PandaSchedConfig

Introduction

The schedconfig table is filled with information from a variety of sources. It is filled by autopilot/pilotController.py which will overwrite most manual changes to the Db. It is used by the Panda server, AutoPilot and the pilot itself from the WN. Each row represents a queue to which pilots can be submitted. There is 1 or more queues per site.

Modifying schedconfig

For information on modifying schedconfig see the ' Managing the AutoPilot databases' section of the AutoPilot twiki.

Sources

Sources of information are:
  • Top-level EGEE BDII: site info, releases, GOC name
  • TiersOfATLAS: ddm site, SRM endpoint fo SE host and path
  • pilotController.py: cloud structure, local overrides, e.g. stage-in tool dccp
  • SAM Db: site availability, maintenace info - Not yet implemented

The ddm site to GOC site mapping is currently done manually in pilotController.p but can be done from ToA 'alternateName'.

Table fields

schedconfig is a table in the PandaMetaDB Db. I try to split the 57 fields into groups.

Identifiers

  • siteid - Randomly assigned human name for a site, migrating to GOC name for EGEE sites.
  • nickname - constructed from the GOC name, the short CE hostname, queue and jobmanager type, e.g. RAL-LCG2-lcgce02-grid700M-lcgpbs
  • cloud

Data Handling

  • ddm - ddm site name as in TiersOfATLAS(ToA)
  • lfchost - LFC_HOST hostname, usually the at T1 of the cloud. Taken from ToA.
  • se - a comma seperated list of the SE(s) local to the site srm://heplnx204.pp.rl.ac.uk
  • seprodpath - Output files are stored in this physical directory. Default is base atlas path on SE plus /dq2. Can vary for local disk pool reasons.
  • lfcprodpath - The long lfn starts with /grid/atlas/dq2
  • sepath - as above but for analysis, base plus /users/pathena
  • lfcpath - /grid/atlas/users/pathena
  • copytool -
  • copytoolin -
  • copyprefix -
  • copyprefixin -

Pilot Submission

  • gatekeeper - the hostname
  • jobmanager - the part after jobmanager-, e.g. lcgpbs,pbs,sge,lcgcondor
  • queue - concatenation of above, e.g. lcgce02.gridpp.rl.ac.uk/jobmanager-lcgpbs ( WHY BOTH?)
  • localqueue - the queue name, often "atlas"
  • nqueue - the number of pilots to keep queued

Pilot modifiers

  • status [online|offline|draining] - the value seen by the pilot is an OR of site status from schedconfig and cloud status from the cloud table.
    • draining - pilots ask jor jobs, server gives jobs, no new jobs assigned to the site
    • offline - pilots do not ask for jobs, server does not give job(?), no new jobs assigned to the site. The pilot sees offline if the schedconfig.status='offline' OR the cloud status from the cloud table is offline.
  • recoverdir
  • retry
  • envsetup
  • envsetupin


Major updates:
-- RodWalker - 26 Nov 2007



Responsible: RodWalker

Never reviewed

Edit | Attach | Watch | Print version | History: r3 < r2 < r1 | Backlinks | Raw View | WYSIWYG | More topic actions
Topic revision: r3 - 2008-04-17 - RodWalker
 
    • 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