zfs discuss - Dec 2005 - Inappropriate ZFS diagnostic message

On page 116 in section 9.3.2 of
http://www.opensolaris.org/os/community/zfs/docs/zfsadmin_1205.pdf is:
"A complete zpool status output looks similar to the following:
# zpool status tank
pool: tank
state: ONLINE
reason: Data corruption detected.
action: Remove corrupted data or restore from backup.
..."
The most obvious interpretation of that output is that tank is online because
data corruption is detected. In this case the generic "reason-action"
pair of field names is inappropriate; "reason" should be
"problems", especially considering the fact mentioned in section
9.3.2.1 that the "reason" field is omitted if there are no problems.
Alternatively, if the generic "reason-action" pair must be maintained,
then insert between "state" and "reason" the line
"healthy: no".

Another issue:
Why is "zpool status tank" reporting the pool name as "tank"
but the first name under config as "test"?

Yet another issue:
How were 4 errors detected if there were no checksum errors on the pool or any
of the pool''s devices?
This message posted from opensolaris.org

On Tue, Dec 13, 2005 at 10:43:31AM -0800, Andrew wrote:
>
> The most obvious interpretation of that output is that tank is online
> because data corruption is detected. In this case the generic
> "reason-action" pair of field names is inappropriate;
"reason" should
> be "problems", especially considering the fact mentioned in
section
> 9.3.2.1 that the "reason" field is omitted if there are no
problems.
> Alternatively, if the generic "reason-action" pair must be
maintained,
> then insert between "state" and "reason" the line
"healthy: no".
This is precise terminology for just this reason.  The terms
ONLINE/FAULTED/DEGRADED have specific ramifications.  In this case, the
pool experienced an uncorrectible error from a device but was able to
recover due to replicated data available.

The pool is still online.  It is still providing the necessary data with
the necessary of replication.  However, it is one of the _devices_ which
is degraded.  So there is a problem with one of the devices in the pool,
though it doesn''t affect the overall health of the pool.

I agree that the term ''reason'' is a little misleading in this
case.  We
had originally had it termed "status", does this make any more sense.
I don''t think that having a "healthy" field would be any
benefit.  Once
again, the status of the pool (which includes device status, ongoing
resilvering, etc) is distinct from its state (which is precise under the
set of FMA guidelines).
> Why is "zpool status tank" reporting the pool name as
"tank" but the
> first name under config as "test"?
Sounds like a documentation bug.  We''ll get it fixed in the next rev.
> How were 4 errors detected if there were no checksum errors on the
> pool or any of the pool''s devices?
Probably another doc bug.

- Eric

--
Eric Schrock, Solaris Kernel Development       http://blogs.sun.com/eschrock