Bug ID 1549549: Blades in the "none" partition may cause kubernetes services to fail.

Last Modified: Jan 30, 2025

Affected Product(s):
F5OS F5OS-C(all modules)

Fixed In:
F5OS-C 1.8.0

Opened: Feb 20, 2024

Severity: 3-Major

Symptoms

If blades in a chassis a assigned to the none partition, it is possible that kubernetes services may get scheduled on that blade, and fail because they cannot find the correct container version for the service. This can cause the kubernetes cluster to fail, and specific services in the cluster to fail.

Impact

The kubernetes cluster may show as failed, or the kubevirt or multus services may not operate correctly if their services land on one of the blades assigned to the none partition. This can cause existing tenants to fail, and new tenant deployments to fail.

Conditions

This can happen when there are one or more blades assigned to the none partition, and other blades and controllers in the chassis are rebooted. These reboots can cause the kubernetes services to get re-assigned to the blade in the none partition.

Workaround

The workaround is to move the blades in the none partition into a dummy partition that has a valid software version and is enabled. This will allow the blades to correctly start the kubernetes services assigned to those blades.

Fix Information

Blades moved to the none partition are now marked as Non-Schedulable to that kubernetes will not try to schedule any services on them.

Behavior Change

Guides & references

K10134038: F5 Bug Tracker Filter Names and Tips