Link to home
Start Free TrialLog in
Avatar of Pete
PeteFlag for United Kingdom of Great Britain and Northern Ireland

asked on

Replace good RAID disks

I have a RAID 5 array of 4 x 4TB disks, I want to swap each disk with a 10TB disk allowing rebuilding between each, then expand the array size at the end to utilise all the new disk space. I’m using Intel Raid web console 2.

I’m hoping the general idea is ok? I’m doing this to prevent data loss, otherwise it’s destroy the array and replace disks.

How do I remove safely and replace a good/configured disk?
Avatar of John Tsioumpris
John Tsioumpris
Flag of Greece image

Your usable space is around 12 TB ...so grab a 10 TB disk and make a backup of everything (better 2 just to be sure...)..then start the process..if everything goes well then OK...if not you will have backup and you could make a RAID 5 initially with 3 Hdds (10TB) while keeping the 4th(10TB) with the data until you have copied and verified the data on the "olds" 4TB
Avatar of Pete

ASKER

Ok thanks, the data is actually a DPM repository, I don't think that can just be copied pasted? DPM has lots of funny hidden partitions.

How do I remove a good disk for the array and make it rebuild to new?
Everything depends upon the hardware RAID controller and what it allows you to do.  What controller is it?
Avatar of Pete

ASKER

RMS3CC080 8 Port (integrated)
Avatar of Member_2_231077
Member_2_231077

RAID 5 will protect you against the odd unreadable block with 10TB disks but the chances of surviving a disk failure without data corruption are minimal. The "swap each disk in turn" procedure was valid with 147GB disks and took a few hours, I'd hate to think how long it will take with 4TB ones. At a guess a day per disk, can you afford 4 days without parity?

If it were mine I would physically add one 10TB disk and migrate it to RAID 6, then do the small for large swaps. AFAIK it's LSI's stack so that would probably work.
That idea is theoretically sound, however it may fail. For it to work, ALL current drives need to be PERFECT. The rebuild involves reading every sector of each drive, to create the new parity data. If you have a single hard read error, data will be corrupted. During this time of intense drive thrashing, you of course have no parity protection in place, so a single failure means data loss. Also, it is not uncommon for some RAID adaptors to turn a  blind eye to a few bad sectors on a drive and just keep using party to correct the problem, this will result in a "punctured array" when a drive is replaced.

All up, if your hardware can handle a few days of intense thrashing, and has no bad sectors this procedure will work, but if anything goes wrong, you will lose data.

If the data is important, I would be looking at performing a backup first.

More on punctured arrays here:
http://www.theprojectbot.com/what-is-a-punctured-raid-array/
What's the URE on the old disks and new disks? Depending on the URE specs on those drives, you might expect the rebuild to fail in this process, if the RAID fails with the 10 TB drives.

https://www.zdnet.com/google-amp/article/why-raid-5-still-works-usually/

https://www.zdnet.com/article/why-raid-5-stops-working-in-2009/

I suggest RAID 10 or more disks and use RAID 6. What's the use case?
The OP mentioned that this array is a "DPM repository", ie a backup. Usual procedure is to have a Disk-Disk-"something else" backup, where "something else" could be a tape library, a removable USB drive, or some cloud storage. For this usage, I would be happy with RAID5. Should the shit hit the fan, just rebuild a new array, and copy the data back from the offsite storage. Would mean being without new backups for a few days, but this is often an acceptable risk.
ASKER CERTIFIED SOLUTION
Avatar of Davis McCarn
Davis McCarn
Flag of United States of America image

Link to home
membership
This solution is only available to members.
To access this solution, you must be a member of Experts Exchange.
Start Free Trial
Avatar of Pete

ASKER

Thanks for all the ideas, I've jiggled a few things around and now have the whole server at my disposal for Veeam. My estimated storage after year 1 is 25TB so I need to arrange my 10TB disks in a way that will be resilient and offer some future expansion space.

I have 8 drives to use (but don't necessarily need to use them all) what RAID would you recommend.

7 x 10GB - RAID 6 to give 50GB space?
Yes, that would work fine assuming it will be fast enough. RAID 6 is a bit slower than RAID 5 but the RAID penalty is offset assuming you have a decent amount of cache on the controller.
Avatar of Pete

ASKER

hey Davis McCarn the plan is now to add some identical drives to the existing RAID5 array, then make it RAID6. Do you think this is possible without data loss?
I could answer that but since you want Davis to answer I'll unsubscribe.
Avatar of Pete

ASKER

hey andyalder i would like you to answer of course, I was just replying directly to Davis post with info on the specific controller..
SOLUTION
Link to home
membership
This solution is only available to members.
To access this solution, you must be a member of Experts Exchange.
Start Free Trial
Read section 2.4.13 of the second manual I posted which says you can expand a RAID 5 to RAID 6.
Avatar of Pete

ASKER

You guys...