[2019-07-16] CI failures on master, missing pull policy
Summary
master
branch CI jobs have been failing in the specs due to migrations being out of date for several days (see specs
jobs). The reports show Database Schema - current: 20190529142545, codebase: 20190703130053
, though we know for a fact master
of GitLab Rails is much newer than this.
This appears to be caused by the lack of global.imagePullPolicy: Always
. Without this, pullPolicy
falls back to the server default, which is IfNotPresent
. Thus, jobs running on a node with an old image would result in the use of out of date images.
Steps to reproduce
https://gitlab.com/charts/gitlab/-/jobs/252868868
Configuration used
global:
hosts:
domain: helm-charts.win
hostSuffix: ""
ingress:
annotations:
external-dns.alpha.kubernetes.io/ttl: 10
configureCertmanager: false
tls:
secretName: helm-charts-win-tls
operator:
bootstrap: false
enabled: true
Current behavior
migrations
Jobs running gitlab-rails-ee:latest
images are only running updates to 20190529142545
vs 20190703130053
Expected behavior
Migrations run with latest images, fully up to date.
Versions
- Chart:
master
- Platform:
- Cloud: GKE
- Kubernetes: (
kubectl version
)- Client:
v1.12.7
- Server:
v1.11.10-gke.5
- Client:
- Helm: (
helm version
)- Client:
v2.12.1
- Server:
v2.12.3
- Client: