Experimental pure-Docker Sourcegraph cluster deployment reference
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Type Name Latest commit message Commit time
Failed to load latest commit information.
README.md
deploy-frontend-internal.sh
deploy-frontend.sh
deploy-github-proxy.sh
deploy-gitserver.sh
deploy-management-console.sh
deploy-pgsql.sh
deploy-query-runner.sh
deploy-redis-cache.sh
deploy-redis-store.sh
deploy-repo-updater.sh
deploy-searcher.sh
deploy-symbols.sh
deploy-syntect-server.sh
deploy-zoekt-indexserver.sh
deploy-zoekt-webserver.sh
deploy.sh
teardown.sh

README.md

Pure-Docker Sourcegraph cluster deployment reference

sourcegraph: search

There are two recommended ways to deploy Sourcegraph:

But what if your organization wants a multi-machine deployment without using Kubernetes? What if you use a different container management platform, for example? This project aims to solve that, by providing a pure-Docker deployment option.

The goal is that anyone using a container management platform other than Kubernetes (Netflix's Titus, Apache's Mesos, etc.) would be able to use this repository as a reference for how to deploy Sourcegraph.

⚠️ Alpha release

This project is using an alpha release of Sourcegraph v3.0.0. It is not yet heavily tested, and there may be major bugs or regressions.

When Sourcegraph v3.0.0 preview is released, this repository will be updated and code intelligence will work.

Deploying

First clone the repository, then:

./deploy.sh

Visit http://localhost:3080 to visit the running Sourcegraph instance!

Tearing down the deployment

./teardown.sh

System topology

To understand the system topology:

  1. Look at deploy.sh to get an overview of services.
  2. Every service (deploy-*.sh) has documentation inline indicating:
    • What the service does / provides.
    • What ports the service exposes.
    • What other services it talks to (see environment variables).

Service system requirements

Every service (deploy-*.sh) documents inline what the system requirements are (CPU/Memory/Disk allocation). For example, the frontend service.

Scaling / replicas

To scale the cluster deployment, you will need to:

  1. Deploy more instances of gitserver, searcher, and symbols services as desired.
  2. Configure the frontend and frontend-internal to communicate with the new instances.
  3. You're done! You do not need to configure or restart any other services.

Configuring HTTPS

See https://docs.sourcegraph.com/admin/nginx#nginx-for-other-sourcegraph-clusters-e-g-pure-docker

Configuring SSH cloning

  1. Ensure your Site admin > External services code host configuration is configured to use SSH. For example, by setting the gitURLType field to "ssh". You may still need to configure an access token or other codehost authentication method in order for Sourcegraph to discover your repositories.

Alternatively, you may use the OTHER codehost type under External services, which allows you to directly specify Git repository URLs for cloning.

  1. Provide your gitserver instances with your SSH / Git configuration (usually just .ssh/id_rsa, .ssh/id_rsa.pub, and .ssh/known_hosts -- but you can also provide other files like .netrc, .gitconfig, etc. if needed) by mounting it into the sourcegraph users home directory in the gitserver containers. For example, by adding the following flag:
-v ~/my-sourcegraph-ssh-config/.ssh:/home/sourcegraph/.ssh

All future Git cloning operations will use the credentials configured there.

If you wish, you can test that cloning with your configuration is working by performing the clone in a gitserver container shell, e.g. first acquire shell access:

$ docker exec -it gitserver-0 sh

Then try cloning the repository:

$ git clone ssh://git@myhost.com/my/repo /tmp/my-repo

Questions

Open an issue or contact us (support@sourcegraph.com), we are happy to answer any questions!