portal issue

classic Classic list List threaded Threaded
9 messages Options
Reply | Threaded
Open this post in threaded view
|

portal issue

Mike Duijvelaar
Hello,

Has anyone seen the next problem:

If you run a high loadtest in portal, that sometimes porlets are mingled. The titel of the page is e.g. portlet A(psml), but content is from portlet B(psml).

Met vriendelijke groet,
Mike Duijvelaar
 
ANWB Online Beheer
 
Gebouw B midden; B330
Wassenaarseweg 220
Den Haag
 
Tel: (intern) toestel 7325/7301
Mob: 06-50233023
**********************************************************************************************
Disclaimer

E-mail wordt door ANWB niet gebruikt voor het aangaan van externe verplichtingen.
Deze e-mail is uitsluitend bestemd voor geadresseerde(n). Indien deze e-mail
onverhoopt niet voor u is bestemd dan verzoeken wij u vriendelijk contact op te
nemen met de afzender en daarna het bericht te vernietigen. Deze e-mail mag niet
worden doorgestuurd, openbaar gemaakt of verveelvoudigd worden zonder de
toestemming van de afzender.
ANWB betracht grote zorgvuldigheid bij het verzenden van e-mails. ANWB kan
echter niet garanderen dat deze e-mail juist, volledig, tijdig en virusvrij wordt
overgebracht. In een dergelijk geval is ANWB op geen enkele wijze aansprakelijk
voor enige schade, direct dan wel indirect, in welke vorm dan ook.

ANWB B.V.
**********************************************************************************************

_______________________________________________
Hippoportal-dev mailing list
[hidden email]
http://lists.hippo.nl/mailman/listinfo/hippoportal-dev
Reply | Threaded
Open this post in threaded view
|

Re: portal issue

Nick_S
A little clarification about what we are seeing.

We have psml A with some portlets and psml B with some portlets. It
seems that, under load, sometimes a request which retrieves page A
gets the content page A with one portlet not on page A but on page B
and the portlet that is supposed to be there, isn't there. It even
seems that this content is from another request.

This is just a wild guess, but it seems that the portlet handling
intermingles outputstreams between requests.

Is this a known issue?

Nick Stolwijk
~Java Developer~

Iprofs BV.
Claus Sluterweg 125
2012 WS Haarlem
www.iprofs.nl



2009/6/26 Mike Duijvelaar <[hidden email]>:

> Hello,
>
> Has anyone seen the next problem:
>
> If you run a high loadtest in portal, that sometimes porlets are mingled. The titel of the page is e.g. portlet A(psml), but content is from portlet B(psml).
>
> Met vriendelijke groet,
> Mike Duijvelaar
>
> ANWB Online Beheer
>
> Gebouw B midden; B330
> Wassenaarseweg 220
> Den Haag
>
> Tel: (intern) toestel 7325/7301
> Mob: 06-50233023
> **********************************************************************************************
> Disclaimer
>
> E-mail wordt door ANWB niet gebruikt voor het aangaan van externe verplichtingen.
> Deze e-mail is uitsluitend bestemd voor geadresseerde(n). Indien deze e-mail
> onverhoopt niet voor u is bestemd dan verzoeken wij u vriendelijk contact op te
> nemen met de afzender en daarna het bericht te vernietigen. Deze e-mail mag niet
> worden doorgestuurd, openbaar gemaakt of verveelvoudigd worden zonder de
> toestemming van de afzender.
> ANWB betracht grote zorgvuldigheid bij het verzenden van e-mails. ANWB kan
> echter niet garanderen dat deze e-mail juist, volledig, tijdig en virusvrij wordt
> overgebracht. In een dergelijk geval is ANWB op geen enkele wijze aansprakelijk
> voor enige schade, direct dan wel indirect, in welke vorm dan ook.
>
> ANWB B.V.
> **********************************************************************************************
>
> _______________________________________________
> Hippoportal-dev mailing list
> [hidden email]
> http://lists.hippo.nl/mailman/listinfo/hippoportal-dev
>
_______________________________________________
Hippoportal-dev mailing list
[hidden email]
http://lists.hippo.nl/mailman/listinfo/hippoportal-dev
Reply | Threaded
Open this post in threaded view
|

Re: portal issue

Ate Douma-3
Nick Stolwijk wrote:

> A little clarification about what we are seeing.
>
> We have psml A with some portlets and psml B with some portlets. It
> seems that, under load, sometimes a request which retrieves page A
> gets the content page A with one portlet not on page A but on page B
> and the portlet that is supposed to be there, isn't there. It even
> seems that this content is from another request.
>
> This is just a wild guess, but it seems that the portlet handling
> intermingles outputstreams between requests.
>
> Is this a known issue?

No, not really, except possibly when the following (incorrect) configuration is used:
- if portletA (on Page A) has the same fragment id as portlet B (on Page B)
- portlet caching is enabled for one or both these portlets (e.g. content cache managed by the *portal*)

If the above is the case, the portal might deliver the content for the portlet directly from its internal cache which is keyed off the
fragment id.
Thus: if you have enabled portlet caching you should *not* reuse its fragment id across psml pages.

Regards,

Ate

>
> Nick Stolwijk
> ~Java Developer~
>
> Iprofs BV.
> Claus Sluterweg 125
> 2012 WS Haarlem
> www.iprofs.nl
>
>
>
> 2009/6/26 Mike Duijvelaar <[hidden email]>:
>> Hello,
>>
>> Has anyone seen the next problem:
>>
>> If you run a high loadtest in portal, that sometimes porlets are mingled. The titel of the page is e.g. portlet A(psml), but content is from portlet B(psml).
>>
>> Met vriendelijke groet,
>> Mike Duijvelaar
>>
>> ANWB Online Beheer
>>
>> Gebouw B midden; B330
>> Wassenaarseweg 220
>> Den Haag
>>
>> Tel: (intern) toestel 7325/7301
>> Mob: 06-50233023
>> **********************************************************************************************
>> Disclaimer
>>
>> E-mail wordt door ANWB niet gebruikt voor het aangaan van externe verplichtingen.
>> Deze e-mail is uitsluitend bestemd voor geadresseerde(n). Indien deze e-mail
>> onverhoopt niet voor u is bestemd dan verzoeken wij u vriendelijk contact op te
>> nemen met de afzender en daarna het bericht te vernietigen. Deze e-mail mag niet
>> worden doorgestuurd, openbaar gemaakt of verveelvoudigd worden zonder de
>> toestemming van de afzender.
>> ANWB betracht grote zorgvuldigheid bij het verzenden van e-mails. ANWB kan
>> echter niet garanderen dat deze e-mail juist, volledig, tijdig en virusvrij wordt
>> overgebracht. In een dergelijk geval is ANWB op geen enkele wijze aansprakelijk
>> voor enige schade, direct dan wel indirect, in welke vorm dan ook.
>>
>> ANWB B.V.
>> **********************************************************************************************
>>
>> _______________________________________________
>> Hippoportal-dev mailing list
>> [hidden email]
>> http://lists.hippo.nl/mailman/listinfo/hippoportal-dev
>>
> _______________________________________________
> Hippoportal-dev mailing list
> [hidden email]
> http://lists.hippo.nl/mailman/listinfo/hippoportal-dev


--
[hidden email]  www.onehippo.com

Amsterdam     - Hippo B.V. Oosteinde 11 1017 WT Amsterdam +31(0)20-5224466
San Francisco - Hippo USA Inc. 101 H Street, suite Q Petaluma CA 94952-5100 +1(877) 41-HIPPO

_______________________________________________
Hippoportal-dev mailing list
[hidden email]
http://lists.hippo.nl/mailman/listinfo/hippoportal-dev
Reply | Threaded
Open this post in threaded view
|

Re: portal issue

Wouter Zelle
Hi Ate,

Portlet caching is off for all of these portlets (<expiration-cache>0</expiration-cache>).

Regards,

Wouter


On Mon, Jun 29, 2009 at 11:45, Ate Douma <[hidden email]> wrote:
Nick Stolwijk wrote:
A little clarification about what we are seeing.

We have psml A with some portlets and psml B with some portlets. It
seems that, under load, sometimes a request which retrieves page A
gets the content page A with one portlet not on page A but on page B
and the portlet that is supposed to be there, isn't there. It even
seems that this content is from another request.

This is just a wild guess, but it seems that the portlet handling
intermingles outputstreams between requests.

Is this a known issue?

No, not really, except possibly when the following (incorrect) configuration is used:
- if portletA (on Page A) has the same fragment id as portlet B (on Page B)
- portlet caching is enabled for one or both these portlets (e.g. content cache managed by the *portal*)

If the above is the case, the portal might deliver the content for the portlet directly from its internal cache which is keyed off the fragment id.
Thus: if you have enabled portlet caching you should *not* reuse its fragment id across psml pages.

Regards,

Ate



Nick Stolwijk
~Java Developer~

Iprofs BV.
Claus Sluterweg 125
2012 WS Haarlem
www.iprofs.nl



2009/6/26 Mike Duijvelaar <[hidden email]>:
Hello,

Has anyone seen the next problem:

If you run a high loadtest in portal, that sometimes porlets are mingled. The titel of the page is e.g. portlet A(psml), but content is from portlet B(psml).

Met vriendelijke groet,
Mike Duijvelaar

ANWB Online Beheer

Gebouw B midden; B330
Wassenaarseweg 220
Den Haag

Tel: (intern) toestel 7325/7301
Mob: 06-50233023
**********************************************************************************************
Disclaimer

E-mail wordt door ANWB niet gebruikt voor het aangaan van externe verplichtingen.
Deze e-mail is uitsluitend bestemd voor geadresseerde(n). Indien deze e-mail
onverhoopt niet voor u is bestemd dan verzoeken wij u vriendelijk contact op te
nemen met de afzender en daarna het bericht te vernietigen. Deze e-mail mag niet
worden doorgestuurd, openbaar gemaakt of verveelvoudigd worden zonder de
toestemming van de afzender.
ANWB betracht grote zorgvuldigheid bij het verzenden van e-mails. ANWB kan
echter niet garanderen dat deze e-mail juist, volledig, tijdig en virusvrij wordt
overgebracht. In een dergelijk geval is ANWB op geen enkele wijze aansprakelijk
voor enige schade, direct dan wel indirect, in welke vorm dan ook.

ANWB B.V.
**********************************************************************************************

_______________________________________________
Hippoportal-dev mailing list
[hidden email]
http://lists.hippo.nl/mailman/listinfo/hippoportal-dev

_______________________________________________
Hippoportal-dev mailing list
[hidden email]
http://lists.hippo.nl/mailman/listinfo/hippoportal-dev


--
[hidden email]  www.onehippo.com

Amsterdam     - Hippo B.V. Oosteinde 11 1017 WT Amsterdam +31(0)20-5224466
San Francisco - Hippo USA Inc. 101 H Street, suite Q Petaluma CA 94952-5100 +1(877) 41-HIPPO


_______________________________________________
Hippoportal-dev mailing list
[hidden email]
http://lists.hippo.nl/mailman/listinfo/hippoportal-dev


_______________________________________________
Hippoportal-dev mailing list
[hidden email]
http://lists.hippo.nl/mailman/listinfo/hippoportal-dev
Reply | Threaded
Open this post in threaded view
|

Re: portal issue

Nick_S
In reply to this post by Ate Douma-3
We have the convention for fragment id's like
<projectname>-<page-name>-<page-part> so no double fragment id's (I
have just double checked them). Also the portlet cache is disabled,
afaik.

We'll keep investigating the issue. If someone has any additional
tips, they're welcome.

With regards,

Nick Stolwijk
~Java Developer~

Iprofs BV.
Claus Sluterweg 125
2012 WS Haarlem
www.iprofs.nl



On Mon, Jun 29, 2009 at 11:45 AM, Ate Douma<[hidden email]> wrote:

> Nick Stolwijk wrote:
>>
>> A little clarification about what we are seeing.
>>
>> We have psml A with some portlets and psml B with some portlets. It
>> seems that, under load, sometimes a request which retrieves page A
>> gets the content page A with one portlet not on page A but on page B
>> and the portlet that is supposed to be there, isn't there. It even
>> seems that this content is from another request.
>>
>> This is just a wild guess, but it seems that the portlet handling
>> intermingles outputstreams between requests.
>>
>> Is this a known issue?
>
> No, not really, except possibly when the following (incorrect) configuration
> is used:
> - if portletA (on Page A) has the same fragment id as portlet B (on Page B)
> - portlet caching is enabled for one or both these portlets (e.g. content
> cache managed by the *portal*)
>
> If the above is the case, the portal might deliver the content for the
> portlet directly from its internal cache which is keyed off the fragment id.
> Thus: if you have enabled portlet caching you should *not* reuse its
> fragment id across psml pages.
>
> Regards,
>
> Ate
>
>>
>> Nick Stolwijk
>> ~Java Developer~
>>
>> Iprofs BV.
>> Claus Sluterweg 125
>> 2012 WS Haarlem
>> www.iprofs.nl
>>
>>
>>
>> 2009/6/26 Mike Duijvelaar <[hidden email]>:
>>>
>>> Hello,
>>>
>>> Has anyone seen the next problem:
>>>
>>> If you run a high loadtest in portal, that sometimes porlets are mingled.
>>> The titel of the page is e.g. portlet A(psml), but content is from portlet
>>> B(psml).
>>>
>>> Met vriendelijke groet,
>>> Mike Duijvelaar
>>>
>>> ANWB Online Beheer
>>>
>>> Gebouw B midden; B330
>>> Wassenaarseweg 220
>>> Den Haag
>>>
>>> Tel: (intern) toestel 7325/7301
>>> Mob: 06-50233023
>>>
>>> **********************************************************************************************
>>> Disclaimer
>>>
>>> E-mail wordt door ANWB niet gebruikt voor het aangaan van externe
>>> verplichtingen.
>>> Deze e-mail is uitsluitend bestemd voor geadresseerde(n). Indien deze
>>> e-mail
>>> onverhoopt niet voor u is bestemd dan verzoeken wij u vriendelijk contact
>>> op te
>>> nemen met de afzender en daarna het bericht te vernietigen. Deze e-mail
>>> mag niet
>>> worden doorgestuurd, openbaar gemaakt of verveelvoudigd worden zonder de
>>> toestemming van de afzender.
>>> ANWB betracht grote zorgvuldigheid bij het verzenden van e-mails. ANWB
>>> kan
>>> echter niet garanderen dat deze e-mail juist, volledig, tijdig en
>>> virusvrij wordt
>>> overgebracht. In een dergelijk geval is ANWB op geen enkele wijze
>>> aansprakelijk
>>> voor enige schade, direct dan wel indirect, in welke vorm dan ook.
>>>
>>> ANWB B.V.
>>>
>>> **********************************************************************************************
>>>
>>> _______________________________________________
>>> Hippoportal-dev mailing list
>>> [hidden email]
>>> http://lists.hippo.nl/mailman/listinfo/hippoportal-dev
>>>
>> _______________________________________________
>> Hippoportal-dev mailing list
>> [hidden email]
>> http://lists.hippo.nl/mailman/listinfo/hippoportal-dev
>
>
> --
> [hidden email]  www.onehippo.com
>
> Amsterdam     - Hippo B.V. Oosteinde 11 1017 WT Amsterdam +31(0)20-5224466
> San Francisco - Hippo USA Inc. 101 H Street, suite Q Petaluma CA 94952-5100
> +1(877) 41-HIPPO
>
> _______________________________________________
> Hippoportal-dev mailing list
> [hidden email]
> http://lists.hippo.nl/mailman/listinfo/hippoportal-dev
>
_______________________________________________
Hippoportal-dev mailing list
[hidden email]
http://lists.hippo.nl/mailman/listinfo/hippoportal-dev
Reply | Threaded
Open this post in threaded view
|

Re: portal issue

Woonsan Ko-3
Hi Nick,

If a portlet class or its ancestor class has member variables to store
some states and the member variables are set on requests, it would
cause a thread-safety problem.

Just two cents,

Woonsan


On Mon, Jun 29, 2009 at 11:56 AM, Nick Stolwijk<[hidden email]> wrote:

> We have the convention for fragment id's like
> <projectname>-<page-name>-<page-part> so no double fragment id's (I
> have just double checked them). Also the portlet cache is disabled,
> afaik.
>
> We'll keep investigating the issue. If someone has any additional
> tips, they're welcome.
>
> With regards,
>
> Nick Stolwijk
> ~Java Developer~
>
> Iprofs BV.
> Claus Sluterweg 125
> 2012 WS Haarlem
> www.iprofs.nl
>
>
>
> On Mon, Jun 29, 2009 at 11:45 AM, Ate Douma<[hidden email]> wrote:
>> Nick Stolwijk wrote:
>>>
>>> A little clarification about what we are seeing.
>>>
>>> We have psml A with some portlets and psml B with some portlets. It
>>> seems that, under load, sometimes a request which retrieves page A
>>> gets the content page A with one portlet not on page A but on page B
>>> and the portlet that is supposed to be there, isn't there. It even
>>> seems that this content is from another request.
>>>
>>> This is just a wild guess, but it seems that the portlet handling
>>> intermingles outputstreams between requests.
>>>
>>> Is this a known issue?
>>
>> No, not really, except possibly when the following (incorrect) configuration
>> is used:
>> - if portletA (on Page A) has the same fragment id as portlet B (on Page B)
>> - portlet caching is enabled for one or both these portlets (e.g. content
>> cache managed by the *portal*)
>>
>> If the above is the case, the portal might deliver the content for the
>> portlet directly from its internal cache which is keyed off the fragment id.
>> Thus: if you have enabled portlet caching you should *not* reuse its
>> fragment id across psml pages.
>>
>> Regards,
>>
>> Ate
>>
>>>
>>> Nick Stolwijk
>>> ~Java Developer~
>>>
>>> Iprofs BV.
>>> Claus Sluterweg 125
>>> 2012 WS Haarlem
>>> www.iprofs.nl
>>>
>>>
>>>
>>> 2009/6/26 Mike Duijvelaar <[hidden email]>:
>>>>
>>>> Hello,
>>>>
>>>> Has anyone seen the next problem:
>>>>
>>>> If you run a high loadtest in portal, that sometimes porlets are mingled.
>>>> The titel of the page is e.g. portlet A(psml), but content is from portlet
>>>> B(psml).
>>>>
>>>> Met vriendelijke groet,
>>>> Mike Duijvelaar
>>>>
>>>> ANWB Online Beheer
>>>>
>>>> Gebouw B midden; B330
>>>> Wassenaarseweg 220
>>>> Den Haag
>>>>
>>>> Tel: (intern) toestel 7325/7301
>>>> Mob: 06-50233023
>>>>
>>>> **********************************************************************************************
>>>> Disclaimer
>>>>
>>>> E-mail wordt door ANWB niet gebruikt voor het aangaan van externe
>>>> verplichtingen.
>>>> Deze e-mail is uitsluitend bestemd voor geadresseerde(n). Indien deze
>>>> e-mail
>>>> onverhoopt niet voor u is bestemd dan verzoeken wij u vriendelijk contact
>>>> op te
>>>> nemen met de afzender en daarna het bericht te vernietigen. Deze e-mail
>>>> mag niet
>>>> worden doorgestuurd, openbaar gemaakt of verveelvoudigd worden zonder de
>>>> toestemming van de afzender.
>>>> ANWB betracht grote zorgvuldigheid bij het verzenden van e-mails. ANWB
>>>> kan
>>>> echter niet garanderen dat deze e-mail juist, volledig, tijdig en
>>>> virusvrij wordt
>>>> overgebracht. In een dergelijk geval is ANWB op geen enkele wijze
>>>> aansprakelijk
>>>> voor enige schade, direct dan wel indirect, in welke vorm dan ook.
>>>>
>>>> ANWB B.V.
>>>>
>>>> **********************************************************************************************
>>>>
>>>> _______________________________________________
>>>> Hippoportal-dev mailing list
>>>> [hidden email]
>>>> http://lists.hippo.nl/mailman/listinfo/hippoportal-dev
>>>>
>>> _______________________________________________
>>> Hippoportal-dev mailing list
>>> [hidden email]
>>> http://lists.hippo.nl/mailman/listinfo/hippoportal-dev
>>
>>
>> --
>> [hidden email]  www.onehippo.com
>>
>> Amsterdam     - Hippo B.V. Oosteinde 11 1017 WT Amsterdam +31(0)20-5224466
>> San Francisco - Hippo USA Inc. 101 H Street, suite Q Petaluma CA 94952-5100
>> +1(877) 41-HIPPO
>>
>> _______________________________________________
>> Hippoportal-dev mailing list
>> [hidden email]
>> http://lists.hippo.nl/mailman/listinfo/hippoportal-dev
>>
> _______________________________________________
> Hippoportal-dev mailing list
> [hidden email]
> http://lists.hippo.nl/mailman/listinfo/hippoportal-dev
>



--
[hidden email]     www.onehippo.com
EUROPE • AMSTERDAM - Hippo B.V. Oosteinde 11 1017 WT Amsterdam +31(0)20-5224466
NORTH AMERICA • SAN FRANCISCO - Hippo USA Inc. 185 H Street, Suite B
Petaluma CA 94952 +1 (877) 414-4776
_______________________________________________
Hippoportal-dev mailing list
[hidden email]
http://lists.hippo.nl/mailman/listinfo/hippoportal-dev
Reply | Threaded
Open this post in threaded view
|

Re: portal issue

Nick_S
That is correct, but our problem is with two completely different
portlets, which normally only exists on two different psml's. But once
in a while, the content of portlet B is shown on page A, instead of
portlet A.

With regards,

Nick Stolwijk
~Java Developer~

Iprofs BV.
Claus Sluterweg 125
2012 WS Haarlem
www.iprofs.nl



On Mon, Jun 29, 2009 at 12:16 PM, Woonsan Ko<[hidden email]> wrote:

> Hi Nick,
>
> If a portlet class or its ancestor class has member variables to store
> some states and the member variables are set on requests, it would
> cause a thread-safety problem.
>
> Just two cents,
>
> Woonsan
>
>
> On Mon, Jun 29, 2009 at 11:56 AM, Nick Stolwijk<[hidden email]> wrote:
>> We have the convention for fragment id's like
>> <projectname>-<page-name>-<page-part> so no double fragment id's (I
>> have just double checked them). Also the portlet cache is disabled,
>> afaik.
>>
>> We'll keep investigating the issue. If someone has any additional
>> tips, they're welcome.
>>
>> With regards,
>>
>> Nick Stolwijk
>> ~Java Developer~
>>
>> Iprofs BV.
>> Claus Sluterweg 125
>> 2012 WS Haarlem
>> www.iprofs.nl
>>
>>
>>
>> On Mon, Jun 29, 2009 at 11:45 AM, Ate Douma<[hidden email]> wrote:
>>> Nick Stolwijk wrote:
>>>>
>>>> A little clarification about what we are seeing.
>>>>
>>>> We have psml A with some portlets and psml B with some portlets. It
>>>> seems that, under load, sometimes a request which retrieves page A
>>>> gets the content page A with one portlet not on page A but on page B
>>>> and the portlet that is supposed to be there, isn't there. It even
>>>> seems that this content is from another request.
>>>>
>>>> This is just a wild guess, but it seems that the portlet handling
>>>> intermingles outputstreams between requests.
>>>>
>>>> Is this a known issue?
>>>
>>> No, not really, except possibly when the following (incorrect) configuration
>>> is used:
>>> - if portletA (on Page A) has the same fragment id as portlet B (on Page B)
>>> - portlet caching is enabled for one or both these portlets (e.g. content
>>> cache managed by the *portal*)
>>>
>>> If the above is the case, the portal might deliver the content for the
>>> portlet directly from its internal cache which is keyed off the fragment id.
>>> Thus: if you have enabled portlet caching you should *not* reuse its
>>> fragment id across psml pages.
>>>
>>> Regards,
>>>
>>> Ate
>>>
>>>>
>>>> Nick Stolwijk
>>>> ~Java Developer~
>>>>
>>>> Iprofs BV.
>>>> Claus Sluterweg 125
>>>> 2012 WS Haarlem
>>>> www.iprofs.nl
>>>>
>>>>
>>>>
>>>> 2009/6/26 Mike Duijvelaar <[hidden email]>:
>>>>>
>>>>> Hello,
>>>>>
>>>>> Has anyone seen the next problem:
>>>>>
>>>>> If you run a high loadtest in portal, that sometimes porlets are mingled.
>>>>> The titel of the page is e.g. portlet A(psml), but content is from portlet
>>>>> B(psml).
>>>>>
>>>>> Met vriendelijke groet,
>>>>> Mike Duijvelaar
>>>>>
>>>>> ANWB Online Beheer
>>>>>
>>>>> Gebouw B midden; B330
>>>>> Wassenaarseweg 220
>>>>> Den Haag
>>>>>
>>>>> Tel: (intern) toestel 7325/7301
>>>>> Mob: 06-50233023
>>>>>
>>>>> **********************************************************************************************
>>>>> Disclaimer
>>>>>
>>>>> E-mail wordt door ANWB niet gebruikt voor het aangaan van externe
>>>>> verplichtingen.
>>>>> Deze e-mail is uitsluitend bestemd voor geadresseerde(n). Indien deze
>>>>> e-mail
>>>>> onverhoopt niet voor u is bestemd dan verzoeken wij u vriendelijk contact
>>>>> op te
>>>>> nemen met de afzender en daarna het bericht te vernietigen. Deze e-mail
>>>>> mag niet
>>>>> worden doorgestuurd, openbaar gemaakt of verveelvoudigd worden zonder de
>>>>> toestemming van de afzender.
>>>>> ANWB betracht grote zorgvuldigheid bij het verzenden van e-mails. ANWB
>>>>> kan
>>>>> echter niet garanderen dat deze e-mail juist, volledig, tijdig en
>>>>> virusvrij wordt
>>>>> overgebracht. In een dergelijk geval is ANWB op geen enkele wijze
>>>>> aansprakelijk
>>>>> voor enige schade, direct dan wel indirect, in welke vorm dan ook.
>>>>>
>>>>> ANWB B.V.
>>>>>
>>>>> **********************************************************************************************
>>>>>
>>>>> _______________________________________________
>>>>> Hippoportal-dev mailing list
>>>>> [hidden email]
>>>>> http://lists.hippo.nl/mailman/listinfo/hippoportal-dev
>>>>>
>>>> _______________________________________________
>>>> Hippoportal-dev mailing list
>>>> [hidden email]
>>>> http://lists.hippo.nl/mailman/listinfo/hippoportal-dev
>>>
>>>
>>> --
>>> [hidden email]  www.onehippo.com
>>>
>>> Amsterdam     - Hippo B.V. Oosteinde 11 1017 WT Amsterdam +31(0)20-5224466
>>> San Francisco - Hippo USA Inc. 101 H Street, suite Q Petaluma CA 94952-5100
>>> +1(877) 41-HIPPO
>>>
>>> _______________________________________________
>>> Hippoportal-dev mailing list
>>> [hidden email]
>>> http://lists.hippo.nl/mailman/listinfo/hippoportal-dev
>>>
>> _______________________________________________
>> Hippoportal-dev mailing list
>> [hidden email]
>> http://lists.hippo.nl/mailman/listinfo/hippoportal-dev
>>
>
>
>
> --
> [hidden email]     www.onehippo.com
> EUROPE • AMSTERDAM - Hippo B.V. Oosteinde 11 1017 WT Amsterdam +31(0)20-5224466
> NORTH AMERICA • SAN FRANCISCO - Hippo USA Inc. 185 H Street, Suite B
> Petaluma CA 94952 +1 (877) 414-4776
> _______________________________________________
> Hippoportal-dev mailing list
> [hidden email]
> http://lists.hippo.nl/mailman/listinfo/hippoportal-dev
>
_______________________________________________
Hippoportal-dev mailing list
[hidden email]
http://lists.hippo.nl/mailman/listinfo/hippoportal-dev
Reply | Threaded
Open this post in threaded view
|

Re: portal issue

Nick_S
It looks like we have found our problem. It was in one of our
extensions to the hippo portal, which saved the request context in a
member variable. And this caused the wrong PSML to be loaded and thus
the wrong portlets being rendered.

With regards,

Nick Stolwijk
~Java Developer~

Iprofs BV.
Claus Sluterweg 125
2012 WS Haarlem
www.iprofs.nl



On Mon, Jun 29, 2009 at 12:46 PM, Nick Stolwijk<[hidden email]> wrote:

> That is correct, but our problem is with two completely different
> portlets, which normally only exists on two different psml's. But once
> in a while, the content of portlet B is shown on page A, instead of
> portlet A.
>
> With regards,
>
> Nick Stolwijk
> ~Java Developer~
>
> Iprofs BV.
> Claus Sluterweg 125
> 2012 WS Haarlem
> www.iprofs.nl
>
>
>
> On Mon, Jun 29, 2009 at 12:16 PM, Woonsan Ko<[hidden email]> wrote:
>> Hi Nick,
>>
>> If a portlet class or its ancestor class has member variables to store
>> some states and the member variables are set on requests, it would
>> cause a thread-safety problem.
>>
>> Just two cents,
>>
>> Woonsan
>>
>>
>> On Mon, Jun 29, 2009 at 11:56 AM, Nick Stolwijk<[hidden email]> wrote:
>>> We have the convention for fragment id's like
>>> <projectname>-<page-name>-<page-part> so no double fragment id's (I
>>> have just double checked them). Also the portlet cache is disabled,
>>> afaik.
>>>
>>> We'll keep investigating the issue. If someone has any additional
>>> tips, they're welcome.
>>>
>>> With regards,
>>>
>>> Nick Stolwijk
>>> ~Java Developer~
>>>
>>> Iprofs BV.
>>> Claus Sluterweg 125
>>> 2012 WS Haarlem
>>> www.iprofs.nl
>>>
>>>
>>>
>>> On Mon, Jun 29, 2009 at 11:45 AM, Ate Douma<[hidden email]> wrote:
>>>> Nick Stolwijk wrote:
>>>>>
>>>>> A little clarification about what we are seeing.
>>>>>
>>>>> We have psml A with some portlets and psml B with some portlets. It
>>>>> seems that, under load, sometimes a request which retrieves page A
>>>>> gets the content page A with one portlet not on page A but on page B
>>>>> and the portlet that is supposed to be there, isn't there. It even
>>>>> seems that this content is from another request.
>>>>>
>>>>> This is just a wild guess, but it seems that the portlet handling
>>>>> intermingles outputstreams between requests.
>>>>>
>>>>> Is this a known issue?
>>>>
>>>> No, not really, except possibly when the following (incorrect) configuration
>>>> is used:
>>>> - if portletA (on Page A) has the same fragment id as portlet B (on Page B)
>>>> - portlet caching is enabled for one or both these portlets (e.g. content
>>>> cache managed by the *portal*)
>>>>
>>>> If the above is the case, the portal might deliver the content for the
>>>> portlet directly from its internal cache which is keyed off the fragment id.
>>>> Thus: if you have enabled portlet caching you should *not* reuse its
>>>> fragment id across psml pages.
>>>>
>>>> Regards,
>>>>
>>>> Ate
>>>>
>>>>>
>>>>> Nick Stolwijk
>>>>> ~Java Developer~
>>>>>
>>>>> Iprofs BV.
>>>>> Claus Sluterweg 125
>>>>> 2012 WS Haarlem
>>>>> www.iprofs.nl
>>>>>
>>>>>
>>>>>
>>>>> 2009/6/26 Mike Duijvelaar <[hidden email]>:
>>>>>>
>>>>>> Hello,
>>>>>>
>>>>>> Has anyone seen the next problem:
>>>>>>
>>>>>> If you run a high loadtest in portal, that sometimes porlets are mingled.
>>>>>> The titel of the page is e.g. portlet A(psml), but content is from portlet
>>>>>> B(psml).
>>>>>>
>>>>>> Met vriendelijke groet,
>>>>>> Mike Duijvelaar
>>>>>>
>>>>>> ANWB Online Beheer
>>>>>>
>>>>>> Gebouw B midden; B330
>>>>>> Wassenaarseweg 220
>>>>>> Den Haag
>>>>>>
>>>>>> Tel: (intern) toestel 7325/7301
>>>>>> Mob: 06-50233023
>>>>>>
>>>>>> **********************************************************************************************
>>>>>> Disclaimer
>>>>>>
>>>>>> E-mail wordt door ANWB niet gebruikt voor het aangaan van externe
>>>>>> verplichtingen.
>>>>>> Deze e-mail is uitsluitend bestemd voor geadresseerde(n). Indien deze
>>>>>> e-mail
>>>>>> onverhoopt niet voor u is bestemd dan verzoeken wij u vriendelijk contact
>>>>>> op te
>>>>>> nemen met de afzender en daarna het bericht te vernietigen. Deze e-mail
>>>>>> mag niet
>>>>>> worden doorgestuurd, openbaar gemaakt of verveelvoudigd worden zonder de
>>>>>> toestemming van de afzender.
>>>>>> ANWB betracht grote zorgvuldigheid bij het verzenden van e-mails. ANWB
>>>>>> kan
>>>>>> echter niet garanderen dat deze e-mail juist, volledig, tijdig en
>>>>>> virusvrij wordt
>>>>>> overgebracht. In een dergelijk geval is ANWB op geen enkele wijze
>>>>>> aansprakelijk
>>>>>> voor enige schade, direct dan wel indirect, in welke vorm dan ook.
>>>>>>
>>>>>> ANWB B.V.
>>>>>>
>>>>>> **********************************************************************************************
>>>>>>
>>>>>> _______________________________________________
>>>>>> Hippoportal-dev mailing list
>>>>>> [hidden email]
>>>>>> http://lists.hippo.nl/mailman/listinfo/hippoportal-dev
>>>>>>
>>>>> _______________________________________________
>>>>> Hippoportal-dev mailing list
>>>>> [hidden email]
>>>>> http://lists.hippo.nl/mailman/listinfo/hippoportal-dev
>>>>
>>>>
>>>> --
>>>> [hidden email]  www.onehippo.com
>>>>
>>>> Amsterdam     - Hippo B.V. Oosteinde 11 1017 WT Amsterdam +31(0)20-5224466
>>>> San Francisco - Hippo USA Inc. 101 H Street, suite Q Petaluma CA 94952-5100
>>>> +1(877) 41-HIPPO
>>>>
>>>> _______________________________________________
>>>> Hippoportal-dev mailing list
>>>> [hidden email]
>>>> http://lists.hippo.nl/mailman/listinfo/hippoportal-dev
>>>>
>>> _______________________________________________
>>> Hippoportal-dev mailing list
>>> [hidden email]
>>> http://lists.hippo.nl/mailman/listinfo/hippoportal-dev
>>>
>>
>>
>>
>> --
>> [hidden email]     www.onehippo.com
>> EUROPE • AMSTERDAM - Hippo B.V. Oosteinde 11 1017 WT Amsterdam +31(0)20-5224466
>> NORTH AMERICA • SAN FRANCISCO - Hippo USA Inc. 185 H Street, Suite B
>> Petaluma CA 94952 +1 (877) 414-4776
>> _______________________________________________
>> Hippoportal-dev mailing list
>> [hidden email]
>> http://lists.hippo.nl/mailman/listinfo/hippoportal-dev
>>
>
_______________________________________________
Hippoportal-dev mailing list
[hidden email]
http://lists.hippo.nl/mailman/listinfo/hippoportal-dev
Reply | Threaded
Open this post in threaded view
|

Re: portal issue

Ate Douma-3
Nick Stolwijk wrote:
> It looks like we have found our problem. It was in one of our
> extensions to the hippo portal, which saved the request context in a
> member variable. And this caused the wrong PSML to be loaded and thus
> the wrong portlets being rendered.
Ok, thanks for reporting back Nick.

We'll regard this issue then as resolved and out-of-scope/unrelated to Hippo Portal.

Regards,

Ate

>
> With regards,
>
> Nick Stolwijk
> ~Java Developer~
>
> Iprofs BV.
> Claus Sluterweg 125
> 2012 WS Haarlem
> www.iprofs.nl
>
>
>
> On Mon, Jun 29, 2009 at 12:46 PM, Nick Stolwijk<[hidden email]> wrote:
>> That is correct, but our problem is with two completely different
>> portlets, which normally only exists on two different psml's. But once
>> in a while, the content of portlet B is shown on page A, instead of
>> portlet A.
>>
>> With regards,
>>
>> Nick Stolwijk
>> ~Java Developer~
>>
>> Iprofs BV.
>> Claus Sluterweg 125
>> 2012 WS Haarlem
>> www.iprofs.nl
>>
>>
>>
>> On Mon, Jun 29, 2009 at 12:16 PM, Woonsan Ko<[hidden email]> wrote:
>>> Hi Nick,
>>>
>>> If a portlet class or its ancestor class has member variables to store
>>> some states and the member variables are set on requests, it would
>>> cause a thread-safety problem.
>>>
>>> Just two cents,
>>>
>>> Woonsan
>>>
>>>
>>> On Mon, Jun 29, 2009 at 11:56 AM, Nick Stolwijk<[hidden email]> wrote:
>>>> We have the convention for fragment id's like
>>>> <projectname>-<page-name>-<page-part> so no double fragment id's (I
>>>> have just double checked them). Also the portlet cache is disabled,
>>>> afaik.
>>>>
>>>> We'll keep investigating the issue. If someone has any additional
>>>> tips, they're welcome.
>>>>
>>>> With regards,
>>>>
>>>> Nick Stolwijk
>>>> ~Java Developer~
>>>>
>>>> Iprofs BV.
>>>> Claus Sluterweg 125
>>>> 2012 WS Haarlem
>>>> www.iprofs.nl
>>>>
>>>>
>>>>
>>>> On Mon, Jun 29, 2009 at 11:45 AM, Ate Douma<[hidden email]> wrote:
>>>>> Nick Stolwijk wrote:
>>>>>> A little clarification about what we are seeing.
>>>>>>
>>>>>> We have psml A with some portlets and psml B with some portlets. It
>>>>>> seems that, under load, sometimes a request which retrieves page A
>>>>>> gets the content page A with one portlet not on page A but on page B
>>>>>> and the portlet that is supposed to be there, isn't there. It even
>>>>>> seems that this content is from another request.
>>>>>>
>>>>>> This is just a wild guess, but it seems that the portlet handling
>>>>>> intermingles outputstreams between requests.
>>>>>>
>>>>>> Is this a known issue?
>>>>> No, not really, except possibly when the following (incorrect) configuration
>>>>> is used:
>>>>> - if portletA (on Page A) has the same fragment id as portlet B (on Page B)
>>>>> - portlet caching is enabled for one or both these portlets (e.g. content
>>>>> cache managed by the *portal*)
>>>>>
>>>>> If the above is the case, the portal might deliver the content for the
>>>>> portlet directly from its internal cache which is keyed off the fragment id.
>>>>> Thus: if you have enabled portlet caching you should *not* reuse its
>>>>> fragment id across psml pages.
>>>>>
>>>>> Regards,
>>>>>
>>>>> Ate
>>>>>
>>>>>> Nick Stolwijk
>>>>>> ~Java Developer~
>>>>>>
>>>>>> Iprofs BV.
>>>>>> Claus Sluterweg 125
>>>>>> 2012 WS Haarlem
>>>>>> www.iprofs.nl
>>>>>>
>>>>>>
>>>>>>
>>>>>> 2009/6/26 Mike Duijvelaar <[hidden email]>:
>>>>>>> Hello,
>>>>>>>
>>>>>>> Has anyone seen the next problem:
>>>>>>>
>>>>>>> If you run a high loadtest in portal, that sometimes porlets are mingled.
>>>>>>> The titel of the page is e.g. portlet A(psml), but content is from portlet
>>>>>>> B(psml).
>>>>>>>
>>>>>>> Met vriendelijke groet,
>>>>>>> Mike Duijvelaar
>>>>>>>
>>>>>>> ANWB Online Beheer
>>>>>>>
>>>>>>> Gebouw B midden; B330
>>>>>>> Wassenaarseweg 220
>>>>>>> Den Haag
>>>>>>>
>>>>>>> Tel: (intern) toestel 7325/7301
>>>>>>> Mob: 06-50233023
>>>>>>>
>>>>>>> **********************************************************************************************
>>>>>>> Disclaimer
>>>>>>>
>>>>>>> E-mail wordt door ANWB niet gebruikt voor het aangaan van externe
>>>>>>> verplichtingen.
>>>>>>> Deze e-mail is uitsluitend bestemd voor geadresseerde(n). Indien deze
>>>>>>> e-mail
>>>>>>> onverhoopt niet voor u is bestemd dan verzoeken wij u vriendelijk contact
>>>>>>> op te
>>>>>>> nemen met de afzender en daarna het bericht te vernietigen. Deze e-mail
>>>>>>> mag niet
>>>>>>> worden doorgestuurd, openbaar gemaakt of verveelvoudigd worden zonder de
>>>>>>> toestemming van de afzender.
>>>>>>> ANWB betracht grote zorgvuldigheid bij het verzenden van e-mails. ANWB
>>>>>>> kan
>>>>>>> echter niet garanderen dat deze e-mail juist, volledig, tijdig en
>>>>>>> virusvrij wordt
>>>>>>> overgebracht. In een dergelijk geval is ANWB op geen enkele wijze
>>>>>>> aansprakelijk
>>>>>>> voor enige schade, direct dan wel indirect, in welke vorm dan ook.
>>>>>>>
>>>>>>> ANWB B.V.
>>>>>>>
>>>>>>> **********************************************************************************************
>>>>>>>
>>>>>>> _______________________________________________
>>>>>>> Hippoportal-dev mailing list
>>>>>>> [hidden email]
>>>>>>> http://lists.hippo.nl/mailman/listinfo/hippoportal-dev
>>>>>>>
>>>>>> _______________________________________________
>>>>>> Hippoportal-dev mailing list
>>>>>> [hidden email]
>>>>>> http://lists.hippo.nl/mailman/listinfo/hippoportal-dev
>>>>>
>>>>> --
>>>>> [hidden email]  www.onehippo.com
>>>>>
>>>>> Amsterdam     - Hippo B.V. Oosteinde 11 1017 WT Amsterdam +31(0)20-5224466
>>>>> San Francisco - Hippo USA Inc. 101 H Street, suite Q Petaluma CA 94952-5100
>>>>> +1(877) 41-HIPPO
>>>>>
>>>>> _______________________________________________
>>>>> Hippoportal-dev mailing list
>>>>> [hidden email]
>>>>> http://lists.hippo.nl/mailman/listinfo/hippoportal-dev
>>>>>
>>>> _______________________________________________
>>>> Hippoportal-dev mailing list
>>>> [hidden email]
>>>> http://lists.hippo.nl/mailman/listinfo/hippoportal-dev
>>>>
>>>
>>>
>>> --
>>> [hidden email]     www.onehippo.com
>>> EUROPE • AMSTERDAM - Hippo B.V. Oosteinde 11 1017 WT Amsterdam +31(0)20-5224466
>>> NORTH AMERICA • SAN FRANCISCO - Hippo USA Inc. 185 H Street, Suite B
>>> Petaluma CA 94952 +1 (877) 414-4776
>>> _______________________________________________
>>> Hippoportal-dev mailing list
>>> [hidden email]
>>> http://lists.hippo.nl/mailman/listinfo/hippoportal-dev
>>>
> _______________________________________________
> Hippoportal-dev mailing list
> [hidden email]
> http://lists.hippo.nl/mailman/listinfo/hippoportal-dev


--
[hidden email]  www.onehippo.com

Amsterdam     - Hippo B.V. Oosteinde 11 1017 WT Amsterdam +31(0)20-5224466
San Francisco - Hippo USA Inc. 101 H Street, suite Q Petaluma CA 94952-5100 +1(877) 41-HIPPO


_______________________________________________
Hippoportal-dev mailing list
[hidden email]
http://lists.hippo.nl/mailman/listinfo/hippoportal-dev