[torqueusers] hanging jobs/communication error with mixed network environments

Alan Mosca alan at clusterresources.com
Fri Jul 6 04:04:18 MDT 2007

I know it might sound a bit like over-engineering on a small cluster,
but I generally recommend using DNS (bind9 possibly) with a local domain
configured. Where a mix of public/private ips appears, you can start
using DNS views and forget about it :)

 Alan Mosca
 EMEA Systems Engineer
 Cluster Resources, Ltd.

Adrian Knoth wrote:
> On Thu, Jul 05, 2007 at 06:33:00PM +0100, Alan Mosca wrote:
>>  I'm taking a bit of a blind guess, but probably the hostnames you have
>> set up in the "server_priv/nodes" file, are pointing to the public
>> interfaces (where available), so maybe fixing that by either changing
> You're right. I just said "racl00" or "racl01", and /etc/hosts contained
> multiple lines (public and private addresses) for these entries.
> I've now dedicated internal names (like racl00-svc, racl01-svc) for the
> private addresses, thus fixing the problem.
> Thanks for clarification.

More information about the torqueusers mailing list