upgrade k8s without deleting etcd data #1038
Conversation
| kubectl create configmap host-configuration --from-file=host-configuration/ --dry-run -o yaml | kubectl apply --overwrite=true -f - |
There was a problem hiding this comment.
What‘s effect of --dry-run here? And what's effect of --overwrite here?
If I remove --overwrite, what will happen? I know when we label the node with overwrite, if the key already exists, it will be overwritten.
But when applying some object, without the --overwrite, it still works. So I wanna know what's the purpose of --overwrite here.
There was a problem hiding this comment.
The --dry-run here is to let create command generate the configuration without executing it. Since we need to read config from host-configuration file, the apply command cannot support reading config from a file. So I leverage the create dry run to generate the config and then apply it to the object.
The --overwrite flag is to resolve the conflict when applying new configuration by overwriting the old config with the new one. If we don't enable overwrite here, when merge the new config with the live config, if conflicts occur, the apply will fail and kubectl will report an error. Here I enable --overwrite is to make sure the new config can be successfully applied to the live object in case the object has been restored by k8s.
The etcd data will not be deleted by default when clean the cluster. New configurations will applied to the live objects if the objects is restored by k8s when the cluster is started.