You are viewing a plain text version of this content. The canonical link for it is here.
Posted to users@jackrabbit.apache.org by Andrey Adamovich <an...@yahoo.com> on 2009/10/05 11:47:40 UTC

PM concurrency issues

Hi guys!

 
I have recently created an issue about concurrency problems with Oracle PM: https://issues.apache.org/jira/browse/JCR-2345 

But this seems to be a long running issue with JR as in this issue (https://issues.apache.org/jira/browse/JCR-314) Jukka says:

"Note that without solving the concurrency issues in persistence
managers (both db and bundle) we're still stuck with fully serialized
backend access. But this is one step in the correct direction; the
second step would be to start removing the synchronization on the
persistence managers."

We are using JR 1.5.5 and I haven't noticed any big changes in JR's SVN till version 1.6.0. So this is still an issue?

Am I right? Or is there a possible solution to that? Anything I can do about this? Will removing syncronisation blocks from Oracle PM be very dangerous step?

Best regards,
Andrey 



      

Re: PM concurrency issues

Posted by Andrey Adamovich <an...@yahoo.com>.
Is this than a major issue for JR 1.+ versions?

 Andrey 




________________________________
From: Ian Boston <ie...@tfd.co.uk>
To: users@jackrabbit.apache.org
Sent: Monday, 5 October, 2009 13:27:38
Subject: Re: PM concurrency issues


On 5 Oct 2009, at 10:47, Andrey Adamovich wrote:

> Hi guys!
> 
> 
> I have recently created an issue about concurrency problems with Oracle PM: https://issues.apache.org/jira/browse/JCR-2345
> 
> But this seems to be a long running issue with JR as in this issue (https://issues.apache.org/jira/browse/JCR-314) Jukka says:
> 
> "Note that without solving the concurrency issues in persistence
> managers (both db and bundle) we're still stuck with fully serialized
> backend access. But this is one step in the correct direction; the
> second step would be to start removing the synchronization on the
> persistence managers."



I would be interested to know if this is still a problem with the DB based PM's in general ?
Are there any other PM's out there suitable for production work that dont have a DB as a bottleneck, because AFAICT, each PM has 1 DB connection serializing the writes to the DB?
A more sophisticated DB PM might have worker threads to increase throughput ?

Lots of questions, no answers (sorry)

Ian


> 
> We are using JR 1.5.5 and I haven't noticed any big changes in JR's SVN till version 1.6.0. So this is still an issue?
> 
> Am I right? Or is there a possible solution to that? Anything I can do about this? Will removing syncronisation blocks from Oracle PM be very dangerous step?
> 
> Best regards,
> Andrey
> 
> 
> 


      

Re: PM concurrency issues

Posted by Ian Boston <ie...@tfd.co.uk>.
On 5 Oct 2009, at 10:47, Andrey Adamovich wrote:

> Hi guys!
>
>
> I have recently created an issue about concurrency problems with  
> Oracle PM: https://issues.apache.org/jira/browse/JCR-2345
>
> But this seems to be a long running issue with JR as in this issue (https://issues.apache.org/jira/browse/JCR-314 
> ) Jukka says:
>
> "Note that without solving the concurrency issues in persistence
> managers (both db and bundle) we're still stuck with fully serialized
> backend access. But this is one step in the correct direction; the
> second step would be to start removing the synchronization on the
> persistence managers."



I would be interested to know if this is still a problem with the DB  
based PM's in general ?
Are there any other PM's out there suitable for production work that  
dont have a DB as a bottleneck, because AFAICT, each PM has 1 DB  
connection serializing the writes to the DB?
A more sophisticated DB PM might have worker threads to increase  
throughput ?

Lots of questions, no answers (sorry)

Ian


>
> We are using JR 1.5.5 and I haven't noticed any big changes in JR's  
> SVN till version 1.6.0. So this is still an issue?
>
> Am I right? Or is there a possible solution to that? Anything I can  
> do about this? Will removing syncronisation blocks from Oracle PM be  
> very dangerous step?
>
> Best regards,
> Andrey
>
>
>