Scheduler Delay

Scheduler delay happens when a process is ready to run but does not get CPU time quickly enough.

Symptom

service has CPU available overall, but individual requests pause unexpectedly

What kprobe captures

kprobe records scheduler switches and connects them to service activity, CPU, process, thread, and request context.

Causal pattern

batch-worker runs on CPU 2
  -> payment-api thread waits
  -> sys_read resumes late
  -> request exceeds timeout

What to check

• CPU core involved
• process that ran before the delayed service
• cgroup or pod responsible
• whether the delay repeats across requests