Experimental Physics and
Industrial Control System

"J. Lewis Muir" <[email protected]> · Sat, 9 Nov 2013 07:01:02 -0600

On 11/8/13 6:11 PM, Maren Purves wrote:
> Lewis and Andrew,
>
> for me at least this thread kinda "fell out of the sky".

Hi, Maren.

Sorry about that.  That's because this thread was started a while ago.
Here's the start of it in the Tech-Talk archive:

  http://www.aps.anl.gov/epics/tech-talk/2013/msg01068.php

> Without knowing what you're running on, Lewis, our experience is that
> with vxWorks not having garbage collection if you create and drop
> connections to somewhere you'll at some point find yourself unable to
> connect to anything anymore.  We have one IOC that I reboot weekly
> to prevent exactly that to happen, and we have separated that part
> (monitoring weather and the like) out to prevent this from causing
> more serious problems than image quality degradation.
>
> Not sure I'm helping any here, but that's the very first thing that
> came to my mind.
>
> Aloha,
> Maren

I appreciate your comments.  The IOC hardware and software is stated at
the bottom of the first message in this thread (referenced above).

What kind of connections are you talking about?  TCP and UDP?  This
would include EPICS CA connections too, right?  So, you're saying that
an EPICS IOC cannot run indefinitely on VxWorks when regularly creating
and dropping TCP or UDP connections?

I have other IOCs that this problem IOC communicates with.  One of the
IOCs is a soft IOC that is usually stopped and started around once a
day.  If any other IOC is rebooted that this problem IOC communicates
with, that would be a dropped and then created connection.  This problem
IOC runs two asyn TCP/IP servers that are connected to around once a day
(and disconnected from).  Are you saying this won't work indefinitely on
VxWorks?

Is there any way to detect that the IOC is about to get into this bad
state?

Thanks,

Lewis

Experimental Physics and Industrial Control System

Experimental Physics and
Industrial Control System