[zfs-discuss] Is a scrub a data read-only or read-write operation?

Michael Kjörling michael at kjorling.se
Sun Jan 26 08:30:44 EST 2014


On 25 Jan 2014 22:55 -0200, from durval.menezes at gmail.com (Durval Menezes):
> On Jan 25, 2014 10:43 PM, "Cédric Lemarchand" <cedric.lemarchand at ixblue.com>
> wrote:
>> But yes, there are some writes too, I don't know what and why exactly.
> 
> Wild guess: scrub progress/checkpoint counter?

Yes; I think I covered those in my original question. :-) I suppose
the real question is what, if anything, _other than that sort of pool
metadata_ gets updated or rewritten during a scrub. Particularly, what
if anything _relating to the user-stored data_ gets updated or
rewritten during a scrub?

Obviously if an error is detected then depending on the level of
redundancy available that fact will be recorded and/or the data
healed.

However, for the purposes of my question, let's play the naivity game
and for a moment assume a _perfect_ storage device, defined as one
that when asked to read a specific location never presents anything
but exactly what was last written to that location. Yes, that includes
the entire path from RAM to physical storage media and back, including
cabling, drive firmware, etc.

-- 
Michael Kjörling • http://michael.kjorling.semichael at kjorling.se
OpenPGP B501AC6429EF4514 http://michael.kjorling.se/public-keys/pgp
                “People who think they know everything really annoy
                those of us who know we don’t.” (Bjarne Stroustrup)

To unsubscribe from this group and stop receiving emails from it, send an email to zfs-discuss+unsubscribe at zfsonlinux.org.



More information about the zfs-discuss mailing list