ZFS RAID-Z panic on vdev failure + subsequent panics and hangs

[Boris Kochergin]

Boris Kochergin wrote:
> Pawel Jakub Dawidek wrote:
>> On Fri, Aug 07, 2009 at 04:00:05PM -0400, Boris Kochergin wrote:
>>  
>>> Pawel Jakub Dawidek wrote:
>>>    
>>>> On Fri, Aug 07, 2009 at 03:34:34PM -0400, Boris Kochergin wrote:
>>>>  
>>>>      
>>>>> Pawel Jakub Dawidek wrote:
>>>>>           
>>>>>> Yeah, that's strange indeed. Could you try:
>>>>>>
>>>>>>     print ab->b_arc_node.list_prev
>>>>>>     print ab->b_arc_node.list_next
>>>>>>
>>>>>>
>>>>>>                
>>>>> (kgdb) print ab->b_arc_node.list_prev
>>>>> $1 = (struct list_node *) 0x1
>>>>>            
>>>> Yeah, list_prev is corrupted. If it panics on you everytime, I could
>>>> send you a patch which will try to catch where the corruption occurs.
>>>>
>>>>  
>>>>       
>>> I eventually get the arc_evict panic every time I successfully 
>>> manage to mount the filesystem, but it usually panics (with the 
>>> other backtrace) as soon as I try to mount it, or mount just hangs. 
>>> I'll gladly try the patch, though--the data on the array is 
>>> important to me. Thanks.
>>>     
>>
>> To get the data from there you could also try to 'zfs send' it without
>> mounting the dataset at all (just in case).
>>
>>   
> Sorry for the delay. I had to find another machine to move the disks 
> into so that I could continue experimenting. Anyway, the filesystem 
> didn't have any snapshots I could send, so I tried creating one with 
> "zfs snapshot home at 1" and the machine hung.
>
> FYI, In the new machine, all disks (including the one with the / 
> filesystem) retain their device names.
>
> -Boris
> _______________________________________________
> freebsd-fs at freebsd.org mailing list
> http://lists.freebsd.org/mailman/listinfo/freebsd-fs
> To unsubscribe, send any mail to "freebsd-fs-unsubscribe at freebsd.org"
Some more panics using RELENG_8 sources from yesterday: 
http://acm.poly.edu/~spawk/zfs/. The one in panic3.txt happens much more 
often than the other ones. If any brave soul wants to look into it, I 
can provide NFS/geom_gate/whatever access to the disk images (or actual 
disks, if there's a difference) so that they can recreate the problem on 
a local machine.

-Boris