[918] in Release_7.7_team
FrameMaker, rpcbind, and SGIs
daemon@ATHENA.MIT.EDU (Dorothy Bowe)
Wed Apr 2 14:29:00 1997
To: release-team@MIT.EDU
Date: Wed, 02 Apr 1997 14:28:56 EST
From: Dorothy Bowe <dot@MIT.EDU>
Over the weekend we began to hear scattered reports of SGIs being unable
to talk to the FrameMaker license server. The reports were consistent
with earlier problems we experienced with rpcbind on the SGI and
portmapper on the Suns. jhawk did some debugging and sent the following
mail to ops and myself:
> Twice this evening users from the W20 cluster logged in on SGIs
> reported problems wherein starting Frame hung. There was no reason
> to believe this was a license problem, as Frame never mapped a window.
>
> I looked at one of these, wherein trace (watchmaker) indicated Frame
> kept continually SIGALRM-ing.
>
> Pulling a rabbit out of my hat, I observed that the rpcbind/portmapper
> on the machine did not respond to "/usr/etc/rpcinfo -p". Restarting
> rpcbind fixed this problem.
>
> In a survey of all the SGI's in the host table in W20, these
> machines have nonresponsive rpcbinds, per rpcinfo -p:
>
> 18.187.0.122
> 18.187.0.107
> 18.187.0.124
> 18.187.0.129
> 18.187.0.89
> 18.187.0.103
> 18.187.0.126
> 18.187.0.121
> 18.187.0.127
> 18.187.0.86
>
> It's been speculated that this is due to a release bug.
> Perhaps someone else feels inclined to go get a stack trace
> from these rpcbind processes and disassemble them in time
> for tomorrow's release team meeting.
>
Jonathon and I talked about this a little and agreed that it might be
related to the patch release, though we have no idea how or why. With
yesterday's release team meeting snowed out, we hope that this message
will at least make release team members aware of the problem and alert
for possible solutions.
Dot
