Solaris 8, afs3.6 2.6 problem

[Touretsky, Gregory]

It's known bug in AFS 3.6 2.5 (aka patch 1) server code.
File server marks client as down when it gets too much requests from it.
As Transarc promise, it will be solved in p2 (to be release this month)
Till then we downgraded all file servers to AFS 3.6 2.3 (previous release),
thus solving the problem

Gregory Touretsky
IDC Computing / Systems Engineering Group
Unix Server Platforms
gregory.touretsky@intel.com
> (+) 972-4-865-6377, Fax: 04-865-5999
iNET: 465-6377, M/S: IDC-1B



-----Original Message-----
From: Mary P Washburn [mailto:mpmw@silk.Stanford.EDU]
Sent: Tuesday, April 03, 2001 5:15 PM
To: info-afs@transarc.com
Subject: Solaris 8, afs3.6 2.6 problem





In the past 3 months we have upgraded our AFS servers to solaris 8
running afs3.6 2.6.  Recently we are experiencing a problem where a
random client machine will lose sight of one or more afs servers;
rebooting the client machine, purging the cache, or other client
tricks do not fix the problem.  The client machines are a variety of
hardware, OS levels, and AFS client versions.  It appears that the
problem is with the AFS server and some internal database where that
client's name or IP are cached incorrectly.  The only way to fix the
problem is to restart bos on the problem AFS servers.  bos restart
-all for the AFS server causing the particular "connection timeout
problem" provides a fix, but restarting bos causes a 5 (or more)
minute outage or slowdown for anyone using AFS on campus, since we
have 20 AFS servers and the connection timeout problems affect any one
of these machines, this is not a comfortable fix.  Have any other
users experienced this problem?  Are there any other fixes aside from
bos restart?  We are considering downgrading our AFS server software
to afs3.6 2.0.


Thanks for any help you can provide.

  Mary Washburn
  Leland Systems
   Computing Systems and Services
   Stanford University
   Stanford, CA   94305-3090
   Telephone:  (650) 723-0334
   Fax:	    (650) 725-9121