►
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
There
cool
everybody
see
the
document.
I
assume
alrighty,
welcome
everybody.
It
is
not
February
2nd,
it
is
March
9th
2023.
This
is
the
cluster
API
Azure
provider
meeting.
We
are
a
sub-project
of
Sig
cluster
lifecycle
and
therefore
we
try
to
adhere
to
Sig
cluster
life
cycles
to
the
general
kubernetes
Community
code
of
conduct,
which,
as
you
probably
know,
boils
down
to.
Please
let
give
everybody
a
chance
to
talk,
try
to
use
the
raise
hand
feature
and
generally,
let's
be
nice
to
each
other.
A
At
the
beginning,
we
have
a
second
where,
if,
if
you're
new
here
or
want
to
introduce
yourself,
I
can
be
quiet
for
a
second
and
go
ahead
and
do
so.
A
I
think
I
recognize
all
the
faces
here,
but
just
so
we're
following
our
protocol,
all
right
done
and
if
you
want
to
add
your
name
to
the
attendee
list
above
that
would
be
great
and
let's
move
on
to
open
discussion.
A
The
1.80
release
is
out
woohoo.
Everybody
here,
I'm
pretty
sure,
contributed
to
this
in
Fairly
significant
ways.
So,
let's
all
Pat
ourselves
on
the
back.
A
There's
some
pretty
big
features,
and
this
is
a
fairly
significant
minor
release.
As
everybody
knows,.
B
Yeah
I
just
want
to
just
call
attention
to
the
two
announcements
that
are
at
the
top
of
the
release
page
and
make
sure
everyone
reads
those
if
they're
going
to
upgrade
the
first
one
is
AKs
clusters
or
graduated
out
of
experimental.
This
is
huge.
This
has
been
in
the
work
for
a
month,
if
not
over
a
year
this.
What
this
means
for
existing
users
is
that
one
you
no
longer
need
to
use
the
manage
cluster
feature
flag
and
two.
B
B
B
So
for
those
of
you
who
are
not
aware,
there's
been
for
ongoing
in
kubernetes
to
move
all
the
cloud
providers
specific
code
out
of
the
kubernetes
project
and
make
kubernetes
truly
Cloud
agnostic
such
that
cloud
providers,
release
their
components
for
cloud
controller
manager,
Etc
like
separately,
and
some
of
our
templates
were
still
using
the
Azure
cloud
provider,
which
is
the
entry
provider,
and
so
we
have
now
moved
everything
to
use
out
of
troop.
This
does
not
impact
existing
clusters.
So,
if
you
are,
you
know
using
Cabazon,
you
upgrade.
B
This
won't
affect
your
clusters,
but
it
does
impact
you
if
you're,
using
our
reference
templates
or
flavors
to
create
clusters
in
the
sense
that
now
the
flavors
use
the
out
of
tree
copywriter,
which
means
that
it
does
not
install
a
cloud
provider
by
default.
The
copywriter
is
now
an
add-on,
which
means
you
need
to
take
a
separate
step
just
like
cni
to
install
the
collab
provider
their
instructions
in
the
docs.
Please
take
a
look,
and
let
us
know
if
you
have
questions
questions.
A
Cool
anybody
have
any
questions
about
the
release.
This
document
is
release.
Notes
are
pretty
clear,
but
again
it's
pretty
good
work.
We
should
all
be
pretty
proud
of
this
release.
A
Let's
move
on
the
next
topic,
I
wonder
if
we
should
take
this
to
the
end
since
jack
isn't
here
and
we
can
give
just
in
case,
we
want
to
run
out
of
time.
We
might
run
out
of
time
I'm
going
to
do
so
so
Dane.
If
you
want
to
talk
about
support,
bring
your
own
network
interface.
Let's
do
that.
First
and
I'll
move
this
down.
C
Surely
so
this
was
just
something
I
kind
of
wanted
to
throw
out
here
for
a
very
brief
discussion
before
I
went
and
created
an
issue
for
it
to
be
there.
I
haven't
looked
for
an
issue.
Maybe
there
is
one
I,
don't
think
there's
one,
but
the
idea
is
fairly
simple.
C
In
certain
use
cases
you
may
need
to
pre-create
network
interfaces
if,
for
instance,
virtual
machines
need
stable,
IPS
things
like
that,
and
this
is
something
that
we
use
and
we've
used
in
other
Cloud
providers,
so
we'd
be
interested
in
bringing
that
capability
to
cap,
z
and
I.
Think
the
main
question
I
had
before
we
went
about
creating
an
issue
and
potentially
starting
on
any
code,
was
just
around
how
that
should
function.
I
did
look
at
some
of
the
existing
bring
your
own
things.
C
It's
pretty
clear,
there's
some
some
tagging
to
represent
ownership.
Whether
or
not
cap
Z
owns
it
and
created
it
and
should
tear
it
down
or
not.
That
looks
pretty
straightforward
and
then
it
looked
like
there
were
a
couple.
C
Different
conventions
used
across
resources,
sometimes
where
resources
seem
to
be
specifying
entire
IDs,
such
as
maybe
an
image
ID
or
in
other
cases,
just
like
a
network
interface
name
and
then
the
resource
Group
and
subscription
ID
are
inferred
from
presumably
the
Azure
cluster
object
and
I
just
wanted
to
throw
that
out
there
before
we
started
like
adding
Fields
what
folks
thought
was
most
idiomatic
for
capsie.
B
I
think
in
this
case,
you
probably
want
to
look
at
v-nets
as
being
the
kind
of
model
for
this,
since
it's
the
probably
the
closest
example,
I
think
images
are
a
little
bit
different,
not
necessarily
for
good
reasons,
but
I,
wouldn't
necessarily
look
at
that,
as
the
reference
I
think
Venus
is
what
you
want
to
be
looking
at
and
you're
right
that
we
do
some
tagging
to
kind
of
say,
like
this
resource
is
owned
by
kab
Z
and
the
the
ideas
that
we
never
delete
things
we
don't
create.
B
So
we
only
clean
up
the
resources
that
cavity
created
and
that's
what
allows
us
to
know
whether
it
was
created
by
cabs
or
not
for
by
this
specific
cluster
too.
So
if
it
was
created
by
another
cluster,
we
don't
want
to
delete
it
either
because
it
might
be
shared
by
clusters
and
that
cluster
might
still
be
using
it.
B
So
yeah
I,
don't
see
any
concern.
I
think
like
starting
with
an
issue
is
probably
the
best
way
to
start
are
the
do
you
know
if
the
network
interfaces
are
already
getting
tagged
right
now,.
B
Okay,
Goods
because
otherwise
that
might
add
some
tricky
backwards
compatibility
handling,
because
if
they
are
being
tagged
then
we
can,
we
already
know
which
ones
were
created
in
the
past.
So
that's
good
I.
A
So
the
goal
here
is
to
open
a
proposal
or
an
issue
about
this.
Okay,
great
anybody
have
any
more
questions
for
Dave
about.
This
seems
like
a
great
idea.
I
was
thinking
somehow
this
might
overlap
with
the
Osa
stuff,
because
I
know
John.
You
talked
a
lot
about
bring
your
own,
but
maybe
that's
premature
or
maybe
not.
A
Cecile
I'm
glad
you
put
this
on
here,
because
I
was
going
to
bring
it
up.
It
seems
like
all
the
PRS
are
failing.
Yeah
sorry
go.
B
Ahead,
all
right
can
I
I
just
have
one
more
question
for
Dan
before
we
move
on,
I
was
actually
I'm
wondering.
Could
you
just
talk
a
bit
more
about?
What's
your
use
case
for
this,
because
for
v-nets
it's
pretty
obvious
people
want
to
share,
like
you
know,
v-netsam
different
clusters,
different
resource
groups
for
network
interfaces.
Are
you
doing
some
kind
of
sharing
like?
What's
what's
the
what's
the
scenario
here.
C
It's
it's
a
fairly
simple
use
case.
Actually,
the
main
requirement
is
simply
that
the
VM
have
a
stable
IP,
even
if
the
VM
is
replaced.
C
C
They
don't
necessarily
need
to
be
specific
IPS.
They
just
need
to
be
stable.
Then
we
put
the
ID
of
that
network
interface
into,
in
that
case
the
AWS
machine
template,
so
that
when
the
machine
deployment
creates
the
machine,
it
always
comes
up
with
that,
attaching
that
existing
network
interface
and
it
always
comes
up
with
the
same
IP
in
this
case.
It's
because,
like
I,
can
get
this
specific
application.
C
B
Got
it,
and
is
there
no
way
of
doing
that
same
thing
with
like
cabsy?
Creating
the
interface
like,
let's
say
saying,
like
I,
want
a
static
IP
for
my
or
it
needs
to
be
always
the
same
one,
even
if
the
machine
is
replaced
right,
so
you
said
correct
got
it
okay,
which
wouldn't
work
unless
you
specified
a
deterministic,
IP
I,
guess
which
I
don't
know
if
it
lets
you
do
that.
B
Okay,
cool
well,
that
might
be
just
worth
checking
just
before
we
start
doing
all
this
work,
but
but
yeah
otherwise
sounds
good.
Thanks
for
explaining.
A
C
Yeah,
so
we
just
let
it
create
and
the
in
theory
if
there
were
a
way
to
specify
that,
then
we
could
probably
attempt
to
create
logic
that
tries
to
find
a
free,
IP
and
and
sets
that
when
those
templates
are
created,
that
would
be
a
lot
more
Logic.
On
the
you
know.
On
the
cluster
operator
side,
it's
also
error
prone,
because
there
is,
unfortunately
a
window
when
machines
are
replaced
that
that
IP
is
technically
free.
If
you
actually
delete
Nyx
by
leaving
the
Knicks
always
up,
those
IPS
are
effectively
reserved,
nothing
can
grab
them.
D
A
No
okay,
Cecile
this.
This
test
has
been
failing
all
right,
yeah.
B
I,
don't
really
have
anything
to
add
so
far,
I
haven't
looked
at
it.
I
just
saw
this
just
logging
on
this
morning
and
right
before
coming
to
this
call,
but
yeah
thanks,
Michael
I,
don't
know
if
Michael's
here,
but
thanks
Michael
for
opening
the
issue
but
yeah,
basically
we're
seeing
all
the
end-to-end
tests
feeling
trying
to
pull
the
cute
proxy
image
for
Windows,
so
something's
wrong.
There
something
happened
environmentally
overnight,
probably,
and
we
have
to
look
into
what
that
is.
A
B
And
if
anyone
sees
this
in
their
PR,
it's
just
FYI.
This
is
known
and
if
anyone
knows
what's
going
on
or
has
it
an
idea
of
how
to
fix
it,
please
speak
up.
D
D
Yeah,
basically,
it's
just
there's
a
lot
of
issues
or
bugs
that
are
around
that
need
help
to
that
people.
If
you're
able
to
please
take
a
look,
because
there's
I
just
did
a
recent
issue,
triaging
of
all
the
bugs
and
there's
a
few
old
ones
that
seem
to
have
I,
don't
think
anyone
has
looked
at
them
in
a
while.
So
there's
a
lot
of
them
that
are
I,
think
pretty
important
to
work
on.
D
So,
if
you
all
have
any
extra
Cycles
it'd
be
awesome
to
take
for
people
to
work
on
these
there's
also
a
few
others
that
haven't
been
tagged
with
Help
Wanted,
but
I'm
kind
of
just
like
paying
the
issue
Creator
to
see
if
it's
like
still
an
issue
so
as
they
respond
I'll,
add
the
Help
Wanted
tag,
but
for
now
these
are
the
ones
that
need
the
most
attention.
I
think
see:
I'm,
just
a
PSA
for
anyone
who's.
Looking
for
something
to
do.
B
Oh,
no
sorry
I
just
want
to
say
thanks
for
doing
that,
and
also
maybe
we
should
add
some
of
these
to
the
milestone
for
1
8.
Since
we're
about
to
start,
you
know,
planning,
for
that
would
be
a
good
place
to
start
focusing
on
cleaning
up
our
bugs
in
our
issue.
Queue.
D
Yeah
I'll
have
permissions
to
tag
things
as
Milestones,
but
maybe
during
the
Milestone
review
we
can
take
a
look
okay,
so,
like
kind
of
my
next
topic
is
like
kind
of
related.
So
there
has
been
a
few
bugs
I
think
that
are
mislabeled
as
bugs
and
also
like
features
that
are
mislabeled
because
they're
I
don't
think,
there's
an
issue
template
for
stuff
that
isn't
necessarily
a
bug
or
a
feature.
So
it's
not
like.
D
If
it
isn't
a
bug,
then
it's
a
feature
so
I
think
we
should
probably
create
like
a
new
issue
template
for
maybe
cleanup
issues
or
just
any
like
suggestion
issues,
because
right
now
it's
kind
of
just
like
feature
requests
or
bugs
so
yeah
I
haven't
created
an
issue
for
this.
Yet
it's
kind
of
meta
but
I'll
I'll
get
an
issue
out
and
then
I'll
probably
be
working
on
that.
So
just
keep
an
eye
out
when
you're
creating
a
new
issue
make
sure
you
label
it
correctly.
A
Yeah,
it's
true.
We
don't
give
people
enough
options
and
I
have
been
guilty
of
this
good.
D
Idea
and
yeah
I
think
overall,
some
of
the
issue
templates
may
need
a
little
refactoring
because
some
of
them
are
like
kind
of
huge
and
I
end
up
just
deleting,
but
like
most
of
it.
So
maybe
it'll
just
be
part
of
a
larger
effort
to
refactor
those.
A
A
Cool
any
questions
about
templates
or
bugs
should
we
move
on
to
the
next
item
is
something
Jack
wanted
to
bring
up,
but
he
wasn't
able
to
be
here
this
morning,
but
I'm
assuming
it's
self-explanatory
right
now,
our
rules
about
what
we
will
cherry
pick
restrict
us
to
only
cherry
picking
straight
up,
bug
fixes
back
to
the
supported
branches
and
Jack's
wondering
if
it'll
be
okay
to
change
the
rules,
so
that
documentation
is
acceptable
to
cherry
pick.
A
C
A
B
So
you
can,
if
you
know
the
URL,
so
it's
not
discoverable,
but
there
is
a
way
to
get
to
it,
but
the
what
we've
done
recently
is:
we've
pinned
the
like
main,
like
book,
to
be
the
current
latest
release
instead
of
the
main
branch,
because
we
had
a
few
occurrences
of
users
being
really
confused
coming
to
using
like
I'm,
seeing
this
in
the
dock,
I'm
doing
the
exact
same
steps,
but
it's
not
working
and
then
we're
like.
Oh
that's
because
that's
not
released
yet
so
this
feature
doesn't
exist
in
your
release.
B
So
in
order
to
avoid
that,
since
the
book
is
meant
to
be
mostly
user
facing,
we
changed
it
to
be
based
on
the
latest
release
branch,
which
means
that
if
we
like
add
docs
like
mid-cycle
for
something
that
already
exists
or
we
make
like
a
doc
fix,
then
that
won't
actually
get
to
the
book
until
the
next
Miner.
Unless
we
chair
pick
it,
which
is
I,
think
where
this
suggestion
is
coming
from
I'm
plus
one
of
this
I
think
it's
a
good
idea,
mostly
for
Doc
fixes,
though
not
for
like
net
new
features.
A
B
Part
of
that
we
have
to
I,
don't.
B
Done
it
yet
I,
don't
know
if
Jack
did
it
I
Didn't,
Do
It,
so
I'll
follow
up
with
him,
but
yeah
yeah.
That's
something.
B
B
A
A
Okay,
General
agreement
was
in
evidence,
hey,
that's
all
we
had
on
the
agenda,
but
do
people
have
anything
else
they
want
to
bring
up
before
we
look
at
a
Milestone
review.
A
E
Yeah,
just
a
quick
question
about
the
hosting
is
just
basically
the
learning
now
is
just
following
the
directions
at
the
top
and
volunteering
one
week
or
is
there
something
else?
I
should
follow
up
on
ahead
of
time?
I
can't
do
it
next
week,
but
I
was
thinking
perhaps
the
following
week.
I
could
help
out
with
some
of
that.
A
E
Yeah,
if
I
follow
up
and
see
if
there
was
anything
special
I
should
do
ahead
of
time
or.
A
Not
no,
but
that'd
be
great.
If
you
want
to
do
it
in
a
couple
weeks,
we
will
pencil
you
in.
A
A
A
Oh,
let's
close
it
woohoo.
A
So
the
two
sources
of
input
are
going
to
be
everything
we
threw
into
next.
It's
tempting
to
say,
by
default.
This
should
all
go
in
1.9.
Is
that
what
we've
done
in
the
past
Cecile.
B
I,
don't
think
so.
I
think
the
next
is
more
like
a
general
touch.
I'll
look
like
oh
we're,
not
doing
this
right
away,
but
we
sh
don't
do
this
I,
don't
think!
We've
been
very
good
at
triaging
next,
but
I
think
maybe
this
time
we
could
start
from
either
next
and
or
like
the
bugs
that
Willie
was
talking
about
it's
like
we
kind
of
try
to
prioritize
that
before
we
fill
it
up
with
more
stuff,
but
then
yeah
after
that.
B
A
F
A
I
mean
I
think
most
of
the
things
that
are
already
PR's
and
that
haven't
gotten
Rusty
should
go
on
the
milestone
this
Nawaz.
Thank
you
for
following
up
on
that.
By
the
way
this
needs
to
be
addressed
soon.
A
This
is
a
lot
of
effort.
I
know
that,
but
that's
probably
going
to
happen
soon,.
B
E
A
Yeah,
oh
right,
I
guess
I'll
just
scroll
slowly
and
if
we
are
pretty
sure
any
of
these
go
in
the
next
Milestone
and
just
shout
it
out.
A
B
A
B
Is
that
what's
happening
on
April
3rd
I?
Don't
want
to
spread
false
information,
but
yeah
I
know
something's
happening
on
April
3rd
yeah.
A
A
Workload,
identity.
Yes,
let's
hope.
A
E
A
B
To
that,
no
that's
that's
unrelated
I
think
that
issue
has
been
open
for
a
while
I
I
fixed
that
and
then
I
think
someone
asked
to
reopen
it.
I
should
look
into
that.
So
we're
definitely
we're
testing
out
a
tree,
Club
writer
with
Windows
now
and
that's
been
merged
for
over
a
week
without
issues
so
I'll
look
into
that
yeah
should
the
cluster
class
one
be
in
there
really
or
is
that
is
that
too
much
hoping
above
I
think
it's
a
bit
higher
yeah
that
one.
A
B
A
A
little
progress
animation
hold
on
there
we
go,
and
then
we
can
bounce
over
to
this
list
that
Willie
came
up
with
and
see.
If
any
of
these
are
good
candidates.
A
D
I
think
most
of
these
aren't
that
big
they're,
just
like
looking
for
people
to
work
on
so
I'm,
not
sure
if
we
should
just
add
all
of
them
that
aren't
assigned-
or
some
of
them
are
assigned
them.
Okay,
that
one
has
a
PR,
so
that
should
probably
be
good.
F
A
Generator
is
broken,
I
don't
know.
Some
of
these
are
definitely
good
first
issues,
so
I
don't
think
that
they
should
call
the
Milestone,
but
well
does
any
of
these
other
ones
or
should
we
just
call
it
good?
With
these
couple.
D
D
D
A
A
And
then
this
we've
talked
about
it's
already
on
the
milestone.
This
is
assigned
to
someone
but
I'm
not
sure
they're
actually
working
on
it.
Yet
so
I
think
I'll
just
leave
it
there.
A
D
Some
of
the
other
ones
seem
a
little
larger.
I
know
the
one
where
it's
like
e
to
e
tet
scripts
should
not
edit
files
outside.
It
ignores
a
kind
of
a
hard
one,
because
I
think
this
has
been
attempted
before
and
it
doesn't
it's
just
like
it's
just
a
way
with
how
like
e
to
e
test
scripts
need
to
like
change
the
registry,
to
use
like
the
Local
Host,
but
then
until
it
requires
the
GCR
dot
io1.
So
it's
a
little
bit
like
a
conflict.
There
yeah.
E
A
A
Okay,
should
we
call
that
good
for
a
first
pass,
or
do
we
want
to
actually
go
through
all
the
issues?
A
Unless
people
I
mean
obviously
at
any
time,
we're
all
encouraged
to
put
these
on
a
milestone,
if
you
don't
actually
have
permission
to
add
stuff
to
a
milestone,
just
ping
one
of
us,
but
we
should.
We
should
all
be
doing
this
as
a
continuous
activity.
This
is
just
like
because
we
forget
to
do
it
ideally.
B
Sorry,
just
for
remember
I
forgot
to
add
this
earlier,
but
there
are
two
PR's
open
right
now
to
add
two
new
reviewers
to
cap
Z.
One
of
them
is
wooly
and
the
other
one
is
Noah
they're
both
here.
B
B
So
if
anyone
has
you
know,
anyone
I
think
they
want
to
add
before
it
merges.
Please
do
so,
possibly
in
the
you
know
next
hour
or
next
few
minutes,
but
yeah
just
wanted
to
call
this
out
and
congrats
to
willianos,
for
you
know
all
their
help
being
a
reviewer
already.
You
know
and
I
think
this
is
just
recognition
of
their
efforts.
So
thank
you.
B
B
And
also
sorry,
if
anyone
else
who
is
here
would
like
to
become
a
reviewer
we're
definitely
looking
for
more
reviewers,
and
sometimes
you
know,
you
think
you
need
to
be
doing
this
for
a
lot
longer
than
you
actually
do.
I
think
being
a
Reviewer
is
also
part
of
the
learning
and
towards
becoming
an
approver.
So
if
you're
interested
in
you
know
getting
involved,
there
like,
please
feel
free
to
reach
out
to
me
I'm
happy
to
answer
any
questions
and
talk
about
it
and
yeah
it'd
be
great
to
have
more
folks.
A
Okay,
unless
people
have
some
final
comments
or
something
else
to
add.
A
I
guess
not,
let's
call
it
good
for
today
thanks
everybody
and
we'll
see
you
next
week.
Hopefully.