What if: controller fails with a single P4300 G2

[Hardydrewandthenancyboy]

Supposing I have a single P4300 G2 in RAID5 configuration. So if it is 1TB disks giving 16TB native, I might have 12TB ish in usable space with raid5.
There is an active-active controller per node but is the second controller of any use to me if the first controller fails?
If I have network raid across the two nodes then yes but what if I'm using RAID5 across the whole box only?

P.S. This thread has been moved from Storage Area Networks (SAN) (Small and Medium Business) to HP StoreVirtual Storage / LeftHand. - Hp Forum Moderator

[Uwe Zessin]

I am not sure what you mean by "active-active controller per node" - from what I have learned, a P4300/P4500 storage node is a special-branded ProLiant server running SAN/iQ.

Network RAID-5 requires 4 storage nodes.

[Hardydrewandthenancyboy]

thanks for the reply, that was a quote from the quickspecs.
From what you say node=proliant server. So unless we configure network raid across the two Proliant servers that come within the P4300 then we will have a complete failure if one of the two Proliant servers in there fails?

So then we are forced into raid10 (which halves the usable space) or to buy a second P4300 if we dont want the nodes to each be a single point of failure?

[Hardydrewandthenancyboy]

Maybe rather than "network raid" I should say mirroring across the two nodes.

[Patrick Terlisten · Accepted solution]

Hello,

you have to configure the volumes with nRAID 10. In this case, two nodes are sufficient and if one node fails, the data is still available. A controller is a single node. So if HP is talking about dual-controller active-active, they mean at least a cluster with two nodes. A node is a controller. Each node has a Smart Array controller for the disks, so a single or double disk failure (regarding of the raid level) is covered by the Smart Array. A node failure is covered by network RAID.

Nework RAID 5 needs 3+1 nodes, so you need at least four nodes to use nRAID 5. For nRAID 6 (4+2) you need six nodes. nRAID 10 (two nodes), nRAID 10+1 (three nodes) and nRAID 10+2 (four nodes).

Hope this helps.

Regards,
Patrick

[Uwe Zessin]

Agree, the P4000 is a grid storage system and not a tradtional two-controller(module) storage array.

I explain "Dual Active/Active Controllers" so that a (simple) P4000 cluster has two storage nodes
which both provide services (Network RAID (NR-)0 or 10 volumes) to clients. There is no "standby storage node" like on the old MSA1000 with "active/passive" software.

And you are correct that NR-0 does not provide redundancy. If a storage node goes down, even if its temporary - all NR-0 volumes are (temporarily) unavailable.


On the other hand, you can grow pretty easy:

Start with a two-node cluster and NR-10.

If you need more space, expand with a third node - SAN/iQ can make use of the additional capacity AND retain NR-10 on the volumes.

[Hardydrewandthenancyboy]

thanks alot. Can you explain how the data is spread if you have three nodes and NR-10? Surely you would need four nodes?

[Uwe Zessin]

Let's say we have nodes: 1, 2, 3
and our data blocks are: A, B, C, D, ...

Then NR-0 is layed out like:
1-A, 2-B, 3-C
1-D, 2-E, 3-F
1-G, ...

NR-10 is layed out something like:
1-A, 2-A, 3-B
1-C, 2-B, 3-C
1-D, 2-D, 3-E
1-F, 2-E, 3-F

And NR-10+1:
1-A, 2-A, 3-A
1-B, 2-B, 3-B
1-C, 2-C, 3-C

[Hardydrewandthenancyboy]

makes sense, thank you