Open alexandrst88 opened 5 years ago
@jeff-knurek @TAvardit as the Helm experts in the house, can you please help?
@alexandrst88 - The problem is caused by having a rollingUpdate strategy with default settings (MaxSurge 25% maxUnavailable 25%) , one replica and PV. An ability to change these settings was added to chart version 1.10.0
(https://github.com/helm/charts/tree/master/stable/sonatype-nexus), so you can either upgrade your chart and edit the values or update you live deployment manifest to be set to :
maxSurge: 0
maxUnavailable: 1
The chart has no option to set MaxSurge/maxUnavailable. So this is still an issue.
Update: My fault, is is possible to set it. But the documentation is very missleading. Is says the default is "rollingUpdate" which is wrong, the default is "{}" - so we can pass an object here, not a string.
The correct value is:
deploymentStrategy:
type: RollingUpdate
rollingUpdate:
maxSurge: 0
maxUnavailable: 1
Edit: maybe this should be the default to make the chart work?
Actually, if no strategy was specified the default strategy is rollingUpdate
.
Did you experience issue deploying the chart without changing the empty object of deploymentStrategy
?
Yes. The default is maxSurge: 25%, maxUnavailable: 25%. So with 1 replicas, nexus is first started and then stopped leading to this issue.
Yes, I didn't want to change the default deploymentStrategy
but to allow users to change it based on their need.
When upgrade chart via helm, the new pod is stuck, with RollingUpdate strategy due to