solaris assert triggered after panic in ZFS at mount time

[Alexander Leidinger]

Quoting Andriy Gapon <avg at freebsd.org> (from Fri, 13 Jul 2018 14:47:01 +0300):

> On 13/07/2018 14:12, Alexander Leidinger wrote:
>> Hi,
>>
>> This is with r336194.
>>
>> I have a dataset which panics the machine at mount time (during  
>> reboot after a
>> panic):
>> ---snip---
>> panic: solaris assert: dmu_object_claim(zfsvfs->z_os, obj,
>> DMU_OT_PLAIN_FILE_CONTENTS, 0, obj_type, bonuslen, tx) == 0 (0x11  
>> == 0x0), file:
>> /usr/src/sys/cddl/contrib/opensolaris/uts/common/fs/zfs/zfs_znode.c, line:  
>> 819
>
> Hmm, 0x11 == 17 == EEXIST.
> So, this looks like a possible race in ZIL code.  The intent log contains a
> record to create a file with certain ID, but that file already  
> exists on disk.
> I don't know what to recommend.  Maybe try to import discarding one  
> transaction
> group...  If you can afford losing data from that TXG.

Yes, I can afford losing data, it's mainly usr/local of a jail and  
usr/local/etc I have in a backup. I'm rebuilding it and I would like  
to compare that I didn't miss something.

This is a root pool of a server without local access, how do I get rid  
of TXG of a dataset of an already imported pool (no possibility to  
have this pool accessed in an exported state)?

Bye,
Alexander.

-- 
http://www.Leidinger.net Alexander at Leidinger.net: PGP 0x8F31830F9F2772BF
http://www.FreeBSD.org    netchild at FreeBSD.org  : PGP 0x8F31830F9F2772BF
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 819 bytes
Desc: Digitale PGP-Signatur
URL: <http://lists.freebsd.org/pipermail/freebsd-fs/attachments/20180713/263c6820/attachment.sig>