We have two node 4.2 PAM cluster. We use PAM to create SDM tickets from SC and file trigger for start processing from external system. We tested some stress situations:
- stopped PAM and submitted about 1500 new orders to Service Catalog
- we stopped PAM and put to trigger folder about 3000 files for processing
in both cases after we started PAM again PAM started to process orders in queue or files from trigger folder but PAM always tries to start a lot of processes at once and becomes very slow. PAM processes that are processed in normal situation in a few seconds suddenly last a few minutes or longer or PAM freezes and I have abort these processes from GUI. Is it possible to force PAM to not start so many processes at once and rather process less load at once but faster?
You cannot globally throttle the orchestrators, if you request 3000 instances run all at the same time then Process Automation will do its best to run those 3000 instances as quickly as possible.
You can limit number of runs on a process, or multiple process level by using the Resource object and operators.
See the documentation for more details: Resources
I would also suggest reviewing and tuning your engine to ensure it can process as many instances as possible. Some basic tuning details can be found here: Tune CA Process Automation
Hello Michael, thank you for your answer. We were running mail trigger in PAM 3.1 for a few years, PAM always tried to load only 5-10 emails at once although in mailbox were thousands of emails for processing, PAM never tried to load all emails from the mailbox at once. It took a few hours to process all mailbox but it was expected. I was expecting similar behavior in file trigger in PAM 4.2 or in general in PAM 4.2. I don't need to run 3000 instances at the same time, I just need to safely process the queue of requests. I'm going to do more tests and I'm going to try to use resources as you recommended.
Hello Michael,using of resource object and operators really helped to limit maximum of running instances, thank you. I have some finding from my tests.Releasing of resources is very slow if there is a queue of instances waiting for free resources. During my tests I noticed that in some moments there were no running instances at all because all instances that had allocated resources were also in waiting state. When a free resource is assigned then the instance is processed quickly but at the end the instance is in waiting state and the allocated resource is released only after a few minutes. I attached some screenshots. If there is no queue of instances waiting for free resource and instances are processed continuously resources are released quickly. Is there any explanation for this behavior, is it possible to change this?
As part of PAM 4.3, we have fixed good number of Resource scalability issues and were successfully able to improve the performance by many folds. Could you please try your use case with our latest version?
Do let me know if you need further help !!
I will try to upgrade to latest version.
I realize this post is two years since the last one, but I am running 4.3.3 and using Resources to throttle concurrent workflow activity. The "Take Unit" is working fine, but I found a case (currently happening) where I have a number of Workflows currently stuck at the Resource operator trying to perform a "Free Unit" on the resource, just prior to coming to an End. Subsequently, the Resources are not being freed up which is prohibiting any new Workflows from performing. Any troubleshooting I can do to expalin why the Resource operator to Free the Unit is hanging? I've tried suspending/resetting-operator/resuming the hung workflow, but it still doesn't work. Please advise.