preflight · 5 checks

The machine is slow, or something ran out of memory

Memory, CPU or load average was inspected to explain slowness, a stall, or a container or process being OOM killed.

Before you call it done

1. Read the cgroup's limit, not the host's totals.cat /sys/fs/cgroup/memory.max and memory.events # free(1) is not scoped to the cgroupguards against NS-063
2. Rank every process by footprint, not just the one that died.ps -eo rss,comm --sort=-rss | head -20 # count anything spawned in a loopguards against NS-011
3. Check whether load is CPU work or uninterruptible sleep.ps -eo state= | sort | uniq -c # D means blocked on I/O, not runningguards against NS-064
4. Sample CPU over an interval rather than reading a lifetime average.Compare /proc/PID/stat utime+stime across a few secondsguards against NS-065
5. Resolve a PID before killing anything matched by pattern.pgrep -af PATTERN and read it; your own command matches tooguards against NS-012 NS-040 NS-039 NS-036

Plain text: /recipes.txt · all tasks