Description of problem:

Recent versions of Cluster Logging Operator do not start up successfully, when there is no Elasticsearch API available.

If the Elasticsearch Operator is installed into the cluster the Cluster Logging Operator works fine.

Version-Release number of selected component (if applicable):

cluster-logging 5.5.5 (also on release-5.5, release-5.6 and master branch, deployed from source)

OCP 4.11 (also on 4.12)

How reproducible:

Steps to Reproduce:

1. Check that no Elasticsearch Operator is installed. If it had been installed to the cluster before, check that the Elasticsearch CRD is not present anymore.
2. Install Cluster Logging Operator (from catalog or from source) without also installing Elasticsearch Operator

Actual results:

Cluster Logging Operator pod will start to run but fail after a few minutes. This cycle will continue, eventually leading to CrashLoopBackOff state of the pod. The error message before exiting looks like this

{"_ts":"2022-12-09T15:51:29.687141987Z","_level":"0","_component":"cluster-logging-operator","_message":"Manager exited non-zero","_error":{"msg":"failed to wait for clusterlogging caches to sync: timed out waiting for cache to be synced"}} 

Expected results:

Cluster Logging Operator works fine, even without Elasticsearch Operator installed.

Additional info:

• Uninstalling the Elasticsearch Operator using the OpenShift Console does not remove the CRDs from the cluster, so it can be possible to have the Cluster Logging Operator running ok even without Elasticsearch Operator if the CRDs are still on the cluster from a previous install of the Elasticsearch Operator.
• The cause for this behaviour seems to be the addition of the watch during LOG-2999