disaster and recovery
 

I have been researching about disk arrays and on the way have learned some interesting things. I have found out that when building a disk array, the size and type of hard drive have to be identical. Deviating from this will reduce your redundancy. There is really no "add on." When it comes to disk arrays, it is best to build a server that already contains them. By trying to add a disk array after the fact may cause some complications to the OS setup and or missing/corrupt data


During my research I have found servers that range from 1 or 200 dollars all the way up to 30k. It depends on the age and the configuration that you want.

I found that even at the $200 level, there really is no need for a 170 Terrabite 200lb gorilla of a machine. I can not even begin imagine racking the thing since I do not have a proper rack.

Basically, from the hacker in the basement point of view, I do not have the budget, the space or even the right equipment (racks, break out cable etc.) that such an endeavour requires and all in the quest of a disk array and increasing redundancy.

I know that I am basically whining here, and if I do come across as such then I do apologize. Frustrated that I am, such a problem needs a cool head and proper logic solving.

So here is the problem I have a single 40Gig Maxtor HDD that is built into a normal P4 HT 1 gig of DDR ram server. This server has one nick card. one on board graphics card and only one power supply, running Fedora Core 4 Linux. It is the same machine you can find in a small store or in your nearest staples/costco/walmart etc.

This machine has had an uptime of 2 years without any faults and or problems and continues to run even now. It is a httpd/DNS/Mail and Sql server

Of course a good admin looks to the future, thus regardless of the system running fine now, i am still looking for a concrete disaster and recover stratagy.

The way as I see it I do have a few options.

I could buy a Hard Drive and then install said HD, install and configure it back the way it was then take it out. When the first HD dies, I just pop in the second one and carry on. However such a task will increase my down time as I prepare the spare HD with the OS, software configuration and data.

This downtime could result from a few days to a few weeks depending on how things go and will result in loss of services es.

another way is to dd the first hard drive onto a second HDD via usb. when the first HDD dies dd clone back to the new one.

Then there is the idea of building a new server with a disk array in it then while this one runs, prep the new server.


There is an option of adding a storage server to the network for backup and recovery.

Anyways, thats what I have been thinking anyways. I would love for people around here to add their input on what I should do because I have no idea really and I never really actually been in the server market before. Any advice and or wisdom would be great full!

Your server has hot swap hard disks does it ? No so you're going to have down time anyway whilst you change the failed disk.

1) You need to separate the system from the data. This is quite easy the data is stuff that changes regularly and the system is the OS, which once setup will run that way unless you specifically make changes to it.

2) You setup your system so that the system is separate from the data. Separate partitions or logical partitions.

3) You install yourself a second hard disk and set that up to use the same kind of separation.

4) You replicate your system by booting off of a CD and mounting your source and target partitions/LVs and using a command such as 'cd /source ; tar cvpf - . | `cd /target ; tar xvpf -`' this will effectively copy the entire system partition/LV to the 'backup system' partition on the second disk.

5) Reboot your system. Once it comes up write your self a script that will do the same type of command to replicate your data partitions/LVs in the same way as you've just done your system partition/LV.

6) Schedule your script to run as a cron job on your system... you'll want to have tested your script first to be sure it works.

In the event of a disk fail you'll only lose the data that has changed since your script last ran.

Simple really :-)