How to (Publish and) Find SC Resources

Overview

With the start of the service phase of Service Challenge 3, there should no longer such a split between service challenge resources and "normal" LCG production resources. In particular, while there might still be dedicated storage resources, due to network topology and connectivity, computing resources will the same as for LCG production. The LCG information system (lcg-bdii.cern.ch) will be used to publish and find all resources.

This document summarises the steps that a site should follow in order to publish their SC resources, as well as the configuration parameters and tools needed in order to find these resources.

Resource Publication

This consists simply of publishing an extra Runtime variable in the BDII for the CE, setting the SC3 SRM (if a separate one is used) as an additional closeSE for that CE and setting some additional environment variables for the job environment in the worker nodes to allow SC3 jobs to find the location of the SC3 SE.

This is a procedure that will need to be applied that are participating (i.e. Tier-0, Tier-1 and Tier-2 sites) in the service challenges and are providing compute or storage resources.

Procedure

  1. Publish an addition 'GlueHostApplicationSoftwareRunTimeEnvironment' attribute for the CE. This should take the value of LCG_SC3. If using yaim, this can be done by adding LCG_SC3 to the variable $CE_RUNTIMEENV in the site-info.def
  2. Publish the SC3 SRM as a close SE. If using yaim, this can be done by adding another SE to the list of close SEs ($CE_CLOSE_SE) and defining the hostname for that SE (no access point is needed since it's an SRM)
  3. For each supported VO, publish the environment variable $VO_SC3_<VO>_DEFAULT_SE. This is similar to the variable $VO_DEFAULT_SE that is already set. There is currently no direct support to do this in yaim, but it will be added in future versons. Extending config_lcgenv.sh (or adding directly to etc/profile.d/lcgenv.sh would be the easiest way for the moment. If you worry about the change getting overwritten on future configs, just create another profile.d script instead.
  4. For a running job, to swap into "sc-mode" they would just have to do.
     > VO_SC3_<VO>_DEFAULT_SE=$VO_<VO>_DEFAULT_SE

Resource Discovery

From a Job

Within a job, it is possible to select only SC3 compute resources by matching against the GlueHostApplicationSoftwareRunTimeEnvironment with a value of 'LCG_SC3'. the following JDL does this:

xxx

For the gLite FTS

The gLite FTS can publish its endpoint into the information system. For site admins, see FtsBDIIPublishing13 for details on how to publish into the BDII from your T-1 FTS.

Currently this does not publish channel information (will be in gLite FTS 1.4) so, for a Tier-2 site, you must ask your Tier-1 site for details. See TierOneContactDetails for the contact information for your Tier-1. The details for the Tier-0 are at FtsConfigurationSetup13.

Edit | Attach | Watch | Print version | History: r4 < r3 < r2 < r1 | Backlinks | Raw View | WYSIWYG | More topic actions
Topic revision: r4 - 2006-11-28 - LaurenceField
 
    • Cern Search Icon Cern Search
    • TWiki Search Icon TWiki Search
    • Google Search Icon Google Search

    LCG All webs login

This site is powered by the TWiki collaboration platform Powered by PerlCopyright & 2008-2019 by the contributing authors. All material on this collaboration platform is the property of the contributing authors.
Ideas, requests, problems regarding TWiki? Send feedback