RE: [Oscar-users] Client won't finish installing image

Tue, 01 Feb 2005 23:00:10 -0800

Thank you for replying Bernard.  I kind of figured that I might have installed the updates too soon so I'm now in the process of reinstalling the OS again.  This time I'm going to install OSCAR first thing.  I really hope it works this time.

 

How would you recommend I do updates after it's done?  I mean after I get the image pushed out to all the nodes and its working.  Setting up a yum repository or something like that?  Is there a way to add all the updated RPM's into the image before you push it out?  Thanks again for your help!

 

Mike

 

From: Bernard Li [mailto:[EMAIL PROTECTED]
Sent: Tuesday, February 01, 2005 10:45 PM
To: Johnston Michael J Contr AFRL/DES; [email protected]
Subject: RE: [Oscar-users] Client won't finish installing image

 

Hi Johnston:

 

Can you tell us the output of:

 

rpm -qa | grep tftp

 

on your headnode?

 

Typically, we recommend you install OSCAR on a freshly installed Linux and only install updates after OSCAR has been successfully installed - not sure if this is the case, though.

 

Cheers,

 

Bernard

 

From: [EMAIL PROTECTED] on behalf of Johnston Michael J Contr AFRL/DES
Sent: Tue 01/02/2005 8:15 PM
To: '[email protected]'
Subject: [Oscar-users] Client won't finish installing image

First off, I want to thank all you gurus who have had your hand in this.  This is an awesome management tool and I'm excited to get it going...

 

I've been testing OSCAR on a 33 node cluster running RedHat 9.0.  When I set up the server it all went very well and I could build an image and reboot my nodes and have the TFTP boot pick it up and push out the image and they would come up great.  So after this I decided to use OSCAR to help upgraded my planned migration to Fedora C2.  That's where I'm running into problems.

 

I've rebuilt my head node to Fedora 2, updated all the RPM patches and then installed OSCAR.  The nodes all see the TFTP server and start to load the OS.  The problem now is that no node will finish loading its image.  They all die at different spots in the load and I see errors like "rsync: read error: connection reset by peer" and then it dumps out to the "systemimager.org/support" message.

 

I thought it might be my gigabit switch being overloaded, so I connected the head node directly to a slave with a cross-over cable and pushed it out again.  I still have the same problem.  It crashes every time.  It looks like it has something to do with 'rsync.'  Is there anything else I can try?  Thanks for you support.

 

Mike

Reply via email to