►
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
much
better:
okay,
hello,
everybody
Welcome
to
kubernetes
Cluster,
API,
azure,
weekly
office
hours.
A
A
If
you
would
mind
adding
your
name
to
the
attendees
list
here,
that'd
be
great
I'm,
relatively
confident.
We
don't
have
any
new
members
or
attendees
this
time.
So
I'll
check
with
skip
over
that.
A
And
we
can
just
go
right
into
open
discussion
unless
anyone
has
anything
random
to
add
so.
The
first
open
discussion
topic
is
from
John.
A
A
The
larger
thing
here
is
and
I
know-
probably
all
you
are
aware
of
this,
but
just
in
case
you're
not
or
in
case
someone's
watching
the
recording.
We
need
to
move
off
of
the
current
version
of
the
SDK,
because
essentially
it's
be
it's
not
going
to
be
supported
in
another
a
little
over
a
month,
actually
a
little
under
a
month,
so
we
want
to
get
off
of
it
as
soon
as
possible.
We
have
two
ideas
for
how
to
do
that.
A
One
of
them
I'll
show
you
in
a
second
the
other
one
is
a
little
more
ambitious
or
maybe
future
looking,
which
is
this
moving
to
the
thing
called
the
Azure
service
operator,
which
is
a
sort
of
kubernetes
native-ish
way
to
call
Azure
apis.
So
we
don't
need
to
go
too
far
into
it
here,
but
this
is
just
John
wanted
to
make
people
aware
that
this
is
had
a
lot
of
review
and
isn't
a
stage
where
anybody
can
hopefully
jump
in
here
and
have
some
feedback.
A
B
B
A
Cool
all
right,
the
next
topic
is
related.
I
put
it
in
here
because
John
put
that
on
there,
but
basically
I
finally
got
to
a
point
where
I
have
a
proof
of
concept
more
or
less
working,
although
I
want
to
change
something
today.
That
I
think
is
could
be
better,
but
this
is
this
is
how
me
we
move
to
the
Azure
SDK
for
go.
That
sounds
like
it'd,
be
pretty
trivial
and
most
of
it
is,
but
they
changed
some
fundamental
things
like
they
don't
return
a
future
object
now
from
the
asynchronous
apis.
A
They
return
a
polar
object
which
is
similar,
but
it
doesn't
work
quite
the
same
way.
So
I
explained
it
all
here,
but
basically
I
took
the
approach
of
cloning.
Just
copying
the
async
framework
that
we're
using
making
some
adjustments
to
make
it
work
with
go
generics
because
that's
in
their
SDK,
V2,
API
and
and
getting
it
all
to
work
for
just
one
service,
whereas
all
the
other
services
are
still
talking
to
SDK
V1,
so
I
think
that's
reassuring
and
probably
applies
to
the
ASO
proposal
too.
A
B
A
I'm
trying
to
just
use
that
exact
same
mechanism,
because
luckily,
the
only
the
only
basically
the
future
objects,
our
own
internal
future
objects
that
we're
storing
in
the
status
field
just
have
sort
of
a
string
field,
an
untyped
field
where
they
store
a
token
and
so
I'm
just
reusing
that
field
instead
of
storing,
essentially
the
Json
representation
of
an
auto
rest
future,
which
is
what
we
currently
do
in
the
async
I'm.
A
Storing
a
resume
token,
so
it's
even
kind
of
fits
better,
but
at
that
at
that
high
level
that
you're
talking
about
it,
looks
exactly
the
same.
You
actually
that
the
async
framework
looks
the
same.
It's
only
slightly
lower
than
that,
but
things
change
much.
A
No
I
mean
not
in
terms
of
like
our
API
diff
check.
No,
it
doesn't
show
up
here.
A
A
So
hopefully
it's
not
very
disruptive
I'm
going
to
tweak
it
a
little
more
today
and
hopefully
get
something
working
that
actually
simplifies
the
code
a
little
bit
and
and
fixes
a
border
case
I
found
and
then
and
then
we
can
compare
and
then
I'm.
Then
it's
an
open
question,
whether
it's
really
worth
where
it's
worth
putting
our
efforts
in
so
I
think
it'll
be
interesting.
A
To
have
people
look
at
both
I
didn't
actually
write
a
proposal,
although
if
people
want
me
to
I,
think
that
would
be
appropriate
I
just
sort
of
jumped
into
the
code
because
it
didn't
look
like
there
was
that
much
distance
between
the
two.
But
if
it
would
be
more
appropriate
to
have
a
proposal
I'm
happy
to
do
that.
C
This
is
more
of
an
update.
Yesterday
was
I
was
able
to
make
a
good
breakthrough.
I
was
able
to
update
the
cloud
config
of
two
worker
nodes
and
get
Cody
and
spots
up
and
running
so
from
the
Outlook.
The
workload
cluster
looked
good
and
working
so
yay,
but
I
still
need
to
figure
out
to
what
extent
am
I
going
to
test
it.
Of
course,
I'll
be
supporting
the
this
Azure
cni
with
a
test
file
or
something,
but
yesterday
was
at
least
a
good
Milestone
to
complete
that
yeah.
A
A
We
only
have
well
I
mean
we
have
about
two
weeks
on
the
calendar
until
March
7th,
but
in
practice
that
really
means
stuff
bottle
in
soon
next
week,
because
you
know
how
it
goes
at
the
end
and
it's
hard
to
get
stuff
through
CI
and
we
need
reviews.
Some
of
these
are
pretty
substantial.
A
Still
so
just
saying
really,
don't
don't
give
yourself
until
March
7th,
because
then
it's
definitely
not
going
to
merge
I,
don't
know
that
it's
worth
going
through
every
single
one
on
the
that's
still
on
the
Milestone,
but
maybe
if
we
just
scroll
slowly,
if
anyone
has
any
updates
to.
B
One
thing
is:
I:
got
blogged
on
the
workload
identity
part
due
to
One
Challenge
regarding
the
cloud
provider
as
our
thing,
so
cloud
provider
Azure
actually
did
not
get
support
for
cloud
identity.
B
A
No
not
at
all.
This
is
a
huge
amount
of
work
right.
We
all
know
so
you
you
so
basically
you're
saying
you'll
have
a
better
read
on
whether
it
can
really
land
for
1.8
sometime
next
week.
B
A
This
is
related
to
my
SDK
v2pr
and,
assuming
that
we
decide
to
merge
that
I'm
going
to
close
this
and
then
open
like
an
epic
PR
that
says:
okay,
here's
all
the
services
we
want
to
convert
and
make
it.
You
know
we'll
check
them
off
one
by
one,
because
each
of
those
will
be
a
straightforward
but
not
necessarily
trivial
PR.
A
And
here's
the
other
half
workload
identity,
this
I
we
actually
merged
something
yesterday
that
fixed
the
more
fundamental
flex
stuff
because
unfortunately,
I'm
still
stuck
on
this
I
did
a
little
brainstorming
with
Cecile.
But
at
this
point,
I
Pro
I
need
some
help.
If
anyone
has
an
idea,
you
can
it's
easy
to
reproduce
otherwise,
I'll
probably
try
and
grab
some
pairing
time
with
people
soon,
I
don't
have.
B
A
I
don't
want
to
defocus
you
I
know
you
have
a
lot
of
stuff
to
do
yeah,
but
if,
if
you
have
a
chance
to
look
at
it,
it
is
at
least
easy
to
reproduce
with
tilt,
but
I
have
no
clue
what's
going
on
still
at
this
point,
thank
you.
I.
A
Now,
as
I
saw
that
depend
about
actually
updated
the
hack
tools
in
a
PR
this
morning,
so
I
don't
know
if
that
means
this
is
fixed
or
if
we
just
don't
understand.
C
C
A
study
yeah
go
ahead.
A
C
Awesome
I
didn't
get
a
chance
to
look
at
that
beard.
I'll,
probably
look
at
it
and
update
this
issue.
If
you
need
this
in
this
release,
foreign.
A
D
Yeah
no
worries
about
the
like
tilt:
cluster
flavor
command,
I.
Think
someone
I
mean
this
should
be
an
easy
fix,
but
I
think
someone
assigned
it
to
themselves,
but.
A
D
A
Yeah,
we
could
just
ask,
because
swastik
has
worked
on
several
other
er's
yeah,
so
this
should
be
easy
for
them.
I
would
hope.
A
C
D
Ahead:
okay,
so
there's
like
one
there's
been
this
bug
that
I
from
the
configurable
scope
PR
that
I
emerged,
that
was
causing
the
upgrade
test
to
fail
on
test
grid
and
that
PR
is
out
and
has
a
LG
TM.
So
it
just
needs
an
approval,
so
it'll
probably
be
good
to
it.