Maximum space used for an Aggregate

[aclaus]

Hi all, I have a question about how much space of an Aggregate can be used, what the best practice is.

We currently have a SAN with 4 Aggregates.

1. Aggr0 is the system aggregate, 96% is used of this aggregate.
2. Aggr1 is devided between cifs and vols, 99% of this aggregate is being used.
3. Aggr2 is devided between cifs and vols and 95% of this aggregate is being used.
4. Aggr3 still has 6TB of space left and will be fully utilized in the near future.

My question is, is there a maximum usage of space for an aggregate? For the last couple of days I'm getting the message 'Critical: Aggregate aggr1_01 on node "name" is 99% used' and 'Critical: Aggregate aggr2_02 on node "name" is 95% used'. Nothing has happened yet, no deteriozation of speed or anything like that. I think the alarm threshold has been put in by the company who helped us setting up the SAN. But I was wondering what the best practice is for an Aggregate space usages.

Hope to hear from you soon and thank you for your time.

André

[clackamas]

We strive for high utilization and get nervous at 95% or above.  If you are thin provisioning, like we do then I really try not to go above 95%.  With that said we only see performance issues when its 99% or more full.  You may have aggregate snapshots on and you can usually get 5% back.

FAS3070*> snap sched -A

Aggregate aggr64: 0 0 0

FAS3070*> snap reserve -A

Aggregate aggr64: current snapshot reserve is 0% or 0 k-bytes.

If you have aggregate snapreserve, was 5% default for quite some time, you can change that to 0 and then set the aggregate snap shot schedule to 0 0 0.  You don't need aggregate snapshots, they are only used for Metrocluster.

[DOMINIC_WYSS]

above 90% usage you will have performance degradation.

now it depends if you have enabled thin provisioning or not. if you don't use thin provisioning, it may show a 99% full aggregate with df -A,

but you may still have space left.

if you use thin provisioning, then don't go over 90%!

have a look at aggr show_space (-h), there you can see the usable space.

and keep an eye on the snapshots. maybe there are old snapshots to delete on the volumes.