Post AYFMnSkFGmd5ULRqmO by sten@chaos.social
 (DIR) More posts by sten@chaos.social
 (DIR) Post #AYEQmmCde4cACmhI4O by attie@chaos.social
       2023-07-30T20:53:47Z
       
       0 likes, 0 repeats
       
       A couple of weeks ago, I was working on a massively parallel software project. One time, I ran the thing, swap started filling very quickly, I cancelled it, and it left *hundreds* of zombie processes behind.Since then, I've noticed that the swap was regularly being used, well beyond what would make sense. `htop` reported ~30 GiB of free RAM, yet ~8+ GiB was getting into swap.Yesterday I had failed allocs when running some things.Killing the container seems to have fully resolved the issue.
       
 (DIR) Post #AYEQmnVomFYCGYg7bU by attie@chaos.social
       2023-07-30T20:56:02Z
       
       0 likes, 0 repeats
       
       ... I didn't gather enough info before killing the container, but... are zombie processes not reported correctly as part of the "used" memory in Linux? That's the only explanation I can think of that makes sense here...~24h later, and we're still at 0% swap (which is what I'd expect).... and since killing that container, caches & buffers are filling and using memory again.
       
 (DIR) Post #AYEQmor7mWBiQveeS8 by penguin42@mastodon.org.uk
       2023-07-30T21:00:16Z
       
       0 likes, 0 repeats
       
       @attie But I thought zombies didn't use any resources except existing as a process?
       
 (DIR) Post #AYEQqitzrKHH7urMWm by attie@chaos.social
       2023-07-30T21:00:59Z
       
       0 likes, 0 repeats
       
       @penguin42 Yeah, I thought they were basically the exit status & info, waiting for someone to collect... otherwise all resources free'd
       
 (DIR) Post #AYFMnSkFGmd5ULRqmO by sten@chaos.social
       2023-07-31T07:50:16Z
       
       0 likes, 0 repeats
       
       @penguin42 @attie was about to write exactly that. Unless you have so many zombies that you run out of PIDs, they should not be a problem for performance.