Page MenuHomePhabricator

Initial production image build fails helm test
Closed, ResolvedPublic


...because there is no helm chart initially.

This is a catch22: we need a production image to build and test the helm chart, but the pipeline runs helm test as part of creating and pushing a production image.

You can see this failure happen most recently for zotero during the post-merge step:

Had a discussion with @dduvall in IRC:

18:13:29      marxarelli yeah, that sounds right: a service needs a chart before it can be fully tested and deployed via the pipeline
18:13:42      marxarelli i'm wondering if we can create a generic "bootstrap" chart or something
18:14:27      marxarelli just to get a service through the pipeline with a production image that was tested to some extent
18:14:37      marxarelli i.e. the port it listens on was probed or something
18:14:44      marxarelli not sure
18:14:56      thcipriani yeah, that's a pretty good idea
18:15:02      thcipriani a FIXME.tar.gz
18:15:25      thcipriani that does a basic: is this in a crash loop? no? lgtm
18:15:32      marxarelli right

Event Timeline

thcipriani triaged this task as Medium priority.Nov 19 2018, 6:23 PM
thcipriani moved this task from Backlog to Migration on the Release Pipeline board.

In the deployment pipeline meeting, an idea was presented to break this Catch 22 issue. Don't run the helm install if .pipeline/helm.yaml does not exist.

Change 483013 had a related patch set uploaded (by Thcipriani; owner: Thcipriani):
[integration/config@master] service-pipeline: check for .pipeline/helm.yaml

thcipriani claimed this task.

Change 483013 merged by jenkins-bot:
[integration/config@master] service-pipeline: check for .pipeline/helm.yaml