Remove storage node from cluster

[anunnthorn]

Hi, I have 2 P4000(each with raid-5 with no hot spare) they're in a cluster with network raid-10 (2 way mirror), the problem we have in this config is we can only utilize 50% storage capacity, our aim is to remove 1 node from the cluster, reconfig that node for raid-5 with 1 hot-spare, then move data from the other node that's still in the cluster on to this node and then reconfig the 2nd node with raid-5 with 1 hot-spare, this way we will have 2 separate storage node and utilize 100% of the storage capacity.

The question is, if I remove one node from the cluster to reconfig raid-5 with 1 hot-spare, with one node remaining in the cluster will the storage still function normally?

Anyone have done this before? Please share your procedure, thank you.

[oikjn]

Can you clairify.  what exactly are you doing and why do you think you are only utilizing 50% of your storage capacity?

Are you thinking of changing the network raid level from 10 to 0?  If so, that is generally a bad idea unless you don't care about the risk of having to restore from backups or genreal availability.  Also, network raid5 isn't an option with only two nodes, but even if it was, it has performance issues so you really should be sticking with network raid10.

If you are talking about chaing your hardware from having to groups of raid5 to one large group or raid5 and your thought is that adding one hot-spare will cover you, I would suggest against that since the rebuild time on such a large raid group would be so big that you would have a large chance of having an unrecoverable error during the rebuild process.

If you have your system setup as teh default specified by HP, you should keep it that way.  They made is that way for a reason and the storage capacity cost is just part of cost of using their system (which you shoudl have through about before you purchased it).  On the flip side, if you are in need of more storage space ASAP, you can always get another node to add to your cluster which will seamlessly add more capacity and also more performance.

[Rukam]

Network Raid 5 is not a best practise as the nodes have to do dual parity in writing the data and rebuilds take a longer time , SANIQ reads the volumes as network raid 10 as default, so if we use the network raid 5, the nodes have to two extra jobs-

1. One of converting the raid 10 volume read by saniq into raid 5 while writing the data to the volumes finally, and
2. Other of writing the dual parity.

Hence the 30 % of the top layer of the NR% volume is NR10 as Saniq reads and writes data in NR10.

This is a tedious job for the SAN and as compared to NR10 the performance is not very good.

Please check the current volumes if they are thin or full provisioned ?

Thin Provisioned Volumes are most effectively used when you know how a volume will grow over time, the growth is predictable, and users do not need immediate, guaranteed access to the full volume size. We must monitor thin-provisioned volumes and must be prepared to increase storage space by adding or moving arrays to the pool where the volume resides. When properly managed and monitored, thin provisioning can result in lower storage costs due to better disk utilization and can reduce some of the administrative overhead associated with monitoring individual volumes.

Remember we can change from thin to full or from full to thin without a downtime and any effects on the performance of the LHN SAN.

I hope that helps .

[Emilo]

Unless I am mistaken I dont believe you can reconfigure the HW Raid in the manner you propose.

At least not from the CMC. You might be able to do it through the backend but You have to be carefull and mess up the OS logical drive.

You are probably better off calling support and have them take the node out for repair and converting to

Hardware RAID 6 not a hot spare but as close as you can come with the options your are provided in the reconfigure RAID menu.

You are correct you are only getting 50 % of your space but this does provide you with

Redundancy and Performance

There are other ways to to reduce space all of them have thier Pros and Cons.

[anunnthorn]

Thank you for all your comments, after I thought about it the more I understand that HP LeftHand relys on network raid because each node only has one controller, so I was by design to have at lease 2 node, therefore these's isn't a whole lot we can do expect add more node, thank you.