We have multiple AF databases on our test server, which different users can modify.
Last week, PI Analysis Processor service on the same server stopped working (and as you may have guessed on Friday evening!). I tried the following actions:
1- restarting the Analysis service
2- restarting AF Application service, then 1
3- Changing startup type for the Analysis service to manual, then restarting the server itself, then 1
4- Emptying the recalculation queue in "RecalculationProcessor.RequestedRecalculations.xml", then 3
Nothing worked, so I opened a case. After spending a fair bit of time on the phone, I came to my senses and started to investigate each database individually (by disabling all of the analyses and re-enabling them) to see which one of the databases is causing the Analysis service to crash.
I contacted the database owner, rolled back the last changes he had made and restarted the service. Everything started to work again!
Here comes the fun question: Is there an elegant way to look out for troublesome analyses i.e. resource consumption, time taken to execute each analysis or something else? We are lucky to have less than 10 databases and 9000 analyses. What if there are one million analyses in a thousand databases that need to be monitored?
I appreciate any help, and wish you all incident-free weekends...