tcp_acknak.pdl vs mlp_tcp.pdl thoughts, recomendations, etc.

William.

Many years ago we moved from mlp_tcp to tcp_cknak to communicate between Cloverleaf sites on the same server.

We found after multiple power outages during a *really* bad day that messages were ‘lost’ between sites.  On investigation, we proved that killing a site communicating with mlp_tcp would lose the message  – we believed that the message was in the TCP buffer and not in the Cloverleaf recovery database.

We used the same test scripts with the tcp_acknak PDL and could not reproduce the same message loss.

Note that we DID NOT use message acking between the Cloverlewf sites.

I hope this helps

PS

Note that the tcp_acknak PDL as its name indicates actually waits not an ACK (or NAK) from the receiving system.

Just so I am clear here, when you are refering to “between sites” does that indicate communication between Cloverleaf and foreign systems, or other Cloverleaf engines?

Mark.

We have about 70 production sites, with one engine per site and the ‘between sites’ is the comunication between these.

We actually have thread suffixes ‘_out’ and ‘_in’ for  threads that communicate between sites and ‘_snd’ and ‘_rcv’ for external communication

Thanks for the clarification.  The reason I was asking is that we are having a tcp communication issue with a UNIX host, and I am exploring any possible options.  Is the tcp ack/nak used with external systems?

Mark.

I would not recommend tcp_acknak between Cloverleaf and applications ….

We communicate with iSoft’s Clinical Management system and they use an encapsulated version of  the tcp_acknak code and half-duplex communications for the HL7 message and ACKS.

I have provided details in another thread.

We have recently found that messages are lost when there are database locks , so full-duplex would have been far better!!

FYI We also found that message were lost when MSMQ aborted!!