Re: debugging PageExpiredExceptions

[sbrookes2 <se...@shaw.ca>]

ver 1.3.6

I have been tracking this issue for a few months now in our app. Similar to
the following threads:

http://apache-wicket.1842946.n4.nabble.com/PageExpiredException-getting-this-when-the-session-hasn-t-timeout-td1894202.html#a1894216
PageExpiredException - getting this when the session hasn't timeout  
http://apache-wicket.1842946.n4.nabble.com/PageExpiredException-on-ajax-calls-if-apps-pages-are-opened-in-different-tabs-td1884182.html#a1884182
PageExpiredException on ajax calls if apps pages are opened in different
tabs 
http://apache-wicket.1842946.n4.nabble.com/Using-AbstractDefaultAjaxBehavior-on-a-Bookmarkable-page-td1861596.html
Using AbstractDefaultAjaxBehavior on a Bookmarkable page? 

Our design goal is a simple RESTful page navigation with moderate AJAX
interactions on the pages (join/leave groups, rate items, connect to people,
etc).  If a user refreshes the page or clicks the back button we simply
repaint the page using whatever state the spring beans return.  We are using
IndexedParamUrlCodingStrategy and have implemented a relatively simple
navigation infrastructure using simple href/get requests with manually
generated links

eg: http://<server>/web/common/profilehome/userxyz

Where 'common' is the mount name and userxyz is parameter "0".

So far so good.

The pages are stateless and automaticWindowSupport = false (we can't have
the additional parameters on our URL)

I have been seeing these exceptions for a while :
- "Cannot find the rendered page in session"
- "The target page does not exist anymore"

I thought the issue was limited to a cached page not matching the current
state so I implemented a temporary hack in order to improve the user
experience (it was getting very common) by intercepting the PEE and
redirecting to the current request URL (i.e. trigger a page refresh) which
really calmed things down from a user experience perspective but the
messages kept coming in.

It turns out that isn't the only source of these messages as it seems to be
related to using AJAX behaviors on stateless pages causing the pagemap to
either lose or remove references to certain pages so that when an AJAX
request comes in the requesting URL is no longer valid (if
http://apache-wicket.1842946.n4.nabble.com/Using-AbstractDefaultAjaxBehavior-on-a-Bookmarkable-page-td1861596.html 
I understand the previous thread )

I see a few possible paths forward and would like to get some feedback from
the community on them and to see if there is additional information you can
share to help me out here.

1) I looked at the technique in this blog post: 
http://day-to-day-stuff.blogspot.com/2008/10/wicket-extreme-consistent-urls.html
Wicket extreme consistent URLs  and it looks like it may be a possibility
but unfortunately we are using indexed parameters and the technique is based
on named parameters so it would mean quite a bit of work to rework our
navigation implementation so I am reluctant to venture to far down that path
until it is either confirmed by someone here or we run out of other options

2) Is there some setting or pagemap configuration that I don't know about
that would either keep the pages in the store or 'turn off' the page map
altogether?

3) Convert our AJAX calls from wicket based to a more generic 'servlet'
style so that we would construct the AJAX target URL ourselves to remove the
dependency on the native wicket url and hopefully bypass the pagemap
reference altogether.  Would this 'solve' the problem?  It would also be
quite a lot of work but we have been moving in this direction for a while
now anyway in order to get more control over our AJAX for performance
reasons so if this sounds like a viable way to avoid the PEEs then at least
we would be able to commit to it and get on with fixing it.

Wicket has been very good to us for a couple of years but now that I am
trying to implement a 'real' fix I am concerned I may have made a big
architecture mistake given the heavy emphasis on state with Wicket.

Hopefully there is a magic bullet out there someone can share.

Thanks,
Sean
-- 
View this message in context: http://apache-wicket.1842946.n4.nabble.com/debugging-PageExpiredExceptions-tp1882040p2274494.html
Sent from the Wicket - User mailing list archive at Nabble.com.

---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@wicket.apache.org
For additional commands, e-mail: users-help@wicket.apache.org

Re: debugging PageExpiredExceptions

[sbrookes2 <se...@shaw.ca>]

Thanks for the reply Martijn.

I have done some poking around and I think this thread holds the key: 
http://apache-wicket.1842946.n4.nabble.com/PageExpiredException-ajax-request-td1878866.html#a1878866
PageExpiredException - ajax request 

It looks like our 'stateless' page gets rendered with AJAX links in it that
get saved to the pagemap.  If I open another browser window and interact
with the application for a period of time it seems the AJAX link references
in the initial tab are removed from the page map so if I click on any of
them the app returns a PEE.

So it seems the answer is to detect the PEE and then regenerate a fresh
version of the page.  I am still working on that part but at least I have
learned some more about areas of the framework I hadn't spent enough time
with to this point.

Some questions around ajax, stateless/statefull pages and page mounts.
(please forgive me if any of these reflect my ignorance).

- if I have a stateless page(mounted page with an empty/PageParameter
constructor) but add elements/panels with AJAX behaviors does it become
statefull?  Confusingly isStateless() always returns true even though the
page seems to be referenced in the pagemap. I'm still trying to fully
understand the relationship between stateless and statefull in Wicket.

- should either of the following have any effect?
  getRequestCycleSettings().setBufferResponse(false);
 
getRequestCycleSettings().setRenderStrategy(IRequestCycleSettings.ONE_PASS_RENDER);

- setting setAutomaticMultiWindowSupport to true adds the following to new
urls:   "/wicket:pageMapName/wicket-3/".  Is there any way to prevent this
as it mangles the URL effect I am trying to implement?






-- 
View this message in context: http://apache-wicket.1842946.n4.nabble.com/debugging-PageExpiredExceptions-tp1882040p2277017.html
Sent from the Wicket - User mailing list archive at Nabble.com.

---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@wicket.apache.org
For additional commands, e-mail: users-help@wicket.apache.org

Re: debugging PageExpiredExceptions

[Martijn Dashorst <ma...@gmail.com>]

On Thu, Jul 1, 2010 at 2:30 AM, sbrookes2 <se...@shaw.ca> wrote:
> It turns out that isn't the only source of these messages as it seems to be
> related to using AJAX behaviors on stateless pages causing the pagemap to
> either lose or remove references to certain pages so that when an AJAX
> request comes in the requesting URL is no longer valid (if
> http://apache-wicket.1842946.n4.nabble.com/Using-AbstractDefaultAjaxBehavior-on-a-Bookmarkable-page-td1861596.html
> I understand the previous thread )

AFAIK wicket ajax is not supported on stateless pages (roll your own
solutions might very well work though).

Martijn

---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@wicket.apache.org
For additional commands, e-mail: users-help@wicket.apache.org