I had today a problem in production. We have an interface through an ICO (sender HTTP Channel, receiver FTP Channel). I have identified that the receiver Channel could have log into the target FTP, but somehow the communication was so slow that it was waiting until the timeout (300 sec) and give up. The problem was, that I had 200 messages for this channel and I have 5 worker tasks for inbound HTTP communication, so after the fifth has been in the wait-situation, no further HTTP process was processed:
Now I had stopped the channel, waited until the 5 stuck messages have reached the timeout, then the backlog was processing.
I already using message prioritization, but if those messages are in same priority, they are processed one by each other (max. 5 in parallel).
What can be done in order to not hold the complete PI System? I think about maybe only 2 of 5 worker processes are allowed for the same communication channel, etc. ? Other ideas?