Hi Piotr,
the "session table" replication is controlled by just "StateMirror.MirrorSessions" settings and is completely independent of the "network connection" replication of your individual virtual servers / SNATs.
"network connection" mirroring for VS or SNAT can be turned on or off depending on your needs. Once enabled it would replicate some sort of "stateful connection tables" between the devices to allow a silent failover of already established connections. Without "network connection" mirroring, a reestablishment of the connections would be required to continue.
Regarding: I was looking for fast and reliable way to exchange table data between separate devices/VE.
At least im not aware of a fast and reliable way to do this. The things I'm aware of are slow and/or unreliable (aka. using sidebands in combination with webservice calls (slow but reliable) or even data exports/imports (fast but unreliable for write access due to a duplication of information which may causes race condition to ocour)
You should see the performance of table queries like this...
- TMM1 reads data which is held on TMM1 (its very fast without pausing the TCL execution)
- TMM2 reads data which is held on TMM2 (its very fast without pausing the TCL execution)
- TMM1 reads data which is held on TMM2 (its alredy somewhat slow and pauses the TCL execution on TMM1 while waiting for TMM2 to respond)
- TMM1 on device UNIT1 reads data over a sideband from device UNIT2, using TMM2 of UNIT2 as entrypoint to finally access data on TMM1 of UNIT2. (its slow like hell, since you need to call accros the network and maybe jump between TMMs on the destination UNIT at the same time)
Regarding: SOL13478
This article is referencing to "network connection" mirroring (TCP, UDP, SNAT) and not to "session table" mirroring (persistence records, table information). I haven't seen any problems so far with
[table]
information, when moving the traffic groups between the devices nor rebooting a certain device. Its just keeps replicating the contained information...
Cheers, Kai