Title: SAMGrid installation status
1SAMGrid installation status
He-uh-nu KIM UT-Arlington SAR-II meeting Sep
26-27 Univ. of Oklahoma at Norman
2Started with What is SAMGrid?
3SAMGrid from ordinary users point of view
At first, I could not stop asking, Why do we
need all those complicated Globus-related
packages like GRAM for Resource Management, GSI
for security GASS for stderr,stdout transfer,
MDS for Monitoring and Condor/Condor-G? Are they
really necessary?
4Happened to speculate on the meaning of using my
local machines. In order to do D0 analysis,
whats needed is that I have to log on to a
machine, compile my code, execute the compiled
executable, receive any error message, and be
able to monitor the jobs using ps ef or top
and so on,
5But I was too accustomed to realize those basic
things are actually very crucial to do some
significant jobs under Unix, and noticed that
its not so easy to implement these basics when
we are forced to need larger resources than we
have.
??? ???
Being able to use remote resources like local
resources can mean to have those basics
implemented
6So I started to realize the importance of all
those new packages. And then began to learn one
by one. (www-d0/computing/grid/)
There are 4 major components in SAMGrid
- Client Site can be our laptops
- Submission Site samgrid.fnal.gov
- Execution a single pc(heppc32) then cse-farm
- Monitoring Site
7User scriptJDF
Condor-G schedd
Grid Manager GRAM client
Browser
ApacheCGI
Broker
Cluster-OU
Site D0SAR
Gatekeeper
GSI Host-cert Sam-ser-cert
Advertise
Cluster-UTA
Job Mgr-sam
DB
MDS slapd
SAM stations
GGK
Project.sh
Batch System
SBA
Tomcat
SAM stations
Consumer.sh
Worker Node 2 Of BS
Worker Node 1 Of BS
Worker Node
User script
8Test installation on a single PC
- Started with a testbed heppc32.uta.edu
- UPD/UPS, SAM, Condor, Pesonal Certificates
- SBA Condor
- TomcatServlet(Xindice) heppc32.uta.edu7080
- Globus-Gatekeeper
- Advertising confirmed with samgrid.fnal.gov
- MDS set-up LDAP daemon running
- Sent a simple SAM-job Job-monitorring done
- Resource monitoring Many sites vs One site
9Test installation on a CSE-FARM
- UPD/UPS, SAM, Condor were all there
- SBA Condor
- Globus-Gatekeeper(v2.2) fresh one(v2.0) set up
with different port(2119 normal) - Other features, XMLDB,ADV,MDS easily done.
- We wanted to send McFarm-specific request.
- Had no clear idea how to this how McFarm will
interact with SAM, Condor etc. - Now got an idea MF-running command can be
included in the user-script which will
eventually be run by consumer-script - Now working on configuration to achieve that.
10User scriptJDF
Condor-G schedd
Grid Manager GRAM client
Browser
ApacheCGI
Broker
Cluster-OU
Site D0SAR
Gatekeeper
GSI Host-cert Sam-ser-cert
Advertise
Cluster-UTA
Job Mgr-sam
DB
MDS slapd
SAM stations
GGK
Project.sh
Batch System
SBA
Tomcat
SAM stations
Consumer.sh
User script
Worker Node 1 Of MF
Worker Node 2 Of MF
Worker Node
11Summary and Conclusion
SAMGrid installation on a single host and
small-sized farm was successful. Simple SAM job
has been sent to heppc32 Now struggling to send
McFarm request to CSE-farm via SAMGrid. I hope
this can be done very soon.
Many thanks to Gabriele, Igor, Parag!