►
From YouTube: KubeVirt community meeting 2021-03-03
Description
KubeVirt community meeting 2021-03-03
A
Hello,
everybody
and
welcome
to
the
coobert
weekly
meeting,
I'm
your
host
chris
caligari
and
let's
begin,
I
posted
the
cooper
weekly
meeting
agenda
document
into
chat
for
those
that
don't
have
the
link
already
saved
and,
of
course
you
need
to
be
a
member
of
the
cooper
dev
mailing
list
in
order
to
edit
it.
A
Okay,
for
starters,
welcome
everyone
and
I
added
a
new,
a
new
bullet
point
for
new
members
to
introduce
themselves
if
they
would
like.
A
Okay,
nobody
knew
this
week,
then
we
will
move
on.
So
I
I
have
the
first
agenda.
No
bullet
point
here
just
wanted
to
give
everybody
a
friendly
reminder
about
the
github
pages
settings
on
repos
and
cooper.
A
B
Okay,
I
should
be
seeing
my
firefox
one
down
with
the
meeting
minutes,
so
this
is
a
continuation
of
last
week.
We'll
take
a
few
minutes
to
continue
the
discussion.
The
little
bullet
point
out
of
this,
so
I
added
more
details
and
to
the
design
dock
and
for
the
bulk
api
discussion
and
the
design
doc
has
linked
in
there
all
right
so
I'll
switch
over
to
it.
So
I'm
not
going
to
go
through
all
the
details.
I
filled
out
a
number
of
things.
B
I
kind
of
I
attempted
to
mock
just
some
of
the
apis
just
to
like
and
then
put
some
examples
in
just
to
kind
of
get
an
idea,
but
if
folks
can
read
through
that
in
their
own
time
and
I'll
probably
what
I'll
do
is
I'll
send
this
out
in
the
mailing
list
too.
Just
so
it
gets
far
and
wide
and
we
can
get
some
feedback
that
way.
So
mainly
what
I
want
to
discuss
is
that
I,
I
briefly
touched
on
last
time
kind
of
the
implementation
phases.
B
I
guess
if
you
will
so
like
the
I
could
I
haven't
listed
here
in
the
high
level
features
and
have
six
total,
so
two
changes
that
I
made
since
last
time
was.
I
added
this
called
unique,
vm
metadata,
so
this
wasn't
here
before,
and
so
the
idea
here
that
if
I
want
to
create
100
virtual
machines
and
then
I
want
to
add
unique
metadata
to
each
of
them
and
the
idea
is
kind
of
along
the
lines
with
like.
B
I
guess
the
example
would
be
that
if
I
had
a
a
secret
that
I
wanted
to
get
unique
secret,
I
wanted
to
be
distributed
to
each
of
the
virtual
machines
and
I
kind
of
got
the
some
the
like
the
inspiration
for
that
when
I
looked
at
this
from
looking
at
cloud
internet
and
so
I'll
go
to
the
docs
for
kubert.
B
I
was
looking
at
the
reading
through
the
content
section
and
there's
this
section
at
the
bottom
called
future
metadata.
Server
based
data
sources
and
kind
of
that.
I
I
might
understand
kind
of
the
problem
here
is
that
you
have
cloud
in
it
that,
and
you
have
a
lot
of
vms
and
you
want
to
sort
of
dynamically
populate
the
metadata
of
a
cloud
in
it
for
all
these
virtual
machines.
B
So
you
have
a
server,
that's
going
to
return
the
information
so
that
you
can
configure
how
these
virtual
machines
run
so
kind
of
like
a
similar
problem
space
to
what
I'm
getting
at
in
that
so
instead
of
having
a
server,
I
could
essentially
have
in
this
case,
like
I
could
sort
of
up
front
say
like
I
want
unique
secrets
or
something
I
can
just
reference
them
and
then
those
get
populated
to
a
large
bulk
number
of
virtual
machines
during
creation
so
similar
problem
that
I'm
kind
of
looking
at
tackling.
B
So
that's
kind
of
the
the
example
and
that's
the
only
one.
I
reference
in
the
the
section
that
they
talk
about
it,
but
I'll
go
to
that
in
a
second.
The
second
one
is
detach
so
last
time
somebody
become
about
to
detach-
and
I
think
it
actually
kind
of
makes
sense
to
be
in
its
own
phase,
which
I
think
it's
sort
of
something
that
has
a
lot
of
components
to
it.
B
So
I
separate
it
out,
and
each
of
these
has
their
own
section,
where
I
go
into
details
on
on
each
of
them,
but
I
want
to
focus
just
on
the
unique
meta
via
mediative
portion
and
kind
of
feel
questions,
because
I
list
this,
oh
and-
and
I
should
mention
that
so
this
is
kind
of
an
order
how
at
least
I
at
least
from
my
use
case,
how
I
see
them
in
terms
of
priority.
B
So
I'll
move
down
here.
So
before
I
talk
more
about
this.
C
I
get
what
you're
wanting
to
do.
I'm
curious,
if,
like
what's
the
tangible
use
case
here,
is
it?
Is
it
really
for
cloud
in
it?
Is
that
what
you're
wanting
to
use
it
for?
Is
it
the
user
data
portion
of
cloud.net?
You
need
a
unique
user
data
for
each
vm.
What
is
it
yeah.
B
Yeah,
so
it
would
be
providing
like
a
token
or
something
or
whatever
to
as
part
of
the
user
data
section
of
cloud.
C
And,
what's
being
passed
into
that
user
data,
like
I
don't
need,
specifics
is
more
of
it?
Is
it?
Is
it
something
that's
application
specific
that's
being
passed
in
or
is
it
something
more
generic
like
ssh,
keys
or
or
things
like
that
that
need
to
be
unique.
C
B
C
I'm
trying
to
understand
the
workflow
of
how
the
unique
metadata
would
be
generated.
How
would
you
hook
in
to
that
to
give
each
virtual
machine
as
it's
as
it's
being
created
the
unique
bits
something
has
to
generate
right.
B
Yeah,
so
so
this
is
where
I
want
to
go
at
least
with
this
is
that
is
that
the
assumption
would
be
an
assumption
would
be
made
that
since
you're,
since
there's
a
there's
a
template,
a
virtual
machine
there's
will
be
a
reference
to
a
secret
and
the
way
that
I
kind
of
do
it
is.
I
have
a
label
here
so
in
case
any
secret
with
a
label.
B
Will
that
has
this
label
will
status
as
a
that
will
be
passed
on
to
these
virtual
machines?
B
So
I'm
like
exercising
strict
control
over
it
could
be
the
exact
amount
like
100
secrets,
or
there
could
be
just
10
and
and
just
my
phone
would
just
kind
of
like
cycle
through
them,
and
I
mean
that
would
be
error.
So
I
I
think
it
to
me
when
I'm
getting
the
creation
or
of
any
sort
of
the
unique
metadata
would
be
an
external
step.
B
No,
it
would
exist,
they
would
exist.
That's
what
I'm
saying
yeah
like
they
crits
would
have
to
be
there.
This,
as
that
would
be
the
assumption.
You
have
provided
a
c
you're,
providing
you're
referencing
a
label.
B
The
people
would
go
out
and
take
a
look
at
the
secrets
that
match
this
label
and
it
will
just
distribute
whatever
it
finds
to
the
to.
However,
vm's
that
you
want.
C
So
it's
just
taking
like
a
round
robin
approach
to
the
ones
that
finds
the
secrets
it
finds
that
matches
label.
It's
assigning
one
of
them
out
of
this
selection
of
secrets
that
match
the
label
or.
B
Yeah
yeah
yeah,
so
it
would
be
well.
I
would
think
it
would
be
like
a
robin
robin.
I.
I
think
we
don't
want
to
do
random
just
because
we
want
uniqueness,
so
we
would
just
kind
of
loop
through
the
list
and
and
just
keep
allocating
them
until
we
run
out
of
secrets
to
allocate
them.
We
just
keep
going.
If
there's
still
more
virtual
machines,
then
do
the
loop
again.
B
Does
you?
What
do
you
think
about
the
I
guess,
the
the
concept
of
sort
of
common?
I
call
it
kind
of
like
I
call
here
common
overrides
or
something
like
distribute
data.
Is
this
something
that
you
could
be
a
common
pattern?
I
mean
you
mentioned.
Ssh
keys
sounds
like
that,
might
also
be
an
area
that
this
could
be
used.
It's
not
something.
I
specifically
have
an
example
for,
but.
C
Yeah,
so
we
distribute
ssh
keys
via
secrets
as
well,
but
I'm
not
aware
of
a
use
case
where
you'd
want
like
a
pool
that
has
each
maybe
there's
a
use
case
for
a
pool,
and
you
want
a
unique
ssh
key,
that's
stored
in
a
secret.
I
I'm
unsure.
I
was
trying
some
more
piece
apart.
C
What
the
problem
was
that
you're
trying
to
solve
whether
it's
application,
specific
something
that's
unique
to
your
use
case
or
if
it's
something
that
we
can
tackle
more
generically
and
it
does
sound
like
it's
totally
application
specific
yeah.
I've
never
even
encountered
this
pattern
before
it's
really
interesting,
I'm
not
sure
what
to
think
of
it.
Yet
I'd
have
to
okay
yeah,
even.
A
I
agree
with
you
the
ssh
host
keys,
whereas
my
instantly,
my
first
thought
I'm
seeing
this
this
api
bit
and
it's
a
huge
problem
and
and
scaled
out
infrastructures
as
easy
as
it
sounds.
A
B
So
how
would
you,
how
would
you
maybe
talk
a
little
bit
more
about
that
like?
How
would
you
just
if
you're,
if
you're,
launching
a
lot
of
virtual
machines,
how
would
you
normally
distribute
the
authorizations
for
set
clients.
A
A
I
could
give
examples,
but
I
don't
think
I'm
allowed
to,
but
for
for
ssh
keys,
it's
every
every
infrastructure
that
has
more
than
a
handful
of
virtual
machines
goes
through
the
goes
through
this
problem.
A
That's
not
a
very
secure
method
of
handling
keys
and
they
were
looking
at
a
hsa
product
to
to
secure
that
which
has
its
own
set
of
problems
to
deal
with.
B
Okay,
so
I
guess
like
at
a
high
level
like
what
to
kind
of
like
try
and
capture,
because
I
kind
of
I
maybe
I
can
do
it
with
like
a
little
bit
of
a
contrast.
B
So
we
in
this
with
this,
I
guess
feature-
I
talk
about
different
vm
profiles,
so,
like
kind
of
overriding
things
like
cpu
mem
disk
there
being
sort
of
uniqueness
between
virtual
machines,
everything
is
going
to
sort
of
have
its
own
override
in
some
way
and
the
contrast,
this
is
the
way
to
have
sort
of
a
something
that
will
be
added
everywhere,
sort
of
an
override,
that's
added
everywhere.
B
So
it's
essentially
like
a
way
of
of
making
it
so
that
I
don't
have
to
say
say
have
like.
Let's
say
I
had
100
virtual
machines
and
this
in
the
different
vm.
I
wanted
10
with
a
certain
skew
amount
90
with
another.
This
case
would
be.
I
would
have
to
have
100
different
overrides,
because
each
one
is
going
to
be
unique.
It's
sort
of
a
way
to
like
pull
all
that.
A
From
from
a
hierarchy
perspective,
I
don't
think
that
would
be
too
bad
of
a
of
a
thing
to
have
if,
if
you've
ever
dealt
with
puppet,
for
instance
their
what
is
their
their
database
called.
A
I
forget
now,
since
it's
been
a
few
years,
they
have
a
data
source
behind
the
the
web,
server,
a
puppet
that
that
handles
all
of
these
unique
bits
of
data
and
then
passes
them
out
to
clients
and
it's
just
a
key
value
file,
and
you
end
up
with
a
hundred
entries
for
ssh
keys,
and
it's
really
not
bad
at
all.
B
Yeah
I
mean
I
get
like
part
of
part
of
like
this
as
I
can.
I
can
see
this
this
kind
of
feature
quickly
becoming
like
its
own
dsl,
almost
and
I'm
afraid
of
that
and,
like
I.
D
B
I
don't
want
to
like
like
because,
because
to
me
once
they
start
asking
users
to
like,
say
we're
to
ask
user
to
write
all
these
unique
things
you
know
now
you
know
what's
to
stop
someone
to
say
like.
Oh
I,
why
can't
you
support
this
too
and
it's
kind
of
override
and
and
like
I
don't
know,
I'm
trying
to
like
make
like
very
clear
barriers
so
that
it's
there
isn't
like
sort
of
a
language
to
define
like
it's
just
like
okay,
this
is
all
or
nothing,
here's
what
I
want
to
do.
B
If
I
need
to
do
anything
specific
changes
underneath
you
know,
so
there
isn't
sort
of
any
like
we're
just
sort
of
tweaking
bits.
Instead
of
having
like
too
much
yaml
being
thrown
around.
You
know,
I'm
just
thinking
like
yeah,
I'm
sorry
go
ahead.
Sorry.
C
I'm
thinking
about
this
a
little
further
and
I'm
unsure.
So
what
I
would
typically
see
is,
if
you're
wanting
to
distribute
something
unique
to
a
virtual
machine.
What
I've
seen
a
pattern.
I've
seen
in
the
past
is
that
every
virtual
machine
gets
the
exact
same
cloud
in
it
and
they
would
reach
out
to
a
service
on
first
boot
to
retrieve
whatever
unique
token
or
whatever
they
need.
C
So
I'm
nervous
about
using
I'm
apprehensive
it's
about
using
the
infrastructure
like
the
kubernetes
infrastructure,
to
support
this
when
it
might
be
something
that
needs
a
service
specific
to
your
application.
I'm
I'm
kind
of
torn
between
that.
D
I
just
wanted
to
say
that
I
mean
this
is
exactly
what
I
wanted
to
to
mention
here,
because
you
are
talking
about
application,
specific
and
anything
generic,
but
we
are
managing
in
the
project
in
flux,
infrastructure
of
virtual
machines.
We
are
not
trying
to
distribute
things
inside
the
vm,
I
mean
unless
it's
something
very
related
to
the
infrastructure.
So
so
there
are
tools
to
do
that,
which
you
can
you
can
use
it
like
what
david
just
said.
B
Yeah
I
mean,
I
guess
part
of
it
is
because
kubernetes,
because
we
can
store
things
that
are
specific
in
secrets
and
we
can
distribute
them
that
way.
That's
how
I
looked
at
it
as
solving
as
something
that
we
could
do
it
using
the
infrastructure,
but
I
mean
like.
C
You
can
have
the
virtual
machine
even
talk
to
the
api.
I
mean.
Maybe
that
crosses
a
boundary.
I'm
just
gonna,
throw
it
out
for
your
use
case.
You
can
have
the
virtual
machine
actually
talk
to
the
api
server
retrieve
a
random
secret.
B
Yeah
yeah,
I
mean.
E
A
So
would
you
guys
mind
if
you
took
the
debate
over
to
the
mailing
list,
so
we
can
move
on
with
this
meeting
yeah?
I
think
this
is
a
very.
C
Specific
topic
ryan:
would
you
mind
making
a
mailing
list
post
just
about
the
unique
metadata,
and
then
I
think,
I'm
probably
going
to
follow
up
with
trying
to
understand
whether
this
can
be
solved
reasonably
with
an
external
service
or
if
this
is
something
generic
that
belongs
in
cubert.
So
since
it's
so
unique,
that's
like
the
thing
that
signaled
to
me
that
maybe
it
doesn't
belong
in
kubert,
but
I'm
trying
to
explore
it
a
little
bit
further
and
make
sure
them
right.
A
Feature
sounds
really
neat.
I
can't
wait
to
see
it
get
developed.
A
A
Okay,
silence
is
golden.
Do
we
have
any
poll
requests
that
are
worthy
of
discussion.
F
A
F
Let
me
share
the
screen,
can
you
see
it.
F
All
right,
so
the
the
first
one
is
from
a
watermelon
brother
to
have
support
for
hotblock
network
interfaces,
cpu
and
memory.
I
mean
it's
a
huge
list
of
huge
features,
but
I
I
think
we
want
to
have
support
for
all
of
these
eventually.
C
Right,
yeah,
it's
pretty
so,
but
I've
heard
that
there's
talk
of
making
pods
mutable
at
some
point
in
the
future.
Is
that
still?
Okay?
There's
no
discussion
here.
It's
just
a
request:
okay,
yeah.
F
And
he
raises
that
he
wants
to
join
the
comfort
development
team
and
how
can
we
implement
the
health
plug?
So
let's
I
mean
if
it
were
to
treat
this
as
any
other
rfe.
I
guess
the
first
step
would
be
to
split
this
right
and
propose
a
solution
on
the
mailing
list
and
then
get
the
discussion
going.
D
F
F
F
I'm
sorry
for
my
slow
typing
and
I
will
use
our
new
label.
F
Yeah
this
issue
is
daring,
but
why
not.
F
Resize
os
volume
that
was
uploaded
using
vertical
image
upload.
I
went
through
this
one.
The
person
uploaded
an
image
that
had
80
gigs
and
then
they
confirmed
it
has
80
gigabytes
and
the
vm.
Then
they
turned
the
vm
off
wrist
tries
to
resize
the
pvc,
but
it
wasn't
reflected
in
the
vm.
F
Is
this
something
we
want
to
track
under
convert,
or
is
this
more
of
a
cdi
issue?
I
have
no
clue
to
be
honest.
F
So
I
added
the
pvc
so
like.
Maybe
I
can
just
keep
cutting
edit.
C
Yep,
so
it
would
be
a
workflow
something
it
might
even
have
to
be,
since
it's
it's
a
multiple
step,
workflow,
it's
increasing
both
the
pvc
size
and
then
expanding
the
image
inside
of
it,
which
can
take
different
forms.
That
seems
like
it
belongs
in
cdi.
It
would
be
some
sort
of
you
declare.
You
want
to
expand
this
volume
and
a
cdi
controller
would
would
then
act
on
that
begin.
I
I'm
spitballing,
I'm
making
that
up
as
I
go.
F
F
F
F
Yeah,
it
belongs
to
cdi.
F
This
is
so
lazy,
good
issues
solved.
F
I
don't
know
how
do
we
treat
this.
F
I
will
add
a
label
that
it
needs
information
just
to
wait
for
the
person
to
respond,
and
then
we
can
close
it.
F
Failed
due
to
an
sc
critical
network
error.
E
F
F
F
All
right,
then,
we
have
this
one.
The
ipv6
tests
fail
to
reach
out
to
an
external
ip
address.
Eddie,
do
you
want
to
comment
on
this?
One.
D
There
was
a
pr
that
expelled
it,
but
and
it's.
D
F
What's
that
attacking
so
I
mean
so,
is
there
an
action
item
on
this
one,
or
can
we
close
it?
I'm
not
sure
I
follow.
E
D
E
F
F
F
F
F
A
I
think
I
might
be
able
to
comment
on
that.
Let's
see
it
looks
very
basic.
A
A
Yes,
cloud
in
it.
F
Okay,
thanks
alex
observation:
could
you
please
confirm
that.
F
F
Do
we
have
a
logging
that
saves
the
locks
on
the
disk
in
any
of
the
components,
or
are
we
just
flashing
all
to
std
out?
Anybody
knows.
F
E
A
F
C
F
F
Okay,
let's
see
if
they
help
us
with
this
one.
F
F
So
they
used
the
snapchat
feature
and
created
pvc
had
null
as
a
value.
F
Seth,
I
wonder
how
we,
how
comes
we
haven't
catch
this
one,
but.