-
Epic
-
Resolution: Obsolete
-
Critical
-
None
-
None
-
None
-
Leverage cgroups v2 to reserve resources for HyperShift hosted to avoid one cluster impacting other clusters
-
False
-
None
-
False
-
Not Selected
-
To Do
-
OCPSTRAT-576 - HyperShift Core Component Readiness for GA - Part II
-
OCPSTRAT-576HyperShift Core Component Readiness for GA - Part II
-
0
-
0
-
0
We observed during the chaos testing for HyperShift that one hosted cluster going rougue can impact the performance of other hosted clusters running on the same worker node given that they share the same resources ( IO, CPU, Memory and Network ) - https://issues.redhat.com/browse/SDE-2191.
It's important to reserve resources per each hosted cluster to avoid one bad cluster impacting the health and missing the SLOs/SLAs of other clusters. Cgroups v2 which is tech preview in OCP 4.12 can be enabled/leveraged to reserve the resources.
- relates to
-
HOSTEDCP-1429 Expose T-Shirt Size On HostedClusters
- Closed
-
HOSTEDCP-516 KubeletConfig and ContainerRuntimeConfig CRDs should not exist in HyperShift cluster
- Closed