A multinode session does not complete fine between two nodes, the switch of node does not occur when it should according to the session definition.
The following error appear in the second node:
|ERROR|X|IO |pid=p.t| o_module_exc_cycle_recept | Error in RE Cycle of exchanger: 5
|ERROR|X|IO |pid=p.t| o_module_exc_cycle_recept | Error in RE Cycle of exchanger: 600
Wrong design of the session, it was not coherent between both nodes.
This could be checked by analyzing the uxshw ses output from both nodes:
Release : 6.x
Component : DOLLAR UNIVERSE
In order to fix the issue, the following needs to be done:
1. Deploy the correct session to the other node/s so that it has the same definition everywhere and cancel all related Job runs.
2. Stop both nodes and reinitialize the exchanger data files of the area impacted to avoid the exchanger errors in universe.log, ie. for area EXP: