[OpenAFS] Re: Crazy DAFS problem (with log)
Mon, 21 Mar 2011 14:01:23 -0400
On Sun, Mar 20, 2011 at 11:54 PM, Ryan C. Underwood
> On Sun, Mar 20, 2011 at 09:57:22PM -0500, Andrew Deason wrote:
>> Oh, hmm, this starts too late. I guess you don't have the BosLog before
> Nope, wish it rotated more old logs out but haven't looked into this.
>> > 03/20/2011 16:36:17 The volume header file /vicepa/V0536871274.vol is not associated with any actual data (deleted)
>> > 03/20/2011 16:40:14 dispatching child to salvage volume 536871273...
>> > 03/20/2011 16:40:14 2 nVolumesInInodeFile 56
>> > 03/20/2011 16:40:14 CHECKING CLONED VOLUME 536871274.
>> > 03/20/2011 16:40:14 www_logs.readonly (536871274) updated 03/20/2011 16:39
>> > 03/20/2011 16:40:14 Vnode 22: length incorrect; (is 1794950 should be 0)
>> Do you know if this volume was recreated between these times? I'm not
>> sure if the salvager is just lying, but this would indicate it deleted
>> the RO, and then it came back again and again has invalid metadata.
> If by recreated you mean a remsite/addsite, the answer is no.
remite and addsite manipulate the vldb (only) and dont touch the
volume. you'd have to zap the clone, then release.
>> Based on the salsrv logs, I'd be looking at vnodes 264, 266, and 268 in
>> 536870915/6, and 22, 202, and 204 in 536871273/4. That is, see if the
>> 'UFS-Filename' file on disk actually has the length in the 'volinfo'
> If/when it comes back I'll dig into this stuff.
it did, per your next message. post what you find out, when you do.
if you don't find anything interesting, it may be worth trying a
fileserver without HAVE_PIO defined. are you able to rebuild or do you