Re: ZFS deadlocks or panics

Pawel Jakub Dawidek wrote:
On Tue, Jun 26, 2007 at 03:40:38PM +0200, Jan Schmidle wrote:
I'm trying to bild a storagebox with ZFS. Now I have some problems
which seems to be related to ZFS. Perhaps after adding disks to an
existing pool, all consoles trying to execute a command with I/O
operations freezes. Only sometimes I get a panic.

Are you sure the disks are fine?
Not 100%, but as the problem also shows up when I'm destroying
pools, I don't think that there is a Problem. But I will check this.

I played around with the tuning variables, proposed in
http://wiki.freebsd.org/ZFSTuningGuide
but didn't help.

The box is running -CURRENT amd64/SMP in a GENERIC configuration.
Last cvsup and build was "Fri Jun 22"

I also tried with i386 with and without SMP.

I collected some Information and uploaded it to
https://jschmidle.org/upload/

In "panic1" you can see a whole system-bootup resulting in a Panic.
"panic2" is just another lockup during bootup.

Backtraces would tell us if this is ZFS or not, can you include them?
I will post a Backtrace as soon as I get a panic again.

"areca" contains the information about the raid controller in use.
in "dump", I tried to create a coredump during a lockup.
"alltrace", "witness", "ps" also have been created during the lockup.
"sysctl" contains the current settings.

Can anyone recognize where the problem is coming from or is
additional information needed?


The following were created during a panic.
-----------------------------

db> show lockedvnods
Locked vnodes
db>

-----------------------------
db> show alllocks
Process 4144 (vdev:worker da2) thread 0xffffff0004b10000 (100230)
exclusive sleep mutex zfs:vdev:geom:queue r = 0 (0xffffff000707d4b0) locked @ /usr/src/sys/modules/zfs/../../contrib/opensolaris/uts/common/fs/zfs/vdev_geom.c:203
Process 4118 (spa_zio_intr_2) thread 0xffffff000456a680 (100172)
exclusive sleep mutex zio_cache (UMA zone) r = 0 (0xffffff00cff40b70) locked @ /usr/src/sys/vm/uma_core.c:2257
Process 20 (swi2: cambio) thread 0xffffff00010ea9c0 (100017)
exclusive sleep mutex Giant r = 0 (0xffffffff809db980) locked @ /usr/src/sys/cam/cam_xpt.c:7141
db>

-----------------------------------------------------
-----------------------------------------------------


And these during lockup
-----------------------------

db> show lockedvnods
Locked vnodes

0xffffff0004a2dd90: tag syncer, type VNON
usecount 1, writecount 0, refcount 2 mountedhere 0
flags ()
lock type syncer: EXCL (count 1) by thread 0xffffff000430c9c0 (pid 48)
db>

-----------------------------
db> show alllocks
db>

-----------------------------


_______________________________________________
freebsd-current@xxxxxxxxxxx mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-current
To unsubscribe, send any mail to "freebsd-current-unsubscribe@xxxxxxxxxxx"

Relevant Pages

  • Re: Encrypted zfs?
    ... Afterwards they remove all of them, shuffle them, ... ZFS has no trouble ... If you don't export the pool before ... ATA disks. ...
    (freebsd-current)
  • Re: Encrypted zfs?
    ... imported after the drives have been shuffled around. ... ZFS has no trouble ... ATA disks. ... After shutting the machine down (without exporting the pool) and shuffling disks around, it seems that the disks that were originally connected via SATA but that are now on the SAS controller show up as ...
    (freebsd-current)
  • Re: ZFS the perfect FS? if only...
    ... then ZFS would be the ultimate FS. ... Note you need at least 3 disks for raidz and 4 disks for raidz2. ... pool but not in a single mirror, ...
    (freebsd-current)
  • Re: ZFS deadlocks or panics
    ... >> I'm trying to bild a storagebox with ZFS. ... >> existing pool, all consoles trying to execute a command with I/O ... > Are you sure the disks are fine? ...
    (freebsd-current)
  • Re: ZFS deadlocks or panics
    ... >> I'm trying to bild a storagebox with ZFS. ... >> existing pool, all consoles trying to execute a command with I/O ... > Are you sure the disks are fine? ... panics. ...
    (freebsd-current)