►
Description
A Kubernetes community meeting about the Azure provider for Cluster API. Cluster API brings familiar, declarative APIs to Kubernetes cluster creation, configuration, and management.
A
Okay,
welcome
everybody,
it
is
july
21st
2022,
and
this
is
the
cluster
api
provider
for
azure
office
hours
meeting.
We
have
this
meeting
every
two
weeks
on
thursdays,
cluster,
api
and
cluster
api
api
provider.
Azure
are
both
parts
of
sig
cluster
lifecycle
under
the
kubernetes
umbrella
and
as
such
we
comply
with
general
standards
of
conduct.
You
can
read
about
them,
there's
a
link
in
the
doc,
but
basically
it
boils
down
to
everybody
be
kind
to
each
other
and
don't
try
not
to
interrupt.
Try
to
use
the
raised
hand
feature.
A
That
I
think
that's
about
it
here
is
where
we
provide
a
moment
for
new
members
or
attendees
to
say
hello.
I
don't
think
we
have
anyone
who's
new
today,
so
we
can
skip
over
that
unless
somebody
has
something
to
volunteer,
if
you
want
to
add
your
name
to
the
little
attendee
list.
Sorry,
what's
that.
A
If
you
want
to
add
your
name
to
the
attendee
list,
that
would
be
nice.
Otherwise,
let's
get
going.
We
don't
have
much
of
an
agenda
today.
So
probably
this
will
be
short,
but
that
also
means
we
could
just
talk
about
whatever
we're
interested
in
so
first
we
have
a
psa
about
the
release.
Do
you
want
to
do
that?
Cecile.
B
Sure
do
you
want
to
share
your
screen
with
the
agenda,
so
we
can
all
see
it
thanks,
yeah,
so
just
a
quick
announcement
that
since
last
time
we
had
this
office
hours
right
after
we
met,
we
released
1.4.0.
B
So
this
is
a
pretty
significant
release
with
lots
of
features.
A
couple
of
bug
fixes
mostly
focused
around
testing
and
improving
our
integrations
with
testing
of
kubernetes
on
azure
in
pro
so
yeah
check
it
out
reach
out.
If
you
have
any
feedback
or
questions,
and
then
I
guess
next
release
is
plan
four,
so
we
opened
the
new
milestone.
Next
release
is
fun
for
first
week
of
september.
I
believe
so
that's
we're
trying
to
do
this
two-month
cadence
for
now,
we'll
see
if
this
is
sustainable
over
time.
B
So
far,
it's
working
pretty
well,
but
yeah
I'll
actually
be
out
that
week,
but
other
folks
will
be
around
to
take
care
of
everything
around
it.
So
yeah.
A
Nope,
next
up
dane,
do
you
want
to
give
us
an
update
on
configurable
network
interfaces.
C
C
Really
quickly,
we
submitted
that
pr.
It
was
a
pretty
big
pr,
and
I
wanted
to
thank
everyone
because
we
got
some
fantastic
feedback
from
lots
of
people
got
a
lot
of
eyes
on
that
which
we
really
appreciated,
and
we
had
a
first
time
contributor
authoring
most
of
that
pr.
So
that
was
a
a
really
really
great
opportunity
and
there's
the
current
feedback
on
it
is
to
break
it
up
into
two
which
we
are
going
to
work
on.
C
We
got
stalled
because
we
got
pulled
into
some
other
things
for
a
couple
weeks
here,
but
we
should
be
getting
back
to
it.
I
think
in
the
next
week
or
two
and
then
we're
going
to
split
it
up
as
requested
and
try
and
address
all
that
feedback.
B
Yeah,
I
just
want
to
say
thank
you
for
opening
the
pr
like
it's
something
that
I
know
like
several
people
ask
for
in
one
way
or
another,
and
I
think
it's
going
to
unlock
several
use
cases,
so
I'm
excited
to
get
it
in.
I
think
most
of
my
comments
were
around
the
ux
aspect
of
it.
There
is
some
trickiness
around
the
api
just
because
we're
adding
new
fields
which
are
kind
of
already
covered
by
existing
fields,
so
yeah
just
some
some
stuff
to
deal
with
around
that,
but
otherwise,
like
yeah.
B
I
think,
let's,
let's
work
together
to
get
it
in
as
soon
as
possible.
C
A
Cool
anything
else
to
say
about
that.
A
Looks
like
we
have
already
run
off
the
end
of
our
agenda,
so
is
it
worth
doing
a
milestone
review
or
it's
not
too
far
out
at
this
point,
it's
a
seal.
B
Actually,
I
think
if
we
have
time
it
might
be
useful
to
do
like
a
little
milestone
triage
just
because
last
time
we
met
it
was
before
we
released
1.4,
and
so
now
this
is
like
the
first
time
so
right
now
we
have
a
1.5
milestone,
which
just
has
basically
the
stuff
that
we
carried
over
and
maybe
a
few
additional
things,
but
it
might
be
good
to
just
look
at
you
know
what
are
our
current
aims
for
1.5?
B
A
A
Flat
car
support
did
ria
really
open
this.
Oh
my
that's
a
lot,
that's
kind
of
old,
but
yeah,
I'm
pretty
sure,
they're
converging
on
this.
We're
very
close.
I
think
out
of
tree
seems
like
it's
very
close
as
well
right.
B
So
out
of
tree
is
currently
on
pause,
I
would
say
instead
of
like
in
in
in
terms
of
defaulting
it
we
have
like
the
support.
We've
recently
switched
it
to
home,
but
I
think
we're
waiting
for
better
integration
with
our
cluster
api
add-ons
to
be
able
to
install
it
by
default
so
that
the
ux
of
like
creating
a
cluster
isn't
broken
or
it
doesn't
regress
by
integrating
it.
B
I
don't
know
if
it's
worth
moving
out
of
the
milestone,
since
I
don't
think
we're
planning
on
getting
this
done
in
the
next.
I
would
say
this
is
probably
a
good
one
for
the
next
milestone
like
1.6.
B
A
Black's
user
interface
is
obscuring
my
other
controls.
Okay,
make
services
async.
B
So
jonathan
is
currently
working
on
tags
and
agent
pools
the
first
one.
So
after
that,
the
only
two
remaining
are
vmss
skill
sets,
which
we
haven't
touched
kind
of
intentionally
so
far,
because
it's
already
doing
async
but
in
slightly
different
ways.
So
it's
gonna
be
a
pretty
like
it's
a
refactor
basically,
but
it's
not
gonna
change
any
functionality
and
there's
also
a
lot
of
moving
pieces
around
machine
pool.
So
we
didn't
wanna
get
in
the
middle
of
that.
Basically,
and
it's
not
super
high
priority
david,
do
you
have
any
thoughts
on
that?
E
So
maybe
maybe
I
can
carve
out
some
time
in
august
to
take
a
look
at
this
or
maybe
next
week.
I
think
we're
gonna
get.
I
think,
we'll
have
a
little
bit
of
time
to
take
a
look
and
maybe
take
a
first
swipe
at
it.
I
I
wouldn't.
E
I
don't
know
since
like
what
you
said.
It's
already
async,
so
it's
just.
It
would
be
really
nice
to
get
the
code
cleaned
up
to
follow
the
same
patterns
as
everywhere
else
yeah.
Let
me
I,
I
probably
have
the
most
context.
There
it'll
be
a
fun
thing
to
get
into
and
refactor
it
back
to
the
pattern
that
we
used.
B
E
You
know
what
I
didn't
think
about
that.
That's,
let's
not
do
that
until
that
gets
pulled
in
and
then
we
just
do
a
grand
refactor.
A
Okay,
so
tags,
I
guess
tags
on
its
own
stays
in.
A
A
A
Async
tags:
we
just
talked
about
document
release,
cadence
and
support
policy.
Jack
signed
up
for
this,
but
I'm
happy
to
take
it
on
if,
if
he's
busy,
when
he
gets
back
tomorrow,.
B
A
There
configurable
network
interfaces.
We
just
talked
about
thanks
team.
A
This
is
interesting.
I
saw
that
cecile
you
mentioned
this
and
I
was
taking
a
look
at
it.
It
should
be
fixed
soon.
B
A
D
B
Yeah
that
one
I
also
added,
I
think
I
just
fell
through
the
correct
question,
actually
I'll
paint
jack
on
it
and
see
when
he
comes
back
and
see
if
he's
looking
at
it,
cool.
E
Hey
thanks
so
lachlan's
pr
looking
at
it
really
it
was.
It
was
a
pretty
lightweight
pr.
The
only
thing
that
we're
asking
for
is
to
get
some
conditions
in
there
to
or
maybe
even
just
one
condition
to
be
able
to
show,
what's
actually
happening.
So
there's
nothing
that
really
gives
user
feedback
about.
E
Why
this
thing,
isn't
you
know
using
the
desired
state,
so
the
desired
state
is
like
imagine.
You've
requested
10
10
workers
in
the
pool
and
you've
also
turned
on
auto
scaler
for
aks.
We
don't
have
control
of
the
number
of
nodes
there
anymore,
so
we
we
just
don't
try
to
reconcile
the
desired
state
for
the
number
of
nodes,
and
I'm
sure
dane,
has
some
strong
opinions
about
this
one
too.
The
external
scalar
right.
E
You
have
opted
into
external
scaling
or
platform
scaling
and
we
aren't
going
to
set
your
desired
state
here
for
the
number
of
replicas
we're
going
to
allow
the
platform
to
take
advantage
of
it,
and
I'm
not
sure
this
is
the
best
way
to
do
it.
But
it's
definitely
a
way
around.
Some
of
the
pain
that
that
they're
facing,
when
you
know
you
get
the
thrashing
between
the
external
scalar
and
desired
state,
dane
what's
up.
C
C
I've
struggled
a
little
bit
with
how
controversial
the
feature
was
in
cluster
api
itself,
because
it
seemed
like
that
was
a
good
place
to
surface
some
of
that
feedback,
keeping
the
replicas
count
up
to
date,
and
things
like
that.
C
But
that
said
like,
as
I
thought
through,
I
think
one
of
the
alternate
proposed
solutions
which
seemed
somewhat
reasonable
might
be
just
to
use
an
annotation
or
a
spec
field
in
in
the
actual
infrastructure
machine
pool
resources
that
that
tells
it
to
essentially
reverse
that
feedback
loop
and
and
write
to
specta
replicas
in
the
machine
pool
not
much
differently
than
say
cluster,
auto
scaler
with
the
cappy
provider
would
do
it's
just
a
proposal,
and
you
know
it.
C
E
Yeah,
I
I
didn't
think
about
turning
on
the
x,
the
aks,
the
platform
auto
scaler,
and
we
really
have
very
little
control
over
how
that's
reacting,
and
I
I
think
it
raises
the
importance
of
of
modeling
this
appropriately,
so
that
folks
are
pretty.
C
E
What's
going
on
yeah
it
when
I,
when
I
saw
this,
I
was
like
oh
great
catch
and
I
just
thought
of
you
dane.
C
It's
all
good
it'll
it'll
be
nice
to
if
we
can
get
that
in
that
feature
to
move
forward
a
little
bit,
it's
kind
of
I
we've
kind
of
I
haven't
had
time
to
pursue
it.
After
some
of
the
it.
It
took
a
lot
of
work
to
try
to
drive
consensus
around
it
and
a
lot
of
time,
and
it
was,
I
basically
ran
out
of
time
at
the
time
so
but
yeah.
Hopefully
we
can
get
to
a
consensus
around
some
of
that
and
then
once
we
do.
C
A
All
right
anything
else,
you
want
to
say
on
that.
You
are
that
topic.
A
Oh,
I
guess
the
next
thing
would
be
for
the
milestone.
Is
there
anything
that
really
needs
to
go
in
here
that
we
don't
have
listed.
B
It
might
be
good
to
put
the
pr
that
nawaz
was
working
on
in
there.
I
think
it's
a
feature
so
that
I
think
that
is
reasonable
to
what
do
you
think
there
was.
D
Oh
yeah,
so
yeah
I
haven't,
made
good
progress
on
it,
but
I'll
look
into
it
I'll,
look
into
it
and
update
you.
If
we
can
add
this
into
the
release
yeah
so.
D
B
The
milestone
is
more
to
give
an
idea
to
people
of
what
are
the
things
we're
currently
working
on
or
hoping
to
work
on.
It's
not
like
a
strict
commitment
of
like
because
this
is
in
the
milestone
it's
going
to
be
in
the
release,
it's
more
to
give
folks
who
are
like
kind
of
following
for
for
who
don't
have
the
day-to-day
information
of
you
know
what's
happening
right
now
in
canada
and
what's
going
to
be
worked
on
for
the
next
release,.
A
Well,
this
obviously
needs
to
go
in
there
in
the
next
couple
days.
Hopefully,
we
need
to
update
the
to
currentcapy.
A
I'm
gonna,
I'm
gonna,
be
working
on
this
today,
oops.
What's
that
anything
else.
B
Yeah
I
have
one
I'm
working
on
actually,
which
I
don't
know
if
there's
an
issue
for
it,
but
it's
basically
the
pr
that
jack
started
around
calico
using
home
the
work
in
progress
yeah,
so
I'm
I
took
over
that
one
and
I'm
gonna
open
another
pr
for
it
at
some
point
soon.
So
I
don't
know
if
there
was
an
issue.
I
guess
there
is
an
issue
for
it.
Oh
wait,
never
mind
not.
B
Maybe
I'll
create
a
separate
issue
just
for
that
and
put
it
in
there.
Okay,.
B
I
guess
just
an
update
on
that
if
anyone's
interested,
so
I
think
I
found
so
one
of
the
blockers
we
had
or
not
blockers,
but
one
of
the
things
we
were
talking
about
in
this
original
pr
was,
like
you
know,
it'd
be
really
great
if
we
don't
have
to
maintain
our
own
home
chart
and
what
was
brought
up
at
the
time
as
well,
it's
kind
of
difficult
to
implement
the
ipv6
dual
stack
versions
of
it,
but
I
actually
think
I
found
a
way
to
get
it
to
work
with
the
operator.
B
So
we
actually
can
get
around
installing
calico
for,
like
ipv4
ipv6
and
dual
stack
with
the
default
like
publish,
manifest
the
official
publish
manager.
Sorry
official
published
charts
using
just
like
a
slightly
different
values
file.
So
it
ends
up
being
a
lot
cleaner
and
a
lot
less
code
than
having
to
like
put
a
whole
like
custom
home
chart
with
like
if
else
templating
in
there.
A
All
right,
anyone
think
of
anything
else.
We
ought
to
put
in
1.5
milestones.
A
B
I
don't
have
the
answer
to
that.
Yet
it's
next
on
my
list
to
find
out.
I
was
gonna,
look
at
windows
and
see
what
happens
there
so
I'll.
Let
you
know
what
I
find.
B
Yeah
yeah,
I
think
either
it's
our
you
know
already
there.
We
can
just
use
it
out
of
the
box
for
everything
and
that's
great,
if
not
we'll,
just
keep
the
windows
installed
separate
for
now,
as
we're
doing
now
and
wait,
you
know
monitor
until
we
can
switch
over.
A
A
I
guess
not,
let's,
let's
call
it
a
day,
get
a
half
an
hour
back
and
see
you
here
again
in
a
couple
weeks,
thanks
everybody
for
coming.