Re: [Gluster-devel] glusterfs crash with mainline-2.5 patch 260 with dbench

[Amar S. Tumballi]

Hi Dale,
This bug is fixed in patch-270. Again, a big thanks for having a little bit
different spec file than the normal, which made the bug to show up :p

For the information of list, Dale was using io-threads below unify but
above client-protocol in client side, so he had two io-threads below unify.
(below, above corresponds to call flow, as in fuse comes at the top, and
posix comes at the bottom). We used to regularly test with a setup where
io-threads used to be above unify, so never hit on this bug.

Thanks again.
-bulde

On 7/3/07, Dale Dude <address@hidden> wrote:
> I had just turned it on actually. Trying your 'du -h' freezes the whole
> box freezes for a few seconds and glusterfsd cores only:
>
> #0  0x00002aaaab07b35a in strncpy () from /lib/libc.so.6
> #1  0x00002aaaab66a8bd in server_writedir (frame=0x2aaaab908610,
> bound_xl=0x5097f0, params=<value optimized out>) at server-protocol.c:4520
> #2  0x0000000000000000 in ?? ()
>
>
> Harris Landgarten wrote:
> > Dale,
> >
> > Have you had any problems with posix-locks. Yesterday I could not
> complete a du -h with it in the server chain. There have been a lot of fixes
> since but none I saw dealt directly with posix-locks.
> >
> > Harris
> >
> > ----- Original Message -----
> > From: "Dale Dude" <address@hidden>
> > To: "gluster-devel" <address@hidden>
> > Sent: Monday, July 2, 2007 5:32:38 PM (GMT-0500) America/New_York
> > Subject: Re: [Gluster-devel] glusterfs crash with mainline-2.5 patch 260
> with dbench
> >
> > Reply to list. Sorry for the direct email Harris ;)
> >
> > Dale Dude wrote:
> >
> >> Im now just using the below. After taking out iocache on the client
> >> there is no more crash. Thanks much for the ping,  Harris.
> >>
> >> Server: posix, posix-locks, io-threads, server
> >> Client: client, io-threads, RR unify, writebehind, readahead
> >>
> >> Dale
> >>
> >> Harris Landgarten wrote:
> >>
> >>> Dale,
> >>>
> >>> I was only using unify, readahead and writebehind on the client. I
> have io-threads on the server
> >>>
> >>> Harris
> >>>
> >>> ----- Original Message -----
> >>> From: "Dale Dude" <address@hidden>
> >>> To: "gluster-devel" <address@hidden>
> >>> Sent: Monday, July 2, 2007 5:18:40 PM (GMT-0500) America/New_York
> >>> Subject: Re: [Gluster-devel] glusterfs crash with mainline-2.5 patch
> 260 with dbench
> >>>
> >>> Removing performance/io-cache from glusterfs-client.vol solves this.
> Now
> >>> only using writebehind.
> >>>
> >>> Dale
> >>>
> >>> Dale Dude wrote:
> >>>
> >>>
> >>>> Ran again with 261 (which only seems to fix perms?) with same
> results.
> >>>> Using ubuntu dapper (2.6.15-28-amd64-server) with fuse 2.6.5.
> >>>>
> >>>> I only have one dbench test file so I have no choice.
> >>>> -rw-r--r-- 1 root root 26214401 2005-11-19 21:19
> >>>> /usr/share/dbench/client.txt
> >>>>
> >>>>
> >>>>
> >>> _______________________________________________
> >>> Gluster-devel mailing list
> >>> address@hidden
> >>> http://lists.nongnu.org/mailman/listinfo/gluster-devel
> >>>
> >>>
> >>>
> >>>
> > _______________________________________________
> > Gluster-devel mailing list
> > address@hidden
> > http://lists.nongnu.org/mailman/listinfo/gluster-devel
> >
> >
> >
> _______________________________________________
> Gluster-devel mailing list
> address@hidden
> http://lists.nongnu.org/mailman/listinfo/gluster-devel



--
Amar Tumballi
http://amar.80x25.org
[bulde on #gluster/irc.gnu.org]