Hi,
I don't refuse from current alarm too. This way just adds
another mechanism to collect all alarms. So maybe ALARM subsystem should
have 2 ways? One way is narrow what there is now. Another one is wide
that collect all.
I intentionally missed details of implementation. Like ALARM
thread in IOC, ALARM logger server and GUI client. I also missed data
interface between thread and logger, logger and client. I am sure that
correct implementation gives possibility to jump from to edm screen of
alarm device.
> But it just adds one more network dependency to the IOCs
Actually, IOC has one network dependency for all network connections.
But other ends of connections are question. Fortunately, control system
can have small quantity types of tools servers (Error, DB, Alarm,
Archive, File server ...) :)
Another way is new network. Control system has timing, mps. I suppose
this way will be a little expensive ...
Thanks, Andrei.
-----Original Message-----
From: Kay-Uwe Kasemir [mailto:[email protected]]
Sent: Thursday, November 03, 2005 1:24 PM
To: tech talk
Subject: Re: Alarm in Epics
On Nov 3, 2005, at 12:29 , Liyu, Andrei wrote:
> Then checkAlarms() function calls recGblSetSevr( precord, ...)
> function.
>
> It is macros. So we can EASY change macros and new function will write
> message to errlog server (I don't know another suitable central engine
> in Epics).
>
> After that we will have all alarms in one place. I believe that
> advantages are clear.
Hi:
As far as I understand, the current idea is
- each record _determines_ its alarm state,
reflects the result in STAT/SEVR field.
A CA client can ask for data with alarm info.
- the alh alarm handler _monitors_ alarms,
_logs_ them, _displays_ them to users,
allows users to _acknowledge_ alarms, ...
- some part of the acknowledgement is actually
also handled by special fields in the records
which remember the highest no-ack'ed alarm.
There are alternate ideas. The D0 significant event server is one.
I agree that combining monitoring, logging, displaying, ack'ing
all in one GUI tool is too much.
I'm not sure about your idea of _adding_ the logging to the IOC.
Sure, it's a comparably easy thing to add a message log 'write'
to recGblSetSevr.
But it just adds one more network dependency to the IOCs,
and then how do you display the message log,
how do you acknowledge alarms?
If you're serious about re-thinking the alarm handling,
I'd rather remove the special alarm ack fields from the IOCs,
since non-database-CA-servers sometimes don't implement these
anyway, and also split the alh task, resulting in this separation:
- records/IOCs/CA servers determine alarms
- one or more alarm monitor daemons monitor values & alarms,
log them
(maybe as last time & value before alarm, time & value with alarm),
remember the highest non-acked alarm,
provide network API for ack'ing alarms,
where the acknowledgement (w/ user, maybe comment) gets logged.
- GUI clients to the alarm daemon can display the current alarms.
They also offer the option of acknowledging alarms,
though that's probably limited to certain users.
Of course these are just my ideas, I've probably missed many details,
but I think it's clear that just adding a log message doesn't
really improve much.
-Kay
- Navigate by Date:
- Prev:
Re: Running softIoc's in the background Steven Hartman
- Next:
RE: linux-xscale port, SYSFS device support Jeff Hill
- Index:
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
<2005>
2006
2007
2008
2009
2010
2011
2012
2013
2014
2015
2016
2017
2018
2019
2020
2021
2022
2023
2024
- Navigate by Thread:
- Prev:
Re: Alarm in Epics Kay-Uwe Kasemir
- Next:
Base 3.14.7 build error on SUSE machine alcocer
- Index:
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
<2005>
2006
2007
2008
2009
2010
2011
2012
2013
2014
2015
2016
2017
2018
2019
2020
2021
2022
2023
2024
|