I can “host pbsserver1.example.com” and “host pbsserver2.example.com”.
But I cannot *host pbsserver1".
I think I can write FQDN in PBS_SERVER and PBS_PRIMARY, PBS_SECONDARY, because man pbs.conf said:
Hostname of host running the server.
If the short name of the server host resolves to the
correct IP address, you can use the short name for the value of the
PBS_SERVER entry in pbs.conf. If only the FQDN of the server host
resolves to the correct IP address, you must use the FQDN for the
value of PBS_SERVER.
On the other hand, we found that the PBS startup script (/opt/pbs/libexec/init.d/pbs_init.d) converts the FQDN set in PBS_SERVER to a short name and checks if the name resolution can be done with the short name.
My questions are:
Is the current behavior a glitch that does not match the specification? Or is it working according to specifications and there are no problems?
Is there any workaround to start PBS in this environment ?
Not mandatory.
Every hostname resolves to only one IP address and /etc/hosts is correctly populated and is same across the pbs complex i.e., name resolution should be consistent across the entire pbs complex/cluster and the IP resolves to the same canonical name across all the compute nodes/ headnode that runs PBS Pro services or daemons.
Recommendation: The canonical name for each address (first on the line in /etc/hosts) is usually fully qualified (with domain name); the output of “hostname” is usually the short name.