My question is to address all the "snapshot reserve 90%" full alerts we're getting from unified manager. This seems to be an issue in every site.
These root volumes don't seem to have autodelete turned on but they are still within the default 5% reserve space, just nearly full. They have the default policy applied 2 weekly, 6 daily and 6 hourly or whatever it is.
I didn't want to go making changes to the root volumes and aggregates everywhere without knowing what the best practices are.
Can anyone advise me on how to keep the snapshot reserve below 90%?
I am assuming you are talking about SVM root volume. By default, it is 1GB which may sound like a small size but usually it is more than enough for SVM namespace related meta-data storage but due to "5%" default snapshot reserve, for 1GB it comes to very minuscule size i.e 50MB and therefore it is no surprise that most SVM's root vol would have either reached 90 % or overshot their snapshot reserved. Hence, it may be generating those alerts, but it is best to to stick to 'Default' policy and leave it alone. Default is : 2 Weekly, 2 Daily and 6 Hourly and this schedule ensures that snapshots are re-cycled to free up space in the root vol. If you do not wish to get these alerts, then you can either increase the snapshot reserve (as the 5% of 1GB is 50MB only) or just make it 2GB (which will make snapshot reserve 100MB).
On the side note: auto-delete is not turned on by-default for any volumes, and I don't think it is necessary for root-vol. That default schedule is good enough to recycle snapshots.
I have seen similar alerts on SVM root volumes, regards to snapshots, when junction-path volumes have high change rate. It can get quite noisy with alerts tripping on both root & LS volumes for same issue.
My standard is now is deploying Load Sharing mirror volumes set to 3g and SVM root volumes increased to 2G with 10% snapshot reserve.
Its worth burning the couple of gigs to keep the alerts down.