The host on which we run our Integration Server and its Broker hits a couple of performance spikes each day during which it starts swapping for a period of say 15 minutes. Nothing fails, processes just slow down for that period, then pick up speed again afterwards.
When the slowdown occurs, the Integration Server logs the following error:
2005-10-13 03:15:44 EST [ISS.0098.0041C] Unable to connect to Broker. Starting to poll
Some time later, the IS reconnects to the broker:
2005-10-13 03:16:07 EST [ISS.0098.0042D] Successfully reconnected to Broker. Stopped polling
This wouldn’t a problem if we didn’t then get the following sort of message during the disconnected phase, which affects our processing and can cause downstream problems:
2005-10-13 03:16:02 EST [ISS.0098.0064D] Publishing delayed while outbound store is draining. Service: wm.server.publish:publish
We should be able to avoid all this unpleasantness by increasing the IS’s timeout when talking to the broker. Is this possible?