You are viewing a plain text version of this content. The canonical link for it is here.
Posted to fop-dev@xmlgraphics.apache.org by "Luis Bernardo (JIRA)" <ji...@apache.org> on 2014/01/24 01:07:38 UTC

[jira] [Commented] (FOP-2333) knowing Multi threading issues for FOP 1.1

    [ https://issues.apache.org/jira/browse/FOP-2333?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13880528#comment-13880528 ] 

Luis Bernardo commented on FOP-2333:
------------------------------------

Although flagged as an improvement (request?) questions like these are better dealt with in the mailing list. Each FopFactory instantiates a cache for images. Having more than one thread share a FopFactory, and as a consequence share also the cache, may cause concurrency issues, although there are other users of FOP that share FopFactory's between threads and apparently without problem. But the risk is there.

Can you provide a small example that can replicate the high CPU usage after a long continuous run?

> knowing Multi threading issues for FOP 1.1
> ------------------------------------------
>
>                 Key: FOP-2333
>                 URL: https://issues.apache.org/jira/browse/FOP-2333
>             Project: Fop
>          Issue Type: Improvement
>          Components: general
>    Affects Versions: 1.1
>         Environment: Production
>            Reporter: Krishankant singh
>            Priority: Critical
>              Labels: performance
>             Fix For: 1.1
>
>   Original Estimate: 672h
>  Remaining Estimate: 672h
>
> Hi,
> I am a software engineer - java/j2ee
> We have a production application that creates 500 thousand pdf/day and it has implemented fop 0.95  As per current implementation in our code, we have not shared the FopFactory instance for threads. For each thread, we create a new FopFactory  instance, so after 6 or 7 days, the application CPU usage is too high and the application needs to be restarted every 6 - 7 days.
> As we know that creating new  FopFactory instance for each thread is a very costly operation, so to cater to above problem can we share one FopFactory instance for all threads. Are there any problems reported with this approach in fop 1.1 because we can for sure upgrade to fop 1.1 from fop 0.95



--
This message was sent by Atlassian JIRA
(v6.1.5#6160)