Helm 3 install for resouces that exist

when running helm install (helm 3.0.2)

I got the following error: Error: rendered manifests contain a resource that already exists. Unable to continue with install: existing resource conflict: kind: PodSecurityPolicy, namespace: , name: po-kube-state-metrics

But I don't find it and also In the error im not getting the ns, How can I remove it ?

when running kubectl get all --all-namespaces I see all the resources but not the po-kub-state-metrics ... it also happen to other resources, any idea?

I got the same error to: monitoring-grafana entity and the result of kubectl get PodSecurityPolicy --all-namespaces is:

monitoring-grafana false RunAsAny RunAsAny RunAsAny RunAsAny false configMap,emptyDir,projected,secret,do

7 Answers

First of all you need to make sure you've successfully uninstalled the helm release, before reinstalling.

To list all the releases, use:

$ helm list --all --all-namespaces 

To uninstall a release, use:

$ helm uninstall <release-name> -n <namespace> 

You can also use --no-hooks to skip running hooks for the command:

$ helm uninstall <release-name> -n <namespace> --no-hooks 

If uninstalling doesn't solve your problem, you can try the following command to cleanup:

$ helm template <NAME> <CHART> --namespace <NAMESPACE> | kubectl delete -f - 

Sample:

$ helm template happy-panda stable/mariadb --namespace kube-system | kubectl delete -f - 

Now, try installing again.

Update:

Let's consider that your chart name is mon and your release name is po. Since you are in the charts directory (.) like below:

. ├── mon │   ├── Chart.yaml │   ├── README.md │   ├── templates │   │   ├── one.yaml │   │   ├── two.yaml │   │   ├── three.yaml │   │   ├── _helpers.tpl │   │   ├── NOTES.txt │   └── values.yaml 

Then you can skip the helm repo name (i.e. stable) in the helm template command. Helm will use your mon chart from the directory.

$ helm template po mon --namespace mon | kubectl delete -f - 

i've got the same issue while deploying Istio. So i did

kubectl get clusterrole kubectl get clusterrolebinging kubectl delete mutatingwebhookconfiguration istio-sidecar-injector kubectl delete validatingwebhookconfiguration istio-galley kubectl delete namespace <istio-namespace> 

and when deleted all and started, it worked.

I had the same error with CRDs objects. I used this chart on Github, and to prevent this error I used the --skip-crds flag. Maybe the project that you are using has something like this:

So nor the --force or no the other options help. Here is the error that I was getting.

Release "develop-myrelease" does not exist. Installing it now. Error: rendered manifests contain a resource that already exists. Unable to continue with install: existing resource conflict: namespace: , name: develop-myrelease, existing_kind: Kind=ClusterRoleBinding, new_kind: Kind=ClusterRoleBinding 

So I just delete clusterrolebinding and its work.

kubectl get clusterrolebinding | grep develop-myrelease kubectl delete clusterrolebinding develop-myrelease 

and run the deployment again.

for my case able to successfully upgrade my build with --force

Mulhasans-MacBook-Pro:helm-tuts mulhasan$ helm upgrade --install --force api-streamingserver ./api-streamingserver 

This will help for the same Release if you are doing with different release choose a different name for conflicting resources as of now Helmv3.x doesn't have option for CRDs --skip-crds is removed in Helmv3.x

If you are upgrading to helm 3, ensure you can run helm 2 and helm 3 separately. Example

helm2 list helm3 list 

After this, if you will try to install a helm chart within helm 3, that error will pops-up because it exists in helm 2.

Use helm2to3 plugin to upgrade to Helm3:

I follow this exactly and I got no issues

I spent many hours on bugs that are related to the error:

Error: rendered manifests contain a resource that already exists...

I have 3 simple conclusions:

1 ) Resources from previous deployments (via kubectl or helm) might exists in the cluster.

2 ) Use an advanced administrative/debugging tool like k9s or Lens to view ALL cluster resources (instead of kubectl get / helm ls).

3 ) Usally, the resource names which are specified in the error has a meaning - search directly for them and see if they can be deleted.