Problems with dump of an atacontrol-ed RAID1 array

Jonathan Reeder jreeder at
Tue Jan 18 10:31:13 PST 2005

I have an ICH6 SATA RAID controller that I have set up with atacontrol to
behave as a RAID1 array.  The array (ar0) consists of ad4 and ad6.  Last
night I started a dump like so:

# dump -0Luaf /dev/da0s1 /

to *hopefully* dump my root filesystem (ar0s1a) to a usb storage device.
Now here is where things get a little fuzzy...

The dump started just fine, and I was able to stick around until dump
reported that it was about 10% done.  At that point, I left for the night
and then woke up to calls from users saying that they couldn't access
services on that box.  When I got to the terminal, I noticed that the last
couple of dump's output messages on screen had to do with being unable to
read from ar0s1a.  I hit ^C to kill the dump and was asked if I wanted to
abort.  I said yes and that was the end of that terminal.  I couldn't Alt+Fx
or ssh in, so I had to power down the machine and bring it back up.  When I
did, ar0 was gone!  So obviously the boot process told me it couldn't find
its root filesystem and asked me for another one.  I gave it ad4s1a and the
machine came up just fine, albeit without RAID.

So... has anyone had any experience with this?  Does a rigged atacontrol
software RAID not play well with a dump?  I'm trying to figure out:

a) why my dump failed
b) why my ar0 RAID array was destroyed in the process I described above

Lastly, if anyone with experience with atacontrol can tell me the best way
to rebuild my array without losing any data.  I *think* that I need to just
create an "array" that only consists of one of my two drives (probably ad4
since its the one that's live right now), then add a hotspare (ad6), and
rebuild.  Is that right?  The only real step I'm worried about is the
"atacontrol create" statement, because I'm afraid it will wipe the drive
right off the bat.

Thanks a bunch,

Jonathan Reeder

More information about the freebsd-questions mailing list