Run P4300 G2 with six disks?

[MartinBH]

Hi,

Can I run a P4300 G2 with 6 disks instead of the usual 8?  I am assuming that I would shut it down, remove the 2 unwanted disks, then restart it and rebuild RAID.  I don't expect to retain data during the change...

thanks!

Martin

[a_o]

I've not done it the way you describe, but I think rebuilding  RAID is not going to be enough.

At a minimum, there would be a mismatch between the underlying RAID and Storevirtual's expectation of what should be there.

Normally, a P4x00 G2 would be configured as RAID 5. If you were to remove 2 disks, your RAID set would be gone.

However, conceptually,  I don't see why the Quick Restore DVD  should not be able to create a hardware RAID 5 with 6 disks during the restore process.

[MartinBH]

"Rebuild" was not the right word; I meant more like "recreate".

Martin

[oikjn]

unless you are doing this for ALL your nodes in that cluster you are going to have problems with the cluster afterwards.  

Think of the node like a single disk in a raid group.  If that node is suddenly 20% smaller, it is going to make all other nodes 20% smaller as well, so assuming you can manage to get the physical raid group reconfigured, the storage system IS definitely going to cause the above described problem just as it would do with a VSA node.  What you will see is that your usable space in your cluster will decrease significantly... not just by the amound of those disks on that one node, but multiply that ammount by the number of nodes you have in that cluster.

[MartinBH]

So, if I do set up the same config for all the nodes in the cluster, do you think it will work?

thanks!

Martin

[Bryan McMullan]

Theoretically, it should work.  I have some old IBM X3650's that run SAN/iQ 9.0 with six disks.  The only issue I can think of is that system validation may balk as it tries to create the initial virtual drives during setup and may not be able to create what it wants (as it would read 8 bays, yet only see six populated).  The other thing is that system monitoring may throw a hissy fit....

All guesses.  Why would you only want to run six?

[MartinBH]

Thanks for your response.  Sounds like this is not normal practice :-) so I would be living on the edge.

I'm considering this because I have no avenue to replace failed disks and am thinking of keeping a couple out to use if an in-use disk fails.

Martin

[oikjn]

martin,

no avenue because you can't afford a replacement disk or for some other reason?  They come up on ebay relitively cheaply so if you are considering going down to 6 disks, you might as well just keep an eye on ebay for them and snag a few when you can.

What you are asking to do is definitely not normal practice, but as long as the physical raid arrays are configured and not alarming, then it should work, but you are losing so much capacity in the cluster by doing this that nobody would do this unless they didn't pay for the unit originally and had no other alternative.   The store Virtual OS doesn't care what size the disks are that the underlying storage presents it and it will allow you take nodes with different capacities and join them to the same cluster and if you aren't paying attention then you might realize you are wasting what you are wasting.  Each node in CMC has a property line that says "Raw Space" and another that says "Usable Space".  When all your nodes in the cluster are the same size, the Raw and Usable space values are identical.  When you have different sizes of nodes in your cluster, the Usable space for EVERY node changes to be equal to the Raw space value for the SMALLEST node.  What that translates to is your cluster "Available space" is the sum of all your nodes usable space (keep in mind that you should be using NR10 so your actual net space you can provision to present to a host ends up being 1/2 the available space).