01-07-2014 09:27 PM
We have a cluster of 7 Itanium blade servers running OpenVMS 8.3-1H1.
We observed local lock conversion of around 10500 at one time and another around 340000 on the active node. using the T4 data.
I believe that the rates are site specific but what are the "normal" lock rates (lock conversions, ENQ, and DEQs)? What values should we be vigilant?
01-08-2014 07:54 AM
There is no majic number at which the rate would be considered excessive. It all depends upon your environment. If you are seeing transaction delays coincide with these high locking rates, perhaps a review of the application and how locks are used would be in order.
Hope this helps a bit,,,
01-08-2014 07:58 AM
10,000 is nothing to worry about.
100,000 you want to make sure that it is inat reasonable relation to the work performed.
FOr example, when processing an aggregate 20,000 records/sec in an share RMS file it would be fine.
340,000 probably means you are maxed out/bottlenecked.
Details investigation is in order.
Use ANALYZE/SYSTEM extension to see the top locks.
What is the main database used? Main access? Version?
01-08-2014 04:41 PM
Thanks for the replies.
The cluster is using RDB V7.2.
I looked at the T4 graphs for the other days and I can see spikes of more than 100,000 local conversions but not consistent to specific times.
We were trying to relate the reported slow response to ENQs and DEQs but the graphs show otherwise.