online services logistic pp2009
Goal
monitor performance of EMC/L0/L2/cdev/scaler during 2009 run using various custom code (different from PPlot or fast-ofline), talking directly to various STAR/RHIC online systems, and posting results on the web/STARdb with latency of ~5 minutes after end of every STAR production run.
The scheme of all those services is the same. There is dedicated group account on one of the new onl01 - onl14
(.starp.bnl.gov) machines which run a pair of synchronized programs:
* "reader" which accesses (via shared disk or a socket) results of last STAR run or RHIC status, analyses it, and produces locally plots & tables
* "post-It" service which sends recent plots from .starp domain to another location seen by any web browser world wide
- We would like to have one 'onlmon' group account recognized by all 14 machines: onl01-onl14.
Jim Sowinski (sowinski,sowinski@indiana.edu) &
Matt Walker (mattheww,mwalker@MIT.EDU)
will be primary people responsible for this account and make sure all people receive correct encryption keys. - We would like all people listed in item 5.1 to receive private accounts recognized by all 14 machines: onl01-onl14.
- Private & group account will be used in the following distinct way:
*private account will be used to develop & test-run any monitoring service (used to develop new versions).
* 'onlmon' - group account will be used only for the actual official monitoring, code there must always run. The current (working) monitoring code for every service will be fully copied from private account to the one of the following directories:
~onlmon/cdev2009/
~onlmon/L0trg2009/
~onlmon/L2algo2009/
~onlmon/bemcDaq2009/
~onlmon/scaler2009/
~onlmon/zdc2009/
The L2algo2009/ will contain the following subdirectories:
ped,jet,gamma,calib,highEne,W,upsilon
(just a suggestion)
All computation results, log-files, root-histos, PS, PDF, full image of posted web pages will be created and retained in the respective ~onlmin/XXX2009/YYY directory.
(This way it will be enough to copy just this tree to have full dump of all 2009 services).
- The ~onlmon/XXX2009/ directories should be visible at all 14 machines: onl01-onl14
- The following web-ares will be created at physical location: /onlineweb/www/
http://online.star.bnl.gov/L0trg2009
http://online.star.bnl.gov/L2algo2009
http://online.star.bnl.gov/bemcDaq2009
http://online.star.bnl.gov/scaler2009
http://online.star.bnl.gov/zdc2009
http://online.star.bnl.gov/test/
- a unique disk area will be associated with every domain listed above, mounted to all 14 machines: onl01-onl14.
Only user=onlmon will have read & write access to all those area except 'test2009'.
All users listed in this mail will have read/write privileges to the 'test2009' area. - the URL for all those area will be made visible for the world
- The link to all those areas will be added to the STAR online page in the 2009 section ( to be created for this purpose).
http://online.star.bnl.gov
- a unique disk area will be associated with every domain listed above, mounted to all 14 machines: onl01-onl14.
- We would like to have unique access to 4 of the 14 machines (say 02,03,04,06) during 2009 run.
It is fine other people use them sporadically, but no other process should run continuously on those machines during 2009 data taking.
- this is tentative assignment of online tasks for those 4 machines:
onl02: L2algos
* ped: Jan, balewski@ mit.edu
* calib: Ross Corliss (rcorliss,rcorliss@MIT.EDU)
* jet : Brian (pagebs,pagebs@indiana.edu)
* gamma: Endcap *Ilya Selyuzhenkov (seluzhen,ilya.selyuzhenkov@gmail.com)
* gamma: Barrel Michael Betancourt (betan, betan@mit,edu)
* highEne: Ross
* upsilon: Haidong Liu(hdliu,hliu@student.physics.ucdavis.edu)
* W: Joe Seele (seelej,seelej@MIT.EDU)
onl03:
* cdev: Brian
* L0trig: Pibero Djawotho (pibero,pibero@comp.tamu.edu)
onl04:
*bemcDaq, Willie (wleight,wleight@mit.edu)
onl05:
*scalerA, Ernst Sichtermann, (EPSichtermann@lbl.gov)
onl06:
*scalerB, Christopher Jones (cjones,c_jones@MIT.EDU)
*E+BEMC pedestals for offline DB, Justin Stephens (stevens4, stevens4@indiana.edu)onl07: ZDC, Oleksandr Grebenyuk, ogrebenyuk@lbl.gov
- this is tentative assignment of online tasks for those 4 machines:
- Special needs, should be followed up by interested parties:
*onl03 needs to have special library allowing communication with CAD cdev server. For L0trig, the only requirements are access to
the trigger scratch disk via NFS.
* onl02, onl03 need access to online L2 machine
* onl04 needs to be able to connect via socket with Tonko's DAQ machine
* onl05 needs to see disk where scaler files are written.
* the http://online.star.bnl.gov/test2009/ will be used by all to test their code, as needed. - every expert listed in 5.1 will provide a short writeup of the service he maintains on separate Drupal page, available on day 1, listing all code names, disk locations, intermediate file names.
Also full procedure to abort & restart given service. - Offline E+BEMC tower peds will be produced from L2ped histos on a none-online machine accessing info via web-interface to L2ped page. Adam developed code in 2008 for the Barrel, Justin will duplicate it for the Endcap. ( Or perhaps merge & maintain both ?) TBD
- spin.lbl.gov has for several years for near real-time analysis of BBC and other scaler data. Most likely this service will continue, TBD.
- Expected footprint. Based on past years experience the total volume exported to the
/onlineweb/www/ area may approach 1-2 GB.
It will grow very linear with time, say 100MB/week.
The local files on onlNN machines will be ~10x larger, but still initial quota of 1GB per person should be enough. - Estimated disk usage for 15 weeks of good runing , based on 2009 data
End-of-run9 cleanup status, December 2, 2009
service | person |
web-disk (GB) *) |
local-disk (GB) **) |
pp500+200 archived |
remarks |
---|---|---|---|---|---|
cdev | Jim | 0.64 | |||
L0trig | Pibero | ||||
L2algo | Mike | 3.3 | 2.0 | path2 | ok |
BSDM | Will | 4.1 | 21 | path2 | ok |
scaler | James | 8.6 | |||
zdc | Oleksandr | N/A | 3.3 zip | path1 | ok |
PPlot | Renee? | - | ? | ||
*) web-path= /onlineweb/www/
**) local-path= user onlmon
path1= HPSS: /home/ogrebeny/zdc/ zdc2009_pp200.tar & zdc2009_pp500
path2= HPSS: /home/balewski/2009-online-dump/
- balewski's blog
- Login or register to post comments