Best Practice - redundancy

[mattdwyerva]

A friend taught me this 30 years ago when I lost everything (own small company) using incremental backups that were corrupt. We are mostly closed on weekends and backups run late at night.

• Mon/Tues/Wed/Thurs - backup every file that has changed in the last 45 days (modify date less than 30 days). Backup to disk, duplicate to tape.
• Fri - full backup to disk, duplicate to tape.

Tapes used

• 4 tapes in use during the week (a Monday tape that is used every Monday,etc)
• 5 different Friday tapes that rotate and are labeled FRI-1, FRI-2, etc, for the # of the Friday in the month.
• Retire a Friday set once a quarter or so if feeling nervous,or once a year if not.

If we lose the server, we take ANY of the Friday sets (makes no difference) and do a full restore and then take the most recent MTWR tape to be fully restored. If any of the friday sets is corrupt or missing, just grab a different one. If the latest MTWR tape is kaput, you must go back a day (lost 24 hrs), which would be sad, but you know exactly what you lost. Need to keep tapes off-site in case of fire, which we mostly do by dragging back and forth commuting.

Any issues/comments with this process? Not sure it matters, but my current infrastructure is one host in use (+one avail for replication) with 6 guests (Dell 2 cpu rack server, VMWare, GB network, no WAN backup, closed on weekends). I have never done replication, but might in the future (going virtual freed up a physical box).

[mattdwyerva]

Reading further (sorry, should have done that first), if I understand VBR backups, my method is impossible (can't backup all files in last 45 days using CBT), but I could do either full backups every day or synthetic full backups every day.... or use incrementals and pray. hmm. Maybe I will push my daily incremental (probably under 10 gb) to the cloud as well as tapes to get redundancy there, but pretty much screwed if my most recent full backup tape is bad for some reason? How do people deal with that? It must happen on occasion.

[veremin]

I think you’re aware about the so-called “3-2-1” rule that one should bear in mind while thinking about backup plan:

3 Copies -> 2 Different Media -> 1 different location -> 0 errors at test restore.

The very idea is it everything possible should be done in order to guarantee security of data against any potential failure that might happen practically with every device, and tapes are not exceptions here. Taking abovementioned points into account, you should think about tapes as only one part of the whole backup strategy, and not as the universal panacea.

So, it might be worth specifying local backup jobs first (local repository), using whatever backup method you want to, and, then, copying the backup data to external media (tape) and different location (cloud), as well.

Thanks.