Sorry for the late response. I just want to say that you was right! I told my customer what was possibly the reason, but he wants me to verify it by NetAPP. And guess what? They came up with the same KB article.
Beside the possiblity to just delete the volume they came with another possibility.
I paste what they wrote so that it may help somebody else out there!
The ‘sis start -s <vol> command discussed in the KB is resource-intensive. To avoid the performance hit,
you can just reset the sis process for the volume. That will clear out all metadata and allow the
fingerprint database to start fresh manually. It will not eliminate the deduplication that has already
Stop deduplication for the volume if it is active or queued. Afterward, it can be restarted:
filer> sis stop -a /vol/volume_name
The sis reset command:
filer> priv set diag
filer*> sis reset /vol/volume_name
filer> priv set admin
The ‘sis reset’ commands should run in a matter of seconds, a minute at worst.
*Note*: ‘sis reset’ is interactive and will ask some confirmation questions.
This will delete all the volume’s metadata in the volume and aggregate, but will not affect the
currently deduplicated data. However, it will require a rebuild of the tables from scratch.