►
From YouTube: KubeVirt Community Meeting 2023-02-15
Description
Meeting Notes: https://docs.google.com/document/d/1nE09vQWcCTW-9Ohe9oCldWrE0he-T_YFJ5D1xNzMtg4/
B
B
All
right,
it
looks
like
we
have
attendees
logging
attendance,
so
I
appreciate
that
and
while
you're
in
there,
if
you
have
anything
to
add
to
the
agenda,
either
agenda
notes
open
floor
or
any
PRS
bugs
or
mailing
list
items
that
you
would
like
special
attention
on.
Please
go
ahead
and
add
those
now
I'll
continue
to
watch
for
any
late
additions.
While
we
go
through
things,
so
don't
worry
if
we
don't
see
it
on
the
first
round
through.
B
Of
course,
do
we
have
any
new
attendees
on
the
call
who
would
like
to
introduce
themselves
whether
you're
new
on
this
call
or
just
I've,
been
listening
in
and
haven't
introduced
yourself
before
we'd
love
the
opportunity
to
go
ahead
and
say
hello
and
welcome
you
to
the
Cooper
community.
B
It
looks
like
my
art,
we
have
a
an
item
about
tests
passing
you
want
to
speak
to
that.
C
Okay,
so
am
I
audible.
First
of
all,
yes,
oh
yeah,
so
there's
this
test
about.
It's
really
really
old.
It's
testing,
apparently,
okay,
so
for
context.
There's
something
called
owner
references
and
if
an
object
has
an
owner
and
this
owner
does
not
exist,
then
kubernetes
deletes
that
object
and
we
use
that
to
make
sure
cleanup
happens
of
like
extra
items
like
pods
and
PVCs.
C
I,
guess
and
there's
like
a
esoteric
feature
where
you
can
delete
with
Cascade
Falls
were
If,
the
child
objects
have
an
owner,
it'll
remove
the
owner
first
and
then
they
can
keep
existing.
So
you
know
in
this
case
we
are
deleting
a
VM
object
and
we
expect
normally,
the
VMI
is
owned
by
the
VM
if
it
was
created
by
a
VM
and
also
the
PVC
and
the
data
volume,
and
we
checked
that
if
you
delete
the
VM,
then
the
VMI
with
the
Cascade
forced
and
the
VMI
exists.
C
And
so,
if
you
try
to
remove
the
owner
like
manually
on
a
VMI
that
is
owned
by
a
VM,
then
keyword
will
re-add
it,
and
sometimes
it
decides
to
also
delete
the
VMI
itself
explicitly
so
I
don't
know
if
it
is
kind
of
surprising
It,
Ever
Was,
and
it
survived
this
long.
This
test
so
I'm,
not
so
pretty
sure,
like
it's
kind
of
weird,
to
remove
a
test.
C
That's
several
years
old
and
say
that
it
shouldn't
have
ever
worked
because
it
obviously
does
and
then
for
a
really
long
time
does
anyone
have
like,
since
the
behavior
to
re-adopt
orphans
is
so
old.
It's
not
something.
That
seems
like
a
good
idea
to
change.
You
know
I,
don't
know
I
wonder
if
anyone
has
any
opinions
on
this
and
just
generally
I
would
like
to
bring
attention
to
this
issue.
A
D
A
I
definitely
can't
speak
authoritatively
about
this,
because
it's
not
something
that
I've
worked
on
or
that
familiar
with,
but
it
would
make
sense
to
me
that
you
wouldn't
want
to
have
an
orphaned
machine
lying
around.
So
I
could
see
why
you
might
want
to
make
sure
that
someone
owns
that
VM,
but
the
same
time.
It's
weird
that
there's
this
inconsistent
behavior
that
you
pointed
out.
C
B
A
I
mean
you
would
think,
because
so
and
maybe
I'll
just
go
and
look
at
it,
and
maybe
we
can
talk
about
it,
a
bit
more
on
there,
but
this
is
just
more
again
kind
of
abstract
or
hypothetically,
but
you
think
that
you
know
you
would
want
that
to
kind
of
be
the
default
Behavior.
If
you
kill
a
parent,
you
don't
want
to
have
any
any
children,
VMS
left
behind
without
explicitly
stating
so
but
yeah
I'll
I'll
pull
up
and
see.
If
I
can
we
can
have
a
discussion
about
it
on
there.
C
So
in
this
case
delete
without
the
Cascade
false
thing
which
doesn't
leave
the
VMI
and
other
objects
around
that's
the
default.
This
is
some
kind
of
alternative
delete
Behavior,
where
it
orphans
it's
supposed
to
often
resources
that
are
previously
owned
by
it.
C
Like
the
test
itself
is
testing
that
you
created
the
VMI
by
setting
running
on
a
VM,
I
guess
and
if
you
delete
the
VM
you're
still
left
with
the
VMI.
That's
all
you
can
still
use
that.
C
If
you
can,
you
can
create
a
VMI
without
the
VM
object.
I
think
if
I'm
not
100
sure
about
this.
A
Good
question
again:
I'll
I'll
do
what
I
can
to
look
at
that
again,
I'm
not
super
familiar
with
it.
It's
not
something.
I've
worked
on
just
again,
you
know
aspect
of
just
looking
at
it
as
an
objective.
Third
party
seems
kind
of.
A
C
B
All
right
and
then
it
looks
like
the
nest
next
item
is
about
arm
64
supported
releases.
Do
you
want
to
go
ahead
and
speak
to
that.
B
E
Okay,
So,
currently
the
sale
testing
of
64
works.
Well,
so
I
want
to
discuss
about
how
what
is
the
next
step?
We
need
to
do
to
make
QB
words
officially
released,
arm
64
version
in
my
image
and
the
boundaries
so
I
list
some
some
points
here.
An
array
helped
me
to
at
least
some
more
like
do.
We
need
to
add
more
E3
testing.
F
Yeah
one
related
question
I
had
is
that?
Can
we
can
we
start
the
existing
test
suit
with
arm?
Is
that
possible.
E
Waiting
elbow
some
core
tests-
ins,
sync
compute,
I'm
664
for
now,
but
for
Network
and
the
storage
there
is
no
test.
Wait,
wait.
Wait,
do
not
have
the
ETV
test
on
64
for
now.
G
So
I
think
that
you
I
think
you
need
to
First
that
we
have
the
list
of
conformance
tests
which
some
some
some
vendors
are
using
to
run
instead
of
running
three
hours
of
10
hours
of
test.
They
just
run
this
this
guys.
Yes,
so
you
could
try
to
just.
We
could
try
to
say
that
this
is
the
representative
test
or
features
that
need
to
run,
but
I
also
I'm
not
really
sure
that
it's
it
is
enough,
but
it
is
a
starting
point
and
maybe
it
is
worth
opening.
G
You
know
a
thread
or
or
part
of
some
document
of
which
features
you
want
to
cover,
and
then
anything
that
is
not
supported.
You
will
just
say
it's
an
exception
that
doesn't
work,
but
that's.
D
A
general
I'm.
H
Hey
hi,
Howard,
hi
I
have
a
question
for
you,
so
the
what
is
it
like
would
you
say,
are
the
remaining
barriers
for
you
to
to
make
the
remaining
test.
Suites
work
like
the
remaining
core
test.
Suites,
like
the
network
storage.
Is
that
something
that
is
it
kind
of
your
approach?
So
you
think
that
we
that
we
just
that
we
need
to
compute
and
that
that
should
be
enough
and
then
or
is
it
like?
Are
you
still
trying
to
get
these
Network
and
storage
Lanes
to
pass
with
our.
E
I
want
to
verify
the
network
and
storage
on
64.,
but
not
of
all
the
tests.
Yes,
so
I
would
like
to
have
a
like
some
list
or,
like
AdWords
said,
like
the
com
make
the
conformance
test
works
on
first,
okay,
I'm,
not
sure
it's
a
conformance
test
includes
some
networking
tests
or
storage
test.
H
So
I
guess
what
I?
What
I
would
advise
on
this,
and
is
that,
like,
when
kind
of
when
we
look
when
looking
at
some
of
the
PRS
for
some
of
the
stuff
or
even
just
up
here
in
general
right,
there
are
the
the
core
lanes
that
run
on
every
PR
and
there
is
the
arm
tests
that
run
in
there
having
the
having
additional
tests
that
include
network
storage,
I
think
really
empowers
the
case
of
having
an
official
release
for
arm
and
even
if
it's
just
a
subset,
I
think
kind
of
the
the
question
here
becomes.
H
You
know
if
I
was
a
user
and
I
was
to
show
up
and
wanted
to
look
at
keyword
and
I'm
wondering
what
is
the
extent
that
arm
is
supported,
I
I
think,
if
we're
going
to
say
it's
the
official
release,
what
I'd
expect
is
I'd
want
to
know
the
answer
to
that
question.
What
is
you
know?
What
is
the
extent
of
the
support,
and
so
that
would
mean
like
you
know.
What
are
the
features
that
that
are
supported
that
are
document
and
they're
documented?
H
That's
you
know,
essentially,
what's
the
standard
is
on
for
x86
and
that
the
the
code
or
the
features
that
are
supported
are
tested
in
in
CI
Lanes,
so
I
think,
like
the
where
it's
what
it
sounds
like,
what
the
what
there's
still
is
a
gap
is
that
there
are
some
tests
that
we
haven't
done.
We
haven't
confirmed
the
features
notice.
The
features
that
we
that
we
want
to
support
and
I
don't
know
the
extent
of
whether
it's
been
documented.
E
H
E
E
Okay,
so
the
signal
one
supports
long
I
think
it's
hybrid,
the
right,
coursing
VMI
API,
so
Ray,
you
are
doing
this
right
now
right.
H
Yeah,
so
this
is
something
we
we're
looking
at
yeah,
so
this
is
for
people
who
don't
know
like
this
is
there's
a
threads
field
on
the
vmis,
and
the
assumption
is
at
least
I'm
pretty
sure
the
assumption
is
hyper
as
hyper
threats
and
an
arm
course.
You
don't
have
that,
and
so
we
need
to
come
up
with
a
solution
for
how
we
deal
with
this,
how
we
assign
CPUs
so
yeah
like
we're,
we're
looking
at
this
right
now
and
we
have
some.
E
Occur
and
a
sort
of
a
hybrid
support,
all
right
yeah.
This
is
a
good
future,
but
I
don't
think
it's.
This
will
block
officially
released
of
64
image
right
yeah.
H
I
agree,
Howard,
I,
think
the
reason
I'm
mentioning
this
is
because
I
think
I've
seen
like
four
issues
or
more
on
this,
and
and
this
is
a
PR-
it's
already
open
for
it
and
it's
already
been
a
lot
of
work
done
for
it
and
it's
been
reviewed
extensively,
so
I
I,
don't
think
it's
a
blocker
I
just
wanted
to
call
it
out
that,
because
people
have
had
a
lot
of
interest
in
it
because
they're
dealing
I,
don't
know
they're,
maybe
using
arm
computer
arm
control
plane
in
the
next
86
for
the
other.
H
E
I
see
okay
and
there's
a
last
one.
Do
we
need
to
make
support
on
Cydia
CDI.
I
I
Far
as
I
as
far
as
I
know,
Alexander
Wells
is
there's
more
of
the
built
stuff,
but
pretty
sure
we
build
the
arm
and
images
what
other?
What
else
do
you
need,
assuming
that
that
exists.
E
Yes,
I
know
CBI
is
important,
but
is
it
so
if
a
way
currently
for
now,
we
do
not
support
CDI
Arma
64
and
where
this
block
the
officially
released
for
qbivert
image
or
banneries.
E
I
Yeah
I
guess
this
is
something
we
can
bring
up
with
Alexander
Wells.
If
you
put
a
note
there
as
far
so.
I
Yeah,
when
he
do,
you
have
any
idea
what.
I
C
Various
reduced
no
testing
specifically
on
arm
for
CGI,
so
it
would
be
weird
to
say
that
it
is
supported.
G
I
I
actually
understood
the
question
differently,
so
I
understood
I
I
thought
that
the
question
is
like
can
covert,
be
be
released
as
supporting
arms
64
without
CDI
support.
G
I
Yeah
yeah
and
the
we
do
check
yeah.
You
just
can't
use
any
of
the
data
volume
stuff.
Obviously,
but
yeah
you
should
be
able
to
have
keeper
without
CDI.
F
I
have
a
question,
so
if
we
release
keyword
without
CDI,
how
do
we
prevent
users
from
like?
What
is
the
error
message
that
will
say?
Oh
because
I'm
using
CDI
is
not
supported
and
hence
I
should
not
use
this
well.
I
Yeah
so
I
think
if
you
create
a
VM
that
has
data
volume
templates,
we'll
check
to
see
that
I'm
pretty
sure
that
we
check
it
in
the
web
hook.
But
obviously,
even
if
the
VM
gets
created,
the
for
controller
won't
be
able
to
create
the
data
volumes
later.
I
So
your
VM
will
never
get
provisioned.
But
I
think
that
there
is
a
check
in
the
web
hook.
You
can.
You
can
check,
but.
I
Basically,
the
the
that's.
The
main
interaction
with
the
cubert
API
is
that
there
are
the
data
volume
templates
embedded
in
a
VM,
and
it
would
be
easy
enough
to
validate
on
the
web
hook
if
CVI
is
installed.
F
Make
sense
so
if
we
say
we
can
release
it
without,
we
can
release
the
arms
about
without
CDI.
Is
that
something
we
would
have
to
add,
or
it's
already
present.
F
I
think
when
one
more
Point
regarding
this
so
for
for
features
that
we
say
that
we
are
not
going
to
support
for,
does
it
make
sense
to
have
similar
web
hook
tests
or
checks
to
make
sure
that
we
give
the
user
the
correct
error
while
creating
VMS.
F
Like
this
is,
this
is
in
parallel
to
the
point
that
Ryan
was
mentioning,
that
we
need
a
document
that
describes
the
supported
and
not
supported
features.
Maybe
one
more
step
forward
would
be
to
add
that
webhook
check,
validation
to
make
sure
users
get
correct,
error
messages.
G
G
G
Doesn't
I
mean
it's?
We
are
in
a
reconcile
situation,
so,
first
of
all
you
cannot
install.
You
will
not
be
able
to
deploy
CDI.
If
then,
if
you
don't
have
both
to
deploy
them
and
and
the
nodes
may
be,
part
of
them
may
be
arm
and
part
of
them
may
not
be
harmed,
so
it
will
it
I.
Guess
the
the
need
will
be
that
the
somewhere.
Some
things
cannot
run
on
some
nodes,
something
like
that
and
some
feature.
G
If
a
VMI
lands
on
on
some
specific
node,
it
will
not
just
not
be
able
to
either
start
or
use
the
feature,
so
it
will
just
blow
up,
but.
D
F
So
what
I'm
sorry
I
think
is
that,
instead
of
having
that
VMI
go
in
into
a
state
where
it
will
not
start
up,
can
we
block
at
the
API
level
where,
if
we
have
a
list
of
fields
that
are
supported,
we
can
have
add
that
check
during
creation
and
webhook
validation,
state.
G
No
I
think
we
don't
do
this
even
now,
because
again
it's
like
the
VMI
will
not
be
scheduled.
I
mean,
let's,
let's
say:
let's
assume
that
you
want
to
run
it
somewhere
and
you
only
have
and
it
cannot
run
on
any
on
any
existing
nodes
because,
for
example,
the
hardware
type,
so
it
will
just
not
be
scheduled
because
it
has
nowhere
to
to
land
so
I,
don't
know
how
can
you
check
it
at
the
webbook?
G
H
It's
like
it
would
be
a
feature
we
supported
x86
that
we
don't
support
an
arm
like
one
example-
and
this
isn't
this
isn't
a
case,
but
let's
just
imagine
that
we
didn't
support
srirv,
attaching
srov,
Network
or
something,
and
the
VMI
spec
has
an
sreb
Network
on
it
and
as
part
of
the
VMI
it
has.
H
It
has
the
architecture
defined
for
for
arm,
for
whatever
reason,
so
in
that
case
we
know,
SRI
review
is
not
supported,
so
the
question
is:
should
we
should
we
deny
it
at
the
API
level,
because
we
know
it's
an
invalid
VMI
because
we're
not
supporting
it
because
we
know
arm
isn't
going
to
support
this.
H
H
Yeah,
so
this
is
where
okay,
so
this
is
where
it
gets
to
I
think
this
is
where
it
gets
into
that
hybrid
support
PR,
because
that
hybrid
support
PR
allows
you
to
actually
specify
that's
a
new
architecture
field
that
allows
you
to
specify
which
node
to
run
on.
So
we
can
actually
figure
out
ahead
of
time
if
we're
going
to
be
targeting
an
arm,
node
or
not.
G
D
G
I
get
I,
guess
you
you
are
so
there
is
a
actually
I
think
there
is
some
kind
of
disagreement
or
an
argument
about.
Should
you
what
should
you
do
at
the
Web
book
level
and
what
you
you
should
do
at
the
reconcile
level?
So
some
say
that
in
the
workbook
level
you
should
do
this
or
this
validation
and
some
claim
that
you
should
just
let
them
the
all
the
things
to
just
flow
and
if
it
is
not
able
to
do
that,
then
scream
that
it's
impossible
or
something
like
that,
but
yeah
yeah.
H
I
think
maybe
we
need
to
this
is
something
we
can
replace.
I
think
we
have
to
like
I
think
we
can't
do
this
until
we
at
least
have
the
hybrid
support
PR,
because
I
don't
think
it
can
be.
We
can
even
discuss
it
I
think
to
to
your
point,
because
we
have
no
way
of
knowing,
because
we
won't
know
the
runtime.
We
won't
know
where
we're
scheduled.
That's
that's
much
farther
down
the
pipeline
than
the
API.
H
We
would
only
know
if
we,
if
we
added
on
the
VMI
spec
and
then
it's
really
a
matter
of
like
whether
it
makes
sense
from
a
policy
standpoint
and
that's
something
we
can
discuss,
but
we
don't
I,
don't
think
we
have
the
proper
tools
to
to
at
least
have
this
discussion.
The
LA
so
I
think
we
need
to
wait.
I
Yeah
I,
agree
and
I
think
that,
even
if,
like
for
example
with
like
with.
I
Populate
a
a
data
volume
on
a
you
know
on
a
non-arm
node
in
that
data
and
that
PPC
could
be
used
on
a
arm
node
later
or
something
like
that.
So
that
would
be
hard
to
validate
yeah
or
no
upfront,
but
I
did
look
into
the
code
quickly
and
it
doesn't
look
like
our
web
hook.
Does
any
checking
and
I
think
that
may
be
consistent
with
like
what
Edward
was
saying
about
upfront,
validation
versus
just
letting
things
go
yeah,
it
doesn't
look
like
we
do
any
validation.
I
With
the
data
volume
the
TDI
is
installed
in
a
web
hook
level.
The
assumption
is
probably
about
someone
could
later.
You
know
see
that
the
VM
didn't
get
provisioned
and
install
CDI
later
or
something
like
that.
F
So
ran
a
follow-up
question
to
that
is,
like
I
think
you
said
that
we
need
to
figure
out
how
we
would
be
able
to
make
these
kinds
of
validation,
because
the
info
doesn't
exist.
F
So
can
we
say
that,
like
figuring
out
that
parties,
at
least
before
the
release
is
blocker
for
release
because,
like
the
concern
I
have,
is,
if
we
don't
add
any
validation,
whether
this
is
arm
or
not,
and
what
features
are
supported
then
I
fear
like
we
will
start
seeing
issues
where,
like
users
are
using,
some
features
on
x86
and
same
users
are
using
that
feature
in
am
it
works
on
one
and
it
doesn't
work
on
the
other
and.
H
Well,
there's
there's
sort
of
a
sort
of
balance
here
that
one
could
argue
that
if
we
document
this
well
enough,
then
any
user.
Doesn't
you
know
if
the
user
is
not
following
the
documentation
of
what
we're
saying
is
supported?
Then
then,
that's
unfortunate
and
but
I
understand
also
the
argument
of
like
okay
worked
in
X
A6,
so
I
expect
this
to
just
work
everywhere
and
and
then
and
I'm
not
going
to
read
the
documentation
that
that
could
happen
too.
So
I
I
could
see
both
sides
of
this
so
I
mean
I.
H
Think
what
you're
saying
is
like
where
maybe
what
we
need
to
do
is
we
need
to
say
that
the
hybrid
support
part
of
this
is
a
blocker
and
then
the
discussion
of
of
whether
this
is
a
should
be
a
web
work
or
not
need
to
be
the
blocker
for
for
the
for
the
arm
release
right,
yeah.
F
That
we
can
defer
yeah
future
releases,
but
at
least
the
hybrid
support,
or
or
at
least
a
plan
of
how
we're
going
to
validate
these
things.
Okay,.
H
Yeah
I
mean
I,
guess:
okay,
I
I,
don't
know,
I
mean
I,
think
I
guess
we
could
go
either
way.
I
mean
to
me,
like
I
I.
H
Think
at
the
end
of
the
day
that
this
hybrid
support
thing
is
likely
to
not
end
up
blocking
I
think
it's
going
to
be
Verge
soon,
so
I'm
sort
of
like
I
kind
of
want
to
wait
for
before
we
decide
anything
because
I
kind
of
think
it's
going
to
resolve
itself
because
it's
very
far
along
and
I
think
it's
going
to
probably
emerge
before
we
decide
on
when
we're
going
to
release
in
our
efficient
release.
So
I
think
we're
going
to
have
our
answer
anyway,
regardless
of
whether
we
call
it
a
block
or
not.
B
B
B
B
Go
and
but
go
get
was
working
at
one
point:
wasn't
it
I,
don't
remember
what
the
command
is.
Anyone
on
the
call
remember
what
the
go
get
command
is
to
pull
in
the
vert
CTL
and
build
it
locally
on
arm.
B
I'm
going
to
say
about
one
and
insert
here
in
a
bit
we're
definitely
missing
that
overall.