled.

Determine primary Gitaly node

To determine the primary node of a repository, use the praefect metadata subcommand.

View repository metadata

Gitaly Cluster maintains a metadata database about the repositories stored on the cluster. Use the praefect metadata subcommand to inspect the metadata for troubleshooting.

You can retrieve a repository's metadata by its Praefect-assigned repository ID:

sudo -u git -- /opt/gitlab/embedded/bin/praefect -config /var/opt/gitlab/praefect/config.toml metadata -repository-id <repository-id>

When the physical path on the physical storage starts with @cluster, you can find the repository ID in the physical path.

You can also retrieve a repository's metadata by its virtual storage and relative path:

sudo -u git -- /opt/gitlab/embedded/bin/praefect -config /var/opt/gitlab/praefect/config.toml metadata -virtual-storage <virtual-storage> -relative-path <relative-path>

Examples

To retrieve the metadata for a repository with a Praefect-assigned repository ID of 1:

sudo -u git -- /opt/gitlab/embedded/bin/praefect -config /var/opt/gitlab/praefect/config.toml metadata -repository-id 1

To retrieve the metadata for a repository with virtual storage default and relative path @hashed/b1/7e/b17ef6d19c7a5b1ee83b907c595526dcb1eb06db8227d650d5dda0a9f4ce8cd9.git:

sudo -u git -- /opt/gitlab/embedded/bin/praefect -config /var/opt/gitlab/praefect/config.toml metadata -virtual-storage default -relative-path @hashed/b1/7e/b17ef6d19c7a5b1ee83b907c595526dcb1eb06db8227d650d5dda0a9f4ce8cd9.git

Either of these examples retrieve the following metadata for an example repository:

Repository ID: 54771
Virtual Storage: "default"
Relative Path: "@hashed/b1/7e/b17ef6d19c7a5b1ee83b907c595526dcb1eb06db8227d650d5dda0a9f4ce8cd9.git"
Replica Path: "@hashed/b1/7e/b17ef6d19c7a5b1ee83b907c595526dcb1eb06db8227d650d5dda0a9f4ce8cd9.git"
Primary: "gitaly-1"
Generation: 1
Replicas:
- Storage: "gitaly-1"
  Assigned: true
  Generation: 1, fully up to date
  Healthy: true
  Valid Primary: true
  Verified At: 2021-04-01 10:04:20 +0000 UTC
- Storage: "gitaly-2"
  Assigned: true
  Generation: 0, behind by 1 changes
  Healthy: true
  Valid Primary: false
  Verified At: unverified
- Storage: "gitaly-3"
  Assigned: true
  Generation: replica not yet created
  Healthy: false
  Valid Primary: false
  Verified At: unverified

Available metadata

The metadata retrieved by praefect metadata includes the fields in the following tables.

For each replica, the following metadata is available:

Command fails with 'repository not found'

If the supplied value for -virtual-storage is incorrect, the command returns the following error:

get metadata: rpc error: code = NotFound desc = repository not found

The documented examples specify -virtual-storage default. Check the Praefect server setting praefect['configuration'][:virtual_storage] in /etc/gitlab/gitlab.rb.

Check that repositories are in sync

Is some cases the Praefect database can get out of sync with the underlying Gitaly nodes. To check that a given repository is fully synced on all nodes, run the gitlab:praefect:replicas Rake task on your Rails node. This Rake task checksums the repository on all Gitaly nodes.

The Praefect dataloss command only checks the state of the repository in the Praefect database, and cannot be relied to detect sync problems in this scenario.

dataloss command shows @failed-geo-sync repositories as out of sync

@failed-geo-sync is a legacy path that was used on GitLab 16.1 and earlier by Geo when project synchronization failed and has been deprecated.

On GitLab 16.2 and later, you can safely delete this path. The @failed-geo-sync directories are located under the repository path on the Gitaly nodes.

Relation does not exist errors

By default Praefect database tables are created automatically by gitlab-ctl reconfigure task.

However, the Praefect database tables are not created on initial reconfigure and can throw errors that relations do not exist if either:

• The gitlab-ctl reconfigure command isn't executed.
• Errors occur during the execution.

For example:

• ERROR: relation "node_status" does not exist at character 13

• ERROR: relation "replication_queue_lock" does not exist at character 40

• This error:

  {"level":"error","msg":"Error updating node: pq: relation \"node_status\" does not exist","pid":210882,"praefectName":"gitlab1x4m:0.0.0.0:2305","time":"2021-04-01T19:26:19.473Z","virtual_storage":"praefect-cluster-1"}

To solve this, the database schema migration can be done using sql-migrate sub-command of the praefect command:

$ sudo -u git -- /opt/gitlab/embedded/bin/praefect -config /var/opt/gitlab/praefect/config.toml sql-migrate
praefect sql-migrate: OK (applied 21 migrations)

Requests fail with 'repository scoped: invalid Repository' errors

This indicates that the virtual storage name used in the Praefect configuration does not match the storage name used in gitaly['configuration'][:storage][<index>][:name] setting for GitLab.

Resolve this by matching the virtual storage names used in Praefect and GitLab configuration.

Gitaly Cluster performance issues on cloud platforms

Praefect does not require a lot of CPU or memory, and can run on small virtual machines. Cloud services may place other limits on the resources that small VMs can use, such as disk IO and network traffic.

Praefect nodes generate a lot of network traffic. The following symptoms can be observed if their network bandwidth has been throttled by the cloud service:

• Poor performance of Git operations.
• High network latency.
• High memory use by Praefect.

Possible solutions:

• Provision larger VMs to gain access to larger network traffic allowances.
• Use your cloud service's monitoring and logging to check that the Praefect nodes are not exhausting their traffic allowances.

gitlab-ctl reconfigure fails with a Praefect configuration error

If gitlab-ctl reconfigure fails, you might see this error:

STDOUT: praefect: configuration error: error reading config file: toml: cannot store TOML string into a Go int

This error occurs when praefect['database_port'] or praefect['database_direct_port'] are configured as a string instead of an integer.

Common replication errors

The following are some common replication errors with possible solutions.

Lock file exists

Lock files are used to prevent multiple updates to the same ref. Sometimes lock files become stale, and replication fails with the error error: cannot lock ref.

To clear stale *.lock files, you can trigger OptimizeRepositoryRequest on the Rails console:

p = Project.find <Project ID>
client = Gitlab::GitalyClient::RepositoryService.new(p.repository)
client.optimize_repository

If triggering OptimizeRepositoryRequest does not work, inspect the files manually to confirm the creation date and decide if the *.lock file can be manually removed. Any lock files created over 24 hours ago are safe to remove.

Git fsck errors

Gitaly repositories with invalid objects can lead to replication failures with errors in Gitaly logs such as:

• exit status 128, stderr: "fatal: git upload-pack: not our ref".
• "fatal: bad object 58....e0f... ssh://gitaly/internal.git did not send all necessary objects.

As long one of the Gitaly nodes still has a healthy copy of the repository, these issues can be fixed by:

1. Removing the repository from the Praefect database.
2. Using the Praefect track-repository subcommand to re-track it.

This will use the copy of the repository from the authoritative Gitaly node to overwrite the copies on all other Gitaly nodes. Be sure a recent backup of the repository has been made before running these commands.

1. Move the bad repository out of place:

   run `mv <REPOSITORY_PATH> <REPOSITORY_PATH>.backup`

   For example:

   mv /var/opt/gitlab/git-data/repositories/@cluster/repositories/de/74/2335 /var/opt/gitlab/git-data/repositories/@cluster/repositories/de/74/2335.backup

2. Run the Praefect commands to trigger replication:

   # Validate you have the correct repository.
   sudo -u git -- /opt/gitlab/embedded/bin/praefect -config /var/opt/gitlab/praefect/config.toml remove-repository -virtual-storage gitaly -relative-path '<relative_path>' -db-only
   
   # Run again with '--apply' flag to remove repository from the Praefect tracking database
   sudo -u git -- /opt/gitlab/embedded/bin/praefect -config /var/opt/gitlab/praefect/config.toml remove-repository -virtual-storage gitaly -relative-path '<relative_path>' -db-only --apply
   
   # Re-track the repository, overwriting the secondary nodes
   sudo -u git -- /opt/gitlab/embedded/bin/praefect -config /var/opt/gitlab/praefect/config.toml track-repository -virtual-storage gitaly -authoritative-storage '<healthy_gitaly>' -relative-path '<relative_path>' -replica-path '<replica_path>'-replicate-immediately

Replication fails silently

If the Praefect dataloss shows repositories partially unavailable, and accept-dataloss command fails to synchronize the repository with no error present on the logs, this could be due to a mismatch in Praefect database in the repository_id field of the storage_repositories table. To check for a mismatch:

1. Connect to the Praefect database.

2. Run the following query:

   select * from storage_repositories where relative_path = '<relative-path>';

   Replace <relative-path> with the repository path beginning with @hashed.

Alternate directory does not exists

GitLab uses the Git alternates mechanism for deduplication. alternates is a text file that points to the objects directory on a @pool repository to fetch objects. If this file points to an invalid path, replication can fail with one of the following the errors:

• "error":"no alternates directory exists", "warning","msg":"alternates file does not point to valid git repository"
• "error":"unexpected alternates content:
• remote: error: unable to normalize alternate object path

To investigate the cause of this error:

1. Check if the project is part of a pool by using the Rails console:

   project = Project.find_by_id(<project id>)
   project.pool_repository

2. Check if the pool repository path exists on disk and that it matches the alternates file content.

3. Check if the path in the alternates file is reachable from the objects directory in the project.

After performing these checks, reach out to GitLab Support with the information collected.