HA issues with "TIBCO RV Adapter for TIBCO FTL"
I'm executing the following scenario:
Sender RV (custom) -> "TIBCO RV Adapter for TIBCO FTL" -> "FTL Cluster" (3 nodes, same machine for testing purposes) -> Receiver FTL (custom)
I'm logging everything I send and receive to be able to verify integrity. Sender is sending ~3mb/s using ~1024 bytes packages. I am using "Shared Memory" transport for the endpoints and "Auto" for communication among clusters/stores.
1) If nothing is restarted in this scenario, integrity check is OK.
2) If I keep restarting "Receiver FTL" and handle duplicated packages properly, integrity check is OK. Durable does its job.
3) If I keep restarting cluster elements to test HA using these steps:
- Kill cluster element 1, wait 4 seconds, start cluster element 1, wait 15 seconds, kill cluster element 2, etc.
I can see little pauses (a backup node takes leadership, clients reconnect), and everything keeps working. However, my integrity check FAILS, because I lose about 13% of the packages in "Receiver FTL" during several points of the transmission.
4) If instead of "publisher_settings": "store_send_noconfirm" I switch to "publisher_settings": "store_confirm_send", integrity check is OK, however in this configuration I experience severe performance issues (I'm unable to keep receiver rate at 3mb/s in a very powerful physical server).
So my questions are:
a) Is this behavior OK? 100% of delivered packages is only possible in "store_confirm_send"?
b) Is it expected to not be able to handle 3mb/s in "store_confirm_send" or I'm configuring something wrong?
I can provide more information about the scenario if necessary.
Thank you in advance.