In this page you will find information about the settings that are used on GitLab.com.
SSH host keys fingerprints
Below are the fingerprints for GitLab.com's SSH host keys.
Add the following to
.ssh/known_hosts to skip manual fingerprint
confirmation in SSH:
gitlab.com ssh-ed25519 AAAAC3NzaC1lZDI1NTE5AAAAIAfuCHKVTjquxvt6CM6tdG4SLp1Btn/nOeHHE5UOzRdf gitlab.com ssh-rsa AAAAB3NzaC1yc2EAAAADAQABAAABAQCsj2bNKTBSpIYDEGk9KxsGh3mySTRgMtXL583qmBpzeQ+jqCMRgBqB98u3z++J1sKlXHWfM9dyhSevkMwSbhoR8XIq/U0tCNyokEi/ueaBMCvbcTHhO7FcwzY92WK4Yt0aGROY5qX2UKSeOvuP4D6TPqKF1onrSzH9bx9XUf2lEdWT/ia1NEKjunUqu1xOB/StKDHMoX4/OKyIzuS0q/T1zOATthvasJFoPrAjkohTyaDUz2LN5JoH839hViyEG82yB+MjcFV5MU3N1l1QL3cVUCh93xSaua1N85qivl+siMkPGbO5xR/En4iEY6K2XPASUEMaieWVNTRCtJ4S8H+9 gitlab.com ecdsa-sha2-nistp256 AAAAE2VjZHNhLXNoYTItbmlzdHAyNTYAAAAIbmlzdHAyNTYAAABBBFSMqzJeV9rUzU4kWitGjeR4PWSa29SPqJ1fVkhtj3Hw9xjLVXVYrU9QlYWrOLXBpQ6KWjbjTDTdDkoohFzgbEY=
GitLab.com sends emails from the
mg.gitlab.com domain via Mailgun and has
its own dedicated IP address (
Alternative SSH port
GitLab.com can be reached via a different SSH port for
~/.ssh/config is the following:
Host gitlab.com Hostname altssh.gitlab.com User git Port 443 PreferredAuthentications publickey IdentityFile ~/.ssh/gitlab
Below are the settings for GitLab Pages.
|Custom domains support||yes||no|
|TLS certificates support||yes||no|
|Maximum size (uncompressed)||1G||100M|
NOTE: Note: The maximum size of your Pages site is regulated by the artifacts maximum size which is part of GitLab CI/CD.
Below are the current settings regarding GitLab CI/CD.
|Artifacts maximum size (uncompressed)||1G||100M|
|Artifacts expiry time||kept forever||deleted after 30 days unless otherwise specified|
|Scheduled Pipeline Cron||
Repository size limit
The maximum size your Git repository is allowed to be, including LFS. If you are near or over the size limit, you can reduce your repository size with Git.
|Repository size including LFS||10G||Unlimited|
GitLab.com, CI/CD, and related services are deployed into Google Cloud Platform (GCP). Any IP based firewall can be configured by looking up all IP address ranges or CIDR blocks for GCP.
Static endpoints are being considered.
Shared Runners on GitLab.com run in autoscale mode and powered by Google Cloud Platform. Autoscaling means reduced waiting times to spin up CI/CD jobs, and isolated VMs for each project, thus maximizing security. They're free to use for public open source projects and limited to 2000 CI minutes per month per group for private projects. More minutes can be purchased, if needed. Read about all GitLab.com plans.
All your CI/CD jobs run on n1-standard-1 instances with 3.75GB of RAM, CoreOS and the latest Docker Engine installed. Instances provide 1 vCPU and 25GB of HDD disk space. The default region of the VMs is US East1. Each instance is used only for one job, this ensures any sensitive data left on the system can't be accessed by other people their CI jobs.
gitlab-shared-runners-manager-X.gitlab.com fleet of Runners are dedicated for GitLab projects as well as community forks of them. They use a slightly larger machine type (n1-standard-2) and have a bigger SSD disk size. They will not run untagged jobs and unlike the general fleet of shared Runners, the instances are re-used up to 40 times.
Jobs handled by the shared Runners on GitLab.com (
will be timed out after 3 hours, regardless of the timeout configured in a
project. Check the issues 4010 and 4070 for the reference.
Below are the shared Runners settings.
|GitLab Runner||Runner versions dashboard||-|
|Default Docker image||
The full contents of our
Google Cloud Platform
concurrent = X check_interval = 1 metrics_server = "X" sentry_dsn = "X" [[runners]] name = "docker-auto-scale" request_concurrency = X url = "https://gitlab.com/" token = "SHARED_RUNNER_TOKEN" executor = "docker+machine" environment = [ "DOCKER_DRIVER=overlay2", "DOCKER_TLS_CERTDIR=" ] limit = X [runners.docker] image = "ruby:2.5" privileged = true volumes = [ "/certs/client", "/dummy-sys-class-dmi-id:/sys/class/dmi/id:ro" # Make kaniko builds work on GCP. ] [runners.machine] IdleCount = 50 IdleTime = 3600 OffPeakPeriods = ["* * * * * sat,sun *"] OffPeakTimezone = "UTC" OffPeakIdleCount = 15 OffPeakIdleTime = 3600 MaxBuilds = 1 # For security reasons we delete the VM after job has finished so it's not reused. MachineName = "srm-%s" MachineDriver = "google" MachineOptions = [ "google-project=PROJECT", "google-disk-size=25", "google-machine-type=n1-standard-1", "google-username=core", "google-tags=gitlab-com,srm", "google-use-internal-ip", "google-zone=us-east1-d", "engine-opt=mtu=1460", # Set MTU for container interface, for more information check https://gitlab.com/gitlab-org/gitlab-runner/issues/3214#note_82892928 "google-machine-image=PROJECT/global/images/IMAGE", "engine-opt=ipv6", # This will create IPv6 interfaces in the containers. "engine-opt=fixed-cidr-v6=fc00::/7", "google-operation-backoff-initial-interval=2" # Custom flag from forked docker-machine, for more information check https://github.com/docker/machine/pull/4600 ] [runners.cache] Type = "gcs" Shared = true [runners.cache.gcs] CredentialsFile = "/path/to/file" BucketName = "bucket-name"
GitLab.com runs Sidekiq with arguments
and the following environment variables:
SIDEKIQ_MEMORY_KILLER_MAX_RSS setting is
16000000 on Sidekiq import
nodes and Sidekiq export nodes.
Periodically executed jobs by Sidekiq, to self-heal GitLab, do external synchronizations, run scheduled pipelines, etc.:
GitLab.com being a fairly large installation of GitLab means we have changed various PostgreSQL settings to better suit our needs. For example, we use streaming replication and servers in hot-standby mode to balance queries across different database servers.
The list of GitLab.com specific settings (and their defaults) is as follows:
|effective_cache_size||338688MB||Based on how much memory is available|
|shared_buffers||112896MB||Based on how much memory is available|
|shmall||30146560||Based on the server's capabilities|
|shmmax||123480309760||Based on the server's capabilities|
Some of these settings are in the process being adjusted. For example, the value
shared_buffers is quite high and as such we are looking into adjusting it.
More information on this particular change can be found at
https://gitlab.com/gitlab-com/infrastructure/issues/1555. An up to date list
of proposed changes can be found at
GitLab.com adjusts the memory limits for the unicorn-worker-killer gem.
GitLab.com-specific rate limits
NOTE: Note: See Rate limits for administrator documentation.
IP blocks usually happen when GitLab.com receives unusual traffic from a single IP address that the system views as potentially malicious based on rate limit settings. After the unusual traffic ceases, the IP address will be automatically released depending on the type of block, as described below.
If you receive a
403 Forbidden error for all requests to GitLab.com, please
check for any automated processes that may be triggering a block. For
assistance, contact GitLab Support
with details, such as the affected IP address.
HAProxy API throttle
GitLab.com responds with HTTP status code
429 to API requests that exceed 10
per second per IP address.
The following example headers are included for all API requests:
RateLimit-Limit: 600 RateLimit-Observed: 6 RateLimit-Remaining: 594 RateLimit-Reset: 1563325137 RateLimit-ResetTime: Wed, 17 Jul 2019 00:58:57 GMT
- Search for
rate_limit_sessions_per_secondin GitLab.com's current HAProxy settings.
Rack Attack initializer
Details of rate limits enforced by Rack Attack.
Protected paths throttle
GitLab.com responds with HTTP status code
429 to POST requests at protected
paths that exceed 10 requests per minute per IP address.
See the source below for which paths are protected. This includes user creation, user confirmation, user sign in, and password reset.
This header is included in responses to blocked requests:
See Protected Paths for more details.
Git and container registry failed authentication ban
GitLab.com responds with HTTP status code
403 for 1 hour, if 30 failed
authentication requests were received in a 3-minute period from a single IP address.
This applies only to Git requests and container registry (
- Is reset by requests that authenticate successfully. For example, 29 failed authentication requests followed by 1 successful request, followed by 29 more failed authentication requests would not trigger a ban.
- Does not apply to JWT requests authenticated by
No response headers are provided.
Admin Area settings
- Has rate limits on raw endpoints set to the default.
- Does not have the user and IP rate limits settings enabled.
We use Fluentd to parse our logs. Fluentd sends our logs to Stackdriver Logging and Cloud Pub/Sub. Stackdriver is used for storing logs long-term in Google Cold Storage (GCS). Cloud Pub/Sub is used to forward logs to an Elastic cluster using pubsubbeat.
You can view more information in our runbooks such as:
- A detailed list of what we're logging
- Our current log retention policies
- A diagram of our logging infrastructure
GitLab.com at scale
In addition to the GitLab Enterprise Edition Omnibus install, GitLab.com uses the following applications and settings to achieve scale. All settings are publicly available at chef cookbooks.
We use Elasticsearch and Kibana for part of our monitoring solution:
We use Fluentd to unify our GitLab logs:
Prometheus complete our monitoring stack:
For the visualization of monitoring data:
Open source error tracking:
High Performance TCP/HTTP Load Balancer: