in my experience all these session leaks have usually been caused by
bugs in user code - such as using a singleton to store some state.
however, once you can give us a test case that reproduces this we will
be happy to fix it.

-igor

On Wed, Dec 8, 2010 at 9:44 PM, Jason Lea <[email protected]> wrote:
> Hi,
>
> We have wicket 1.4.12 on our site at the moment, and over the last few weeks
> we had occasional reports of users seeing pages that belonged to another
> user (eg when logged in, the username on the page was for another user).
>
> The panel that displays the username was using a property model which was
> holding the Session object - after seeing some discussion on the list we
> know this is bad and will change it.  But in our case memory wasn't blowing
> up, and it did have the username in the session.  If we used the current
> session users might not have noticed the bug.
>
> We made a change to our base page to try and catch the problem.  When a page
> is created with a session we stored the session id and username in the page
> as fields.  Then onBeforeRender we check the current session's id against
> the id stored in the page - when they didn't match we would log the problem
> and redirect to an error page so the user does not see another user's data.
>
> From this we got 5 cases over a day.  Most of the cases occurred when user A
> had logged out between 10 secs to 1 minute, user B would see a page
> belonging to user A.  In some cases the user A had not logged out, but a new
> page had been instantiated in the same time frame.  My guess here would be
> it has to do with evicting pages when the pagemap is full - logging out
> removes everything, creating pages will remove pages when there is no more
> space.
>
> 2 days ago we checked where our disk page store files were stored and it had
> about 500Mb of files there (which is consistent with our settings).  We
> deleted files that were more than 2 days old - which reduced this to 40Mb.
>  We didn't get any new cases occurring, but will keep looking.
>
> We can manage the situation by leaving this code to redirect on error and
> monitor the number of cases we get.  We will look at going to Wicket 1.4.14
> soon.
>
> Is there some place we could/should add logging (DiskPageStore?) that might
> detect this problem earlier?
>
> --
> Jason Lea
>
>
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: [email protected]
> For additional commands, e-mail: [email protected]
>
>

---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to