Testing backups

[kaphc]

I've just implemented swappable backups in a small primary school and it got me thinking about how to test and verify backups. At the moment, all I do is periodically create a test file before the next backup, wait until the backup has run, then delete the test file and check I can restore it from the backup drive.

 

I'm sure there's more I could be doing, and I'm sure this might open up a whole can of worms .... but can anyone provide any tips on best practice?

[glennda]

It completely depends on a. what you are backing up and B: how you are backing up

[kaphc]

Good point! To clarify - these are server backups I am referring to, i.e. in Windows Server 2003 System State plus volumes and in Windows Server 2008 Bare metal recovery, System State plus volumes.

 

The backups are in various different schools which is why I'm working with two different Server o/s and different backup methods. They use Windows Server Backup (apart from one which uses Cobian) either directly or by calling it as a scheduled task.

 

Hope this helps to provide more detail to what was quite a vague question :-)

[glennda]

Well to be honest a full DR plan really is required for each of the sites - I know it sounds like a huge task but to be honest its not really.

 

The plan should include:

1. What happens with backups i.e how they are backed up - to what, does it go offsite etc etc

2. How to restore normal files (pretty much what you test now).

3. What should happen if the building burnt down - what would be required to restore things and the steps to do it i.e new server (specs etc)

4. Actually run this test and ensure you could essentially restore the entire network back to how it is now.

[Heebeejeebee]

4. Actually run this test and ensure you could essentially restore the entire network back to how it is now.

 

What! Burn the school down to test the backups

 

HBJB

[glennda]

What! Burn the school down to test the backups

 

HBJB

 

Not quite . I don't tend to work much in schools now but the main way you are going to find out if you plan works is actually testing it. I have done loads of DR tests and only 1 has worked 100% first time and you don't want to be finding these issues out when you are under the cosh of getting things in place.

[dhicks]

but can anyone provide any tips on best practice?

 

My ideal would be to have a server in a separate physical location (to cater for fire / flood / etc) that mirrors the state of your live servers, either in real-time or via a nightly job to clone a VM image from the live server to the backup server. Then, at any point you should be able to pull the power out of the UPS powering your main server and have it automatically detect there's a power problem then shut down after a minute or two and start up the backup VM images on the backup server. Files should be user-restorable back at least a couple of weeks from a versioned file server.

[kaphc]

To add to the mix, these are all small primary schools - in a couple of them, we have "old" servers that were replaced which I haven't disposed of, just as a really last-ditch option. But there's never enough budget or spare decent kit around to have extra servers.

 

Each school has a back-up / DR policy so that's in place. I'm just implementing swappable backups in a couple of them. The admin staff also run backups of the finance & pupil management software. The DR policy lists priorities for restoring services.

 

To actually do a full restore, each primary has maintenance with the County Council which would cover the full restore of the server. Until I've sat down now and thought this out, it hadn't really occurred to me though that there would be a mix of the County Council / external suppliers involved and it would be useful to map out who would do what. This will be a really useful exercise to do and make sure we have all the relevant responsibilities and requirements mapped out plus easy access to contact details etc.