Experimental Physics and
Industrial Control System

"Kevin Tsubota" <[email protected]> · Wed, 13 Aug 2008 08:07:47 -1000

Thank you Ralph, that makes more sense and in line with what we would've
expected... the more clients the more server posts.

I'll run more isolated tests but I feel more at ease about it now.

Mahalo,
Kevin

-----Original Message-----
From: Ralph Lange [mailto:[email protected]] 
Sent: Tuesday, August 12, 2008 9:36 PM
To: Kevin Tsubota
Cc: EPICS Tech Talk
Subject: Re: EPICS Gateway question

Hi Kevin,

I assume that the description you are quoting is ambiguous or not exact.

 From my understanding, the server post rate is measured after splitting

up and partly duplicating the event stream to multiple clients of the 
gateway. I.e. if the gateway looks at one channel at 1Hz and has two 
clients watching that channel, the server event rate will be one, the 
server post rate will be two.

You could set up a mini gateway and pull up two commandline camonitors 
to check the logic behind those rates.

I think you're fine. I hope.

Cheers,
Ralph

Kevin Tsubota wrote:
>
> Hi,
>
>  
>
> I'm working on deployment of an EPICS Gateway version 2.0.3.0 in our 
> system here.  I have the gateway running on a solaris-8 computer built

> against EPICS R3.14.9 with gcc 3.3.6.   The gateway is serving 2 
> VxWorks IOCs on a separate VLAN running EPICS R3.13.10 and VxWorkd
5.3.1.
>
> According to the following description for serverPostRate:
>
> > This is the rate in Hz at which events are posted to CAS. If the 
> Gateway is keeping up, this will be very close to the serverEventRate.

> If not, it will be higher and all posted events will not have been 
> processed.
>
> It's not good when the serverPostRate is higher than the 
> serverEventRate. When I look my gateway stats I see that my 
> servoPostRate is a lot higher than my serverEventRate, sometimes by as

> much as 2x, see attached gateway_rates.pdf.  The data was logged once 
> every 10sec.  My load and cpu frac seems okay, see attached 
> gateway_load.pdf.
>
>  
>
> Is there a way see what's causing this and what can I do about it?
>
>  
>
> Thank you in advance.
>
>  
>
> Kevin Tsubota
>
> W.M. Keck Observatory
>
>  
>

Experimental Physics and Industrial Control System

Experimental Physics and
Industrial Control System