Hi, Thanks both of you for the info and docs provided, really appreciated. I've read some 8.2.2 Clustered Documentations already. I will create a HA pair on the new filers with new SAS disks and join those 2 new filers to the cluster as a front end, but the back end with disks will be on its own HA cluster only. (referred to the config on https://library.netapp.com/ecm/ecm_download_file/ECMP1367947 (page 13)). "HA pairs are components of the cluster, and both nodes in the HA pair are connected to other nodes in the cluster through the data and cluster networks. But only the nodes in the HA pair can takeover each other's storage."
I'm still slightly confused if I must disable a HA on the existing node at first and then join each filer to the cluster or not. My goal is to join the new filers to the cluster so that the volume migration will go easily rather than using a snapmirror which can be slower. Please let me know
Yes, documentation is rather confusing as it mixes "HA" to mean "HA pair" (actually storage failover) and "HA mode" (special epsilon-less mode for two node cluster). You have to disable HA mode. It may be done automatically by wizard when you join the third node in 8.3 - at least, express guide does not explicitly mentions it. But in any case you should check and verify that it is disabled.
SnapMirror won't be really slower, as volume migration between HA pairs will still be performed over 10G ethernet. But volume move in a cluster will be completely transparent to clients - there won't be any cut-over phase comparing with SM.