[OpenAFS] Solaris 8 Fileserver hangs after upgrade to 1.2.5

Erwin Broschinski broschi@id.ethz.ch
Tue, 06 Aug 2002 17:18:25 +0200 (MEST)


Hi

2 weeks ago, I have upgraded a sun4u_58 AFS fileserver from IBM afs3.6 2.26 to
OpenAFS 1.2.5. (The client software remained IBM).
The server was running fine, for about 29 hours. Then there was no access to
the volumes on that server. (Some binaries located in volumes on that server
were hanging when started and '/bin/ls -l' froze.)
As root I could (localauth) bos uninstall. A following restart of the
bosserver to activate the 'old' binaries was hanging, until a kill -9 of the
fileserver process. This freed the bos restart and the fileserver started
salvaging. It was then up and running again with the old server binaries in
/usr/afs/bin.

(SalvageLog had some entries that I can't remember to have seen before:
08/06/2002 16:53:33 dir vnode 1049: special old unlink-while-referenced file
.__afsB90E is deleted (vnode 36346))


To make sure, it had something to do with the AFS upgrade, I retried on
Monday. And sic! same thing happened again now.

OpenAFS runs currently on 2 other Suns with Sol. 2.6 without problems, but on
machines with replicas only. All the other servers in our cell are still IBM
afs3.6 2.26

Any idea?


Erwin