New "wait" on gitlab_charts may hit timeout on fresh install
!700 (merged) added a "wait" parameter to the helm chart deploy to make ansible wait until the chart is actually deployed before proceeding, to attempt to deal with a few transient problems ascribed to pods not being ready yet.
However, this is passed to helm
directly (as --wait
) and in that case the default timeout (overridden with --timeout
on helm) is 5 minutes. This is insufficient time for deploying the GitLab helm chart for the first time, in at least some small cases (it might be better in more highly provisioned environments, but the dev environment where this happens is not exactly tiny). The result is a Error: context deadline exceeded
See some discussion on !700 (comment 978065969).
In short, it appears we'll need an explicit timeout
with a value large enough to work reliably but without making it wait too long if there are real problems with the deploy.