Nodes in 'NotReady' state are not removed from their pool(s) when using ServiceType Loadbalancer

[marcriddle]

Setup Details

CIS Version : 2.13.1
Build: f5networks/k8s-bigip-ctlr:latest
BIGIP Version: BIG-IP 15.1.5.1 Build 0.0.14 Point Release 1
AS3 Version: 3.45.0
Agent Mode: AS3
Orchestration: K8S
Orchestration Version: v1.25.9
Pool Mode: Nodeport

Description

Nodes in 'NotReady' state are not removed from pool when using ServiceType Loadbalancer. If nodes are in this state when a service is created, they are not added to the pool that is created, but if they transition to this state they are not removed from the pool. Additionally, when new nodes are added to a cluster or 'NotReady' nodes transition to 'Ready' they are not added to the pool.

We have some automated processes that build, destroy, and reboot nodes, and part of this lifecycle involves the nodes transitioning to 'NotReady' before being restarted/destroyed. Ideally, they would be removed from any bigip pools at this point, as if not they'll need to fail health checks before being marked down, resulting in some number of failed requests to that node.

Steps To Reproduce

1. Create service of type 'LoadBalancer'
2. Stop kubelet on a node that belongs to the pool that was created, or otherwise transition node to 'NotReady'
3. Observe that node is still a member of the pool

Expected Result

Nodes marked 'NotReady' should be removed from pools that they belong to.

Actual Result

Nodes remain in pool and are not removed

[trinaths]

@marcriddle Please mail automation_toolchain_pm@f5.com this issue with required CIS configuration and resource examples to reproduce at our end.

[marcriddle]

Sent an email to the above address with details, please let me know if any additional information is needed.

[trinaths]

Created [CONTCNTR-4083] for internal PD tracking. ETA ~ 2.14

[vklohiya]

Fixed with #2855. Fix will be available in CIS 2.14.0 release.

[vklohiya]

Already fixed. Hence closing the issue.