Martin Koch wrote:
In general you can also have even more domains or hostnames A and or SRV pointing for different reasons to your vcs.
Like the additional A record for your VCS. For example a 4pc.vcs.domain.com which is an a record poiting to one of the
vcs just for the polycom systems and then for example have a lower ttl just on that a record so you can quick change it to the other a record if the the first vcs dies. (not sure maybe they support the alternate gatekeeper so you would not need to).
Or you want to control on which cluster peer specific systems register, then you could have different
sets of priorities and or weight for two pairs of srv sets.
The example above a relatively simplistic with just a couple of VCSs in a test cluster. The A records provide Round Robin DNS as a fail back and yes, the Polycoms do use the alternate gatekeeper via H.323 RAS, but if one is down prior to registration, then the DNS A records do thier job.
Actually, what we are looking to do is something similar to what you suggest. We actually are looking to group 2 clusters of VCS-E's (a 4 and a 3) in different geographical locations, but connected via a very high speed core network (approx max 8ms round trip time between each VCS-to-VCS). At the moment they are independent and have several VCS-Cs that register with each. Essentially, I was going to create different DNS zones which referenced the different location such that:
e.g.
location-a.cluster1.domain.com
location-b.cluster1.domain.com
location-c.cluster1.domain.com
location-d.cluster1.domain.com
location-e.cluster2.domain.com
location-f.cluster2.domain.com
location-g.cluster2.domain.com
Each zone would have the relevant SRV and A records pointing to all the expressways, but with different weighting and priorities when referencing each of the individual peers
This would then mean that I could still logically get users to user their "local" DNS zone so registrations would (in the first instance) take place with their most local peer, but then fail over to other member of the local cluster - and further finally to the other peers in the other cluster.
Further, each root domain used for placing actual calls will be completely different such that:
e.g.
organisation1.com
organisation2.com
organisation2.com
....
Each of these DNS zones will have similar SRV and records pointing again to the individual peers (after you have now put me right about the fact that you can't do recursive SRV lookups - shame because this would have taken away a chunk of work), and weighted according to locality.
Many thanks Martin for the great feedback.
