We have an HP ProLiant ML350 G5 server with an 8-bay RAID 5 array configured on a Smart Array E200i controller. The system volume and data are on the same logical drive. The drives are 72GB each and we are running out of space. All the bays are used up. We have considered external storage expansion and using an archive drive to move the data temporarily while we upgrade the array, but given it's a system drive, that could get tricky.
We came up with this plan and would like some validation this will work:
- Verify full data backup is completed. Remove 72gb drives from bays 1-4, breaking the array.
- Put in new drives into bays 1-4 (600gb drives). We are fully aware that adding these drives to the existing array will not utilize all the space, yet.
- Let the array rebuild.
- Remove drives from bays 5-8, breaking the array again.
- Now, all the data is also on the first 4 600GB drives, so we can expand the array to the full size and remove the last 4 drives from the logical drive.
The part I am mainly unsure of is how to remove the last 4 drives from the array logically and whether or not this will work at all. Thanks for any insight.
UPDATE: Since I can't mark a comment as the answer, I'll update the post. In the end, we had no other solution than to backup all the data, replace the drives, recreate the arrays and restore. This is relatively nerve racking when you don't do it often, so we put a test plan together and used Macrium Server to backup to a USB 3 drive and restored from that once the new drives were in place. Before replacing the drives we made sure we could access all the backups and had a plan B. It took a while, but it did work and now they have plenty of space available. Thanks.
Why not just replace all eight disks with 300GB drives (cheaper and make more sense for your server)? You can remove them one-by-one and handle the array transformation online.
There's no concept of "breaking the array" in a parity RAID setup. That would potentially work in a RAID 1+0 configuration, but that's not the case here.