Re: How to achieve per ORB/thread client connection timeouts

Am 19.06.2026 um 10:30 schrieb Duncan Grisby:
> On Thu, 2026-06-18 at 22:42 +0200, Thomas Braun via omniORB-list wrote:
> 
> 
> [...]
>> const int NARROW_CLNT_TIMEOUT = 100;
>> const int DB_RECONNECT_TIMEOUT = 20000;
> 
>> The obvious issue is that the client connect timeout change is
>> global,
>> so while the RAII object is in place also other threads connecting to
>> other ORBs see the modified connect timeout.
>>
>> Am I missing something or is a per thread/ORB client connect timeout
>> not possible? Or am I misunderstanding things?
> 
> I take it from the numbers you show that you want to allow opening a
> new connection to the overloaded server to take a lot longer than
> normal. Is it really the case that you need to give it such a long time
> just to accept the new incoming connection, or do you really need an
> overall long timeout for the call?

 From my understanding we want to have a connection timeout. The use 
case is one server which, when the whole distributed system is 
restarted, has much more than 10k clients trying to connect to it. And 
that can get slowish...

> 
> As I expect you know, you can have a per thread call timeout:
> 
> https://omniorb.sourceforge.io/omni43/omniORB/omniORB006.html#sec95
> 
> But, as you say, the overriding connect timeout does not have a per-
> thread version. It would not be especially hard to add a per-thread
> version of the connect timeout, but it is not something that is
> currently available.

If it's possible to have a per thread connection timeout in a future 
omniORB version, that would be splendid!

Thanks,
Thomas