Understanding Metrocluster

[winged canyon]

Hello, I'm a new netapp end-user, currently learning about metrocluster and it's capabilities but got stuck on a few questions I couldn't figure the answer for.

(Referring to the image attached)
If a node in site A would fail, then I could still access the Diskshelf in site A via the node in site B. But why would I want to access those disks? Aren't they duplicated to site B already?
And the second question I have is how would a user that is conneceted to node in site A, access the node in site B when a failover happens to site A?

Would greatly appreciate any answers and help 🙏

[stiff totem]

If site A fails - you would switchover to site B. jeither automatically or manually. the data is syncmirrored to the other side as well as the cluster has SVMs on standby.

if just node A at site A fails, Node B at site A would do a normal takeover.

if you're in switchover and say node A at site B fails. node B at site B would take over all the workload from site A and B.

[mystic junco]

1. You don't really choose directly which diskshelf you access. You simply access volumes which are located on something called an aggregate which is made out of RAID groups. This aggregate is mirrored between the sites. There is a copy on site A and a copy on site B (called "plex" in ONTAP-language). If only the nodes on site A fail the replicated vservers (together with all the IPs and volumes) are started on site B. Since the shelves on site A are still up the aggregate continues to be in a mirrored state. So every write is still written on both nodes. Depending on the configuration the reads are handled only by the local copy of the aggregate or from both sites.
   The advantage simply is that your data is still being mirrored. If in that scenario for example the shelves in site B fail you could still access everything.

2. All configuration of the vservers (IPs, etc.) is constantly replicated between the sites. So if site A goes down the replicated IPs go online on site B. So your users wouldn't notice anything. They would not know that they now access site B instead of site A.

Also be aware that your screenshot shows a Metrocluster-FC where the backend has a Fibre Channel fabric. If you get a new Metrocluster you most likely will get a Metrocluster-IP nowadays where the backend is IP-based. The cabling is a bit different there but for the clients it's the same.