ebpf agent cannot run successfully in okd 4.14 cluster

[kaolaaz163]

I am using a cluster of OKD 4.14，Both loki operator and network observability operator have been successfully installed. When running ebpf agent, its status is shown as CrashLoopBackOff.

The pod of ebpf agent has the following error log. Can anyone help me figure out what the problem is?

[github-actions]

Congratulations for contributing your first netobserv-operator issue

[jotak]

Hi @kaolaaz163
Which version of netobserv are you using? Did you install a release from operatorhub, or directly from operator repo / main branch ? Can you also tell what's the docker image used in the agent pods?

[jotak]

cc @msherif1234

[kaolaaz163]

@jotak The environment information I use is as follows.I installed operator directly through OperatorHub.

OpenShift: OKD 4.14.0-0.okd-2023-11-14-101924
Network Observability Operator: 1.4.2
Loki Operator: 5.8.0
ebpf agent image:registry.redhat.io/network-observability/network-observability-ebpf-agent-rhel9@sha256:5dee0d22ca62ab56c1b3d07147a31602d9106a24c423740a75405a82eece8a7f

[kaolaaz163]

Can anyone help me figure out what the problem is?

[msherif1234]

This is a known kernel bug in 6.5.5 that I filed and fix is in review
https://lore.kernel.org/bpf/20231030132145.20867-1-shung-hsi.yu@suse.com/

[jotak]

As a side comment, we don't systematically test OKD fcos (based on fedora) so for better guarantees of stability I'd rather suggest to use OKD scos (based on centos stream), which is on par with the rhel-based distros that we support. Cf for instance the scos-stable list here: https://amd64.origin.releases.ci.openshift.org/#4-scos-stable