I get a nasty memory leak while polling an ABB Freelance (ver. 6.1) OPC server with OSI's PI client interface. It takes about 3 days to eat up ~400MB of memory. Has anyone run into a similiar problem with a Freelance opc server?
Haven't been too successful with tech support from ABB or OSI, and I'm getting tired of worrying about the opc server bombing out at 3:00 am... Help if you can!

 

Paul,

I have recently completed a project that is quite simular to yours and resulting in the same problem. Although the DCS vendor was different, the problem was the same.

The vendor software development group identified the problem in their OPC server software, not the client feeding the data to the server.

In this instance, the problem was traced to how the server processed the data being sent by the client and the deallocation of used memory. We were sending too much data too fast and the server could not process the received data before a new set of data was received resulting in the excessive memory leak.

Once the server receives the data, it must propagate it throughout all nodes on the DCS network. In our case, it could not propagate it fast enough.

You can run a simple test to see if you have the same problem as ours. Decrease the number of tags being sent from the client and decrease the
update frequency to the server.

Good Luck,
Craig
CSE Engineering, Inc.
[email protected]

 

Thank you VERY much for your reply. I will give this a shot and post the results.

Paul

 

Hallo Paul,
can you give me more details for your problem.
I am using Freelance 2000 (V6.1) with OPC connected to InTouch wonderware.My systems consists of four PLCs and three PCs connected through Ethernet using OPC.
Regards
Tasos

 

Paul,

I'm the Support Manager for OSIsoft. I'm sorry you weren't able to get help from our support department. I researched your call and discovered that we were waiting to hear back with information about Freelance's OPC server. We
can work with them directly to try to track down the cause of this memory leak. I'm having one of our Engineers write you in order to get this moving. If you've discovered anything from the test discussed in this thread, let us know.

Thanks much,
Steve Nye
Techsupport Manager, OSIsoft

 

Wow, I'm impressed! I was contacted by one of your support staff and the problem has since been resolved. Thank you (and I respectfully retract my earlier comments about your TS department).

For the other readers: The OPC server was giving me values for each point on every scan, letting the client decide whether or not the data
was changed. Configuring the points as advised/update on change solved the problem.

Thanks again,

Paul Felsheim
Ticona Celstran, Inc.