A Session switches the Management Unit on its Uprocs via a HDP or modifying the Management Unit of the succesor Uproc, like for example: {A } or {S } or SPECIFIC_MU and is working fine switching the execution Node as pointed by the Management Unit Dependencies of the MUs A* with S* and viceversa.
After an unknown event or modification occurred at a specific date, the session stops execution in one of the nodes and does not continue to execute in the node where the Depending Management Unit points.
How to find out what the problem is and to fix it?
Release : 6.x
Component : DOLLAR UNIVERSE
Missing objects: on one case it was due to missing MU Dependencies on the involved Management Units on both Nodes, on another case it could be due to missing uprocs from the Session on both nodes.
In order to find out what the problem could be, the first thing to do is consulting the Audit Trail of both Dollar Universe Nodes, looking for any modifications performed around the day that the Session stopped working as expected, filtering on Object Type:
MU, MU Dependence, Session
In this case, we would find that a necessary MU Dependency had been removed that day, so recreate the necessary MU Depencies on both nodes to fix the problem, but a modification on the Session or MU could also produce a similar problem.
As a golden rule, please always verify that Object Definitions ( Sessions, Uprocs, MU Dependencies, MUs ) are coherent on both Nodes.
The same could occur in a multinode session if the session was somehow modified in one of the nodes and not in the other.
For example, in this case, the uproc A launched in MU A should trigger two son Uprocs B and C in MU MUB but these two uprocs are not executed at all.
In the Job Log of Uproc A we would find these errors:
ERR GADATPRO :
Procedure unknown in this catalogue
ERR GADATPRO :
Procedure unknown in this catalogue
The solution consists on deploying again the Session from the reference node to both nodes and click in Expand to make sure that all Objects are coherent in both Nodes.