bmurray at snf.stanford.edu
Tue Mar 19 15:48:37 PST 2002
On Tue, 19 Mar 2002, John Shott wrote:
> That is great news!!! ... is the condition that you were looking for a case
> where the enterLab: looking up ... and the enterLab: adding current log
> entries were separated by other thread activity?
That's correct. It was very similar to the first bug I fixed. It was
another situation where I was synchronizing on the wrong object, in this
case a PreparedStatement, for a very brief moment instead of synchronizing
on the sql Connection. So it took just the right sequence of events to
cause the problem.
> Also, I notice that we are at a total of something like 6500 transactions ...
> this seems like a lower number than we were seeing previously, but, it
> appears, as if the various heartbeat frequencies haven't changed. Is my
> "mental calibration" scale simply off.
This is considerably lower. I was incorrect in assuming that the problem
was related to the number of transactions. It's really most closely related
to load. Basically we needed a local client to register and call enterLab
while another thread had the database locked. That happened three times
on Friday. Although Friday was a bad day, the logs provided me with enough
information to find a bug. It's definitely a bug, and I'm pretty sure it's
More information about the coral