Re: "interesting" crash : 3.18.44, huge xfs, nfs

[Emmanuel Florac <eflorac@intellique.com>]

[-- Attachment #1: Type: text/plain, Size: 1709 bytes --]

Le Tue, 4 Apr 2017 12:45:07 -0500
Eric Sandeen <sandeen@sandeen.net> écrivait:

> > 
> > I don't see why or where we'd use filestreams there... What's the
> > use case actually?  
> 
> well, it's in your crash backtrace, so you tell me ;)
> 

I remember now... There writing files in long sequences, so that's what
the filestream option is for. It worked fine for a few months though...

> >>>
> >>> The machine goes on crashing on disk access... xfs_repair 4.9 is
> >>> running now, and after that we'll reboot with a current kernel
> >>> (4.4.59). Any advice?    
> >>
> >> Yep, go back in time & get an xfs_metadump before you repair
> >> it.  :(  
> > 
> > You know how it goes it crashes, users scream, administrator jumps
> > on xfs_repair, and the rest is history.
> >    
> >> Please do at least capture repair output.
> >>  
> > 
> > Well the only thing I know is "there was nothing special in
> > xfs_repair output"... I'll tell the admin to catch the output next
> > time if necessary...  
> 
> So it crashed on every access, and repair fixed it, but repair said
> "nothing special" in the process?  :(
> 

It crashes, it reboots without complaining, so there aren't any grave
obvious upfront problems at least...

Well we'll see if it crashes again with a 4.4.59... Just in case we
removed the filestreams option in the fstab.


-- 
------------------------------------------------------------------------
Emmanuel Florac     |   Direction technique
                    |   Intellique
                    |	<eflorac@intellique.com>
                    |   +33 1 78 94 84 02
------------------------------------------------------------------------

[-- Attachment #2: Signature digitale OpenPGP --]
[-- Type: application/pgp-signature, Size: 181 bytes --]