Reconfigured disk RAID

[djuuic]

We currently have 4 nodes in a management group. There are two clusters - one with 2 P4300s and one with 2 P4500s. We have all of our volumes set to network RAID 10.

We'd like to reconfigure the disk RAID levels on the P4500 nodes from RAID 10 to 5. I was looking at the documentation and it looks like we would be able to remove one node from the cluster, then from the management group, then reconfigure the RAID on that node and add it back into the management group and cluster. Then repeat that for the other node. Is this right?

Since our volumes are set to network RAID 10 and we only have two nodes in the cluster, will it let us gracefully remove one of the nodes from that cluster?

[IainS]

I think you may want to do a little more reading of the docs first.

You will probably find that your SAN nodes are already in RAID5 for the disks in that array. I'd suggest you leave them at that although I believe you can change this to other forms of RAID for the entire node (and not for some of the disks - it has to be all that node).

Your volumes on the SAN have a form of network raid that is independent of the RAID internal to the node storage. This is either Network Raid 0 or Network Raid 10 for two nodes. Note I tend of think of it as RAID 1 really - it doesn't seem to really be RAID 10 equivalent. To do network RAID 5 you need to have I think at least 4 nodes and since yours are in separate clusters and no doubt are different sizes you can't really do that, nor do I think it is a good idea to do that.

What is the end goal you are trying to achieve?

And you can remove one node from a volume that is network raid 10 but you then only have one copy so no redundancy (outside of the RAID internal to the node). You will also need to be sure that you have a FOM setup so that you don't end up with neither node knowing which is the active one (split brain).

[djuuic]

--
You will probably find that your SAN nodes are already in RAID5 for the disks in that array.
--

Nope. The disks are in RAID 10 on the P4500s. It got set to this when they were first set up because of performance concerns - had bad experiences with RAID 5 on the old SAN. Now we think that was probably unnecessary, and we could use more space, so we wanted to convert them to RAID 5. That was the end goal.

Network RAID on the volumes is set to RAID 10 and we don't want to change that.

I'm starting to think that this isn't worth the risk, since it'll take hours (maybe days) for the nodes to do all the restriping necessary...

[IainS]

OK well not having done that I can't comment authoritatively, however I'm sure you can do that no problem.

RAID 10 on the node vs RAID 5 is a huge space cost so it is probably well worth converting one node at a time to RAID5. You will get some write performance decrease but perhaps not a very noticeable amount.

You need to decide if you want to risk running without the redundancy for the time to convert. It might not take that long to change RAID types though - but I imagine the resync of the RAID10 volumes could.

[teledata · Accepted solution]

To avoid having to re-stripe your cluster four times; twice-per-node (once when you remove the node, and then again when you re-add it back in) you should perform a "Repair Node".

This will allow you to remove the node, utilize the 2-way replication in its absence, reconfigure the HW RAID to 5, then re-add the node, then it will perform a re-stripe.

This will cut your rebuild time in half.

This of course will require that all your volumes are Network-RAID 10

[Jitun]

You could also move the volume to the other cluster and then changing the RAID to 5 on the Nodes and then moving the volumes back into this cluster.

[djuuic]

Thanks for the information.