[Halld-offline] [EXTERNAL] Re: Building versions for Nacer, notes on analysis launch 1 (of 4).

Wed May 6 17:59:47 EDT 2020

I want to note briefly what has to be true for the analysis version set to appear on the webform.

1)there must be an entry in the correlation xml file

2) every package must exist on the OSG. Rather, the local directory that gets syncd

The database gets recreated once a day early in the morning. So if magically it’s happens now it’ll be available tomorrow. I can trigger it manually if needed in a pinch

Thomas Britton

On May 6, 2020, at 5:53 PM, jlab <nacer at jlab.org> wrote:

Hi all,

Great ! seems we agree on the recon version, now what about analysis.xml version (the analysis launch for 2016ver20 wiki says analysis-2017_01-ver20.xml) bu MCwrapper does not find it !

— Nacer

On 6. May 2020, at 23:26, Alexander Austregesilo <aaustreg at jlab.org<mailto:aaustreg at jlab.org>> wrote:

There is no such xml file, and if there was, it would not be useful for your purpose. He just has to use the latest recon-2017_01-ver03_*.xml, same as for spring 2017. We could provide a link to fulfill the naming convention.

On 5/6/2020 5:19 PM, Mark Ito wrote:
Sean and Alex,

Wow. You guys have good memories!

I will go on to 2 of 4 until someone finds that file.

  -- Mark

On 5/6/20 5:06 PM, Sean Dobbs wrote:
Mark,

If I remember correctly, 2016-02ver06 was reconstructed with the same
software as 2017-01ver03 at roughly the same time.  If we check the
website:
https://halldweb.jlab.org/cgi-bin/data_monitoring/monitoring/dataVersions.py
recon2016-02ver06 was reconstructed with the version set
version_recon_2016_02_ver06.xml  (I think there was a typo on the
page).  The link is broken, but if we can track the file down, that
would answer the question.

Data for analysis launches should be stored in here as well. but
perhaps that's a problem for version 2 of this database.

---Sean

On Wed, May 6, 2020 at 4:57 PM Mark Ito <marki at jlab.org<mailto:marki at jlab.org>> wrote:
Folks,

The first analysis launch on Nacer's list of four is "2016ver20". Here
is the web page:

https://halldweb.jlab.org/wiki-private/index.php/Spring_2016_Analysis_Launch#Version20

There is no version information there, but it does name the REST
production run it was produced from. That is REST ver06. From that page:

    REST ver06 (August 2018), runs 11366 - 11555 ('golden period')

That reconstruction launch appears on this web page:

https://halldweb.jlab.org/wiki-private/index.php/Spring_2016_Dataset_Summary#Full_REST_Production

It is the fourth line from the bottom. The link on that line leads to
this page:

https://halldweb.jlab.org/data_monitoring/recon/summary_swif_output_recon_2016-02_ver06_batch01.html

which does not contain any version information.

If I list tags from halld_recon that match *recon*2016* I get:

    [marki at markdesk4 halld_recon]$ git tag | grep 2016 | grep recon
    recon-2016_02-ver01
    recon-2016_02-ver02
    recon-2016_02-ver03
    recon-2016_02-ver04

If I do the same search for tags in sim-recon, I get:

    [marki at markdesk4 sim-recon]$ git tag | grep 2016 | grep recon
    recon-2016_02-ver01
    recon-2016_02-ver02
    recon-2016_02-ver03
    recon-2016_02-ver04

Recall that halld_recon started as a clone of sim-recon. So we expect
tags that were applied to sim-recon to appear in both sim-recon and
halld_recon. So these four tags were all initially applied when we were
still using sim-recon. Note there is no tag for recon-2016_02-ver06.

I have been making backups of sim-recon and halld_recon and writing them
to tape before deleting them from the group disk. If I look in the tape
library I see:

    lorentz:marki:releases> ls *2016_02*
sim-recon-recon-2016_02-ver03_Linux_CentOS6-x86_64-gcc4.9.2.tar.gz
sim-recon-recon-2016_02-ver03_Linux_CentOS7-x86_64-gcc4.8.5.tar.gz
sim-recon-recon-2016_02-ver03_Linux_RHEL6-x86_64-gcc4.9.2.tar.gz
sim-recon-recon-2016_02-ver03_Linux_RHEL7-x86_64-gcc4.8.5.tar.gz
sim-recon-recon-2016_02-ver04-Linux_CentOS7-x86_64-gcc4.8.5.tar.gz

so again no 2016_02-ver06. And there is no CentOS7.7 build and no
container build either. Also note that launches are were _never_ run
from builds on the group disk (except for on the OSG and at NERSC). The
launch-meister always builds his own version on /work/halld. But I think
that that is the least of our worries.

My best guess is that ver06 was built with the ver04 tag, and perhaps
run with updated calibration constants or software settings. So I can
try to build the ver04 tag. I think that that should work.

Now which version of halld_sim should I use to go with it? Two choices:

A) the latest halld_sim

B) the simulation code that is contained in the old sim-recon

and perhaps there are others. In case (A) there is no guarantee that a
modern halld_sim will build against an old halld_recon, at least not
without a fight. In case (B), that is a very old version of the simulation.

Another question: which version of hdgeant4 should I use. Recall hdgean4
depends on halld_recon.

One final point. The intention is to get all of this running with
MCwrapper. There may or may not be issues there.

Thoughts?

    -- Mark

P. S. For the record, the other three analysis launches Nacer mentions
are 2017ver21, 2018_08ver02, and 2018_01_ver03.

_______________________________________________
Halld-offline mailing list
Halld-offline at jlab.org<mailto:Halld-offline at jlab.org>
https://mailman.jlab.org/mailman/listinfo/halld-offline
_______________________________________________
Halld-offline mailing list
Halld-offline at jlab.org<mailto:Halld-offline at jlab.org>
https://mailman.jlab.org/mailman/listinfo/halld-offline

_______________________________________________
Halld-offline mailing list
Halld-offline at jlab.org
https://mailman.jlab.org/mailman/listinfo/halld-offline
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://mailman.jlab.org/pipermail/halld-offline/attachments/20200506/d2963886/attachment-0002.html>