[OpenAFS-devel] Got one of those interesting "bunch of servers won't talk to a client" situations right at the moment...

[Nathan Neulinger]

I sent Derrick a capture, unfortunately, I didn't have the -s parm set,
so it was the short default, and by the time I got his note, the machine
had already recovered on it's own.

-- Nathan

On Fri, 2005-03-18 at 17:22 -0500, Jeffrey Hutzelman wrote:
> 
> On Thursday, March 17, 2005 09:52:00 PM -0500 Derrick J Brashear 
> <shadow@dementia.org> wrote:
> 
> > On Thu, 17 Mar 2005, Neulinger, Nathan wrote:
> >
> >> Client and server are both running builds from within the past few
> >> months.
> >>
> >>
> >> -bash-2.05b# /usr/afsws/bin/fs checks
> >> These servers unavailable due to network or server problems:
> >> afs-fs1.cc.umr.edu afs-fs17.cc.umr.edu afs-fs7.cc.umr.edu
> >
> > I would have suggested it was the bug Tom Keiser sent us a patch for in
> > 1.3.79 but...
> >
> >> In a network trace, a few of the servers are sending back rx abort
> >> packets.
> >
> > This suggests otherwise. Looking at your tcpdump output I think you might
> > have more than one problem, possibly one which is this:
> > http://www.openafs.org/cgi-bin/wdelta/STABLE14-fix-multirx-checkservers-2
> > 0050216
> > (explaining why valid replies are seemingly ignored)
> >
> >> 14:42:01.771456 afs-fs1.cc.umr.edu.afs3-fileserver >
> >> sysinst.cc.umr.edu.afs3-callback:  rx abort (32)
> >> 14:42:01.774793 afs-fs7.cc.umr.edu.afs3-fileserver >
> >> sysinst.cc.umr.edu.afs3-callback:  rx abort (32)
> >
> >> afs-fs7.cc.umr.edu.afs3-fileserver:  rx data fs call get-time (32)
> >> 14:42:04.858423 sysinst.cc.umr.edu.afs3-callback >
> >> afs-fs1.cc.umr.edu.afs3-fileserver:  rx data fs call get-time (32)
> >> 14:42:05.321829 afs-fs1.cc.umr.edu.afs3-fileserver >
> >> sysinst.cc.umr.edu.afs3-callback:  rx abort (32)
> >> 14:42:05.835009 afs-fs7.cc.umr.edu.afs3-fileserver >
> >> sysinst.cc.umr.edu.afs3-callback:  rx abort (32)
> >
> > And this is something else. Can I see raw tcpdump output? I want to look
> > more clossely at the aborts.
> 
> By "raw", I assume Derrick means as produced by 'tcpdump -s 1500 -x'.
> Which would indeed be interesting.
> 
> -- Jeff
> _______________________________________________
> OpenAFS-devel mailing list
> OpenAFS-devel@openafs.org
> https://lists.openafs.org/mailman/listinfo/openafs-devel
> 
-- 

------------------------------------------------------------
Nathan Neulinger                       EMail:  nneul@umr.edu
University of Missouri - Rolla         Phone: (573) 341-6679
UMR Information Technology             Fax: (573) 341-4216