gmirror: degraded @ 100%

[Dmitry Morozovsky]

On Mon, 19 Feb 2007, Oliver Fromme wrote:

OF> Dmitry Morozovsky <marck at rinet.ru> wrote:
OF>  > root at office:/usr/local/etc# gmirror status m0g
OF>  >       Name    Status  Components
OF>  > mirror/m0g  DEGRADED  ad4g
OF>  >                       ad6g (100%)
OF> 
OF> It seems that the second disk is broken and locks up the
OF> channel near the end of the disk.
OF> 
OF> You could try to remove the drive from the mirror and
OF> run dd(1) on it.  If the dd command also hangs near the
OF> end of the disk, then you should definitely get a new
OF> disk drive.
OF> 
OF> If the dd command finishes without errors (and displays
OF> the correct number of transfered blocks), then I guess
OF> there must be a bug in gmirror.

This was my first guess, and I did exactly the same steps.

Actually, there was kernel panic just a couple minutes after my first mail, but 
kernel has refused to either dump or even reboot. Offending process was, as 
expected, g_mirror m0g, but I was unable to write down parameters (serial 
console was defunct at the moment).

What is wondering me, after reboot m0g finishes syncing successfully. I did 
recoverdisk /dev/ad6 /dev/null, and it finished without any error. Now 
recoverdisk /dev/ad4 /dev/null if working.

What is also making me nervous: smartctl on both bisks reports bad SMART 
checksum, and refuses to start any tests, reporting test already in progress, 
40% remaining.

I suppose I'll change both drives ASAP.

However, I'm still conserning geom_mirror bug(s) in corner cases, where some 
[not-so]-error conditions exist.

Sincerely,
D.Marck                                     [DM5020, MCK-RIPE, DM3-RIPN]
------------------------------------------------------------------------
*** Dmitry Morozovsky --- D.Marck --- Wild Woozle --- marck at rinet.ru ***
------------------------------------------------------------------------