We are also seeing this. Approximately 10 transactions in 60,000. Of the reasons listed:
1 Teradata Database restart
2 Transient network failure
3 Faulty network hardware
1 Is not the case since it was a performance test in the frist place and has never not happened during such.
2 and 3 seem unlikely since the net in question is designed for performance testing
What I would like to ask is if misconfigured connection pooling could cause this issue? If a pooled connection went bad and the system then tried to re-use that bad connection for example, could something like this happen?
What we've got going on is a Container (Tomcat) managed JNDI connection pool. Problem is that MyBatis also has it's own built-in connection pool that may not be playing well with our arrangement. When the app first went in for performance testing it was discovered that we effectively had no pooling despite having set the configuration in Tomcat's JNDI configuration settings. I have a new configuration that works according to our profiler but still hasn't been tested by the performance team.
↧