This page describes the 5.0.0 version of the production environment based on Lenny version of Debian distribution for
x86_64 machines. It intends to explain how this environment was built and how to use it with Kadeploy.
Kernel version 2.6.26-2-amd64 from Debian for x86_64
see the Environment Management page to see the building process of the environment, you can checkout debian-x64-prod-5.0.0 tag.
this is the raw git log, as it can be seen in the git repository.
f331ec5 [g5k-checks] recipes 0a14efb [prod-nancy] add g5k-checks 6f0ef02 [prod-nancy] add g5k-checks 5451552 [packages] Remove avahi-daemon package 5b06d63 [setup/locale] Installs locale en_US-UTF8 8ceeacd [Rakefile] use API 2.0 instead of sid version fc724fd [prod] Use API 2.0 instead of sid 815212b [openipmi] removes memory-manager => solves segfaults with openib interfaces 37c214a [envs] Add Prod for Bordeaux Lille Luxembourg Lyon and Toulouse, modify grenoble bf1783f [oar] Add OAR keys for Bordeaux Lille Lyon Nancy and Toulouse 9c74545 [envs] prod : Add in json description for rennes, recipes removed for test 22f0e45 [Rakefile] Remove gem budle in setup task 8c179ee [Rakefile] Add envs:push and scripts:push dependencies to envs:generate 3e0b5a9 [scripts] Download evironment description b11851f [setup] Add field in kaenv.dsc 48473ec [envs] Add description fields a70c466 [drivers] Correct Logger for netxtreme2 recipe 47be90a Add stress package in packages cookbook 30c8d4f Correct production environment description file for nancy bb8eae6 Add stderr in log file when generate production environment 01be8db [Rakefile] Correct paths in envs generation task ae11423 [openmpi] Changed version of source archive to apply a patch 24e1bcb [Rakefile] Move Rakefile at TOPDIR and implement task envs:generate 4b3d055 [envs] correct rennes prod env 5d172f3 [package] Add r-base package 69be3d5 [setup] add generation scripts, and environment notion 2edf876 [packages] Added package ia32-libs, see bug#2917 fd2c493 [setup] Separate udev configuration from misc reciepe 3f7e2dd [setup] Add nscd configuration d103119 [packages] Add psmisc package 2b230ec [Rakefile] Add drivers cookbook 6053dd8 [oar,setup] Correct ulimit bug (bug#2779) b495d77 ajout des cles oar de grenoble. NT. 321bd97 [setup] Set localtime to Europe/Paris 4e4bd9d Merge branch master of ssh://git.grid5000.fr/srv/git/repos/chef-node-bootstrap 219ded0 [g5kparts] Switched to version 1.0.1 to correct a bug with NFS checks 8d2e73c [setup] a generic motd b8fade4 [drivers] Add drivers cookbook for Netxtreme II support (cluster suno) d0d520c [setup] Remove some others udev permissions a2848cf [myrinet] Change default installation path 138ce10 [g5kparts] Updated to use version 1.0 222d296 [oar] Adapted /etc/default/oar-node configuration file for g5k-parts 1.0 requirements 9a024dc [myrinet] Corrected a small typo in the filename of ifup script 72f1b2d [oar] Added fixed OAR SSH host keys 2990c09 Merge branch master of ssh://git.grid5000.fr/srv/git/repos/chef-node-bootstrap d8c8f48 [myrinet] Added a case statement in template to manage different netmasks between sites 88dca5c [Rakefile] correct task push_bootstrap 0d63a5e [myrinet] Add package syslinux for command gethostip 657defe [openmpi] Implement openmpi cookbook 794fbe7 [packages] remove lam and mpich 90ca6fa Add logs 66de384 [oar] set version 2.4 by default, add logs 19829f4 [myrinet] rewrite myrinet cookbook 73437fc [ofed] rewrite ofed cookbook 6134b28 [packages] Add mysql-client, libjson-perl and libjson-xs-perl 274ed51 [kernel] Change kernel for suno cluster 986293a [oar] Added ssh related files for Orsay fc7b843 Change root password, add apt proxy, add hearders for ofed 2f43fed [misc] Correct problem with ulimit and memlock value 864be93 [kernel] Changes to support suno cluster 383e193 [oar] Add sophia configuration bf652ce [ganglia] Bug fix 99bfc25 Implement OAR 2.4 installation 2bfcf50 Add g5kparts eacc0ba Infiniband devices permissions ec94cf5 A lot of improvement... ;-) c4a1df8 Add sensors for temperature management a96b7fd Install nex version of oar 2.2.16 a69137d Add packages 96aec68 Fixe problem with sun and non-interactiv licence agreement c90390e Remove ssh_host_key df814d4 Restart nscd 6380186 Upgrade package debian-archive-keyring d942437 Correct some development packages and add a hack for java licence agreement 568bf86 Add packages needed for production environment 89c72e0 Add ganglia cookbook into the bootstrap archive 0cab8b3 Fix an attribute problem in setup cookbook 66fea39 Add support of oar 2.4 bd08f83 Add ganglia cookbook 5489ab3 Configure pam 68bedab Add author comment into openib script f0788d4 Add ldap and fstab to setup cookbook 31a00dc Implement ldap authentification e71412a Implement fstab configuration 93f5e3b Implement oar client installation 45d96c3 Initial commit of chef-node-bootstrap
this sections intend to highlight the various points you should know about this images, please consider checking out the Chef repository for more accurate and complete informations.
we are compiling openmpi without the memory manager, this causes a segfault when no myrinet card is present as of openmpi 1.4.1. this has been reported to the openmpi team and will be included in the 1.4.3 release. please check this thread on the openmpi mailing list for further details
some udev rules applies group/users permissions on devices, some of these groups are not created on the system. this is a problem because at boot the system fallbacks on ldap to find those group/users, which isn't reachable because udev starts way before the network.
we choose to remove these group/owner specifications from the rules to solve the problem.
we set the max open files to 8192 and max locked memory to unlimited.
commit 6053dd8 and 2f43fed