BUG: drift detection manager is installed erroneously on addon-controller restart for non drift detection clusters

[kylewuolle]

Problem Description

If a sveltos cluster does not have drift detection enabled and the addon controller is restarted then the drift-detection-manager deployment is created because the skipUpgrading function does not ensure that a cluster actually has drift detection enabled. This is because skipUpgrading just checks for the existence of the ResourceSummary CRD but not whether the drift detection is actually deployed.

System Information

CLUSTERAPI VERSION: 1.12.5
SVELTOS VERSION: 1.8.0
KUBERNETES VERSION: 1.33.1

[gianlucam76]

Thank you @kylewuolle

[gianlucam76]

Hi @kylewuolle can you please help me understand this scenario.

When Sveltos has to deploy drift-detection-manager, Sveltos uses this function deployDriftDetectionManagerInCluster

and that function calls deployDriftDetectionCRDs which deploys ResourceSummary CRD (using deployResourceSummaryCRD)

After that drift-detection-manager gets deployed.

So going to restart path. When Sveltos is restarted to make sure the correct version of drift-detection-manager is present on each managed cluster, it checks for the existence of the ResourceSummary CRD. If present it means drift-detection manager should also be there.

The only cluster were addon-controller deploys an extra drift-detection-manager is the cluster were Sveltos is running (as there the CRD is present from installation).

Is your bug for the management cluster?

[kylewuolle]

@gianlucam76 Yes it is!