IRIDIA cluster backup files

This page contains information on how to stop the cluster and on what files to backup before reinstalling everything.

Stopping the current operation

Log on to majorana as root, do a qstat to see if anyone is using the cluster, if not stop the Sun Grid Engine (SGE):

/etc/init.d/rc.d/sge stop

Check if there is any mail on the servers (/var/spool/mail)

Backing up important files

If there are any MySQL database running on one of the servers back it up using mysqldump, e.g.:

mysqldump --allow-keywords --all-databases -u root -p > mysqldbbackup.txt

Backup the follow directories from majorana:

/home                    (Users' home directories)
/etc                     (User passwords, conf files for NIS, NFS, ..) 
/usr/local/sge           (Sun Grid Engine)
/usr/src/ganglia         (Ganglia)  
/usr/lib/fai             (Fully automated installation)
/usr/share/fai           (Fully automated installation)
/usr/local/share/fai     (Fully automated installation)
/root                    (Scripts etc.)

Note: there is no /usr/src/ganglia directory on majorana or polyphemus! In my log file I wrote to backup the main package only, because configuration files are already taken care of in /etc

Backup the following directories from polyphemus:

/etc                     (Misc configuration files)
/root/flexlmd            (FlexLM license manager files)
/var/flexlm              (More FlexLM)
/etc/init.d/flexlmd      (More FlexLM)
/etc/rc*/*80flexlmd      (More FlexLM)
/etc/flexlm              (More FlexLM)
/root                    (Scripts etc.)

Backup the following directories from one diskless node and one rack node:

/etc

Note: Some directories you can simply tar and compress, while that might take a bit too long/too much space for the home directories. In those cases rsync can prove useful.