[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Qemu-devel] [libvirt] [PATCH v6 2/2] vl: fix use of --daemonize wit

From: Daniel P . Berrangé
Subject: Re: [Qemu-devel] [libvirt] [PATCH v6 2/2] vl: fix use of --daemonize with --preconfig
Date: Tue, 12 Jun 2018 14:17:08 +0100
User-agent: Mutt/1.9.5 (2018-04-13)

On Tue, Jun 12, 2018 at 03:04:42PM +0200, Michal Privoznik wrote:
> On 06/12/2018 02:42 PM, Igor Mammedov wrote:
> >>>
> >>> Do we really need to make -daemonize and -preconfig work
> >>> together?  libvirt uses -daemonize only for its initial
> >>> capability probing, which shouldn't require -preconfig at all.
> >>>
> >>> Even on that case, I wonder why libvirt doesn't simply create a
> >>> server socket and waits for QEMU to connect instead of using
> >>> -daemonize as a sync point.
> >>>   
> >>
> >> because libvirt views qemu as well behaved daemon. Should anything go
> >> wrong libvirt reads qemu's stderr and reports error to upper layers.
> > We can keep daemonizing flow in QEMU as it's now.
> > But Eduardo's idea about libvirt created socked + letting QEMU connect to it
> > has a merit. It should fix current deadlock issue with as monitor
> > won't be depending on lead exit event.
> Not sure about the benefits. Currently, libvirt spawns qemu, waits for
> monitor to show up (currently, the timeout dynamic depending on some
> black magic involving guest RAM size) and if it does not show up in time
> it kills qemu. The same algorithm must be kept in place even for case
> when libvirt would pass pre-opened socket to qemu in case qemu deadlocks
> before being able to communicate over qmp. The only advantage I see is
> libvirt would not need to label the socket (set uid:gid, selinux, ...).

As mentioned in my other reply, we already do FD passing, and that code
has intentionally got rid of the timeout, because timeouts cause false
failures to launch QEMU. This is a particular problem when using many
disks that are encrypted, since LUKS encryption has a minimum 1 second
delay on opening each disk, so with many disks we're at risk of hitting
the timeout even when QEMU is still starting normally.

I don't see a reason to special case startup with timeouts to deal
with hangs, while ignoring the possibility of hangs after initial

> On the other hand, since it would be libvirt creating the socket what
> would happen on libvirtd restart?

We're creating a *listener* socket, not a client connection, so on
restart we simply connect again as normal.

|: 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 :|

reply via email to

[Prev in Thread] Current Thread [Next in Thread]