[torqueusers] trusted hosts for torque with a separate head node
kissi at uni-mainz.de
Sat Jul 31 21:02:25 MDT 2010
Dear torque users and developers,
I have a question regarding the list of trusted hosts for torque.
Namely, the head node (called gadolinium in my case) it is an old dual-core Xeon machine.
This node is having the functionality of running the torque server, but it is also the
main NFS server, so I explicitly want to exclude this box from the list of compute nodes, which
are octocore Xeons. Thus, gadolinium it is not included in the server_priv/nodes file.
On the other hand I am monitoring all nodes with pbs_mom and Ganglia (head node included).
The problem is that the log messages are full with the entries:
Aug 1 04:17:46 gadolinium PBS_Server: LOG_ERROR::is_request, bad attempt to
connect from 192.168.199.254:1023 (address not trusted - check entry in server_priv/nodes)
The /etc/hosts /etc/ssh/ssh_known_hosts files should be set fine, and if I add gadolinium
to the server_priv/nodes even with a single CPU than the error messages stop.
192.168.199.254 is the IP address of eth0 of gadolinium, eth1 is behind a firewall.
I tried to add gadolinium with the np=0 option, but than torque does not start because it
cannot create a node with zero cores.
Otherwise things are running fine up to now, but there is a new line generated in
the /var/log/messages in every two seconds, which is very disturbing if one would like
to look for high battery temperatures in RAID controllers and other more important
system messages. I have already browsed through the mailing list archives, and it
seems that I am not alone with such problems, however, I could not find any solution
for the issue.
It would be nice of you if you could help me to solve this problem or at least to give me
some hint where to look for a workaround.
Probably the most simple solution would be to stop pbs_mom on
the head node, but than I could not monitor the activity of the head node
online with Ganglia.
More information about the torqueusers