smart self-test vs 7200.12/ich9r/ahci

[Andriy Gapon]

on 13/04/2009 18:51 Andriy Gapon said the following:
> on 13/04/2009 18:48 Sean Bruno said the following:
>> The "self-test" never goes beyond "90%" complete?
> 
> Yes, exactly. Even for the short one, which is supposed to complete in 1
> minute:

Another data point.
This disk was in an experimental 3-way mirror zpool, two other devices were
Hitachi HDP725050GLA360 and WDC WD5000AAKS-00A7B2. Self-tests were scheduled at
the same time for all 3 disks in smartd.conf. The Hitachi and Western Digital
disks were able to complete the tests before start of a work day, but the Seagate
disk was stuck at 90%.
Now I removed it from the mirror (designated it as a hot-spare) and now the
self-tests work as expected for it.
Makes me wonder: what was that access pattern that prevented self-test progress of
the Seagate disk but did not affect the other two.

-- 
Andriy Gapon