Page

Main page on modifying BEgrid templates (will be moved from here)

The first issue is see a policy to define what is updated by whom. We should try to start from the original files and try to fix the site dependencies in the following order:

1. try gloabla variable first, with or without default value
2. specifiy additional include pro_something_local; and provide all sites with their additional templates
3. as a last solution make a copy of the template and put it in the site specific structure (this one is last because it's most difficult to maintain)

• Structure

Stucture clusters/site/standard used as from grif

• Hardware

In Standard, collect all part info, but we need some naming convention All machines go into site (that way, clusters on the same site can easily share hardware + this shouldn't be in standard anyway)

Problem 1: iihe uses modified machine hardware, (ie includes smp and architecture), gent doesn't. 
We must come up with a similar but better solution for this) Probably through functions or global variables.

Problem2: fix cciss in aii

• • Network

iihe sets boot flags for nics in machine profiles, i added boot ?= false as defaults in the nic templates

• OS

added the rpms from both sites, gent kernel 2.6.13 is in pro_sl305_smdesmet, for iihe, this is in pro_site_kern26. fix this for iihe...

• components

 Problem: aii MUSt be cleaned out and made equal for both sites

 Problem: check and upgrade all versions of components to the same version

 Problem: iihe uses ncm-network and ncm-netdriver. it's recommended to other sites also.

• pan (previously called standard)

for iihe we made some changes mostly due to network (ie private subnet + extra components). normally these should cause no conflict for other site try to integrate them in the defaults. or make them begrid defaults.

• repository

currently no central repository. someone should look at the webproxy server config. will be fixed ;)

• Site

in first attempt, added all old site config files to cluster. some of these should end back in site/somwhere/site or even in standard/site (like functions etc)

• lcg

the number of files in all lcg related directories should be reduced

  RPMS most of the stuff is correct
  enabled fm on most machines
 
  VO this needs to be cleaned out. iihe also uses different than standard approach to add vo. is this still relevant??
  VOS also must be cleaned out (try to fit as much as possible VOS in BEgrid for easier support on other sites)

 SOURCE: default from 2.6.0-16 in standard, diff against that in clusters
 There's an error in pro_lcg2_user_edginfo in ugent site !
 Most differences between ugent and standard are because of different vo addressing
 iihe source contains lots of old garbage. fix this. also some subtle changes due to dcache/selist etc. 

try to compare and generalise as much as possible.

we need a central repo asap. for the moment i'm using both ugent and iihe