[PATCH] libvirt-guests.sh: fix log output with new systemd

[Daniel P. Berrangé]

On Thu, Apr 30, 2020 at 02:42:21PM +0200, Christian Ehrhardt wrote:
> On Thu, Apr 30, 2020 at 2:10 PM Michal Privoznik <mprivozn at redhat.com> wrote:
> >
> > On 4/30/20 11:59 AM, Christian Ehrhardt wrote:
> > > Newer systemd is too smart, it detects the PIDs of the gettext calls
> > > and due to that a log lists libvirt-guests.sh with many different PIDs.
> > >
> > > Furthermore it struggles to collect the output in time, so it can be
> > > truncated and overall looks like:
> > >     libvirt-guests.sh[37986]: Running guests on default URI:
> > >     libvirt-guests.sh[37995]: Running guests on
> > >     libvirt-guests.sh[37977]: R
> > >
> > > Gather the gettext result into a local variable and printing
> > > the value from libvirt-guests.sh itself fixes both issues.
> > >
> > > Fixes: https://bugs.launchpad.net/ubuntu/+source/libvirt/+bug/1875708
> > >
> > > Signed-off-by: Christian Ehrhardt <christian.ehrhardt at canonical.com>
> > > ---
> > >   tools/libvirt-guests.sh.in | 74 +++++++++++++++++++++-----------------
> > >   1 file changed, 42 insertions(+), 32 deletions(-)
> > >
> > > diff --git a/tools/libvirt-guests.sh.in b/tools/libvirt-guests.sh.in
> > > index 7af24dab3b..324abe3623 100644
> > > --- a/tools/libvirt-guests.sh.in
> > > +++ b/tools/libvirt-guests.sh.in
> > > @@ -28,6 +28,16 @@ test ! -r "$sysconfdir"/rc.d/init.d/functions ||
> > >   # Make sure this file is recognized as having translations: _("dummy")
> > >   . "@bindir@"/gettext.sh
> > >
> > > +# Avoid output being listed under gettext PID and being truncated
> > > +unbuffered_gettext () {
> > > +    msg="$(gettext "$1")"
> > > +    echo "$msg"
> > > +}
> > > +unbuffered_eval_gettext () {
> > > +    msg="$(eval_gettext "$1")"
> > > +    echo "$msg"
> > > +}
> > > +
> >
> > Ah, is this the case where gettext prints the text in many sequences
> > whereas echo does it all at once? Or what do you think is happening
> > here? I don't quite understand how buffering steps into this.
> 
> You have two effects:
> - systemd tries to detect and report which PID the output is from when
>   something it printed. Therefore it appears to a user as if the script
>   itself would have been invoked multiple times while actually is will be
>   the PIDs of the different gettext calls
> - I assume that to get this content it is hitchhiking the output streams
>   somehow - the gettext processes are very very short lived and it seems
>   when they are exiting some output of them might be lost.

This doesn't pass a sniff test to me.

systemd is reading stdout/stderr of the libvirt-guests script. How
would it ever tell that certain lines from stdout/err are coming from
different PIDs. Systemd isn't involved in spawning the gettext processes.
It can see they come & go via cgroups pids, but that's about it.

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 :|