Date: Wed, 15 Jul 2009 22:42:04 +0100 (BST) From: Chris Hedley <freebsd-current@chrishedley.com> To: Peter Schuller <peter.schuller@infidyne.com> Cc: Randy Bush <randy@psg.com>, freebsd-current@freebsd.org, Freddie Cash <fjwcash@gmail.com> Subject: Re: ZFS pool corrupted on upgrade of -current (probably sata renaming) Message-ID: <alpine.BSF.2.00.0907152238120.12181@teapot.cbhnet> In-Reply-To: <20090715200342.GA89750@hyperion.scode.org> References: <alpine.BSF.2.00.0907132009040.2027@teapot.cbhnet> <alpine.BSF.2.00.0907142318520.2686@teapot.cbhnet> <b269bc570907141550u6854bc8eh6ea73fe9bd3e788a@mail.gmail.com> <m21voi1ufz.wl%randy@psg.com> <b269bc570907150922l3ded8a76id12ea72801abb3c7@mail.gmail.com> <20090715200342.GA89750@hyperion.scode.org>
next in thread | previous in thread | raw e-mail | index | archive | help
On Wed, 15 Jul 2009, Peter Schuller wrote: > While I can see this working "most of the time" - is there any reason > to believe it is guaranteed to? glabel keeps meta data at the end of > the device; is it guaranteed that ZFS is not using that part of the > device actively? (For example by having a policy to reserve some > amount at the end.) I think that would be my worry; I did wonder about trying the same with my own installation, but even if it were possible (or advisable for other reasons; I'm not sure it is in either case) I am worried that at some point something or other might want to use that last block with interesting consequences. I've just been having fun and games moving over to the gpart scheme and having to remember to include an extra block to keep the label information, since gmirror forbids me to add a labeled partition due to it being too small (in which case maybe ZFS doesn't use the full partition if it allows it to be re-added even though it's shrunk by a block)... but I'm going to rebuild my ZFS array from scratch. Chris.
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?alpine.BSF.2.00.0907152238120.12181>