We seem to have a hard time replacing our old physical machines with new physical machines, despite the fact that we have tons and tons of newer, faster physical machines.
But wait, don't we do this fancy "virtualization" thing that's supposed to make things like that easier?
I propose, as a stopgap measure, since the transition to Etch is taking much longer than anticipated, that we proactively attempt to move Sarge domUs to Dom0s running Etch.
Caveats
I forget the exact reason, but we can't run the sarge domains under the old kernel. So, we need to use the new kernel (2.6.24-1-xen-686).
This has the following consequences:
- Need to add etch-cclub to /etc/apt/sources.list, for the kernel and openafs modules.
- Need /etc/apt/preferences to ensure packages come from the right places
- There seems to be a problem with /dev/pts (one solution is to install udev)
- Need to edit /etc/inittab; run a getty on hvc0, rather than tty1
Apt Preferences
Package: * Pin: release v=4* Pin-Priority: 50 Package: * Pin: release a=etch-cclub Pin-Priority: 50 Package: linux-modules-*-xen-* Pin: release v=4* Pin-Priority: 999 Package: openafs-modules-* Pin: release v=4* Pin-Priority: 999
Packages to Install
- linux-modules-2.6.24-1-xen-686
- openafs-modules-2.6.24-1-xen-686 (if DomU runs AFS)
- udev (brute force way to fix the /dev/pts issue)