Hi Aaron,
All of the VS's always on a single LTM/TMM instance that is either 9.4.8 or 10.1 (depending on the datacenter)
The multiple VS's are just to allow multiple externalized ports access to the same internal pool of servers. It would be fantastic if you can confirm that incrementing access to the global var is serialized for multiple VS's incrementing the same global counter
On the CMP...All of our LTM HA implementations involved active/passive pairs of LTMs...either 6400's or Viprions. So only one is ever active, and the other is a warm standby. It sounds like this is for some level of shared processing by both nodes in the LTM pair, but since we aren't doing that its probably moot. We have used global vars for sometime, just wanted to check that if multiple VS's are incrementing the same global var on the same TMM instance, that it wont produce a race condition or another anomaly.
Curious...assuming all in bound connections/requests are serialized into single consecutive set of requests for multiple VS's using one IP, but different ports, how does LTM handle breaking the "tie" if two connections/requests arrive at literally the same time on two of the VS's?
Thanks
M