►
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
As
you
probably
know,
we're
abiding
by
the
cncf
code
of
conduct,
which
you
can
read
about,
but
basically
boils
down
to
everybody,
be
kind
to
each
other
and
try
to
use
the
raise
hands
feature
in
Zoom.
So
we
don't
talk
over
each
other
yeah
and.
A
If
you
would
like
to
add
your
name
to
the
attendees
list,
just
so
people
can
Network
later
or
whatever.
Please
do
so.
A
This
is
also
time
where,
if
somebody
is
new
here,
it'd
be
great.
If
you
want
to
say
hi
and
identify
yourself,
I'll
be
quiet
for
a
minute,
so
you
can
just
unmute
and
do
that
if
there's
anyone
who
wants
to.
B
Hi,
this
is
Jorge
Luna
from
Microsoft.
B
B
Oh
sorry,
okay
go
ahead.
Everyone
is
like
hello.
I
can
delay
I,
actually
work
with
Joaquin
Norway,
so
glad
to
be
here,
awesome,
hey
everybody!
This
is
lucky
because
I
also
work
with
AKA
and
Jorge
glad
to
be
here.
A
All
right
not
much
of
an
agenda
today,
but
we'll
see
things
often
evolve.
Cecile.
Do
you
want
to
talk
about
the
patch
releases.
C
C
They
each
have
the
same
content,
which
are
three
bug
fixes
that
were
identified
over
the
past
few
weeks,
thanks
ajitosh
for
making
some
of
the
fixes
in
there
and
also
pushing
for
release
so
yeah
check
it
out.
And
then,
if
you
have
any
questions,
please
let
us
know.
A
There
cool
any
questions
about
the
patches.
B
D
Would
so
we
have
a
light
agenda
today,
but
we
have
lots
of
folks
here.
Should
we
maybe
do
some
1.6
milestone.
D
Jamming
would
that
be
we'll
put
that
up
for
a
vote?
A
I
mean
Milestone
review
is
one
of
the
things
we
do
at
the
end
and
we
haven't
really
done
that.
So,
oh
hey,
what's
going
on.
D
Yeah
and
Cecile
asks
if
there
are
any
other
topics
that
we
want
to
discuss.
Ad
hoc
I
could
give
four
questions
yes
time
permitting.
I
can
give
some
updates
on
the
Azure,
managed
cluster
story,
but
I
think
Milestone
1.6
would
be
probably
the
most
the
best
use
of
our
time.
We're
actually
not
that
far
away.
Given
we
do
bi-monthly
releases.
D
B
D
C
That
no
I
think
we
still
need
to
I
think
the
main
missing
part
is
the
cluster
template.
So
for
non-managed
clusters
we
had
to
add
Azure
cluster
template
in
order
to
get
cluster
class
working.
I
think
we'll
have
to
do
the
equivalent
change
with
managed
clusters,
so
Azure
managed
cluster
template.
They
still
a
missing
piece
and
yeah
I
think
we
need
an
issue
tracking
that
and
we
don't
have
one
right
now
so
feel
free
to
open
one
she'd
love
to
see
that
happen.
B
Awesome-
and
so
she
just
mentioned
the
the
templates-
are
what
would
be
required.
The
cluster,
the
cluster
templates
for
using
cluster
class.
C
Sorry
by
template,
I
don't
mean,
like
templates
and
yaml
template
just
the
there's,
a
a
crd
called
clustered
Azure
cluster
template,
which
is
basically
a
something
that
cluster
class
uses
to
then
make
individual
clusters
from
that
template,
and
so
we
need
the
equivalent
for
manage
cluster.
So
we'd
be
adding
that
crd
I
can
send
you
a
link
to
the
cluster
template.
If
that
helps.
But.
B
C
D
Yep
thanks
Cecil.
Could
you
actually
do
you
mind,
adding
some
context
as
to
your
use
case.
B
Ahead
yeah,
the
main
use
case
currently
is
use
using
life
cycle
hooks,
though
using
the
runtime
by
cycle
Hooks
and
from
talking
on
the
from
the
information
I
got
from
the
cluster
API
slack
channel
that
that's
only
supported
when
you
use
cluster
class,
so
using
just
the
cluster
crd
currently
does
not
work
with
the
life
cycle.
Hooks
extension
feature
so
we're
hoping
to
get
that
working
for
some
of
the
work
we're
doing.
D
A
Cool
any
other
things
you
want
to
talk
about,
or
should
we
go
back
to
1.6.
D
So
date
it
looks
like
we
released
105
on
I
was
due
on
the
7th
I'm,
not
sure
if
we
actually
release
it
on
the
seventh,
but
we're
a
little
one
had
add
two
months.
Does
that?
Does
that
mean
November
9th?
D
C
I
guess
in
the
spirit
of
retro
comments,
does
anyone
else
have
any
like
I
guess
any
thoughts
on
how
we're
currently
doing
Cadence
like?
Is
that
working
well
or
do
we
find
that
releases
are
coming
too
soon
and
we're
not
having
time
to
get
enough
in
there
or
is
it
good,
I'm
personally,
finding
it
working
pretty
well,
but
I,
don't
know
if
others
have
thoughts.
A
It
it
was
a
little
challenging,
as
Jack
pointed
out,
to
get
1.55
done
exactly
on
the
date
we
were
trying
to
and
then,
of
course,
we
ran
into
some
things
and
then
we
realized
we
hadn't
implemented,
server-side
apply
and
a
couple
other
things
and
I.
Imagine
it'll
off
it'll
usually
be
like
that,
but
it
seemed
like
a
healthy
exercise.
Even
though
it
we
slipped
by
a
couple
days,
I
I
liked
being
focused
on
a
date.
D
D
Just
because
there
were
some
breaking
changes
in
in
Cappy
1.2
that
we,
with
the
help
of
Stefan
mother
folks,
we
spent
a
couple
days
trying
to
get
as
much
of
those
baked
into
our
1.5
release
as
possible,
so
that
I,
don't
think
we'll
encounter
every
release.
But
we
will
encounter
that
from
time
to
time
and
I
I
think
that
was
super
healthy
for
the
project.
A
Any
other
opinions
on
the
release,
Cadence,
it's
still
sort
of
in
review.
Have
we
been
doing
patch
releases?
Well
enough,
we've
been
doing
them
almost
once
a
week
and
kind
of
on
demand,
which
is
a
little
more
than
we
used
to.
D
If
you
want
to
volunteer,
maybe
I
can
pull
the
room.
What
is
your,
what
is
the
the
Cadence
that
you
actually
consume
new
releases?
Are
you
able
to
aggressively
update
your
your
environments
with
the
latest
cap
C
bits
reliably.
E
Yes,
so
far
like
we
are
good.
We
are
able
to
consume
and
update
it
and
yeah,
but,
like
I,
don't
know
like
like
how
do
we
support
dot
releases?
So
maybe
you
know
for
fixing
bugs
like
thanks.
First
of
all,
thank
you
Cecil
for
helping
out
on
that
dot
release.
E
E
You
know
guidelines
to
that
because
this
time
like
because
of
that
work,
there
was
some
Downstream
pipeline
that
got
blocked
for
you
know
couple
of
days
but
yeah
we
got
the
thought
release.
So
thanks.
D
D
D
D
We,
we
have
recently
agreed
to
sort
of
pull
the
community
every
week,
so
I
think
the
convention
would
be
during
this
meeting
and
based
on
community
desire.
I
mean
this,
isn't
the
exclusive
justification
for
releasing
a
patch
release,
but
it's
probably
the
most
common,
so
we
would
pull
the
community.
So
we
can
do
that
right
now,
as
an
exercise
and
say:
do
we
want
to
plan
on
releasing
a
1.5.2
early
next
week
and
So
based
on
community
feedback?
D
That
would
go
some
ways
so
the
way
toward
determining
that.
So
it's
as
frequent
as
once
a
week,
we
could
cut
a
patch
release
from
the
head
commit
of
the
the
release:
the
M
minus
two
release
branches.
So
right
now
we're
1.5
and
1.4
Cecile.
C
Yeah
I
had
mentioned
that
last
time
we
talked
about
this
I
still
think
we
should
avoid
doing
any
sort
of
consensus
or
Community
polling
solely
in
the
office
hours.
Just
because
not
everyone
is
able
to
make
it
so
I
think
we
should
start
with
a
slack
async
thread.
Maybe
the
automated
reminder:
every
week,
that's
the
day
before
the
office
hours
and
then
at
office
hours
we
can
go,
look
at
the
responses
and
say:
okay.
Anyone
else
have
anything
to
add.
Should
we
do
the
release
based
on
that
information?
D
D
Okay
cool,
so
so
my
point
really
Ashton,
since
we're
having
a
conversation
here
is:
are
you
finding
it
effective
to
update
your
environments
from
1.5.0
to
1.5.1
on
a
Cadence
as
frequently
as
every
week
or
every
two
weeks,
not
that
you
are
representative
of
everyone
using
capsi
but
I'm
just
curious?
If
that's,
if,
if
we're
doing,
there's
any
gaps
in
actually
being
able
to
get
these
releases
into
folks
hands
as
quickly
as
we're
releasing
them.
E
I,
don't
have
an
immediate
answer
to
this
stack.
I.
Think
I
also
need
like
a
little
more
context
on
how
frequent
you
know.
Things
are
Downstream,
so
maybe
I'll
get
back
to
your
answer.
You
know.
Okay,.
D
Yeah,
it's
not
I,
just
I'm
honestly
just
curious,
as
we
have
folks
here
in
the
community.
If
these
weekly
releases
are
getting
are
able
to
get
into
hands
because
there's
a
a
whole
practical
world
that
is
sort
of
hidden
from
us,
as
the
developers
and
I
want
to
make
us
aware,
if,
if
there's
something
better
that
we
could
do
to
make
these
releases
more
easily
consumable.
A
Cool,
so
do
you
think
we're
we're
a
152
patch
release?
Should
we
go
ahead
today
and
then
try
and
create
a
select,
pull
and
see
if
we
can
kick
it
off
that
way
or.
A
C
A
A
I
mean
if
it
seems
obvious
that
we
meet
the
criteria.
We
have
some
important
fixes
that
we
want
to
release.
Do
we
need
to
go
through
the
polling
process?
We
said
we
were
going
to
anyway,
but
go
ahead.
Cecile.
C
I
think,
given
that
we
just
released
yesterday,
maybe
we
should
plan
on
doing
another
one
early
next
week,
just
to
give
a
I,
don't
I
mean
unless
people
think
this
one
can't
wait,
but
I
don't
think
we're
at
the
like.
This
isn't
a
regression.
This
is
a
bug.
That's
been
there
for
a
long
time,
so
I
don't
think
an
extra
day
or
two
will
make
a
huge
difference
and
then
I
think
after
that,
maybe
we
can
set
up
the
regular
polling
for
the
following
weeks.
D
Check
check
yeah
so
just
to
be
that's
a
great
point.
We
actually
discussed
a
few
weeks
ago
in
in
office
hours
and
sort
of
at
least
the
folks
present
at
that
office.
Hours
sort
of
voted
to
prefer
not
releasing
at
the
end
of
the
week,
so
I
think
there's
a
assumption
when
we
talk
about
this
polling
happening
toward
the
end
of
the
week
that
we
would
give
folks
a
few
days
to
weigh
in
and
that
the
actual
release
would
be
cut
on,
like
say
a
Tuesday,
or
something
like
that.
D
A
A
D
B
A
A
I
guess
we've
done
that
one
to
death
at
this
point?
Probably
so
so
we're
gonna,
so
we're
gonna
do
a
patch
release
on
Monday.
It
sounds
like
and
then
try
and
start
this
notification
or
polling
thing
next
week
for
the
subsequent
patch
release.
D
No
I
think
we
should
pull
and
then
let
give
the
community
an
opportunity
to
say.
Please
don't
cut
this
release
because
there's
another
bug
we
want
to
fix.
So
if
you
can
wait
till
Wednesday,
that
would
be
hugely
helpful.
I
think
we
want
to
give
someone
a
chance
to
to
express
that
particular
opinion.
Okay,.
A
Well,
I
can
try
and
whip
up
some
slack
thing
today
that
says:
hey
we're
thinking
of
doing
a
release.
That
would
be
next
Monday,
please
let
us
know
if
that's
a
good
or
bad
idea,
and
then
we
can
make
that
the
template
for
going
forward.
A
D
It
be
helpful,
maybe,
in
the
background
asynchronously
I
will
be
on
a
different
on
my
own
browser
window
retroactively
tagging
things
that
haven't
been
tagged
a
milestone
at
all,
yet
to
better
reflect
the
work.
We've
already
done,
that
it
will
land
in
1.6.
So
there's
a
bunch
of
features
in
Maine
that
don't
have
this
Milestone
tag
because
it
wasn't
existent
when
that
featured.
D
A
D
C
I
was
just
going
to
ask:
is
there
anything
still
in
the
1.5
Milestone
that
should
be
moved
over?
Was
that
one
closed
or
should
we
close?
It.
C
Oh
and
then
I
was
also
going
to
say
if
anyone
has
any
yards
or
issues
or
specifically,
issues
that
they're
tracking
that
they
would
like
to
see
get
into
the
next
release.
Maybe
it'd
be
helpful
to
paste
those
links
in
the
chat
or
in
the
agenda
somewhere.
So
we
can
take
a
look
at
those
and
kind
of
crowdsource.
A
Sounds
good
there's
nothing
in
1.5,
so
we
can
go
ahead
and
close
it.
A
A
Machine
pull
bootstrap
token
does
not
get
refreshed
automatically
I,
don't
know,
I
hesitate
to
drill
into
any
of
these
because
then
we'll
all
be
here
for
another
half
hour,
so
I'm
just
going
to
stay
at
this
level,
and
if
people
want
to
shout
out
the
numbers
of
the
bug
or
put
it
in
chat
or
something
like
that,
we
can
use
that
as
a
process
to.
F
Yeah
that
one's
related
to
when
you
like,
run
the
test
clip
locally
it
like
changes,
a
bunch
of
files
that
you
don't
want
to
commit
and
a
lot
of
people
like
commit
like
by
mistake.
So
that's
something
yeah
and
also
like
one
of
the
files
that
changes
messes
up
tilt.
So
it
like
a
lot
of
people
might
not
understand
why
their
tilt
is
not
working
after
they
run
like
an
end-to-end
test
script.
A
F
Yeah
sure
this
is
might
be
more
complicated
than
before,
because
then
on
the
surface,
because
this
you'll
mentioned
they
try
to
fix
it
before.
But
it
was
like
some
design
that
didn't
allow
it
to
be
changed,
but
I'm
happy
to
like
work
on
it
and
see
if
I
can
get
it
done.
B
A
C
C
The
person
who
opened
the
issue
actually
said
they
were
about
to
open
a
PR.
So
if.
B
E
Yes,
this
is
a
issue
of
that
I.
Think
Jack
has
already
raised
the
pr
we
discussed
it
I
think
in
Lost
office
hours
we
need
to
actually
get
server
side,
apply,
integrated
and.
E
For
that,
we
need
to
put
tags
on
the
go
structures
so
that
server
side
apply
works.
Fine,
otherwise,
you
know
tldr
is
that
the
topology
controller
from
the
capicide
as
well?
As
you
know,
the
cap
Z
controller,
they
kind
of
compete
with
each
other
and
end
up
updating
the
object.
You
know
continuously
and
more
details
and
findings.
I
have
already
pasted
in
the
issue
so
but
I.
E
There
is
some
comment
on
the
pr
and
we
are
not
sure,
often
of
a
field
that
could
act
like
a
primary
key.
Something
like
that.
So.
F
D
I
think
it
should
definitely
go
into
the
release
Milestone
for
1.6
and
it
will
be
back
ported
once
we
so
I
actually
think
we
just
need
to
lean
in
a
little
bit
and
there's
some
good
conversation.
I
feel
like
we're
almost
there
in
the
pr
threat.
Cecile
might
offer
some
additional
thoughts,
but
do
you
agree
with
that?
Cecile
yeah.
C
A
Yeah,
thanks
for
adding
at
the
Milestone
I
didn't
I
just
realized.
It
was
already
there
and
the
next
one
is
as
well
so
we'll
skip
over.
That
rate,
limiting
and
429
back
off
for
azure
in
general
sounds
like
a
big
deal.
D
A
A
A
Help
wanted
so
it's
pretty
much
by
definition,
not
going
to
be
in
the
milestone:
yeah,
that's
definitely
like
V
Next
or
something
adding
the
cubescape
scanner
to
all
the
tests.
Aks
preview
features.
We
talked
about
this
a
lot
recently.
A
A
Sorry
I'm
having
some
JavaScript
weirdness
here
show
resource
diffs
for
AKs
clusters,
help
wanted
so
I'm
gonna
assume
if
it's
Help
Wanted
we're
not
putting
on
a
milestone.
D
A
C
A
A
A
C
D
I
think
I
just
put
that
in
the
Milestone,
so
I
am
going
through
the
prq
and
adding
those
to
the
milestone
in
the
background
just
FYI,
so
the
well
release
1.6
Milestone
should
have
some
stuff
in
it.
You
added
this
one
already
I
added
the
P
I'm
going
to
the
pr
queue
so
I
added
the
pr
to
the
Milestone
I'll.
Do.
C
One
actually
Lily
was
wearing
it,
but
I
just
closed
it
because
it
was
fixed
okay
by
John,
Str
ICT,
the
one
we're
talking
about
releasing.
A
Awesome,
oh
yeah
spot
instance.
Eviction
policy
should
be
configurable.
D
So
this
is
really
a
matter
of
if
anyone
feel
free
to
speak
up.
If
you
have
context
as
a
matter
of
just
wrapping,
the
VMS
Azure
vmss
spot
feature
surface
area
inside
our
crd
configuration
surface
area
yeah,
so
I
probably
am
not
allowed
to
do
this
work,
but
this
seems
like
pretty
valuable
stuff.
D
C
C
D
I
feel
like
I
treat
the
Milestone
as
a
default
prioritization
signal
and
because
we
don't
have
a
prioritization,
we
don't
we
don't
prioritize
issues
in
any
other
way.
So
I'm
not
saying
that
that's
how
we
should
be
doing
it,
but
when
I
add
something
to
the
Milestone,
it's
a
way
of
expressing
priority.
D
So
when
I'm,
if
I'm
a
someone
who's
like
okay
I
want
to
pop
an
issue
off
the
queue
there
really
is
no
way
to
do
that
other
than
maybe
personal
preference.
But
if,
if,
if
you
go
to
the
Milestone
first,
you
see
there's
a
help
wanted
issue
in
the
Milestone.
That
would
be
signal
for
that
user
to
do
that,
work.
First
of
all,.
A
Yeah
my
my
personal
take
is
it's
aspirational,
but
should
be
achievable,
and
so
then
I
resist
putting
stuff
on
here
that
we
don't
have
anyone
committed
to,
but
I,
don't
think
it
I
I
don't
have
any
real
problem
with
putting
on
the
milestone.
C
A
D
A
Really
good
this
one,
though,
is
something
I,
think
John
and
I
just
found
when
we
were
looking
through
test
grid.
This
might
be
the
same
bug,
so
it's
actually
killing
the
new
tests.
So
this
needs
to
go
on
the
Milestone
I'm
going
to
say
unless
it's
unrelated,
but
we
just
there's
a
there's,
an
error
about
the
name
being
longer
than
63
characters
for.
D
Right,
but
we
can
we
could.
We
could
fix
that
in
the
test
by
using
fewer
characters.
This
PR
or
this
issue
suggests
that
we
might
programmatically,
truncate
and
and
I
think
that's
more
controversial,
so
yeah.
A
D
D
D
D
E
Yes,
so
are
you
I
started
on
one
of
the
issue
that
Sam
had
been
working
in
past?
That
is
like.
Let
me
paste
that
issue
on
chat.
It's
like
long
pending
one
to
secure
the
bootstrap
data
and
I
was
thinking
like.
Should
we
mark
it
as
like
1.6
Milestone,
because
there
has
already
been
a
PR
raised
by
sham
and
I
think
it
should
be
doable
and
I
know
Cecil.
E
You
have
asked
you
kind
of
raise
a
design
document,
because
the
pr
was
mammoth
so
I
did
some
work,
but
got
sidetracked
with
other
activities
and
I
should
be
like
acting
on
it.
What
do
you
think.
C
C
If
we're
gonna
do
a
puzzle,
I
don't
know
if
we're
gonna
and
given
this
is
a
pretty
significant
change.
I,
don't
know
if
we're
gonna
have
time
to
merge
the
proposal
and
the
pr
given
that
the
release
is
gone
for
a
bit
over
a
month
from
now
and
there's
keep
going
in
the
middle,
but
I
think
we
can
try
yeah.
E
C
D
Matt
you
can
assign
that
top
issue
to
myself
and
if
ashitosh
agrees
to
him
as
well
and
I'll
work
with
you,
Ashton
I
I,
actually
wasn't
aware
that
you
had
an
SSA
Repro
so
that
that,
if
you
have
Cycles
to
be
hugely
helpful
for
you
to
test,
if
you
can't
the
changes
we're
making
in
that
other
PR
that
I'm
driving
sure.
D
D
A
Right
that
was
the
end
of
the
agenda,
but
do
people
have
anything
else
you
want
to
talk
about
before
we
sign
off.
A
All
right,
cool,
we'll
see
everybody
next
week
same
time,
thanks
for
the
help
with
the
Milestone
and
everything
that's
great.