Closed maci0 closed 8 years ago
@maci0 What version of ZFS are you using? It would also be helpful to see the console logs, they should indicate which threads are blocked and why.
@behlendorf im using version 0.6.2
no idea how to get the console log, journald doesnt log anymore when this happens. this whole system just freezes and never recovers
sysrq doesnt really work either
@behlendorf I think this should be pretty easy to reproduce anyway.
@maci0 OK, we'll look it to it when we get a chance. Although this will probably be after some additional memory management work is completed.
@maci0 I am already working on a fix for this. The plan is to have it in 0.6.3.
@ryao , awesome :)
for now I have limited the arc cache to prevent my system from freezing randomly. once 0.6.3 is released i will test again
I confirm this bug on ubuntu saucy. Here are the steps to repeat it, and following that a syslog trace showing the problem. You can use this to test whether your fix is enough. This PPA contains the current stable 0.6.1 release.
To repeat: 1: Create a zpool with at least two partitions in a mirror, and zvol 1.5 times system memory. mkswap and set the system to use that as it's only swap space. Create another zpool with a mirror and another zvol, put a file system on it and run a script that changes the content all the time.
Create another script that does snapshots on the zvol with the changing files. Fill up system memory so the swap is engaged. Sooner or later you will get this core dump and a hang. There is no recovery.
...
Nov 18 20:00:01 noc3 CRON[18437]: (root) CMD (zfs-auto-snapshot -q -g --label=frequent --keep=4 //)
Nov 18 20:00:35 noc3 zfs-auto-snap: @zfs-auto-snap_frequent-2013-11-18-2000, 1 created, 1 destroyed, 0 warnings.
Nov 18 20:11:24 noc3 kernel: [36240.776022] INFO: task kworker/2:0:35 blocked for more than 120 seconds.
Nov 18 20:11:24 noc3 kernel: [36240.776028] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
Nov 18 20:11:24 noc3 kernel: [36240.776031] kworker/2:0 D ffff88041fc94580 0 35 2 0x00000000
Nov 18 20:11:24 noc3 kernel: [36240.776064] Workqueue: events async_pf_execute [kvm]
Nov 18 20:11:24 noc3 kernel: [36240.776066] ffff88040968db28 0000000000000046 ffff88040968dfd8 0000000000014580
Nov 18 20:11:24 noc3 kernel: [36240.776071] ffff88040968dfd8 0000000000014580 ffff880409691770 ffff88041fc94e28
Nov 18 20:11:24 noc3 kernel: [36240.776080] ffff88040968dbb0 0000000000000002 ffffffff8113ead0 ffff88040968dba0
Nov 18 20:11:24 noc3 kernel: [36240.776083] Call Trace:
Nov 18 20:11:24 noc3 kernel: [36240.776091] [
@maci0 @hcoin #2484 might help here.
Closing as stale, refer to the swap section of the wiki for additional configuration information.
as soon as I run
my system freezes. same happens when I just use it normally and some java process happens to eat up most of my memory