[omniORB] Open handles are exhausted on Linux

[Matej Kenda]

Hi,

In the application that we develop the following scenario happens on
Linux from time to time:

A process containing CORBA servants is hit by many requests from
different CORBA clients in a short time period. Each of the clients
terminates after sending one request and receiving the answer.

The number of open handles in the servant gets exhausted and the
subsequent outgoing calls fail with CORBA::TRANSIENT
(TRANSIENT_NoUsableProfile). There are a number of sockets in the state
CLOSE_WAIT at that moment.

Increasing the number of open handles helps, but I would like to know if
there is any of the omniORB parameters that I can use to prevent this
situation from happening?

I have tried to set the following parameters, but without any success.

oneCallPerConnection = 0
threadPerConnectionPolicy = 1
threadPerConnectionUpperLimit = 20
threadPerConnectionLowerLimit = 15

TIA,

Matej

-- 
Matej Kenda, Lead Engineer
HERMES SoftLab (www.hermes-softlab.com)
Erjavčeva 2, 5000 Nova Gorica, Slovenia