[torqueusers] Adding a new PBS node without restarting the server
sorrillo at jlab.org
Tue Jun 23 12:45:25 MDT 2009
We already have many compute nodes in our cluster and I wanted to add
one more. I wanted to see if I can do this without restarting the
as it is very fickle at our site. Can this be done?
Our PBS installation is under /var/spool/torque. Also, we already have a
standard procedure for installing all the appropriate client pieces on
the client new machine.
The pbs version is 2.1.10.
To add a new node the the queue:
1. Install the PBS mom on the client.
2. Check the correctness of /var/spool/torque/mom_priv/config on the
PBS client node.
On the server:
3. Ensure that the new host resolves properly in DNS.
4 Ensure that the new host is in /etc/hosts.equiv.
5. Add the node to /var/spool/toque/server_priv/nodes and restart the
pbs_server OR qmgr -c 'create node node_name np=8' and not restart?
How do I specify a particular queue using qmgr as we have several which
this server manages.
6. Start the pbs_mom on the client at this point.
7. Check that pbsnodes -a, reports the new node.
Are these steps correct?
More information about the torqueusers