We are a bit stumped.
Once in a while ps
/w
hang forever (at least an hour) on our server, and it is on reading /proc/pid/cmdline
for a process (happend twice today, verified with strace cat
directly on cmdline
).
Only things that fixes it is a kill -9
.
There are no containers running on the server, and no OOM conditions (actually >3TB memory free). No NFS or other network filesystems.
Running 5.19.0-1019-aws #20~22.04.1-Ubuntu SMP Thu Jan 26 11:24:23 UTC 2023 x86_64 x86_64 x86_64 GNU/Linux
, Ubuntu 22, its a high core-count and high memory NUMA system.
The offending processes are at S or Z states during the hang.
We read the following but it didn't help us:
ChatGPT4 is stumped to.
Any ideas? Thanks!