I am patiently waiting for 14.04 to release and stabilize (June?) and I will then likely upgrade my server. I've decided to add a new level of automation to it by having it automatically update itself at regular intervals along with the ability to rollback if an update fails. Since I use btrfs, snapshot-ing is easy and takes little space.
Here's the gist of my idea:
Daily (weekly?) cronjob will snapshot, then update and dist-upgrade.
A kernel count will be maintained and auto-remove all but the current and two previous kernels.
Snapshots will be limited to the two most recent also.
All the above is rather easy. I'm wondering if I should have the system reboot itself whenever a new kernel is installed or if it should just email a notification requesting a reboot. I'm leaning toward the latter so I don't leave myself without a server when I'm away.
Topics for discussion are:
I'm interested to hear what I'm missing...
Here's the gist of my idea:
Daily (weekly?) cronjob will snapshot, then update and dist-upgrade.
A kernel count will be maintained and auto-remove all but the current and two previous kernels.
Snapshots will be limited to the two most recent also.
All the above is rather easy. I'm wondering if I should have the system reboot itself whenever a new kernel is installed or if it should just email a notification requesting a reboot. I'm leaning toward the latter so I don't leave myself without a server when I'm away.
Topics for discussion are:
What are the possible pitfalls?
How often is often enough or too often to update and reboot?
Can or should recovery be automated also? In other words, does a failed boot result in auto-reboot to previous state?
Should some packages be excluded from auto-update? For example: a failed grub will leave the system unbootable and unrecoverable without manual intervention, so should grub be excluded (grub package update, not update-grub)?
How often is often enough or too often to update and reboot?
Can or should recovery be automated also? In other words, does a failed boot result in auto-reboot to previous state?
Should some packages be excluded from auto-update? For example: a failed grub will leave the system unbootable and unrecoverable without manual intervention, so should grub be excluded (grub package update, not update-grub)?
I'm interested to hear what I'm missing...
Comment