Open haooliveira84 opened 4 months ago
tagging @ishitasequeira from https://github.com/argoproj/argo-cd/pull/15036
Anyone looking to this?
same /similar: https://github.com/argoproj/argo-cd/issues/18467
Temp band-aid is to kick the argocd-application-controller
via kubectl rollout restart statefulset argocd-application-controller -n argocd
I had this issue as well, I thought that it happened because the cluster ownership registration in the configmap was not being expired. I didn't look into it further then because I ran it on prod and people started complaining :)
Maybe the application controllers can use a Kubernetes lease for every cluster instead of a configmap to keep track of the ownership between application controllers.
Our configmap stays healthy after restarts, I'll keep it running for a while and see where it breaks. (3 instances currently)
❯ k get cm argocd-app-controller-shard-cm -o yaml | k neat
apiVersion: v1
data:
shardControllerMapping: '[{"ShardNumber":0,"ControllerName":"argocd-application-controller-bb77ddc6-vlfhk","HeartbeatTime":"2024-09-18T07:06:37Z"},{"ShardNumber":1,"ControllerName":"argocd-application-controller-bb77ddc6-7fvrd","HeartbeatTime":"2024-09-18T07:06:36Z"},{"ShardNumber":2,"ControllerName":"argocd-application-controller-bb77ddc6-t8gpj","HeartbeatTime":"2024-09-18T07:06:35Z"}]'
kind: ConfigMap
metadata:
name: argocd-app-controller-shard-cm
namespace: argocd
Checklist:
argocd version
.Describe the bug I've updated the ArgoCD from v2.9.5 to 2.11.0 with
dynamicClusterDistribution: enabled
and the ArgoCD Applications don't sync every are stuck in refreshing, I see the application controller does not have many logs.To Reproduce 1 - Install Argocd with helm-chart with version: 6.9.3 2 - Edit the
values.yaml
and configure the controller todynamicClusterDistribution: enabled
Expected behavior
The application controller balances the clusters with yours pods. Sync stuck resolved