StoreVirtual Storage
cancel
Showing results for 
Search instead for 
Did you mean: 

Network RAID clarification

ALgon
Advisor

Network RAID clarification

Network RAID clarification

 

Hello,

 

We have 8 P4500 G2 nodes configured in a multisite cluster - so 4 nodes on each site.

 

My understanding is that volumes configured as network raid 10 would allow us to lose an entire site and the volume would remain online.  Is this correct?

 

With raid 10+2, we could lose an entire site plus a single node at the remaining site?

 

So, we have just purchased 6 P4730 G3 nodes and configured this as a multisite cluster - so 3 nodes on each site.

 

With network raid 10, I would still assume that we can lose up to 3 nodes, or an entire site and the volume would remain online.  However, when testing this, we were only able to shut down two nodes and when attempting to shut down the third node it warned us the doing this would take the volume offline.  Can someone help clarify this a little more for me? 

 

With 3 nodes on each site, would we need to configure the volumes as 10+1 to allow for a complete site outage?

 

Many thanks,

2 REPLIES
oikjn
Honored Contributor

Re: Network RAID clarification

Your assumption about NR10 is kind of correct.

 

There is technically one mirror of each data point in NR10 so you could lose UP TO 50% of  your nodes and not have a problem, but it has to be the "right" nodes.  If you lose the two nodes that have the same data on it, you will go offline.

 

If you setup your management group correctly and identified the sites for the nodes correctly, it will make sure that your mirror copies are on different sites so you could lose one site and keep your data active (assuming FOM is setup at a 3rd site).  

 

for multi-site environments setup correctly, NR10 can be considered site redundancy.  If you want site+node redundancy, you need NR10+2, or NR10+1 if you can handle one site w/out the node redundancy.

 

Your statement about the test you did with the warning is a bit odd for a correctly configured setup.  Did you have the correct number of managers running?  How about the FOM?  If you look at the node availability tab it will tell you what is affected and why its affected.  If you had 6 nodes in the cluster and you were trying to shut off the third node, it should not have been a problem unless you were just picking random nodes to shutoff and you happened to pick a node that was mirrored to one of the nodes you just shut off.

Bart_Heungens
Honored Contributor

Re: Network RAID clarification

Hi,

 

Take a look here, there are some pictures that explain how the blocks are spread across the nodes

http://www.bitcon.be/?p=2537

 

The trick is to put the odd nodes on 1 site and the even nodes on the other site... Naming convention of the nodes can be rather important to avoid confusion...

You can check the node order at cluster level (and change it also which is not recommended...)

 

 

Kr,

Bart

--------------------------------------------------------------------------------
If my post was useful, clik on my KUDOS! "White Star" !
My blog: http://blog.bitcon.be