Copying mirrored partitions - will this work?
kraduk at googlemail.com
Wed Mar 31 11:05:48 UTC 2010
On 31 March 2010 10:22, Mike Clarke <jmc-freebsd2 at milibyte.co.uk> wrote:
> On Saturday 20 March 2010, Mike Clarke wrote:
> > I'm currently running 8.0-RELEASE and am considering experimenting
> > with 8.0-STABLE. I'd like to preserve my existing system in case
> > things go pear-shaped so I'll copy the entire system onto a spare
> > slice and then use csup to upgrade the copy to STABLE. Normally I'd
> > go through the steps of bsdlabel, newfs and then dump|restore to
> > create the copy but I'm wondering if I can take advantage of my
> > recently created gmirror to cut down the work.
> > I have two 500GB disks, /dev/ad4 and /dev/ad8, each partitioned into
> > 4 slices of 88, 88, 42 and 259GB. My system is installed on the first
> > slices (ad4s1 and ad8s1) which are mirrored as /dev/mirror/gm0. The
> > second slices (ad4s2 and ad8s2) are currently unused. My thoughts are
> > to temporarily add ad4s2 into gm0 with "gmirror insert gm0 ad4s2" and
> > wait for the mirror to synchronise. I should then be able to remove
> > the temporary addition with "gmirror remove gm0 /dev/ad4s2" at which
> > point ad4s2 should be a duplicate of the original system and I can
> > then go ahead and create a new mirror with "gmirror label -b load gm1
> > ad4s2" and "gmirror insert gm1 ad8s2". After editing /etc/fstab in
> > the new mirror to use gm1 instead of gm0 I should then be able to
> > boot into the system on slice 2 and upgrade it to STABLE while still
> > keeping my original system to fall back to if required.
> > Is this approach of moving disks from one mirror to another workable,
> > or have I missed something that would lead me into deep trouble? I
> > don't mind unduly if I make a mess of the second slice and have to
> > start again but I don't want to lose the contents of my original
> > system on slice 1.
> I decided to give it a try and the process went through very smoothly.
> It was much less tedious than bsdlabel -> newfs -> dump|restore, and
> quicker too. The mirror synchronised at a bit over 100 MB/sec but dump|
> restore only gave me about 10 MB/sec.
> The system has now been running for a bit over a week without any
> problems with either the original or cloned slices so I'm quite
> confident that things are OK.
> Mike Clarke
> freebsd-questions at freebsd.org mailing list
> To unsubscribe, send any mail to "
> freebsd-questions-unsubscribe at freebsd.org"
ive cloned many systems in this way before, and it does work. Not just
gmirror two, it should work with any mirroring, hardware or software. One
thing to remember though is that it works at the block level. Therefore if
the drive if very big with a small % of data on it might be quicker to copy
There is always an exception though. If you are using zfs, then only the
data on the drive is copied. This is because zfs works at block, and file
system levels, and therefore is aware of what is allocated on the disks.
One thing about your dump/restore speed. Did you play around with larger
block sizes? Increasing it should give you better throughput.
More information about the freebsd-questions