I am setting up a fileserver that uses a Promise EX16300 raid controller that currently has 4 750gb drives on it using raid6 on a Windows 2003 SP2 Server. My idea for the design is since the EX16300 has OCE (Online Capacity Expansion) I should be able to add drives and just grow the total size as more space is needed. This should also amortize out the cost of drives since 750gb drives in 18 months should cost half what the do now.
The problem I have is I don't really have any way to back up my data so am pinning my hopes on tight write access security and raid 6. This also means that once I start adding data to the array I have no easy way of pulling it off to make major changes to the config if I get it wrong, so I wanted to run it by someone else first.

I am setting it up to use:
Type: Dynamic
Partition Style: GUID/GPT
Layout: Simple
Format: NTFS

Does this all sound right? Will this allow me to grow both the array and the volume larger then 2TB? (It will be booting off a separate drive.)

And my final question is how reliable do people feel OCE on this controller is? Do I have a major risk of corrupting my array every time I add a drive? What experiences do people have with it or just OCE in general? I know it takes a long time to add drives. My test from 3 to 4 drives took 6-7 hours and I assume the more drives the longer it takes.
Any help or suggestions would be greatly appreciated.

Ill also will be interested to hear responses on the OCE as this is what i wish to do with my controller.

Evilgeek, that partition setup will work with larger than 2TB i learnt this the hard way. I currently have 5x500gb in RAID 5 = 2TB (1.81tb formatted) however, instead of using GPT i used NTFS which doesnt support larger than 2TB partitions. As a result i cannot convert to GPT without clearing the partition of data first meaning the ~1.2TB of data i have on there i need to backup :S ouch that hurts.

Currently im planning to buy ~4 new 500gb drives, install these on the mainboard and copy data off manually so each drive is full. From there i am going to wipe the array and partition as GPT. I will then copy all the data back from 3 of the 500gb (as this will be all it will take to backup whats there now)... then finally once thats done, i will install the 4th clean drive on the array and try out the OCE. If i loose the data on the array i still have a backup, if it works then i will continue to add 2 more of the drives (although this section is risky because then i wont have a data backup, but at least i have verified OCE functionality).

IMO whatever you do with an array you should back it up first if you dont want to loose anything because there are manny horror stories out there.

GPT is a partitioning scheme, NTFS is a file system. It's not a choice between the two, they don't have anything to do with each other.

Instead of GPT partitioning, you used MBR partitioning, which is limited to 2TB disk size and volume size. To convert the disk to GPT, you're correct in that you have to delete the volumes that are on it.

There is no reason to make the disk dynamic. It only complicates matters. You can use a GPT disk with >2TB partitions and it can still be a basic disk.

To grow the array is a 3-step process:

1. Physical: Install the new drive to add to your array.
2. Logical: Use your RAID controller management software to perform the online capacity expansion (OCE). This will increase the volume size.
3. File system: You then have to increase the partition size on the array using the command-line DISKPART utility (Windows XP64 & Windows Server 2003) or you can use Disk Management in Vista.

Risk of corruption can be minimized through several methods:

- Run a full surface scan check of any new drive you're going to add to the array before adding it. Even if the drive is new out of the box, you need to make sure the drive is good before the RAID controller begins the OCE on it. The OCE operation is one of the most intensive operations a drive can undergo, and if the drive is borderline, that's when its going to fail. I've only personally seen ONE RAID controller where the array data survived when a drive died during an OCE, that was a 3Ware 9650SE.

- Make sure the computer is on a UPS during the OCE.

- It's a good idea to have a RAID controller that has a battery back-up module for it's cache.

- It might be a good idea to disable the on-board write cache during OCE if the RAID controller doesn't do that automatically (most do).

- OCE can be a very long operation. I've seen OCE sometimes take 48-72 hours on some controllers to add a drive to a RAID 5 array. Prepare in advance to have the computer committed and doing nothing but the OCE during that period of time.

- Obviously, if there's something on the array that you can't afford to lose, it needs to be backed up before you start an OCE.

Thank you, good advice. I had chosen Dynamic because I thought it would be simpler to grow. After you pointed out I shouldn't, I did more reading and I now agree with you, no need to use it unless I have to. That was exactly the kind of comment I was looking for.

The surface scan is also good advice; I will put that in my notes.
I have a good dedicated 1500va UPS so that shouldn't be a big issue.
I have looked around a few times and can't find the battery backup unit (BBU) for the EX16300 in stock anywhere. Its part F29BBUP20000000 and costs $115… (who said this was a cheap hobby). Maybe I can get it though Promise directly.
During my testing I remember the controller making a note that it disables write cache during rebuild and OCE, but I will also add it to my notes for my first rebuild.

A few final questions, do you think that RAID 6 complicates things over 5 as far as OCE goes, or will it help mitigate some of the risk like it would in other failure situations?
The same with adding more then one drive at once. It seems like if the process it risky, I would want to do it as few times as possible. AKA, instead of doing it 10 more times with a single drive, just doing it 5 more times adding 2 drives at a time may be better. (how is that for an example of truthyness.)
As for backups, unless someone comes up with a really cheap way of backing up a few TB, it will be a no go. I'm sure I will reply to this in a year or so with much gnashing of teeth and pulling of hair, but that’s life.
Again thanks for the help and advice.

I don't think RAID 6 would complicate things over RAID 5, and would give you an additional layer of protection. Of course, you lose 2 drives worth of space in the array instead of 1.

https://usm.channelonline.com/nerdn [...] D000002119

In stock for $119.

Thanks again, I just ordered one. Never heard of NerdNY but they get ok reviews.
The extra cost of a drive on raid 6 is irritating, but it is an up front cost that at this point I have paid, from here on out each 750gb drive I buy increases the size by 750gb. I also view it like a hot spare without the temptation to use it for something else… :-)

Thanks someJoe7777 helped me out a fair bit as well.

Ill use alot of this advise with my RAID extension and hopefully i wont have hair pulling either.

Just as a follow up to this thread (mostly for my own notes for the next time I do this), I have finally run out of space and added a new drive. I went from 4 750gb drives in raid 6 to 5 drives. The process went without a hitch; however it was as expected very slow. I scanned the surface of the new drive, then used the controllers Migration function to add a new drive. This took around 70 hours, however about 55 hours in I noticed that the migration rate was set to medium not high. The diskpart command only took about 1 sec to run (fast enough that I thought it failed.

Notes for next time:
Make sure settings for rebuild and expansion are set to high.
Suspend monthly disk surface scan if it is scheduled for next 4 days.
Make sure nothing is in event logs (system's and controller's).
Make sure windows update auto restart is disabled.

Scan new drive for surface issues.
Under Logical Drive 1's migration tab, select the new drive. Start now.
Wait forever…
Continue waiting forever…
When finished run command line utility diskpart:

DISKPART> list volume
Volume ### Ltr Label Fs Type Size Status Info
---------- --- ----------- ----- ---------- ------- --------- --------
Volume 0 D Media NTFS Partition 1397 GB Healthy
Volume 1 C NTFS Partition 75 GB Healthy System
DISKPART> select volume 0
Volume 0 is the selected volume.
DISKPART> list volume
Volume ### Ltr Label Fs Type Size Status Info
---------- --- ----------- ----- ---------- ------- --------- --------
* Volume 0 D Media NTFS Partition 1397 GB Healthy
Volume 1 C NTFS Partition 75 GB Healthy System
DISKPART> list disk
Disk ### Status Size Free Dyn Gpt
-------- ---------- ------- ------- --- ---
Disk 0 Online 75 GB 8033 KB
* Disk 1 Online 2095 GB 698 GB *
DISKPART> extend
DiskPart successfully extended the volume.
DISKPART> list disk
Disk ### Status Size Free Dyn Gpt
-------- ---------- ------- ------- --- ---
Disk 0 Online 75 GB 8033 KB
* Disk 1 Online 2095 GB 0 B *
DISKPART> list volume
Volume ### Ltr Label Fs Type Size Status Info
---------- --- ----------- ----- ---------- ------- --------- --------
* Volume 0 D Media NTFS Partition 2095 GB Healthy
Volume 1 C NTFS Partition 75 GB Healthy System
DISKPART> exit

The layout needs to be changed from simple to expanded if you want it to span more than one drive

??? I'm not spanning drives I am growing an array from a 4 drive to 5 drive array. Once the controller grows the array diskpart is used to expand the volume, not span it onto another drive.

Hey evilgeek, good to hear everything went well.

In my first reply i explained what i had to do to mine and everything went well with it too.

4 new drives, copied all the data off wiped the array, rebuilt it. Copied all data over. OCE from 7 to 8 disks in RAID5... took about 3 days i think. all good

I have a similar problem trying to expand my RAID5 set.

I use a highpoint RAID card that support OCE.

I currently have over 6TB of data over many disks, all raid5. They're using simple MBR partition and are split as 3x2TB volumes.

So essentially I have 3 2TB NTFS drives as far as XP is concerned.

My main issue is that I'm running against the 2TB limit. I don't use XP x64 and so I can't use GPT.

I'll like to add another 2TB of space to one of my volumes. So I will have a single 4TB drives (the others will stay at 2TB).

I was wondering why can't I use dynamic disks to simply "expand" a virtual disk across the volumes?

It sounds like the simplest way to do this. I'll "convert" the target 2TB drive to dynamic, then do a OCE of 2 TB of additional space, then create a new 2TB volume, then add that to the dynamic disk.

It seems to be the best time saver, since I don't have to do a complete system re-install (to install XP x64 or even linux) , complete data backup/restore cycle.

I don't plan to upgrade (downgrade) to VISTA. If XP x32 really runs out of livespan in the next 5 years, I'll directly jump to Linux, (but that'll be a long and painful migration with all my data as such...)

Any advice?