I hope this document is a help to someone. We were having problems
with our Dynamo servers grinding to a halt with over 40 threads per
CPU or DRP. In addition thread deaths were out of control. I was
unable to find a centralized summary of our particular problem and
wrote up this document which contains the sum of my findings. The root
cause was traced back to the DB connection pool and the overuse of
blocking=true.
Let me know if this helps you out in any way or if you find errors or
additions that can be made.
This document will describe the uses of parameters used to control the
Dynamo database pool. Using these parameters in conjunction should be
able to handle the issue of a slow database, a database down, a
database that went down and came back up leaving invalid connections
in the pool, a table lock, heavy traffic on the application or a slow
down in the network. They have been tested in a production environment
against Oracle, Sybase and MS SQL-Server.
http://job.wowak.com/dynamodbpool.html
Thanks
brad johnson
http://wowak.com