[OpenIndiana-discuss] Replacing an unavailable hd...

[Jim Klimov]

On 2014-06-16 11:05, John Doe via openindiana-discuss wrote:
> Hum... the server is a bit less than 2 years old, and all disks are
> plugged on the same back-plane, so I would be surprised if it was a
> cabling issue.  And, since I have spares, I prefer to replace the
> suspect disk to be safe and test it later...  If it is indeed a
> cabling issue, the new disk will also look failed I guess.

Not necessarily. In case of oxydation, the problem is intermittent
and will probably be fixed by the act of re-plugging, which scratches
off the oxyde films on contacts and brings the metal parts back into
better contact. If this is the case, however, then it is likely that
the situation will recur roughly each year, and can be "fixed" in the
same way.

Thanks to ZFS, at least, if there are any botched data packets between
the HDD and HBA due to weak signal/interference/noise when the film
is growing but does not yet preclude communications completely, then
you'd see checksum errors on the device (if there are mismatches
indeed) and the pool will try to amend that automatically.

It may also be possible that SAS/SCSI protocol also includes some error
correction and automatic retries for such in-flight mistakes, though
there is likely nothing like that for SATA/IDE protocols.

HTH,
Jim Klimov