-
Bug
-
Resolution: Unresolved
-
Critical
-
None
-
4.11.z
-
Important
-
No
-
False
-
-
N/A
-
Release Note Not Required
Description of problem:
During the upgrade to OCP 4.11 customers are experiencing problems related to an increasing memory requested by control plane components. This was a not expected behavior. Customers had to add more memory to master nodes in order to complete successfully the upgrade. Based on this solution https://access.redhat.com/solutions/6977471 this was expected, but customers read the official documentation instead of searching a possible future problem into the KB.
Version-Release number of selected component (if applicable):
How reproducible:
Considering the same workload for example in OCP 4.10 that is perfectly handled with effort using master nodes 4x16, upgrading to OCP 4.11 the upgrade could be stuck
Steps to Reproduce:
1. Upgrade to OCP 4.11
Actual results:
Upgrade to OCP 4.11 is stuck, ETCD pods are continuously restarted and different CO are degraded
Expected results:
Cluster upgrade is performed correctly
Additional info:
Please highlight this more memory usage in the official 4.11 documentation. Also mention the new alert HighOverallControlPlaneMemory, our customers should be aware that it is good to stay under 60% of memory usage across master nodes
