Hi Loïc, thanks for the pointer. Its kind of the opposite extreme to dropping just everything. I need to know the file name that is in cache. I'm looking for a middle way, say, "drop_caches -u USER" that drops all caches of files owned by user USER. This way I could try dropping caches for a bunch of users who are *not* running a job. I guess I have to wait for the jobs to end. Best regards, ================= Frank Schilder AIT Risø Campus Bygning 109, rum S14 ________________________________________ From: Loïc Tortay <tortay@xxxxxxxxxxx> Sent: Tuesday, October 17, 2023 3:40 PM To: Frank Schilder; ceph-users@xxxxxxx Subject: Re: Re: stuck MDS warning: Client HOST failing to respond to cache pressure On 17/10/2023 11:27, Frank Schilder wrote: > Hi Stefan, > > probably. Its 2 compute nodes and there are jobs running. Our epilogue script will drop the caches, at which point I indeed expect the warning to disappear. We have no time limit on these nodes though, so this can be a while. I was hoping there was an alternative to that, say, a user-level command that I could execute on the client without possibly affecting other users jobs. > Hello, If you know the names of the files to flush from the cache (from /proc/$PID/fd, lsof, batch job script, ...), you can use something like https://github.com/tortay/cache-toys/blob/master/drop-from-pagecache.c on the client. See comments line 16 to 22 of the source code for caveats/limitations. Loïc. -- | Loīc Tortay <tortay@xxxxxxxxxxx> - IN2P3 Computing Centre | _______________________________________________ ceph-users mailing list -- ceph-users@xxxxxxx To unsubscribe send an email to ceph-users-leave@xxxxxxx
 |