IOC memory leak

[bickley@cds005.cebaf.gov (Matt Bickley)]

Hello, all,
	We have an interesting problem at CEBAF and I wondered if anyone has any
insight into it:  For a while we have been aware of a leak in free memory when
running under EPICS version 3.11.  On some IOCs the memory disappears at a rate
of about 100 KB/day and on other IOCs it is much slower.  This concerns us, but
is manageable while we are still commissioning the accelerator.  We believed
that there was a correlation between the rate of the memory leak on an IOC and
the demand placed on that IOC by channel access.  In order to get a better
handle on the problem we have been archiving the total free memory and the
largest available block of memory on every operational IOC at CEBAF.  We were
surprised to discover this week that the leak continued unabated over the
four-day Thanksgiving holiday, despite there being no CA connect/disconnect
activity.
	While some of the memory leakage may be due to CEBAF-developed applications,
we believe that there is probably a memory leak in some core EPICS software.
We believe this because all of our IOCs lose memory, and no single application
runs on all of CEBAF's IOCs.
	We plan to do more extensive analysis of the leak starting at the beginning
of our next shutdown, and will let everyone know what we find.  In the
meantime, has anyone else noted this same symptom?  Does anyone have good ideas
for a likely source of such a leak?  We'd like to hear your ideas.


Matthew Bickley                                      Email: bickley@cebaf.gov
Computer Scientist                                   Voice mail: 804-249-7347
CEBAF