It happens many times that a server run for month or years and then the system administrator has to reboot it and it doesn’t come up again because the power supply(or other hardware) is damaged. At this point someone says with the voice of a teacher: “Never change a running system”. It’s the same with software-updates. The system runs many years without any updates, and then we have to upgrade it and nothing works. And again, someone from behind says with the voice of a teacher: “Never change a running system”. Those are the days where my left eye starts to twitch and I tend to get crazy…
If we presume that those things can happen, we should handle them. For example: During the day our server is busy but in the night nobody works on it and we can just reboot it(or power off). If it doesn’t come up, we can fix it at this time, when nobody else needs this server.
If we do our software-updates regulary, the changes aren’t too big and therefore the problems after updates aren’t too big too. It’s generally wisely to upgrade software as often and as fast as possible because of security-fixes. Of course there might be other problems with new features and changes in release-upgrades, but at some time we also have to do it. And we shouldn’t wait for too long. If we have to upgrade through many major-releases, things might get much more difficult and complex.
In Conclusion I would say, the strategy of my choice is: “Change a running system controlled and regulary to prevent bigger troubles”
