• Simple back up procedure
• Scaling backups
• What data needs to be backed up?
  • PostgreSQL databases
  • Git repositories
  • Blobs
    • Object storage
  • Container registry
  • Storing configuration files
  • Other data
• Command-line interface
  • Requirements
  • Backup command
  • Backup options
    • Backup strategy option
    • Backup filename
    • Backup compression
      • Default compression: Gzip with fastest method
      • Gzip with slowest method
      • No compression
      • Parallel compression with pigz
      • Parallel compression with zstd
    • Confirm archive can be transferred
    • Excluding specific data from the backup
    • Skipping tar creation
    • Create server-side repository backups
    • Back up Git repositories concurrently
    • Incremental repository backups
    • Back up specific repository storages
    • Back up specific repositories
    • Upload backups to a remote (cloud) storage
      • Using Amazon S3
      • S3 Encrypted Buckets
      • Digital Ocean Spaces
      • Other S3 Providers
      • Using Google Cloud Storage
      • Using Azure Blob storage
      • Specifying a custom directory for backups
    • Skip uploading backups to remote storage
    • Upload to locally-mounted shares
      • Avoid conflicting configuration
      • Configure uploads to locally-mounted shares
    • Backup archive permissions
    • Configuring cron to make daily backups
    • Limit backup lifetime for local files (prune old backups)
    • Back up and restore for installations using PgBouncer
      • Bypassing PgBouncer
    • gitaly-backup for repository backup and restore
• Alternative backup strategies
  • Back up repository data separately
    • Prevent writes and copy the Git repository data
    • Online backup through marking repositories as read-only (experimental)

Back up GitLab

The exact procedure for backing up GitLab depends on many factors. Your particular deployment’s usage and configuration determine what kind of data exists, where it is located, and how much there is. These factors influence your options for how to perform a back up, how to store it, and how to restore it.

Simple back up procedure

As a rough guideline, if you are using a 1k reference architecture with less than 100 GB of data, then follow these steps:

1. Run the backup command.
2. Back up object storage, if applicable.
3. Manually back up configuration files.

Scaling backups

As the volume of GitLab data grows, the backup command takes longer to execute. Backup options such as back up Git repositories concurrently and incremental repository backups can help to reduce execution time. At some point, the backup command becomes impractical by itself. For example, it can take 24 hours or more.

In some cases, architecture changes may be warranted to allow backups to scale. If you are using a GitLab reference architecture, see Back up and restore large reference architectures.

For more information, see alternative backup strategies.

What data needs to be backed up?

• PostgreSQL databases
• Git repositories
• Blobs
• Container registry
• Configuration files
• Other data

PostgreSQL databases

In the simplest case, GitLab has one PostgreSQL database in one PostgreSQL server on the same VM as all other GitLab services. But depending on configuration, GitLab may use multiple PostgreSQL databases in multiple PostgreSQL servers.

In general, this data is the single source of truth for most user-generated content in the Web interface, such as issue and merge request content, comments, permissions, and credentials.

PostgreSQL also holds some cached data like HTML-rendered Markdown, and by default, merge request diffs. However, merge request diffs can also be configured to be offloaded to the file system or object storage, see Blobs.

Gitaly Cluster’s Praefect service uses a PostgreSQL database as a single source of truth to manage its Gitaly nodes.

A common PostgreSQL utility, pg_dump, produces a backup file which can be used to restore a PostgreSQL database. The backup command uses this utility under the hood.

Unfortunately, the larger the database, the longer it takes pg_dump to execute. Depending on your situation, the duration becomes impractical at some point (days, for example). If your database is over 100 GB, pg_dump, and by extension the backup command, is likely not usable. For more information, see alternative backup strategies.

Git repositories

A GitLab instance can have one or more repository shards. Each shard is a Gitaly instance or Gitaly Cluster that is responsible for allowing access and operations on the locally stored Git repositories. Gitaly can run on a machine:

• With a single disk.
• With multiple disks mounted as a single mount-point (like with a RAID array).
• Using LVM.

Each project can have up to 3 different repositories:

• A project repository, where the source code is stored.
• A wiki repository, where the wiki content is stored.
• A design repository, where design artifacts are indexed (assets are actually in LFS).

They all live in the same shard and share the same base name with a -wiki and -design suffix for Wiki and Design Repository cases.

Personal and project snippets, and group wiki content, are stored in Git repositories.

Project forks are deduplicated in live a GitLab site using pool repositories.

The backup command produces a Git bundle for each repository and tars them all up. This duplicates pool repository data into every fork. In our testing, 100 GB of Git repositories took a little over 2 hours to back up and upload to S3. At around 400 GB of Git data, the backup command is likely not viable for regular backups. For more information, see alternative backup strategies.

Blobs

GitLab stores blobs (or files) such as issue attachments or LFS objects into either:

• The file system in a specific location.
• An Object Storage solution. Object Storage solutions can be:
  • Cloud based like Amazon S3 and Google Cloud Storage.
  • Hosted by you (like MinIO).
  • A Storage Appliance that exposes an Object Storage-compatible API.

Object storage

The backup command doesn’t back up blobs that aren’t stored on the file system. If you’re using object storage, be sure to enable backups with your object storage provider. For example, see:

• Amazon S3 backups
• Google Cloud Storage Transfer Service and Google Cloud Storage Object Versioning

Container registry

GitLab container registry storage can be configured in either:

• The file system in a specific location.
• An Object Storage solution. Object Storage solutions can be:
  • Cloud based like Amazon S3 and Google Cloud Storage.
  • Hosted by you (like MinIO).
  • A Storage Appliance that exposes an Object Storage-compatible API.

The backup command does not back up registry data when they are stored in Object Storage.

Storing configuration files

You should back up the configuration directory. At the very minimum, you must back up:

You may also want to back up any TLS keys and certificates (/etc/gitlab/ssl, /etc/gitlab/trusted-certs), and your SSH host keys to avoid man-in-the-middle attack warnings if you have to perform a full machine restore.

In the unlikely event that the secrets file is lost, see When the secrets file is lost.

Other data

GitLab uses Redis both as a cache store and to hold persistent data for our background jobs system, Sidekiq. The provided backup command does not back up Redis data. This means that in order to take a consistent backup with the backup command, there must be no pending or running background jobs. It is possible to manually back up Redis.

Elasticsearch is an optional database for advanced search. It can improve search in both source-code level, and user generated content in issues, merge requests, and discussions. The backup command does not back up Elasticsearch data. Elasticsearch data can be regenerated from PostgreSQL data after a restore. It is possible to manually back up Elasticsearch.

Command-line interface

GitLab provides a command-line interface to back up your entire instance, including:

• Database
• Attachments
• Git repositories data
• CI/CD job output logs
• CI/CD job artifacts
• LFS objects
• Terraform states
• Container registry images
• GitLab Pages content
• Packages
• Snippets
• Group wikis
• Project-level Secure Files (introduced in GitLab 16.1)
• External merge request diffs (introduced in GitLab 17.1)

Backups do not include:

• Mattermost data
• Redis (and thus Sidekiq jobs)
• Object storage on Linux package (Omnibus) / Docker / Self-compiled installations
• Global server hooks
• File hooks

Requirements

To be able to back up and restore, ensure that Rsync is installed on your system. If you installed GitLab:

• Using the Linux package, Rsync is already installed.

• Using self-compiled, check if rsync is installed. If Rsync is not installed, install it. For example:

  # Debian/Ubuntu
  sudo apt-get install rsync
  
  # RHEL/CentOS
  sudo yum install rsync
  

Backup command

sudo gitlab-backup create

docker exec -t <container name> gitlab-backup create

sudo -u git -H bundle exec rake gitlab:backup:create RAILS_ENV=production

If your GitLab deployment has multiple nodes, you need to pick a node for running the backup command. You must ensure that the designated node:

• is persistent, and not subject to auto-scaling.
• has the GitLab Rails application already installed. If Puma or Sidekiq is running, then Rails is installed.
• has sufficient storage and memory to produce the backup file.

Example output:

Dumping database tables:
- Dumping table events... [DONE]
- Dumping table issues... [DONE]
- Dumping table keys... [DONE]
- Dumping table merge_requests... [DONE]
- Dumping table milestones... [DONE]
- Dumping table namespaces... [DONE]
- Dumping table notes... [DONE]
- Dumping table projects... [DONE]
- Dumping table protected_branches... [DONE]
- Dumping table schema_migrations... [DONE]
- Dumping table services... [DONE]
- Dumping table snippets... [DONE]
- Dumping table taggings... [DONE]
- Dumping table tags... [DONE]
- Dumping table users... [DONE]
- Dumping table users_projects... [DONE]
- Dumping table web_hooks... [DONE]
- Dumping table wikis... [DONE]
Dumping repositories:
- Dumping repository abcd... [DONE]
Creating backup archive: <backup-id>_gitlab_backup.tar [DONE]
Deleting tmp directories...[DONE]
Deleting old backups... [SKIPPING]

For detailed information about the backup process, see Backup archive process.

Backup options

The command-line tool GitLab provides to back up your instance can accept more options.

Backup strategy option

The default backup strategy is to essentially stream data from the respective data locations to the backup using the Linux command tar and gzip. This works fine in most cases, but can cause problems when data is rapidly changing.

When data changes while tar is reading it, the error file changed as we read it may occur, and causes the backup process to fail. In that case, you can use the backup strategy called copy. The strategy copies data files to a temporary location before calling tar and gzip, avoiding the error.

A side-effect is that the backup process takes up to an additional 1X disk space. The process does its best to clean up the temporary files at each stage so the problem doesn’t compound, but it could be a considerable change for large installations.

To use the copy strategy instead of the default streaming strategy, specify STRATEGY=copy in the Rake task command. For example:

sudo gitlab-backup create STRATEGY=copy

Backup filename

Backup files are created with filenames according to specific defaults. However, you can override the <backup-id> portion of the filename by setting the BACKUP environment variable. For example:

sudo gitlab-backup create BACKUP=dump

The resulting file is named dump_gitlab_backup.tar. This is useful for systems that make use of rsync and incremental backups, and results in considerably faster transfer speeds.

Backup compression

By default, Gzip fast compression is applied during backup of:

• PostgreSQL database dumps.
• blobs, for example uploads, job artifacts, external merge request diffs.

The default command is gzip -c -1. You can override this command with COMPRESS_CMD. Similarly, you can override the decompression command with DECOMPRESS_CMD.

Caveats:

• The compression command is used in a pipeline, so your custom command must output to stdout.
• If you specify a command that is not packaged with GitLab, then you must install it yourself.
• The resultant filenames will still end in .gz.
• The default decompression command, used during restore, is gzip -cd. Therefore if you override the compression command to use a format that cannot be decompressed by gzip -cd, you must override the decompression command during restore.
• Do not place environment variables after the backup command. For example, gitlab-backup create COMPRESS_CMD="pigz -c --best" doesn’t work as intended.

Default compression: Gzip with fastest method

gitlab-backup create

Gzip with slowest method

COMPRESS_CMD="gzip -c --best" gitlab-backup create

If gzip was used for backup, then restore does not require any options:

gitlab-backup restore

No compression

If your backup destination has built-in automatic compression, then you may wish to skip compression.

The tee command pipes stdin to stdout.

COMPRESS_CMD=tee gitlab-backup create

And on restore:

DECOMPRESS_CMD=tee gitlab-backup restore

Parallel compression with pigz

An example of compressing backups with pigz using 4 processes:

COMPRESS_CMD="pigz --compress --stdout --fast --processes=4" sudo gitlab-backup create

Because pigz compresses to the gzip format, it is not required to use pigz to decompress backups which were compressed by pigz. However, it can still have a performance benefit over gzip. An example of decompressing backups with pigz:

DECOMPRESS_CMD="pigz --decompress --stdout" sudo gitlab-backup restore

Parallel compression with zstd

An example of compressing backups with zstd using 4 threads:

COMPRESS_CMD="zstd --compress --stdout --fast --threads=4" sudo gitlab-backup create

An example of decompressing backups with zstd:

DECOMPRESS_CMD="zstd --decompress --stdout" sudo gitlab-backup restore

Confirm archive can be transferred

To ensure the generated archive is transferable by rsync, you can set the GZIP_RSYNCABLE=yes option. This sets the --rsyncable option to gzip, which is useful only in combination with setting the Backup filename option.

The --rsyncable option in gzip isn’t guaranteed to be available on all distributions. To verify that it’s available in your distribution, run gzip --help or consult the man pages.

sudo gitlab-backup create BACKUP=dump GZIP_RSYNCABLE=yes

Excluding specific data from the backup

Depending on your installation type, slightly different components can be skipped on backup creation.

sudo gitlab-backup create SKIP=db,uploads

sudo -u git -H bundle exec rake gitlab:backup:create SKIP=db,uploads RAILS_ENV=production

SKIP= is also used to:

• Skip creation of the tar file (SKIP=tar).
• Skip uploading the backup to remote storage (SKIP=remote).

Skipping tar creation

The last part of creating a backup is generation of a .tar file containing all the parts. In some cases, creating a .tar file might be wasted effort or even directly harmful, so you can skip this step by adding tar to the SKIP environment variable. Example use-cases:

• When the backup is picked up by other backup software.
• To speed up incremental backups by avoiding having to extract the backup every time. (In this case, PREVIOUS_BACKUP and BACKUP must not be specified, otherwise the specified backup is extracted, but no .tar file is generated at the end.)

Adding tar to the SKIP variable leaves the files and directories containing the backup in the directory used for the intermediate files. These files are overwritten when a new backup is created, so you should make sure they are copied elsewhere, because you can only have one backup on the system.

sudo gitlab-backup create SKIP=tar

sudo -u git -H bundle exec rake gitlab:backup:create SKIP=tar RAILS_ENV=production

Create server-side repository backups

Instead of storing large repository backups in the backup archive, 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.

1. Configure a server-side backup destination in Gitaly.
2. Create a back up using the repositories server-side option. See the following examples.

sudo gitlab-backup create REPOSITORIES_SERVER_SIDE=true

sudo -u git -H bundle exec rake gitlab:backup:create REPOSITORIES_SERVER_SIDE=true

kubectl exec <Toolbox pod name> -it -- backup-utility --repositories-server-side

Back up Git repositories concurrently

When using multiple repository storages, repositories can be backed up or restored concurrently to help fully use CPU time. The following variables are available to modify the default behavior of the Rake task:

• GITLAB_BACKUP_MAX_CONCURRENCY: The maximum number of projects to back up at the same time. Defaults to the number of logical CPUs.
• GITLAB_BACKUP_MAX_STORAGE_CONCURRENCY: The maximum number of projects to back up at the same time on each storage. This allows the repository backups to be spread across storages. Defaults to 2.

For example, with 4 repository storages:

sudo gitlab-backup create GITLAB_BACKUP_MAX_CONCURRENCY=4 GITLAB_BACKUP_MAX_STORAGE_CONCURRENCY=1

sudo -u git -H bundle exec rake gitlab:backup:create GITLAB_BACKUP_MAX_CONCURRENCY=4 GITLAB_BACKUP_MAX_STORAGE_CONCURRENCY=1

Incremental repository backups

Incremental repository backups can be faster than full repository backups because they only pack changes since the last backup into the backup bundle for each repository. The incremental backup archives are not linked to each other: each archive is a self-contained backup of the instance. There must be an existing backup to create an incremental backup from.

Use the PREVIOUS_BACKUP=<backup-id> option to choose the backup to use. By default, a backup file is created as documented in the Backup ID section. You can override the <backup-id> portion of the filename by setting the BACKUP environment variable.

To create an incremental backup, run:

sudo gitlab-backup create INCREMENTAL=yes PREVIOUS_BACKUP=<backup-id>

To create an untarred incremental backup from a tarred backup, use SKIP=tar:

sudo gitlab-backup create INCREMENTAL=yes SKIP=tar

Back up specific repository storages

When using multiple repository storages, repositories from specific repository storages can be backed up separately using the REPOSITORIES_STORAGES option. The option accepts a comma-separated list of storage names.

For example:

sudo gitlab-backup create REPOSITORIES_STORAGES=storage1,storage2

sudo -u git -H bundle exec rake gitlab:backup:create REPOSITORIES_STORAGES=storage1,storage2

Back up specific repositories

You can back up specific repositories using the REPOSITORIES_PATHS option. Similarly, you can use SKIP_REPOSITORIES_PATHS to skip certain repositories. Both options accept a comma-separated list of project or group paths. If you specify a group path, all repositories in all projects in the group and descendent groups are included or skipped, depending on which option you used.

For example, to back up all repositories for all projects in Group A (group-a), the repository for Project C in Group B (group-b/project-c), and skip the Project D in Group A (group-a/project-d):

  sudo gitlab-backup create REPOSITORIES_PATHS=group-a,group-b/project-c SKIP_REPOSITORIES_PATHS=group-a/project-d

  sudo -u git -H bundle exec rake gitlab:backup:create REPOSITORIES_PATHS=group-a,group-b/project-c SKIP_REPOSITORIES_PATHS=group-a/project-d

Upload backups to a remote (cloud) storage

You can let the backup script upload (using the Fog library) the .tar file it creates. In the following example, we use Amazon S3 for storage, but Fog also lets you use other storage providers. GitLab also imports cloud drivers for AWS, Google, and Aliyun. A local driver is also available.

Read more about using object storage with GitLab.

Using Amazon S3

For Linux package (Omnibus):

1. Add the following to /etc/gitlab/gitlab.rb:

   gitlab_rails['backup_upload_connection'] = {
     'provider' => 'AWS',
     'region' => 'eu-west-1',
     'aws_access_key_id' => 'AKIAKIAKI',
     'aws_secret_access_key' => 'secret123'
     # If using an IAM Profile, don't configure aws_access_key_id & aws_secret_access_key
     # 'use_iam_profile' => true
   }
   gitlab_rails['backup_upload_remote_directory'] = 'my.s3.bucket'
   # Consider using multipart uploads when file size reaches 100MB. Enter a number in bytes.
   # gitlab_rails['backup_multipart_chunk_size'] = 104857600
   

2. Reconfigure GitLab for the changes to take effect

S3 Encrypted Buckets

AWS supports these modes for server side encryption:

• Amazon S3-Managed Keys (SSE-S3)
• Customer Master Keys (CMKs) Stored in AWS Key Management Service (SSE-KMS)
• Customer-Provided Keys (SSE-C)

Use your mode of choice with GitLab. Each mode has similar, but slightly different, configuration methods.

SSE-S3

To enable SSE-S3, in the backup storage options set the server_side_encryption field to AES256. For example, in the Linux package (Omnibus):

gitlab_rails['backup_upload_storage_options'] = {
  'server_side_encryption' => 'AES256'
}

SSE-KMS

To enable SSE-KMS, you need the KMS key via its Amazon Resource Name (ARN) in the arn:aws:kms:region:acct-id:key/key-id format. Under the backup_upload_storage_options configuration setting, set:

• server_side_encryption to aws:kms.
• server_side_encryption_kms_key_id to the ARN of the key.

For example, in the Linux package (Omnibus):

gitlab_rails['backup_upload_storage_options'] = {
  'server_side_encryption' => 'aws:kms',
  'server_side_encryption_kms_key_id' => 'arn:aws:<YOUR KMS KEY ID>:'
}

SSE-C

SSE-C requires you to set these encryption options:

• backup_encryption: AES256.
• backup_encryption_key: Unencoded, 32-byte (256 bits) key. The upload fails if this isn’t exactly 32 bytes.

For example, in the Linux package (Omnibus):

gitlab_rails['backup_encryption'] = 'AES256'
gitlab_rails['backup_encryption_key'] = '<YOUR 32-BYTE KEY HERE>'

If the key contains binary characters and cannot be encoded in UTF-8, instead, specify the key with the GITLAB_BACKUP_ENCRYPTION_KEY environment variable. For example:

gitlab_rails['env'] = { 'GITLAB_BACKUP_ENCRYPTION_KEY' => "\xDE\xAD\xBE\xEF" * 8 }

Digital Ocean Spaces

This example can be used for a bucket in Amsterdam (AMS3):

1. Add the following to /etc/gitlab/gitlab.rb:

   gitlab_rails['backup_upload_connection'] = {
     'provider' => 'AWS',
     'region' => 'ams3',
     'aws_access_key_id' => 'AKIAKIAKI',
     'aws_secret_access_key' => 'secret123',
     'endpoint'              => 'https://ams3.digitaloceanspaces.com'
   }
   gitlab_rails['backup_upload_remote_directory'] = 'my.s3.bucket'
   

2. Reconfigure GitLab for the changes to take effect

If you see a 400 Bad Request error message when using Digital Ocean Spaces, the cause may be the use of backup encryption. Because Digital Ocean Spaces doesn’t support encryption, remove or comment the line that contains gitlab_rails['backup_encryption'].

Other S3 Providers

Not all S3 providers are fully compatible with the Fog library. For example, if you see a 411 Length Required error message after attempting to upload, you may need to downgrade the aws_signature_version value from the default value to 2, due to this issue.

For self-compiled installations:

1. Edit home/git/gitlab/config/gitlab.yml:

     backup:
       # snip
       upload:
         # Fog storage connection settings, see https://fog.io/storage/ .
         connection:
           provider: AWS
           region: eu-west-1
           aws_access_key_id: AKIAKIAKI
           aws_secret_access_key: 'secret123'
           # If using an IAM Profile, leave aws_access_key_id & aws_secret_access_key empty
           # ie. aws_access_key_id: ''
           # use_iam_profile: 'true'
         # The remote 'directory' to store your backups. For S3, this would be the bucket name.
         remote_directory: 'my.s3.bucket'
         # Specifies Amazon S3 storage class to use for backups, this is optional
         # storage_class: 'STANDARD'
         #
         # Turns on AWS Server-Side Encryption with Amazon Customer-Provided Encryption Keys for backups, this is optional
         #   'encryption' must be set in order for this to have any effect.
         #   'encryption_key' should be set to the 256-bit encryption key for Amazon S3 to use to encrypt or decrypt.
         #   To avoid storing the key on disk, the key can also be specified via the `GITLAB_BACKUP_ENCRYPTION_KEY` your data.
         # encryption: 'AES256'
         # encryption_key: '<key>'
         #
         #
         # Turns on AWS Server-Side Encryption with Amazon S3-Managed keys (optional)
         # https://docs.aws.amazon.com/AmazonS3/latest/userguide/serv-side-encryption.html
         # For SSE-S3, set 'server_side_encryption' to 'AES256'.
         # For SS3-KMS, set 'server_side_encryption' to 'aws:kms'. Set
         # 'server_side_encryption_kms_key_id' to the ARN of customer master key.
         # storage_options:
         #   server_side_encryption: 'aws:kms'
         #   server_side_encryption_kms_key_id: 'arn:aws:kms:YOUR-KEY-ID-HERE'
   

2. Restart GitLab for the changes to take effect

If you’re uploading your backups to S3, you should create a new IAM user with restricted access rights. To give the upload user access only for uploading backups create the following IAM profile, replacing my.s3.bucket with the name of your bucket:

{
  "Version": "2012-10-17",
  "Statement": [
    {
      "Sid": "Stmt1412062044000",
      "Effect": "Allow",
      "Action": [
        "s3:AbortMultipartUpload",
        "s3:GetBucketAcl",
        "s3:GetBucketLocation",
        "s3:GetObject",
        "s3:GetObjectAcl",
        "s3:ListBucketMultipartUploads",
        "s3:PutObject",
        "s3:PutObjectAcl"
      ],
      "Resource": [
        "arn:aws:s3:::my.s3.bucket/*"
      ]
    },
    {
      "Sid": "Stmt1412062097000",
      "Effect": "Allow",
      "Action": [
        "s3:GetBucketLocation",
        "s3:ListAllMyBuckets"
      ],
      "Resource": [
        "*"
      ]
    },
    {
      "Sid": "Stmt1412062128000",
      "Effect": "Allow",
      "Action": [
        "s3:ListBucket"
      ],
      "Resource": [
        "arn:aws:s3:::my.s3.bucket"
      ]
    }
  ]
}

Using Google Cloud Storage

To use Google Cloud Storage to save backups, you must first create an access key from the Google console:

1. Go to the Google storage settings page.
2. Select Interoperability, and then create an access key.
3. Make note of the Access Key and Secret and replace them in the following configurations.
4. In the buckets advanced settings ensure the Access Control option Set object-level and bucket-level permissions is selected.
5. Ensure you have already created a bucket.

For the Linux package (Omnibus):

1. Edit /etc/gitlab/gitlab.rb:

   gitlab_rails['backup_upload_connection'] = {
     'provider' => 'Google',
     'google_storage_access_key_id' => 'Access Key',
     'google_storage_secret_access_key' => 'Secret',
   
     ## If you have CNAME buckets (foo.example.com), you might run into SSL issues
     ## when uploading backups ("hostname foo.example.com.storage.googleapis.com
     ## does not match the server certificate"). In that case, uncomment the following
     ## setting. See: https://github.com/fog/fog/issues/2834
     #'path_style' => true
   }
   gitlab_rails['backup_upload_remote_directory'] = 'my.google.bucket'
   

2. Reconfigure GitLab for the changes to take effect

For self-compiled installations:

1. Edit home/git/gitlab/config/gitlab.yml:

     backup:
       upload:
         connection:
           provider: 'Google'
           google_storage_access_key_id: 'Access Key'
           google_storage_secret_access_key: 'Secret'
         remote_directory: 'my.google.bucket'
   

2. Restart GitLab for the changes to take effect

Using Azure Blob storage

For more details, see the table of Azure parameters.

Specifying a custom directory for backups

This option works only for remote storage. If you want to group your backups, you can pass a DIRECTORY environment variable:

sudo gitlab-backup create DIRECTORY=daily
sudo gitlab-backup create DIRECTORY=weekly

Skip uploading backups to remote storage

If you have configured GitLab to upload backups in a remote storage, you can use the SKIP=remote option to skip uploading your backups to the remote storage.

sudo gitlab-backup create SKIP=remote

sudo -u git -H bundle exec rake gitlab:backup:create SKIP=remote RAILS_ENV=production

Upload to locally-mounted shares

You can send backups to a locally-mounted share (for example, NFS,CIFS, or SMB) using the Fog Local storage provider.

To do this, you must set the following configuration keys:

• backup_upload_connection.local_root: mounted directory that backups are copied to.
• backup_upload_remote_directory: subdirectory of the backup_upload_connection.local_root directory. It is created if it doesn’t exist. If you want to copy the tarballs to the root of your mounted directory, use ..

When mounted, the directory set in the local_root key must be owned by either:

• The git user. So, mounting with the uid= of the git user for CIFS and SMB.
• The user that you are executing the backup tasks as. For the Linux package (Omnibus), this is the git user.

Because file system performance may affect overall GitLab performance, we don’t recommend using cloud-based file systems for storage.

Avoid conflicting configuration

Don’t set the following configuration keys to the same path:

• gitlab_rails['backup_path'] (backup.path for self-compiled installations).
• gitlab_rails['backup_upload_connection'].local_root (backup.upload.connection.local_root for self-compiled installations).

The backup_path configuration key sets the local location of the backup file. The upload configuration key is intended for use when the backup file is uploaded to a separate server, perhaps for archival purposes.

If these configuration keys are set to the same location, the upload feature fails because a backup already exists at the upload location. This failure causes the upload feature to delete the backup because it assumes it’s a residual file remaining after the failed upload attempt.

Configure uploads to locally-mounted shares

Backup archive permissions

The backup archives created by GitLab (1393513186_2014_02_27_gitlab_backup.tar) have the owner/group git/git and 0600 permissions by default. This is meant to avoid other system users reading GitLab data. If you need the backup archives to have different permissions, you can use the archive_permissions setting.

Configuring cron to make daily backups

You can schedule a cron job that backs up your repositories and GitLab metadata.

The CRON=1 environment setting directs the backup script to hide all progress output if there aren’t any errors. This is recommended to reduce cron spam. When troubleshooting backup problems, however, replace CRON=1 with --trace to log verbosely.

Limit backup lifetime for local files (prune old backups)

To prevent regular backups from using all your disk space, you may want to set a limited lifetime for backups. The next time the backup task runs, backups older than the backup_keep_time are pruned.

This configuration option manages only local files. GitLab doesn’t prune old files stored in a third-party object storage because the user may not have permission to list and delete files. It’s recommended that you configure the appropriate retention policy for your object storage (for example, AWS S3).

Back up and restore for installations using PgBouncer

Do not back up or restore GitLab through a PgBouncer connection. These tasks must bypass PgBouncer and connect directly to the PostgreSQL primary database node, or they cause a GitLab outage.

When the GitLab backup or restore task is used with PgBouncer, the following error message is shown:

ActiveRecord::StatementInvalid: PG::UndefinedTable

Each time the GitLab backup runs, GitLab starts generating 500 errors and errors about missing tables will be logged by PostgreSQL:

ERROR: relation "tablename" does not exist at character 123

This happens because the task uses pg_dump, which sets a null search path and explicitly includes the schema in every SQL query to address CVE-2018-1058.

Because connections are reused with PgBouncer in transaction pooling mode, PostgreSQL fails to search the default public schema. As a result, this clearing of the search path causes tables and columns to appear missing.

Bypassing PgBouncer

There are two ways to fix this:

1. Use environment variables to override the database settings for the backup task.
2. Reconfigure a node to connect directly to the PostgreSQL primary database node.

Environment variable overrides

By default, GitLab uses the database configuration stored in a configuration file (database.yml). However, you can override the database settings for the backup and restore task by setting environment variables that are prefixed with GITLAB_BACKUP_:

• GITLAB_BACKUP_PGHOST
• GITLAB_BACKUP_PGUSER
• GITLAB_BACKUP_PGPORT
• GITLAB_BACKUP_PGPASSWORD
• GITLAB_BACKUP_PGSSLMODE
• GITLAB_BACKUP_PGSSLKEY
• GITLAB_BACKUP_PGSSLCERT
• GITLAB_BACKUP_PGSSLROOTCERT
• GITLAB_BACKUP_PGSSLCRL
• GITLAB_BACKUP_PGSSLCOMPRESSION

For example, to override the database host and port to use 192.168.1.10 and port 5432 with the Linux package (Omnibus):

sudo GITLAB_BACKUP_PGHOST=192.168.1.10 GITLAB_BACKUP_PGPORT=5432 /opt/gitlab/bin/gitlab-backup create

If you run GitLab on multiple databases, you can override database settings by including the database name in the environment variable. For example if your main and ci databases are hosted on different database servers, you would append their name after the GITLAB_BACKUP_ prefix, leaving the PG* names as is:

sudo GITLAB_BACKUP_MAIN_PGHOST=192.168.1.10 GITLAB_BACKUP_CI_PGHOST=192.168.1.12 /opt/gitlab/bin/gitlab-backup create

See the PostgreSQL documentation for more details on what these parameters do.

gitaly-backup for repository backup and restore

The gitaly-backup binary is used by the backup Rake task to create and restore repository backups from Gitaly. gitaly-backup replaces the previous backup method that directly calls RPCs on Gitaly from GitLab.

The backup Rake task must be able to find this executable. In most cases, you don’t need to change the path to the binary as it should work fine with the default path /opt/gitlab/embedded/bin/gitaly-backup. If you have a specific reason to change the path, it can be configured in the Linux package (Omnibus):

1. Add the following to /etc/gitlab/gitlab.rb:

   gitlab_rails['backup_gitaly_backup_path'] = '/path/to/gitaly-backup'
   

2. Reconfigure GitLab for the changes to take effect.

Alternative backup strategies

Because every deployment may have different capabilities, you should first review what data needs to be backed up to better understand if, and how, you can leverage them.

For example, if you use Amazon RDS, you might choose to use its built-in backup and restore features to handle your GitLab PostgreSQL data, and exclude PostgreSQL data when using the backup command.

In the following cases, consider using file system data transfer or snapshots as part of your backup strategy:

• Your GitLab instance contains a lot of Git repository data and the GitLab backup script is too slow.
• Your GitLab instance has a lot of forked projects and the regular backup task duplicates the Git data for all of them.
• Your GitLab instance has a problem and using the regular backup and import Rake tasks isn’t possible.

When considering using file system data transfer or snapshots:

• Don’t use these methods to migrate from one operating system to another. The operating systems of the source and destination should be as similar as possible. For example, don’t use these methods to migrate from Ubuntu to RHEL.
• Data consistency is very important. You should stop GitLab (sudo gitlab-ctl stop) before doing a file system transfer (with rsync, for example) or taking a snapshot to ensure all data in memory is flushed to disk. GitLab consists of multiple subsystems (Gitaly, database, file storage) that have their own buffers, queues, and storage layers. GitLab transactions can span these subsystems, which results in parts of a transaction taking different paths to disk. On live systems, file system transfers and snapshot runs fail to capture parts of the transaction still in memory.

Example: Amazon Elastic Block Store (EBS)

• A GitLab server using the Linux package (Omnibus) hosted on Amazon AWS.
• An EBS drive containing an ext4 file system is mounted at /var/opt/gitlab.
• In this case you could make an application backup by taking an EBS snapshot.
• The backup includes all repositories, uploads and PostgreSQL data.

Example: Logical Volume Manager (LVM) snapshots + rsync

• A GitLab server using the Linux package (Omnibus), with an LVM logical volume mounted at /var/opt/gitlab.
• Replicating the /var/opt/gitlab directory using rsync would not be reliable because too many files would change while rsync is running.
• Instead of rsync-ing /var/opt/gitlab, we create a temporary LVM snapshot, which we mount as a read-only file system at /mnt/gitlab_backup.
• Now we can have a longer running rsync job which creates a consistent replica on the remote server.
• The replica includes all repositories, uploads and PostgreSQL data.

If you’re running GitLab on a virtualized server, you can possibly also create VM snapshots of the entire GitLab server. It’s not uncommon however for a VM snapshot to require you to power down the server, which limits this solution’s practical use.

Back up repository data separately

First, ensure you back up existing GitLab data while skipping repositories:

sudo gitlab-backup create SKIP=repositories

sudo -u git -H bundle exec rake gitlab:backup:create SKIP=repositories RAILS_ENV=production

For manually backing up the Git repository data on disk, there are multiple possible strategies:

• Use snapshots, such as the previous examples of Amazon EBS drive snapshots, or LVM snapshots + rsync.
• Use GitLab Geo and rely on the repository data on a Geo secondary site.
• Prevent writes and copy the Git repository data.
• Create an online backup by marking repositories as read-only (experimental).

Prevent writes and copy the Git repository data

Git repositories must be copied in a consistent way. They should not be copied during concurrent write operations, as this can lead to inconsistencies or corruption issues. For more details, issue #270422 has a longer discussion explaining the potential problems.

To prevent writes to the Git repository data, there are two possible approaches:

• Use maintenance mode to place GitLab in a read-only state.

• Create explicit downtime by stopping all Gitaly services before backing up the repositories:

  sudo gitlab-ctl stop gitaly
  # execute git data copy step
  sudo gitlab-ctl start gitaly
  

You can copy Git repository data using any method, as long as writes are prevented on the data being copied (to prevent inconsistencies and corruption issues). In order of preference and safety, the recommended methods are:

1. Use rsync with archive-mode, delete, and checksum options, for example:

   rsync -aR --delete --checksum source destination # be extra safe with the order as it will delete existing data if inverted
   

2. Use a tar pipe to copy the entire repository’s directory to another server or location.

3. Use sftp, scp, cp, or any other copying method.

Online backup through marking repositories as read-only (experimental)

One way of backing up repositories without requiring instance-wide downtime is to programmatically mark projects as read-only while copying the underlying data.

There are a few possible downsides to this:

• Repositories are read-only for a period of time that scales with the size of the repository.
• Backups take a longer time to complete due to marking each project as read-only, potentially leading to inconsistencies. For example, a possible date discrepancy between the last data available for the first project that gets backed up compared to the last project that gets backed up.
• Fork networks should be entirely read-only while the projects inside get backed up to prevent potential changes to the pool repository.

There is an experimental script that attempts to automate this process in the Geo team Runbooks project.