cDOT - convert 2 node cluster to 2 node stretched Metrocluster

[Fambo]

Hello,

We have a lab setup that consist of a 2 node cDOT cluster running 8.3 soon to be ONTAP9. We would like to see if we can convert this 2 node HA pair into a stretched MetroCluster. Yes, both nodes and shelves are all in the same rack – this will be for testing only.

Current equipment –

Controller -         FAS8040 HA pair – 2 in 1

Shelves/Desk – 1@DS4246 w/ 3TB SATA drives

                         3@DS2246 w/ 900GB SAS drives

Cluster Switch – CN1601

I believe since we have the 8040 we will need to add the VC card for NVRAM exchange. I think the hurdle maybe the shelves.

1. Does the DS4246 with SATA drives pose a problem with the conversion? Could I assigned 12 of the disk to each node/site? If the SATA can’t be used within the MetroCluster can I set the ownership to just one node/site and accept that I don’t have site-site protection?
2. Everything is in the same rack and cabled for MPHA with copper SAS cables, do we have to go to the optical SAS cables or can we use the copper? Once again this is not for live use and performance is not a concern.

Tear down and rebuild – I believe a new MetroCluster comes with some level of factory pre-configuration. Would there be anything that would prevent us from tearing down the existing cDOT cluster and converting to a 2 node MetroCluster?

Thanks in advance for you expertise and guidiance

[aborzenkov]

SATA disks are supported. I do not think cDOT MC allows non-mirrored storage, you need to plan your disk space accordingly. Logically, SAS optical cables are needed only to connect components across long distance, what follows, if SAS copper length is enough, I do not see why you cannot use them.

MetroCluster is simply two single node clusters with relationship setup, so nothing prevents you from doing it in the field. You will need second cluster license though to build second cluster.

[niels]

Hi Fambo,

there are some things you need in order to physically build a MetroCluster:

- a second chassis as the controllers must be physically seperated, otherwise HA will communicate over the backplane rather than the FC-VI card

- two FC-VI cards 16Gbit (X1928A) + SFPs

- two seperate storage stacks

You can use Cu SAS cabling. It's fully supported and sufficient in a single-rack configuration.

But the storage needs to be seperated in two stacks. ONTAP will complain in case the local and remote pool is attached to the same SAS ports.

With ONTAP 9 unmirrored aggregates will be supported. This gives you an option to use the SATA disks as well as one of the SAS shelves as unmirrored aggregate(s).

The other two SAS shelves, each in a different stack, would need to be split to 12x pool0 for the local node and 12x pool1 for the remote node. So I'd recommend to setup a RAID4 root aggregate (1+1) which would leave you with 9 disks for a RAID-DP data aggregate (7+2) and a spare per pool.

You won't need any cluster network switch as with a 2-node MCC each site is just a single-node cluster.

For that you need a second cluster base license. For a lab environment you can basically use the same license code at both sides (but that will also generate the same cluster S/N for both clusters - we have that running this way in the Customer Briefing Center in Munich without issues so far)

You should wipe the cluster and setup the 2-node MCC from scratch according to the documentation:

Clustered Data ONTAP ® 8.3 - MetroCluster ™ Installation and Configuration Guide

This would guide you through the process of configuring the controllers for 2-node MCC.

But as this is for 8.3.x, you won't be able to have unmirrored aggregates. Unfortunately I cannot point you to the ONTAP 9 docu yet. So I'd suggest to eaither wait until it's available or start off with a fully mirrored 2-node MCC in 8.3.2 and don't assign one of the SAS and the SATA shelf to any node.

regards, Niels

---

If this post resolved your issue, please help others by selecting ACCEPT AS SOLUTION or adding a KUDO or both.

[MohammadAli]

I want to convert two single node clusters FAS8040 with cDOT 8.3.2 in two different sites with production data on each site. I have all the required HW and SW with licenses. But only confusion I have if this conversion will affect my production data? Do I need to migrate my data to somewhere else before I start converting it to Metro Cluster, and then migrate back to the Metro Cluster?

One of my friend is telling I have to do the following:

1-      Install transition gear to migrate data from FAS8040s to itself.

2-      Migrate data form FAS8040s to transition gear.

3-      Add metrocluster parts ( FC-SWITCHES  , SAS-Bridges , FC-VI cards ,

and cables ).

4-      Do metro cluster configurations and testing.

5-      Redo data migration from the transition gear to metrocluster

systems.

Is this correct? I never heard about Transition Gear.

Regards,

Mohammad Ali

[niels]

Hi,

your friend is correct.

You cannot configure a MetroCluster as long as it has SVMs, volumes and data present.

The data need to be evacuated from the existing system in order to wipe the systems and re-configure them as MetroCluster.

The "transition gear" is nothing else than a third (and in your case possibly fourth) system you temporarily move your production data to.

Sorry - there is no other way to convert two single node systems into a 2-node MetroCluster.

Kind regards, Niels

-----

If this post resolved your issue, please help others by selecting ACCEPT AS SOLUTION or adding a KUDO or both.

[_Sunil]

Hi, 

I have simple question. 

In our enviroment, we have 6-Node FAS8060 C-DOT cluster for prod and DR has been configured with 2-Node FAS8040 C-DOT cluster.  With this situation is there any possiblities that we can convert the current setup into fabric metrocluster?

Thanks. 

[sgrant]

Hi,

Unfortauntely, you could only use the Production FAS8060 systems in the configuration, since you must have an matching HA-Pair at each site (NVRAM etc). You could therefore have a 4-node MetroCluster with 1*FAS8060 HA-Pair at each site. You can of course have a seperate 2 node MetroCluster with a single node at each site for the remaining systems.

https://library.netapp.com/ecm/ecm_download_file/ECMLP2494091 page 10.

You'd also need some switches for the storage as well as the MetroCluster license.

Thanks,

Grant.