You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@chemistry.apache.org by "Lu, Wentao" <We...@bchydro.com> on 2014/07/01 00:15:47 UTC

RE: memory leak on OpenCMIS bridge?

Hi Florian,

We are testing SimpleForwardingCmisService.java under "org.apache.chemistry.opencmis.bridge.httpsession.sample" package, so I guess it is HttpSession implementation.  Correct me if I am wrong, our current the architecture is 1 ESB->Load Balancer->2 OpenCMIS Bridge Servers->2 CMIS Servers, I am wondering if we enable cookies(session stickiness on LB), then all transactions could possible go to one OpenCMIS Bridge server only. 

Our current session time is 5 minutes in web.xml, this may help to reduce the impact of "lots of unused HTTP session". I am thinking if we should reduce Chemistry client cache TTL from default 2 hours to shorter time or if it is possible we can disable client cache completely?


Thanks
Wentao


-----Original Message-----
From: Florian Müller [mailto:fmui@apache.org] 
Sent: 2014, June 30 1:16 PM
To: dev@chemistry.apache.org
Cc: Lu, Wentao
Subject: Re: memory leak on OpenCMIS bridge?

Hi Wentao,

Are you using the HttpSession or the LruCache implementation?
If you are using the HttpSession implementation, make sure that the clients have cookies enabled. Otherwise you are collecting a lot of unused HTTP sessions on the server side that fill up your memory.


- Florian


> Thanks Florian for your help again.
> 
> I've uploaded the screenshot into Google Doc  
> https://drive.google.com/file/d/0B3EnTtzgQ8P5R0VOOE9iU0N2a0E/edit?usp=
> sharing
> 
> Could you help to have a look?
> 
> 
> Wentao
> 
> -----Original Message-----
> From: Florian Müller [mailto:fmui@apache.org]
> Sent: 2014, June 30 12:38 PM
> To: dev@chemistry.apache.org
> Cc: Lu, Wentao
> Subject: Re: memory leak on OpenCMIS bridge?
> 
> Hi Wentao,
> 
> Attachments and images are not forwarded to this mailing list. We cannot see your screenshot.
> 
> But we (SAP) are using the OpenCMIS Bridge in a productive environment with a lot of requests and sessions. We haven't encountered any memory leaks, yet.
> 
> 
> - Florian
> 
> 
>> Hi,
>>
>>  
>>
>> We are load testing "OpenCMIS Bridge 0.10.0" but looks like it has 
>> “memory leak” issue. I am wondering if that because of the OpenCMIS 
>> client cache or something else? Here is the screenshot from IBM 
>> HeapAnalyzer.
>>
>>  
>>
>>  
>>
>> Thanks
>>
>> Wentao
>>
>> ---------------------------------------------------------------------
>> -
>> -- This email and its attachments are intended solely for the 
>> personal use of the individual or entity named above. Any use of this 
>> communication by an unintended recipient is strictly prohibited. If 
>> you have received this email in error, any publication, use, 
>> reproduction, disclosure or dissemination of its contents is strictly 
>> prohibited. Please immediately delete this message and its 
>> attachments from your computer and servers.
>> We would also appreciate if you would contact us by a collect call or 
>> return email to notify us of this error. Thank you for your cooperation.
>> -BCHydroDisclaimerID5.2.8.1541

Re: memory leak on OpenCMIS bridge?

Posted by "Guendisch, Dieter" <di...@sap.com>.
Florian meant that you should make sure that your CMIS clients should
utilize cookies (not your Loadbalancer).
If the CMIS clients do not use cookies, then every single request might
end up in a new http session.

On 01/07/14 00:15, "Lu, Wentao" <We...@bchydro.com> wrote:

>Hi Florian,
>
>We are testing SimpleForwardingCmisService.java under
>"org.apache.chemistry.opencmis.bridge.httpsession.sample" package, so I
>guess it is HttpSession implementation.  Correct me if I am wrong, our
>current the architecture is 1 ESB->Load Balancer->2 OpenCMIS Bridge
>Servers->2 CMIS Servers, I am wondering if we enable cookies(session
>stickiness on LB), then all transactions could possible go to one
>OpenCMIS Bridge server only.
>
>Our current session time is 5 minutes in web.xml, this may help to reduce
>the impact of "lots of unused HTTP session". I am thinking if we should
>reduce Chemistry client cache TTL from default 2 hours to shorter time or
>if it is possible we can disable client cache completely?
>
>
>Thanks
>Wentao
>
>
>-----Original Message-----
>From: Florian Müller [mailto:fmui@apache.org]
>Sent: 2014, June 30 1:16 PM
>To: dev@chemistry.apache.org
>Cc: Lu, Wentao
>Subject: Re: memory leak on OpenCMIS bridge?
>
>Hi Wentao,
>
>Are you using the HttpSession or the LruCache implementation?
>If you are using the HttpSession implementation, make sure that the
>clients have cookies enabled. Otherwise you are collecting a lot of
>unused HTTP sessions on the server side that fill up your memory.
>
>
>- Florian
>
>
>> Thanks Florian for your help again.
>> 
>> I've uploaded the screenshot into Google Doc
>> https://drive.google.com/file/d/0B3EnTtzgQ8P5R0VOOE9iU0N2a0E/edit?usp=
>> sharing
>> 
>> Could you help to have a look?
>> 
>> 
>> Wentao
>> 
>> -----Original Message-----
>> From: Florian Müller [mailto:fmui@apache.org]
>> Sent: 2014, June 30 12:38 PM
>> To: dev@chemistry.apache.org
>> Cc: Lu, Wentao
>> Subject: Re: memory leak on OpenCMIS bridge?
>> 
>> Hi Wentao,
>> 
>> Attachments and images are not forwarded to this mailing list. We
>>cannot see your screenshot.
>> 
>> But we (SAP) are using the OpenCMIS Bridge in a productive environment
>>with a lot of requests and sessions. We haven't encountered any memory
>>leaks, yet.
>> 
>> 
>> - Florian
>> 
>> 
>>> Hi,
>>>
>>>  
>>>
>>> We are load testing "OpenCMIS Bridge 0.10.0" but looks like it has
>>> ³memory leak² issue. I am wondering if that because of the OpenCMIS
>>> client cache or something else? Here is the screenshot from IBM
>>> HeapAnalyzer.
>>>
>>>  
>>>
>>>  
>>>
>>> Thanks
>>>
>>> Wentao
>>>
>>> ---------------------------------------------------------------------
>>> -
>>> -- This email and its attachments are intended solely for the
>>> personal use of the individual or entity named above. Any use of this
>>> communication by an unintended recipient is strictly prohibited. If
>>> you have received this email in error, any publication, use,
>>> reproduction, disclosure or dissemination of its contents is strictly
>>> prohibited. Please immediately delete this message and its
>>> attachments from your computer and servers.
>>> We would also appreciate if you would contact us by a collect call or
>>> return email to notify us of this error. Thank you for your
>>>cooperation.
>>> -BCHydroDisclaimerID5.2.8.1541
>