Unexplained "dryRun is not supported" error for deploying Application targetted at OCM cluster

[vyta]

Describe the bug

Application workflow fails with "cannot create object: dryRun is not supported)" with OCM managedclusters

To Reproduce

1. Create hub cluster w/ vela + OCM hub + gateway manager
2. Create member cluster and join
3. Create Application definition with membercluster listed in topology
4. Observe application workflow fail with:

Message: run step(provider=multicluster,do=deploy): Found 1 errors. [(error encountered in cluster cluster1: Dispatch: pre-dispatch dryrun failed: Found 1 errors. [(cannot create object: dryRun is not supported)])]

Expected behavior

Should detect cluster and deploy application

KubeVela Version

CLI Version: v1.7.0
Core Version: 1.7.0

Cluster information

1.24.0

[Somefive]

Could you see what cluster-gateway image version are you using? You could run kubectl get deployment -n vela-system -owide and see it.

[vyta]

Could you see what cluster-gateway image version are you using? You could run kubectl get deployment -n vela-system -owide and see it.

v1.7.0-alpha.3

[Somefive]

I guess it might be because the OCM cluster proxy does not support the dryrun request proxy. Could you try to add --set featureGates.preDispatchDryRun=false to the helm install process. It will let the KubeVela controller skip the dryrun phase for dispatching resources.

[testinfected]

We're still getting that error with Vela 1.9.9

Our cluster gateway is 1.9.0-alpha2. @Somefive Any idea where to look to understand/fix this issue?

Thanks

[testinfected]

We saw that Cluster Gateway 1.9.0 was released, but that does not solve the issue.

Would any of the Kubela maintainers ( @FogDong ? @wonderflow ? @Somefive ?) please give a clue on how to solve this issue.

Thanks guys,

[FogDong]

Hi, have you tried adding --set featureGates.preDispatchDryRun=false to the helm install process?

[testinfected]

Hey @FogDong yes that works although we'd like to keep the pre-dispatch dry run if possible. Are you telling me it's not compatible with multicluster deployment?

[FogDong]

Yeah I think right now OCM cluster proxy still does not support the dryrun request proxy. Any insights from @Somefive ?

[testinfected]

@FogDong to clarify we're not using OCM ManagedCluster, just regular cluster join. Is it still expected that dryRun does not work?

[FogDong]

In a standard cluster setup, the dryRun option should indeed function as expected. I will investigate this issue further if it is not working as expected.

[testinfected]

@FogDong it's not working correctly for us. We use normal vela join and don't have any OCM addons installed.

We have to disable dry runs unfortunately. Any clues on what could be the problem here?

[testinfected]

Bump, any idea @FogDong? Thanks in advance