Gitaly reference

Gitaly is configured via a TOML configuration file. Unlike installations from source, in Omnibus GitLab, you would not edit this file directly.

The configuration file is passed as an argument to the gitaly executable. This is usually done by either Omnibus GitLab or your init script.

An example config file can be found in the Gitaly project.


At the top level, config.toml defines the items described on the table below.

socket_pathstringyes (if listen_addr is not set)A path which Gitaly should open a Unix socket.
listen_addrstringyes (if socket_path is not set)TCP address for Gitaly to listen on.
tls_listen_addrstringnoTCP over TLS address for Gitaly to listen on.
bin_dirstringyesDirectory containing Gitaly’s executables.
prometheus_listen_addrstringnoTCP listen address for Prometheus metrics. If not set, no Prometheus listener is started.

For example:

socket_path = "/home/git/gitlab/tmp/sockets/private/gitaly.socket"
listen_addr = "localhost:9999"
tls_listen_addr = "localhost:8888"
bin_dir = "/home/git/gitaly"
prometheus_listen_addr = "localhost:9236"


Gitaly can be configured to reject requests that do not contain a specific bearer token in their headers. This is a security measure to be used when serving requests over TCP:

# A non-empty token enables authentication.
token = "the secret token"

Authentication is disabled when the token setting in config.toml is absent or an empty string.

It is possible to temporarily disable authentication with the transitioning setting. This allows you to monitor if all clients are authenticating correctly without causing a service outage for clients that are not configured correctly yet:

token = "the secret token"
transitioning = true
Warning: Remember to disable transitioning when you are done changing your token settings.

All authentication attempts are counted in Prometheus under the gitaly_authentications_total metric.


Gitaly supports TLS encryption. You will need to bring your own certificates as this isn’t provided automatically.

certificate_pathstringnoPath to the certificate.
key_pathstringnoPath to the key.
tls_listen_addr = "localhost:8888"

certificate_path = '/home/git/cert.cert'
key_path = '/home/git/key.pem'

Read more about TLS in Gitaly.


GitLab repositories are grouped into directories known as “storages” (e.g., /home/git/repositories) containing bare repositories managed by GitLab with names (e.g., default).

These names and paths are also defined in the gitlab.yml configuration file of GitLab. When you run Gitaly on the same machine as GitLab, which is the default and recommended configuration, storage paths defined in Gitaly’s config.toml must match those in gitlab.yml.

storagearrayyesAn array of storage shards.
pathstringyesThe path to the storage shard.
namestringyesThe name of the storage shard.

For example:

path = "/path/to/storage/repositories"
name = "my_shard"

path = "/path/to/other/repositories"
name = "other_storage"


The following values can be set in the [git] section of the configuration file.

bin_pathstringnoPath to Git binary. If not set, will be resolved using PATH.
catfile_cache_sizeintegernoMaximum number of cached cat-file processes. Default is 100.

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”. In order 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 are seeing errors complaining 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 normal 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]))


A Gitaly process uses one or more gitaly-ruby helper processes to execute RPC’s implemented in Ruby instead of Go. The [gitaly-ruby] section of the config file contains settings for these helper processes.

These processes are known to occasionally suffer from memory leaks. Gitaly restarts its gitaly-ruby helpers when their memory exceeds the max_rss limit.

dirstringyesPath to where gitaly-ruby is installed (needed to boot the process).
max_rssintegernoResident set size limit that triggers a gitaly-ruby restart, in bytes. Default is 200000000 (200MB).
graceful_restart_timeoutstringnoGrace period before a gitaly-ruby process is forcibly terminated after exceeding max_rss. Default is 10m (10 minutes).
restart_delaystringnoTime that gitaly-ruby memory must remain high before a restart. Default is 5m (5 minutes).
num_workersintegernoNumber of gitaly-ruby worker processes. Try increasing this number in case of ResourceExhausted errors. Default is 2, minimum is 2.
linguist_languages_pathstringnoOverride for dynamic languages.json discovery. Defaults to an empty string (use of dynamic discovery).


dir = "/home/git/gitaly/ruby"
max_rss = 200000000
graceful_restart_timeout = "10m"
restart_delay = "5m"
num_workers = 2

GitLab Shell

For historical reasons GitLab Shell contains the Git hooks that allow GitLab to validate and react to Git pushes. Because Gitaly “owns” Git pushes, GitLab Shell must therefore be installed alongside Gitaly. This will be simplified in the future.

dirstringyesThe directory where GitLab Shell is installed.


dir = "/home/git/gitlab-shell"


You can optionally configure Gitaly to record histogram latencies on GRPC method calls in Prometheus.

grpc_latency_bucketsarraynoPrometheus stores each observation in a bucket, which means you’d get an approximation of latency. Optimizing the buckets gives more control over the accuracy of the approximation.


prometheus_listen_addr = "localhost:9236"

grpc_latency_buckets = [0.001, 0.005, 0.025, 0.1, 0.5, 1.0, 10.0, 30.0, 60.0, 300.0, 1500.0]


The following values configure logging in Gitaly under the [logging] section.

formatstringnoLog format: text or json. Default: text.
levelstringnoLog level: debug, info, warn, error, fatal, or panic. Default: info.
sentry_dsnstringnoSentry DSN for exception monitoring.
sentry_environmentstringnoSentry Environment for exception monitoring.
ruby_sentry_dsnstringnoSentry DSN for gitaly-ruby exception monitoring.

While the main Gitaly application logs go to stdout, there are some extra log files that go to a configured directory, like the GitLab Shell logs. GitLab Shell does not support panic or trace level logs. panic will fall back to error, while trace will fall back to debug. Any other invalid log levels will default to info.


level = "warn"
dir = "/home/gitaly/logs"
format = "json"
sentry_dsn = "https://<key>:<secret><project>"
ruby_sentry_dsn = "https://<key>:<secret><project>"


You can adjust the concurrency of each RPC endpoint.

concurrencyarrayyesAn array of RPC endpoints.
rpcstringnoThe name of the RPC endpoint (/gitaly.RepositoryService/GarbageCollect).
max_per_repointegernoConcurrency per RPC per repo.


rpc = "/gitaly.RepositoryService/GarbageCollect"
max_per_repo = 1