bmurray at snf.stanford.edu
Wed Mar 13 00:50:35 PST 2002
John and Mike,
It turns out that resetting the JDBC connection without restarting the
servers is extremely difficult in our application. However, as I was
attempting to make this change, I found a bug which certainly could be
causing the problem we've been seeing. It's been in my persistence layer
since I wrote it in 1999. It just takes a really heavy load to expose
the problem (and impatient users that start multiple Coral sessions quickly).
I believe I have fixed the bug, but there are a few other code segments that
I should also check for any related problems. I have installed the fix on
rosen with the debugging output still included. I restarted the servers a
few minutes ago. Let's see if this fixes the problem.
PS Given the complexity of resetting the JDBC connection, we may not
wish to do this if the bug fix resolves the problem. If we do
wish to proceed with this, it might safer just to restart the
PPS If I can hide from Mahnaz until tomorrow afternoon, I would like to
review a couple of other code segments tomorrow morning. My short
1) Confirm that I got the bug.
2) Install new kernel on spf.
3) Finish converting stored procedures to PostgreSQL functions.
4) Set up Mahnaz's new machine.
More information about the coral