Customer is experiencing some pods with message: error loading seccomp filter: errno 524. The problem started on Sep, 16ht after deploying new application pods. One worker node (#3) was more affected (may be due to skew or workoad). Problem is similar to the KCS https://access.redhat.com/solutions/7030968 ! We fixed the issue increasing from the default CoreOS bpf_jit_limit value to a higher value (changed from 5... to 8... or about 60% increased).

From

sudo cat /proc/sys/net/core/bpf_jit_limit
528482304

To

sudo cat /proc/sys/net/core/bpf_jit_limit
828482304

Customer wants to confirm why this issue happened in 4.18.z as the issue was solved in the 4.12 / 4.13 versions.

The worker #3 also had > 20K defunct processes due to the failures on pod start. They will check a way to improve the app pod debugging.

Also we want to know if it is possible to check online the current usage values of bpf_jit_limit to predict the increasing if necessary. If the kernel has his capability to confirm somewhere (/proc files etc).

Ticket: 04267020 hsa must-gather, strace output, and other log outputs.