[Top][All Lists]
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [Gluster-devel] glusterfs and email store problem.
From: |
Anand Avati |
Subject: |
Re: [Gluster-devel] glusterfs and email store problem. |
Date: |
Thu, 8 Nov 2007 07:47:02 +0530 |
Jeff,
this is a known limitation of fuse, which hopefully might get fixed in the
future.
avati
2007/11/8, address@hidden <address@hidden>:
>
> Daniel,
>
> I have used strace before. I attached it with forking to the parent
> dovecot process and reproduced the issue.
>
> http://gluster.pastebin.com/m731ca211
> I pasted in in the above pastebin
>
> Thanks!
>
> -JPH
>
> Daniel van Ham Colchete wrote:
> > Sorry pal, my bad... I read it wrong :). You are using posix-locks.
> >
> > Is there a way to run the command getting the error through strace
> > when it occours? Strace shows every call made to the filesystem by the
> > software and show you the return value of each call.
> >
> > With strace -a you can attach it to a process, them you have to force
> > the error, and them check to see what's going on. If you can't find
> > it, past the result at http://gluster.pastebin.com/
> > <http://gluster.pastebin.com/> and show it to us please. Try to narrow
> > the moment of the error as such as possible in the strace log as it
> > can get quiet big. Using strace -tt or -ttt (I'm not sure) will help
> > you with that.
> >
> > Best,
> > Daniel
> >
> > On Nov 7, 2007 4:48 PM, address@hidden <mailto:address@hidden>
> > <address@hidden <mailto:address@hidden>> wrote:
> >
> > I am obviously new to glusterfs, however, I thought I had enabled
> > posix-locks?
> >
> >
> > volume posix-locks-knworksmail
> > type features/posix-locks
> > option mandatory on
> > subvolumes knworksmail
> > end-volume
> >
> > or am I missing something?
> >
> > -JPH
> >
> >
> > Daniel van Ham Colchete wrote:
> >> Jeff,
> >>
> >> reading the dovecot website, I saw this: Dovecot allows mailboxes
> >> and their indexes to be modified by multiple computers at the
> >> same time, while still performing well. This means that Dovecot
> >> works with NFS and clustered filesystems.
> >>
> >> The only way of doing this is using locks (flock or fnctl). Try
> >> activating posix-locks.
> >>
> >> I had a similar problem with maildrop recently. Because fnctl
> >> wasn't working it wouldn't change one file and report a
> >> filesystem error.
> >>
> >> Although, this doesn't explain the error message in the log, this
> >> is one problem you also have to solve.
> >>
> >> Best,
> >> Daniel
> >>
> >> On Nov 7, 2007 4:12 PM, address@hidden <mailto:address@hidden>
> >> <address@hidden <mailto:address@hidden>> wrote:
> >>
> >> I am hoping someone can shed some light on this issue for me.
> >>
> >> version info first:
> >>
> >> server OS: CentOS release 4.5 (Final)
> >> fuse: fuse-2.7.0-glfs5
> >> glusterfs: glusterfs-1.3.7
> >>
> >> client OS: CentOS release 4.5 (Final)
> >> fuse: fuse-2.7.0-glfs5
> >> glusterfs: glusterfs-1.3.7
> >>
> >> Mount:
> >> glusterfs on /mnt/glusterfs type fuse
> >>
> (rw,nosuid,nodev,allow_other,default_permissions,max_read=1048576)
> >>
> >> Configuration contents listed below issue.
> >>
> >> Issue:
> >> Looking at the logs on my mail (dovecot) server, I see the
> >> following errors:
> >> mmap() failed with index file
> >> /opt/GFS/postfix/vmail/jeff@ bofus.org/.Trash/.imap.index
> >> <http://bofus.org/.Trash/.imap.index>: No such device
> >> mmap() failed with custom flags file
> >> /opt/GFS/postfix/vmail/address@hidden/.Trash/.customflags
> >> <http://bofus.org/.Trash/.customflags>: No such device
> >>
> >> These of course are on the gluster mount, and the files
> >> really do exist:
> >> -rw------- 1 vmail vmail 6816 Nov 5 21:07
> >> /opt/GFS/postfix/vmail/jeff@ bofus.org/.Trash/.imap.index
> >> -rw------- <http://bofus.org/.Trash/.imap.index-rw-------> 1
> >> vmail vmail 100 Oct 15 14:11
> >> /opt/GFS/postfix/vmail/address@hidden/.Trash/.customflags
> >> <http://bofus.org/.Trash/.customflags>
> >>
> >> I was not using posix-locks at first and this same type issue
> >> came up
> >> but with the .subscription file. I am not sure whether
> including
> >> posix-locks or the restart/remount required to enable it
> >> fixed this
> >> issue for the .subscription file.
> >>
> >> This does not happen when I use a plain ext3 local disk
> >> mountpoint.
> >> Only on glusterfs mountpoint.
> >> Does anyone know why the files say "No such device" when they
> are
> >> clearly there on the filesystem?
> >>
> >> Thanks for any assistance!
> >>
> >> -Jeff Humes
> >>
> >>
> >>
> >>
> >>
> >> #################################
> >> # server config:
> >> volume knworksmail
> >> type storage/posix
> >> option directory /glusterfs/knworksmail
> >> end-volume
> >>
> >> volume posix-locks-knworksmail
> >> type features/posix-locks
> >> option mandatory on
> >> subvolumes knworksmail
> >> end-volume
> >>
> >> volume server
> >> type protocol/server
> >> option transport-type tcp/server
> >> subvolumes posix-locks-knworksmail
> >> option auth.ip.knworksmail.allow *
> >> option auth.ip.posix-locks-knworksmail.allow *
> >> end-volume
> >>
> >> volume writebehind
> >> type performance/write-behind
> >> option aggregate-size 1MB
> >> option flush-behind on
> >> subvolumes knworksmail
> >> end-volume
> >>
> >> #################################
> >> # client config:
> >> volume gluster01
> >> type protocol/client
> >> option transport-type tcp/client
> >> option remote-host 10.1.2.226 <http://10.1.2.226>
> >> #option remote-subvolume knworksmail
> >> option remote-subvolume posix-locks-knworksmail
> >> end-volume
> >>
> >> volume writebehind
> >> type performance/write-behind
> >> option aggregate-size 131072
> >> subvolumes gluster01
> >> end-volume
> >>
> >>
> >>
> >>
> >> _______________________________________________
> >> Gluster-devel mailing list
> >> address@hidden <mailto:address@hidden>
> >> http://lists.nongnu.org/mailman/listinfo/gluster-devel
> >>
> >>
> >
> >
>
> _______________________________________________
> Gluster-devel mailing list
> address@hidden
> http://lists.nongnu.org/mailman/listinfo/gluster-devel
>
--
It always takes longer than you expect, even when you take into account
Hofstadter's Law.
-- Hofstadter's Law