Changelog VanGogh Cluster

Please put newest item first

See also ChangeLogVanGogh2004 and ChangeLogVanGogh2005

Build Ruby 1.8.5 from source

2006-11-06 by Hans Blom
Why: The SL package version has problems with remote Java access at some hosts
Where: On vangogh5-8
How:

Start-stop Dragon daemons from init

2006-10-04 by Hans Blom
Why: To start the Dragon daemons at system boot
Where: On vangogh5-8
How:

• Created init.d script /etc/init.d/dragon, using as template other scripts in that directory. In that script the chkconfig service "dragon" will be switched on for the run-levels 3-5 using the script /usr/local/dragon/bin/dragon-no_log_rsvpd.sh.
• Run the command chkconfig --add dragon to add the "dragon" service to chkconfig.

Use /dev/null as log file for RSVPD

2006-10-04 by Hans Blom
Why: The original log file /var/log/RSVPD.log was growing too fast and could not be circularized
Where: On vangogh5-8
How:

• Stopped the "dragon"services with /usr/local/dragon/bin/dragon.sh.
• Copied /usr/local/dragon/bin/dragon.sh to /usr/local/dragon/bin/dragon-no_log_rsvpd.sh and replaced in the new file the line:
  RSVP_ARGS="-c $ETC_DIR/RSVPD.conf -d -o /var/log/RSVPD.log -L select"
  by:
  RSVP_ARGS="-c $ETC_DIR/RSVPD.conf -d -o /dev/null -L select"
• Restarted the "dragon" daemons with /usr/local/dragon/bin/dragon-no_log_rsvpd.sh start-vlsr.
• Removed the old log file /var/log/RSVPD.log.

Cfengine installation

2006-02-16 by Martin Pels (SARA)
Why: Easier clustermanagement
Where: On all VanGogh nodes
How:

• Used RPM from rpmfind.net
• Server on vangogh0: configuration in /data/cfengine
• Client on vangogh1
• Pulled new image of vangogh1
• Installed new image on other nodes

Cluster reinstall with Scientific Linux 4.2

2006-02-15 by Martin Pels (SARA)
Why: Current distro is too old
Where: On all VanGogh nodes
How:

• Did a custom install of SL4.2 from CD on vangogh0 and 1, deselecting everything we don't really need
• Installed NIS-server on vangogh0
• Installed NTP on vangogh0 and vangogh1
• Added a couple of users to NIS and disabled SSH-rootlogins
• Changed firewall configuration to match all running services (can be found in /etc/sysconfig/iptables)
• Installed a Systemimager (http://www.systemimager.org∞) server on vangogh0:
  • Needed RPMs:
    • perl-AppConfig systemconfigurator systemimager-common systemimager-i386boot-standard systemimager-server (use installscript)
    • Perl-XML-Simple (via rpmfind.net)
    • tftp-server (via apt)
  • Moved systemimager libdirectory to /data
  • Configured TFTP server in /data/tftpboot
  • Installed SARA's pxeconfig scripts (pxeconfigd, pxeconfig en hexls) to allow for installation using PXE-boot (the official PXE RPM apparantly has issues with this distro)
• Installed a systemimager client on vangogh1:
  • Needed RPMs:
    • perl-AppConfig systemconfigurator systemimager-client systemimager-common (use installscript)
• Pulled an image of vangogh1:
  • On vangogh1: si_prepareclient --server vangogh0
  • On vangogh0: si_getimage --golden-client vangogh1 --image vangogh
  • Added vangogh2-8 in the addclients util
• Made sure PXE is selected first in the BIOS boot sequence on vangogh1-8
• Installed the image from vangogh1 on vangogh2-8

Reinstalling a node

• Start pxeconfig and enter the IP('s) of the node(s) you wish to reinstall
• Reboot the node
• Get a cup of coffee
• When you return the node will be beeping, asking for a reboot. Now first delete the file that corresponds to the reinstalled node from /data/tftpboot/pxelinux.cfg on vangogh0 (if you don't the node will simply reinstall again), and reboot the node.
• That's it!

Backup of /home and /space in "rembrandt0"

2006-01-10 by Hans Blom
Why: The cluster will be reinstalled
Where: On all VanGogh nodes
How: See the corresponding item at 2006-01-10 in ChangeLogRembrandt.

Install Scientific Linux

2006-04-01 by SARA
Why: Because the current distro is too old
Where: On all VanGogh nodes
How: Not happend yet.

Categories

CategoryLogs