Our organization has an aggregate that hosts several volumes used as NFS datastores for our VMWare environment. These volumes have 2 days worth of daily snapshots and are deduplicated. They deduplicate extremely well. The volumes are roughly 1.3t in size and deduplicate down to roughly 800-900g.
We were getting close to hitting our theoretical IOPS maximum for the aggregate so we decided to add a shelf of disks to the aggregate before migrating any more virtual machines to the datastores hosted on this aggregate. Time to reallocate.
We are running ONTAP 8.1.1 which says it can reallocate snapshotted and deduplicated blocks by using the -p reallocate switch. In previous versions of ONTAP, my understanding is that it was impossible to reallocate deduplicated blocks. If you dont use the -p switch with ONTAP 8.1.1, only non-deduplicated blocks will be reallocated. Given the great deduplication ratio on the volume/datastore (and snapshots) we decided it was in our best interest to use the -p switch.
Is there any way to estimate how long a reallocate start -f -p /vol/<volunmeName> will take to run?
Also, is there any documentation regarding the number of phases/levels when running a reallocation? I had no idea if there were 100 phases or just 4 phases.
Additionally, the percentages displayed by the reallocate status command are misleading because they are specific to that phase/level. Without an overall reallocate "percentage complete" metric, I have no idea how far along the reallocate truly is.
The plan was to finish reallocating all of the volumes on the aggregate over the weekend. Unfortunately, it took 15.5 hours to reallocate one of the 1.3t volumes so we werent able to get through all of them over the course of the weekend. Without knowing how long it would take to reallocate the remaining volumes, we decided to wait until the following weekend to finish up. We are not comfortable running a reallocate during the day as the volumes are fairly busy and reallocates seem to hit the filer processor/disk pretty hard. We also do not want to run more than 1 reallocate at a time.
So in effect, the process of adding a shelf, and taking full advantage of it, looks like it is going to take weeks as we spend 15+ hours reallocating each volume over the new disks in the aggregate. I'd like to know how to estimate how long a reallocate will take so I can better manage the process of adding additional shelves.