I need help with the pooling code again

[Berin Loritsch <bl...@apache.org>]

It seems that with the new working (and blocking) lock, that somewhere
in the course of high load operations in the ComponentPool, I get a
deadlock.  It happens on quickly repetitive tasks.

As a sample, I ran Apache JMeter against Cocoon 2 today testing the
ResourceReader which was occasionally dying due to race conditions.
I fixed that by fixing Lock.  However, when I moved into testing
the normal Generate->Transform->Serialize process, the Pool went
into deadlock.

The strange thing is, I tried two settings:

12 threads with 1000 ms delay (1000ms deviation): all is well on my machine
12 threads with 500ms delay (500ms deviation): deadlock

So I tried it again with 6 threads (500ms variation), and again
I got deadlock.  Somewhere the ComponentPool is dying in a manner
that a Lock is not released--but not throwing any exceptions.

If need be I can revert to yesterday's code, but we are so much
better off if we can get this code to work.

---------------------------------------------------------------------
To unsubscribe, e-mail: avalon-dev-unsubscribe@jakarta.apache.org
For additional commands, e-mail: avalon-dev-help@jakarta.apache.org