On Fri, Mar 25, 2011 at 10:28 AM, Milos Negovanovic < milos.negovanovic@xxxxxxxxx> wrote: > On Fri, Mar 25, 2011 at 05:24:37PM +0100, Cédric Girard wrote: > > No. But what I understood from what Thomas said is: as you need to reboot > > your server anyway from time to time to apply security updates, you may > > decide to switch to an even more often updated kernel, if your > architecture > > permit it (reboot != service interruption). > > For most people running arch on their servers reboot == service > interruption > > Am I wrong here? > > -- > Milos Negovanovic > milos.negovanovic@xxxxxxxxx > Arch on a server is a tricky thing, of course creating a viable long term infrastructure architecture is also a tricky thing, the problem is that MANY sys admins adhere to the idea from the 80s, that you set up a server once and then leave it be for years on end. This is no longer possible, and this idea about system management is not only outdated, but very dangerous. So a viable infrastructure requires %100 failover capacity for all services, so that reboots are routine, and never dangerous, in my deployments we reboot systems VERY frequently, often not even into new kernels, but just to continually verify the failover viability. As for a home server, I generally set up a cron job on those to do a pacman -Syu every night, detect if the kernel has changed, and then reboot automatically. No problems with it so for (This is also how my main company build servers work)
