2011-10-23 03:31 AM
We are running several metro clusters in NFS and CIFS environment in our work place but haven't got a chance to configure one in SAN. I have couple of questions regarding Active/Active Metrocluster in SAN which are
1. How can a host access LUN on primary site. Do we have to buy seperate FC switches other then the metro cluster switches for data network for SAN configuration ?
2. How can a host access LUN in the event of disaster. Let suppose a NetApp FAS6280 (primary site system) mother failed at some time. So, in that case how can a host access LUN (single image) at secondary site ?
2011-10-23 04:07 AM
1. The MetroCluster FC switches are dedicated to interconnection of NetApp components. It is not supported to use them for any other type of traffic, including general purpose SAN connection from hosts.
2. MetroCluster is the same active/active system so standard SAN configuration applies. Your hosts are assumed to be connected to both controllers and to use multipathing software. So in your example if motherboard (I believe, that is what you mean) in one controller fails, another controller initiates takeover and makes resources available to all connected hosts. There is no difference to standard active/actve configuration.
2011-10-23 09:22 AM
All correct.. I wish we had more MetroClusters but don't sell that many in the US. I hear EMEA has a higher install base for this.
Even if we could mix SAN and MetroCluster traffic (we can't correctly answered already), I wouldn't want to put SAN traffic on the MetroCluster fabric. However, I noticed in the 8.1 release notes and ha guide that the MetroCluster FC switches can now be shared for 2 MetroCluster pairs (NetApp connections only like you said)... I don't have any customers with 2 MetrocCusters but if someone needs more than one cluster this will save some cost.
MetroCluster support for shared-switches configuration
Beginning with Data ONTAP 8.1, MetroCluster supports shared-switches configuration. In a shared-switches configuration, two fabric-attached MetroClusters share the same four switches and the ISLs between them.
2011-10-23 10:04 PM
Thank you for your reply.
Regarding the second question what I still dont get is "Do we have to have a fc path to the remote site" ? assuming that we have host only on one site. Cause as far as I understand, for single_image to work you need paths to both the controllers. This would also mean that they need to extend their SAN fabric to the other site? Or else how could we connect the host to the remote node.
Hope it makes sense.
2011-10-23 10:10 PM
Yes, you need to have FC path to remote site.
If all you want to achieve is physical data redundancy, may be you could consider standard active/active pair on one site with SnapMirror to another site. This does not require having extended fabric across both sites.