[Gluster-users] Strange issues with du and df

[Steve]

-------- Original-Nachricht --------
> Datum: Tue, 31 Mar 2009 15:26:00 +0300
> Von: Stas Oskin <stas.oskin at gmail.com>
> An: gluster-users <gluster-users at gluster.org>
> Betreff: Re: [Gluster-users] Strange issues with du and df

> Hi.
> Ok, df was incorrect because one of servers went into <defunct> (zombie)
> state, and df waits until the software done writing or releases writing
> handles.
> 
> I'm giving GlusterFS a 3rd try, but these issues are very critical:
> 
> 1) 0 size files
>
Welcome to the club.


> 2) AFR losts some of the files - after a day of running there is a
> mismatch
> between the one server and the second one
>
Do you mean that you loose the whole file or just the content of the file?
Somehow lately the whole AFR functionality is borked for me. I tried to reduce everything to the minimum (no performance translators or anything other not needed) and still I am loosing data. Restarting server and client can lead to total loss of content for some files in my case.


> 3) Possible memory leak
>
Could you provide valgrind dumps showing where the leaks are occurring?


> 4) Server going to defunct state
> 
> 
> Any idea how it would be the best to diagnose and advance with these
> issues?
> 
> Thanks!
> 
> 2009/3/31 Stas Oskin <stas.oskin at gmail.com>
> 
> > Hi.
> >
> > I'm receiving very strange results with AFR:
> >
> > Client:
> >
> > df -h:
> > Filesystem            Size  Used Avail Use% Mounted on
> > glusterfs              31G   28G 1020M  97% /mnt/media
> >
> > du -sh for mounted directory:
> > 1007M  (ompletely different)
> >
> >
> > Server 1:
> > df -h:
> > Filesystem            Size  Used Avail Use% Mounted on
> > /dev/hda4              31G   28G 1021M  97% /media
> >
> > du -sh for exported directory:
> > 1007M (completely different here too!)
> >
> >
> > Server 2:
> > Filesystem            Size  Used Avail Use% Mounted on
> > /dev/hda4              31G  1.1G   28G   4% /media
> >
> > du -sh for exported directory:
> > 920M
> >
> >
> > 1) It's just like the AFR completely broken.
> > 2) What's more strange, that the server 1 has completely different stats
> > for 2 commands - first time that I see it.
> > 3) This happened after the disk space on GlusterFS got completely full.
> >
> >
> > I also have some files with 0 sizes, so the other mentioned problem
> exist
> > too. The used version is 2RC7.
> >
> > How can these issues be tracked down? They are very serious,
> > and basically influent the whole cluster reliability.
> >
> > Regards.
> >

-- 
Psssst! Schon vom neuen GMX MultiMessenger gehört? Der kann`s mit allen: http://www.gmx.net/de/go/multimessenger01