Open lyesbit opened 8 months ago
One thing to note is that the error message, Namespace "336dc7c4-53f6-4a01-9125-b5dd1d37e3c3-p" (...) is not managed
is not necessarily because the Namespace doesn't exist, but rather that the Namespace is not defined in the list of Namespaces that Argo CD has access to on the cluster, via the cluster secret. See the namespaces
field of that Secret.
Right. However that still does not explain why that one un-syncable resource makes argocd not being able to sync the other healthy/syncable resources that belong to the same application/set, Any explanation ?
Any update here ?
ArgoCD versions 2.10 and below have reached EOL. Can you upgrade and tell us if the issue is still present, please?
Problem still persists
Argo CD: v2.11.6+089247d Build Date: 2024-07-23T00:41:35Z Go Version: go1.21.11 (Red Hat 1.21.11-1.module+el8.10.0+21986+2112108a) Platform: linux/amd64 jsonnet: v0.20.0 kustomize: v5.2.1 unknown Helm: v3.14.4+g81c902a kubectl: v0.26.11
Checklist:
argocd version
.Describe the bug
We're experiencing some annoying bug/feature of argocd. It can be broken down like this:
We have an applicationset, lets call it customer-network-policies (definition below in the screenshot part)
So basicaly argocd makes use of the generator to loop through a directory called dev-cluster/network-policies and generate one Network Policy per file.
Let's imagine for instance that the so called folder dev-cluster/networkpolicies/ contains 3 network policies files, and, for the sake of the example, let's also imagine that each netpol is created within its own namespace. When everything works fine the Argo CD UI renders something like this:
So far so good.
Problem occurs here:
To Reproduce
It very much looks like that one unhealthy network policy which cannot be created for whatever reason (in this case because the namespace does not exist) causes argocd to go bunkers and it cannot even sync/refresh the other network policies belonging to the same application/application set which are absolutely fine.
We've tested this at length and this behaviour is very much reproducible, but it feels wrong.
Expected behavior
One "unhealthy" resource failing to be created/updated/synced should not have any consequences on the other "healthy" resources. ArgoCD should be able to move forward with the healthy one and only display an error message for the unhealthy one.
Screenshots
Definition of the ApplicationSet
Error Occuring when including just one net pol that cannot be synced/created
ArgoCD logs following error
Version 2.8.4