ContextQMD
Back to Gitlabhq

Disaster recovery for GitLab Dedicated and GitLab Dedicated for Government

doc/administration/dedicated/disaster_recovery.md

19.3.06.2 KB
Original Source

{{< details >}}

  • Tier: Ultimate
  • Offering: GitLab Dedicated, GitLab Dedicated for Government

{{< /details >}}

GitLab Dedicated and GitLab Dedicated for Government provide disaster recovery to restore your instance if your primary region becomes unavailable. Recovery objectives, replication behavior, and the failover process depend on whether your environment has a secondary region configured for Geo-based failover.

Recovery objectives

{{< tabs >}}

{{< tab title="GitLab Dedicated" >}}

To be eligible for the full recovery objectives:

If no secondary region is configured, recovery is limited to backup restoration.

GitLab Dedicated provides disaster recovery with these recovery objectives:

  • Recovery Time Objective (RTO): Service is restored to your secondary region in eight hours or less.
  • Recovery Point Objective (RPO): Data loss is limited to a maximum of four hours of the most recent changes, depending on when the disaster occurs relative to the last backup.

{{< /tab >}}

{{< tab title="GitLab Dedicated for Government" >}}

GitLab Dedicated for Government does not use Geo-based secondary-region failover. Recovery relies on backup restoration. The recovery targets match those for GitLab Dedicated environments without a secondary region configured:

  • Recovery Time Objective (RTO): Service is restored within eight hours. Larger repositories or databases can extend recovery time.
  • Recovery Point Objective (RPO): Data loss is limited to a maximum of four hours of the most recent changes.

{{< /tab >}}

{{< /tabs >}}

Geo replication

{{< tabs >}}

{{< tab title="GitLab Dedicated" >}}

When you create your instance, you select a primary region and a secondary region for your environment. Geo continuously replicates data between these regions, including:

  • Database content
  • Repository storage
  • Object storage

{{< /tab >}}

{{< tab title="GitLab Dedicated for Government" >}}

GitLab Dedicated for Government does not support Geo. Your instance runs in a single AWS GovCloud region (US-West), with backups replicated to a separate AWS GovCloud region (US-East) for redundancy.

{{< /tab >}}

{{< /tabs >}}

Automated backups

GitLab performs automated backups of all GitLab Dedicated datastores (including databases and Git repositories) every four hours (six times daily) by creating snapshots.

Backups are tested, retained for 30 days, and stored in your chosen secondary region. They are also geographically replicated by AWS for additional protection.

Database backups:

  • Use continuous log-based backups in the primary region for point-in-time recovery.
  • Stream replication to the secondary region to provide a near-real-time copy.

Object storage backups use geographical replication and versioning to provide backup protection.

The four-hour backup frequency supports the Recovery Point Objective (RPO) to ensure you lose no more than four hours of data.

Disaster coverage

Disaster recovery covers these scenarios with guaranteed recovery objectives:

  • Partial region outage (for example, availability zone failure)
  • Complete outage of your primary region

These scenarios are covered on a best-effort basis without guaranteed recovery objectives:

  • Loss of both primary and secondary regions
  • Global internet outages
  • Data corruption issues

Service limitations

Disaster recovery has these service limitations:

  • Advanced search indexes are not continuously replicated. After failover, these indexes are rebuilt when the secondary region is promoted. Basic search remains available during rebuilding.
  • ClickHouse Cloud is provisioned only in the primary region. Features that require this service might be unavailable if the primary region is completely down.
  • Production preview environments do not have secondary instances.
  • Hosted runners are supported only in the primary region and cannot be rebuilt in the secondary instance.
  • Some regions have limited feature availability due to AWS service constraints. For more information, see supported regions. These feature limitations do not affect disaster recovery capabilities or RTO and RPO targets.

GitLab does not provide:

  • Programmatic monitoring of failover events
  • Customer-initiated disaster recovery testing

Failover process

{{< tabs >}}

{{< tab title="GitLab Dedicated" >}}

When your instance becomes unavailable due to a complete region failure or critical component failure that cannot be quickly recovered, the GitLab Dedicated team:

  1. Gets alerted by monitoring systems.
  2. Investigates if failover is required.
  3. If failover is required:
    1. Notifies you that failover is in progress.
    2. Promotes the secondary region to primary.
    3. Updates DNS records for <customer>.gitlab-dedicated.com to point to the newly promoted region.
    4. Notifies you when failover completes.

If you use PrivateLink, you must update your internal networking configuration to target the PrivateLink endpoint for the secondary region. To minimize downtime, configure equivalent PrivateLink endpoints in your secondary region before a disaster occurs.

The failover process typically completes in 90 minutes or less. Throughout the process, GitLab communicates with you through one or more of:

  • Your operational contact information in Switchboard
  • Slack
  • Support tickets

GitLab may establish a temporary Slack channel and Zoom bridge to coordinate with your team throughout the recovery process.

{{< /tab >}}

{{< tab title="GitLab Dedicated for Government" >}}

GitLab Dedicated for Government does not have a secondary region to fail over to. If your instance becomes unavailable due to a complete region failure, the GitLab Dedicated team restores your instance from the most recent backup in your backup region.

Throughout the process, GitLab communicates with you through one or more of:

  • Slack
  • Support tickets

{{< /tab >}}

{{< /tabs >}}