01-28-2010 04:31 PM
This particular issue isn't server specific, but I am posting in this forum because it just so happens to be a brand new Lenovo RD120 I am working on.
Also, sorry for the urgent status but I have a very small amount of time to find the answer.
My issue concerns how to config RAID 5 redundancy and get 2 partitions, in the best possible manner, considering disaster recovery, and performance (but I really am not concerned about performance since this server is way over speced for the function it will be performing). The disks are 250GB X 3 so for simplicity let's just say total space is 500GB (of course total space is less due to misc. overhead).
I want to have 2 partitions on my future Windows 2003 installation; A "C" drive which will hold the OS, and an "E" drive which will hold data. I have 2 ways to accomplish this:
1) Config RAID 5 array with a single container to present to the OS. Have the OS installation create a partition 100GB for itself within that single container, then install the OS. Once installed, create a second partition for the remaining 400GB.
2) Config RAID 5 array with 2 containers to present to the OS: 100GB and 400GB. Install the OS to the 100GB logical disk, and use the 400GB logical disk for "E".
I've Googled this every which way, and cannot come up with a difinitive answer. I also checked with our local server "expert" who says it does not matter which way to do it, and that he would do method #1.
Now I understand that both ways will accomplish what I am trying to do. HOWEVER, in terms of disaster recovery, I am leaning toward method #2 because we had a scenario 6 months ago where a server was configged this way and we lost container #1 (the OS), but container #2 remained intact (the data). We rebuilt container #1, then restored the OS, and it was beautiful; no loss of data from container #2.
SO, does having multiple RAID 5 containers introduce any negative concerns such as:
-More likely for RAID5 config to go corrupt (during events such as hard power loss, etc).
-Performance loss (I am not concerned about this but I would like to know)
-Premature hard drive failure (just a guess but I thought I'd mention it).
Solved! Go to Solution.
01-28-2010 05:12 PM
i've been working with RAID for nearly two decades by now and have seen just about everything in terms of failures. installing an OS on a data array is a bad idea. i wouldn't suggest #1 or 2 whatsoever. instead, go with #3.
3) install your OS on a separate drive or array. this could be a RAID 1 array made from two 250GB drives, allowing you enough space for the OS and swap space. if you ever need to reinstall your OS, it's not tied to your data array. this is invaluable.
then, store your data on a second RAID 1 (or RAID 10) array comprised of two (or four) 250GB or 500GB drives.
a 3-drive RAID 5 array is great until you lose two drives and everything is gone.
01-29-2010 03:50 AM - edited 01-29-2010 03:59 AM
Thank you for this information; your point is well taken. Unfortunately I do not have the luxury of time to be able to order several more drives to accomplish method #3.
We have had very good success with 3-drive RAID5 installations over the years. We currently have 8 servers running this way. We've lost drives here and there, but never have lost multiple drives, knock on wood
01-29-2010 11:02 AM
if you're looking for performance then neither #1 nor #2 will be ideal. #3 is both faster and safer.
either way, good luck with your project!
02-12-2010 11:45 AM
Just to add to the topic as maybe anyone else might benefit from it.
I would suggest similar what Eric proposed. If you already have 3x250, instead of RAID 5 please buy additional 250GB HDD and use 2x250 RAID 1 for OS and 2x250 RAID 1 for data.
Also for data part I would suggest (if there's available bay) another 250GB disk as hot spare.
Disaster recovery is always more expensive than 1 spare HDD, esp. when server's lifecycle is 3-5 years.