On Wed, Jul 31, 2019 at 3:03 AM Vincent Blut <vincent.deb...@free.fr> wrote: > > On Tue, Jul 30, 2019 at 05:35:52PM -0700, Ross Boylan wrote: > >[Ross] > >> >I've run into other problems with services starting before all > >> >filesystems were mounted; I wonder if that's an issue here (not on the > >> >machine right now). > >> >i.e., /usr isn't mounted when timesync first checks for chrony, and so > >> >it thinks things are OK. > >> > >[Vincent] > >> I don’t think it’s feasible as the -.mount unit is unconditionally > >> active. As for separate /usr partition, that’s the role of the initramfs > >> to mount it. > > > >Unfortunately, this is one area I can speak from authority: it is > >absolutely possible for services to start before all critical mounts > >have happened. Bug#933139 has gory details. Among other issues, bind > >attempted to start before /var was mounted. > > Sure Ross, I do not dispute that. My comment referred only to /usr. > I'm not sure I understand "I don't think it's feasible as the -.mount unit is unconditionally active...." I thought "it's feasible" referred to the idea that /usr might not be mounted, and you were saying it would have to be mounted. Since /var doesn't have to be mounted, my assumption is that /usr doesn't have to be mounted for the same reasons, if it's a separate logical volume. I can now confirm it is separate on my system.
As I said, even if all my suppositions are true they may have nothing to do with current problem, though they would explain why timesyncd might be able to start. Ross > >As for how or if systemd and initramfs are integrated, I don't know. > >I am using an initrd, and it didn't prevent the problem just > >mentioned.
