Running the command sudo -u mwpresync /usr/bin/scap stage-train -Dfull_image_build:True --yes auto, produced multiple Error: unknown command "diff" for "helm" errors.
The full stack trace is at P45285
We noticed that the helm environment variables haven't been set for the mwpresync user, which looks like the cause of this failure.
| Subject | Repo | Branch | Lines +/- | |
|---|---|---|---|---|
| profile::mediawiki::deployment::server: Don't pass HELM_* vars to train presync | operations/puppet | production | +0 -7 |
| Reference | Source Branch | Dest Branch | Author | Title | |
|---|---|---|---|---|---|
| repos/releng/scap!102 | review/dancy/T331479 | master | dancy | kubernetes.py: Extract HELM_* env vars from /etc/profile.d/kube-env.sh if available |
We ran into this issue back in December, the reason of the problem being the K8s env is set for login shells only: https://phabricator.wikimedia.org/T324553
In that ticket ops recommended as a workaround to either source the env explicitly or preserve the env with -E when calling sudo: https://phabricator.wikimedia.org/T324553#8447006
Would be nice if we could source the K8s env for mwpresync as well though.
dancy opened https://gitlab.wikimedia.org/repos/releng/scap/-/merge_requests/102
kubernetes.py: Extra HELM_* env vars from /etc/profile.d/kube-env.sh if available
jnuche merged https://gitlab.wikimedia.org/repos/releng/scap/-/merge_requests/102
kubernetes.py: Extract HELM_* env vars from /etc/profile.d/kube-env.sh if available
Change 900731 had a related patch set uploaded (by Ahmon Dancy; author: Ahmon Dancy):
[operations/puppet@production] profile::mediawiki::deployment::server: Don't pass HELM_* vars to train presync
Change 900731 merged by Clément Goubert:
[operations/puppet@production] profile::mediawiki::deployment::server: Don't pass HELM_* vars to train presync