You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Please provide an in-depth description of the question you have:
We're about to release our Flink on Karmada setup to DEV, but I've noticed that resources linger on the Karmada control-plane between subsequent helm uninstall and helm installs. For example, in our DEV setup, we install the FlinkDeployment CRD onto the Karmada control-plane along with related propagation policies, federatedresourcequotas, and cluster objects.
When I do a helm uninstall, the karmada-system namespace is deleted and all karmada related resources are taken down with it. My expectation is that when we do this helm uninstall, that all karmada related resources should be deleted as well.
However we try to install karmada again with helm install karmada -n karmada-system, we find the previous resources we had installed on the karmada control-plane still exist. I assume this might be Karmada's way of ensuring no data is lost if there the etcd pod is deleted, or something goes wrong with the karmada setup.
Does Karmada save some sort of metadata outside the scope of the helm installation that would cause resources to persist between installations? And if so, where is this data stored?
How can I reproduce this?:
Install karmada: helm install karmada -n karmada-system. Verify all Karmada pods are up and running.
Install a CRD to the Karmada control-plane. (You can also install a propagationpolicy, or any other karmada specific resource).
I remember we had a similar discussion about preventing etcd data from deleting during the development on karmada init and deinit. Not sure about the helm chart. @chaosi-zju Can you help to confirm it?
For karmadactl init the etcd data patch is specified by --etcd-data and defaults to /var/lik/karmada-etcd:
--etcd-data='/var/lib/karmada-etcd':
etcd data path,valid in hostPath mode.
Hi Karmada community!
Please provide an in-depth description of the question you have:
We're about to release our Flink on Karmada setup to DEV, but I've noticed that resources linger on the Karmada control-plane between subsequent helm uninstall and helm installs. For example, in our DEV setup, we install the FlinkDeployment CRD onto the Karmada control-plane along with related propagation policies, federatedresourcequotas, and cluster objects.
When I do a helm uninstall, the karmada-system namespace is deleted and all karmada related resources are taken down with it. My expectation is that when we do this helm uninstall, that all karmada related resources should be deleted as well.
However we try to install karmada again with
helm install karmada -n karmada-system
, we find the previous resources we had installed on the karmada control-plane still exist. I assume this might be Karmada's way of ensuring no data is lost if there the etcd pod is deleted, or something goes wrong with the karmada setup.Does Karmada save some sort of metadata outside the scope of the helm installation that would cause resources to persist between installations? And if so, where is this data stored?
How can I reproduce this?:
Install karmada:
helm install karmada -n karmada-system
. Verify all Karmada pods are up and running.Install a CRD to the Karmada control-plane. (You can also install a propagationpolicy, or any other karmada specific resource).
Uninstall karmada
helm uninstall karmada -n karmada-system
.Reinstall karmada. You will notice that the resources that were installed in step two still persist in the newly installed Karmada installation.
Environment: Kubernetes running on a Linux amd64 machine
The text was updated successfully, but these errors were encountered: