Skip site navigation (1)Skip section navigation (2)
Date:      Tue, 3 Dec 2013 13:35:00 -0700 (MST)
From:      Warren Block <>
To:        Julien Cigar <>
Cc:        "Mike." <>,
Subject:   Re: gmirror, gpart and MBR vs GPT in the Handbook
Message-ID:  <>
In-Reply-To: <20131203173700.GD86280@mordor.lan>
References:  <201311301303210813.05DE187E@smtp.24cl.home> <> <201312011121580096.005D00FB@smtp.24cl.home> <20131202103122.GN66981@mordor.lan> <> <20131203092604.GA86280@mordor.lan> <> <20131203173700.GD86280@mordor.lan>

Next in thread | Previous in thread | Raw E-Mail | Index | Archive | Help
On Tue, 3 Dec 2013, Julien Cigar wrote:

> On Tue, Dec 03, 2013 at 09:27:12AM -0700, Warren Block wrote:
>> On Tue, 3 Dec 2013, Julien Cigar wrote:
>>> On Mon, Dec 02, 2013 at 08:56:13AM -0700, Warren Block wrote:
>>>> On Mon, 2 Dec 2013, Julien Cigar wrote:
>>>>> If you want to use GPT with gmirror you may want to mirror each
>>>>> partition instead of the whole disk.
>>>>> For example on my box I have the following:
>>>> There are a couple of potential problems with that.  The big one is when
>>>> a disk fails and is replaced.  If you're not careful, the rebuild of all
>>>> those mirrored partitions will start at the same time.  Head contention
>>>> will bring that to a near-standstill.  It also puts a heavy load on the
>>>> drive that still works.  Hopefully it is not the same model and age as
>>>> the one that failed, or its "warranty timer" may also be close to
>>>> expiring.
>>> That's true, I turned off autosynchronization of stale components to
>>> avoid this kind of bad scenario ..
>>>> A less-serious problem is that only the partitions are mirrored.  That
>>>> leaves out metadata like the partition tables and bootcode, but those
>>>> typically do not change very often and might not be a problem.  The
>>>> admin has to remember to manually install such things on a new
>>>> replacement disk, though.
>>> It's just a matter of gpart backup / gpart restore, right ?
>> That will get the partition tables but not the bootcode or PMBR.  A
>> failed disk could then leave a mirror which has all the data but is
>> unbootable.
> ok, so I've to be sure that:
> $> gpart bootcode -b /boot/pmbr -p gptboot -i 1 adaX
> is executed for each disk involved in the mirror, right ..?

That's correct.

Want to link to this message? Use this URL: <>