Brian, after reading and reading and reading, I am going to investigate this
patch more. Amy's comment in DERBY-1905
As described more in the discussions for DERBY-2130, I removed an
"if-block" from OptimizerImpl that actually causes an infinite loop in
lang/innerjoin.sql (I hacked around that before running derbyall). So that has
to be resolved.
Is interesting and I am wondering if this was not hit somehow. After testing
with "derby.optimizer.noTimeout=true" I have satisfied to myself anyways, that
it is not a bad plan or the computation of all plans that is the issue but
rather some sort of looping, retrying, etc.
I need to investigate more on this "jumping" and the patch and my stack traces
and the code.
Thank you for pointing this issue out!
Brett
-----Original Message-----
From: Bryan Pendleton [mailto:[email protected]]
Sent: Friday, March 14, 2014 10:10 AM
To: [email protected]
Subject: Re: I need help diagnosing a Derby engine problem where threads are
not progressing
>>> My assumption is that there is a possibility that the optimization
>>> may get into a state where it may take a very long time to complete
>>> or may even be looping.
>>> So what I would like to explore is under what conditions could the
>>> optimizer get into a state where it might loop or make no progress
>>> for a very long period of time.
Are you able to try the patch for DERBY-2130 in your environment?
https://issues.apache.org/jira/browse/DERBY-2130
At the time that I was wrestling with this problem (years ago, sorry), I was
experiencing great instability in optimization times.
Sometimes it would run in a few seconds.
Sometimes it would take 15 minutes.
If you can provide any feedback on whether the DERBY-2130 change alters the
behaviors that you see, that would be great!
bryan
