COO OLM install/update fails due to perses-operator OOM #954
Description
When installing COO on a medium large OpenShift cluster (<50 nodes / <1000 namespaces / <10000 pods) the install fails because the perses-operator crashloops due to OOM, and the install never terminates.
Simply doubling the resource requests/limits in the generated ClusterServiceVersion is working, but has to be repeated at every install/update.
One possible fix I could see for this is the ability to set resource requests/limits for all operator pods inside the Subscription. Ideally a flag to disable certain operators could be exposed as well, for our installation (as an example) we don't need the perses operator und I set replicas=0 in side the CSV, but this has to be repeated at each install/upgrade as well.