Skip to content

Sync & store code host versions and send in ping data

Administrator requested to merge mrn/ping-code-host-versions into main

Created by: mrnugget

This fixes https://github.com/sourcegraph/sourcegraph/issues/20033.

Approved RFC on metrics collection: RFC 417 Review: Add code host versions to pings PR to update the Big Query schema: https://github.com/sourcegraph/analytics/pull/211

What we have here works roughly as follows:

  • Every 24 hours we iterate over the list of external services
  • We group those external services by "unique code host identifier" (see comments in code) to avoid sending >1 requests to the same code host instance
  • We send a request to each* unique code host instance to get the version (*currently only GitHub, Gitlab, Bitbucket Server, since other versions are irrelevant to us in Batch Changes and I don't think querying the versions of gitolite, phabricator, etc. makes a lot of sense right now)
  • We store the versions in Redis
  • We only store external service kind (GitHub, Gitlab, etc) next to the version
  • Whenever ping data is sent up to Sourcegraph.com these versions are included, as a JSON array that would look roughly like this:
[
  {
    "external_service_kind": "GITHUB",
    "version": "unknown"
  },
  {
    "external_service_kind": "GITHUB",
    "version": "2.22.6"
  },
  {
    "external_service_kind": "BITBUCKETSERVER",
    "version": "7.11.2"
  },
  {
    "external_service_kind": "GITLAB",
    "version": "12.7.2-ee"
  }
]

Merge request reports