Date: Tue, 19 Oct 2010 14:03:11 -0700 From: Freddie Cash <fjwcash@gmail.com> To: freebsd-fs@freebsd.org Subject: Re: ZFS 'read-only' device / pool scan / import? Message-ID: <AANLkTi=DikO4-N8BG4U0WBX-7ypbPkVCR8=vHSaeN3qV@mail.gmail.com> In-Reply-To: <4CBDFFF6.5080701@digiware.nl> References: <AE519076FDEA1259C5DEA689@HexaDeca64.dmpriest.net.uk> <20101019151602.GA61733@icarus.home.lan> <7BEF90D9F4D4CB985F3573C3@HexaDeca64.dmpriest.net.uk> <4CBDFFF6.5080701@digiware.nl>
next in thread | previous in thread | raw e-mail | index | archive | help
On Tue, Oct 19, 2010 at 1:30 PM, Willem Jan Withagen <wjw@digiware.nl> wrot= e: > On 2010-10-19 17:30, Karl Pielorz wrote: > >> As there is such a large aspect of human error (and controller >> behaviour), I don't think it's worth digging into any deeper. It's the >> first pool we've ever "lost" under ZFS, and like I said a combination of >> the controller collapsing devices, and humans replacing wrong disks, >> 'twas doomed to fail from the start. >> >> We've replaced failed drives on this system before - but never rebooted >> after a failure, before a replacement - and never replaced the wrong >> drive :) >> >> Definitely a good advert for backups though :) > > I'm running my ZFS stuff on a 3ware and an areca controller, and they onc= e > in a while forgot their order of disks during booting. > (the 3ware got fixed by a bios upgrade) > The areca just keeps reordering no matter how hard you like to tell it > otherwise. > > But GPT really proves useful since reallocation of disks does not result = in > a different device in the gpt directory. > > eg.: > =C2=A0pool: zroot > =C2=A0state: ONLINE > =C2=A0scrub: none requested > config: > > =C2=A0 =C2=A0 =C2=A0 =C2=A0NAME =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 STATE = =C2=A0 =C2=A0 READ WRITE CKSUM > =C2=A0 =C2=A0 =C2=A0 =C2=A0zroot =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0ONLINE= =C2=A0 =C2=A0 =C2=A0 0 =C2=A0 =C2=A0 0 =C2=A0 =C2=A0 0 > =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0mirror =C2=A0 =C2=A0 =C2=A0 ONLINE =C2= =A0 =C2=A0 =C2=A0 0 =C2=A0 =C2=A0 0 =C2=A0 =C2=A0 0 > =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0gpt/root4 =C2=A0ONLINE =C2=A0 = =C2=A0 =C2=A0 0 =C2=A0 =C2=A0 0 =C2=A0 =C2=A0 0 > =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0 =C2=A0gpt/root6 =C2=A0ONLINE =C2=A0 = =C2=A0 =C2=A0 0 =C2=A0 =C2=A0 0 =C2=A0 =C2=A0 0 > > I could even migrate a disk from the 3ware controller to the std SATA > interfaces without losing the the gpt-label. While using labelled devices (be it glabel, gpt label, or whatever) certainly helps keep things ordered and working. However, ZFS also labels the devices in the pool. A simple "zpool export poolname" followed by a "zpool import poolname" will scan the metadata on the drives, find all the devices in the pool, re-order things internally, and carry on. Back when I started with ZFS, I used unlabelled drives (also on 3Ware controllers) and made the mistake once of booting with a failed drive removed. Pool came up faulted saying all the drives after the missing one were also faulted. Thought I lost the whole pool. However, some digging online showed the export/import info, and I was able to continue on without losing any data. --=20 Freddie Cash fjwcash@gmail.com
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?AANLkTi=DikO4-N8BG4U0WBX-7ypbPkVCR8=vHSaeN3qV>