Advice on FFI daily + FB weekend

[mattdwyerva]

Historically, I have used disk to disk syn fulls or reverse incremental (1 hr) on weekdays (M/T/W/R) with 12 restore points, and then used a script to push the latest VBK (800GB) to LTO tape (circa 2 hrs). I then run active full every Friday night (12+ hrs), and push that to tape with same script. I have daily tapes and 5 sets of weekend tapes (1st wkend of month, 2nd, 3, 4, 5). Thinking about adding a cloud connect vendor, and trying to understand my options. I have about 800 GB of active data on 3 VMs (one physical box) and perhaps 25 GB of changes per day.

Syn fulls (SF) and Reverse Inc's (RI) both have a higher risk of corruption than active fulls (AF), yes? Similar for Forever Forward Incremental (FFI). Since active fulls require compression, I am not sure why they are safer, but weirdly feel safer.

I may be overthinking this, but I am mulling changing to

DAILY - Daily job would be FFI that goes to disk with secondary target to tape. Looks like about 20 restore points would fit on my LTO-5 as secondary target (script won't work as no "latest" vbk). Presumably this could also be connected to cloud (huh, can I have 2 secondaries?).

WEEKEND - Weekend job (Fri night) would be an active full with 1 restore point that goes to disk and then disk to tape.

Any comments?

[mattdwyerva]

Well, 1 in favor, 0 opposed, 42 abstentions: the ayes have it <gavel thump>.

For lurkers, I switched on my new host and the first backup (full, 800 GB) took 3 hrs instead of 12. New Dell R730 with Samsung SM863 ssds are faster than the old PowerEdge 2950 with SAS 15K hdd. Not surprising - just noting the times.

[alanbolte]

Active fulls are safer in the sense that if for some reason a host or SAN sends us bad data on Monday, and then CBT doesn't indicate that the particular data blocks containing bad data should be backed up again, then the synthetic full on Friday (and maybe even next Friday as well) will contain the bad data, whereas an active full would not (unless we get bad data every time). How often does this happen? Clearly not very often at all, but it's frequently impossible to identify how a restore point became corrupt in the first place, so creating statistics for this particular type of failure is also not possible.

FFI to tape works a little bit differently than what you're used to with RI to tape (if you're using the Veeam tape feature instead of writing a new script) - I recommend a close read of the user guide sections on virtual fulls and tape retention.

[mattdwyerva]

I plan to be using veeam feature, so thanks for the warning. I was thinking I could simply specify full+incr in tape job definition and get all the files to tape every day, but I guess I better RTFM.

[alanbolte]

if you want to do that, you can use a files to tape job instead of a backup to tape job, and have it run fulls every day.