Stumbled upon a potential "pseudo-feature"?

[pkelly_sts]

I know if you put 20 IT people in a room to discuss backup strategies, you'd likely get 20 different sets of requirements, but I know for some people the equivalent of "Differential" backups is a potentially useful feature and I think I may have almost stumbled upon a pseudo-version of such a feature (or, at least, the makings of) when I spotted some unexpected behaviour entirely of my own doing via a silly config mistake.

I just noticed that a (test) tape job was writing more than I was expecting it to and upon closer examination I found that it was writing the same .vib files as I already saw it writing yesterday.

When I looked into it I spotted that I'd rather stupidly configured the Media Pool (incremental-specific in this case) to "Do not protect data (cyclically overwrite tapes as required)". I probably did it for a reason originally but that reason now escapes me :-/

Once I realised what was going on I set a more appropriate protection option for the pool but whilst doing so it occurred to me that what I was almost effectively getting was the equivalent of a "Differential" backup of the incrementals. The obvious flaw in this was that, although it was re-writing the growing set of increments, the act of doing so was also overwriting the only version of those files on tape, but I figured that it shouldn't be a massive stretch to build in some logic to emulate this process, whilst actually not overwriting the tape.

Perhaps it's equivalent to "back-up latest chain as differential" whereby the most recent full, plus it's subsequent incrementals, will always be on the same tape.

I dunno, maybe I'm talking daft ideas but it struck me as an interesting scenario so I thought I'd at least share my madness

[Dima P.]

Hey Paul,

onfigured the Media Pool (incremental-specific in this case) to "Do not protect data (cyclically overwrite tapes as required)"

Here is what’s going on: when the tape is erased with all it’s incrementals, tape job picks the entire chain of incrementals again to make the latest restore point operational.

Perhaps it's equivalent to "back-up latest chain as differential" whereby the most recent full, plus it's subsequent incrementals, will always be on the same tape.

It's hard to predict when data is appended or when tape will be overwritten but what bothers me the most - it affects tape life time.