left sequoia
Let's say we have two differnt type of data for FabricPool tiering, one is for all backups / volumes, the other type is all snapshots located on all the other dev/prod volumes. To distingush these two types of data,
Currently, there is one bucket by default. If somebody please show me steps / docs on creating second bucket and how to use it?
thin bay
you can't distinguish data on a volume basis... FP buckets are attached to aggregates... you can tag volume data
if your data is segregated by aggregate, then you just need to have a different bucket for for the two types of data, but there's really no point to any of this. Multiple buckets just cause more headaches when you move volumes that are have FP in different buckets
brisk ibex
Multiple buckets just cause more headaches when you move volumes that are have FP in different buckets
civic breach
> Multiple buckets just cause more headaches when you move volumes that are have...
unless you're paying for egress/ingress, there's no real "heaache" involved. On the other hand, optimized volume move (if multiple aggregates share an FP bucket) had some nasty data-loss bugs in previous versions so you want to make sure you're on the newest ONTAP patch for your release before implementing that scenario
thin bay
there's a lot more load on the grid with all of the ingress/egress... i moved a few PB that way, but it was between systems
civic breach
yes, if you do a lot of vol moves, using a single bucket is better. Sure, it depends on your workload, but from my experience, the number of customers who do more than like 5 volume moves per year is vanishingly small
thin bay
it was a migration/consolidation
brisk ibex
Having 1 bucket per cluster could benefit of optimized vol move also when refreshing performance tier hardware or consolidate as previously mentioned.
@civic breach have you any bug reference in order to verify the right Ontap patch release?
left sequoia
Your points are well taken. But, What if I move a volume to the other aggregate belong to two different buckets? Would it still work but just take long time to move objects from one bucket to the other along with moving data in performance tiers?
brisk ibex
@left sequoia yes, cold data will be rehidrated and then copied back to cloud tier based on the tiering policy. Performance tier must have room to accomodate cold data. Throughput will depend on multiple factors.
civic breach
Having 1 bucket per cluster could benefit of optimized vol move also when refres...
The one I was thinking of is SU414 (https://kb.netapp.com/Support_Bulletins/Customer_Bulletins/SU414)