Panda Integration Testing

Introduction

This document describes testing and QA activities in support of Panda development.

Installation

How to run?

  • cd into windmill installation directory
  • start panda server ?
  • start executor: >./launch_executor
  • type print or help
  • in a separate window, cd into windmill directory, then: >./launch_supervisor
  • type print or help

Testing

Test jobs without using prodDB

Useful for debugging the system. Nothing is read from or written to the production database. First, define jobs as comma separated list of parameters, one job per line in panda/supExecuteJobs.asc, which already contains several example jobs. Next, change the following flags in data/windmill xml:
   <fakejobs>1</fakejobs>
   <fakedb>1</fakedb>
   <oracle>0</oracle>

I have tested this with the 'rocinante' executor simulator - works fine.

Test jobs without using official tables in prodDB

This is also useful for debugging the system. Job definitions are read from panda/supExecuteJobs.asc, as above. Job execution status is recorded in a separate table in the production database. Note - this table is sporadicaly flushed. You will need to change the following flags in data/windmill xml (all other parameters use those in windmill.panda):
   <fakejobs>1</fakejobs>

I have tested this with the 'rocinante' executor simulator - works fine.

Test real jobs without executing them on the grid

During this test, real jobs will be fetched from the ATLAS database, but their execution will be faked by the 'rocinante' executor, which simulates running jobs on the grid.

Set <exetype>rocinante</exetype> in windmill.xml. Set <maxJobs> and 'Polling' parameters to smaller values ('rocinante' is quick - jobs finish in a few minutes).

Note - 'rocinante' cannot handle jobs without output files, for example the 'sleep' test jobs. It creates output xml with lfn='None'.

Test real panda 'sleep' jobs

100 jobs are defined now. Need a volunteer to run them and update this twiki with results.

Note - 'sleep' jobs have no input or output files. log files are defined to be the concatenation of stdout and stderror - which should be handled by the pilot job and named as defined in job xml. They should be catalogued in dq2?

Testing Team

Everyone!


Major updates:
-- KaushikDe - 14 Sep 2005



Responsible: KaushikDe

Edit | Attach | Watch | Print version | History: r4 < r3 < r2 < r1 | Backlinks | Raw View | WYSIWYG | More topic actions
Topic revision: r4 - 2007-02-02 - StephenHaywood
 
    • Cern Search Icon Cern Search
    • TWiki Search Icon TWiki Search
    • Google Search Icon Google Search

    PanDA All webs login

This site is powered by the TWiki collaboration platform Powered by PerlCopyright & 2008-2020 by the contributing authors. All material on this collaboration platform is the property of the contributing authors.
or Ideas, requests, problems regarding TWiki? use Discourse or Send feedback