• About the Gitaly token
• Run Gitaly on its own server
  • Network architecture
  • Install Gitaly
  • Configure Gitaly servers
    • Configure authentication
    • Configure Gitaly server
  • Configure Gitaly clients
    • Mixed configuration
  • GitLab requires a default repository storage
  • Disable Gitaly where not required (optional)
• Control groups
  • Configure repository cgroups (new method)
  • Configure repository cgroups (legacy method)
  • Configuring oversubscription
• Background repository optimization
• Rotate Gitaly authentication token
  • Verify authentication monitoring
  • Enable “auth transitioning” mode
  • Update Gitaly authentication token
  • Ensure there are no authentication failures
  • Disable “auth transitioning” mode
  • Verify authentication is enforced
• Pack-objects cache
  • Configure the cache
    • enabled defaults to false
    • Cache storage directory dir
    • Cache eviction window max_age
    • Minimum key occurrences min_occurrences
  • Observe the cache
• cat-file cache
• Configure commit signing for GitLab UI commits
• Generate configuration using an external command
• Configure server-side backups
  • Configure Azure Blob storage
  • Configure Google Cloud storage
  • Configure S3 storage
    • Configure S3-compatible servers

Configure Gitaly

Configure Gitaly in one of two ways:

The following configuration options are also available:

• Enabling TLS support.
• Limiting RPC concurrency.
• Limiting pack-objects concurrency.

About the Gitaly token

The token referred to throughout the Gitaly documentation is just an arbitrary password selected by the administrator. It is unrelated to tokens created for the GitLab API or other similar web API tokens.

Run Gitaly on its own server

By default, Gitaly is run on the same server as Gitaly clients and is configured as above. Single-server installations are best served by this default configuration used by:

• Linux package installations.
• Self-compiled installations.

However, Gitaly can be deployed to its own server, which can benefit GitLab installations that span multiple machines.

The process for setting up Gitaly on its own server is:

1. Install Gitaly.
2. Configure authentication.
3. Configure Gitaly servers.
4. Configure Gitaly clients.
5. Disable Gitaly where not required (optional).

Network architecture

The following list depicts the network architecture of Gitaly:

• GitLab Rails shards repositories into repository storages.
• /config/gitlab.yml contains a map from storage names to (Gitaly address, Gitaly token) pairs.
• The storage name -> (Gitaly address, Gitaly token) map in /config/gitlab.yml is the single source of truth for the Gitaly network topology.
• A (Gitaly address, Gitaly token) corresponds to a Gitaly server.
• A Gitaly server hosts one or more storages.
• A Gitaly client can use one or more Gitaly servers.
• Gitaly addresses must be specified in such a way that they resolve correctly for all Gitaly clients.
• Gitaly clients are:
  • Puma.
  • Sidekiq.
  • GitLab Workhorse.
  • GitLab Shell.
  • Elasticsearch indexer.
  • Gitaly itself.
• A Gitaly server must be able to make RPC calls to itself by using its own (Gitaly address, Gitaly token) pair as specified in /config/gitlab.yml.
• Authentication is done through a static token which is shared among the Gitaly and GitLab Rails nodes.

The following digraph illustrates communication between Gitaly servers and GitLab Rails showing the default ports for HTTP and HTTPs communication.

In the following sections, we describe how to configure two Gitaly servers with secret token abc123secret:

• gitaly1.internal.
• gitaly2.internal.

We assume your GitLab installation has three repository storages:

• default.
• storage1.
• storage2.

You can use as few as one server with one repository storage if desired.

Install Gitaly

Install Gitaly on each Gitaly server using either:

• A Linux package installation. Download and install the Linux package you want but do not provide the EXTERNAL_URL= value.
• A self-compiled installation. Follow the steps at Install Gitaly.

Configure Gitaly servers

To configure Gitaly servers, you must:

• Configure authentication.
• Configure storage paths.
• Enable the network listener.

The git user must be able to read, write, and set permissions on the configured storage path.

To avoid downtime while rotating the Gitaly token, you can temporarily disable authentication using the gitaly['auth_transitioning'] setting. For more information, see the documentation on enabling “auth transitioning mode”.

Configure authentication

Gitaly and GitLab use two shared secrets for authentication:

• Gitaly token: used to authenticate gRPC requests to Gitaly.
• GitLab Shell token: used for authentication callbacks from GitLab Shell to the GitLab internal API.

Configure Gitaly server

Configure Gitaly server.

Configure Gitaly clients

As the final step, you must update Gitaly clients to switch from using local Gitaly service to use the Gitaly servers you just configured.

This can be risky because anything that prevents your Gitaly clients from reaching the Gitaly servers causes all Gitaly requests to fail. For example, any sort of network, firewall, or name resolution problems.

Gitaly makes the following assumptions:

• Your gitaly1.internal Gitaly server can be reached at gitaly1.internal:8075 from your Gitaly clients, and that Gitaly server can read, write, and set permissions on /var/opt/gitlab/git-data and /mnt/gitlab/git-data.
• Your gitaly2.internal Gitaly server can be reached at gitaly2.internal:8075 from your Gitaly clients, and that Gitaly server can read, write, and set permissions on /srv/gitlab/git-data.
• Your gitaly1.internal and gitaly2.internal Gitaly servers can reach each other.

You can’t define Gitaly servers with some as a local Gitaly server (without gitaly_address) and some as remote server (with gitaly_address) unless you use mixed configuration.

Configure Gitaly clients in one of two ways. These instructions are for unencrypted connections but you can also enable TLS support:

When you tail the Gitaly logs on your Gitaly server, you should see requests coming in. One sure way to trigger a Gitaly request is to clone a repository from GitLab over HTTP or HTTPS.

Mixed configuration

GitLab can reside on the same server as one of many Gitaly servers, but doesn’t support configuration that mixes local and remote configuration. The following setup is incorrect, because:

• All addresses must be reachable from the other Gitaly servers.
• storage1 is assigned a Unix socket for gitaly_address which is invalid for some of the Gitaly servers.

git_data_dirs({
  'default' => { 'gitaly_address' => 'tcp://gitaly1.internal:8075' },
  'storage1' => { 'path' => '/mnt/gitlab/git-data' },
  'storage2' => { 'gitaly_address' => 'tcp://gitaly2.internal:8075' },
})

To combine local and remote Gitaly servers, use an external address for the local Gitaly server. For example:

git_data_dirs({
  'default' => { 'gitaly_address' => 'tcp://gitaly1.internal:8075' },
  # Address of the GitLab server that also has Gitaly running on it
  'storage1' => { 'gitaly_address' => 'tcp://gitlab.internal:8075' },
  'storage2' => { 'gitaly_address' => 'tcp://gitaly2.internal:8075' },
})

gitaly['configuration'] = {
  # ...
  #
  # Make Gitaly accept connections on all network interfaces
  listen_addr: '0.0.0.0:8075',
  # Or for TLS
  tls_listen_addr: '0.0.0.0:9999',
  tls: {
    certificate_path:  '/etc/gitlab/ssl/cert.pem',
    key_path: '/etc/gitlab/ssl/key.pem',
  },
  storage: [
    {
      name: 'storage1',
      path: '/mnt/gitlab/git-data/repositories',
    },
  ],
}

path can be included only for storage shards on the local Gitaly server. If it’s excluded, default Git storage directory is used for that storage shard.

GitLab requires a default repository storage

When adding Gitaly servers to an environment, you might want to replace the original default Gitaly service. However, you can’t reconfigure the GitLab application servers to remove the default entry from git_data_dirs because GitLab requires a git_data_dirs entry called default. Read more about this limitation.

To work around the limitation:

1. Define an additional storage location on the new Gitaly service and configure the additional storage to be default.
2. In the Admin area, set default to a weight of zero to prevent repositories being stored there.

Disable Gitaly where not required (optional)

If you run Gitaly as a remote service, consider disabling the local Gitaly service that runs on your GitLab server by default, and run it only where required.

Disabling Gitaly on the GitLab instance makes sense only when you run GitLab in a custom cluster configuration, where Gitaly runs on a separate machine from the GitLab instance. Disabling Gitaly on all machines in the cluster is not a valid configuration (some machines much act as Gitaly servers).

Disable Gitaly on a GitLab server in one of two ways:

Control groups

When enabling cgroups for memory, you should ensure that no swap is configured on the Gitaly nodes as processes may switch to using that instead of being terminated. This situation could lead to notably compromised performance.

You can use control groups (cgroups) in Linux to impose limits on how much memory and CPU can be consumed by Gitaly processes. See the cgroups Linux man page for more information. cgroups can help protect the system against unexpected resource exhaustion because of over consumption of memory and CPU.

Some Git operations can consume notable resources up to the point of exhaustion in situations such as:

• Unexpectedly high traffic.
• Operations running against large repositories that don’t follow best practices.

As a hard protection, it’s possible to use cgroups that configure the kernel to terminate these operations before they hog up all system resources and cause instability.

Gitaly has built-in cgroups control. When configured, Gitaly assigns Git processes to a cgroup based on the repository the Git command is operating in. These cgroups are called repository cgroups. Each repository cgroup:

• Has a memory and CPU limit.
• Contains the Git processes for a single repository.
• Uses a consistent hash to ensure a Git process for a given repository always ends up in the same cgroup.

When a repository cgroup reaches its:

• Memory limit, the kernel looks through the processes for a candidate to kill.
• CPU limit, processes are not killed, but the processes are prevented from consuming more CPU than allowed.

Configure repository cgroups (new method)

To configure repository cgroups in Gitaly using the new method, use the following settings for the new configuration method to gitaly['configuration'][:cgroups] in /etc/gitlab/gitlab.rb:

• mountpoint is where the parent cgroup directory is mounted. Defaults to /sys/fs/cgroup.
• hierarchy_root is the parent cgroup under which Gitaly creates groups, and is expected to be owned by the user and group Gitaly runs as. A Linux package installation creates the set of directories mountpoint/<cpu|memory>/hierarchy_root when Gitaly starts.
• memory_bytes is the total memory limit that is imposed collectively on all Git processes that Gitaly spawns. 0 implies no limit.
• cpu_shares is the CPU limit that is imposed collectively on all Git processes that Gitaly spawns. 0 implies no limit. The maximum is 1024 shares, which represents 100% of CPU.
• cpu_quota_us is the cfs_quota_us to throttle the cgroups’ processes if they exceed this quota value. We set cfs_period_us to 100ms so 1 core is 100000. 0 implies no limit.
• repositories.count is the number of cgroups in the cgroups pool. Each time a new Git command is spawned, Gitaly assigns it to one of these cgroups based on the repository the command is for. A circular hashing algorithm assigns Git commands to these cgroups, so a Git command for a repository is always assigned to the same cgroup.
• repositories.memory_bytes is the total memory limit imposed on all Git processes contained in a repository cgroup. 0 implies no limit. This value cannot exceed that of the top level memory_bytes.
• repositories.cpu_shares is the CPU limit that is imposed on all Git processes contained in a repository cgroup. 0 implies no limit. The maximum is 1024 shares, which represents 100% of CPU. This value cannot exceed that of the top levelcpu_shares.
• repositories.cpu_quota_us is the cfs_quota_us that is imposed on all Git processes contained in a repository cgroup. A Git process can’t use more then the given quota. We set cfs_period_us to 100ms so 1 core is 100000. 0 implies no limit.

For example:

# in /etc/gitlab/gitlab.rb
gitaly['configuration'] = {
  # ...
  cgroups: {
    mountpoint: '/sys/fs/cgroup',
    hierarchy_root: 'gitaly',
    memory_bytes: 64424509440, # 60gb
    cpu_shares: 1024,
    cpu_quota_us: 400000 # 4 cores
    repositories: {
      count: 1000,
      memory_bytes: 32212254720, # 20gb
      cpu_shares: 512,
      cpu_quota_us: 200000 # 2 cores
    },
  },
}

Configure repository cgroups (legacy method)

To configure repository cgroups in Gitaly using the legacy method, use the following settings in /etc/gitlab/gitlab.rb:

• cgroups_count is the number of cgroups created. Each time a new command is spawned, Gitaly assigns it to one of these cgroups based on the command line arguments of the command. A circular hashing algorithm assigns commands to these cgroups.
• cgroups_mountpoint is where the parent cgroup directory is mounted. Defaults to /sys/fs/cgroup.
• cgroups_hierarchy_root is the parent cgroup under which Gitaly creates groups, and is expected to be owned by the user and group Gitaly runs as. A Linux package installation creates the set of directories mountpoint/<cpu|memory>/hierarchy_root when Gitaly starts.
• cgroups_memory_enabled enables or disables the memory limit on cgroups.
• cgroups_memory_bytes is the total memory limit each cgroup imposes on the processes added to it.
• cgroups_cpu_enabled enables or disables the CPU limit on cgroups.
• cgroups_cpu_shares is the CPU limit each cgroup imposes on the processes added to it. The maximum is 1024 shares, which represents 100% of CPU.

For example:

# in /etc/gitlab/gitlab.rb
gitaly['cgroups_count'] = 1000
gitaly['cgroups_mountpoint'] = "/sys/fs/cgroup"
gitaly['cgroups_hierarchy_root'] = "gitaly"
gitaly['cgroups_memory_limit'] = 32212254720
gitaly['cgroups_memory_enabled'] = true
gitaly['cgroups_cpu_shares'] = 1024
gitaly['cgroups_cpu_enabled'] = true

Configuring oversubscription

In the previous example using the new configuration method:

• The top level memory limit is capped at 60 GB.
• Each of the 1000 cgroups in the repositories pool is capped at 20 GB.

This configuration leads to “oversubscription”. Each cgroup in the pool has a much larger capacity than 1/1000th of the top-level memory limit.

This strategy has two main benefits:

• It gives the host protection from overall memory starvation (OOM), because the memory limit of the top-level cgroup can be set to a threshold smaller than the host’s capacity. Processes outside of that cgroup are not at risk of OOM.
• It allows each individual cgroup in the pool to burst up to a generous upper bound (in this example 20 GB) that is smaller than the limit of the parent cgroup, but substantially larger than 1/N of the parent’s limit. In this example, up to 3 child cgroups can concurrently burst up to their max. In general, all 1000 cgroups would use much less than the 20 GB.

Background repository optimization

Empty directories and unneeded configuration settings may accumulate in a repository and slow down Git operations. Gitaly can schedule a daily background task with a maximum duration to clean up these items and improve performance.

Configure background repository optimization in one of two ways:

gitaly['configuration'] = {
  # ...
  daily_maintenance: {
    # ...
    start_hour: 4,
    start_minute: 30,
    duration: '30m',
    storages: ['default'],
  },
}

[daily_maintenance]
start_hour = 4
start_minute = 30
duration = '30m'
storages = ["default"]

Rotate Gitaly authentication token

Rotating credentials in a production environment often requires downtime, causes outages, or both.

However, you can rotate Gitaly credentials without a service interruption. Rotating a Gitaly authentication token involves:

• Verifying authentication monitoring.
• Enabling “auth transitioning” mode.
• Updating Gitaly authentication tokens.
• Ensuring there are no authentication failures.
• Disabling “auth transitioning” mode.
• Verifying authentication is enforced.

This procedure also works if you are running GitLab on a single server. In that case, “Gitaly server” and “Gitaly client” refers to the same machine.

Verify authentication monitoring

Before rotating a Gitaly authentication token, verify that you can monitor the authentication behavior of your GitLab installation using Prometheus.

You can then continue the rest of the procedure.

Enable “auth transitioning” mode

Temporarily disable Gitaly authentication on the Gitaly servers by putting them into “auth transitioning” mode as follows:

# in /etc/gitlab/gitlab.rb
gitaly['configuration'] = {
  # ...
  auth: {
    # ...
    transitioning: true,
  },
}

After you have made this change, your Prometheus query should return something like:

{enforced="false",status="would be ok"}  4424.985419441742

Because enforced="false", it is safe to start rolling out the new token.

Update Gitaly authentication token

To update to a new Gitaly authentication token, on each Gitaly client and Gitaly server:

1. Update the configuration:

   # in /etc/gitlab/gitlab.rb
   gitaly['configuration'] = {
      # ...
      auth: {
         # ...
         token: '<new secret token>',
      },
   }
   

2. Restart Gitaly:

   gitlab-ctl restart gitaly
   

If you run your Prometheus query while this change is being rolled out, you see non-zero values for the enforced="false",status="denied" counter.

Ensure there are no authentication failures

After the new token is set, and all services involved have been restarted, you will temporarily see a mix of:

• status="would be ok".
• status="denied".

After the new token is picked up by all Gitaly clients and Gitaly servers, the only non-zero rate should be enforced="false",status="would be ok".

Disable “auth transitioning” mode

To re-enable Gitaly authentication, disable “auth transitioning” mode. Update the configuration on your Gitaly servers as follows:

# in /etc/gitlab/gitlab.rb
gitaly['configuration'] = {
  # ...
  auth: {
    # ...
    transitioning: false,
  },
}

Verify authentication is enforced

Refresh your Prometheus query. You should now see a similar result as you did at the start. For example:

{enforced="true",status="ok"}  4424.985419441742

enforced="true" means that authentication is being enforced.

Pack-objects cache

Gitaly, the service that provides storage for Git repositories, can be configured to cache a short rolling window of Git fetch responses. This can reduce server load when your server receives lots of CI fetch traffic.

The pack-objects cache wraps git pack-objects, an internal part of Git that gets invoked indirectly by using the PostUploadPack and SSHUploadPack Gitaly RPCs. Gitaly runs PostUploadPack when a user does a Git fetch by using HTTP, or SSHUploadPack when a user does a Git fetch by using SSH. When the cache is enabled, anything that uses PostUploadPack or SSHUploadPack can benefit from it. It is orthogonal to:

• The transport (HTTP or SSH).
• Git protocol version (v0 or v2).
• The type of fetch, such as full clones, incremental fetches, shallow clones, or partial clones.

The strength of this cache is its ability to deduplicate concurrent identical fetches. It:

• Can benefit GitLab instances where your users run CI/CD pipelines with many concurrent jobs. There should be a noticeable reduction in server CPU utilization.
• Does not benefit unique fetches at all. For example, if you run a spot check by cloning a repository to your local computer, you are unlikely to see a benefit from this cache because your fetch is probably unique.

The pack-objects cache is a local cache. It:

• Stores its metadata in the memory of the Gitaly process it is enabled in.
• Stores the actual Git data it is caching in files on local storage.

Using local files has the benefit that the operating system may automatically keep parts of the pack-objects cache files in RAM, making it faster.

Because the pack-objects cache can lead to a significant increase in disk write IO, it is off by default. In GitLab 15.11 and later, the write workload is approximately 50% lower, but the cache is still disabled by default.

Configure the cache

These configuration settings are available for the pack-objects cache. Each setting is discussed in greater detail below.

In /etc/gitlab/gitlab.rb, set:

gitaly['configuration'] = {
  # ...
  pack_objects_cache: {
    # ...
    enabled: true,
    # dir: '/var/opt/gitlab/git-data/repositories/+gitaly/PackObjectsCache',
    # max_age: '5m',
    # min_occurrences: 1,
  },
}

enabled defaults to false

The cache is disabled by default because in some cases, it can create an extreme increase in the number of bytes written to disk. On GitLab.com, we have verified that our repository storage disks can handle this extra workload, but we felt we cannot assume this is true everywhere.

Cache storage directory dir

The cache needs a directory to store its files in. This directory should be:

• In a file system with enough space. If the cache file system runs out of space, all fetches start failing.
• On a disk with enough IO bandwidth. If the cache disk runs out of IO bandwidth, all fetches, and probably the entire server, slows down.

By default, the cache storage directory is set to a subdirectory of the first Gitaly storage defined in the configuration file.

Multiple Gitaly processes can use the same directory for cache storage. Each Gitaly process uses a unique random string as part of the cache filenames it creates. This means:

• They do not collide.
• They do not reuse another process’s files.

While the default directory puts the cache files in the same file system as your repository data, this is not requirement. You can put the cache files on a different file system if that works better for your infrastructure.

The amount of IO bandwidth required from the disk depends on:

• The size and shape of the repositories on your Gitaly server.
• The kind of traffic your users generate.

You can use the gitaly_pack_objects_generated_bytes_total metric as a pessimistic estimate, pretending your cache hit ratio is 0%.

The amount of space required depends on:

• The bytes per second that your users pull from the cache.
• The size of the max_age cache eviction window.

If your users pull 100 MB/s and you use a 5 minute window, then on average you have 5*60*100MB = 30GB of data in your cache directory. This average is an expected average, not a guarantee. Peak size may exceed this average.

Cache eviction window max_age

The max_age configuration setting lets you control the chance of a cache hit and the average amount of storage used by cache files. Entries older than max_age get deleted from the disk.

Eviction does not interfere with ongoing requests. It is OK for max_age to be less than the time it takes to do a fetch over a slow connection because Unix filesystems do not truly delete a file until all processes that are reading the deleted file have closed it.

Minimum key occurrences min_occurrences

The min_occurrences setting controls how often an identical request must occur before we create a new cache entry. The default value is 1, meaning that unique requests do not get written into the cache.

If you:

• Increase this number, your cache hit rate goes down and the cache uses less disk space.
• Decrease this number, your cache hit rate goes up and the cache uses more disk space.

You should set min_occurrences to 1. On GitLab.com, going from 0 to 1 saved us 50% cache disk space while barely affecting the cache hit rate.

Observe the cache

You can observe the cache using metrics and in the following logged information. These logs are part of the gRPC logs and can be discovered when a call is executed.

In the case of a:

• Cache miss, Gitaly logs both a pack_objects_cache.generated_bytes and pack_objects_cache.served_bytes message. Gitaly also logs some more detailed statistics of pack-object generation.
• Cache hit, Gitaly logs only a pack_objects_cache.served_bytes message.

Example:

{
  "bytes":26186490,
  "correlation_id":"01F1MY8JXC3FZN14JBG1H42G9F",
  "grpc.meta.deadline_type":"none",
  "grpc.method":"PackObjectsHook",
  "grpc.request.fullMethod":"/gitaly.HookService/PackObjectsHook",
  "grpc.request.glProjectPath":"root/gitlab-workhorse",
  "grpc.request.glRepository":"project-2",
  "grpc.request.repoPath":"@hashed/d4/73/d4735e3a265e16eee03f59718b9b5d03019c07d8b6c51f90da3a666eec13ab35.git",
  "grpc.request.repoStorage":"default",
  "grpc.request.topLevelGroup":"@hashed",
  "grpc.service":"gitaly.HookService",
  "grpc.start_time":"2021-03-25T14:57:52.747Z",
  "level":"info",
  "msg":"finished unary call with code OK",
  "peer.address":"@",
  "pid":20961,
  "span.kind":"server",
  "system":"grpc",
  "time":"2021-03-25T14:57:53.543Z",
  "pack_objects.compression_statistics": "Total 145991 (delta 68), reused 6 (delta 2), pack-reused 145911",
  "pack_objects.enumerate_objects_ms": 170,
  "pack_objects.prepare_pack_ms": 7,
  "pack_objects.write_pack_file_ms": 786,
  "pack_objects.written_object_count": 145991,
  "pack_objects_cache.generated_bytes": 49533030,
  "pack_objects_cache.hit": "false",
  "pack_objects_cache.key": "123456789",
  "pack_objects_cache.served_bytes": 49533030,
  "peer.address": "127.0.0.1",
  "pid": 8813,
}

cat-file cache

A lot of Gitaly RPCs need to look up Git objects from repositories. Most of the time we use git cat-file --batch processes for that. For better performance, Gitaly can re-use these git cat-file processes across RPC calls. Previously used processes are kept around in a “Git cat-file cache”. To control how much system resources this uses, we have a maximum number of cat-file processes that can go into the cache.

The default limit is 100 cat-files, which constitute a pair of git cat-file --batch and git cat-file --batch-check processes. If you see errors about “too many open files”, or an inability to create new processes, you may want to lower this limit.

Ideally, the number should be large enough to handle standard traffic. If you raise the limit, you should measure the cache hit ratio before and after. If the hit ratio does not improve, the higher limit is probably not making a meaningful difference. Here is an example Prometheus query to see the hit rate:

sum(rate(gitaly_catfile_cache_total{type="hit"}[5m])) / sum(rate(gitaly_catfile_cache_total{type=~"(hit)|(miss)"}[5m]))

Configure the cat-file cache in the Gitaly configuration file.

Configure commit signing for GitLab UI commits

By default, Gitaly doesn’t sign commits made using GitLab UI. For example, commits made using:

• Web editor.
• Web IDE.
• Merge requests.

You can configure Gitaly to sign commits made with the GitLab UI.

By default, Gitaly sets the author of a commit as the committer. In this case, it is harder to Verify commits locally because the signature belongs to neither the author nor the committer of the commit.

You can configure Gitaly to reflect that a commit has been committed by your instance by setting committer_email and committer_name. For example, on GitLab.com these configuration options are set to noreply@gitlab.com and GitLab.

rotated_signing_keys is a list of keys to use for verification only. Gitaly tries to verify a web commit using the configured signing_key, and then uses the rotated keys one by one until it succeeds. Set the rotated_signing_keys option when either:

• The signing key is rotated.
• You want to specify multiple keys to migrate projects from other instances and want to display their web commits as Verified.

Configure Gitaly to sign commits made with the GitLab UI in one of two ways:

Generate configuration using an external command

You can generate parts of the Gitaly configuration using an external command. You might do this:

• To configure nodes without having to distribute the full configuration to each of them.
• To configure using auto-discovery of the node’s settings. For example, using DNS entries.
• To configure secrets at startup of the node, so that don’t need to be visible in plain text.

To generate configuration using an external command, you must provide a script that dumps the desired configuration of the Gitaly node in JSON format to its standard output.

For example, the following command configures the HTTP password used to connect to the GitLab internal API using an AWS secret:

#!/usr/bin/env ruby
require 'json'
JSON.generate({"gitlab": {"http_settings": {"password": `aws get-secret-value --secret-id ...`}}})

You must then make the script path known to Gitaly in one of two ways:

gitaly['configuration'] = {
    config_command: '/path/to/config_command',
}

config_command = "/path/to/config_command"

After configuration, Gitaly executes the command on startup and parses its standard output as JSON. The resulting configuration is then merged back into the other Gitaly configuration.

Gitaly fails to start up if either:

• The configuration command fails.
• The output produced by the command cannot be parsed as valid JSON.

Configure server-side backups

Repository backups can be configured so that the Gitaly node that hosts each repository is responsible for creating the backup and streaming it to object storage. This helps reduce the network resources required to create and restore a backup.

Each Gitaly node must be configured to connect to object storage for backups.

After configuring server-side backups, you can create a server-side repository backup.

Configure Azure Blob storage

How you configure Azure Blob storage for backups depends on the type of installation you have. For self-compiled installations, you must set the AZURE_STORAGE_ACCOUNT and AZURE_STORAGE_KEY environment variables outside of GitLab.

gitaly['env'] = {
    'AZURE_STORAGE_ACCOUNT' => 'azure_storage_account',
    'AZURE_STORAGE_KEY' => 'azure_storage_key' # or 'AZURE_STORAGE_SAS_TOKEN'
}
gitaly['configuration'] = {
    backup: {
        go_cloud_url: 'azblob://<bucket>'
    }
}

[backup]
go_cloud_url = "azblob://<bucket>"

Configure Google Cloud storage

Google Cloud storage (GCP) authenticates using Application Default Credentials. Set up Application Default Credentials on each Gitaly server using either:

• The gcloud auth application-default login command.
• The GOOGLE_APPLICATION_CREDENTIALS environment variable. For self-compiled installations, set the environment variable outside of GitLab.

For more information, see Application Default Credentials.

The destination bucket is configured using the go_cloud_url option.

gitaly['env'] = {
    'GOOGLE_APPLICATION_CREDENTIALS' => '/path/to/service.json'
}
gitaly['configuration'] = {
    backup: {
        go_cloud_url: 'gs://<bucket>'
    }
}

[backup]
go_cloud_url = "gs://<bucket>"

Configure S3 storage

To configure S3 storage authentication:

• If you authenticate with the AWS CLI, you can use the default AWS session.
• Otherwise, you can use the AWS_ACCESS_KEY_ID and AWS_SECRET_ACCESS_KEY environment variables. For self-compiled installations, set the environment variables outside of GitLab.

For more information, see AWS Session documentation.

The destination bucket and region are configured using the go_cloud_url option.

gitaly['env'] = {
    'AWS_ACCESS_KEY_ID' => 'aws_access_key_id',
    'AWS_SECRET_ACCESS_KEY' => 'aws_secret_access_key'
}
gitaly['configuration'] = {
    backup: {
        go_cloud_url: 's3://<bucket>?region=us-west-1'
    }
}

[backup]
go_cloud_url = "s3://<bucket>?region=us-west-1"

Configure S3-compatible servers

S3-compatible servers such as MinIO are configured similarly to S3 with the addition of the endpoint parameter.

The following parameters are supported:

• region: The AWS region.
• endpoint: The endpoint URL.
• disabledSSL: A value of true disables SSL.
• s3ForcePathStyle: A value of true forces path-style addressing.

gitaly['env'] = {
    'AWS_ACCESS_KEY_ID' => 'minio_access_key_id',
    'AWS_SECRET_ACCESS_KEY' => 'minio_secret_access_key'
}
gitaly['configuration'] = {
    backup: {
        go_cloud_url: 's3://<bucket>?region=minio&endpoint=my.minio.local:8080&disableSSL=true&s3ForcePathStyle=true'
    }
}

[backup]
go_cloud_url = "s3://<bucket>?region=minio&endpoint=my.minio.local:8080&disableSSL=true&s3ForcePathStyle=true"