I am wondering about the snapshot topic for a svm root volume (especially using NFS and CIFS protocol), because I did not find any best practices related to such matter.
I have various clusters running CDOT 8.3.2 and ONTAP 9.2P1. Most of the devices are FAS2552, but I have 8040 & 8080EX as well. For each svm I am configuring 2 Load-Sharing mirrors in order to protect the svm root volumes, this work pretty good. My concerns are the following :
- Is there any best practice concerning the snapshot policy for an svm root volume ? Should it be default (6x hourly, 2x daily, 2x weekly) or is it not necessary to have this amount of snap. copies.
- what should be the snapshot reserve size ? Current configuration is more often 5% but some clusters have 10%. What should be set for these LS mirrors ?
I don't believe there is a recommendation on number of snapshots to retain - that would be something for each admin to decide on the basis of their own risk assessment. The default policy is probably fine. The volumes additionally are pretty small, so snap reserve of about 10% is fine.
Finally - LS mirrors provide the most benefit for scale-out clusters of 4 or more nodes. In a simple HA two node cluster, storage HA will address the risk without the slight disadvantages in delays for NFS export policy changes.