Repository storage types

GitLab can be configured to use one or multiple repository storages. These storages can be:

In GitLab:

  • Repository storages are configured in:
    • /etc/gitlab/gitlab.rb by the git_data_dirs({}) configuration hash for Linux package installations.
    • gitlab.yml by the repositories.storages key for self-compiled installations.
  • The default repository storage is available in any installations that haven’t customized it. By default, it points to a Gitaly node.

The repository storage types documented here apply to any repository storage defined in git_data_dirs({}) or repositories.storages.

Hashed storage

Version history
  • Introduced in GitLab 10.0.
  • Made the default for new installations in GitLab 12.0.
  • Enabled by default for new and renamed projects in GitLab 13.0.

Hashed storage stores projects on disk in a location based on a hash of the project’s ID. Hashed storage is different to legacy storage where a project is stored based on:

  • The project’s URL.
  • The folder structure where the repository is stored on disk.

This makes the folder structure immutable and eliminates the need to synchronize state from URLs to disk structure. This means that renaming a group, user, or project:

  • Costs only the database transaction.
  • Takes effect immediately.

The hash also helps spread the repositories more evenly on the disk. The top-level directory contains fewer folders than the total number of top-level namespaces.

The hash format is based on the hexadecimal representation of a SHA256, calculated with SHA256(project.id). The top-level folder uses the first two characters, followed by another folder with the next two characters. They are both stored in a special @hashed folder so they can co-exist with existing legacy storage projects. For example:

# Project's repository:
"@hashed/#{hash[0..1]}/#{hash[2..3]}/#{hash}.git"

# Wiki's repository:
"@hashed/#{hash[0..1]}/#{hash[2..3]}/#{hash}.wiki.git"

Translate hashed storage paths

Troubleshooting problems with the Git repositories, adding hooks, and other tasks requires you translate between the human-readable project name and the hashed storage path. You can translate:

From project name to hashed path

Administrators can look up a project’s hashed path from its name or ID using:

To look up a project’s hash path in the Admin Area:

  1. On the left sidebar, expand the top-most chevron ().
  2. Select Admin Area.
  3. On the left sidebar, select Overview > Projects and select the project.

The Gitaly relative path is displayed there and looks similar to:

"@hashed/b1/7e/b17ef6d19c7a5b1ee83b907c595526dcb1eb06db8227d650d5dda0a9f4ce8cd9.git"

To look up a project’s hash path using a Rails console:

  1. Start a Rails console.
  2. Run a command similar to this example (use either the project’s ID or its name):

    Project.find(16).disk_path
    Project.find_by_full_path('group/project').disk_path
    

From hashed path to project name

Administrators can look up a project’s name from its hashed storage path using:

  • A Rails console.
  • The config file in the *.git directory.

To look up a project’s name using the Rails console:

  1. Start a Rails console.
  2. Run a command similar to this example:

    ProjectRepository.find_by(disk_path: '@hashed/b1/7e/b17ef6d19c7a5b1ee83b907c595526dcb1eb06db8227d650d5dda0a9f4ce8cd9').project
    

The quoted string in that command is the directory tree you can find on your GitLab server. For example, on a default Linux package installation this would be /var/opt/gitlab/git-data/repositories/@hashed/b1/7e/b17ef6d19c7a5b1ee83b907c595526dcb1eb06db8227d650d5dda0a9f4ce8cd9.git with .git from the end of the directory name removed.

The output includes the project ID and the project name. For example:

=> #<Project id:16 it/supportteam/ticketsystem>

To look up a project’s name using the config file in the *.git directory:

  1. Locate the *.git directory. This directory is located in /var/opt/gitlab/git-data/repositories/@hashed/, where the first four characters of the hash are the first two directories in the path under @hashed/. For example, on a default Linux package installation the *.git directory of the hash b17eb17ef6d19c7a5b1ee83b907c595526dcb1eb06db8227d650d5dda0a9f4ce8cd9 would be /var/opt/gitlab/git-data/repositories/@hashed/b1/7e/b17ef6d19c7a5b1ee83b907c595526dcb1eb06db8227d650d5dda0a9f4ce8cd9.git.
  2. Open the config file and locate the fullpath= key under [gitlab].

Hashed object pools

Introduced in GitLab 12.1.

Object pools are repositories used to deduplicate forks of public and internal projects and contain the objects from the source project. Using objects/info/alternates, the source project and forks use the object pool for shared objects. For more information, see How Git object deduplication works in GitLab.

Objects are moved from the source project to the object pool when housekeeping is run on the source project. Object pool repositories are stored similarly to regular repositories in a directory called @pools instead of @hashed

# object pool paths
"@pools/#{hash[0..1]}/#{hash[2..3]}/#{hash}.git"
caution
Do not run git prune or git gc in object pool repositories, which are stored in the @pools directory. This can cause data loss in the regular repositories that depend on the object pool.

Group wiki storage

Introduced in GitLab 13.5.

Unlike project wikis that are stored in the @hashed directory, group wikis are stored in a directory called @groups. Like project wikis, group wikis follow the hashed storage folder convention, but use a hash of the group ID rather than the project ID.

For example:

# group wiki paths
"@groups/#{hash[0..1]}/#{hash[2..3]}/#{hash}.wiki.git"

Gitaly Cluster storage

If Gitaly Cluster is used, Praefect manages storage locations. For more information, see Praefect-generated replica paths.

Object storage support

This table shows which storable objects are storable in each storage type:

Storable objectLegacy storageHashed storageS3 compatibleGitLab version
RepositoryYesYes-10.0
AttachmentsYesYes-10.2
AvatarsYesNo--
PagesYesNo--
Docker RegistryYesNo--
CI/CD job logsNoNo--
CI/CD artifactsNoNoYes9.4 / 10.6
CI/CD cacheNoNoYes-
LFS objectsYesSimilarYes10.0 / 10.7
Repository poolsNoYes-11.6

Files stored in an S3-compatible endpoint can have the same advantages as hashed storage, as long as they are not prefixed with #{namespace}/#{project_name}. This is true for CI/CD cache and LFS objects.

Avatars

Each file is stored in a directory that matches the id assigned to it in the database. The filename is always avatar.png for user avatars. When an avatar is replaced, the Upload model is destroyed and a new one takes place with a different id.

CI/CD artifacts

CI/CD artifacts are:

LFS objects

LFS Objects in GitLab implement a similar storage pattern using two characters and two-level folders, following the Git implementation:

"shared/lfs-objects/#{oid[0..1}/#{oid[2..3]}/#{oid[4..-1]}"

# Based on object `oid`: `8909029eb962194cfb326259411b22ae3f4a814b5be4f80651735aeef9f3229c`, path will be:
"shared/lfs-objects/89/09/029eb962194cfb326259411b22ae3f4a814b5be4f80651735aeef9f3229c"

LFS objects are also S3-compatible.

Legacy storage

caution
In GitLab 13.0, legacy storage is deprecated. If you haven’t migrated to hashed storage yet, check the migration instructions. Support for legacy storage is scheduled to be removed in GitLab 14.0. In GitLab 13.0 and later, switching new projects to legacy storage is not possible. The option to choose between hashed and legacy storage in the Admin Area is disabled.

Legacy storage was the storage behavior prior to version GitLab 10.0. For historical reasons, GitLab replicated the same mapping structure from the projects URLs:

  • Project’s repository: #{namespace}/#{project_name}.git.
  • Project’s wiki: #{namespace}/#{project_name}.wiki.git.

This structure enabled you to migrate from existing solutions to GitLab, and for Administrators to find where the repository was stored. This approach also had some drawbacks:

  • Storage location concentrated a large number of top-level namespaces. The impact could be reduced by multiple repository storage paths.
  • Because backups were a snapshot of the same URL mapping, if you tried to recover a very old backup, you needed to verify whether any project had taken the place of an old removed or renamed project sharing the same URL. This meant that mygroup/myproject from your backup may not have been the same original project that was at that same URL today.
  • Any change in the URL needed to be reflected on disk (when groups, users, or projects were renamed. This could add a lot of load in big installations, especially if using any type of network-based file system.