Changes between Version 5 and Version 6 of PXEInstall

Timestamp:

02/23/12 13:29:39 (13 years ago)

Author:

jdreed

Comment:

--

PXEInstall

@@ -2 +2 @@
 
 This page covers PXE installation.  The actual [wiki:Installation Debathena Installer] is covered in a [wiki:Installation separate page].
+
+== tl;dr ==
+
+PXE serves up an initial installation script that asks the user some questions, kexec's into a new network install using the answers to those questions, and then runs install-debathena.sh after the base system has been installed.
 
 == Infrastructure ==
@@ -64 +68 @@
  * If you want to install to something other than the first IDE/SATA disk, you lose.
  * The stage1 installer.sh script must never exit (it kexec's out, or loops forever with a shell).  If it does exit, the install will fall through to whatever the PXE server initially gave us, which is undoubtedly wrong.
+
+== The Stage 2 Install ==
+
+The stage 2 install is designed to be invoked from 2 places: The stage 1 install, and the hopefully-obsolete-by-the-time-you-read-this auto-upgrade code.  
+
+First, we do some hardware-specific hacks (I'm looking at you, Realtek).
+
+Then we parse the kernel's command-line, looking for options we care about.  In particular, we support:
+ * `debathena/clusteraddr`: DEPRECATED, an IP address, which also serves as a flag indicating that this is a cluster machine. (Note the different prefix, this is not a typo).
+ * `da/pxetype`: cluster, workstation, vanilla, etc.  We fallback to `debathena/pxetype` but will stop doing that.
+ * `da/i`: beta or production
+ * `da/m`: mirror site
+ * `da/part`: auto or manual partitioning
+ 
+For the fallback values (clusteraddr or debathena/pxetype), we set some defaults of installertype=production and mirrorsite=mirrors.mit.edu.  Eventually, we should stop doing this, and just error out if the necessary arguments aren't there.  We already barf if pxetype isn't present.
+
+We re-run net configuration in case we get a debathena/clusteraddr, but we do everything automatically.  We will stop doing this eventually.
+
+We now build-up a pre-seed file by copying the netcfg arguments, and setting the mirror hostname.  We also copy in the contents of `preseed.common`.  We add `preseed.autoinstall` (possibly skipping the partman section) for cluster machines.  We then fetch the actual [wiki:Installation Debathena Installer] (the beta version, if necessary), set pxe-install-flag, and exit out into the main installer, which will use these preseed values.
+
+=== Other files in the stage2 installer ===
+
+ * athena/{netparams,masks} -- these will go away when we stop re-doing netcfg in stage 2
+ * check-disks.sh -- Run by the partman component.  Find the first IDE/SATA disk, and ensure that the disk itself is big enough for the install.  This is defined as "20GB" by a variable at the top of this script.
+ * lvm-cluster-machine.partman -- a partman "recipe" for the layout of a typical cluster machine.  See /usr/share/doc/debian-installer/devel/partman-auto-recipe.txt.gz
+ * postinstall.sh - Run for debathena installation to do some sanity checking and run the debathena installer.
+ * preseed.autoinstall - For things that should be set for cluster machines
+ * preseed.common - Generally useful preseed values
+
+