A quick update can be done by
apt-get update apt-get upgrade
For general installation, updates, and so on
aptitude is a good choice.
Anything that shouldn't be updated? Like stuff with custom patches? Does apt have a history of clobbering configs or anything?
We run cron-apt to give us warnings via root's email when packages need to be updated. This should be installed on every box. Here's the change you need to make to the configuration:
--- /etc/cron-apt/config.orig 2009-02-25 05:41:42.000000000 +0000 +++ /etc/cron-apt/config 2009-02-25 05:44:11.000000000 +0000 @@ -94,7 +94,7 @@ # output (send mail when output is generated) # always (always send mail) # (else never send mail) -# MAILON="error" +MAILON="changes" # Value: error (syslog on error runs) # upgrade (when packages is upgraded)
apt-get update uname -a # Figure out which kernel series this machine needs apt-cache search kernel-image | grep somethingwhichidentifiestheseries # Find the latest kernel in the series apt-get install kernel-image-latestversion-series shutdown -r N
so for example:
uberjoe@davros:~$ uname -a Linux davros 2.6.8-2-sparc64 #1 Fri Jun 10 00:38:40 UTC 2005 sparc64 GNU/Linux uberjoe@davros:~$ apt-cache search kernel-image | grep sparc64 kernel-image-2.6.8-2-sparc64 - Linux kernel binary image for UltraSPARC (sparc64) systems kernel-image-2.6.8-3-sparc64 - Linux kernel binary image for UltraSPARC (sparc64) systems uberjoe@davros:~$ sudo apt-get install kernel-image-2.6.8-3-sparc64 boring debian package installation stuff uberjoe@davros:~$ sudo vim /etc/silo.conf uberjoe@davros:~$ sudo shutdown -r now
Silo should be set up to load the new kernel magically. If not, some haxing may be required to get the new kernel to work. It's probably best to try and set it up like so:
root=/dev/hda2 partition=1 default=debian read-only timeout=100 image=/boot/vmlinuz label=debian append="UDEV_DISABLED=yes" initrd=/boot/initrd.img image=/boot/vmlinuz label=Linux append="UDEV_DISABLED=yes" initrd=/boot/initrd.img image=/boot/vmlinuz.old label=LinuxOLD append="UDEV_DISABLED=yes" initrd=/boot/initrd.img.old
This is more or less how debian has it set up by default, but with devfs turned off, because we don't need it. You might want to check which drive root is on. Debian should update the symlinks in /boot automagically, and all should be well.
Anything tied to specific kernel versions?
The stock Debian stable kernels have a bug in the serial drivers that make the console on Ultra 5s unusable. We previously fixed this with a slightly later version of the package (2.6.8-16), but it looks like the security update (2.6.8-15sarge1) should have the console fix, along with a bunch of security fixes. I haven't tested this version yet.