kubectl rollout restart interrupted ?

[kfir]

hello

This is only an estimation, but I guess argocd with selfheal turned on, is interrupting manual rollout restart

when I trying to rollout the deployment, the newly created pods gets terminating after 20-40 seconds

when I do the same thing on deployment that doesnt managed by argo, I dont get the interruption

Don’t know if it’s the plugin, argo itself, or completely other think

[Travis Glenn Hansen]

Seems likely unrelated. argocd simply invokes kubectl apply behind the scenes.

[kfir]

hi @travisghansen

when we issue :

kubectl rollout restart deploy agent

What he does under the hood is adding annotation name:
kubectl.kubernetes.io/restartedAt

I guess what he does is:
When I dont use helmfile plugin, argocd selfheal have mechanism to ignore this annotation when he trying to auto-sync git repository by selfheal

But in the plugin , argocd doesn't go through git and matching the annotations, he just apply everything he got templated from helmfile plugin

[kfir]

Im still investigate it, mabye im wrong
But its seems like the plugin doesnt wait for the rollout to finish

Edit:
mabye its argocd itself that does that

[kfir]

@travisghansen My mistake
its doesnt related to the plugin at all

argocd doesnt know to handle empty annotations while rollout
hence, its chart problem