Geo

Tier: Premium, Ultimate Offering: Self-managed

Geo is the solution for widely distributed development teams and for providing a warm-standby as part of a disaster recovery strategy.

caution
Geo undergoes significant changes from release to release. Upgrades are supported and documented, but you should ensure that you’re using the right version of the documentation for your installation.

Fetching large repositories can take a long time for teams and runners located far from a single GitLab instance.

Geo provides local caches that can be placed geographically close to remote teams which can serve read requests. This can reduce the time it takes to clone and fetch large repositories, speeding up development and increasing the productivity of your remote teams.

Geo secondary sites transparently proxy write requests to the primary site. All Geo sites can be configured to respond to a single GitLab URL, to deliver a consistent, seamless, and comprehensive experience whichever site the user lands on.

To make sure you’re using the right version of the documentation, go to the Geo page on GitLab.com and choose the appropriate release from the Switch branch/tag dropdown list. For example, v15.7.6-ee.

Geo uses a set of defined terms that are described in the Geo Glossary. Be sure to familiarize yourself with those terms.

Use cases

Implementing Geo provides the following benefits:

  • Reduce from minutes to seconds the time taken for your distributed developers to clone and fetch large repositories and projects.
  • Enable all of your developers to contribute ideas and work in parallel, no matter where they are.
  • Balance the read load between your primary and secondary sites.

In addition, it:

  • Can be used for cloning and fetching projects, in addition to reading any data available in the GitLab web interface (see limitations).
  • Overcomes slow connections between distant offices, saving time by improving speed for distributed teams.
  • Helps reducing the loading time for automated tasks, custom integrations, and internal workflows.
  • Can quickly fail over to a secondary site in a disaster recovery scenario.
  • Allows planned failover to a secondary site.

Geo provides:

  • A complete GitLab experience on Secondary sites: Maintain one primary GitLab site while enabling secondary sites with full read and write and UI experience for each of your distributed teams.
  • Authentication system hooks: Secondary sites receive all authentication data (like user accounts and logins) from the primary instance.

Gitaly Cluster

Geo should not be confused with Gitaly Cluster. For more information about the difference between Geo and Gitaly Cluster, see Comparison to Geo.

How it works

This is a brief summary of how Geo works in your GitLab environment. For a more detailed information, see the Geo Development page.

Your Geo instance can be used for cloning and fetching projects, in addition to reading any data. This makes working with large repositories over large distances much faster.

Geo overview

When Geo is enabled, the:

  • Original instance is known as the primary site.
  • Replicating sites are known as secondary sites.

Keep in mind that:

  • Secondary sites talk to the primary site to:
    • Get user data for logins (API).
    • Replicate repositories, LFS Objects, and Attachments (HTTPS + JWT).
  • The primary site talks to the secondary sites for viewing replication details. The primary does a GraphQL query against the secondary site for sync and verification data (API).
  • You can push directly to a secondary site (for both HTTP and SSH, including Git LFS), and it will proxy the requests to the primary site.
  • There are limitations when using Geo.

Architecture

The following diagram illustrates the underlying architecture of Geo.

Geo architecture

In this diagram:

  • There is the primary site and the details of one secondary site.
  • Writes to the database can only be performed on the primary site. A secondary site receives database updates by using PostgreSQL streaming replication.
  • If present, the LDAP server should be configured to replicate for Disaster Recovery scenarios.
  • A secondary site performs different type of synchronizations against the primary site, using a special authorization protected by JWT:
    • Repositories are cloned/updated via Git over HTTPS.
    • Attachments, LFS objects, and other files are downloaded via HTTPS using a private API endpoint.

From the perspective of a user performing Git operations:

  • The primary site behaves as a full read-write GitLab instance.
  • Secondary sites behave as full read-write GitLab instances. Secondary sites transparently proxy all operations to the primary site, with some notable exceptions. In particular, Git fetches are served by the secondary site when it is up-to-date.

From the perspective of a user browsing the GitLab UI, or using the API:

  • The primary site behaves as a full read-write GitLab instance.
  • Secondary sites behave as full read-write GitLab instances. Secondary sites transparently proxy all operations to the primary site, with some notable exceptions. In particular, web UI assets are served by the secondary site.

To simplify the diagram, some necessary components are omitted.

A secondary site needs two different PostgreSQL databases:

In secondary sites, there is an additional daemon: Geo Log Cursor.

Requirements for running Geo

The following are required to run Geo:

Additionally, check the GitLab minimum requirements, and use the latest version of GitLab for a better experience.

Firewall rules

The following table lists basic ports that must be open between the primary and secondary sites for Geo. To simplify failovers, you should open ports in both directions.

Source site Source port Destination site Destination port Protocol
Primary Any Secondary 80 TCP (HTTP)
Primary Any Secondary 443 TCP (HTTPS)
Secondary Any Primary 80 TCP (HTTP)
Secondary Any Primary 443 TCP (HTTPS)
Secondary Any Primary 5432 TCP

See the full list of ports used by GitLab in Package defaults

note
Web terminal support requires your load balancer to correctly handle WebSocket connections. When using HTTP or HTTPS proxying, your load balancer must be configured to pass through the Connection and Upgrade hop-by-hop headers. See the web terminal integration guide for more details.
note
When using HTTPS protocol for port 443, you must add an SSL certificate to the load balancers. If you wish to terminate SSL at the GitLab application server instead, use TCP protocol.
note
If you are only using HTTPS for external/internal URLs, it is not necessary to open port 80 in the firewall.

Internal URL

HTTP requests from any Geo secondary site to the primary Geo site use the Internal URL of the primary Geo site. If this is not explicitly defined in the primary Geo site settings in the Admin area, the public URL of the primary site is used.

To update the internal URL of the primary Geo site:

  1. On the left sidebar, at the bottom, select Admin.
  2. Select Geo > Sites.
  3. Select Edit on the primary site.
  4. Change the Internal URL, then select Save changes.

Geo Tracking Database

The tracking database instance is used as metadata to control what needs to be updated on the local instance. For example:

  • Download new assets.
  • Fetch new LFS Objects.
  • Fetch changes from a repository that has recently been updated.

Because the replicated database instance is read-only, we need this additional database instance for each secondary site.

Geo Log Cursor

This daemon:

  • Reads a log of events replicated by the primary site to the secondary database instance.
  • Updates the Geo Tracking Database instance with changes that must be executed.

When something is marked to be updated in the tracking database instance, asynchronous jobs running on the secondary site execute the required operations and update the state.

This new architecture allows GitLab to be resilient to connectivity issues between the sites. It doesn’t matter how long the secondary site is disconnected from the primary site as it is able to replay all the events in the correct order and become synchronized with the primary site again.

Limitations

caution
This list of limitations only reflects the latest version of GitLab. If you are using an older version, extra limitations may be in place.
  • Pushing directly to a secondary site redirects (for HTTP) or proxies (for SSH) the request to the primary site instead of handling it directly. The limitation is that you cannot use Git over HTTP with credentials embedded in the URI, for example, https://user:personal-access-token@secondary.tld. For more information, see how to use a Geo Site.
  • The primary site has to be online for OAuth login to happen. Existing sessions and Git are not affected. Support for the secondary site to use an OAuth provider independent from the primary is being planned.
  • The installation takes multiple manual steps that together can take about an hour depending on circumstances. Consider using the GitLab Environment Toolkit Terraform and Ansible scripts to deploy and operate production GitLab instances based on our Reference Architectures, including automation of common daily tasks. Epic 1465 proposes to improve Geo installation even more.
  • Real-time updates of issues/merge requests (for example, via long polling) doesn’t work on secondary sites where http proxying is disabled.
  • Selective synchronization only limits what repositories and files are replicated. The entire PostgreSQL data is still replicated. Selective synchronization is not built to accommodate compliance / export control use cases.
  • Pages access control doesn’t work on secondaries. See GitLab issue #9336 for details.
  • Disaster recovery for deployments that have multiple secondary sites causes downtime due to the need to re-initialize PostgreSQL streaming replication on all non-promoted secondaries to follow the new primary site.
  • For Git over SSH, to make the project clone URL display correctly regardless of which site you are browsing, secondary sites must use the same port as the primary. GitLab issue #339262 proposes to remove this limitation.
  • Git push over SSH against a secondary site does not work for pushes over 1.86 GB. GitLab issue #413109 tracks this bug.
  • Backups cannot be run on Geo secondary sites.
  • Git push with options over SSH against a secondary site does not work and terminates the connection. For more information, see issue 417186.
  • The Geo secondary site does not accelerate (serve) the clone request for the first stage of the pipeline in most cases. Later stages are not guaranteed to be served by the secondary site either, for example if the Git change is large, bandwidth is small, or pipeline stages are short. In general, it does serve the clone request for subsequent stages. Issue 446176 discusses the reasons for this and proposes an enhancement to increase the chance that Runner clone requests are served from the secondary site.
  • When a single Git repository receives pushes at a high-enough rate, the secondary site’s local copy can be perpetually out-of-date. This causes all Git fetches of that repository to be forwarded to the primary site. See GitLab issue #455870.
  • Proxying is implemented only in the GitLab application in the Puma service or Web service, so other services do not benefit from this behavior. You should use a separate URL to ensure requests are always sent to the primary. These services include:
  • With a unified URL, Let’s Encrypt can’t generate certificates unless it can reach both IPs through the same domain. To use TLS certificates with Let’s Encrypt, you can manually point the domain to one of the Geo sites, generate the certificate, then copy it to all other sites.
  • When a secondary site uses a separate URL from the primary site, signing in the secondary site using SAML is only supported if the SAML Identity Provider (IdP) allows an application to be configured with multiple callback URLs.
  • Git clone and fetch requests with option --depth over SSH against a secondary site does not work and hangs indefinitely if the secondary site is not up to date at the time the request is initiated. This is due to problems related to translating Git SSH to Git https during proxying. For more information, see issue 391980. A new workflow that does not involve the aforementioned translation step is now available for Linux-packaged GitLab Geo secondary sites which can be enabled with a feature flag. For more details, see comment in issue 454707. The fix for Cloud Native GitLab Geo secondary sites is tracked in issue 5641.
  • Some customers have reported that git fetch over SSH when the secondary site is out of date hangs and/or times out and fails. git clone requests over SSH are not impacted. For more information, see issue 454707. A fix available for Linux-packaged GitLab Geo secondary sites which can be enabled with a feature flag. For more details, see comment in issue 454707. The fix for Cloud Native GitLab Geo secondary sites is tracked in issue 5641.

Limitations on replication/verification

There is a complete list of all GitLab data types and existing support for replication and verification.

Setup instructions

For setup instructions, see Setting up Geo.

Post-installation documentation

After installing GitLab on the secondary sites and performing the initial configuration, see the following documentation for post-installation information.

Configuring Geo

For information on configuring Geo, see Geo configuration.

Upgrading Geo

For information on how to update your Geo sites to the latest GitLab version, see Upgrading the Geo sites.

Pausing and resuming replication

caution
Pausing and resuming of replication is only supported for Geo installations using a Linux package-managed database. External databases are not supported.

In some circumstances, like during upgrades or a planned failover, it is desirable to pause replication between the primary and secondary.

If you plan to allow user activity on your secondary sites during the upgrade, do not pause replication for a zero-downtime upgrade. While paused, the secondary site gets more and more out-of-date. One known effect is that more and more Git fetches get redirected or proxied to the primary site. There may be additional unknown effects.

Pausing and resuming replication is done through a command-line tool from a specific node in the secondary site. Depending on your database architecture, this will target either the postgresql or patroni service:

  • If you are using a single node for all services on your secondary site, you must run the commands on this single node.
  • If you have a standalone PostgreSQL node on your secondary site, you must run the commands on this standalone PostgreSQL node.
  • If your secondary site is using a Patroni cluster, you must run these commands on the secondary Patroni standby leader node.

If you aren’t using a single node for all services on your secondary site, ensure that the /etc/gitlab/gitlab.rb on your PostgreSQL or Patroni nodes contains the configuration line gitlab_rails['geo_node_name'] = 'node_name', where node_name is the same as the geo_node_name on the application node.

To Pause: (from secondary site)

Also, be aware that if PostgreSQL is restarted after pausing replication (either by restarting the VM or restarting the service with gitlab-ctl restart postgresql), PostgreSQL automatically resumes replication, which is something you wouldn’t want during an upgrade or in a planned failover scenario. 

gitlab-ctl geo-replication-pause

To Resume: (from secondary site) 

gitlab-ctl geo-replication-resume

Configuring Geo for multiple nodes

For information on configuring Geo for multiple nodes, see Geo for multiple servers.

Configuring Geo with Object Storage

For information on configuring Geo with Object storage, see Geo with Object storage.

Disaster Recovery

For information on using Geo in disaster recovery situations to mitigate data-loss and restore services, see Disaster Recovery.

Replicating the container registry

For more information on how to replicate the container registry, see Container registry for a secondary site.

Set up a unified URL for Geo sites

For an example of how to set up a single, location-aware URL with AWS Route53 or Google Cloud DNS, see Set up a unified URL for Geo sites.

Single Sign On (SSO)

For more information on configuring Single Sign-On (SSO), see Geo with Single Sign-On (SSO).

LDAP

For more information on configuring LDAP, see Geo with Single Sign-On (SSO) > LDAP.

Security Review

For more information on Geo security, see Geo security review.

Tuning Geo

For more information on tuning Geo, see Tuning Geo.

Backfill

When a secondary site is set up, it starts replicating missing data from the primary site in a process known as backfill. You can monitor the synchronization process on each Geo site from the primary site’s Geo Nodes dashboard in your browser.

Failures that happen during a backfill are scheduled to be retried at the end of the backfill.

Runners

Remove Geo site

For more information on removing a Geo site, see Removing secondary Geo sites.

Disable Geo

To find out how to disable Geo, see Disabling Geo.

Frequently Asked Questions

For answers to common questions, see the Geo FAQ.

Log files

Geo stores structured log messages in a geo.log file.

For more information on how to access and consume Geo logs, see the Geo section in the log system documentation.

Troubleshooting

For troubleshooting steps, see Geo Troubleshooting.