Closed johscheuer closed 2 years ago
I think if we're going to update our deployments to use a new flag, we will need to update the version of the image we're using. Alternatively, we could defer the updates to the sample deployments into a follow-up task from the original PR that adds the flag. I don't think we want our own deployments to always run on the latest tag, so I'm hesitant to recommend that to others through our samples.
I think to defer that change make sense (and opening an issue to track it). My point here is that we currently already install the latest version from the CRD and if I would run those comments I would expect to get the latest state from main, at least that's what the path implies.
kubectl apply -f https://raw.githubusercontent.com/FoundationDB/fdb-kubernetes-operator/main/config/crd/bases/apps.foundationdb.org_foundationdbclusters.yaml
kubectl apply -f https://raw.githubusercontent.com/FoundationDB/fdb-kubernetes-operator/main/config/crd/bases/apps.foundationdb.org_foundationdbbackups.yaml
kubectl apply -f https://raw.githubusercontent.com/FoundationDB/fdb-kubernetes-operator/main/config/crd/bases/apps.foundationdb.org_foundationdbrestores.yaml
kubectl apply -f https://raw.githubusercontent.com/foundationdb/fdb-kubernetes-operator/main/config/samples/deployment.yaml
If a user really wants to install a specific version not just for playing around they should install those files from tag. Otherwise the CRD and the operators capabilities might not match and the CRD has some fields that are unknown to the operator.
What happened?
When installing the operator with
kubectl apply -f https://raw.githubusercontent.com/foundationdb/fdb-kubernetes-operator/master/config/samples/deployment.yaml
the operator deployment will be in a crash loop:The flag was introduced in https://github.com/FoundationDB/fdb-kubernetes-operator/pull/1046 and all deployments were. adjusted but the issue is that those deployments refer to a tag version of the operator.
I think it would be more consistent to use
latest
in our deployments from the main branch and only use a specific version in our release tags. Otherwise we potentially install a newer CRD with features that are not supported in that operator deployment.What did you expect to happen?
That the following steps bring up the operator in a running state:
How can we reproduce it (as minimally and precisely as possible)?
Run:
Anything else we need to know?
-
FDB Kubernetes operator
-
Kubernetes version
-
Cloud provider
-