on Tue, Dec 16, 2003 at 08:19:05PM -0200, Leandro Guimar?es Faria Corcete Dutra (ldutra@prefeitura.sp.gov.br) wrote:
> Em Ter, 2003-12-16 ??s 17:30, Karsten M. Self escreveu:
> > on Tue, Dec 16, 2003 at 11:57:34AM -0200, Leandro Guimar?es Faria
> > Corcete Dutra (ldutra@prefeitura.sp.gov.br) wrote:
> >
> > > Nautilus will freeze on all terminals at the same time, and refuse to
> > > display icons, open windows or refresh already open windows. It's
> > > impossible also to quit the Gnome session, but most other buttons are
> > > working, so one can open apps and use them to open files anyway.
> >
> > Are do you have any network-mounted drives/partitions? What filesystem
> > are you using for these (e.g.: NFS, samba, ...?). Network timeouts are
> > a great way to confound apps.
>
> Yes, I had... inherited some mounts of NFS from client machines, bad
> practice indeed. On your advice have eliminated them. Now I would
> need to wait and see if the problem reoccurs...
If these are read-only mounts, you might try the "soft" mount option.
If you do this with _writeable_ mounts, you risk badly corrupting data.
> There is also a SMB mount but it is user level, I don't think I can
> eliminate that.
>
> > In general, if you're trying to track down odd application errors,
> > 'strace' is your friend.
>
> OK, but how should I use strace with nautilus? Perhaps put it in a
> script at /usr/bin/nautilus with redirection to a log file?
>
>
> > > The weirder of all is that I can't kill Nautilus after it freezes.
> > > There are some defunct processes, but be it a killall -KILL or a kill
> > > -KILL no process dies, and I'm forced to reboot the server.
> >
> > 'kill -9'? That's terminate with extreme prejudice, but is one way of
> > determining whether or not the problem is the application or I/O layer.
>
> Yes, because -HUP and -TERM did nothing I was forced to -KILL.
>
> But what exactly -KILL not working means?
Files hung in an I/O wait state aren't killable. This is a
characteristic of most Unix-like systems (it's possible that some more
recent Unix-like OSs have worked around this limitation). However, one
of the classically few reasons for having to reboot a 'Nix box is to
clear bad or hung mounts, SCSI accesses, etc.
>
> > You might also want to try running 'lsof' or 'fuser' to see what file(s)
> > are being held open.
>
> But check what? I am used to fuser or lsof a file or directory, but
> what should be the parameter here?
RTFM and try it a few times.
Peace.
--
Karsten M. Self <kmself@ix.netcom.com> http://kmself.home.netcom.com/
What Part of "Gestalt" don't you understand?
Bush/Cheney '04: Less CIA -- More CYA
Attachment:
pgpDdi_wBeabq.pgp
Description: PGP signature