Latest revision as of 19:10, 21 April 2012

Although it is always possible to resurrect a machine back to life by reinstalling it, it is not always suitable to reinstall from scratch. Sometimes the best approach is to try to repair, or "resurrect," a broken system so that is it fully functional again. This document will show you how to resurrect a broken Funtoo system without reinstalling everything from scratch.

[edit] Building binary packages

The best approach for critical system repair is to boot on a SystemRescueCD or other similar LiveCD, and use this as working platform to resurrect your system.

For creating binary packages, you can use any of the following for a source environment:

• Use the most recent Funtoo stage3 for your architecture
• If available, use a recent system backup (tar or cpio archive) or snapshot

On a working Linux system, which can be either your broken system booted with a LiveCD, or any type of Linux system on your network with similar processor, you will want to build a chroot environment using the "source environment" you selected above, and use this as a platform for building binary packages to restore your system. Once these packages are created, they can be copied to your broken system and installed using the steps later in this document.

The binary package creation environment would typically be set up as follows:

# install -d /mnt/rescue
# tar xpvf backup.tar.bz2 -C /mnt/rescue
# cp /etc/resolv.conf /mnt/rescue/etc
# mount --bind /proc /mnt/rescue/proc
# mount --bind /sys /mnt/rescue/sys
# mount --bind /dev /mnt/rescue/dev
# mount --bind /dev/pts /mnt/rescue/dev/pts
# chroot /mnt/rescue
# source /etc/profile
# env-update

No matter of the way you jump in a functional Funtoo/Gentoo environment, the magic command to create a binary package archive once inside is to use the quickpkg command. quickpkg will capture the package in the exact form it is actually deployed on the environment and create an archive of it placed in /usr/portage/packages/<package-category>/<package-name>-<package-version>.tbz2.

In the following example capture everything installed within the "source environment" that is related sys-devel/gcc (4.4.5 is present on the system) is captured in a single archive named gcc-4.4.5.tbz2 located in /usr/portage/packages/sys-devel:

# quickpkg sys-devel/gcc

If you need to recompile a package instead of archiving an already deployed version (and of course without installing it on your "source environment"), just do:

# emerge --buildpkgonly sys-devel/gcc

[edit] Restoring the binary packages on the broken system

There are a couple of methods that can be used to restore binary packages to a broken system.

[edit] Chroot/Emerge Method

This first approach can be used for lightly damaged systems that still have a functional Portage and to which you can still chroot and perform all basic Linux commands. To use this method, you would mount your broken system to /mnt/broken using steps similar to the way we set up /mnt/rescue, above.

Before or after chrooting, copy the binary packages created in the step above in the exact same location on your broken system (e.g. in /usr/portage/packages/sys-devel in the case of sys-devel/gcc).

Once chrooted inside your system, you will be able to merge your packages using emerge as follows:

# emerge -k  sys-devel/gcc

[edit] Tbz2 Extract Method

This alternate method for installing binary packages is simpler and does not require the use of the chroot command. To prepare for using this approach, you just need to mount all key filesystems to /mnt/broken as follows -- and bind mounts are not necessary:

# install -d /mnt/broken
# mount /dev/sdaX /mnt/broken
# mount /dev/sdaY /mnt/broken/usr

Then, use the following commands to extract the .tbz2 archive to your broken filesystem that you mounted at /mnt/broken:

# tar xjpvf gcc-4.4.5.tbz2 -C /mnt/broken

You will see a note about the trailing garbage at the end of the file being ignored. This is normal -- tar is ignoring the Portage .tbz2 metadata that is tacked on to the end of the file.

At this point, you can set up bind mounts (see /mnt/rescue example earlier in this document for those steps), chroot inside /mnt/broken, and perform a few tests to determine if your issue has been resolved.

[edit] Fixing broken portage

sometimes sys-apps/portage may fail and the above binpkg mentioned way to ressurect system will not work because of portage being broken itself. To fix portage manually the following steps required:

# cd /tmp
tmp # wget https://www.github.com/funtoo/portage-funtoo/tarball/funtoo-2.3.3-r4
tmp # tar xf funtoo-2.3.3-r4
tmp # cd funtoo-portage-funtoo-84fe4a7
funtoo-portage-funtoo-84fe4a7 # rm -rf /usr/lib/portage/*
funtoo-portage-funtoo-84fe4a7 # cp -a bin pym /usr/lib/portage/
funtoo-portage-funtoo-84fe4a7 # ln -s /usr/lib/portage/bin/emerge /usr/bin/emerge

This will back a working portage again. To ensure everything is consistent re-emerge portage itself immediately.

# emerge portage 

File collision warning, if any, can be skipped at this point.

[edit] Critical System Packages

Several packages at the heart of your Funtoo system, these mainly are:

[edit] Various weird issues

P1: I have a local distfiles mirror and wget complains about not being able to resolve 'localhost'.
S1: libnss (which handles name resolution) is probably damaged or suffers of inconsistencies, in your /etc/make.conf, change 'localhost' in GENTOO_MIRRORS for 127.0.0.1 (IPv4) or ::1 (IPv6)

P2: Same problem as described in P1 but I use another machine on my network or a public mirror on the Internet
S2: See the solution given in S1 but with providing the IP of the machine you are downloading for.