Repository Storage Types
Introduced in GitLab 10.0.
Legacy Storage
Legacy Storage is the storage behavior prior to version 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 made it simple to migrate from existing solutions to GitLab and easy for Administrators to find where the repository is stored.
On the other hand this has some drawbacks:
Storage location will concentrate huge amount of top-level namespaces. The impact can be reduced by the introduction of multiple storage paths.
Because backups are a snapshot of the same URL mapping, if you try to recover a
very old backup, you need to verify whether any project has taken the place of
an old removed or renamed project sharing the same URL. This means that
mygroup/myproject from your backup may not be the same original project that
is at that same URL today.
Any change in the URL will need to be reflected on disk (when groups / users or projects are renamed). This can add a lot of load in big installations, especially if using any type of network based filesystem.
For GitLab Geo in particular: Geo does work with legacy storage, but in some edge cases due to race conditions it can lead to errors when a project is renamed multiple times in short succession, or a project is deleted and recreated under the same name very quickly. We expect these race events to be rare, and we have not observed a race condition side-effect happening yet.
This pattern also exists in other objects stored in GitLab, like issue Attachments, GitLab Pages artifacts, Docker Containers for the integrated Registry, etc.
Hashed Storage
Warning: Hashed storage is in Beta. For the latest updates, check the associated issue and please report any problems you encounter.
Hashed Storage is the new storage behavior we are rolling out with 10.0. Instead of coupling project URL and the folder structure where the repository will be stored on disk, we are coupling a hash, based on the project's ID. This makes the folder structure immutable, and therefore eliminates any requirement to synchronize state from URLs to disk structure. This means that renaming a group, user, or project will cost only the database transaction, and will take effect immediately.
The hash also helps to spread the repositories more evenly on the disk, so the top-level directory will contain less folders than the total amount of top-level namespaces.
The hash format is based on the hexadecimal representation of SHA256:
SHA256(project.id). The top-level folder uses the first 2 characters, followed
by another folder with the next 2 characters. They are both stored in a special
@hashed folder, to be able to co-exist with existing Legacy Storage projects:
# 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"How to migrate to Hashed Storage
In GitLab, go to Admin > Settings, find the Repository Storage section and select "Create new projects using hashed storage paths".
To migrate your existing projects to the new storage type, check the specific rake tasks.
Rollback
There is no automated rollback implemented. Below are the steps required to rollback from each storage migration.
The rollback has to be performed in the reverse order. To get into "Legacy" state, you need to rollback Attachments first, then Project.
Also note that if Geo is enabled, after the migration was triggered, an event is generated to replicate the operation on any Secondary node. That means the on disk changes will also need to be performed on these nodes as well. Database changes will propagate without issues.
You must make sure the migration event was already processed or otherwise it may migrate the files back to Hashed state again.
Attachments
To rollback single Attachment migration, rename aa/bb/abcdef1234567890... folder back to namespace/project.
Both folder names can be generated by the FileUploader.absolute_base_dir(project), you
just need to switch the version from the project back to the previous one.
project.storage_version
# => 2
FileUploader.absolute_base_dir(project)
# => "/opt/gitlab/embedded/service/gitlab-rails/public/uploads/@hashed/d4/73/d4735e3a265e16eee03f59718b9b5d03019c07d8b6c51f90da3a666eec13ab35"
project.storage_version = 1
FileUploader.absolute_base_dir(project)
# => "/opt/gitlab/embedded/service/gitlab-rails/public/uploads/gitlab/gitlab-shell-renamed"Project
To rollback single Project migration, move @hashed/aa/bb/aabbcdef1234567890abcdef.git and @hashed/aa/bb/aabbcdef1234567890abcdef.wiki.git
back to namespace/project.git and namespace/project.wiki.git respectively and switch the version from the project back to null.
Hashed Storage coverage
We are incrementally moving every storable object in GitLab to the Hashed Storage pattern. You can check the current coverage status below (and also see the issue).
Note that things stored in an S3 compatible endpoint will not have the downsides
mentioned earlier, if they are not prefixed with #{namespace}/#{project_name},
which is true for CI Cache and LFS Objects.
| Storable Object | Legacy Storage | Hashed Storage | S3 Compatible | GitLab Version |
|---|---|---|---|---|
| Repository | Yes | Yes | - | 10.0 |
| Attachments | Yes | Yes | - | 10.2 |
| Avatars | Yes | No | - | - |
| Pages | Yes | No | - | - |
| Docker Registry | Yes | No | - | - |
| CI Build Logs | No | No | - | - |
| CI Artifacts | No | No | Yes | 9.4 / 10.6 |
| CI Cache | No | No | Yes | - |
| LFS Objects | Yes | Similar | Yes | 10.0 / 10.7 |
Implementation Details
Avatars
Each file is stored in a folder with its id from the database. The filename is always avatar.png for user avatars.
When avatar is replaced, Upload model is destroyed and a new one takes place with different id.
CI Artifacts
CI Artifacts are S3 compatible since 9.4 (GitLab Premium), and available in GitLab Core since 10.6.
LFS Objects
LFS Objects implements a similar storage pattern using 2 chars, 2 level folders, following git own 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"They are also S3 compatible since 10.0 (GitLab Premium), and available in GitLab Core since 10.7.