►
Description
Meeting notes https://docs.google.com/document/d/1ushaVqAKYnZ2VN_aa3GyKlS4kEd6bSug13xaXOakAQI/edit#
A
All
right
welcome.
Everyone
today
is
wednesday,
the
11th
of
may
2022-
and
this
is
the
cluster
api
project.
Meeting
cluster
api
follows
the
kubernetes
sig
community
guidelines,
which
basically
means,
if
you'd
like
to
say
something.
Please
raise
your
hand
and
in
general,
please
try
to
be
kind
and
treat
others
as
you
would
expect
to
be
treated
so
with
that.
Let's
get
started
here
so
give
a
few
minutes
to
welcome
new
attendees.
A
If
anyone
would
like
to
introduce
themselves
or
is
new
to
the
community,
please
just
feel
free
to
unmute
your
mic
and
say
hello
to
everybody.
A
All
right,
I
am
not
seeing
any
unmutes,
so
I
guess
it
may
be-
maybe
next
time
so
let's
go
to
the
open
proposal
readouts
here
and
see.
Does
anyone
have
updates
that
they
would
like
to
deliver
on
any
of
the
open
proposals?
Yeah
matt
go
ahead.
B
Thanks
mike
I'm,
basically,
I
know
things
were
a
little
slow
last
week
because
vmware
was
kind
of
off
site
or
whatever
you're
doing,
and
but
I'm
wondering
if
machine
pull
machines,
if
we
could
start
lazy
consensus.
B
At
this
point
we
had
a
couple
lgtms
in
the
past
and
I
think
I've
addressed
outstanding
stuff,
so
just
kind
of
putting
that
out
there
if
we
could
get
if
we
think
we're
there
if
we
think
we're
still
waiting
for
changes.
A
A
It's
all
good
vince
go
ahead.
I.
D
Was
gonna
say
like
there's
a
bunch
of
proposals
out
that
have
been
kind
of
lingering
like
without
comments
for
a
while,
so
we
could
also
just
give
like
us
like
a
week,
even
though
next
week
is
cubecon,
I'm
sure
that
you
know
we
have
some
time
this
week,
where
we
can
wrap
it
up
this
week
or
at
the
end
of
next
week,
just
two
weeks.
You
know
we're
almost
going
to
june,
so.
A
A
E
E
Oh,
that
link
is
we
have
a
pr
I'll
update
the
link
thanks.
A
Stefan,
so
I
guess,
is
there
any
objection
to
starting
the
same
lazy
consensus
on
these
proposals
as
well.
A
F
No,
I
I
generally
agree
to
unblock
this.
This
block
just
another
point
is
that,
given
that
we
are
targeting
the
first
fc
data
for
one
to
two
in
the
middle
of
the
next
month,
it
is
kind
of
an
hour
or
or
never
for
this
release
so
yeah.
Let's
keep
all
this
in
mind.
A
Yeah,
that's
a
really
good
point,
okay,
so
the
the
only
thing
is
someone
you
know
we
just
have
to
make
sure
to
to
get
these
merge
next
week.
Then,
if
there
are
no
points
that
come
up
between
now
and
then
all
right,
anyone
else
want
to
speak
on
one
of
the
open
proposals.
F
F
A
Oh
actually,
I'm
in
I'm
just
in
viewing
mode
could
someone
else
remove
that
from
the
dock?
Please
thank
you.
Okay,
so
not
seeing
any
other
hands
for
the
open
proposals.
So,
let's
move
on
to
the
discussion
topics.
Stefan,
it
looks
like
you're
up
first
with
the
unblock
merge.
E
F
Okay,
thank
you.
So
if
you
remember
some
time
ago,
we
discussed
an
issue
with
the
class
when
the
there
are
two
controllers
co-authoring
the
same
slides.
The
the
use
case
was
a
subnet
in
kappa
being
created
by
that
plate
and
then
being
completed
by
the
controller
yeah
the
issues
as
a
lot
of
documentation,
and
so
we
investigated
the
topic
discussed
a
little
bit
experimented
and
tltr.
F
Is
that
the
we
think
that
the
best
solution
to
solve
this
problem
is
to
change
the
topology
controller
in
cluster,
repeating
in
cluster
bi
in
order
to
use
server
side
apply.
These
brings
a
couple
of
other
digester.
The
the
first
and
the
most
important
one
is
that
the
copy
basically
became
consistent
with
the
rest
of
ecosystem
in
managing
such
such
situation.
Where
controller
are
co-authoring
the
same
object-
and
this
is
definitely
a
plus-
that's
that's-
make
the
behavior
consistent
people
can
look
at
the
string,
documentation
yeah.
F
F
These
are
but
yeah.
These
are,
but
also
all
the
other
solutions
that
we
explored
as
a
a
small
important
provider
as
well,
so
when
necessary.
So
when,
when
do
you
want
that,
a
slice
could
be
outlet
by
two
controller
and
controller?
Basically,
and
you
want
this
slide
to
to
behave
in
a
certain
way,
it
is
necessary
that
you
go
in
to
into
your
fei
types
and
you
are
the
the
merker,
the
mark,
the
golang
mark
list
type
equal
map
and
and
and
the
key.
F
So
I
don't
think
this
requires
course
a
new
release
from
the
provider,
but
at
the
same
time
adding
this
kind
of
annotation
is
getting
more
and
more.
A
best
practice
like
like
joel
joel
was
explaining
us
yeah
in
a
recent
thread
about
how
we
do
api
in
custom
pri.
So
I
want
just
to
make
everyone
aware
that
we
are
going
down
this
path
provider
if
want
the
topology
controller,
behaves
properly,
have
to
start
looking
at
these.
F
As
far
as
I
know,
there
are
two
use
case,
one
in
kappa
and
one
in
kappa
z,
but
the
impact
is
pretty
limited
and
I
think
the
change
makes
sense
just
to
complete
the
picture
yeah.
There
is
some
work
to
make
this
happen.
There
is
also
a
small
downside.
Is
that.
F
We
cannot
use
a
server
side
applying
in
topology
dry
run,
because
the
policy
dry
run
does
not
use
a
real
the
api
server
and-
and
so
the
idea
for
now
is
that
to
preserve
topology
that
iran
in
the
in
the
current
form,
even
if,
with
some
caveats
limitation,
that
we
are
going
to
document
and
then
research,
a
better
long-term
solution,
we've
take
some
time
to
do
it,
and
the
last
caveat
is
that
we
still
have
to
understand
what
happens
to
cluster
created
before
using
server
side
apply.
F
They
use
the
same
identity,
they
they
all
qualify
as
a
manager.
So
it
is
really
hard
to
understand
what
change
were
done
by
copy
and
what,
by
the
controller.
So
we
have
to
figure
it
out
this
this
detail,
but
I
think
that
the
direction
is
good
and
yeah.
I
want
to
broaden
the
topic
to
this
audience.
A
All
right,
thanks
fabrizio,
are
there
any.
I
see,
cecile's
got
her
hand
up,
go
ahead.
G
H
I
can
speak
to
that.
There
are
no
plans
of
changing
the
current
behavior
as
of
now,
so
we
will
probably
do
this
change.
Okay,
thank
you.
A
All
right,
cecile,
you
still
have
your
hand
up.
Did
you
have
another
question
or
okay,
I
guess
fabrizio?
Is
there
any
call
to
action
here
for
the
community
like
should,
should
we
take
a
look
at
this
6495
or
is
this?
Is
there
something
else
maybe.
F
I'm
I'm
working
on
pr
christian
is
that
being
it
is
an
interesting
topic.
It
was
also
for
me
a
good
opportunity
to
learn
to
dig
into
server-side
apply.
So
if
someone
is
interested
in
the
topic
feel
free
to
reach
out
and
and
yeah
or
help
us
or
just
last
question,
it
is
a
good
moment
because
they
are
working
for
it.
F
I
saw
a
question
for
cecile:
is
this
required
for
all
eyes
so
that
the
answer
is
the
default
behavior
for
crd
is
that
the
slides
are
treated
as
atomic,
that
that
means
that
either
one
controller
manage
it
or
the
other?
Okay,
you
have
to
add
the
mark,
the
marker.
Only
if
the
slide
is
do
you
plan
that
this
light
is
called
outright.
F
So
the
marker
is
required
when,
yes,
of
course,
it
is
used
also
for
service
for
kubercatol
and
and
every
operation
that
use
server
side
apply.
But
my
point
is
that
it
it
really.
The
marker
really
depends
by
the
behavior
that
you
expect
on
that
specific
slide.
Otherwise,
you
can
rely
on
the
default
behavior,
which
is
atomic.
A
All
right,
I
guess
to
that
kind
of
answer.
The
question.
G
I
think
so
just
to
clarify
it.
Your
last
comments,
if
I'm
you're
saying
that
I
would
complain,
but
it
wouldn't
be
blocking
right,
it
would
just
be
a.
E
C
C
E
A
Okay
cool,
so
I
guess
yeah
if
anyone's
interested
in
this
topic,
take
a
look
at
the
links
and
and
ask
lots
of
questions
to
fabrizio
right
all
right.
So
next
up
we
have
richard
talking
about
the
kubecon
meeting
next
week.
So
take
it
away
richard
okay,.
C
It's
just
a
quick
one
to
say
that
the
doodle
was
still
open
to
vote
for
a
convenient
time
for
next
week.
The
current
most
popular
votings
are
for
thursday
between
eight
and
nine,
so
breakfast
meetup
and
then
on
wednesday
in
the
morning
again
between
eight
nine
or
in
the
evening.
Just
to
note
that
I
didn't
actually
put
tuesday
in
there
just
in
case
anyone
was
going
to
any
of
the
co-located
events,
but
people
thought
you
know
tuesday
during
the
day
was
a
better
time.
Then
yeah
but
that'd
be
cool.
C
And
then
I
was
wondering
when
we
should
we
close
the
doodle
and
actually
agree
on
the
time.
Whether
that's
you
know
leave
the
poll
open
for
the
rest
of
the
day
and
then
decide
tomorrow,
just
be
good
to
get
your
thoughts.
A
Well,
I
guess
yeah.
We
should
probably
close
it
off
before
the
end
of
this
week,
so
people
can
kind
of
make
plans
for
next
week.
Does
anybody
have
any
thoughts
about
this,
or
is
there
any
objection
to
closing
the
poll
tomorrow?
I
guess.
A
C
Yeah
yeah
we'll
do
and
then
I'll
reply
on
the
the
threads
that
we
already
have
going
in
in
the
channel
as
well.
A
Awesome
and
yeah
it'll
be
great
to
see
everybody
there
fabrizio
you've
got
your
hand
up.
F
Just
for
thank
you
for
thanking
richa
for
trying
these
for
the,
let
me
say,
the
the
the
informal
gathering
by
the
way,
I
will
be
there
both
of
the
contributors
summing
and
and
the
kubacon.
If
someone
want
to
catch
up,
just
send
me
a
messages.
Luck.
A
Cool
sounds
good.
I
will
also
be
there
for
the
contributor
summit,
so
I'll
be
looking
for
you
for
breezy.
A
All
right,
thank
you.
Richard
alberto
you've
got
the
next
topic
with
in
place.
Node
upgrade
so
take
it
away.
I
I
I
share
some
thoughts
there,
and
this
is
also
kind
of
a
recurring
topic
for
permanent
environments.
So
yeah
some
people
reach
out
to
me,
kate,
goldberg
from
microsoft,
and
some
other
folks
show
interest.
So
I
went
ahead
and
put
this
draft
in
place
to
have
a
kind
of
a
starting
point
to
to
about
about
this
this
topic
and
try
to
move
forward
with
it.
It's
pretty
basic
at
the
moment,
just
sharing
some
of
the
stuff
that
we
are
doing
downstream,
but
yeah.
I
Hopefully
we
can
use
it
again
as
a
starting
point
to
to
work
on
a
solution
together.
Main
idea
would
be
to
have
a
dock
operator
or
controller
for
this
to
manage
this.
That
should
be
able
to
work
orthogonally
to
copy
and
eventually
have
some
kind
of
integration
point
between
both
so
yeah.
That's
the
the
idea,
so
please
feel
free
to
chime
in
that
talk
or
give
me
on
slack.
If
you
are
interested
in
this
topic,.
F
I'm
definitely
interested
in
this
token
and
yeah.
We
last
week
we
had
the
chance
to
meet
together
and
yeah.
This
was
on
one
of
the
topic
that
we
would
like
to
work
as
well.
So
after
kubercon,
I
will
definitely
jump
in
and
take
a
look.
Thank
you
for
starting
this
effort.
A
Okay,
I'm
not
seeing
any
hands
coming
up.
So
thanks
thanks
alberto-
and
I
guess
if
folks
are
interested,
take
a
look
at
the
draft
here.
A
So
next
we
have
jonathan
with
a
link
to
cappy
add-on
management
meeting.
So
jonathan,
why
don't
you
take
it
away.
J
Yeah,
so
I
just
wanted
to
give
a
quick
reminder:
I'm
working
on
a
cappy
add-on,
orchestration
proposal
with
fabrizio
and
jack
francis.
We
had
a
sync
up
meeting
a
couple
weeks
ago,
where
we
gave
a
demo
of
a
proof
of
concept
and
we
showed
using
a
controller
to
install
helm
charts.
So
I
just
wanted
to
remind
everyone
about
that
and
drop
a
link
with
the
recording,
and
you
guys
can
expect
the
demo
at
office
hours
after
coupon.
A
We've
gotten
a
very
ecstatic
awesome
from
vincent
chat,
so
I
I
suck
at
that
thanks
jonathan,
so
I
guess
now
we'll
move
into
the
provider
updates
section
got
the
I've
got
a
update
from
kubemark.
We
just
released
version
0.3.0
today,
which
includes
support
for
1.0
plus
of
cluster
api,
so
that
was
a
big,
a
big
thing
for
us,
and
it
also
includes
a
reference
implementation
of
the
of
the
scale
from
xero
for
the
cluster
auto
scaler,
which
is
currently
on
review
in
the
auto
scaler
richard.
C
Yeah
this
just
an
update
on
the
demo
we
gave-
I
don't
know,
probably
about
a
month
ago
or
two
months
ago
about
the
the
micro
vm
provider.
Now
it's
now
actually
public.
It
was
private
originally,
so
you
can
take
a
look.
There
is
improvement
needed
on
the
the
the
end-to-end
testing
side
of
things
so
but
yeah
if
anyone's
interested
in
micro,
vms,
specifically
firecracker
and
cloud
hypervisor,
feel
free
to
have
a
play
with
it.
A
Very
cool
awesome
to
see
that
hitting
the
open
source
thanks
richard
and
I
don't
know
who
put
in
the
capsi
update
here,
would
it
does
anyone
want
to
speak
on
this,
or
should
I
just
read
it
out.
G
A
Awesome,
thank
you
cecile,
and
that
is
I
really
like
your
release
notes
here.
They
look.
They
look
really
nice.
Oh.
G
We're
actually
using
the
kubernetes
release,
notes
so
like
the
same
release.
Note
tool
as
kubernetes
itself.
A
All
right,
so
that's
kind
of
the
end
of
our
agenda.
Does
anyone
have
any
ad
hoc
topics
they
like
to
they'd
like
to
bring
up
or
are
there
any
new
attendees
who
missed
the
intro
at
the
beginning?
I
would
like
to
say
hi.
K
Yes,
that's
me
actually,
my
name
is
george
malika
and
I
work
with
microsoft,
I'm
actually
on
kate's
team
and
it's
actually
my
first
time.
So
everything
is
very
new
for
me,
but
thank
you
guys
for
including
me.
A
I
see
yuri
is
unneeded
unmuted
as
mike,
but
we're
not
hearing
you
yuri.
Oh.
L
Yeah
darn,
it
double
muted
button,
hello
folks,
also
at
microsoft,
cherry
apple
been
working
with
george
and
kate
and
cecile
and
so
nice
to
meet
you
and
and
and
just
also
very
interested
in
being
place
update.
I
didn't
raise
my
hand
earlier,
but
something
I'll
try
to
participate
as
well.
A
All
right,
I
am
not
seeing
any
other
mics
unmuting,
so
I
guess
that
brings
us
to
the
end
of
our
meeting
for
this
week.
Thanks
everybody,
there
will
be
no
meeting
next
week,
as
will
some
of
us
will
be
at
kubecon.
So
I
guess
we'll
see
you
in
two
weeks.