disk scrub generating too much i/o

[ALESSIOTORELLI]

Hi,

our FAS3160 filer reports a crazy i/o load during disk scrub interval.

On an aggregate with 7 SATA disks, DFM reports a throughput of 4000 blocks/sec, which I think is impossible.

We also have our monitoring system, which uses NetApp API, reporting the same result.

How can this be possible?

Any help appreciated

[ALESSIOTORELLI]

anyone has an idea about this issue?

[richardm]

Hi Alessio,

I've seen a Filer run with around 5000 IOPs during a disk scrub reported by Balance as an "Internal Process " but there was also contention with Snap Deletes, backups and volumes not being reallocated.  The first few spindles were trying to service 450+ IOPs a piece!  All I can suggest is to turn off features such as ASIS, snaps, etc and see if your Filer responds better during the scrub.

Good Luck!