File - GARUDA Access Portal

File - GARUDA Access Portal
Kepler Workflow on GARUDA Grid
User Guide
ver 1.0
June 2010
Submitted to
Zeus Numerix Pvt. Ltd.
&
EADS (Bangalore, India)
CENTRE FOR DEVELOPMENT OF ADVANCED COMPUTING
C-DAC Knowledge Park,
No. 1, Old Madras Road, Byappanahalli,
Bangalore – 560 038.
Tel : +91-80-2534 1909
Fax: + 91-80-2524 7724
www.cdacb.in
Table of Contents
Chapter 1 - Introduction...................................................................................................1
1.0
1.1
1.2
1.3
1.4
KEPLER ...............................................................................................................1
Kepler Features.........................................................................................1
GARUDA Grid.........................................................................................1
Kepler on GARUDA Grid........................................................................2
Pre-requisite software ...............................................................................2
Chapter 2 - Usage Instructions.........................................................................................3
2.0
2.1
................................................................3
Procedure for Job Submission on GARUDA ...................................................4
PROCEDURE FOR ACCESS TO GARUDA
Chapter 3 - Testing ............................................................................................................8
Chapter 4 - Frequently Asked Questions ......................................................................10
Appendix 1 – Detailed Kepler User Manual .................................................................13
Appendix 2 – GARUDA Compute Reservation Commands .......................................14
Appendix 3 - Resource Specification Language ..........................................................15
Appendix 4 - MyProxy ....................................................................................................16
Kepler Workflow User Guide, June 2010, Version 1.0
Page i
Chapter 1 - Introduction
1.0
Kepler
Kepler is free and open source, scientific workflow designed to help scientists,
analysts, and computer programmers create, execute, and share models and analyses
across a broad range of scientific and engineering disciplines. The Kepler software is
developed and maintained by the cross-project Kepler collaboration, which is led by a
team consisting of several key institutions that originated the project: UC Davis, UC
Santa Barbara, and UC San Diego. Kepler is a java-based application that is
maintained for the Windows, OSX, and Linux operating systems. The Kepler Project
supports the official code-base for Kepler development, as well as provides user
manuals, mechanism to report bugs, suggest enhancements, discussion forum, etc.
Kepler allows the user to create a workflow and execute the same from GUI. Once
created a workflow can also be run from command line as it allows saving a workflow
in xml or kar format and executed via command line.
1.1
1.2
Kepler Features
Kepler provides a graphical user interface and a run-time engine that can execute
workflows either from within the graphical interface or from a command line.
Kepler uses a director/actor metaphor to visually represent the various components
of a workflow. A director controls (or directs) the execution of a workflow. The
actors take their execution instructions from the director. In other words, actors
specify what processing occurs while the director specifies when it occurs.
Actors are ready-to-use processing components that can be easily customized,
connected and then run to perform complex applications efficiently. Kepler has a
searchable library containing nearly 350 actors.
Kepler workflows can be nested, allowing complex tasks to be composed from
simpler components.
Kepler workflows can leverage the computational power of grid technologies (e.g.,
Globus, SRB, etc). It supports job submission actor for Globus, MyProxy actor for
authentication, and GridFTP actor for file transfer.
Kepler workflows and customized components can be saved, reused, and shared
with colleagues using the Kepler ARchive format (KAR).
GARUDA Grid
GARUDA is a nation wide grid spread across 17 cities of the country, comprising of
collaboration of science researchers, computational nodes, mass storage and scientific
instruments, with an aim to provide the technological advances required to enable data
and compute intensive science for the 21st century.
GARUDA is based on the Globus Toolkit 4.0.7. The Globus Alliance has adopted the
Open Grid Standards Infrastructure (OGSI) standards based on convergence of Web
Kepler Workflow User Guide, June 2010, Version 1.0
Page 1
Services and Grid Computing technology. The various technology components of
GARUDA include a Portal for access and job submission, program development tools,
scheduler, reservation manger, and storage resource manager, and tools for monitoring
and management of the system. Access to GARUDA is through certificates obtained
from the Indian Grid Certification Authority (IGCA).
The GARUDA grid has a grid head node to which various compute resources are
connected. The cluster under consideration for this project is composed of - Xeon
Quad Core Dual CPU Processor X5460 nodes, having 16GB RAM, connected by
Infiniband with RHEL 5.1 on Rocks v5.0. The GARUDA grid can be accessed from
Internet through the gateway machine.
1.3
Kepler on GARUDA Grid
Kepler development version (2.0) has been identified based on discussion with the
Kepler Forum, as it provides Job Submission Actors for Globus. The Kepler source
code was compiled on the grid head node (referred to as Gridfs – IP as
203.200.36.236). The Kepler actor packages are deployed on Gridfs at
/usr/local/EADS_AAOW.
The job submission actor requires inputs:
Job script content - which is an RSL script,
Cluster name,
Globus path, and
Scheduler type.
The executable name needs to be mentioned in the rsl script tag as
<executable>executablename</executable>.
For Kepler to be visible from remote machine, a visual sharing tool like XManager
needs to be used. The X11 on gridfs machine has been enabled for desktop sharing.
1.4
Pre-requisite software
•
JAVA ver 1.6
•
ANT ver 1.8
•
Xmanager 2.0 or greater on client machine
Kepler Workflow User Guide, June 2010, Version 1.0
Page 2
Chapter 2 - Usage Instructions
2.0
Procedure for Access to Kepler on GARUDA
1. Kepler is installed at grid headnode (gridfs - 203.200.36.236) at
/usr/local/EADS_AAOW
–
ssh 203.200.36.236 -l username
2. For remotely accessing the Kepler GUI following are the steps:
a)
Invoke the Xmanager client on local desktop (Xstart)
b)
In the Xstart dialog box, the Execution Command field is used to
specify whether to open complete remote desktop or only the remote
terminal. Hence, Execution Command can have:
c)
–
Execution Command = /usr/bin/gnome-session --display
$DISPLAY
–
Execution Command =
xterm
Fill out the Xstart dialog box and Run
Figure 1: A sample Xstart Dialog Box
Kepler Workflow User Guide, June 2010, Version 1.0
Page 3
d)
e)
Set the Environment variable and Path as below:
–
export JAVA_HOME=/usr/local/jdk1.6.0_10
–
export ANT_HOME=/usr/local/EADS_AAOW/ant/apache-ant-1.8.0
–
export PATH=$PATH:$ANT_HOME/bin
Invoke Kepler through
$./kepler.sh
2.1
Procedure for Job Submission on GARUDA
Step 1: Reserving the resources
When a person needs to submit a job to GARUDA, he should first identify the
resource. RESERVATION of the GARUDA resources needs to be performed. The
Reservation ID obtained should be saved to be mentioned in the RSL file or Job script
content.
Reservation command:
$ garuda_compute_reserve gg-blr.tfg "2010-04-20T11:30:00" "2010-0421T18:58:28" 2
Output of Reservation Command:
$ R714.gg-blr.tfg
Step 2: Check the credentials
Submitting a job requires a valid proxy which is nothing but a short lived credential.
The user’s credentials should be alive otherwise it needs to be invoked.
Proxy information can be obtained with the help of “grid-proxy-info” command
$ grid-proxy-info
subject : /DC=IN/DC=GARUDAINDIA/O=C-DAC/OU=CTSF/CN=Karuna
([email protected])/CN=237891045/CN=2012224773/CN=744875709
issuer
: /DC=IN/DC=GARUDAINDIA/O=C-DAC/OU=CTSF/CN=Karuna
([email protected])/CN=237891045/CN=2012224773
identity : /DC=IN/DC=GARUDAINDIA/O=C-DAC/OU=CTSF/CN=Karuna
([email protected])
type
: Proxy draft (pre-RFC) compliant impersonation proxy
strength : 1024 bits
path
: /tmp/x509up_u502
timeleft : 0:00:00
Kepler Workflow User Guide, June 2010, Version 1.0
Page 4
Proxy can be created by the “grid-proxy-init” command
$ grid-proxy-init
Your identity: /DC=IN/DC=GARUDAINDIA/O=C-DAC/OU=CTSF/CN=Karuna
([email protected])
Enter GRID pass phrase for this identity:
Creating proxy .............................................
Done
Your proxy is valid until: Tue Jun 8 02:38:47 2010
The job will get a limited proxy
Step 3: Upload credentials to MyProxy
Now user has valid credentials to submit a job in GARUDA grid. Certificates should
be available in My-proxy server.
$ myproxy-init
Your identity: /DC=IN/DC=GARUDAINDIA/O=C-DAC/OU=CTSF/CN=Karuna
([email protected])
Enter GRID pass phrase for this identity:
Creating proxy ...............................................
Done
Proxy Verify OK
Your proxy is valid until: Mon Jun 14 14:40:18 2010
Enter MyProxy pass phrase:
Verifying - Enter MyProxy pass phrase:
A proxy valid for 168 hours (7.0 days) for user gridmon now
exists on hyd01.
For retrieving the certificates.
$ myproxy-logon -s hyd01
Enter MyProxy pass phrase:
A credential has been received for user gridmon in
/tmp/x509up_u502.
For complete usage instructions of MyProxy please refer Appendix 4.
Kepler Workflow User Guide, June 2010, Version 1.0
Page 5
Step 4: Create Job Description
The job description file has to be input into the Globus job submission actor. Globus
uses a Resource Specification Language (RSL) script to describe the executable, input,
output and other parameters of the job.
An Example RSL file is shown below:
<job>
<executable>/bin/echo</executable>
<argument>Welcome to C-DAC</argument>
<environment>
<name>LD_LIBRARY_PATH</name>
<value>/home/soademo/lib</value>
</environment>
<queue>batch</queue>
</job>
For complete syntax of the Resource Specification Language (RSL) Script file to be
used on Globus based GARUDA grid please refer Appendix 3.
For each job submission actor in the workflow the corresponding job description file
can be obtained by tweaking the sample RSL above. The RSL has to be given as input
in the ‘Job Script Content’ of the Kepler GlobusWSJob actor.
Step 5: Create the workflow using Kepler actors
Kepler uses the GlobusWSJob actor for submitting jobs to Globus.
GlobusWSJob actor parameters are (as shown in figure)
1 . Job Script Content
2. Globus Host
3. Batch Mode
4. Job Scheduler Type
5. GLOBUS_LOCATION Path
6. AXISClientConfigFilePath
Kepler Workflow User Guide, June 2010, Version 1.0
Page 6
1. Job Script Content: This parameter gives the details of the job to be executed by
the particular actor. The complete RSL file can be copy-paste into this field. Please
ensure that no extra white spaces are included inadvertently.
2. Globus Host: This parameter indicates which machine (cluster) in which the job
should execute. As it is desired to run on the GG-BLR cluster of Garuda grid, please
mention ‘gg-blr.tfg’ in this field.
3. Batch Mode: The default is taken as ‘batch’ job.
4. Job Scheduler Type: Job scheduling on the GG-BLR cluster is accomplished by
the Portable Batch System (PBS). The appropriate job scheduler should be given in
this parameter.
5. GLOBUS_LOCATIO: Path: The absolute path on the cluster (GG-BLR in this
case) where Globus is installed.
6. AXISClientConfigFilePath: The absolute path of the client-config.wsdd in the
GLOBUS_LOCATION
Step 6: Executing the Workflow
Once the workflow is created it can be run from the GUI, click on the play(||>)
button from the toolbar.
Also the workflow can be run through the command line also with the following
command:
ant run-workflow-no-gui -Dworkflow=<saved workflow name>
E.g.
ant run-workflow-no-gui -Dworkflow=<sample.xml>
ant run-workflow-no-gui -Dworkflow=<myworkflow.kar>
ote: A workflow can be saved from the File->Save menu in .xml or .kar
Kepler Workflow User Guide, June 2010, Version 1.0
Page 7
Chapter 3 - Testing
Exhaustive testing has been done with the Kepler deployed on GARUDA. Workflows
containing combination of both serial and parallel (MPI) jobs were tried out. Also
Kepler was tested for access and job submission from different user logins.
Testing of the Kepler tool has been carried out with respect to the following:
•
Job submission Actor of Kepler for both sequential and parallel job submission
•
MPI and Pthread jobs
•
RSL with Reservation ID
•
GARUDA certificates, MyProxy
•
Small workflow submission
•
Both GUI and command line
•
Kepler toolbars through remote visualization
•
Access via Internet
Figure 1: A sample workflow using SDF* director using
GloubsWSJob and GridFtpCopy actor.
* SDF – Synchronous Dataflow (Refer Chapter 4 FAQ 5)
Kepler Workflow User Guide, June 2010, Version 1.0
Page 8
Figure 2: Submitting of jobs through GUI
Figure 3: Submitting of jobs using P,* director.
* PN – Process Network (Refer Chapter 4 FAQ 5)
Kepler Workflow User Guide, June 2010, Version 1.0
Page 9
Chapter 4 - Frequently Asked Questions
1. How can I start kepler?
Login to gateway m/c through ssh and to gridfs m/s
$ cd /usr/local/EADS_AAOW/
$./kepler.sh
Before this pls check if the JAVA_HOME is set to usr/local/jdk1.6.0_1
And ANT_HOME is set to =/usr/local/EADS_AAOW/ant/apache-ant-1.8.0
And path set as $ANT_HOME/bin
2. How can I invoke the grid actors.
Since the globus4.0.7 is not present in the kepler repository, it has to
instantiated onto the workflow canvas.
be
manually
To instantiate a component: on the menu bar
Tools ->Instantiate component->Type in the correct class name of the
to be instantiated
component
For instance some of the globus actors and their classnames
Actor
classname
GlobusWSJob
org.kepler.actor.globus.wsgram.GlobusWSJob
MyProxy
org.kepler.actor.MyProxy
GridFtpCopy
org.kepler.actor.globus.GridFtpCopy
3. How to connect the actors
Once the actors have been instantiated on the workflow canvas, each actor has to be
connected, which can be done by clicking on of the port and dragging it until the other
actor’s port.
4. How can an actor be configured?
Each actor has to be configured, which can be done either by double-clicking the actor
or by right-click->configure actor. Enter the appropriate values and commit the
entries.
5. What is a director?
Kepler uses a director/actor metaphor to visually represent the various components of
a workflow. A director controls (or directs) the execution of a workflow, just as a film
director oversees a cast and crew. The actors take their execution instructions from the
Kepler Workflow User Guide, June 2010, Version 1.0
Page 10
director. In other words, actors specify what processing occurs while the director
specifies when it occurs.
Each workflow must have a director that controls the execution of the workflow using
a particular model of computation. For example, workflow execution can be
synchronous, with processing occurring one component at a time in a pre-calculated
sequence (SDF Director). Alternatively, workflow components can execute in parallel,
with one or more components running simultaneously (which might be the case with a
PN Director). For more details refer Kepler User Manual (Appendix 1).
6. How to run a workflow?
To run the workflow, click on the play(||>) button from the toolbar. Otherwise the
workflow can be run through the command line also with the following command
ant run-workflow-no-gui -Dworkflow=Sample.xml
7.
Can I run the workflow from command line?
With the following comman, gui will not be invoked and you can run workflow ant
run-workflow-no-gui -Dworkflow=Sample.xml
8. How to make an RSL file?
A sample rsl file is listed:
<job>
<executable>testgg</executable>
<environment>
<name>GARUDA_RESV_ID</name>
<value>R714.gg-blr.tfg</value
</environment>
<stdout>/home/tools/gridmon/kepler/stdout/testgg23.out.wrapper<
/stdout>
<stderr>/home/tools/gridmon/kepler/stdout/testgg23.err.wrapper<
/stderr>
<count>8</count>
<jobType>mpi</jobType>
<queue>default</queue>
</job>
9. What is the command to obtain the reservation ID.
garuda_compute_reserve gg-blr.tfg "2010-04-20T11:30:00" "2010-04-21T18:58:28" 2
Refer Appendix 2
10. What are the parameters of the GlobusWSjob actor?
GlobusWSJob actor parameters are
1 . Job Script Content
Kepler Workflow User Guide, June 2010, Version 1.0
Page 11
2. Globus Host
3. Batch Mode
4. Job Scheduler Type
5. GLOBUS_LOCATION Path
6. AXISClientConfigFilePath="/usr/local/GARUDA/GLOBUS-4.0.7/clientconfig.wsdd"
11. Can workflows be saved and run at later point of time?
It can be saved and run as ant run-workflow-no-gui -Dworkflow=Test.xml
Kepler Workflow User Guide, June 2010, Version 1.0
Page 12
Appendix 1 – Detailed Kepler User Manual
The project website (http://www.kepler-project.org/) contains the complete source
code and documentation for Kepler Workflow software.
The Kepler User manual is available at: https://kepler-project.org/users/documentation
Kepler Workflow User Guide, June 2010, Version 1.0
Page 13
Appendix 2 – GARUDA Compute Reservation Commands
GARUDA grid compute reservations commands
The commands related to grid compute node reservation are located at
/opt/garudaresv/bin directory of gridfs.ctsf.cdac.org.in machine.
Command for getting the free resources in a particular time slot.
/opt/garudaresv/bin/garuda_freeresources can be used to find the free
compute nodes available in a particular time slot. This command will return the cluster
FQDN with number of free cpus.
Usage: garuda_freeresources "<start_time>" "<end_time>"
Example: garuda_freeresources "2008-08-05T14:38:28" "2008-08-05T14:58:28"
The date and time that appears as single digit must be made to double digit by
prepending ‘0’. For example, August 5, 2008 must be specified as 2008-08-05 and not
2008-8-5. Same is applicable to time also.
Making a compute reservation
The command /opt/garudaresv/bin/garuda_compute_reserve can be used to create a
compute reservation. This command will return the grid reservation id for the
corresponding reservation slot.
Usage: garuda_compute_reserve <Cluster_FQDN> "<start_time>" "<end_time>"
<number_of_CPU>
Example: garuda_compute_reserve gg-hyd.cdac.org.in "2008-08-05T14:38:28" "200808-05T14:58:28" 1
Job submission with reservation id
The command /opt/garudaresv/bin/garuda_job_submit can be used to submit jobs in a
cluster corresponding to a compute node reservation. The command must be used to
submit job only when the corresponding reservation slot is running or active.
Usage: garuda_job_submit <garuda_resv_id> <gridway_job_template>
Example: garuda_job_submit 99.gg-hyd.cdac.org.in /tmp/test.jt
Modifying a grid compute reservation
/opt/garudaresv/bin/garuda_compute_reserve_modify command can be used to modify
an unexpired compute node reservation.
Usage: garuda_compute_reserve_modify <garuda_resv_id> <number_of_nodes>
Example: garuda_compute_reserve_modify 99.gg-hyd.cdac.org.in 1
Kepler Workflow User Guide, June 2010, Version 1.0
Page 14
Appendix 3 - Resource Specification Language (RSL)
Common notation for exchange of information between components. RSL provides
•
Resource requirements: Machine type, number of nodes, memory, etc.
•
Job configuration: executable, path, arguments,environment, etc.
Important RSL tags
Argument
count
Directory
environment
Executable
job
jobType
stdin
stdout
stderr
queue
Simple RSL example
<job>
<executable>/bin/echo</executable>
<argument>Welcome to C-DAC</argument>
<environment>
<name>LD_LIBRARY_PATH</name>
<value>/home/soademo/lib</value>
</environment>
<queue>batch</queue>
</job>
RSL (MPI Job)
<job>
<executable>/tmp/hello</executable>
<argument>C-DAC</argument>
<directory>/home/soademo/tools</directory>
<stdin>/dev/null</stdin>
<stdout>stdout.${GLOBUS_USER_NAME} </stdout>
<stderr>stderr.execution</stderr>
<count>2</count>
<jobType>mpi</jobType>
</job>
For complete syntax of the Resource Specification Language (RSL) refer
http://www.globus.org/toolkit/docs/4.0/execution/wsgram/schemas/gram_job_descript
ion.html
Kepler Workflow User Guide, June 2010, Version 1.0
Page 15
Appendix 4 - MyProxy
MyProxy Usage Instructions
MyProxy is a open source software for managing X.509 Public Key Infrastructure
(PKI) security credentials (certificates and private keys). It combines an online
credential repository with an online certificate authority. It allows users to securely
obtain credentials when and where needed. The advantage of using MyProxy is that
the user need not keep their Certificates on Grid; instead they can upload and get
certificate from the secured Myproxy Server whenever required.
Storing a credential in the MyProxy repository
Rather than storing your X.509 credentials (certificate and private key) on each
machine you use, you can store them in a MyProxy repository and retrieve a proxy
credential from the MyProxy repository when needed.
To store a credential in the MyProxy repository, run the myproxy-init command on a
computer where your Grid credentials are located. For example:
$ myproxy-info
username: gridmon
owner: /DC=IN/DC=GARUDAINDIA/O=C-DAC/OU=CTSF/CN=Karuna
([email protected])
timeleft: 0:00:00
$ myproxy-init
Your identity: /DC=IN/DC=GARUDAINDIA/O=C-DAC/OU=CTSF/CN=Karuna
([email protected])
Enter GRID pass phrase for this identity:
Creating proxy ..........................................Done
Proxy Verify OK
Your proxy is valid until: Mon Jun 14 14:40:18 2010
Enter MyProxy pass phrase:
Verifying - Enter MyProxy pass phrase:
A proxy valid for 168 hours (7.0 days) for user gridmon now
exists on hyd01.
The myproxy-init command prompts first for the pass phrase of your private key
(similar to grid-proxy-init) and then prompts twice for a new pass phrase to use to
secure the credentials on the MyProxy server. By default, the credential is stored under
your Unix username for 7 days and can be used to retrieve credentials with 12 hour
lifetimes.
Kepler Workflow User Guide, June 2010, Version 1.0
Page 16
Retrieving a credential from the MyProxy repository
Once you've stored a credential in the MyProxy repository, you can retrieve a proxy
credential whenever you need one with the myproxy-logon command. For example:
$ myproxy-logon -s hyd01
Enter MyProxy pass phrase:
A
credential
has
been
/tmp/x509up_u502.
received
for
user
gridmon
in
The myproxy-logon command prompts for the pass phrase you set previously with
myproxy-init, retrieves a proxy credential for you, and stores it in the correct default
location for use with other Globus Toolkit programs.
The detailed list of MyProxy Command reference can be found at
http://www.globus.org/toolkit/docs/4.0/security/myproxy/Cred_Mgmt_MyProxy_Inter
face_Commandline_Frag.html or MyProxy Information
Kepler Workflow User Guide, June 2010, Version 1.0
Page 17
Was this manual useful for you? yes no
Thank you for your participation!

* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project

Download PDF

advertisement