►
From YouTube: KubeVirt community meeting 2020-11-25
Description
Recording of the KubeVirt Community meeting held on 2020-11-25 16:00 CET
A
B
B
B
B
E
A
B
E
B
Great
okay,
so
I
plan
just
to
kind
of
walk
through
high
level.
What
I
want
to
do
what
the
goal
is-
and
maybe
we've
already
seen
some
comments
on
here,
and
maybe
we
can
even
talk
about
something
as
I
mentioned
them.
So
the
the
goal
of
this
proposal
is,
as
I
talked
about,
is
save
and
restore
and
to
give
a
little
background
in
liver.
B
This
would
be
like
we
have
a
running
virtual
machine,
but
you
want
to
essentially
suspend
you
probably
heard
that
term
a
lot
or
like
sleep
so
that
the
virtual
machine
is
its
execution,
is
maintained
and
then
bringing
it
back
online.
So
like
kind
of
restarting
it
or
never,
really
starting
it
kind
of
restoring
it
so
like
bringing
out
asleep,
bringing
out
hibernate
and
resuming
the
execution
the
existing
work
in
place.
B
Is
there
some
work
that
mark
did
for
pause
and
then
pause
slightly
different
state
pause
like
it
stops
the
execution
of
cpu
instructions
and
things
like
that
are
running
to
disk.
So
it's
not
really
in
like
a
stopped
state
or
an
off
state
or
a
completely
like
hibernated
state,
so
slightly
different
use
case
there,
and
also,
if
you
look
at
stop
in
general,
the
qbu
process
does
stop,
but
there's
no
saved
state
associated
with
it.
B
B
So
I
for
my
research
with
liver,
like
when
I
kind
of
I
did
actually
like
a
little
poc
of
this
in
my
own
environment,
where,
by
default
without
any
sort
of
parameters
passed
to
save
and
deliver
the
community
process,
will
in
fact
exit
there
are
parameters
you
can
pass.
I
don't
know
what
they
all
are,
so
maybe
there's
another
expert
out
there.
That
knows
a
little
bit
more,
so
perhaps
it's
not
completely
necessary
for
that
community
process
to
exit.
B
But
when
I
look
at
the
I
look
at
the
diagram
of
liver,
save
is
sort
of
like
a
state
just
like
pause.
Where,
like
like
you,
I
don't
know
if
you'd
want
like.
I
don't
know
how
like
entirely,
how
liver
treats
it,
but
I'd.
Imagine,
there's
sort
of
this
place
where
liver
is
going
to
do
the
stopping
of
execution,
so
it
can
actually
save
the
ramp,
but
it
doesn't
mean
that
the
chemo
process
entirely
exits
and
the
pod
therefore
exits
gratefully.
You
may
just
do
sort
of
a
positive
execution
and
then
resume.
F
B
So
you
can
see
here,
like
you
have
a
vmi
it's
running,
I
have
a
pvc
bound
to
it.
That
is
the
disk,
and
what
I
would
do
is
I'd
execute.
I
do
it
here
with
this
vert
ctl
save
and
they-
and
I
do
they
get
on
the
vmi
and
you'll-
see
that
what
ends
up
happening
is
the
the
in
the
series
I'm
talking
about
where
the
keyboard
process
exits,
the
vmi
goes
into
succeeded,
state
and
the
pods
go
and
succeeded
say,
and
this
makes
sense.
B
This
is
consistent
with
kubernetes
like
as
if
you
had
essentially
had
a
pod
that
our
container
that
just
exited
zero.
This
is
the
what
is
happening
here.
It's
just
a
graceful
exit,
so
the
actual
resources
remain,
but
the
the
vmi
and
the
pod
stay
around.
So
the
so
the
state,
the
state
per
se
is
it's.
Just
a
and
kubernetes
world
is
just
an
exit.
There
could
be
a
condition
that
we
could
have
that
says.
Something's
happening
like
pause.
B
G
G
But
I
think
it
makes
sense
in
general
to
have
an
independent
task
to
to
condense,
display
information
for
cubecuttle
in
the
vmi
itself,
so
that
we
can
independent
of
the
state
machine
at
more
user-friendly
states
like
terminating
saving
post,
whatever
directly
in
the
status
field.
It
would.
It
would
not
be
the
face
status
with
them,
but
it's
kind
of
a
condensed
thing
just
for
video
for
display.
B
Yeah,
I
agree
with
you
yeah
there
are
things
that
we
can
do
and
I
I
think
you're
not
wholly
spelled
out
in
here
in
this
proposal.
I
think
that's
sort
of
some
of
the
details
that
I
still
need
to
add
and
roman.
You
already
left
a
comment
there
and
that's
something
I
do
need
to
to
do
and
I
will
add
those
things,
but
I
guess
my
goal
was
with
this
and
kind
of
explaining
it
is.
B
G
F
I
actually
want
to
elaborate
on
my
question
because
you
answered
about
the
pod
on
the
pod
level.
It's
clear!
We
need
to
remember
that
this
is
a
kind
of
a
higher
level
system.
This
will
covet
will
be
plugged
into
systems
where
you
have
billing
and
vm
level
management
and
so
on.
So
if
I
read
correctly
your
proposal,
you
want
the
vm
to
stay,
because
you
want
all
the
resources
to
still
be
hooked
into
the
vm
right,
so
it
won't
be
rescheduled
to
another
node
or
so
to
speak.
So
it's
still
there.
F
The
customer
is
still
paying
for
that
everything.
So
that
was
my
question.
Somehow
the
customer
need
to
understand
that
the
vm
is
there.
It
might
be
that
the
pod
exited
and
so
on,
but
the
vm
is
still
there
he's
paying
for
that,
for
example,
because
you
still
occupy
the
resources,
if
you,
for
example,
use
gpu
or
memory
or
whatever
you
want
to
kind
of
schedule,
all
those
resources
to
that
vm
right
or
I
go
to.
B
Yeah,
that's
that's
a
good!
That's
a
good
point!
So
this
this
is
some.
This
is
an
area
where
I
think,
like
you,
you're
you're,
giving
a
good
argument.
Sorry,
one
of
the
sides
of
this-
and
you
said
billing
like
that,
you're
still
maintaining
those
resources,
and
I
totally
hear
your
perspective
on
that
and
one
of
the
perspectives
that
I
talk
about
here.
I
think
it's
in
I'm
not
sure
which
section
it's
in.
B
I
think
I'm
talking
a
little
bit
further
down
here,
yeah
and
object,
cleanup
sort
of
the
other
perspective
of
it
and
and
which
is
that
when,
when
somebody
is
doing
this
save
you
know
what
what
it
would.
What
did
what
about
the
state
where
I'm
doing
the
save-
and
I
basically
just
wanted
to
bring
it
back
up
like
I'm
doing
a
quick
clone
of
an
offline
virtual
machine
and
then
and
then
I
bring
it
back
up.
F
Now
for
you,
but
but
but
how
can
you,
if
you
remove
the
vmware,
how
can
you
ensure
that
you
restore
the
the
vm
to
the
same
place
when
you
say
to
a
to
someone
that
is
saving
and
restore?
They
expect
to
be
restored
to
the
exact
same
position.
You
may
even
encounter
a
situation
where
you
try
to
restore
and
you
cannot
because
the
unique
resources
that
you
used
already
occupied-
and
this
is
a
very
ryan-
sorry
finish-
a
sentence.
Sorry
for
that.
No,
no!
G
Right
azure
here
it
would
be
interesting
for
me
to
have
a
little
bit
more
differentiations
because
so,
when
I
think
about
this,
the
safe,
but
the
bmi
should
still
be
running,
I
guess
the
the
advantage
would
be
that
the
ram
is
the
rem
usage
would
actually
be
smaller.
So
maybe
you
can
scatter
something
on
the
nodes
in
between.
But
apart
from
that,
where
is
the
difference
from
pause?
That
would
be
better
for
me
because
when,
as
soon
as
you
stop
at
part,
the
resources
are
not
no
longer
bound
to
a
vmi.
G
B
Okay,
so
you're
saying
hold
on
so
you're
saying
that
when
I,
when
this
pod
reaches
succeeded,
state
all
the
resources
that,
like
let's
say,
let's
say
I
assigned
like
a
pin
to
cpu
or
have
a
device
plug-in
or
whatever
it's
all
released.
Yes,
it's
all
released,
yeah,
okay!
So
then
this
this
is
no
different
than
if
it
was
deleted.
That's
what
you're
saying.
G
Yeah
so
I
mean
so,
if
you
think
about
attached
devices,
we
had
the
discussion
with
daniel
barangia
in
the
document,
so
I
don't
think
that
the
vmi
or
the
domain
itself
cares
too
much
if
it
gets
a
similar
but
different
device
again,
but
I
mean
there
may
be
device
states
in
some
points,
but
I'm
not
sure
what
the
details
are
here.
I
guess
it
depends
on
the
use
case,
but
so
it
like
it's
designed
here,
the
pot
going
to
succeed
it
it
it's
clearly.
It's
clearly
different.
B
Okay,
yeah,
okay,
so
my
understanding
was.
Is
that
because
I
I
do,
I
do
acknowledge
that
there
are
two
states
which
is
that
it
would
be
nice
to
be
able
to
maintain
everything
like
all
the
everything
attached
to
the
vmi
okay.
So
I
thought
this
would
have
this
main.
This
held
that
standard.
This
would
hold
on
to
the
objects.
Okay,
so
you're,
singing,
okay,.
G
G
B
So
it
sounds
like
maybe
instead
of
this,
since,
if
everything
is
freed
up,
all
the
resources
are
freed
up
when
we
hit
the
succeeded
set,
then
maybe
we
do
need
to
do
a
pause
followed
by
a
save
to
reach
that
one
use
case.
Well,
the
second
one
being
which
I
don't
know.
If
it's
realistic,
I
mean
it
sounds
like
that.
Maybe
it
isn't
that
you
do
a
save
and
an
exit,
and
then
things
are
released.
B
F
What
I
I
originally
saw
that
the
difference
in
use
case
is
that
you
actually
even
want
to
allow
live
snapshot
like
live
snapshot
of
the
overall
vm,
and
you
add
also
the
memory
state
right.
We
put
some
point
in
time
snapshot,
you
continue
with
the
vm
and
then
you
say:
oh,
I
want
to
go
back
to
the
kind
of
past
and
you
reload
everything
included,
including
the
memory
state
and
everything
which
is
very
valuable.
So
in
that
case
I
see
it's
very
valuable,
but
if
you
actually
stopping,
then
it's
identical
to
pause.
G
I
guess
it's
it's
an
improvement
as
a
yeah,
so
the
safe
restorer
has
this
one
use
case
that,
where
compared
to
pause,
where
you
can
say,
okay,
I
want
to
continue
this
game
tomorrow,
but
now
for
now,
I
don't
need
it,
and
all
resources
should
be
available
for
something
else.
Like
memories
of
you,
mostly
right.
H
So
sorry
can,
can
I
suggest
that
I
think
one
of
the
problems
with
this
is
that
the
use
case
is
not
clearly.
I
mean
what
what's
the
objective
of
this.
What
what
you
want
to
reach
in
the
end
and
then
maybe
it
can
become
like
an
improvement
of
pause
or
an
option
to
pause
or
a
new
like
a
new
save
command.
But
but
if
you
don't
have
the
use
case,
it
will
be.
It's
like.
We
are
talking
about
technicalities
that
that
that
means.
B
The
two-
I
guess
the
two
things
I'm
looking
for
is
being
able
to
so
the
the
save,
as
I
re
that
was
essentially
I
originally
outlined,
whereas
the
save
stops
the
virtual
machine
and
we
hold
on
to
the
parameter
file,
and
we
don't
really
care
about
those
resources
being
freed
up
to
me.
That
sounds
like
like
a
case
as
long
as
it's
possible
to
do
the
rest.
B
It
sounds
like
there
may
be
some
trouble
there,
but
I
think
that
would
be
something
that'd
be
helpful
because
I'm
essentially
in
this
state
the
execution
is
frozen.
I
understand
what
the
vmi
animal
is.
I
know
the
resources
I
need
to
bring
back
up.
So
that's
one
thing,
the
second
one
which
is,
which
is
what
I
was
trying
to
spell
out
here,
which
is
with
with
like
the
state
of
those
pods
which
is
we
have
a
vmi.
B
B
A
snapshot
of
well,
I
mean
I
guess
we
can
talk
yeah
like
so
it
yeah
snapshot
of
a
virtual
machine,
doesn't
matter
like
whether
it's
if
the
virtual
yeah,
so
I
guess
the
vmi
would
be
live
virtual
machine.
So
snapchat
of
a
live
virtual
machine
take
all
those
resources
and
then
yeah
the
snapchat
would
be
the
copy
and
then
being
able
to
then
take
that
copy
and
launch
the
a
new
one
of
a
new
identical
copy
in
a
different
name.
Space.
I
I
F
H
H
F
But
added
the
point
is
that
in
the
in
a
user
perspective-
and
I
think
your
point
of
what
is
the
use
case
is
important
because
from
user
perspective,
when
you
tell
him
create
a
new
vm,
he
expecting
maybe
that
in
the
worst
case
he
it's
not
scheduled
right.
There
is
no
place,
but
if
he
save-
and
I
want
to
restore
you-
expect
it
to
to
be
restored,
it's
very
strange
to
save
and
then
restore,
and
you
get
an
answer.
Oh
I
couldn't
I
I
didn't
find
the
place.
H
A
A
I
don't
know
if
that's
the
correct
choice:
if
a
vm
is
named
spaced,
it's
got
certain
like
there's
implications
to
that
and
there's
a
reason
to
do
a
namespace.
If
you're
only
allowed
to
see
you
know
the
foo,
namespace
and
but
but
vm
staves
are
not
namespaced.
Would
that
not
mean
that
you
could
then
see
other
namespaces
save
files?
B
I
need
to
think
a
little
bit
more
where
I
used
this
case,
so
that
wait.
Let
me
let
me
try
to
so
they,
if
you
save
you,
save
the
ram
file
in
in
a
pvc
that
pvc
can
be
whoever
can
access
whatever
that
that
pvc
is
kind
of
what
to
use
it
to
attach
to
a
device,
that's
kind
of
like
I
don't
know,
that's
kind
of
where,
where
I
am,
maybe
you
can
repeat
your
use
case,
and
I
just
that's.
A
G
B
Yeah,
so
I
let
me
give
an
example,
because
I
think
that
price
service
better
so
like,
for
example,
a
good
example,
would
be
like
forensics.
Let's
say
I
have
virtual
machine
running
and
whatever
it's
misbehaving,
something's
wrong
with
it,
and
I-
and
I
can
tell
so
I
want
to
diagnose
the
problem,
so
I
have
it
running
in
a
namespace.
B
I
want
to
freeze
its
entire
execution.
I
want
to
freeze
its
disk.
I
then
want
to
say
make
a
copy
of
it
hold
on
to
it
for
later
and
then
put
it
into
my
secure
namespace
in
like
an
offline
environment,
diagnose
the
problem
and
see
if
there's
something
I
need
to
look
at
and
to
roll
out
across
my
foot
or
something
like
that.
That.
F
B
You'd
move
the
virtual
machine
to
sort
of
like
another
namespace
sort
of
maybe
it'd
be
offline
whatever
it
be
because
you
don't
want
when
you
run
this
execution,
let's
say
you're
assuming
it's
a
problem.
You
don't
want.
Maybe
you
don't
want
this
going
out
to
the
internet
or
something
it's
a
total
like
as
if
it
was
a
different
environment.
G
I
Yeah
absolutely
there's
another
thing
with
the
with
forensics:
it's
there's
something.
That's
called
pv
panic
that
triggers
core
dump
and
saves
the
score
down
when
it
feels
that
vm
is
behaving.
When
there's
a
there's
a
and
with
these
two
options
we
can
definitely
provide.
I
mean
save
options
for
pvc
or
something
like
that
like
on
the
fly
when
it
happens,
and
we
can
we
can
detect
when
it
happens.
So
these
are
really
good
ideas,
but
I
mean
it
doesn't
fit
under
this
moment
into
this
model
of
save
and.
I
A
A
Yes,
I
think
that
would
be
a
decently
close,
so
in
a
physical
machine.
Hibernate
means,
of
course,
to
save
the
ram,
and
you
know
basically
pause
in
this
case.
It's
more
just
think
of
it,
as
the
virtual
machine's
clock
stop
ticking
from
one
second
to
the
next.
You
can
just
start
execution
again,
so
there's
no
moving
of
memory
around
specifically
when
you
say
pause
itself
in
this
context,.
E
G
G
What
was
it
called?
Snapchat
snapshot?
S
snapshot
create
s
command,
I'm
not
sure
what
api
calls
that
are,
but
it's
also
different,
safe,
even
restored,
and
there
you
could
take
a
live
memory
dump
and
then
start
the
vm
again.
It
could
be
part
of
the
flow
also.
I
can
also
share
it
with
you,
then,
on
the
document.
B
Yeah,
I
think
so
I
guess
maybe
to
like
talk
about
what's
next,
like
the,
I
think
why
don't
I
add,
a
leverage
on
the
use
case
in
this
document.
It
sounds
like
I'm
in
the
area
of
what
I'm
looking
for
I
mean
it
sounds
like
restores
got
a
lot
of
challenges
to
that
save
seems
reasonably
close.
I
need
to
expand
a
little
bit
on
the
pause
aspect
of
it
in
that
use
case,
because
I
think
what
I
have
here
is
slightly
off,
so
why
don't?
B
D
Okay,
yeah,
and
maybe,
if
I
mean
beyond
beyond
the
use
case,
I
think
it
would
be
also
good
to
make
a
section
or
like
clearly
elaborate
on
the
difference
between
this
approach
and
a
pose
and
a
snapshot,
basically
things
that
already
exist.
So
what
was
this
compared
to
posing
or
snapshotting
or
the
combination
of
pose
so.
E
H
I
hope
this
one
will
be
very
quick,
so
there
was
some
movement
in
the
pr's
with
regarding
ghost
x,
several
piazzas
were
sent
and
all
kind
of
fixes
done
in
this,
and
I
was
wondering:
was
there
any
any
centralized
document
or
suggesting
this
move,
and
if
not,
can
we
have
one?
Because
there
are
some
points
there?
A
You
know
that's
a
very
good
point.
We
actually
there
does
exist
a
document.
I
think
it
was
internal
to
red
hat
in
the
their
issue
tracking
system,
but
now
that
you
mentioned
it,
I
don't
think
there's
any
reason.
It
needs
to
be
there
as
opposed
to
public,
because
there's
nothing
going
on.
That's
that
shouldn't
be
discussed
at
a
higher
level.
A
H
K
F
Yeah,
so
so
we
we
ca
yeah,
we
can
definitely
add
an
issue
still
republic
should
have
also
added
what
are
the
plans
regarding
how
it
will
be
handled
in
the
pr
testing
right?
Is
it
going
to
be
just
a
warning
or
fail
the
whole
pr,
and
all
of
that
we
will
cover
all
of
that
in
the
issue.
I
can
open
an
issue
and
start
the
discussion.
F
H
G
From
my
experience,
if
I,
if
it's
pretty
ugly,
we
want
to
go
in
general
if
just
for
discussing
something
which
is
but
which
you
want
to
write
down.
Google
docs
are
a
good
starting
point
and
then
moving
over
to
the
github
issue
when
it's
more
about
finalizing
it,
because
then
yeah
it's
easier
to
do
the
initial
discussions
in
google
docs
from
my
experience,
but
I
don't
mind
too
much.
H
I
think
what
I
think
that
last
discussion
here,
so
I
hope
that
pr
will
get
in.
If
anyone
here
has
the
he
likes
it
also,
it
will
be
great
if
you
can
approve
it,
but
the
the
part
of
adding
the
cidr.
I
think
we
discuss
it
here
a
few
weeks
back
and
we
said
that
we
will
have
a
separate
section
that
would
include
all
the
information
from
the
from
the
guest
agent
and
later
on
roman
mentioned,
that
it
should
be
formatted.
H
G
C
C
So
another
topic:
I
have
the
rest
coverage
lens.
It's
been
a
while,
since
I
created
the
old
pr,
so
I
saw
that
daniel
replaced
my
old
pr
and
created
a
new
one.
So
I
understand
daniel.
You
want
to
merge
it,
and
I
was
wondering
if
your
test
is
tested
this
recently
and
another
thing
one
note
from
my
site:
this
tool
considers
status
fields
too
and
my
I
must
change
it
to
only
spec
fields
that
it
gives
good
percentage
and
real
values.