Experimental Physics and
Industrial Control System

<[email protected]> · Fri, 7 Mar 2014 03:55:44 +0000

Hi Rolf,

Thanks for the suggestion. I agree that
1) we don’t want yet another server on top of BEAST server, added complexity
2) the problem we have with calc record for alarms is that when an operator gets an alarm, it is not immediately obvious which PV caused the problem. Also if the IOC is not running when it’s in maintenance mode, you don’t have access to this calc record either.

My knowledge of IOCs and PVs is rudimentary, so I don’t understand Rolf’s solution of having a PV for each alarm group (which can be controlled by an operator or IOC logic). Is this PV is going to live in the very IOC which is going to be put in maintenance mode? Maintenance mode for us means we will stop the IOC, then start it after some time. So while this IOC is not running, we don’t have access to this PV.

I don’t understand why disabling the Area or System by setting ENABLED_IND flag in the alarm database is not desirable. Whenever you set ENABLED_IND you have to set all its children’s ENABLED_IND flag as well. It’s a manual process, an operator has to explicitly do it. So there is no confusion of whether the IOC crashed or stopped by the operator.

Thanks,

Xinyu

On 7 Mar 2014, at 4:39 am, <[email protected]> <[email protected]> wrote:

> ------------------------------
> 
> Message: 10
> Date: Thu, 06 Mar 2014 08:49:47 -0800
> From: Rolf Keitel <[email protected]>
> To: tech_talk <[email protected]>
> Subject: Re: CSS BEAST alarm hierarchy
> Message-ID: <[email protected]>
> Content-Type: text/plain; charset=windows-1252; format=flowed
> 
> I have some problems with these approaches (although I may be missing 
> something, I haven't looked at beast for a while):
> 
> With Kay's approach I see the danger of introducing an obfuscation layer 
> in the IOCs which separates the alarm logs from the alarm sources. 
> Operator screens will not show transient alarm conditions.
> Igor's solution introduces another layer of complexity which we don't 
> need on our machines.
> 
> I would like to have an alarm solution which involves only the alarm 
> handler and the IOCs. I would like to have the option to define with 
> each alarm group a PV which can be used to disable that group. This PV 
> could be under operator control (for disabling alarms for section 
> maintenance) or controlled by IOC logic of arbitrary complexity.
> 
> - rolf -
> 
> On 3/6/2014 6:39 AM, Kasemir, Kay wrote:
>> Hi:
>> 
>> Conceptually, I think that is exactly the right approach.
>> 
>> It's usually wrong to simply add 100 raw "temperature" PVs as alarms. It's better to have an operator display that shows these alarms, with summary PVs for sections of the machine, options to disable them, and those summary PVs are then used as alarm triggers.
>> 
>> For the displays, we already have good tools: EDM, BOY, ? editors where you can create displays that suit the purpose.
>> 
>> For the logic, we have CALC records, which have limitations for their number of input links, length of the formula, and also nothing in there that is specific to alarms.
>> So often you end up with one "formula" that has to be split into several calc records.
>> Or you put the SEVR of other records into the INPx fields because you really want to summarize based on the raw alarm states, not the raw values.
>> 
>> Do you have a web link, paper, ? on your intermediate layer?
>> Maybe it would also be worthwhile to create a new "alarm calc" record that has many inputs plus formulas that are aware of not just an INPuts value but its alarm severity?
>> 
>> Thanks,
>> Kay
>> 
>> 
>> On Mar 6, 2014, at 9:26 AM, Igor Kriznar <[email protected]> wrote:
>>> Hi,
>>> 
>>> at ANKA we have implemented filtering, which filters alarms regarding
>>> the operation mode. When machine is in shutdown or maintenance, most of
>>> alarms are suppressed, before they reach CSS BEAST.
>>> 
>>> All our alarms are feed into our intermediate alarm server, which
>>> decided if they should go further to BEAST or be suppressed or be
>>> somehow otherwise processed. For example have summary alarm PV, which
>>> sums alarms from whole subtree.
>>> Our intermediate alarm server is based on JCA and looks like bunch of
>>> plain PV channels which are used by BEAST. No changes were made to
>>> BEAST, everything is in configuration.
>>> 
>>> Enable feature of BEAST I see more like permanent switch, if you want to
>>> disable misbehaving alarm source.
>>> 
>>> Hope this helps,
>>> 
>>> Igor
>>> 
>>> 
>>> On 05.03.2014 04:39, [email protected] wrote:
>>>> Hi
>>>> 
>>>> We are currently using BEAST. There is the ability to put alarm system in Maintenance mode. But it puts the whole entire system in Maintenance mode. We would like to be able to put an Area or a System in maintenance mode.
>>>> 
>>>> I?m wondering if there is any plan to implement such feature. Or would anyone else find this feature useful?
>>>> 
>>>> We have started investigating this implementation. Essentially the ENABLED_IND database field should be move to ALARM_TREE table from PV table, and isEnable() should be a base class method.
>>>> 
>>>> Thanks,
>>>> 
>>>> Xinyu WU
>>>> ASKAP Computing
>>>> Australia Telescope National Facility
>>>> CSIRO Astronomy and Space Science
>>>> phone: +61 2 9372 4727
>>>> postal: PO Box 76 Epping, NSW. 1710
>>>> 
>>>> 
>>>> 
>>> 
>>> 
>> 
> 
> 
> -- 
> +-------------------------------------------------------+
> | Rolf Keitel, Ph.D.                Tel: (604) 222-7453 |
> | TRIUMF                                                |
> | Vancouver, B.C., Canada                               |
> +-------------------------------------------------------+
> 
> 

Xinyu WU
ASKAP Computing                       
Australia Telescope National Facility 
CSIRO Astronomy and Space Science
phone: +61 2 9372 4727
postal: PO Box 76 Epping, NSW. 1710

Experimental Physics and Industrial Control System

Experimental Physics and
Industrial Control System