* Philippe Vogel wrote on Sat, Jun 29, 2002 at 10:34 +0200:
Never update packages and then reboot, if it's not needed.
I don't think so. Better risk the need for manual attention after a failed reboot on a scheduled, known point that after i.e. a power loss! Otherwise you risk that the servers will not boot up successfully after an power- or UPS failure. In that case, you'll not only have to work around the usual things, but additionally against the wrong configuration - well, and this maybe weeks after you changed something! I saw big troubles when booting up complete server centers in after such failures. It's important to know the boot order, it may happend that failures occure since hosts other servers depend on need just to much time to boot (DNS-Servers, for instance). In such a case you may stand in front of a system out of corder completly, just since a DNS server *was* not ready. This is very difficult to trace! Maybe it's a good practise to reboot after such changes. Maybe in the evening on a scheduled, communicated time. Noone should blame you if you reboot something in such cases, but if anything fails later, they will!
Then restart services by hand and only reboot, if kernel is patched. I handle things with caution, since I had several bad experiances.
This has disadvantages. We had a nice detail problem once upon a time. For some unknown reason, on one host the /etc/rc.config file was destroyed (some binary contents). Well, it was a big problem to restore it from backup, we had the last working on a history tape only. All the time the server was down! Another indident: rc.config has to START_WHATVER="". The first was "yes", and on first view nobody noted that at the end of file there was another entry with "no". Starting per Hand worked of course... Well, this took manual investigation at a time we had much other trouble... oki, Steffen -- Dieses Schreiben wurde maschinell erstellt, es trägt daher weder Unterschrift noch Siegel.