Open carlostaffe opened 2 years ago
Hi @carlostaffe , sorry for getting back late on this one.
Which version of Argo Operator did you upgrade from ? Is it < 0.1.0
to 0.4.0
? We introduced some breaking changes in 0.1.0
. After 0.1.0
the API is stable.
Hi Iam, As far as I remember ArgoCD operator installed last year was 0.0.15 . Thanks for ask.
Describe the bug An Argo-CD Operator upgrade to a 0.4.0 version (provided by Argo CD Community) led to an Argo-CD malfunction, did not sync any app anymore and a lot of application resources were not showed in UI console. Taking a look to ArgoCD application controller logging showed this message:
level=error msg="Failed to cache app resources: error setting app resource tree: dial tcp 172.30.138.187:6379: connect: connection refused"
With that endpoint, all services were ckecked.So, found "prd-redis-ha-haproxy" service had no pods as endpoint. "prd-redis-ha-haproxy" replicaSet had 1 as desired state and 0 current. Logs showed this cause:Error creating: pods "prd-redis-ha-haproxy-54575b64d9-" is forbidden: error looking up service account argocd/argocd-redis-ha: serviceaccount "argocd-redis-ha" not found
Service account provided by Argo-CD is "prd-argocd-redis-ha" instead of "argocd-redis-ha", So in order to fix manually this issue, we had to modify prd-redis-ha-proxy ReplicaSet from:Service Account: argocd-redis-ha -> Service Account: prd-argocd-redis-ha
To Reproduce Argo-CD is running in an Openshift Hub cluster. An Openshift cluster update from 4.8 to 4.9.46 version can reproduce the issue. This push an Argo-CD Operator update to a 0.4.0 version.
Expected behavior Deploying a new version of Argo-CD should not cause Redis proxy connection issues. It should not be necessary to manually repair replicaset entity.