The RTI_JMS_ConfigOpsManual.pdf file in section 4.1 recommends segregating systems and subsystems into domains. That is what our designers have done. They have designed our application with 5 domains (may want to expand to 6).
My problem in implementation is that our RTI DDS installation will only allow me 4 domains no matter how high I set the resource_limits.max_objects_per_thread. At 1st I could only get 2 domains (max_objects_per_thread = 1024). After raising max_objects_per_thread I was able to get 2 more domains (max_objects_per_thread = 2048). Now I have raised it very high, but I still cannot get more than 4 domains (max_objects_per_thread = 8192).
Is there some installation parameter for resources that is limiting me to 4 domains? Is there any way to raise the resource limits besides having the application set max_objects_per_thread?
I am using Red Hat release 6.6 (Santiago) on a Dell Precision M6800.
The free -m commands shows 32 gigabytes of memory with 30 gigabytes free.
Responding with 1 to the top command shows 8 CPUs.
I do not see why my resources are being limited. I am thinking it is some sort of installation parameter that was set for RTI DDS.
Hello Herb,
Do you mean you have a single process that is creating 5 different DomainParticipants, each using a different DomainId? What are the DomainId values?
What do you mean by "RTI DDS installation will only allow me 4 domains"? Do you mean that the create_domainparticipant() operation fails? Or do you mean something else? If this is the case what error(s) are you getting? You may want to turn on a higher verbosity level (see the NDDSConfigLogger class) and report any errors you get...
Gerardo
Gerardo,
We have a singleton pattern implemented through an enum that creates the DomainParticipants for all of the processes and keeps track of them in a Java Map. It uses the com.rti.dds.domain.DomainParticipantFactoryQos to create the DomainParticipants, and it enforces only 1 domain per participant. The domain IDs are 0, 8, 9, 10, 11, . . .
After creating a participant with DomainParticipantFactory.get_instance().create_participant_with_profile() and setting subscriber, publisher, datareader and datawriter, there is a call to DomainParticipant.enable which flows through to com.rti.dds.infrastructure.EntityImpl. There is a call there to DDS_Entity_Enable and then the RETCODE_ERROR.check_return_code is (Unknown Source).
I only get this on the 4th DomainParticipant that is created. Actually it may be the 5th domain since we keep domain 0 as the framework domain.
Herb
Hello Herb,
It appears you are hitting another resource limit. It could well be somthing else other than the
max_objects_per_thread. Perhaps it is the shared memory handles as some other posts on participant creation failures indicate. A fast way to troubleshoot it would be if we could get some of the error messages that would normally be printed to the console when the error occurs. Are you not seeing any error messages?Perhaps you can configure a higher verbosity level with a call to:
Logger.get_instance().set_verbosity(ogVerbosity.NDDS_CONFIG_LOG_VERBOSITY_WARNING);Logger.get_instance().set_output_file()I was also suspecting the possibility of a shared memory problem. We switched our transport to UDPv4 instead of SHMEM and things started working. I guess we will have to investigate our shared memory setup more, or just rely on the UDPv4 transport.
Thanks.
If it is shared memory you can try some of the settings suggested in these posts:
https://community.rti.com/kb/what-are-possible-solutions-common-shared-memory-issues
https://community.rti.com/kb/what-causes-rtiosapisharedmemorymutexcreateos-semget-failure-error-0x1c-error-message
https://community.rti.com/content/forum-topic/failure-creating-domain-participant
Gerardo