Deadlocks in Metastorm Database

Now that we've pushed our code to production and it's taking a sufficient beating, we've been seeing a lot of deadlocks in our process. I believe it stems from having Solution Tables with stored procs querying these tables during a flag action.

All the deadlocks are related to the eWait row attempting to run a DELETE while the stored proc for the Solution table is executing. We are moving our action events from the flag into an intermediate system stage to avoid this issue, but was wondering if anyone else has experienced this and whether or not it's written down somewhere about 'best practices'.

...aaron

Find more posts tagged with

BPM

Comments

Nils Wyldbore

Aaron,

Stu White did quite a good document here /awd/forums/questions/3409569 regarding improving performance in v9.

Apart from that, we spend a lot of time reviewing our SQL queries and ensuring that we use locking hints as often as possible (NOLOCK / ROWLOCK etc). After that, the only option is to run a trace on your database and start working on adding appropriate indexes (even on MBPM tables).

I believe there was also an issue recently regarding EFOLDERID fields being of differing types between solution tables and MBPM tables, which will also cause performance degradation. Check that they are of the same type. Sure there is a post on here somewhere too.

Hope that helps?

Nils.

Aaron Van

Plural,

Thanks for the reply. Yea, I've seen the performance document before, but thanks for the link.

We generally have a no NOLOCK policy around here since the issue can usually be solved another way. We've added indexes to the Solution Tables but haven't yet modified any of the internal tables. We moved the stored procs to system stages following the flags and pushed as a hot-fix to PROD so time will tell if that will alleviate the situation.

thanks,

...aaron