Server 2003 Best Practice

[wkim623]

All,

I've started at a new company and I noticed that the servers they have installed had only one partition. So both the OS and data is on the same physical partition.

I was trying to find the white paper again or any info about why this is a bad way to build it this way.

I usually build my servers with two physical partitions, keeping the OS and data away from each other. Or is this an ok way to do a server build?

I just remember reading way back when I was a younger guy that this was the best practice to separate the System and Data either virtually or physically.

Any links or info would be great.

 

[technome]

I usually build my servers with two physical partitions, keeping the OS and data away from each other. Or is this an OK way to do a server build?"

Ran into a network 3 weeks ago which had a single partition for the OS and data; absolutely no organization . Well a hacker gained access to the network, and I needed to rebuilt the system from scratch. In the least it took me three times the amount of time to transfer the data over to a workstation as a backup (no backup unit existed, of course); I had to find the data.
During the rebuild I created separate partitions, and transferred only the data and some program info back to the newly created server. This took at least three times the normal time it takes, as I had to double check every directory for data. No less the hacker planted key loggers everywhere.
My point, something which should have taken approx 3 hours took over 18 gruesome pressurized hours, with the client blaming me for taking so long.

If the moron who originally build this server had separate partitions, I could have quickly transferred the directories in \program files, over to another machine, wiped the OS partition, and re-installed the OS without moving the data.
As a note, this was the worst setup network I ran across in 27 years. Actually the disk was setup with two partition, a c: partition of 1.86 Gigs ( thats right, a partition less than 2 Gigs), the CDrom was D: and the partition with everything on it was E:
.... Setup as a OS raid via Dynamic disks on a server with a hardware raid interface.


........................................
Chernobyl disaster..a must see pictorial

http://www.kiddofspeed.com/default.htm

 

[wkim623]

Thanks for posting your experience, that definately would help out with my argument of why we need to rebuild it now..rather than later. just wanted to have more ammo to bring to my manager.

 

[technome]

The safest setup is to use a separate hardware raid 1 array for the OS (relatively small disks) and another array for the data. Raid 1 is safest redundant disk array, odds against a raid 1 having both physical disks fail in a short period of time are close to astronomical, aside from a bug in disk/controller firmware, massive power anomaly, fire flood.


........................................
Chernobyl disaster..a must see pictorial

http://www.kiddofspeed.com/default.htm

 

[LWComputingMVP]

Consider this - by putting the data on a separate set of disks, you can EXTEND the volume as needed later by adding drives. Typically, the OS is recommended to reside on a RAID 1 whereas file sharing data sits on an expandable RAID 5. You cannot easily extend the C: drive - in many cases it is difficult at best and near impossible without SIGNIFICANT amounts of time, effort, and risk. Even if you can't put 5 drives in this to create a RAID 1/5 combination, the D: partition can still be extended in most cases while C: cannot. You're imposing an unnecessary hard limit on yourself.

A user could, in theory, fill the drive and cause the whole system to crash. Even with Quotas, there no certainty that a misbehaved OS component or third party application won't cause the drive to fill up with useless junk and cause problems for users.

 

[wkim623]

Thanks for all your input guys. Those points would definately help out the reasons why.