[Bug 293750] mv process hangs at 100% CPU (State R) in cache_fplookup / zap_lookup on ZFS

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=293750

--- Comment #2 from Chris St Denis <[email protected]> ---
Additional info:
www/www contained approximately 76,000 web files spread across a tree of 867
directories (average depth, around 5).

No files actually moved.

I have a 783G ktrace.out file.

# kdump -f ktrace.out | head -n 1000
 84123 mv       NAMI  "www/www"
 84123 mv       NAMI  "./www"
<same 2 lines repeat for all 1000 lines>



System did not shutdown cleanly. 

Seemed to hang after "All butters synced". I sent a ctrl+alt+del which seemed
to get it moving again for a bit. 

At the end it would hang on the following lines for >1 minute each.

bufdaemon wait error:35
bufspacedaemon wait error:35
bufspacedaemon wait error:35

At this point I gave up and power cycled the server (It's a production server,
and I couldn't let this drag on any longer). Hopefully the information I
collected will be useful.

I have ran a scrub which completed with no errors in 3:36 which seems amazingly
fast for ~215GB of data, but it is a mirror of nvme drives so may not be
unreasonable.

-- 
You are receiving this mail because:
You are the assignee for the bug.