On Tuesday 13 Jan 2004 2:41 pm, Wade Chandler wrote: > Keith G. Robertson-Turner wrote: > > On Mon, 12 Jan 2004 23:44:16 +0000, Rui Miguel Seabra wrote: > >>On Mon, 2004-01-12 at 23:36, Don wrote: > >>>With MS Windows, it seems a reboot is required after nearly every > >>>software update. > >> > >>In principle, the only update for which a reboot is needed is Linux > >> (since it is the most common kernel of the GNU system), but even this > >> may change in the future. > > > > That would be good, it would certainly massively cut downtime on servers. > > > > The only way I can think of implementing this would be to perform a kind > > of quick suspend/resume, where the "resume" remaps to a new running > > kernel, but surely all services and current tasks would need to be > > restarted too. > > > > - > > K. > > Remember, if you are updating packages for programs which already have > the .so or other file loaded, the only way to get them to start using > that new code is to restart them. So, if you do a full system update, > it may be faster to reboot, switch run levels and back again, or create > scripts to restart the pieces you update often. I usually just reboot > after an update. It saves me the headache of remembering. Unless you > are using an encrypted file system or some other type of password > protected startup you could automate this. Though most server updates > aren't a good idea to automate. You might break functionality your > server applications use by not reading change logs and readmes. > > Wade All wise words, experience has also taught me a reboot prior to a large system update is a prudent policy, especially if the server has been up for a few months, just to ensure no gremlins have crept in. Then at least you know the system was good prior to the changes. -- Alan D
