no, but you are welcome to test and ping us back with the results :)

-igor

On Mon, Dec 13, 2010 at 8:14 PM, Michal Kurtak <michal.kur...@gmail.com> wrote:
> And one more thing. do you have some benchmarks comparing
> DiskDataStore, AsynchronousDataStore to Pagemaps stored in http
> sessions (as in earlier wicket versions)? I mean writing pages to
> disks comparing to writing them to http session (and replicating big
> sessions)?
>
> 2010/12/13 Michal Kurtak <michal.kur...@gmail.com>:
>> Hmm... This solutions seems really good. sorry for bothering
>>
>> michal
>>
>>
>> 2010/12/13 Igor Vaynberg <igor.vaynb...@gmail.com>:
>>> when the page that is stored in http session is deserialized on
>>> another node it is also spooled to disk. so in a cluster all nodes
>>> have identical spooled pages - so everything should work.
>>>
>>> -igor
>>>
>>> On Mon, Dec 13, 2010 at 8:00 PM, Michal Kurtak <michal.kur...@gmail.com> 
>>> wrote:
>>>> Thank you for your replies Igor.
>>>>
>>>> I have another question regarding to clustered environment. Suppose we
>>>> have versioned pages opened in two windows. If one node goes down,
>>>> only the last page user made request from is stored in http session
>>>> and replicated to other nodes. So i suppose that only one page would
>>>> work and the state from the second page is lost.
>>>>
>>>> So multi-window support for stateful pages (even versioned) wouldn't
>>>> work in clustered environment. Is this true?
>>>>
>>>> I think that even weird javascript solution for multiple windows would
>>>> be good (it could be turned off by default), because we are unable to
>>>> migrate our existing wicket projects, or create stateful application
>>>> in multiple windows/tabs running in the cluster. This was supported
>>>> out-of-box by wicket previous versions. I know that there is a
>>>> possibility to provide implementation IPageManager, but there is
>>>> missing support for browser window detection so its a bit more
>>>> complicated to implement IPageManager supporting multiple windows in
>>>> clustered environment.
>>>>
>>>> BR,
>>>> Michal
>>>>
>>>>
>>>> 2010/12/13 Igor Vaynberg <igor.vaynb...@gmail.com>:
>>>>> On Sat, Dec 11, 2010 at 12:04 PM, Michal Kurtak <michal.kur...@gmail.com> 
>>>>> wrote:
>>>>>> Hi folks,
>>>>>>
>>>>>> I know that there has been a lot of written about pagestores and
>>>>>> multi-window support in wicket 1.5, but i have several other
>>>>>> questions:
>>>>>>
>>>>>> 1. Is multi-window supported for non-versioned pages?
>>>>>>
>>>>>> If page is versioned everything works ok, but i always get
>>>>>> StalePageExceptions when i use multiple windows/tabs with
>>>>>> non-versioned pages.
>>>>>> I think that StalePageException is useful in some cases when you wanna
>>>>>> be sure that user has same page in all windows, but in older versions
>>>>>> of wicket it was possible to have non versioned pages and multi-window
>>>>>> support together (e.g. we have an application written in wicket 1.2.6
>>>>>> that uses non-versioned pages in multiple windows)
>>>>>
>>>>> no, it is not supported. supporting it has been problematic even in
>>>>> 1.4.x series where it required javascript and introduced a strange
>>>>> redirect that happened as soon as the page was loaded. unversioned
>>>>> pages also cause weird issues with the back button. if you want
>>>>> something that is unversioned then i think building it as stateless
>>>>> makes more sense.
>>>>>
>>>>>> 2. How page stores work in clustered environment?
>>>>>>
>>>>>> In older versions of wicket we used pagemaps stored in HttpSession.
>>>>>> When one node in cluster refuses to handle request, it was
>>>>>> successfully handled by another nod, because HttpSession was
>>>>>> replicated to another node. We have used non-versioned pages (no back
>>>>>> button support needed) and in one pagemap there was max 5 pages.
>>>>>> Wicket provided this functionality out-of-box.
>>>>>
>>>>> in both 1.4.x and 1.5.x by default we store the current page in
>>>>> session and the rest are spooled to disk. the current page is
>>>>> replicated to all nodes since it is stored in session, so any node can
>>>>> pick up a request.
>>>>>
>>>>>>
>>>>>> 3. How to achieve this in wicket 1.5?
>>>>>>
>>>>>> I have found PersistentPageManager which uses IPageStore to store
>>>>>> pages. Pages are stored on disk by default, but there is also thread
>>>>>> safe SerializedPagesCache with SoftReferences. Theres no out-of-box
>>>>>> solution to store pages in HttpSession.
>>>>>>
>>>>>> I have also found file page-management.txt in org.apache.wicket.page
>>>>>> package. It contains proposal of other pagamanagers and multi-window
>>>>>> support for non-versioned pages.
>>>>>>
>>>>>> The proposal contains classes
>>>>>>
>>>>>> PersistentPageManager with DiskPageStore and
>>>>>> SecondLevelCacheSessionStore for versioned pages
>>>>>> SessionManager for holding non versioned pages in HttpSession.
>>>>>>
>>>>>> 4. Can we expect that this proposal will be implemented in wicket 1.5
>>>>>> final version?
>>>>>
>>>>> no. the current version is feature complete and we are concentrating
>>>>> on bugfixing. a more sophisticated page management may be introduced
>>>>> in the future. we have concentrated on the 99% usecase and simplified
>>>>> the code for that. you can implement your own page storage strategies
>>>>> as you see fit.
>>>>>
>>>>> -igor
>>>>>
>>>>>>
>>>>>> Thanks for your replies.
>>>>>>
>>>>>> BR,
>>>>>> Michal Kurtak
>>>>>>
>>>>>> ---------------------------------------------------------------------
>>>>>> To unsubscribe, e-mail: users-unsubscr...@wicket.apache.org
>>>>>> For additional commands, e-mail: users-h...@wicket.apache.org
>>>>>>
>>>>>>
>>>>>
>>>>> ---------------------------------------------------------------------
>>>>> To unsubscribe, e-mail: users-unsubscr...@wicket.apache.org
>>>>> For additional commands, e-mail: users-h...@wicket.apache.org
>>>>>
>>>>>
>>>>
>>>> ---------------------------------------------------------------------
>>>> To unsubscribe, e-mail: users-unsubscr...@wicket.apache.org
>>>> For additional commands, e-mail: users-h...@wicket.apache.org
>>>>
>>>>
>>>
>>> ---------------------------------------------------------------------
>>> To unsubscribe, e-mail: users-unsubscr...@wicket.apache.org
>>> For additional commands, e-mail: users-h...@wicket.apache.org
>>>
>>>
>>
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: users-unsubscr...@wicket.apache.org
> For additional commands, e-mail: users-h...@wicket.apache.org
>
>

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

Reply via email to