I read in this document https://library.netapp.com/ecm/ecm_download_file/ECMP1653502 that root volume of SVM needs to be configured for load-sharing mirror to prevent outage to the namespace when there is an outage on a node or duing failover. I have a two node (HA pair) cluster, even withought load-sharing mirror, when I perform a takeover of the the node containing the root volume of my SVM, I am still able to access the root of the namespace (\\cifsservername\c$). I have version 8.3.1.
Nevertheless, I am for following best practice, I just want to know when the mirror would help prevent an outage.
LS mirror is recommended for disaster recovery. If root volume is lost, you can promote LS copy to root volume. This is not transparent and requires manual administrator intervention. Of course, as long as root volume is not available clients cannot access data.
Is it safe to say that the main benefit aside from load-balancing of read access, is that in the event of a disaster where the root volume got deleted, that I will have another copy of it? Without the shared-mirror, I was still able to access the root volume even when I performed a failover that brought down the node hosting the root volume.
Also, I assume it is still a good idea to mirror the SVM's root volume to a DR netapp storage that is not part of the same cluster.
I do not think there is any load balancing in root volume access; and even if it were, root volume should be used for junction points only, so benefit is marginal.
For real disaster recovery you should probably look at SVM replication.
It is always possible to replace root volume with another one and manually mount everything - it just takes time (and proper documentation). So LS mirrors saves it for you. Having root mirrored to another cluster does not help, because to restore it you need running SVM and it cannot run without root volume - catch 22.