Re: [condor-users] 6.6.0 upgrade

[Robert Krzaczek <krz@xxxxxxxxxxx>]

On Tuesday, Jan 13, 2004, at 11:14 America/New_York, Mike Smorul wrote:
>  We've tried installing 6.6.0 on several RedHat 9 clients. Our central
> manager is running 6.4.7 for now. After running for a few hours, all 
> the
> 6.6.0 hosts will disappear on the central manager even though all the
> daemons are running fine.

We see a very similar situation on a mix of Sparc Solaris systems under 
Condor 6.6.0.  In our situation, we see the machines disappear one hour 
after bringing the flock online.  The central manager is a Solaris 8 
MU6 system, the rest of the flock are a mix of Solaris 8 and 9.

> The only messages about their disappearance is in the CollectorLog

I found other errors in our log files; for example, the central 
manager's negotiator reported

---------- Started Negotiation Cycle ----------
Phase 1:  Obtaining ads from collector ...
  Getting all public ads ...
Couldn't fetch ads: communication error
Aborting negotiation cycle

I plan on re-upgrading our flock to 6.6.0 (it's currently downgraded 
back to 6.4.7) to check into this further.  As I'm reviewing the logs I 
captured from our brief 6.6.0 run, I'm seeing other messages that might 
be red herrings, or might be further symptoms ("DC_AUTHENTICATE attempt 
to open invalid session..."); I want to see if those errors happen 
before or at the one hour failure point.

\bob


--
\def\bob{Bob Krzaczek, RIT Center for Imaging Science, krz@xxxxxxxxxxx}

Condor Support Information:
http://www.cs.wisc.edu/condor/condor-support/
To Unsubscribe, send mail to majordomo@xxxxxxxxxxx with
unsubscribe condor-users <your_email_address>