Multi-Instance flows hang on Clustered Central environment. - 10.51
Have a case open with HP, but figured I might ask here as well.
A very simple flow that utilizes the multi-instance function works fine if run locally through studio or if run against a central server that is not clustered.
This issue only occurs when running with a clustered setup (2xCentral, MSSQL Backend, 4xRAS)
From what I can tell it is reproducible with any multi-instance step. It's not limited to our particular flow.
The flow just hangs and remains that way forever until I manually shut down central and clean out the database.
If anyone from OO DEV team is listening, case # is: 4653722447
We have the same issue but the occurences are not predictable, nor reproducable. We are on 10.50 in Central and HP thinks the problem occurs when a RAS crashes and comunication with Central is broken. We're on 10.50 and HP claims the problem is remedied in 10.51 but it looks like you're already there with no joy.
I've had a cluster environment running on 10.51 and ran a some multi-instance steps and encountered no issues. I would require more details from you on this issue:
- how many instances were you running?
- when the flows are stuck in running did you check your db's table for oo_triggers to see the state of the split_join_trigger? If yes was this in blocked mode?
- the MI steps were run on RAS workers or the central workers?
- with the flows stuck in running did canceling them stop them? In 10.51 the cancel was improved in order to be able to stop stuck executions.
Now strictly speaking without knowing the exact flows you are using i think that perhaps you are opening more multi-instances than the number of threads that are available for the worker which may cause the issues you are describing. If this is the case you can either impose a threshold on the muti instance step to limit the number of multi instances that are active at a time or using the OO tunning guide increase the number of threads available for the workers.
Hope this helps,