Automatic retry of terminated processes

[MariiaSoboleva]

Hello all,

in our company we have faced with the situation, when for Customer it is really critical to lose every minute in the possible processing time of Robot.
How we went to that critical discussion:  the virtual machines on which we are running automated process were unpredictable rebooted. Process was terminated, nobody was informed and some processing time was lost.
I guess, you all know that if BluePrism process is running on virtual machine, and this machine will be restarted during process execution, Process will receive status 'terminated' and QueueItem will be marked as FAIL with reason  'Automatically set at Cleanup'. After Virtual machine was restarted, BluePrism Listener will be available again, but process which was running on this virtual machine needs to be re-started manually again.

Is it some best practices / solutions how to automatically restart terminated process which was not finished because of any 'external' reasons (not general errors into the process)? 
Any best practices/solution how to notify developer in case if process was automatically terminated because of any external reasons (and not general errors into the process)?

Thank you in advance!

Some details about our use- case:

1 process; 4 virtual machines; 1 Queue;
1 process is running in parallel on 4 machines and work with 1 Queue to process more orders.
We don't have pool of available virtual machines, we use all of them and they are depending on Customer. We can not use 'free' virtual machine of another Customer.

------------------------------
Mariia Soboleva
RPA Developer
T-Systems
UTC
------------------------------

[JohnPass]

Hi Mariia,

This situation is one that can be difficult to approach depending on the nature of the failure. Failures that occur outside of the Blue Prism product may require manual intervention to resolve and so automatically restarting a process may actually be more harmful in this circumstance.

There's is an 'on fail' setting in the scheduler which can be used, however, as above there is a potential risk of creating an infinite loop that cannot be stopped without restarting the Blue Prism server. Outside of this, you could create a 'monitoring process' that checks on the session and restarts another if it has failed. This would be the most reliable way to do this, however, the process itself would require the use of a concurrent session (A license).

I hope this helps!

------------------------------
John Pass
Senior Customer Support Analyst
Blue Prism
------------------------------

[MariiaSoboleva]

Hello dear John,

many thanks for you reply. The nature of failure in my case is  outside of the BluePrism. 
I also thought about additional monitoring process, but that will not help in my case, because if virtual machine on which this monitoring process is running will be unpredictable restarted, than process will be terminated like all my previous processes and will not be automatically started again.

BR, Mariia

------------------------------
Mariia Soboleva
RPA Developer
T-Systems
UTC
------------------------------

[NicholasZejdlik]

If you have access to a server outside of your Blue Prism environment, you could periodically send "alive" signals from your process, and have the server notify you if it hasn't received one for a while.

Another alternative would be to write a Windows service and set it to launch when Windows starts up. It could either send a notification to you so you'd know the VM needs to be logged in, or it could initiate the login process via the Blue Prism command line. A script in the user's startup folder could then launch Blue Prism and resume the process.

I've used both of these solutions in the past, and they were successful as far as letting me know when a VM is down. Nowadays, I have an external service that monitors the bots which automatically logs them in and run processes, which saves a lot of time and hassle on issues like this.

------------------------------
Nicholas Zejdlik
RPA Developer
------------------------------

[MariiaSoboleva]

Hello Nicholas,

thank you for your reply. We also build Windows agent which starts via Startup and send 'alive' signs. And as well , we have monitoring outside of BluePrism and working on remote start and stop BP processes from additional platform to control such unpredictable behavior. But all of that, let's say - external solutions.For me it is strange that such a big pioneer of RPA doesn't have a 'common' solution for this problem. I hoped that should be something here, because it is not that untypical.

BR, Mariia

------------------------------
Mariia Soboleva
RPA Developer
T-Systems
UTC
------------------------------