On Mon, Jul 29, 2019 at 01:50:22PM +0100, Daniel P. Berrangé wrote:
On Mon, Jul 29, 2019 at 01:30:42PM +0100, Daniel P. Berrangé wrote:
On Sun, Jul 28, 2019 at 04:42:52PM +0200, Andrea Bolognani wrote:
On Tue, 2019-07-23 at 17:02 +0100, Daniel P. Berrangé wrote: [...]
- We can make virtproxyd and the virtXXXd per-driver daemons all have "Conflicts: libvirtd.service" in their systemd unit files. This will guarantee that libvirtd is never started at the same time, as this would result in two daemons running the same driver. Fortunately drivers use locking to protect themselves, but it is better to avoid starting a daemon we know will conflict.
I feel like this will need to be tested extensively to make sure we're always doing the right thing, including on non-systemd hosts.
Testing is quite easy - just try to start the two units and make sure only one ends up running. Similarly for non-systemd hosts, start both daemons & see that only one succeeds - the others fail with lock conflict.
+++ b/src/remote/virtproxyd.service.in @@ -0,0 +1,24 @@ +[Unit] +Description=Virtualization daemon +Conflicts=libvirtd.service +Requires=virtproxyd.socket +Requires=virtproxyd-ro.socket +Requires=virtproxyd-admin.socket +After=network.target +After=dbus.service +After=apparmor.service +After=local-fs.target +After=remote-fs.target +Documentation=man:libvirtd(8) +Documentation=https://libvirt.org
There are a few non-obvious changes between libvirtd.service.in and this file:
-Requires=virtlogd.socket -Requires=virtlockd.socket -Wants=systemd-machined.service -Before=libvirt-guests.service -After=iscsid.service -After=systemd-logind.service -After=systemd-machined.service
I can see why we'd move the relationships with iscsid and virtlockd to virtstoraged, except looking ahead to patch 23 I see you haven't actually done that; either way, I'm not so convinced about the remaining changes. Care to explain the rationale behind them?
virtproxdy contains no drivers, so it doesn't need to depend on any of these services.
virtdstoraged/qemud/lxcd should have gained some of these though.
I should have killed dbus.service and remote-fs.service too.
Doh, not. dbus must always be present for polkit to work. remote-fs.service is ok as that's only needed for the hypervisor services (to access disk images). Regards, Daniel -- |: https://berrange.com -o- https://www.flickr.com/photos/dberrange :| |: https://libvirt.org -o- https://fstop138.berrange.com :| |: https://entangle-photo.org -o- https://www.instagram.com/dberrange :|