Experimental Physics and
Industrial Control System

"Leicester, PJ \(Pete\)" <[email protected]> · Mon, 10 Oct 2005 16:58:54 +0100

We are running a test gateway at Diamond and have experienced problems
when doing large transfers (EPICS_CA_MAX_ARRAY_BYTES=3000000). I have
posted to techtalk on previous occasions regarding the problem. Ken
Evans kindly responded drawing my attention to Mantis bug #176 however
we are running an EPICS cvs snapshot that post dates this fix. Clearly
running with a cvs snapshot is not ideal and may in itself be a problem.

Currently I am hoping that EPICS 3.14.8 appears very soon so that I can
try it with the new Gateway release. Since we already have the #176 fix
I am not sure this will solve our problems but I am hopeful.

Pete Leicester
Diamond Light Source

-----Original Message-----
From: Martin L. Smith [mailto:[email protected]] 
Sent: 10 October 2005 14:25
To: Dalesio, Leo `Bob`
Cc: [email protected]
Subject: Re: Channel Access Gateway (Problems at the SLS)

I have been administering the PV gateways (40 of them) at the APS since
the beginning of the PV gateway. There have been issues along the way,
however, at this time I feel that for at least the past year the PV
gateways at APS have been performing rather well.

There are currently no problems that I have seen that can be attributed
directly to the current PV gateway (version 2). I have encountered
problems in the past, but there has been nothing to prove to me that
they were being caused by the PV gateway.

Marty Smith

Dalesio, Leo `Bob` wrote:
> Hi All,
> 
> A change of title would be a bit more precise (as well as sensative).
> 
> Airing issues is what a collaboration meeting does best when things 
> are going well. It is a chance to explain what is going wrong for 
> others. It would have been nice to see if anyone else is experiencing 
> any of the reported problems. Is anyone?
> 
> Some of the things reported in this talk as bugs are simply the way 
> that the gateway must work to actually limit traffic. For instance, if

> you connect to the same channel 5 times, one is periodic and the other

> is moniitor - the monitor is considered to be the most up to date 
> value. If a large MDEL makes the value less useful for the application

> that wants to do a get - use a smaller MDEL to the gateway. It is fine

> to point out how a particular tool works in ways that were not 
> expected by the user. It gives others a heads up.
> 
> It would be nice to go beyond the imperfect way we communicate and 
> have a list of the issues posted and a response to whether or not the 
> situation is the intentional implementation of the gateway and some 
> pointers on how to get the tool to provide what you want, if the 
> situation is a site specific problem that cannot be reproduced and 
> perhaps some pointers fromt the community on a way to get more 
> diagnostic information to further characterize the problem, and 
> finally isolate anything that is a problem with the gateway so that 
> people with very limited time that have provided very valuable support

> to the community can provide some help.
> 
> We have precedent for having these people come to your site and help 
> you to diagnose the issue. This is normally done at the expense of the

> site requesting support. Even though it may be a problem in the code 
> authored elsewhere, it is really important to remember that these are 
> very large codes, supported almost entirely at the expense of some 
> other project.
> 
> Bob
> 

Experimental Physics and Industrial Control System

Experimental Physics and
Industrial Control System