Rudi Ahlers wrote on Thu, 18 Feb 2010 10:54:11 +0200: > The server booted up, ran fsck, then each VM, as it booted up ran fsck as > well - which just slowed down the whole process since there's a 5 minute > delay in starting each VM. Why would you autostart a VM only every 5 minutes? Or did you mean the next one only started once the earlier one had finished fsck? As mine aren't doing this I haven't ever seen that. But by the time most users could reconnect, 2+ > hours have lapsed. this particular server wasn't rebooted in 274 days, Which means it ran without many kernel security updates for a long time. > But, how does one get past this? I know we need to reboot from time to time, > but more than often it's (preferably) not sooner than 6 - 10 months, so fsck > will run. You use tune2fs on the VM filesystems as Mogens explains. Then you don't have any extra downtime for the VMs. When I reboot servers it takes about ten extra ping losses for the dom0 fsck if it is time. For the VMs you then run it twice a year manually (or by script) from dom0. Take the VM down, mount the LVM partitions, run fsck, unmount, start again. Should take less than 5 minutes per VM. Kai -- Get your web at Conactive Internet Services: http://www.conactive.com