Modification to the design of PBS_SERVER_INSTANCES switch


I have modified the following design which we floated some time back for Multi-Server, where in we are saying PBS Server instance is a short host name / FQDN and also this value should not be derived from PBS_SERER conf variable. Please take a look at the following.

I think your last sentence is not necessary and could actually be confusing. PBS_SERVER was ALWAYS the name of the cluster, not a server instance. For example, in a traditional failover setup, the two instances of pbs servers have different names depicted by pbs_primary and pbs_secondary, yet the PBS_SERVER value is the same (demonstrating the fact that that switch always stood as a representation of the entire cluster, and not a server instance). There are some sites who configure a non-host name as PBS_SERVER when they have a failover setup as well.

Besides, in the case of a single server configuration (in mutli-server), the code could deduce that the internal value of PBS_SERVER_INSTANCES (if not specified explicitly) is based on PBS_SERVER - no?

Thanks, @subhasisb for the detailed explanation. I will remove the last sentence which is confusing. Yes, as you said PBS_SERVER may be a non-hostname in case of failover.

Yes in the case of a single server configuration(in Multi-server), we are deducing the internal value of PBS_SERVER_INSTANCES if not specified using PBS_SERVER only.
I have removed the last line in design doc. Please check.

1 Like