Experimental Physics and
Industrial Control System

"Jeff Hill" <[email protected]> · Fri, 6 Aug 2004 13:17:10 -0600

If the IOC with the SEVERITY that does not update meets the following
requirements then I think that Ken may very well have come up with a correct
assessment of the situation. If so, then this is bug 52 in Mantis which, as
Ken mentioned, is fixed in R3.13.10 (and R3.14.5).

o running less than R3.13.10 (or less than R3.14.5) 
o was at some time in the past very low on network buffers

This problem was probably not identified in the past because, with older
versions of vxWorks, the network buffer pool was expanded when insufficient
to keep up with load. With newer versions of vxWorks the network buffer
pools size is determined when the vxWorks kernel is compiled, and does not
change at runtime.

Jeff

> -----Original Message-----
> From: Kenneth Evans, Jr. [mailto:[email protected]]
> Sent: Friday, August 06, 2004 11:24 AM
> To: Tech-Talk
> Cc: Jeff Hill
> Subject: Re: CA Gateway problem
> 
> We have had problems of that nature that derive from a bug in RSRV, the
> server in the IOCs, for 3.13 IOCs.  The IOC loses contact with the
> client
> (e.g. Gateway) and marks the circuit as down but doesn't close the
> socket.
> The Gateway thinks it is still connected and continues sending out its
> cached data.  There is no way to make the Gateway reconnect except to
> restart it.  All channels in the circuit are affected.  (A circuit is
> all
> the connections between one client and one server, all associated with
> the
> same TCP/IP socket.)  Other Gateway circuits are OK.  The problem seems
> to
> be associated with a lack of MBUFs in the IOC at some time.  There may
> be no
> memory problems later when you look at it, however.  This problem was
> fixed
> in 3.13.10, and RSRV now closes the socket so that the Gateway knows it
> is
> disconnected.
> 
> 	-Ken

Experimental Physics and Industrial Control System

Experimental Physics and
Industrial Control System