[omniORB] Per thread for per connection?

GAO,YAN (HP-Singapore,ex3) yan_gao@hp.com
Fri, 4 May 2001 18:12:09 +0800


Hi, Duncan,

Thanks for your message. I now increase the the number of thread. Could you
tell me after how long time if not in used, the thread will be closed. Can I
control the behaviour? If possible can I modify the source to make it as
short as possible?

I create a corba server A by using the persistent policy like 
    pl[0] = root_poa->create_lifespan_policy(PortableServer::PERSISTENT);
    pl[1] = root_poa->create_id_assignment_policy(PortableServer::USER_ID);

Then I use it to fork and execl another corba B. 

I kill the process A and leave the B alive.

next if I try to start A, I will definitely get a core file( receiving
SIGABORT). if I start A after I kill B then everything is OK. So I think
that when forked from A, B inherite some property from A. In this manner,
even A died, the resource ( port number) will be released to the system. Is
there are anything I can do the prevent it from core dumping?

Thanks in advance,

Best Regards,

Gao Yan
> -----Original Message-----
> From:	Duncan Grisby [mailto:dgrisby@uk.research.att.com]
> Sent:	Thursday, May 03, 2001 12:19 AM
> To:	GAO,YAN (HP-Singapore,ex3)
> Cc:	'omniorb-list@uk.research.att.com'
> Subject:	Re: [omniORB] Per thread for per connection? 
> 
> On Wednesday 2 May, "GAO,YAN (HP-Singapore,ex3)" wrote:
> 
> > 1)	Is this true that omni Corba use dedicated thread to each
> > connection. ( Previously I think when method is called, the thread will
> be
> > created to do it, After the end of method call, the thread will close)
> 
> Yes. See chapter 8 of the omniORB manual. Note that unused connections
> (and their associated threads) are closed after a while, so you only
> see lots of threads if there are lots of active clients.
> 
> > 2)	Do you have thread pool support for example?
> 
> Not at present.
> 
> > 3)	Any work around to resolve this issue?
> 
> You can quite probably increase the number of threads your OS will
> permit the server process to have.
> 
> Cheers,
> 
> Duncan.
> 
> -- 
>  -- Duncan Grisby  \  Research Engineer  --
>   -- AT&T Laboratories Cambridge          --
>    -- http://www.uk.research.att.com/~dpg1 --
>