Synthetic Fulls - Opinions

[hutchingsp]

We currently backup to tape, full at a weekend, incrementals every day.

I've been reading the concept of Synthetic Fulls and I'm interested who's doing this and how you have it setup and how you've found it to be.

We have a pair of tape libraries, a primary one with twin drives and a secondary one with a single drive - I'm thinking that if we were to look at synthetic fulls having a secondary copy of the "real" full is a must-have?

Also we have a single remote office on the end of a 2mbps circuit who have a couple hundred gig on a local server which has a media agent and LTO drive - I'd love to look at synthetic fulls on that but there's a big question mark over how to "seed" the initial 200gb full backup given the tiny pipe?

Thanks in advance.

 

[NATD]

1) A synthetic full *IS* a real full. The only difference is how it came to be. You don't need to protect the full any differently based on how it was created.

2) You seed to some kind of removable storage, I use a cheap and cheerful USB2 harddisk - one of the small light ones that run off USB power and are easy to ship with just a layer of bubble wrap in a bag. You run a backup when the library is under the media agent of the remote server, then migrate (just a context menu option) that library to your central media agent and fix up the path if required.

It's all documented in Books Online or you can find variations that work better for you.

 

[hutchingsp]

Thanks for the reply.

When you say a synthetic full *IS* a real full, what I mean with my question is this:

Week 1 I take a conventional full backup, I do the same Weeks 2, 3 and 4.

I need to do a restore from week 3 and when I do so a tape fails, assuming the files were there in weeks 1 or 2 I still have a chance to recover the files.

Week 1 I take a conventional full backup, weeks 2,3,4 I take a synthetic full.

I need to do a restore from week 4, but a tape from the "real" full in week 1 fails - at this point I'm totally screwed aren't I?

Perhaps I've totally mis-understood how the synthetic full mechanism works in which case some help would be much appreciated.

I have two media agents with multi-gbps between them and each with an ultrium library attached.

The primary library has 48 slots and two drives and is physically attached to the file server where all our data is.

The smaller has 24 slots and a single drive and is currently used for weekday incrementals so they're stored away from the data.

Regards the remote office, you mention USB which I guess is for "backup to disk" which we don't have a license for?

I did think one option would be to migrate the existing media to the main library and then look at synthetic fulls, but it would be nice to have a full backup available at the remote site too, and AIUI any synchronous/aux copy copies from media to media so the bottleneck is still the 2mbps WAN link?

 

[Psy053]

I think you are looking at Synthetic Fulls as if they work similar to an Incremental Job, which is not the case, and as NATD said, a Synthetic Full is the same as a "real" full and as such doesn't have dependencies on other backup jobs. So, if you lost Week 1, you would still be able to restore from weeks 2, 3 and 4.


As for your slow link backups, I would suggest that you continue to backup all data locally, and then look at implementing something like Commvault CDR (at a cost) or Microsoft DFS (free) to replicate the user data from the remote site to your local site where it can be included in your regular Inc/Full backup cycle.
The benefit of this is that CDR and DFS can both trickle data over the slow link with little impact to link utilisation. The only downside is that the intial data copy will take some time, however, this shouldn't matter too much as the data is already being protected locally.


Just a couple of questions:
How many clients are you backing up locally?
Why do you need to separate the Incrementals and Fulls by location?

 

[hutchingsp]

Yup my bad, I'd been thinking that if my full backup on week 1 was, say, 20 tapes, that my synthetic full on week 3 would only be 3 tapes of whatever the change delta - clearly not the case it just makes the synthetic from existing fulls and incrementals.

CDR may be an option, we're going to be implementing that shortly between the two local file servers/media agents to avoid needing to hit the tapes so quickly in the event of a disaster.

To answer your questions, only a dozen or so, and we separate solely so that should we lose the primary server/tape library on a weeknight, the incrementals are safe in the "off-site" library - the incremental backup takes no time so whilst I know I could switch it to the primary library and run an aux copy it's small enough not to be an issue.