[OpenIndiana-discuss] zpool iostat

Jim Klimov jimklimov at cos.ru
Wed Jan 16 00:59:52 UTC 2013


On 2013-01-16 00:58, Jason Matthews wrote:
>
> I am confused. Zpool iostat -v worn 5 1000 reports this...
>
> The three disks with goose eggs are offline. I am burning out the last few
> cells on all four SSDs so I can RMA.  However, I am confused why the only
> disk that works reports 149 write operations while the pool reports 465.
>
> How does that work? The system is running 151a


MAYBE, write-coalescing comes into play? Your OS sends several IOs
to the pool, which are seen to be destined for neighboring sectors
and reordered and aggregated accordingly while being sent to the
top-level vdev with a particular known layout (and it is possible
that for SSDs ZFS knows to use different algorithms - such as try
to fill a HW page of cells and reduce wear, but this would only be
known and relevant at tlvdev or leaf-vdev level of the stack)?

This is pure speculation from me, though based on things I've heard
and read casually ;)
//Jim



More information about the OpenIndiana-discuss mailing list