Andrea Venturoli ml at
Sat Mar 28 10:03:13 UTC 2015

On 03/28/15 06:05, CK wrote:
> Regarding the unexpected loss of files from the filesystem under various
> loads, is the appended 'smartctl' data sufficient to make the determination
> that the loss of files while the operating system is in use could be due to
> the condition of the drive?

I half trust smartctl and I half trust the logs: if these give errors, 
the drive is bad (although it could possibly be repaired by the vendor 
The only thing I really trust are the vendors testing tools (Ultimate 
Boor CD might be a good start), but you can't run them while the disk is 
mounted on a running system.

> 1)  I would expect a FreeBSD error to the effect of "unable to read/write
>      /dev/ada0" or "block checksum does not match block data".

Usually I get read errors in the log when a drive is bad.

> 2)  I would expect that all data read/written to from a drive is verfied to be
>      correct by FreeBSD with checksums, and that it is guaranteed to be correct
>      if there are no serious and fatal errors reported by the operating system.

Can't tell, sorry.

> My past experience has only
> taught me that when a drive begins to make very bad noises, this generally
> accompanies obvious and serious problems

When a drive starts making noises it's already too late; SMART was born 
to warn in advance. Whether it works (or to what extent) is an open issue.

> and that a drive fails when the mechanical parts fail,

I've also seen dead drive which were mechanically perfect: it was the 
electronics which died (although I admin this is less frequent).

> SMART overall-health self-assessment test result: PASSED
 > ...
 > SMART Error Log Version: 1
 > No Errors Logged

This is a little reassuring, but don't trust it blindly.
If you suspect something, run the tests.

> Offline data collection status:  (0x84)	Offline data collection activity
> 					was suspended by an interrupting command from host.

Maybe it's normal, I don't know, but I would investigate this.

> Self-test execution status:      (   0)	The previous self-test routine completed
> 					without error or no self-test has ever
> 					been run.
> ...
> No self-tests have been logged.  [To run self-tests, use: smartctl -t]

If you cannot stop the system and run the vendor tools, I strongly 
suggest to at least run these tests.


More information about the freebsd-questions mailing list