140 likes | 366 Views
Tales of a TSM Implementation. Peter Smith Bendigo Bank Ltd. Agenda. Why Did we Move to TSM ? Major Components of our TSM Implementation TSM Architecture Summary of TSM DAILY Backups Summary of TSM MONTHLY Backups How we handle “LongTerm” Backups Summary of TSM Operations
E N D
Tales of a TSM Implementation Peter Smith Bendigo Bank Ltd.
Agenda • Why Did we Move to TSM ? • Major Components of our TSM Implementation • TSM Architecture • Summary of TSM DAILY Backups • Summary of TSM MONTHLY Backups • How we handle “LongTerm” Backups • Summary of TSM Operations • Data Storage Policy • What we’ve learnt along the way • Future Plans
Why did we move to TSM ? Backup Window unmanageable • FileServer (560Gb) & SQLServer (680Gb) taking approx 31 hours each for a month end backup. Leaves little/no room for error. Restores delayed due to transport of tapes. • Backup from last night lives at Vault. Most restores require a trip to vault to retrieve relevant tape. No redundancy of Tapes • Only a single copy of each tape is kept. If a tape becomes corrupt then all data on tape is lost. Cannot rebuild tape.
How is TSM architected ? • Major Components • TSM Server located in BBL Data Center • DR TSM Server located in BCC • Disk Storage Pools on SAN located in BBL Data Center • Primary_Data 200Gb • SQL_Data 10Gb • 3494 ATL with 4 x 3590 Tape Drives located at BCC • TSM Clients
Summary of DAILY TSM Backups • By default data is backed up to a disk based storage pool • Large files eg SQL backup files are sent directly to tape based stgpools • Whilst a file exists on the LAN a backup copy is kept indefinitely • If the LAN copy of a file is deleted the backup copy is kept for 1 year from deletion date. After this date the backup copy is also deleted. • dsmc executed from client to invoke backup
Summary of MONTHLY (LT) TSM Backups • <node> is re-registered with server as <node>_LT • <node>_LT is assigned LT client options file (retain for 7 years) • dsm.lt.opt with “nodename = <node>_LT” is created • Backed up to disk based storage pools • dsmc –nodename=<node>_LT –optfile=dsm.lt.opt executed from client to invoke backup(same .bat file as DAILY backup) • All modified data since the last longterm backup
Summary of TSM “Operations” Daily • Backup client data to disk / tape • Create “offsite” copy of all backed up data • Migrate disk based storage pools to tape • Expire data according to “management class” • Run DR operations • Backup TSM DB to Tape and Disk • Copy TSM DB to DR server • Copy scripts, prepare plan, volhist, devcnfg to DR server • Run “reclamation” for tape usage optimisation Monthly • “colocate” top 10 clients
Data Storage Policy Client Option Sets System / Application disks = BA_7DAYS_MC Data disks = BA_31DAYS_MC <nodes>_LT always configured with LT_FILE_MC (takes precedence over all other MC)
What we’ve learnt along the way (1) • Keep on top of TSM Service Packs • Closely monitor tape drive activity • Rogue activity can quickly deplete “scratch” volumes • Consider an alternative to TSM Scheduler • Don’t rely on the GUI • Get to know the cmd line, basic SQL cmds • Design your own reporting • Research TDP’s carefully • Watch changes to MC very closely • Consider the default MC • Get the “business” to define data retention rules • Have a costing structure
What we’ve learnt along the way (2) • Don’t get caught without scratch tapes • 2 tape drives are not enough • Locate the DB on a disk with room for growth • Perform regular reclamation • NewsGroups / UserGroups are your friends • Be careful with firewalls / VLANS
Future Plans • Implement TDP backups • Build a “costing” model • Performance Tuning • Delegate “restore” authority • TSM in regional hub sites • Consider moving to larger capacity tapes