Richard L. Rhodes wrote:
A bunch of IBM NSeries 3220 (NetApp FAS2240-2)
OnTap 8.1.2p1
Ok so I'm pretty sure this version has wallocV2 algorithm in it. Not 100% though, no time to look it up now sorry
I'm about to start adding 2 disks to a bunch of NetApp systems - 2 disks into each system to expand CIFS shares. I will need to run reallocate to spread the vols in the expanded aggregates. There are around 30 vols in each aggr, some tiny and some a couple TB.
If it's wallocV2 in 8.1.2, then this won't make much different, if any. Mo use. You can just as well leave it and in 1-2 weeks it'll have sorted itself out. The only use for vol reallocate, is to get the layout in the particular FlexVol be more optimised for seq READ. If you don't care about that, then don't bother. Especially if you have high inode density and many inodes in these vols -- then your reallocate jobs may take a VERY LOOONG TIME.
Reallocate start –f –p /vol/<vol> How have you handled reallocate of a large number of vols in a aggr?
Refrain from doing it at all :-) No point.
There's a longer discussion here, but I have no time now, late evening here where I am. You may want to do reallocate start -A -o <aggr_name>. WARNING: this is impactful, it will be noticed, a lot on these 3220, on protocol latency.
It will vastly improve w-chain-lenght in the Aggregate especially if you've been "too full" before on them (>~85%). If you care about this depends on how much W you have in your workload
/M
Would you just start them all and let them run? If not doing them all at once, I could script them one at a time from AIX via ssh, but I don’t see a “-w” (wait) option to allow easy single threading. I would need a loop to monitor for a particular reallocate cmd completes before starting another. I’d rather not get into that much scripting effort.
Note that there is snapmirror and snapvault running every 6 hours on these vols. I don’t want to hit the systems so hard that these have problems.
Thanks!
Rick