Sorry to be replying to my own post. I have news and a new question.
It looks like the timeouts mentioned below are being caused by a piece
of code we wrote a LONG time ago to enforce a limit on number of
simultaneous sessions with the same username. Our monitoring system
"logs in" with the same username as I was using interactively. After we
wrote this function, we immediately wrote (for our system) a bug,
because it didn't work. Now it works (with in-memory sessions, both
Memory and Dynamic).
New question, similar to the old question: Are there some semantic
differences in how in-memory sessions are handled vs. file-based
David Hancock | dhancock@... | 410-266-4384=20
> -----Original Message-----
> From: Hancock, David (DHANCOCK) =20
> Sent: Tuesday, March 15, 2005 9:08 PM
> To: webware-discuss@...
> Subject: Expiration problem after switching to Dynamic sessions
> We've been seeing the occasional KeyError from the SessionStore code
> on our production website. We use 'File'-based sessions. We originally
> started using File because we kept session data on an NFS-mounted
> directory so we could load-balance WebKit over several machines. That
> turned out not to be necessary, and we're using one machine now. (We
> had a lot more session issues using NFS, also.)
> Over the last couple days, we've seen an uptick in the number of
> session KeyErrors, so I thought I'd try switching the Session store to
> Dynamic. Now what's happening is that some sessions are expiring long
> before their timeout, within a minute or two sometimes. If I switch
> back to File, it stops happening.
> Has anybody else seen this behavior? Any ideas for me to use in
> Redhat 7.3, Python 2.3, Webware 0.8.
> Thanks in advance for any help you can give. This is not an immediate
> showstopper, because we can use File-based sessions, but I would like
> to trust Memory and Dynamic sessions, too. I did notice a slight
> "snappiness" improvement of pages using Memory sessions.
> David Hancock | dhancock@... | 410-266-4384=20