Commit 5330598d authored by Eric Eide's avatar Eric Eide

Moved from `doc'.

This is a copy of `doc/updating-RHL.txt' r1.2, modified on Nov 12 2003.
parent 3c8cb167
When transitioning the testbed to a new Red Hat Linux release, there are
a few stumbling blocks, and notes to be aware of:
* Always use the "buddy" system: start with a combined FreeBSD+Linux
testbed image, then wipe out, and install the new distribution to the
Linux slice.
* The Linux install often corrupts the FreeBSD bootlabel, and replaces
the MBR. You might end up having to imaging the Linux slice, and
laying it back down on a reloaded combined scratch image.
* There are many superflous RPMs installed by default, esp. ones for
multimedia, and Xwindows (window managers, gnome, KDE, etc.). Also,
don't forget to install some of the more useful, non-default
packages: tcpdump, ttcp, gated, etc.
* The "typical" Red Hat "Workstation" installation selection does not
install any sort of means to access the node remotely! Ensure that
you at least add openssh-server to the list of packages, or
post-install, before rebooting. Alternatively, FreeBSD has
excellent support for ext2fs, including an fsck.ext2fs that can be
installed from the ports tree. This trick can be used to edit a
broken Linux slice.
* Red Hat CAN be installed via the serial port. The boot floppy
image's LILO works with BIOS redirection. You _must_ choose text
installation, and must pass: console="ttyS0,115200" as kernel
parameters from the lilo prompt.
* Patching the kernel for IPOD is volatile. Minor kernel revisions
will likely break the patch, but hand updating the file(s) is
usually trivial since IPOD is a separate/new ICMP type, with new
handling functions. Beware of the sysctl enums, and virtual
function tables - those are were you are likely to run into trouble.
* The tmcd/startup script client side install targets may break the
image! Be especially wary of the modules.conf file. This may whack
the proper ethernet driver autoloading depending on the logic of the
new distro's startup scripts. Don't forget to update the tag, and
release specifications in the supfile.
* You should download updated RPMS from a redhat mirror, and update
all the installed packages to plug security holes and bugs. It
appears that redhat tries to keep versions/behavior consistent in
its distributions, usually providing updates only to fix bugs and
security holes.
* LILO is a pain. It doesn't redirect to the console correctly. We
are currently able to get around this by using an internal
hacked/bugfixed version that supports our standard 115200 baud rate.
Newer versions of lilo are significantly different, and appear to be
broken as well. Work will need to be done on these versions before
they will be useful in our environment.
* Grab a precompiled CVSup binary, either from the old RHL image, or
from the CVSup website. _Much_ less hassle.
* Newer (circa 11/2003) Emulab images require that elvind be installed.
Building libelvin and elvind for Linux is not really a problem, the
problem is getting the correct sources to build from. We use Elvin
4.0.3 which may no longer be available. Once libelvin and elvind
are installed, the tmcd/linux make will take care of hooking it in
to the startup.
* Elvin libraries are installed in /usr/local/lib. Make sure the default
shared library path (/etc/ld.so.conf) includes this. If necessary, add
it and run ldconfig.
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment