“zpool history” is consuming 100% CPU

[mpfusion]

Since fivteen hours the “zpool history” command is eating up my CPU. Why is that and why does it take so long to finish?

Code:

[root@freenas] /mnt/tank# ps aux | grep zpool
root        2279 100.0  0.0  1704772  2156 ??  R    1:15AM  817:53.33 zpool history tank
root        2280  0.0  0.0    16272  1424 ??  I    1:15AM    0:00.00 egrep ^[0-9\\.\\:\\-]{19} zpool scrub tank$
root      39364  0.0  0.0    16272  1952  0  S+    3:10PM    0:00.00 grep zpool

What is going on here?

 

[cyberjock]

Good question. Is your pool in good health? Whenever I've run zpool history it has taken, at most, 10 seconds. So either you've got a zpool that's older than the universe(and that long history takes a REALLY long time to parse out) or there's something else going on.

Do you have something to say to us about your pool.. perhaps with the words "from the future" and "time machine" and "I'm a badass"? ;)

 

[mpfusion]

The pool is five months old and nothing spectacular, two mirrored vdevs. I have no idea what might cause this command to hang. Is it save to just kill it?

 

[cyberjock]

Yeah. I'd say it's pretty safe to kill it. You could check the zpool iostat 1 command and see if the disks are being used a lot before you kill it. I don't think zpool history writes to the pool at all, but better to check with the iostat before killing it.

 

[mpfusion]

I killed it and the system is back to normal. But that doesn't solve the problem. “zpool history” still hangs. And I assume it's run by FN automatically and the next time it will run the CPU goes up to 100% again.

 

[cyberjock]

Hmm... that's... odd. I have no recommendations since I've never seen this problem before.

 

[mpfusion]

To follow up this thread, I created ticket #4736 and it seems to be fixed in the most recent build.