On 06/18/2019 12:38 AM, McCormick,
Jeremy I. wrote:
[log in to unmask]"
type="cite">
Hi,
I have some good news.
I've been able to put together a platform for running the HPS
recon in parallel from the ET system.
It works using a simple client-server architecture. A server
would run on each machine where the recon should be started. A
command-line client interface can be used to start/stop
instances of the recon on each machine running the server. This
should allow us to run many instances of the recon in parallel
on the new clon machines.
I have tested the system locally on my laptop running 2-4
instances of the recon plus DQM, and I can confirm that events
are being distributed in round-robin fashion between the
different ET stations. DQM plots are written incrementally
every N events from each station and a simple server task
combines them into a single ROOT output file. I have not done
this yet, but it should be straightforward to setup a JSROOT
server in the counting house to monitor these combined plots in
a browser.
Working test scripts for this framework can be found here:
https://github.com/JeremyMcCormick/hps-projects/tree/master/projects/online-recon-test
I know we are probably some time away from getting this into
production in the counting house, but it would be good to start
testing the system as soon as possible. Likely some issues will
crop up in the counting house computing environment. We would
need to setup the et2et component so that events are written to
the secondary ET rings, for instance.
Who would like to take the lead in testing/running/managing this
system in the counting house?
I am working remotely, and I believe this should be something
that is administrated by someone who is local and working in
Hall B.
Thanks!
--Jeremy