Date: Tue, 20 Sep 2005 11:13:59 -0700 From: Drew Tomlinson <drew@mykitchentable.net> To: Jerry McAllister <jerrymc@clunix.cl.msu.edu> Cc: FreeBSD Questions <freebsd-questions@freebsd.org> Subject: Re: Can I Rebuild / and /usr Remotely? Ideas? Message-ID: <43305167.9050603@mykitchentable.net> In-Reply-To: <200509201720.j8KHKXjY026494@clunix.cl.msu.edu> References: <200509201720.j8KHKXjY026494@clunix.cl.msu.edu>
next in thread | previous in thread | raw e-mail | index | archive | help
On 9/20/2005 10:20 AM Jerry McAllister wrote: >>I have a system running 5.4-STABLE. I created a geom_stripe disk for >>/usr. It works fine except that upon reboot, the stripe attempts to >>load itself twice and thus fails. Therefore, since I have no /usr, the >>system comes up in single user mode. At that point I can do 'kldunload >>geom_stripe' and then 'kldload geom_stripe' to get the stripe built and >>finish booting my system. I sent a problem report describing this >>behavior. http://www.freebsd.org/cgi/query-pr.cgi?pr=83521 Please see >>it for details. >> >>Based upon the information in the pr, the commiter suggests clearing all >>metadata on the stripe providers and starting over. >> >>When I built this system, I configured the disks using sysinstall. I >>used the "dangerously dedicated"mode just as I had when I ran the 4.x >>series. I suspect my problems occur because geom_stripe doesn't get >>along well with disks that are "dangerously dedicated". >> >>Anyway, my system has 2 9gb drives (da0 and da1) that I wish to use for >>the main system. I want a 500mb slice as /dev/da0s1a for '/', a 500mb >>slice as /dev/da1s1b for swap, and the rest of each drive as >>/dev/daXs1d. I will build my stripe with /dev/da0s1d and /dev/da1s1d >>and mount it at /usr. Other directories such as /var, /home, etc. will >>be symlinked to /usr/var, /usr/home, etc. >> >> > >I would be inclined to want some swap on da0 - the boot drive - too, but >I guess you don't have too. > > What would the advantage be? My thinking is to put root on one and the same size swap on the other leaving two equal sized partitions (or are they slices? I'll never get that straight) with which to build my stripe set. >>I have remote console access to this machine and want to attempt to >>rebuild the system remotely. If I mess up, it's not too difficult to >>physically get to the box but I would like to avoid it if possible. I >>have another disk on the system (ad0) that is available and large enough >>to hold the contents of both da0 and da1. Can I backup my system, do >>the needed operations on da0 and da1, restore da0 and da1, reboot, and >>still have a working system? I've never used fdisk, bsdlabel, newfs, >>and whatever else I might need from the command line. Besides the man >>pages, are there any guides for what I want to do? Even a simple "first >>this, then this, then this" type of guide will help me get started. >> >> > >As long as there is room on the ad0 drive for all of the dumps there >should be no problem. There will be a dump file for each current >file system. You may also need to have some space to unroll a dump >if the way you are breaking up the file system in to parts with links >is different than the way it is now. for example if you take /usr/local >out our /usr and put it in its own space, you will first need to restore >all of /usr somewhere (maybe in its new space if there is room or on >the ad0 drive if there is not) and then transfer the separate parts to >their new homes - probably using 'tar -cpf'. > > I'm happy with the layout now but needing intervention on each reboot is not acceptable to me. This is the only reason I'm considering this exercise. >The main thing is to think out the pieces - what each file system is and >what order you will need to restore things so the each new file system >is created and ready and the mount points are there when you need them. > > Thank you for your reply. This gives me some direction in which to proceed. Since the layout is as I like now, it seems to me that I would only have to dump / and then dump /usr. The restore order would be first / then /usr. Is that correct? Here's my current fstab: blacklamb# cat /etc/fstab # Device Mountpoint FStype Options Dump Pass# /dev/da1s1b none swap sw 0 0 /dev/da0s1a / ufs rw 1 1 /dev/stripe/data /usr ufs rw 2 2 /dev/acd0 /cdrom cd9660 ro,noauto 0 0 Thanks, Drew -- Visit The Alchemist's Warehouse Magic Tricks, DVDs, Videos, Books, & More! http://www.alchemistswarehouse.com
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?43305167.9050603>