Closed jonathan-innis closed 3 years ago
So this is a delay caused by the helm-controller reconciliation interval? If so can we always keep it short since we don't necessarily need it to be tweaked by the end user?
Yes, it is. I think that we should re-reconcile very quickly while we are checking the status with the argo workflow, but not during standard reconciliation. Otherwise, this is going to put a lot of weight on the api server for no reason
@nitishm I'm picking this up and prioritizing this above testing because otherwise our test cases are going to take forever
Go for it!!
As I'm looking at it, it seems to be that we need to force re-reconcile after updating the HelmRepository/pushing the helm charts, I'm gonna try this, but I suspect this will fix it 😄
Describe the bug In cases where the interval is long and the
HelmRelease
doesn't reconcile immediately, it can take up to theinterval
specified for the helm chart for theHelmRelease
to re-reconcile. In some cases where this time > 5m, this can fail the entire workflow.To Reproduce Steps to reproduce the behavior:
./config/samples/bookinfo.yaml
(this issue is intermittent so you may not see this on the first go-around)Expected behavior To avoid ignoring user-specified reconcile interval, we should install the helm-release and force re-reconcile on a polling interval using the reconcile annotation provided through
HelmRelease
apiAdditional context More information on how to force re-reconcile can be seen within the
flux reconcile
portion of the Flux CLI https://github.com/fluxcd/flux2/blob/main/cmd/flux/reconcile.go