Experimental Physics and
Industrial Control System

Kenneth Kang <[email protected]> · Wed, 25 Aug 1999 11:23:10 -0700 (PDT)

One of the SLAC/BaBar IOCs (CEN-BIP) seems to be experiencing the symptoms
described in the message below.  What is the status of the patch?  How can
we avoid the bug?

===== From Tech Talk Archives =====
From: [email protected] (Jeff Hill)
Date: Wed, 21 Jul 1999 18:52:49 -0600
Subject: bug in EPICS R3.13 and R3.12
Message-ID: <[email protected]>

All,

Last week during testing I found a bug in the channel 
access client library which, while it should occur under 
rare circumstances, is significant because it could result 
in the failure of client side programs. If the problem 
occurs in client programs running in the IOC experimental
evidence indicates that under heavy network communication
load network communication may stall or permanently fail. 
I introduced the bug into R3.12 in August of 1995.

Symptoms on a workstation: 
o client programs stall or hang when the workstation is
experiencing heavy network communication load

Symptoms on an IOC:
o dbCaLink or sequencer task in IOC is temporarily
or permanently using all available CPU at its priority
level.
o network communication to the IOC stalls or possibly
permanently fails under heavy network communication load

A patch for R3.13 will be available soon (I have committed 
a fix to the R3.13.1 branch at APS). My apologies for any 
inconvenience that this has caused.

Jeff
______________________________________________________
Jeffrey O. Hill           Internet     [email protected]
LANL MS H820              Voice        505 665 1831
Los Alamos NM 87545 USA   FAX          505 665 5107

On Wed, 25 Aug 1999 [email protected] wrote:

> Sorry for not informing you about this, it did not just happen after I left
> (otherwise they wouldn't know what to do about it, I left them instructions on
> how to reboot the BIP).  The condition started last Wednesday.  I had added
> several new channels recently, so I thought that was the problem.  But after
> removing them the condition persisted.  I made a change to the injection
> state machine several days before the problem started, so I don't think that
> is causing it.  Maybe Kenneth can take a look at inj.st to see if anything 
> looks funny.  However, the symptom doesn't point to an infinite loop problem,
> where you see one sequencer taking up all the cpu time.  Instead I see processes
> like dbCaLink and seqAux taking up all the cpu time.  There are a large number
> of CA clients connected to the BIP (>60) so I am wondering if there is a client
> out there taking down the IOC.

Experimental Physics and Industrial Control System

Experimental Physics and
Industrial Control System