gmirror failed with error 19.
fbsd-mas-0 at ml.turing-complete.org
Sun Oct 16 15:31:49 UTC 2011
* "Luchesar V. ILIEV" <luchesar.iliev at gmail.com> [2011-10-16 17:25 +0300]:
> On 16/10/2011 14:51, Lev Serebryakov wrote:
> > Hello, Nicolas.
> > You wrote 16 ?????????????? 2011 ??., 15:05:26:
> >> Of course you have to create the gpt inside the gmirror, not on the
> >> same device the gmirror is based on. The warnings occur, when the gpt
> >> is seen on the real disk. But later the gpt on the gmirror is used,
> >> and on the device created by gmirror, gpt uses the last sector of that
> >> device.
> >> I hope this will not change.
> > Problem is, that new BIOSes could refuse to boot from disk with such
> > misplaced (from raw disk's and BIOS' point of view) secondary GPT.
Such a BIOS is of course possible. But I think that FreeBSD should
continue to support such configurations.
> Probably I'm missing something, but why not do the opposite: configure
> any gmirrors on top of the GPT partitions? That's at least what I do for
> my swap volumes, although, admittedly, I'm also heavily using ZFS, which
> means I don't have many partitions (2 to 3 at most).
If you have to resync after a power failure or a crash, gmirror will
start all resyncs at once, which is considerably slower. Sp you have
to disable automatic resyncs and handle it differently. And, with
onnly one mirror per disc, you cannot get into a situation, where some
mirrors use one disk and other mirrors the other one.
More information about the freebsd-geom