TRIUMF Grid System Management Documentation

TRIUMF Grid Back-ups

A complete backup system is not yet in place for our prototype cluster.  To ensure that all critical file systems are backed up for now we do the following:

  • Two nodes - lcfg and lcgui01 - are backed up via TRIUMF's amanda setup. /etc and /home are backed up.  Additionally /var is backed up for lcfg.
  • Once a week, on saturday evening, the script /root/bin/doBackup runs from lcfg. This script cycles through all nodes and backs up the following trees, storing the tarballs at /home/backups/YYYY_MM_DD/ on lcfg:
    • /etc
    • /root
    • /opt/lcg/yaim
    • /var/spool
    • /var/log
    • /opt/globus/var     (except lcggiis01)
    • /opt/edg/var          (except lcfg and condorg)

    Additionally specific nodes have these trees backed up:
    host tree
     lcgui01, lcgce02, condorg
    lcfg /var/www, /var/lib/ganglia
    lcggiis01, lcgce01, lcgce02
    sc4tr /var/lib/lgsql, /opt/pnfsdb
    sc5tr /var/tmp/glite-url-copy-sc3
  • Additionally on sc4tr cyclic backups are run as /root/bin/backup_pnfsdb every hour for database systems mysql, pgsql and pnfsdb.  Backup files are created in /home/backups/pnfsdb/?/?? where ?/?? corresponds to dayofweek/hourofday.  Hence the files are eventually overwritten once a week.  As well the entire tree is rsync-ed to sc1tr:/home/backups/pnfsdb/