Skip to content
GitLab

Reference architecture: up to 2,000 users (FREE SELF)

This page describes GitLab reference architecture for up to 2,000 users. For a full list of reference architectures, see Available reference architectures.

  • Supported users (approximate): 2,000
  • High Availability: No. For a highly-available environment, you can follow a modified 3K reference architecture.
  • Estimated Costs: See cost table
  • Cloud Native Hybrid: Yes
  • Validation and test results: The Quality Engineering team does regular smoke and performance tests to ensure the reference architectures remain compliant
    • Test requests per second (RPS) rates: API: 40 RPS, Web: 4 RPS, Git (Pull): 4 RPS, Git (Push): 1 RPS
    • Latest Results
Service Nodes Configuration GCP AWS Azure
Load balancer3 1 2 vCPU, 1.8 GB memory n1-highcpu-2 c5.large F2s v2
PostgreSQL1 1 2 vCPU, 7.5 GB memory n1-standard-2 m5.large D2s v3
Redis2 1 1 vCPU, 3.75 GB memory n1-standard-1 m5.large D2s v3
Gitaly 1 4 vCPU, 15 GB memory n1-standard-4 m5.xlarge D4s v3
GitLab Rails 2 8 vCPU, 7.2 GB memory n1-highcpu-8 c5.2xlarge F8s v2
Monitoring node 1 2 vCPU, 1.8 GB memory n1-highcpu-2 c5.large F2s v2
Object storage4 - - - - -
NFS server (non-Gitaly) 1 4 vCPU, 3.6 GB memory n1-highcpu-4 c5.xlarge F4s v2
  1. Can be optionally run on reputable third-party external PaaS PostgreSQL solutions. Google Cloud SQL and Amazon RDS are known to work. However, Amazon Aurora is incompatible with load balancing enabled by default in 14.4.0, and Azure Database for PostgreSQL is not recommended due to performance issues. Consul is primarily used for PostgreSQL high availability so can be ignored when using a PostgreSQL PaaS setup. However, Consul is also used optionally by Prometheus for Omnibus auto host discovery.
  2. Can be optionally run as reputable third-party external PaaS Redis solutions. Google Memorystore and AWS ElastiCache are known to work.
  3. Can be optionally run as reputable third-party load balancing services (LB PaaS). AWS ELB is known to work.
  4. Should be run on reputable Cloud Provider or Self Managed solutions. More information can be found in the Configure the object storage section.

NOTE: For all PaaS solutions that involve configuring instances, it is strongly recommended to implement a minimum of three nodes in three different availability zones to align with resilient cloud architecture practices.

@startuml 2k
skinparam linetype ortho

card "**External Load Balancer**" as elb #6a9be7

collections "**GitLab Rails** x3" as gitlab #32CD32
card "**Prometheus + Grafana**" as monitor #7FFFD4
card "**Gitaly**" as gitaly #FF8C00
card "**PostgreSQL**" as postgres #4EA7FF
card "**Redis**" as redis #FF6347
cloud "**Object Storage**" as object_storage #white

elb -[#6a9be7]-> gitlab
elb -[#6a9be7]--> monitor

gitlab -[#32CD32]--> gitaly
gitlab -[#32CD32]--> postgres
gitlab -[#32CD32]-> object_storage
gitlab -[#32CD32]--> redis

monitor .[#7FFFD4]u-> gitlab
monitor .[#7FFFD4]-> gitaly
monitor .[#7FFFD4]-> postgres
monitor .[#7FFFD4,norank]--> redis
monitor .[#7FFFD4,norank]u--> elb

@enduml

Requirements

Before starting, you should take note of the following requirements / guidance for this reference architecture.

Supported CPUs

This reference architecture was built and tested on Google Cloud Platform (GCP) using the Intel Xeon E5 v3 (Haswell) CPU platform as a baseline (Sysbench benchmark).

Newer, similarly sized CPUs are supported and may have improved performance as a result. For Omnibus environments, ARM-based equivalents are also supported.

NOTE: Any "burstable" instance types are not recommended due to inconsistent performance.

Supported infrastructure

As a general guidance, GitLab should run on most infrastructure such as reputable Cloud Providers (AWS, GCP, Azure) and their services, or self managed (ESXi) that meet both the specs detailed above, as well as any requirements in this section. However, this does not constitute a guarantee for every potential permutation.

Be aware of the following specific call outs:

Setup components

To set up GitLab and its components to accommodate up to 2,000 users:

  1. Configure the external load balancing node to handle the load balancing of the GitLab application services nodes.
  2. Configure PostgreSQL, the database for GitLab.
  3. Configure Redis.
  4. Configure Gitaly, which provides access to the Git repositories.
  5. Configure the main GitLab Rails application to run Puma, Workhorse, GitLab Shell, and to serve all frontend requests (which include UI, API, and Git over HTTP/SSH).
  6. Configure Prometheus to monitor your GitLab environment.
  7. Configure the object storage used for shared data objects.
  8. Configure NFS (optional, and not recommended) to have shared disk storage service as an alternative to Gitaly or object storage.
  9. Configure Advanced Search (optional) for faster, more advanced code search across your entire GitLab instance.

Configure the external load balancer

In a multi-node GitLab configuration, you'll need a load balancer to route traffic to the application servers. The specifics on which load balancer to use or its exact configuration is beyond the scope of GitLab documentation. We assume that if you're managing multi-node systems like GitLab, you already have a load balancer of choice and that the routing methods used are distributing calls evenly between all nodes. Some load balancer examples include HAProxy (open-source), F5 Big-IP LTM, and Citrix Net Scaler. This documentation outline the ports and protocols needed for use with GitLab.

This architecture has been tested and validated with HAProxy as the load balancer. Although other load balancers with similar feature sets could also be used, those load balancers have not been validated.

The next question is how you will handle SSL in your environment. There are several different options:

Application node terminates SSL

Configure your load balancer to pass connections on port 443 as TCP instead of HTTP(S). This will pass the connection unaltered to the application node's NGINX service, which has the SSL certificate and listens to port 443.

For details about managing SSL certificates and configuring NGINX, see the HTTPS documentation

Load balancer terminates SSL without backend SSL

Configure your load balancer to use the HTTP(S) protocol instead of TCP. The load balancer will be responsible for both managing SSL certificates and terminating SSL.

Due to communication between the load balancer and GitLab not being secure, you'll need to complete some additional configuration. For details, see the proxied SSL documentation.

Load balancer terminates SSL with backend SSL

Configure your load balancers (or single balancer, if you have only one) to use the HTTP(S) protocol rather than TCP. The load balancers will be responsible for the managing SSL certificates for end users.

Traffic will be secure between the load balancers and NGINX in this scenario, and there's no need to add a configuration for proxied SSL. However, you'll need to add a configuration to GitLab to configure SSL certificates. For details about managing SSL certificates and configuring NGINX, see the HTTPS documentation.

Readiness checks

Ensure the external load balancer only routes to working services with built in monitoring endpoints. The readiness checks all require additional configuration on the nodes being checked, otherwise, the external load balancer will not be able to connect.

Ports

The basic load balancer ports you should use are described in the following table:

Port Backend Port Protocol
80 80 HTTP (1)
443 443 TCP or HTTPS (1) (2)
22 22 TCP
  • (1): 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. For details, see the web terminal integration guide.
  • (2): When using the HTTPS protocol for port 443, you'll need to add an SSL certificate to the load balancers. If you need to terminate SSL at the GitLab application server, use the TCP protocol.

If you're using GitLab Pages with custom domain support you will need some additional port configurations. GitLab Pages requires a separate virtual IP address. Configure DNS to point the pages_external_url from /etc/gitlab/gitlab.rb to the new virtual IP address. For more information, see the GitLab Pages documentation.

Port Backend Port Protocol
80 Varies (1) HTTP
443 Varies (1) TCP (2)
  • (1): The backend port for GitLab Pages depends on the gitlab_pages['external_http'] and gitlab_pages['external_https'] settings. For details, see the GitLab Pages documentation.
  • (2): Port 443 for GitLab Pages must use the TCP protocol. Users can configure custom domains with custom SSL, which wouldn't be possible if SSL was terminated at the load balancer.

Alternate SSH Port

Some organizations have policies against opening SSH port 22. In this case, it may be helpful to configure an alternate SSH hostname that instead allows users to use SSH over port 443. An alternate SSH hostname requires a new virtual IP address compared to the previously described GitLab HTTP configuration.

Configure DNS for an alternate SSH hostname, such as altssh.gitlab.example.com:

LB Port Backend Port Protocol
443 22 TCP

Configure PostgreSQL

In this section, you'll be guided through configuring an external PostgreSQL database to be used with GitLab.

Provide your own PostgreSQL instance

If you're hosting GitLab on a cloud provider, you can optionally use a managed service for PostgreSQL.

A reputable provider or solution should be used for this. Google Cloud SQL and Amazon RDS are known to work. However, Amazon Aurora is incompatible with load balancing enabled by default in 14.4.0, and Azure Database for PostgreSQL is not recommended due to performance issues.

If you use a cloud-managed service, or provide your own PostgreSQL:

  1. Set up PostgreSQL according to the database requirements document.
  2. Create a gitlab username with a password of your choice. The gitlab user needs privileges to create the gitlabhq_production database.
  3. Configure the GitLab application servers with the appropriate details. This step is covered in Configuring the GitLab Rails application.

See Configure GitLab using an external PostgreSQL service for further configuration steps.

Standalone PostgreSQL using Omnibus GitLab

  1. SSH in to the PostgreSQL server.

  2. Download and install the Omnibus GitLab package of your choice. Be sure to follow only installation steps 1 and 2 on the page.

  3. Generate a password hash for PostgreSQL. This assumes you will use the default username of gitlab (recommended). The command will request a password and confirmation. Use the value that is output by this command in the next step as the value of POSTGRESQL_PASSWORD_HASH.

    sudo gitlab-ctl pg-password-md5 gitlab

  4. Edit /etc/gitlab/gitlab.rb and add the contents below, updating placeholder values appropriately.

    • POSTGRESQL_PASSWORD_HASH - The value output from the previous step
    • APPLICATION_SERVER_IP_BLOCKS - A space delimited list of IP subnets or IP addresses of the GitLab application servers that will connect to the database. Example: %w(123.123.123.123/32 123.123.123.234/32) 
    # Disable all components except PostgreSQL related ones
roles(['postgres_role'])

# Set the network addresses that the exporters used for monitoring will listen on
node_exporter['listen_address'] = '0.0.0.0:9100'
postgres_exporter['listen_address'] = '0.0.0.0:9187'
postgres_exporter['dbname'] = 'gitlabhq_production'
postgres_exporter['password'] = 'POSTGRESQL_PASSWORD_HASH'

# Set the PostgreSQL address and port
postgresql['listen_address'] = '0.0.0.0'
postgresql['port'] = 5432

# Replace POSTGRESQL_PASSWORD_HASH with a generated md5 value
postgresql['sql_user_password'] = 'POSTGRESQL_PASSWORD_HASH'

# Replace APPLICATION_SERVER_IP_BLOCK with the CIDR address of the application node
postgresql['trust_auth_cidr_addresses'] = %w(127.0.0.1/32 APPLICATION_SERVER_IP_BLOCK)

# Prevent database migrations from running on upgrade automatically
gitlab_rails['auto_migrate'] = false

  5. Copy the /etc/gitlab/gitlab-secrets.json file from the first Omnibus node you configured and add or replace the file of the same name on this server. If this is the first Omnibus node you are configuring then you can skip this step.

  6. Reconfigure GitLab for the changes to take effect.

  7. Note the PostgreSQL node's IP address or hostname, port, and plain text password. These will be necessary when configuring the GitLab application server later.

Advanced configuration options are supported and can be added if needed.

Configure Redis

In this section, you'll be guided through configuring an external Redis instance to be used with GitLab.

Provide your own Redis instance

Redis version 5.0 or higher is required, as this is what ships with Omnibus GitLab packages starting with GitLab 13.0. Older Redis versions do not support an optional count argument to SPOP which is now required for Merge Trains.

In addition, GitLab makes use of certain commands like UNLINK and USAGE which were introduced only in Redis 4.

Managed Redis from cloud providers such as AWS ElastiCache will work. If these services support high availability, be sure it is not the Redis Cluster type.

Note the Redis node's IP address or hostname, port, and password (if required). These will be necessary when configuring the GitLab application servers later.

Standalone Redis using Omnibus GitLab

The Omnibus GitLab package can be used to configure a standalone Redis server. The steps below are the minimum necessary to configure a Redis server with Omnibus:

  1. SSH in to the Redis server.

  2. Download and install the Omnibus GitLab package of your choice. Be sure to follow only installation steps 1 and 2 on the page.

  3. Edit /etc/gitlab/gitlab.rb and add the contents:

    ## Enable Redis
roles(["redis_master_role"])

redis['bind'] = '0.0.0.0'
redis['port'] = 6379
redis['password'] = 'SECRET_PASSWORD_HERE'

gitlab_rails['enable'] = false

# Set the network addresses that the exporters used for monitoring will listen on
node_exporter['listen_address'] = '0.0.0.0:9100'
redis_exporter['listen_address'] = '0.0.0.0:9121'
redis_exporter['flags'] = {
      'redis.addr' => 'redis://0.0.0.0:6379',
      'redis.password' => 'SECRET_PASSWORD_HERE',
}

  4. Copy the /etc/gitlab/gitlab-secrets.json file from the first Omnibus node you configured and add or replace the file of the same name on this server. If this is the first Omnibus node you are configuring then you can skip this step.

  5. Reconfigure Omnibus GitLab for the changes to take effect.

  6. Note the Redis node's IP address or hostname, port, and Redis password. These will be necessary when configuring the GitLab application servers later.

Advanced configuration options are supported and can be added if needed.

Configure Gitaly

Gitaly server node requirements are dependent on data size, specifically the number of projects and those projects' sizes.

NOTE: The Reference Architecture specs have been designed with good headroom in mind but for Gitaly, increased specs or switching to Gitaly Cluster may be required for notably large data sets or load.

Due to Gitaly having notable input and output requirements, we strongly recommend that all Gitaly nodes use solid-state drives (SSDs). These SSDs should have a throughput of at least 8,000 input/output operations per second (IOPS) for read operations and 2,000 IOPS for write operations. These IOPS values are initial recommendations, and may be adjusted to greater or lesser values depending on the scale of your environment's workload. If you're running the environment on a Cloud provider, refer to their documentation about how to configure IOPS correctly.

Be sure to note the following items:

  • The GitLab Rails application shards repositories into repository storage paths.
  • A Gitaly server can host one or more storage paths.
  • A GitLab server can use one or more Gitaly server nodes.
  • Gitaly addresses must be specified to be correctly resolvable for all Gitaly clients.
  • Gitaly servers must not be exposed to the public internet, as Gitaly's network traffic is unencrypted by default. The use of a firewall is highly recommended to restrict access to the Gitaly server. Another option is to use TLS.

NOTE: The token referred to throughout the Gitaly documentation is an arbitrary password selected by the administrator. This token is unrelated to tokens created for the GitLab API or other similar web API tokens.

The following procedure describes how to configure a single Gitaly server named gitaly1.internal with the secret token gitalysecret. We assume your GitLab installation has two repository storages: default and storage1.

To configure the Gitaly server, on the server node you want to use for Gitaly:

  1. Download and install the Omnibus GitLab package of your choice. Be sure to follow only installation steps 1 and 2 on the page, and do not provide the EXTERNAL_URL value.

  2. Edit the Gitaly server node's /etc/gitlab/gitlab.rb file to configure storage paths, enable the network listener, and to configure the token:

    NOTE: You can't remove the default entry from git_data_dirs because GitLab requires it.

# Avoid running unnecessary services on the Gitaly server
postgresql['enable'] = false
redis['enable'] = false
nginx['enable'] = false
puma['enable'] = false
sidekiq['enable'] = false
gitlab_workhorse['enable'] = false
prometheus['enable'] = false
alertmanager['enable'] = false
grafana['enable'] = false
gitlab_exporter['enable'] = false
gitlab_kas['enable'] = false

# Prevent database migrations from running on upgrade automatically
gitlab_rails['auto_migrate'] = false

# Configure the gitlab-shell API callback URL. Without this, `git push` will
# fail. This can be your 'front door' GitLab URL or an internal load
# balancer.
gitlab_rails['internal_api_url'] = 'https://gitlab.example.com'

# Gitaly
gitaly['enable'] = true

# Make Gitaly accept connections on all network interfaces. You must use
# firewalls to restrict access to this address/port.
# Comment out following line if you only want to support TLS connections
gitaly['listen_addr'] = "0.0.0.0:8075"
gitaly['prometheus_listen_addr'] = "0.0.0.0:9236"

# Gitaly and GitLab use two shared secrets for authentication, one to authenticate gRPC requests
# to Gitaly, and a second for authentication callbacks from GitLab-Shell to the GitLab internal API.
# The following two values must be the same as their respective values
# of the GitLab Rails application setup
gitaly['auth_token'] = 'gitalysecret'
gitlab_shell['secret_token'] = 'shellsecret'

# Set the network addresses that the exporters used for monitoring will listen on
node_exporter['listen_address'] = '0.0.0.0:9100'

git_data_dirs({
  'default' => {
    'path' => '/var/opt/gitlab/git-data'
  },
  'storage1' => {
    'path' => '/mnt/gitlab/git-data'
  },
})

  1. Copy the /etc/gitlab/gitlab-secrets.json file from the first Omnibus node you configured and add or replace the file of the same name on this server. If this is the first Omnibus node you are configuring then you can skip this step.

  2. Reconfigure GitLab for the changes to take effect.

  3. Confirm that Gitaly can perform callbacks to the internal API:

    • For GitLab 15.3 and later, run sudo /opt/gitlab/embedded/bin/gitaly check /var/opt/gitlab/gitaly/config.toml.
    • For GitLab 15.2 and earlier, run sudo /opt/gitlab/embedded/bin/gitaly-hooks check /var/opt/gitlab/gitaly/config.toml.

Gitaly TLS support

Gitaly supports TLS encryption. To be able to communicate with a Gitaly instance that listens for secure connections you will need to use tls:// URL scheme in the gitaly_address of the corresponding storage entry in the GitLab configuration.

You will need to bring your own certificates as this isn't provided automatically. The certificate, or its certificate authority, must be installed on all Gitaly nodes (including the Gitaly node using the certificate) and on all client nodes that communicate with it following the procedure described in GitLab custom certificate configuration.

NOTE: The self-signed certificate must specify the address you use to access the Gitaly server. If you are addressing the Gitaly server by a hostname, add it as a Subject Alternative Name. If you are addressing the Gitaly server by its IP address, you must add it as a Subject Alternative Name to the certificate.

It's possible to configure Gitaly servers with both an unencrypted listening address (listen_addr) and an encrypted listening address (tls_listen_addr) at the same time. This allows you to do a gradual transition from unencrypted to encrypted traffic, if necessary.

To configure Gitaly with TLS:

  1. Create the /etc/gitlab/ssl directory and copy your key and certificate there:

    sudo mkdir -p /etc/gitlab/ssl
sudo chmod 755 /etc/gitlab/ssl
sudo cp key.pem cert.pem /etc/gitlab/ssl/
sudo chmod 644 key.pem cert.pem

  2. Copy the cert to /etc/gitlab/trusted-certs so Gitaly will trust the cert when calling into itself:

    sudo cp /etc/gitlab/ssl/cert.pem /etc/gitlab/trusted-certs/

  3. Edit /etc/gitlab/gitlab.rb and add:

    gitaly['tls_listen_addr'] = "0.0.0.0:9999"
gitaly['certificate_path'] = "/etc/gitlab/ssl/cert.pem"
gitaly['key_path'] = "/etc/gitlab/ssl/key.pem"

  4. Delete gitaly['listen_addr'] to allow only encrypted connections.

  5. Save the file and reconfigure GitLab.

Configure GitLab Rails

This section describes how to configure the GitLab application (Rails) component.

In our architecture, we run each GitLab Rails node using the Puma webserver, and have its number of workers set to 90% of available CPUs, with four threads. For nodes running Rails with other components, the worker value should be reduced accordingly. We've determined that a worker value of 50% achieves a good balance, but this is dependent on workload.

On each node perform the following:

  1. If you're using NFS:

    1. If necessary, install the NFS client utility packages using the following commands:

      # Ubuntu/Debian
apt-get install nfs-common

# CentOS/Red Hat
yum install nfs-utils nfs-utils-lib

    2. Specify the necessary NFS mounts in /etc/fstab. The exact contents of /etc/fstab will depend on how you chose to configure your NFS server. See the NFS documentation for examples and the various options.

    3. Create the shared directories. These may be different depending on your NFS mount locations.

      mkdir -p /var/opt/gitlab/.ssh /var/opt/gitlab/gitlab-rails/uploads /var/opt/gitlab/gitlab-rails/shared /var/opt/gitlab/gitlab-ci/builds /var/opt/gitlab/git-data

  2. Download and install the Omnibus GitLab package of your choice. Be sure to follow only installation steps 1 and 2 on the page.

  3. Create or edit /etc/gitlab/gitlab.rb and use the following configuration. To maintain uniformity of links across nodes, the external_url on the application server should point to the external URL that users will use to access GitLab. This would be the URL of the load balancer which will route traffic to the GitLab application server:

    external_url 'https://gitlab.example.com'

# Gitaly and GitLab use two shared secrets for authentication, one to authenticate gRPC requests
# to Gitaly, and a second for authentication callbacks from GitLab-Shell to the GitLab internal API.
# The following two values must be the same as their respective values
# of the Gitaly setup
gitlab_rails['gitaly_token'] = 'gitalysecret'
gitlab_shell['secret_token'] = 'shellsecret'

git_data_dirs({
  'default' => { 'gitaly_address' => 'tcp://gitaly1.internal:8075' },
  'storage1' => { 'gitaly_address' => 'tcp://gitaly1.internal:8075' },
  'storage2' => { 'gitaly_address' => 'tcp://gitaly2.internal:8075' },
})

## Disable components that will not be on the GitLab application server
roles(['application_role'])
gitaly['enable'] = false
nginx['enable'] = true

## PostgreSQL connection details
gitlab_rails['db_adapter'] = 'postgresql'
gitlab_rails['db_encoding'] = 'unicode'
gitlab_rails['db_host'] = '10.1.0.5' # IP/hostname of database server
gitlab_rails['db_password'] = 'DB password'

## Redis connection details
gitlab_rails['redis_port'] = '6379'
gitlab_rails['redis_host'] = '10.1.0.6' # IP/hostname of Redis server
gitlab_rails['redis_password'] = 'Redis Password'

# Set the network addresses that the exporters used for monitoring will listen on
node_exporter['listen_address'] = '0.0.0.0:9100'
gitlab_workhorse['prometheus_listen_addr'] = '0.0.0.0:9229'
puma['listen'] = '0.0.0.0'
sidekiq['listen_address'] = "0.0.0.0"

# Configure Sidekiq with 2 workers and 10 max concurrency
sidekiq['max_concurrency'] = 10
sidekiq['queue_groups'] = ['*'] * 2

# Add the monitoring node's IP address to the monitoring whitelist and allow it to
# scrape the NGINX metrics. Replace placeholder `monitoring.gitlab.example.com` with
# the address and/or subnets gathered from the monitoring node
gitlab_rails['monitoring_whitelist'] = ['<MONITOR NODE IP>/32', '127.0.0.0/8']
nginx['status']['options']['allow'] = ['<MONITOR NODE IP>/32', '127.0.0.0/8']

# Object Storage
# This is an example for configuring Object Storage on GCP
# Replace this config with your chosen Object Storage provider as desired
gitlab_rails['object_store']['enabled'] = true
gitlab_rails['object_store']['connection'] = {
  'provider' => 'Google',
  'google_project' => '<gcp-project-name>',
  'google_json_key_location' => '<path-to-gcp-service-account-key>'
}
gitlab_rails['object_store']['objects']['artifacts']['bucket'] = "<gcp-artifacts-bucket-name>"
gitlab_rails['object_store']['objects']['external_diffs']['bucket'] = "<gcp-external-diffs-bucket-name>"
gitlab_rails['object_store']['objects']['lfs']['bucket'] = "<gcp-lfs-bucket-name>"
gitlab_rails['object_store']['objects']['uploads']['bucket'] = "<gcp-uploads-bucket-name>"
gitlab_rails['object_store']['objects']['packages']['bucket'] = "<gcp-packages-bucket-name>"
gitlab_rails['object_store']['objects']['dependency_proxy']['bucket'] = "<gcp-dependency-proxy-bucket-name>"
gitlab_rails['object_store']['objects']['terraform_state']['bucket'] = "<gcp-terraform-state-bucket-name>"

gitlab_rails['backup_upload_connection'] = {
  'provider' => 'Google',
  'google_project' => '<gcp-project-name>',
  'google_json_key_location' => '<path-to-gcp-service-account-key>'
}
gitlab_rails['backup_upload_remote_directory'] = "<gcp-backups-state-bucket-name>"

## Uncomment and edit the following options if you have set up NFS
##
## Prevent GitLab from starting if NFS data mounts are not available
##
#high_availability['mountpoint'] = '/var/opt/gitlab/git-data'
##
## Ensure UIDs and GIDs match between servers for permissions via NFS
##
#user['uid'] = 9000
#user['gid'] = 9000
#web_server['uid'] = 9001
#web_server['gid'] = 9001
#registry['uid'] = 9002
#registry['gid'] = 9002

  4. If you're using Gitaly with TLS support, make sure the git_data_dirs entry is configured with tls instead of tcp:

    git_data_dirs({
  'default' => { 'gitaly_address' => 'tls://gitaly1.internal:9999' },
  'storage1' => { 'gitaly_address' => 'tls://gitaly1.internal:9999' },
  'storage2' => { 'gitaly_address' => 'tls://gitaly2.internal:9999' },
})

    1. Copy the cert into /etc/gitlab/trusted-certs:

      sudo cp cert.pem /etc/gitlab/trusted-certs/

  5. Copy the /etc/gitlab/gitlab-secrets.json file from the first Omnibus node you configured and add or replace the file of the same name on this server. If this is the first Omnibus node you are configuring then you can skip this step.

  6. To ensure database migrations are only run during reconfigure and not automatically on upgrade, run:

    sudo touch /etc/gitlab/skip-auto-reconfigure

    Only a single designated node should handle migrations as detailed in the GitLab Rails post-configuration section.

  7. Reconfigure GitLab for the changes to take effect.

  8. Run sudo gitlab-rake gitlab:gitaly:check to confirm the node can connect to Gitaly.

  9. Tail the logs to see the requests:

    sudo gitlab-ctl tail gitaly

When you specify https in the external_url, as in the previous example, GitLab expects that the SSL certificates are in /etc/gitlab/ssl/. If the certificates aren't present, NGINX will fail to start. For more information, see the HTTPS documentation.

GitLab Rails post-configuration

  1. Designate one application node for running database migrations during installation and updates. Initialize the GitLab database and ensure all migrations ran:

    sudo gitlab-rake gitlab:db:configure

    If you encounter a rake aborted! error message stating that PgBouncer is failing to connect to PostgreSQL, it may be that your PgBouncer node's IP address is missing from PostgreSQL's trust_auth_cidr_addresses in gitlab.rb on your database nodes. Before proceeding, see PgBouncer error ERROR: pgbouncer cannot connect to server.

  2. Configure fast lookup of authorized SSH keys in the database.

Configure Prometheus

The Omnibus GitLab package can be used to configure a standalone Monitoring node running Prometheus and Grafana:

  1. SSH in to the Monitoring node.

  2. Download and install the Omnibus GitLab package of your choice. Be sure to follow only installation steps 1 and 2 on the page.

  3. Edit /etc/gitlab/gitlab.rb and add the contents:

    roles(['monitoring_role'])

external_url 'http://gitlab.example.com'

# Prometheus
prometheus['listen_address'] = '0.0.0.0:9090'
prometheus['monitor_kubernetes'] = false

# Grafana
grafana['enable'] = true
grafana['admin_password'] = '<grafana_password>'
grafana['disable_login_form'] = false

# Nginx - For Grafana access
nginx['enable'] = true

  4. Prometheus also needs some scrape configurations to pull all the data from the various nodes where we configured exporters. Assuming that your nodes' IPs are:

    1.1.1.1: postgres
1.1.1.2: redis
1.1.1.3: gitaly1
1.1.1.4: rails1
1.1.1.5: rails2

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

    prometheus['scrape_configs'] = [
  {
     'job_name': 'postgres',
     'static_configs' => [
     'targets' => ['1.1.1.1:9187'],
     ],
  },
  {
     'job_name': 'redis',
     'static_configs' => [
     'targets' => ['1.1.1.2:9121'],
     ],
  },
  {
     'job_name': 'gitaly',
     'static_configs' => [
     'targets' => ['1.1.1.3:9236'],
     ],
  },
  {
     'job_name': 'gitlab-nginx',
     'static_configs' => [
     'targets' => ['1.1.1.4:8060', '1.1.1.5:8060'],
     ],
  },
  {
     'job_name': 'gitlab-workhorse',
     'static_configs' => [
     'targets' => ['1.1.1.4:9229', '1.1.1.5:9229'],
     ],
  },
  {
     'job_name': 'gitlab-rails',
     'metrics_path': '/-/metrics',
     'static_configs' => [
     'targets' => ['1.1.1.4:8080', '1.1.1.5:8080'],
     ],
  },
  {
     'job_name': 'gitlab-sidekiq',
     'static_configs' => [
     'targets' => ['1.1.1.4:8082', '1.1.1.5:8082'],
     ],
  },
  {
     'job_name': 'node',
     'static_configs' => [
     'targets' => ['1.1.1.1:9100', '1.1.1.2:9100', '1.1.1.3:9100', '1.1.1.4:9100', '1.1.1.5:9100'],
     ],
  },
]

  5. Save the file and reconfigure GitLab.

  6. In the GitLab UI, set admin/application_settings/metrics_and_profiling > Metrics - Grafana to /-/grafana to http[s]://<MONITOR NODE>/-/grafana

Configure the object storage

GitLab supports using an object storage service for holding numerous types of data. It's recommended over NFS and in general it's better in larger setups as object storage is typically much more performant, reliable, and scalable.

GitLab has been tested on a number of object storage providers:

There are two ways of specifying object storage configuration in GitLab:

The consolidated form is used in the following examples when available.

NOTE: When using the storage-specific form in GitLab 14.x and earlier, you should enable direct upload mode. The previous background upload mode, which was deprecated in 14.9, requires shared storage such as NFS.

Using separate buckets for each data type is the recommended approach for GitLab. This ensures there are no collisions across the various types of data GitLab stores. There are plans to enable the use of a single bucket in the future.

Enable incremental logging

GitLab Runner returns job logs in chunks which Omnibus GitLab caches temporarily on disk in /var/opt/gitlab/gitlab-ci/builds by default, even when using consolidated object storage. With default configuration, this directory needs to be shared through NFS on any GitLab Rails and Sidekiq nodes.

While sharing the job logs through NFS is supported, it's recommended to avoid the need to use NFS by enabling incremental logging (required when no NFS node has been deployed). Incremental logging uses Redis instead of disk space for temporary caching of job logs.

Configure NFS (optional)

For improved performance, object storage, along with Gitaly, are recommended over using NFS whenever possible.

See how to configure NFS.

WARNING: Engineering support for NFS for Git repositories is deprecated, and technical support is scheduled to be unavailable after the release of GitLab 15.6. No further enhancements are planned for this feature.

Read:

Configure Advanced Search (PREMIUM SELF)

You can leverage Elasticsearch and enable Advanced Search for faster, more advanced code search across your entire GitLab instance.

Elasticsearch cluster design and requirements are dependent on your specific data. For recommended best practices about how to set up your Elasticsearch cluster alongside your instance, read how to choose the optimal cluster configuration.

Cloud Native Hybrid reference architecture with Helm Charts (alternative)

As an alternative approach, you can also run select components of GitLab as Cloud Native in Kubernetes via our official Helm Charts. In this setup, we support running the equivalent of GitLab Rails and Sidekiq nodes in a Kubernetes cluster, named Webservice and Sidekiq respectively. In addition, the following other supporting services are supported: NGINX, Task Runner, Migrations, Prometheus, and Grafana.

Hybrid installations leverage the benefits of both cloud native and traditional compute deployments. With this, stateless components can benefit from cloud native workload management benefits while stateful components are deployed in compute VMs with Omnibus to benefit from increased permanence.

The 2,000 reference architecture is not a highly-available setup. To achieve HA, you can follow a modified 3K reference architecture.

NOTE: This is an advanced setup. Running services in Kubernetes is well known to be complex. This setup is only recommended if you have strong working knowledge and experience in Kubernetes. The rest of this section assumes this.

NOTE: Gitaly Cluster is not supported to be run in Kubernetes. Refer to epic 6127 for more details.

Cluster topology

The following tables and diagram detail the hybrid environment using the same formats as the normal environment above.

First are the components that run in Kubernetes. The recommendation at this time is to use Google Cloud's Kubernetes Engine (GKE) or AWS Elastic Kubernetes Service (EKS) and associated machine types, but the memory and CPU requirements should translate to most other providers. We hope to update this in the future with further specific cloud provider details.

Service Nodes Configuration GCP AWS Min Allocatable CPUs and Memory
Webservice 3 8 vCPU, 7.2 GB memory n1-highcpu-8 c5.2xlarge 23.7 vCPU, 16.9 GB memory
Sidekiq 2 4 vCPU, 15 GB memory n1-standard-4 m5.xlarge 7.8 vCPU, 25.9 GB memory
Supporting services 2 2 vCPU, 7.5 GB memory n1-standard-2 m5.large 1.9 vCPU, 5.5 GB memory
  • For this setup, we recommend and regularly test Google Kubernetes Engine (GKE) and Amazon Elastic Kubernetes Service (EKS). Other Kubernetes services may also work, but your mileage may vary.
  • Nodes configuration is shown as it is forced to ensure pod vCPU / memory ratios and avoid scaling during performance testing.
    • In production deployments, there is no need to assign pods to nodes. A minimum of three nodes in three different availability zones is strongly recommended to align with resilient cloud architecture practices.

Next are the backend components that run on static compute VMs via Omnibus (or External PaaS services where applicable):

Service Nodes Configuration GCP AWS
PostgreSQL1 1 2 vCPU, 7.5 GB memory n1-standard-2 m5.large
Redis2 1 1 vCPU, 3.75 GB memory n1-standard-1 m5.large
Gitaly 1 4 vCPU, 15 GB memory n1-standard-4 m5.xlarge
Object storage3 - - - -
  1. Can be optionally run on reputable third-party external PaaS PostgreSQL solutions. Google Cloud SQL and Amazon RDS are known to work. However, Amazon Aurora is incompatible with load balancing enabled by default in 14.4.0, and Azure Database for PostgreSQL is not recommended due to performance issues. Consul is primarily used for PostgreSQL high availability so can be ignored when using a PostgreSQL PaaS setup. However, Consul is also used optionally by Prometheus for Omnibus auto host discovery.
  2. Can be optionally run on reputable third-party external PaaS Redis solutions. Google Memorystore and AWS ElastiCache are known to work.
  3. Should be run on reputable Cloud Provider or Self Managed solutions. More information can be found in the Configure the object storage section.

NOTE: For all PaaS solutions that involve configuring instances, it is strongly recommended to implement a minimum of three nodes in three different availability zones to align with resilient cloud architecture practices.

@startuml 2k
skinparam linetype ortho

card "Kubernetes via Helm Charts" as kubernetes {
  card "**External Load Balancer**" as elb #6a9be7

  together {
    collections "**Webservice** x3" as gitlab #32CD32
    collections "**Sidekiq** x2" as sidekiq #ff8dd1
  }

  collections "**Supporting Services** x2" as support
}

card "**Gitaly**" as gitaly #FF8C00
card "**PostgreSQL**" as postgres #4EA7FF
card "**Redis**" as redis #FF6347
cloud "**Object Storage**" as object_storage #white

elb -[#6a9be7]-> gitlab

gitlab -[#32CD32]--> gitaly
gitlab -[#32CD32]--> postgres
gitlab -[#32CD32]-> object_storage
gitlab -[#32CD32]--> redis

sidekiq -[#ff8dd1]--> gitaly
sidekiq -[#ff8dd1]-> object_storage
sidekiq -[#ff8dd1]--> postgres
sidekiq -[#ff8dd1]--> redis

@enduml

Resource usage settings

The following formulas help when calculating how many pods may be deployed within resource constraints. The 2k reference architecture example values file documents how to apply the calculated configuration to the Helm Chart.

Webservice

Webservice pods typically need about 1 CPU and 1.25 GB of memory per worker. Each Webservice pod consumes roughly 4 CPUs and 5 GB of memory using the recommended topology because two worker processes are created by default and each pod has other small processes running.

For 2,000 users we recommend a total Puma worker count of around 12. With the provided recommendations this allows the deployment of up to 3 Webservice pods with 4 workers per pod and 1 pod per node. Expand available resources using the ratio of 1 CPU to 1.25 GB of memory per each worker process for each additional Webservice pod.

For further information on resource usage, see the Webservice resources.

Sidekiq

Sidekiq pods should generally have 0.9 CPU and 2 GB of memory.

The provided starting point allows the deployment of up to 4 Sidekiq pods. Expand available resources using the 0.9 CPU to 2 GB memory ratio for each additional pod.

For further information on resource usage, see the Sidekiq resources.

Supporting

The Supporting Node Pool is designed to house all supporting deployments that don't need to be on the Webservice and Sidekiq pools.

This includes various deployments related to the Cloud Provider's implementation and supporting GitLab deployments such as NGINX or GitLab Shell.

If you wish to make any additional deployments, such as for Monitoring, it's recommended to deploy these in this pool where possible and not in the Webservice or Sidekiq pools, as the Supporting pool has been designed specifically to accommodate several additional deployments. However, if your deployments don't fit into the pool as given, you can increase the node pool accordingly.