Neophyte
If it were me, i would prefer to group by the value of an instance level property. It's simply simpler. But that would require discovering the instance level property in the first place. As I understand your issue, the peer group isn't something that's discoverable via SNMP. (If it is discoverable via SNMP, then we just need to augment the existing discovery to discover the peer group for each instance and group them that way.) Since it's not discoverable via SNMP, regex grouping is a logical choice. However, if you're not discovering each peer's peer group, then what is the regex going to evaluate against? How are you going to identify which peers go in which group?
As far as the regex itself, you simply need to specify the group name and the corresponding regex. If an instance matches the regex, it will be in that instance group, like this.
To give you an example, I built a Santa Claus tracker just before the holidays last year. There are instances for each of the 9 reindeer and also for the big man himself. Each instance has an instance level property (auto.species). To group by species, I can simply choose "instance level property" in the group by and provide "auto.species".
To use regex, I simply provide this in the regex field:
Elves="S.*" Reindeer="[DPVCBR].*"
That ends up creating two groups like this:
