After deploying objects from a v5 node to a v6 node, Supervisor Performances are much worse.
Jobs with a file resource dependency in a NFS file will go to Time Overrun as the supervisor is not able to process the whole u_frrv60 file in the 30 seconds allowed by its cycle.
Enabling the trace 0,SURCYCLE we can see that only 50 Files can be checked every 30 seconds.
The following kind of Warning appears in the universe.log:
|WARN |X|IO |pid=p.t| o_module_sur_cycle | Interrupting supervisor job condition checks because the time limit (120 s) was exceeded. The job runs conditions will be check less often than specified until further notice.
Release : 6.x
OS: AIX performances are by far worse than Linux / Windows.
Component : DOLLAR UNIVERSE
Supervisor on version 6 takes too much time to process resources checks during its cycle, specially on AIX machines.
Workaround: manually bypass jobs in Event Wait, try to reduce the number of concurrent jobs in Event Wait waiting for File Resources.
Update to a fix version listed below or a newer version if available.
Dollar Universe 6.10.21 - Released 15th January 2020Dollar Universe 6.8.43 - Available only for AIX as a hotfix
Since version 6.10.21 the supervisor has a multi-thread processing which allows to process up to 20x times more files per second.