OT : Disaster recovery?

[Steve Campbell]

----- Original Message ----- 
From: "Scott Silva" <ssilva at sgvwater.com>
To: <mailscanner at lists.mailscanner.info>
Sent: Wednesday, October 25, 2006 7:20 PM
Subject: Re: OT : Disaster recovery?


> Colocation Colocation spake the following on 10/25/2006 4:05 PM:
>> So I've just spent the past 3 weeks setting up and tweaking my
>> Mailscanner installation. I've done every possible tweak and gone
>> through everything with a magnifying glass and it all looks great.
>>
>> So now my thoughts turn to backups. How best should i protect my
>> investment? I cannot yet justify a second box for redundancy so if all
>> goes wrong i will need to be able to quickly get my mail server back 
>> online.
>>
>> I'm considering taking one of my raid mirrors out and rebuilding the
>> array online. That way i will have a spare incase it all goes wrong....
>> however there are alot of drawbacks to this.
>>
>> My server is equipped with lights out management so really what i want
>> to do is a "bare metal" type backup, that way if anything does go wrong
>> and i happen to be holiday i will still be able to fix it.
>>
>> Dream Scenario : uh oh for x reason my server has totally died and all
>> data is lost. I ssh in, boot off my already connected USB key and
>> reinstall the operating system. (or some kind of restore software?) I
>> then pull my backup from my ISP's san storage and begin the restore. Two
>> hours later my server is back online and serving mail as it should be!
>>
>> Any thoughts? Or am i living in cuckoo land?
>>
>>
>>
> You could use something like mondo for a backup that could give you a bare
> metal restore. You would need the restore program, mindi, on the usb key.
>
> -- 
>

I second this solution. Mondo is great (if you are sure it will work for you 
in the first place)

I have restored using Mondo before and a complete restoration, including CD 
exchanges, would take less than an hour on an older machine with a smaller 
HD. But I have also had Mondo not work on older machines and some newer 
machines. It has gone through quite a bit of revision lately, and my main 
suggestion for using Mondo is to try it first before you put a machine into 
production. If it works then, you're going to be OK. If it doesn't, it may 
be fixable (there used to be a grub problem, but could be overcome by 
rebooting the Mondo CD and running a utility provided as part of the backup, 
for example).

Now, once you have a working OS+Mondo, don't upgrade either. If you do, you 
should try the test backup+restore all over again. And this is the problem I 
have with Mondo - you either need a twin machine out of production to test 
the new release or be able to justify the restore failing for some reason. I 
_have_ upgraded both before (one at a time, not concurrently) with no 
problems, but I don't always have the spare machine to test with. You can do 
a compare against the backup and live system, but I just don't always trust 
that as I don't know exactly what it is comparing.

Mondo was originally for Red Hat, if I'm not mistaken, and was the only 
utility I found that would restore my RH 6.2 server effortlessly (after 
telling it to include my SCSI kernel mods on the backup - note that then the 
SCSI mod was not on the distro and had to be installed separately) when 
other utilities kept failing. It has now become an all-flavor distro 
utility, and it's just hard to make it do it all that without some problems. 
But the dev team works hard to solve a never-ending stream of requests.

If you can test this first and it succeeds, you will enjoy how easy it is.

Steve Campbell