Isilon Cluster migration to new nodes

Question

Hello all - I am looking do to an Isilon cluser 'refresh' from X200 nodes to X210 nodes. Currently there are four X200 nodes we would like to migrate to four brand new X210 nodes and eventually move the original X200 nodes to a DR site. I was looking at Article Number 000304058, 'OneFS 6.5 and later: How to replace nodes in a cluster with nodes of a different type using SmartPools?' which seemed promising.  However, I noted that one of the requirements is that the nodes be of different node types.  I'm assuming X200 to X210 does not qualify.  If they do not qualify, can anyone point me in a direction for doing this 'migration'?  Ideally I would like to be able to add the new nodes into the cluster, transition the data to the new nodes and smartfail out the original nodes. One point not mentioned in that article - would I need or want to renumber the nodes so that there is a 'node 1' once all is done? Thanks in advance for any suggestions or insights.

Peter_Sero · Answer

X200 and X210 are different if drive and other specs differ significantly.

Good news is, when specs are identical or close, you have the /choice/

to declare the node types equivalent or not equivalent. Where the latter

means the cluster will split into two pool as you want it.

All relevant docs are linked here: Node Compatibility (Equivalency) and SSD Compatibility in OneFS 7.2 and Later - Isilon Info Hub

hth

Peter

AdamFox · Answer

Even if your have a system that meets the node compatibility matrix (same drives, equiv memory), by default the cluster will still set up the nodes in a separate node pool unless you tell it to make the nodes compatible. So if you don't do that, you will have 2 node pools and SmartPools can move data between those pools.

So either way it should work. This is the use case behind the reason for the compatibility to be manually turned on.

Hope this helps.

DHCSTI · Answer

Thank you!

DHCSTI · Answer

The specification for the two sets of nodes are very close; neither have SSDs. Original nodes:  X200-2U-Single-24GB-2x1GE-2x10GE SFP+-36TB New nodes: X210-36T/24G/2X10GE 2X1GE    (X210-SATA-010) So, I'm assuming that I'll be able to choose 'not equivalent'.  I look at the link you provided. Thanks! Dawn

DHCSTI · Answer

I do have a follow-up question.  After smartfailling (one at a time) the old nodes out, should I renumber the new nodes to then be 1-4?  I seem to remember once upon a time reading that some things automatically looked to node 1.

Peter_Sero · Answer

Speaking of side effects, never experienced an impact to normal cluster operation when renumbering nodes.

It's just hard to keep track of node identities over time for purposes like incident and performance monitoring where logical node numbers are used instead of node ids.

At least some good news is that in recent versions of OneFS the system logs include both numbers in each log entry.

Cheers

-- Peter

sluetze · Answer

It's not to node 1 but to the node with the lowest Logical Node Number. So you don't have to renumber and i would also not do this because i am scared of side effects

sluetze · Answer

sometimes im scared without logical reason.

DHCSTI · Answer

Thank you. I too am wary of side effects, and would rather not renumber if it is not necessary. Sent from my iPad

Isilon

Was this post helpful?