[Zope] CoreSessionTracking vs. SQLSession

Chris McDonough chrism@digicool.com
Tue, 29 May 2001 12:57:09 -0400


> For a brief period (a day), I had BOTH CoreSessionTracking and
> SQLSession maintaining session information.  At random times, the
> CoreSessionTracking would totally lose ALL session data.  SQLSession did
> not.

Well, one out of two isn't bad.  ;-)

> This is CoreSessionTracking without any timeout set.
> 
> My current understanding of this problem -- it does happen to others
> (not just me), but is very rare.  The problem is not identified, is not
> reproducible by any digicool tests.

To be pedantic, it's not "any digicool tests", it's "any Chris McDonough
tests" (CoreSessionTracking is badly named, I should have thought before
presuming I could smuggle it into the core in a reasonable amount of
time ;-).  But yes.  I wrote several complicated test cases which fire
off a bunch of reader and writer threads and do a funny dance with
aborts and such, and don't see the problem.  I also tried testing
manually, and didn't see any of the symptoms being reported.  
 
> I do have an environment which reproduces the problem within a short
> time, and would be glad to run any version of CoreSessionTracking that
> would help identify the problem.

Thank you.   I think my next step is to instrument CST with debugging
statements that go to the log, and ask someone in the community who is
having these problems to run it for a while and attempt to associate the
debug statements with the problems.  I have no ETA for the
debug-instrumented version.

> Currently I am running with SQLSession -- so far it has not had a single
> problem.

I'm glad you got something to work, I apologize for your problems with
CST.

- C