HA latency.

hi,

cluster members test connectivity to the master unit by sending 'hello packets'. These are repeated quite often per second. IIRC the interval is either 200 ms or 20 ms. If a low number of consecutive packets are lost the slave deducts that the master has failed and fails over.

So, latency on the HA link is no direct problem. I would be much more concerned about packet loss. You can adjust the HA timeout and/or counts in the CLI to make the setup more tolerant but that will mean the failover is delayed even in case it is needed.

Besides, depending on your HA setup, quite a lot of data is exchanged/synced between cluster members (master to slave, that is). The link should be able to sustain that without loss. Otherwise the syncing will never stop which may mean a) incomplete synchronization of configuration and session data and b) instable HA cluster behavior.