[Halld-online] Comments on DAQ system

[Alexander Somov]

Hi David,

I just wanted to say that if you hardcode the host of the EB
(for instance) in the cool db, you'll need to make another
cool configuration or modify the existing one if you decide to change
the host name. As the run control doesn't care about the host
name it can be specified in the start up script (i.e., you can
run it on any hosts without changing the cool db).

Cheers,
        Alex





On Thu, 17 Apr 2014, David Lawrence wrote:

>
> Hi Alex,
>
>  I’m not sure if you’re agreeing or disagreeing with me so I’ll just assume you’re agreeing with me ;)
>
>  In case my point wasn’t clear before, let me just restate it for the wider audience:
>
> The CODA COOL configuration allows us to specify names for each coda component including
> the coda_roc processes running on all of the DAQ crates (ROCs). It also does this for the Event Builder (EB),
> Event Recorder (ER) and Farm Managers (FCS). It does not allow us to specify the host that
> these components are to run on. We currently use a convention that the name of the ROC
> components is the hostname of the computer it is to run on (e.g. ROCBCAL1). This convention
> cannot be used, however for the other non-ROC components that need to run on machines like
> gluon46 or gluonraid1. Therefore, the hostnames that the startup script should use to ssh
> into in order to start up these non-ROC components must come from somewhere else. Either
> we hardcode them into the startup scripts, or we try and embed them in the COOL configuration
> somewhere (possibly as RTVs). The current system of just running them all on whatever host
> the startup script is run on will work for all of the small, detector DAQ systems we’re using now.
> However, we need this additional functionality in the not-too-distant future so I thought it would be
> worth thinking about it just a little now to avoid a lot of refactoring later.
>
> The farm node configuration question is one we will need to address soon as well.
>
> Regards,
> -David
>
> On Apr 17, 2014, at 1:33 PM, Alexander Somov <somov at jlab.org> wrote:
>
>>
>> Hi,
>>
>>> 3. I mentioned this to you earlier, but I’ll just repeat it here. The current system automatically
>>>   starts EB, platform, and rcgui on whatever machine setups.sh is run on. This is not such
>>>   a problem for the platform and rcgui, but things like the EB, ER, and FCS will likely
>>>   be run on specific hosts. We should invent some mechanism for specifying this. I think
>>>   adding this to the COOL configuration might be best since we already use information
>>>   from there to tell where the coda_roc processes are run. However, you may have other
>>>   thoughts.
>>
>> We can start EB/ER using the daq start up script from any predefined hosts
>> similar to coda_roc's. The Run Control is currently using daq component names only (rather than hostnames).
>>
>>
>> I also think that it would make sense to have only one configuration
>> for each specific hardware and handle parameters (like farm settings)
>> through RTVs
>>
>> Cheers,
>>       Alex
>
>