IBM MQ - Best Practice

Hi All,

 

I am reconfiguring our IBM MQ setup in our organization and have done some testing with our MQ guy. One of the interesting behaviours is in the below scenario that I don't believe I can resolve with F5 magic but thought I would ask.

 

The scenario is that there is 2 x MQ servers in a pool and the team want to perform maintenance on each device individually without causing impact to the clients connecting.

 

1. Server is disabled, traffic is drained etc but due to a long life connection there are still connections.
2. Server is patched/rebooted and with the action-service-down, set a reselect it does nothing.
3. Client attempts to send data, but doesn't TCP handshake as that was torn down in step 2 ( right ? ).
4. Client errors out as it fails to get a response.

I have tried changing the action on service down to reject, which resets the TCP connection on the client end but the issue with this is that the application itself seems to not to know that the TCP connection is gone and just fails when trying to send data.

 

So to make this work, I have to basically have the connection from Client to F5 never drop, yet still support long lived connections through the if the server at the backend goes down the client doesn't need to complete a 3 way handshake again to load balance to a new connection.

 

Any ideas ? Configuration below.

 

ltm virtual VS-mqB { destination 10.49.x.x:ibm-mqseries ip-protocol tcp mask 255.255.255.255 pool Pool-mq profiles { tcp-idle-360 { } } source 0.0.0.0/0 source-address-translation { pool SNAT-Pool-VS type snat } translate-address enabled translate-port enabled }

 

ltm pool Pool-mq { load-balancing-mode least-connections-member members { 10.x.x.14:ibm-mqseries { address 10.x.x.14 } 10.x.x.15:ibm-mqseries { address 10.x.x.15 } } monitor tcp service-down-action reselect }