Red Hat

Are you sure? Can you qualify that a bit more? I always thought Raid 0 was stripe, thus giving a performance increase (yes I know not twice the speed etc), and Raid 1 was mirror giving redundancy. Is this something specific to CentOS or MDADM?

Just to clarify the hardware looks like this to the OS (sizes are approximate etc)

979.79GB - HW Raid 5 (made up of 8 discs)
979.79GB - HW Raid 5 (made up of 8 discs)

I would like to combine both physical volumes into this

1959.58 GB SW Raid 0 (made up of 16 discs)

This should use all discs and therefore see an increase in performance while having the ability to lose a drive from each of the Raid 5 arrays without losing the disc. It seems LVM does combine the two physical volumes into one big pool but doesn't stripe but rather appends one volume onto the other meaning unless you fill the first volume you will never ever use the second volume whilst not seeing any performance benefit.

My apologies, I confused raid 0 and raid 1. MDADM raid levels aren't quite the same as the traditional ones and I had to double-check.

You don't get the same order of performance increase with a software RAID that you do with a hardware one, though. Write speeds in particular won't be improved.

No worries, thought I was loosing the plot there for a minute haha

Corona,

RAID 0 does not give you any redundancy. It gives you the potential for double the throughput as you are executing multiple reads and writes to different areas. You are thinking of RAID 1, which is mirroring. RAID 0 is striping without parity. As long as the data is larger than the chunk size, the data is hitting both devices. Think concatenation with interleaving instead of appending.

---------- Post updated at 01:36 PM ---------- Previous update was at 01:35 PM ----------


RAID 5, with larger drives is potentially a disaster waiting to happen. I never suggest RAID 5 without at least a dedicated hot spare (global hot spare is fine as long as there are at least 2 per 5 drives).

However, it is still a poor choice to do this.

I think the hardware can do 'hot spare', the problems are capacity (I needed as much useable space as possible) and internally I think the hardware sees two individual raid controllers making global hot spare not possible. The drives are enterprise class drives so I don't doubt the quality. It's a balance based on the risk vs the importance of the data, and while downtime would be an inconvinience I could recover rapidly from the backups.