►
From YouTube: KubeVirt Community Meeting 2021-12-08
Description
Meeting Notes: https://docs.google.com/document/d/1kyhpWlEPzZtQJSjJlAqhPcn3t0Mt_o0amhpuNPGs1Ls/edit#heading=h.u74oyrl72es0
A
A
And
I'd
like
to
invite
anyone
to
go
ahead
and
introduce
themselves
if
they're
new
to
the
meeting,
I
will
actually
start
off.
I
was
briefly
introduced
last
week,
but
I
am
going
to.
I
will
be
helping
run
some
of
the
community
meetings
going
forward.
My
name
is
kat,
morgan,
previous
canonical
red
hat
and
dell,
currently
working
for
kong
and
leveraging
convert
for
personal
side
projects
and
some
ambitious
other
goals
as
well,
we'll
see
how
how
that
goes,
but
I'm
excited
to
be
part
of
the
community
now
and
getting
back
and.
A
B
So
what
are
we
going
through?
Now?
I
don't
think
ryan
is
here
quite
yet.
Okay,
maybe
we
can
move
that
to
the
end
of
the
agenda.
A
Yeah
we
can
see
if
they're
here
in
a
bit,
we
can
go
ahead
and
go
through
the
er's
mailing
risk
list,
review
and
bug
scrub
and
then
jump
back
to
agenda
notes
and
open
floor
by
chance.
Does
that
sound
good?
Well,
what
did.
B
Edward
have
did
you
want
to
just
move
on
to
edwards.
A
C
Hi,
can
you
hear
me
yeah,
I'm
in
the
middle
of
the
rain
here,
so
I
hope
you
can
hear
I
was.
I
was
like
doing
some
work
on
the
source
code
in
convert
cover
and
I
saw
I
assumed
something
in
the
past
and
it
seemed
not
not
to
be
correct.
C
So
maybe
here
I
will
get
it,
do
you
we
in
order
to
configure
interfaces
networking
we
have
two
two
entries.
C
C
B
I
think
I
can
answer
that,
so
it's
similar
to
the
reason
why
we
have
volumes
and
disks.
We
have
the
back
end
and
kind
of
the
front.
End
separation.
We
see
the
same
thing
with
pods
as
well,
where
they
have
things
like
volumes
and
then
volume
mounts.
So
you
take
the
back
end
thing
the
volume
and
you
describe
how
it's
being
injected
into
that
specific
container.
B
With
the
volume
mounts
we
followed
the
same
approach
when
we
were
designing
the
the
vmi
api
where
we
would
have
volumes
and
then
you
would
attach
it
using
a
specific
disk
and
a
disk
bus
and
things
like
that.
The
same
thing
extended
to
networks
where
we
would
have
a
network
provided
by
some
cni
plug-in
or
whatever,
and
then
we
had
the
binding
mechanism,
I
believe
in
the
interface.
So
that's
describing
how
and
the
fact
that
these
things
are
separated.
B
B
I
don't
know
if
we
felt
confident
that
that
would
be
the
case
at
the
time
and
we
also
for
consistency
lights
to
maintain
that
separation
for
network
the
same
as
we
did
for
storage
and
potentially
other
things
in
the
future
as
well.
But
there's
there's
a
good
argument,
both
ways.
I
think.
E
D
C
C
Network
attachment
definition-
and
that
was
like
I
thought
it
represent
like
it-
was
confusing.
A
I
was
on
mute,
okay,
yeah,
so
just
to
summarize
the
way
I
understood
it
is
we
have
the
two
different
objects.
We
have
the
the
network
and
the
interface
definition
that
the
network
piece
has
a
number
of
different
possible
tunings,
whether
you're
using
multis
or
not,
and
then
the
interface
also
has
several
different,
tunings
so
logically
grouping
those
separately
allows
you
to
maintain
a
clearer
api
spec,
while
maintaining
the
the
features
that
you
need
to
expose
in
each
of
those.
A
A
lot
of
them
can
be
one
to
one,
but
ultimately
there
are
deltas
between
those
two
parts
of
the
spec.
Does
that
answer
the
question
there
and
they
dropped
on
the
agenda.
Edward.
C
B
I
think
at
the
very
least,
we
have
to
have
the
separation
between
the
front
and
in
the
back
end,
so
those
could
have
exist
in
the
same
hierarchy,
where
you'd
have
like
a
an
interface
on
the
domain
which
described
the
the
back
end
and
the
front-end
separate
structures.
But
they
were
all
like
the
same
like
list
unit
in
that
list.
B
That
just
looked
really
odd
in
comparison
to
how
we
did
volumes
and
other
things
to
describe
things
that
way,
because,
for
example,
we
didn't
do
that
with
disks,
we
didn't
have
volumes
being
described
with
the
disc
in
the
same
entity.
We
had
that
split,
so
it
seemed
natural
to
have
the
same
split
for
the.
C
Know
comes
to
the
api
and
wants
to
understand
it.
I
think
in
the
in
the
current
state.
It's
it's
pretty.
I
mean
it's
not
natural,
not
trivial,
because
usually
you
will
set
like
in
any
other
system.
He
wants
to
set
some
properties
like
the
marked
the
mq.
Whatever
he
wants
to
set
up,
he
will
set
it
in
on
an
interface
and
then
what
is
used
for
the
pod
part
that
exactly
that's
like
an
internal
thing
that
needs
to
be
adjusted
to
fit
it
right.
C
It's
I
mean,
except
having
this
interference,
to
be
connected
to
a
powder
to
a
multus
network.
You
don't
really
need
anything
else.
This
is
actually
what
happens
today.
I
guess,
if
I'm
not
mistaken,
we
don't
have
anything
in
the
network
except
saying
if
it's
multus
or
and
what
will
the
multus
network,
and
if
it's
support,
that's
it,
we
don't.
We
don't
add
the
details
it
could
have
been
like.
Instead
of
having
the
name
connecting
between
the
interface
and
the
network,
it
could
have
been.
C
Just
I
don't
know
a
network
property
under
the
internet,
but
anyway
I'm
just
when
I
when
I
went
over
it.
It
was
like
very
odd
at
the
beginning.
A
Across
kubernetes
objects
starts
to
get
a
little
bit
long
in
the
tooth,
but
yeah
yeah
we're
maintaining
the
consistency
with
the
rest
of
the
kubernetes
ecosystem
with
the
current
separation.
So
thank
you
for
dropping
your
agenda.
Note.
Do
we
have.
A
Who
is
this
person
again?
Oh.
A
F
Joined
yeah-
hey,
hey
everybody,
so
I'd!
I
added
this
in
here.
It
looks
like
someone
left.
I
just
saw
this,
so
someone
has
an
example.
So
the
the
the
thing
I
want
to
bring
up
was
some
generating
the
the
runtime
scheme,
which
I
think
is,
I
think
it's
client
go
that
has
this
in
different
languages,
and
so
it
looks
like
cube
already.
Has
that
I
wasn't
aware
this,
but
the
this
is
python,
one
of
the
other
ones
that
the
one
that
I
was
specifically
interested
was
actually
java.
F
We
had
a
use
case
for
that,
but
I
basically
my
ass
was
that
if
there
was
some
precedent
for
this-
and
it
looks
like
there
is
so
I
guess
my
next
question,
what
would
the
process
look
like
if,
if
that
makes
sense
to
people
like
you
know
generating?
I
I've
never
done
this
before,
like
you
know,
what's
what
does
this
look
like.
F
I
don't
know
what
what
do
people
think
is
that
no
is
is
kubert
like
well
I'll
show
you
ask
us
first,
like
do
you
guys,
so
from
my
perspective,
we
have
a
use
case
for
doing
this
in
java.
Does?
Does
that
make
sense
like
to
have
another
client
in
java
to
like?
Is
anyone
opposed
to
that.
A
I
think
what's
interesting
to
me
about
that
is
I
have
been
interested
in
developing
for
coupe
vert
and
using
typescript.
So
I,
if
I
understand
right
correctly,
then
we're
we
could
start
off
by
defining
the
crd
in
a
package.
I
don't
know
about
the
convert
posting
of
that.
That
does.
G
Someone
want
to
speak
to
that
hey.
This
is
fabian,
and
I
think
that
was
somebody
from
from
from
nvidia
speaking
right,
hey
ryan
long
time
to
let's
see,
hey
family
and
hey
catherine
good
to
see
you
taking
us
through
this
fantastic
ride.
Yeah.
I
wonder
about
that.
I
mean
all
the
binnings
are
interesting
right
and
I
think
for
whatever
we
want
to
host
in
cubert.
I
think
we
really
like
these
standards
right.
F
G
I
would
say
the
relaxed
cncf
standards
right,
so
some
kind
of
ci
we
probably
want-
and
we
want
to
have
like
clear
ownership
and
releases
for
a
ci.
For
example,
we
should
ensure
that
there's
some
kind
of
you
know
that
there's
the
basic
assumptions
of
our
of
the
users
would
be
met
right,
so
possibly
even
functional
tests.
We
need
to
see
how
far
you
want
to
drive
it,
but
at
least
something
functional,
maybe
david.
You.
You
want
to
weigh
in.
B
What
do
you
think?
Well,
I
think
yeah
it
may
be
functional
at
the
very
least.
All
I
want
to
see
is
automation
that
ensures
that
this
client
stays
in
sync
with
our
client
go
like
our
our
go,
laying
client.
B
D
B
D
This
is
the
no
I'm
talking
about.
Someone
was
using
the
swagger
to
generate
a
java
client
and
there
was
a
bug
at
some
point
yeah,
and
I
think
the
python
client
that
we
have
that
we
release
is
based
on
the
swagger
too.
I
believe,
oh.
G
B
We
recently
made
some
changes
in
how
we
publish
our
api
and
or
our
golang
api,
and
it's
more
consistent
with
what
we
see
in
other
projects
and
it's
easier
to
consume.
I'd
like
to
understand
what
other
precedent
the
kubernetes
community
has
for
other
clients
like
java
and
python,
make
sure
that
we're
adhering
to
those
as
well.
B
So
I
think
my
ask
for
this
is
certainly
I
think
that
we
would
be
willing
to
host
a
java
client
just
like
we're
going
to
host
a
python
client
I'd
like
some
research
done
to
see
how,
like
the
kubernetes
java
client
is
created
and
maintained.
I
assume
that
there's
one
and
there
might
not
be
one,
but
I
would
sounds
pretty
standard,
and
we
should
be
following
that
precedent
to
make
sure
that
we're
doing
the
same
sort
of
automation.
Things
like
that.
Assuming
that
it's
good,
but
that's
where
I
would
start.
B
F
Okay,
I
think
I
mean
I
think
that
gives
me
always
a
platform
I
think
like
so
what
I
guess.
What
I
can
do
with
this
is.
This
is
just
something
that's
sort
of
that's
something
like
it's
like
on
our
roadmap,
but
I
think
what
I
I
think
would
be
at
least
the
best
step.
F
For
me
to
start
with
is
maybe
doing
so
doing,
generating
the
clients,
maybe
in
its
own
repo,
just
kind
of
testing
out,
seeing
how
it
goes
and
see
how
it,
how
we're
able
to
use
it
and
solve
our
use
case,
and
then
you
know
we
can
maybe
expand.
F
You
know
as
like
you
know
if,
if
everything
works
kind
of
the
way
we
expect
and
we
kind
of
follow
what
the
existing
kubernetes
java
client
has.
Then
maybe
we
can,
you
know
talk
about
some
of
the
other
components
here.
You
know
ci
releases
repromentators
so
on,
and
maybe
we
had
us.
We
have
a
conversation
of
possibly
having
one
in
three.
A
Look
at
that
sorry,
the
cat
was
going
crazy
in
the
background
I
didn't
want
to
mess
with
everyone
on
that
all
right.
So
does
anyone
want
to
speak
to
open
floor.
G
Yeah,
let's,
let's
let
it
not
be
silent,
so
I
think,
what's
coming
up
is
a
couple
of
conferences
right,
I
think
defcon,
I'm
not
sure
if
the
cfp's
still
open
but
def
con
from
foster.
It's
that
time
over
here
again,
oh
first
def
con
actually
closed
to
waltergon,
but
I
think
piatra
sent
an
email
to
cuberdev
that
foster
cfp
is
still
open
in
the
iis
room.
G
So
whenever
you
have
something
nice
to
just
to
say
about
kubert
in
the
sense
of
technical
advancements,
not
cubed
by
itself,
but
what
we
did,
what
we
achieved
or
how
you
can
address
use
cases,
which
is
this
just
a
shout
out
to
to
submit
your
paper
to
to
foster,
which
will
likely
be
in
a
virtual
setting.
G
G
And
then
more
conferences
will
come
up.
If
I
remember
stu,
I
think,
last
year
right,
there
was
like
what
was
it
on
your
content.
We
did
all
things
open
this
this
year,
actually
either
wasn't
there
an
admin
or
user
related?
What
was
the
use
next
or
something
like
this?
It's
also
coming
up.
Usually,
let's
see.
A
And
if
anyone
wants
to
speak
at
any
of
the
conferences,
but
it's
a
little
crutch
high
pulling
the
trigger
there
are
resources
you
can
spin
up
to,
of
course,
help
put
together
a
proposal
and
such
so
don't
don't
be
shy
on
that
and
with
that
and
go
ahead
and
jump
over
to
pull
requests.
So
it
looks
like
we
have
one
call
out
already.
C
I
think
it
just,
I
think
it
just
to
raise
like
a
notice,
not
nothing.
At
least
this
is
what
my
intention
here.
C
E
They
will
be
able
to
comply
with
the
new
host
devices
interface,
and
this
didn't
happen
yet.
F
Yeah
vladik,
I
haven't,
I
have
an
update
on
that
we're
we
have.
We
have
some
work
in
progress
on
that,
so
I
don't
have
a
date
yet,
but
I
think
I
think
it'll
be,
I
think,
we'll
be
handled
for
the
holidays.
At
least
we've
already
started
some
some
development
on
it.
F
So
I
think
in
the
next
few
weeks.
F
Yeah
the
yeah,
so
that's
my
hope.
My
hope
is
that
this
will
be
closed
up
for
the
new
year
and
then
then
we
can
merge
that
pr.
F
A
H
A
H
A
B
You
are
essentially
dumping
to
standard
out
the
cloud
in
it
log.
Well,
just
all
the
cloud
in
it
related
logs.
H
B
H
B
E
E
H
I
want
to
oh
yeah,
so
I
I
want
to
dump
this
audit
log,
because
I
saw
on
few
srv
lanes
failures
that,
for
some
reason,
the
there
is
no
permissions
for
the
vfio
device,
and
I
want
to
see
I
want
to
see
who
is
trying
to
get
it,
but
on
the
audit
log.
But
I
don't
have
the
the
log
form
and-
and
I
want
to
make
sure
that
there
is
no,
that
we
don't
miss
some
essay
linux
policies.
H
B
I'm
not
sure
if
it's
a
problem,
it's
it
might
be
confusing.
So
if
you
look
at
that
log,
let's
say
you're
on
one
of
the
whatever
upstream
ci
hosts
that
audit
log
is
going
to
contain
a
lot
of
things
that
aren't
related
necessarily
to
that
ci
run.
I
think
that
you
can
have
lots
of
things
happening
in
parallel
on
that
machine.
H
A
This
be
behavior
be
consistent
with
something
that
you're
used
to
with
general
kubernetes
and
regular
container
usage
or
with
something
like
openstack,
aws,
and
so
on
is
like.
Does
it
represent
a
behavior
that
you're
used
to
somewhere
else.
H
H
B
H
Oh,
I
I
just
I
didn't
really
remove
it.
He
moved
it
for
a
few
lines.
H
A
A
This
is
about
the
guest
agent
and
migrating
from
the
vmware.
A
Language
old
posts
was
concluded
with
it's
not
a
general
practice
to
scrub
old
content,
just
changing
the
path
going
forward.
A
A
This
one
was
an
interesting
thread
talking
about
how
the
topology
schedule
that
pods
are
scheduled
across
the
cluster.
A
A
A
Yeah,
that's
right
so
someone's
requesting
to
understand
what
flavor
they
are
scheduled
as
inside
the
guest
vm
itself.
A
A
A
Does
anyone
want
to
post
any
bugs
the
bug
scrub
agenda
that
I
can
hone
in
on?
Otherwise
do
a
quick
review.
A
I
B
C
B
I
Yeah,
but
is
it
yeah
I
mean
if
you
don't
create
your
generity
with
manually
or
I
think
vertendra
is
trying
to
create
an
empty
disk.
Img.
I
A
What
I
think
is
interesting
is
the
error
that
he's
getting
has
to
do
with
access
and
other
processes
and
he's
also
using
a
host
path.
So,
depending
on
how
he's
set
up
all
kinds
of
things
could
possibly
jump
in,
I
almost
wonder
if
he
switched
to
a
different
storage
provider.
If
that
would
go
away.
B
B
If
it's
on
host
path,
then
we
have
no
guarantee
what
access
is
being
given
to
that
specific
file,
because
it's
not
being
enforced,
for
example
by
the
pvc
or
anything
like
that.
A
A
A
B
B
He
says
that
it
was
he
suspected.
It
was
memory
at
first
that
we
were
maybe
over
committing
to
memory
or
something
along
those
lines,
but
he
was
able
to
reproduce
it.
Even
after
increasing
the
memory
on
the
nodes.
B
I
don't
know
if
logs
would
be
helpful
to
collect.
I
think
somebody
needs
to
have
access
to
that
environment
and
debug
it
for
themselves,
so
it
needs
an
engineer,
who's
able
to
reproduce
this
and
spend
time
with
it
to
understand
why
it's
happening,
and
it
might
be
something
that
we
have
to
raise
with
the
qmu
community.
Even
it's
unclear.
If
this
is
something
we
can
even
influence
or
if
it's
something
that
kimu
needs
to
address
and
somebody
just
had
to
go
through
and
sort
through
all
this,
that's
really
what
it
comes
down
to.
A
Be
sure
and
check
out
the
changelog
for
48
48.48
release
and
definitely
in
engages
on
the
google
groups
and
mailing
list.
If
there's
any
of
the
conferences
and
upcoming
talks
that
you'd
like
to
schedule
and
or
get
off
with,
and
I
think
that's
a
wrap.