►
From YouTube: KubeVirt Community Weekly Meeting - 2018-11-07
Description
No description was provided for this meeting.
If this is YOUR meeting, an easy way to fix this is to add a description to your video, wherever mtngs.io found it (probably YouTube).
B
B
Currently
it's
under
Fabian's
personal
account,
but
if
we're
going
to
be
bringing
this
under
the
Qbert
umbrella,
it
might
make
sense
to
do
so
and
so
I
set
that
email
I
actually
already
hit
the
button
to
fork
the
repo,
which
is
a
mistake,
because
that's
not
how
we
should
have
done
it
anyway,
so
we
will
be
cleaning
that
up.
If
you
have
noticed
that
the
repo
already
exists.
That's
my
fault,
we'll
fix
that
looking
for
thoughts
or
comments
on
that
in
terms
of
moving
it
over.
If
there's
any
objections
or
thoughts
on
that.
A
Oh
no
just
just
sign,
oh
so
now
it
has
been
disagreement
about
that
feature
and
I
acknowledge
that
although
I
think
the
problem
is,
we
have
now
by
now
I
think
around
50
repositories
and
Cubert,
all
of
them
to
a
different
degree
of
ripeness
and
maturity.
I
want
and-
and
that's
I'm
not
happy
about
that
state,
but
I,
don't
see
a
better
solution
for
that.
I
think
we
want
to
concentrate
stuff
which
is
about
Qbert
in
that
org
and
so
even
for
features
like
this,
like
vm
CTL
or
we
have
different
opinions
on.
A
I
think
it's
right
to
move
it
to
Qbert,
but
what
I
think
we
should
think
about
and
I'd
thought
about.
It
is
how
we,
how
we
differentiate
between
ripeness
of
repositories,
right,
we
have
like
incubating
projects
or
non
right
ones,
and
some
which
we
can
really
recommend
to
use
and
actually
I
wondered
if
we,
if
we,
if
we
choose
a
different
place
to
separate
them
by
location
rights,
of
something
like
Cuban
and
Cuban
incubator,
then
there
are
different
orgs
and
we
lose
visibility.
A
So
I
rather
thought
if,
if
a
first
step
could
be
just
to
add
certain
tags,
I
mean
Cuba.
Github
repositories
can
be
annotated
with
tags.
If
we're
using
text
to
differentiate
like
incubating
projects
from
something
else
just
to
have
some
baseline
informations
about
how
we
define
a
the
ripeness.
Otherwise,
it's
just
a
mass
of
things.
Maybe
there
are
opinions
well.
B
I,
do
actually
have
a
thought
on
that
when
you
look
at
the
queue
the
kubernetes
project,
from
which
we're
sort
of
you
know
a
related
project
to
that,
it
becomes
difficult,
especially
when
api's
are
moved
from
alpha
to
beta,
because
they
physically
move
modules
and
import
paths
break
whatever.
We
do
I'd
like
to
not
do
that.
You
know
if
we
have
some
sort
of
like
metadata
marking
on
a
repository,
that's
great,
but
in
terms
of
physically
moving
things
around
that
causes
a
little
bit
of
pain
for
consumers.
C
So
I
think
the
I
agree
with
with
Stu
and
notes
about
the
namespaces
and
stuff
like
that.
What
I
think
would
work
really
well
is
the
tags
idea
and
actually,
as
far
as
github,
that's
the
only
that's
the
only
solution,
I
know
of
as
far
as
how
to
do
that
unless
you
get
into
a
branching
method
or
something
like
that,
which
is
just
ugly,
so
I
think
your
proposal
on
using
tags
is
actually
a
great
idea.
D
I
might
have
a
different
one
right
now,
I'm,
looking
at
the
list
of
all
the
repositories
and
much
you
see
the
text
but
they're
not
differentiated
in
any
way.
So,
basically,
you
have
three
four
five
tags
per
repo,
but
you
can
actually
add
something
visible
to
the
beginning
of
the
description
to
the
single
line.
Description.
That's
actually
differentiate,
differentiates
it
from
the
tags
and
it
would
set
apart
and
it's
part
of
the
description.
So
you
should
always
see
that,
even
when
you
click
through
to
the
repple,
you.
B
D
E
A
I
would
take
both
because
the
good
thing
about
taxes-
you
can
really
query
them,
so
we
get
really
true.
That's
true,
so
I
think
both
might
make
sense
and
I
would
actually
propose
to
start
simple
and
just
say,
like
you
know,
incubating
and
all
the
others
just
don't
have
it
so
that
you
can
perform
in
creating
projects,
and
we
would
also
say
that
if
it
gets
an
incubating
tag,
then
the
description
needs
to
be
prefixed
with
incubating.
That
will
be.
My
programmatic
sounds.
C
A
A
Yeah,
we
don't
need
to
do
any
relocation.
A
website
update
all
that
nasty
stuff,
okay,
cool,
so
stup.
That
means
I
will
actually
start
with
once
you
have
that,
once
we
move
vm
CTO,
then
I
would
actually
attach
it
tag
it
as
incubating
and
how
to
so
for
other
people's
as
well.
Okay,
cool!
Thank
you.
Alright,.
G
I'm
here
oh
cool,
so
I
finished.
The
work
on
this
part
and
cool
thing
is
from
Cuba
perspective.
We
actually
don't
have
to
do
any
new
code,
only
small
small
native,
to
be
implemented.
What
is
it
was?
This
is
about
it's
about
providing
drivers
just
before
Windows
install
meaning.
If
the
user
has
built
I
all
drive,
he
needs
to
have
an
option
to
install
the
trial
drivers
if
he
has
some
custom
hardware
in
his
datacenter
that
it's
not
recognized
before
the
installation,
it
has
to
have
the
driver
before
so.
G
The
windows
recognizes
it
and
can
use
it.
From
this
part,
we
have
to
split
the
implementation
and
in
two
steps
providing
the
mateo
drivers
and
allowing
cluster
are
means
to
build
the
repository
of
the
drivers.
As
for
the
bit
I/o
cool
thing
is
that
we
can
actually
take
over
the
process
and
distribute
to
the
drivers
ourselves
from
the
google
earth.
The
register
disk
can
be
used.
For
this
part,
it
will
be
fairly
straightforward.
G
This
can
be
done
a
little
bit
differently,
but
again
the
parts
are
already
there.
The
we
can
may
utilize
containerize
data
importer,
which
allows
allows
cluster,
are
meant
to
build
its
own
cluster
repository
with
disks,
and
the
disks
are
then
mounted
as
data
volume
as
long
as
in
both
cases
vio
even
in
the
city
in
the
in
the
other,
drivers
that
is
are
present
before
Windows
installation
windows
will
correctly
recognize
them
and
allows
you
to
install
the
windows
drivers
for
any
hardware
want.
This
is
the
thousand
miles
overview
from
the
feature
I've
been
working
on.
G
A
Ask
a
stupid
question:
yeah,
so
is
it
effectively
that
the
work
you
did
or
envisioned
is
that
you
prepare
some
disk
so
volume
or
container
registry
disk,
and
they
are
attached
as
regular
disks
mm
to
the
VM,
and
there
is
no
code
change
needed
in
Cubert
direct
link.
You
ever
know
okay,
so
the
only
guarantee
thing
effectively
what
the
work
you
did
is
covering
to
provide
these
disks
right
to
provide
the
driver,
disk
I've.
A
H
There's
no
much
to
talk
about
there.
Really
we're
just
enable
this
option
to
set
be
said.
This
addresses
four
disks
and
and
the
network
interfaces
and
are
using
veritime
depending
on
the
chipset.
It's
a
it
different
addresses
can
be
allowed
and
we
expected
a
Liberto
to
provide
the
feedback,
whether
they
are
just
as
correct
or
not.
So
the
error
will
be
floated
from
from
libros
itself.
B
D
D
H
B
So
the
question
is:
when
you
look
at
the
virtual
machine
instance:
yeah
Mille,
it
is
not
for
discs
that
don't
have
a
specific
address
that
they
are
requesting,
we're
leaving
that
blank
and
then
Lippert
will
decide.
Is
there
a
way
to
read
those
addresses
back
from
libvirt
and
actually
find
them
at
the
cluster
level?
B
So
I
don't
think
that
we
actually
do
have
a
mechanism
at
this
point
to
read
those
back
kind
of
my
opinion
on
the
matter
which
is
just
mine,
is
that
if
the
user
doesn't
specify
it
it's
because
the
user
doesn't
care
I,
get
that
we're
talking
about.
You
know
Windows
activation
and
what
like,
and
we
want
to
keep
the
virtual
machine
is
like
as
possible
during
future
incarnations,
but
at
the
end
of
the
day,
if
it
wasn't
specified,
does
it
really
matter.
D
Well,
yes,
it
doesn't
matter
to
the
user,
but
we
already
discussed
this
in
the
past.
If
the
user
exports
the
VM
and
imports
them
imports
it
into
a
second
cluster.
Just
for
the
sake
of
conversation,
Venus
might
start
installing
drivers.
They
might
not
require
reactivation
little
Vladek
said
a
month
ago,
or
so
we
know
so
much
more
tolerant,
frightened
right
now,
not
sure
about
Windows
7.
D
But
recent
versions
are
pretty
tolerant,
but
anyway
they
might
install
drivers
and
they
definitely
do
that
when
you
just
take
your
USB
USB
key
and
plug
it
to
a
different
port.
That's
annoying
and
people
are
finding
bugs,
sometimes
about
that.
So
I
just
wanted
to
know
if
there
is
a
way
to
to
get
it
out,
there
isn't
I'll
have
to
figure
that
out.
We.
H
B
H
I
think
something
like
I
just
say:
it's
me,
I
just
want
to
briefly
mention
that
recently,
we've
merged
the
of
enabled
the
multi
cue
function
for
Everton
net,
and
basically
we
set
the
number
of
Q's
sorry,
the
number
of
virtual
CPUs
on
the
on
the
queues
and
merger.
Recently
the
documentation
for
this
and
then
I,
don't
know
just
try
it
out.
D
H
Don't
know
I,
don't
know,
there
are
a
few
issues
there
that
I
wouldn't
enable
it
just
unconditionally.
It
depends
on
them
on
them
all
kind
of
different
things
depends
on
the
traffic
that
you
expect
to
have
inside
the
guest,
and
there
are
all
kinds
of
different
implications,
also
consumed
MSI's
MSX
vectors
and
depends
on
how
many,
how
many
network
devices
do
you
have
and
how
many
virtual
CPUs
overall
you
have.
So
it's
kind
of
people
need
to
decide
for
themselves
whether
they
need
this
or
not.
H
B
All
right,
thank
you.
Next
up,
vert
control
image,
upload
I
had
added
this
note,
even
though
Mike
Hendrickson
owns
this.
But
basically
the
idea
here
is
that
this
feature
just
barely
missed
0.9,
because
it
was
still
under
review
when
we
cut
the
release.
So
we
are
attempting
the
back
book
that
currently
I
don't
know.
If
there's
any
other
thoughts
on
that
yeah.
A
Just
cite
not
so
I
mean
this
is
about
Qbert
right
and
we
has
read
that
we
plan
to
do
stable
release
based
on
the
police,
zero,
nine
branch
and
that's
what
we
favor
picking
that
feature
into
the
release
year,
nine
run.
Usually
we
don't
want
to
pick
features,
but
that's
sounded
like
a
justified
interest
and
thinks
a
lot
of
convenience
for
the
user.
So
that's
a
reason
where
we
said
we
as
maintainer
said
right
to
pick
their
features
yeah.
A
I
J
K
The
PR
is
still
interview.
There
were
quite
a
bunch
of
useful
comments.
Yesterday,
I
need
to
follow
up
on
that.
There
was
one
comment
about
testing
this
whole
thing.
A
remand
suggests
that
we
should
have
some
node
in
our
CI
cluster
with
the
PCI
device
that
it
would
be
able
to
be
served
by
the
Israeli
device
plugin,
and
then
we
would
have
functional
tests.
That
would
actually
you
know
well.
K
First,
we
would
have
probably
a
provider
that
would
that
would
would
actually
configures
everything,
and
then
we
would
have
a
functional
test
that
would
test
it
and
and
including
a
location
of
the
VMI
with
this
therapy
device
and
then
checking
apparently
that
it's
actually
located
whatnot,
so
I'm
very
happy
to
serve
with
the
with
function.
Tests
I'm,
not
sure
how
I'm
not
sure
whether
it's
in
the
realm
of
possibility
to
get
us
and
node
with
in
our
set
with
with
the
device
and
so
I
wonder
what
is
it
worth
pursuing
or
we
can.
K
A
Yeah
I
mean
I
could
take
a
look
that
we
get
a
box
from
get
a
box
from
our
pool
somehow
with
an
S
or
IV
device
to
be
used
for
CI
testing,
and
we
do
have
conditional
test
like
for
Windows
and
for
maltose
right.
Would
you
wanna
do
that
conditionally?
What
I
wonder
is
how
do
we
emulate
an
asura,
be
capable
device
Oh
somebody's
checking
his
head?
Oh.
K
K
K
K
K
A
Yeah,
actually
do
you
think
you
can
resolve
that
I
mean
we
know
that
we
have
a
pool
of
hardware.
The
question
is
that
we
need
to
sort
out
if
we
can
use
that
for
for
upstream
testing,
because
I
think
we
would
have
devices
right
like
in
our
internal
systems
and
I.
Just
wonder
if
we
can
integrate
that
somehow
yeah
with
with
upstream
testing
I
know,
we
can
take
that
off.
London.
Let's
see,
okay,.
K
Yeah
but
I
will
still
add
the
test.
It's
that's
a
good
suggestion
and
other
than
that
there
is
a
bunch
of
things
parallel
things
happening
in
the
device
in
this
repeat
device
playing
in
that
that
will
enable
us
to
support
some
additional
stuff
like
actress
for
properly
support,
multiple
SRV
networks,
religion,
privileges
in
the
containers,
but
this
is
all,
depending
on
on
the
progress
in
the
upstream,
though
I
think
we're
getting
close
there
now,
which
is
good.
A
K
You
know
and
another
another
thing
that
is
happening
in
the
device.
Plugin
is
adding
some
kind
of
a
mutation
hook
that
would
fill
in
the
limits
of
requests
in
in
the
port
spec
for
us,
because
currently
you
need
two
for
VMI.
You
need
to
attach
it
to
Montrose
network
and
also
specify
the
limits
and
frequent
ten
requests
explicitly,
which
is
not
nice,
because
we
already
have
this
information
at
the
motors
network.
B
K
D
D
K
A
B
F
So
I
have
some
updates.
We
succeeded
to
create
some
new
integrated
environment,
which
includes
two
components
like
prosperity,
external
provider
and
not
the
cover
component.
So
we
also
succeeded
to
run
the
wall
flow.
So
in
the
case,
if
not
start
to
be
non,
responding
from
some
reason,
networking
or
my
the
complete
gun,
not
recoverable.
F
You'll
see
the
change
of
the
node
will
monitor
not
until
call
you
delete
the
relevant
machine
machine
object
and
external
provider
will
run
actuator
of
the
specific
machine
you
power
off.
Just
a
note
and
after
power
on
to
note
the
upon
mention,
I
also
updated.
A
little
did
not
recover
repository
with
some
documentation
and
some.
A
F
From
our
work,
because
when
we
note
lost
connection
with
the
master,
it
just
all
the
pots
with
county-run
all
the
non-responsive
note,
you
have
straight
node
lost,
and
so
the
Nexus
cannot
rerun
them
because
they
can
have
some
persistent
volume
and
put.
The
latest
does
not
really
know
if
it
just
some
complete
problem
or
some
metal
problem
or
really
the
note.
F
You
has
power
off
or
off
state.
So
in
our
case
you
will
run
fencing,
commands
that
will
restart
restart.
You
do
note,
and
in
the
case,
if
it
was
some
network
problem,
this
restart
can
solve
it.
So
so
master
we
lost
our
connection,
need
to
know
and
we
will
shut
down
on
pot
and
we
will
be
stopped
us
from
from
you
from
station
I.
Believe.
F
Extent,
our
not
recovering
controller,
not
just
to
be
able
to
restart
the
mode,
but
also,
if
after
restart,
must
have,
cannot
restore
connection
to
the
north
I.
Basically,
our
controller
can
start
to
shut
down
pots
from
some
way
to
start
to
make
possible
from
master
to
stop
them
to
restart
the
pots
on
some
other.
On
some
other
note
also,
when
master
does
not
connection,
the
Volkhov
connection
is
not
but
I
believe
it.
You
will
need
to
check
how
we
can
do
it.