►
Description
No description was provided for this meeting.
If this is YOUR meeting, an easy way to fix this is to add a description to your video, wherever mtngs.io found it (probably YouTube).
A
Okay,
welcome
everybody
good
morning,
good
afternoon,
good
evening
to
this
edition
of
capc
office
hours,
cluster
api
provider.
Azure
is
a
cncf
sponsored
project,
so
we
adhere
to
the
code
of
conduct,
which
means
we
are
all
kind
to
each
other.
We
raise
hands
and
we
want
to
have
an
opportunity
to
speak
if
anyone's
new,
I'm
scrolling
through
the
list
up
here
feel
free
to
say
hi,
but
I
feel,
like
we
all
know
each
other
josh.
A
But
hello,
hello,
again
welcome!
Welcome
thanks,
hey
david
okay
so
looks
like
matt
may
be
back,
but
I
will
carry
on
I'll
go
ahead
and
share
my
browser.
A
A
So
how
about
we
go
down?
These
aren't
ordered
in
any
specific
way.
Maybe
maybe
they're
ordered
chronologically,
but
I
don't
think
they're
meaningfully
ordered
so
I'll
just
go
down
one
by
one
and
maybe
consider
if
we
want
to
keep
it
in
the
milestone
or
drop
it
or
maybe
anybody
wants
to
say
anything
else
about
it.
So,
starting
with
black
car
support
for
capacity.
I
know
this
is
a
fairly
large
one.
Do
we
want
to
keep
this
in
the
milestone,
or
do
we
want
to
punt
it
to
be
next.
D
C
Oh,
no,
it's
okay,
yeah!
So
for
flat
car
we
merged
community
galleries,
which
was
a
pre-req
prerequisite
in
order
to
use
the
community
gallery
image
that
was
published
for
flat
car,
so
we're
getting
closer.
I
think
at
this
point
the
pr
just
needs
rebasing
and
then
using
the
committee
gallery
image.
C
I
wouldn't
say
it's
a
blocker
for
release
if
it
makes
it
great,
if
not,
then
it
will
just
make
the
next
stream,
and
I
think
that
should
apply
to
most
of
the
issues
here
by
the
way.
Unless
we
have
like
a
regression
or
a
bug,
I
don't
think
we
should
hold
the
release
for
anything.
So
I
think
it's
fine
to
keep
it
for
now.
If
it
doesn't
make
it,
we
can
move
it
to
the
next
one.
A
That
makes
sense,
so
it
sounds
like
generally.
The
idea
is
we
kind
of
kick
things
into
v-next
at
the
11th
hour
when
we're
ready
to
cut
a
release,
so
none
of
these
should
be
considered
blocking.
I
mean
unless
I
suppose
there
are
exceptional
cases,
but
it
sounds
like
you've
looked
through
this
list
recently
and
are
confident
that
none
of
these
are
need
to
have
before
we
release
1.4.
Is
that
right.
C
I
think
it's
worth
going
through
the
list
right
now,
but
just
yeah.
Like
anything,
that's
not
a
bug.
I
I
haven't
checked
this
whole
list,
but
anything
that's
not
a
bug.
I
think
we
should
hold
to
that
same
rule.
A
Okay,
cool:
let's
just
go
down
the
list
then
so
the
second
one
I
can
speak
a
little
bit
too.
There
hasn't
been
any
recent
increased
cloud
controller
manager
coverage
across
the
various
tests,
but
it
would
be
fairly
easy
to
sort
of
pivot
and
prioritize
that
based
on
the
helm
chart
we
have
so
that's
the
status
on
that
I
I
don't
have
the
latest
status
on
the
csi
storage
driver
test,
update
that,
let's
see
if
it's
linked
in
this.
A
Well,
we
don't
need
to
get
super
detailed
information,
but
I
know
that's
in
progress
and
mayor
may
make
it
by
1.4
if
we
release
on
july
6th.
So
that's
a
brief
status
of
that
any
questions
on
this
before
we
keep
going
down.
A
Cool
this
cecile,
can
you
give
a
broad
overview
of
the
progress
that
we've
made,
since
I
guess
1.3,
and
maybe
anything
that
might
land
in
the
next
week
or
two.
C
Yeah,
so
I
don't
think
we'll
close
that
one
like
the
epic,
because
we
won't
get
to
skill
set
in
time.
So
I
think
skelsight
is
the
last
one
remaining,
but
we
have
public
ips
and
so.
C
C
A
Great
great
okay,
cool
that
makes
sense,
so
this
one
has
been,
I
think,
just
basically
participating
in
lots
of
releases
over
the
past
several
months,
because
it's
an
epic
as
you
say,
and
it
looks
like
we've-
got
the
specific
issues
for
that,
so
that
kind
of
takes
out
three
issues
from
this
list.
A
So
the
next
one
is
node
image
upgrades
to
aks
node
pool.
Let's
see
what
we
can
is
there
a
pr
that
links
to
this
looks
like
it's
essentially
not
being
worked
on.
Is
that
true.
C
Yeah
sounds
like
it.
Maybe
we
should
unassign
the
person
who's
assigned
right
now
and
put
it.
A
A
C
A
C
A
A
A
A
A
A
C
A
Okay,
multi-tenancy
within
one
manager
instance
using
using
azure
user
assigned
identities.
C
Yeah
that
one
has
a
pr
the
pr
is
being
reviewed.
I
saw
there
were
comments.
I
need
to
address
the
comments
and
but
yeah,
I
think
it's
fine
to
leave
it
there
for
now.
Oh.
A
Yes,
I
am
familiar
with
npr
cool
yeah
that
one's
almost
done
great,
so
we'll
leave
that
in.
Thank
you
for
doing
that.
That
pr
looks
good
by
the
way
and
matt
do
you
want
to
talk
about
this
issue?
This
sounds
like
that's
related.
A
D
This
is
ready
to
merge.
Okay,
cool
james
just
had
a
comment
at
the
end
about
we
should
introduce
some
more
logs,
so
I
opened
a
different
pr
for
that,
but
this
should
be
able
to
merge
right
now.
I
think.
A
Okay,
awesome
so
that
all
sounds
like
we're
in
control
of
everything.
D
D
C
An
issue
into
the
milestone,
as
you
were
doing,
but
I
just
added
the
public
ip
tags
which
some
user
internal
user
asked
for,
and
so
I'm
working
on
that
I'm
gonna
have
a
pr
open
today
or
tomorrow.
A
Okay,
wow,
that's
nice!
That's
great!
I
know
that
that's
been
asked
for
a
long
time.
So
well
done
on
that
note.
I
was
thinking.
Maybe
the
next
thing
we
do
is
go
through
just
the
issue
queue
generally
and
quickly
see
if
there's
anything
that
should
be
in
the
milestone
that
we
forgot
to
tag.
Does
that
make
sense,
or
has
that
been
done
very
recently?
C
That
sounds
good
or
even
prs.
Maybe
I
don't
know
if
we
wanted
to
peers
and
issues
or
just
issues.
A
Okay
cool:
I
have
a
bad
habit
of
opening
up
pr's
that
don't
link
to
issues.
So
let's
let's
go
in
order
here,
so
I
don't
think
we
need
to
consider
this
as
a
change
on.
A
Yeah
cool,
I
agree
all
right,
so
this
is
already
accounted
for.
This
is
new
and
I
wouldn't
be
surprised
if
this
merged
in
the
next
couple
of
weeks.
I
think
this
maybe
there's
more
complexity
here
than
I'm
than
I
am
anticipating,
but
the
pr
is
in
a
really
good,
like
75
percent
their
state
and
I've
already
started
reviewing
it.
So
I
think,
assuming
that
that
brian
is
serious
about
getting
emerged,
I
would
guess
it'll
merge
before
the
milestone.
Is
it
worth
tagging
this
or
is
it?
C
Yeah,
I
think
so
for
context
that
one
is
actually
something
like
a
lot
of
people
want.
So
if
we
can
get
it
in
soon,
it's
gonna
unblock
azure
cni
without
having
explicit
support
for
azure
cni.
So
it's
a
way
of
supporting
pre-allocating
ips
for
pods,
and
so
that
is
a
prerequisite
for
azure
cmi
right
now
with
self-managed
clusters.
So
yeah,
if
you
know,
as
as
long
as
the
reviews
go
well
and
the
author
is
responsive,
I
think
we
should
be
able
to
get
it
in
soon.
C
C
Formally
so
we
actually
have
so
we
have
a
milestone,
bot
automation,
where
every
pr
that
merges
gets
tagged
with
the
current
the
latest
milestone,
and
so
one
of
the
release
like
tasks
when
you
do.
A
kind
of
release
is
to
go
update
the
milestone
applier
so
that
it
will
start
tagging
with
the
next
milestone
that
we
created
so.
C
A
Got
it
well,
I
think
if
we
feel
like
we
can
support
the
contributions,
then
it's
it
might
level
up
the
sort
of
motivation
and
enthusiasm
around
it.
So
it's
worth
doing.
A
A
I
mean
at
the
expense
of
potentially
adding
more
pressure,
but
or
maybe
the
other
way
to
put
it
is
we
don't
want
to
sort
of
set
expectations
too
highly
where
if
this
does
take
longer,
folks
are
disappointed
that
it
doesn't
get
included
in
1.4
if
it
gets
kicked
out,
but
I
I
think
I
think
we're
fine.
So,
let's
I
would
say
my
vote
is
to
tag
it.
Does
anybody
descend.
C
A
C
C
They
will
set
it
on
the
pr
when
the
pr
merges
issue
is
usually
like
when
we
want
to
check
something
to
do.
That's
not
done
yet.
I
don't
know.
If
there's
a
convention,
though,.
A
A
All
right,
okay,
cool
so
back
to
the
prq.
This
is
also
something
that
landed
recently
that
I
think
will
will
land
soon,
I'm
not
sure.
If
there's
an
issue
associated
with
it.
Let's
see
there
is
so
I'm
going
to
advocate
that
we
tagged
this
with
a
milestone
as
well.
Does
anybody
want
to.
A
A
E
Only
oh,
let's
see
I
just
related
to
ci.
This
is
related
to
the
make
file
cleanup.
So
I
don't
really
know
if
it
falls
into
ci.
A
C
Yeah,
I
think
that
one
is
developer
focused
like
deflo
ci
scripts,
so
not
user
facing
cool.
A
All
right,
thank
you
for
working
on
that
nawaz.
The
next
one
is
mine.
I
will
tag
this
as
a
milestone.
A
We
should,
I
should
actually
file
a
bug.
I'm
going
to.
A
A
I'm
going
to
just
to
make
this
quick.
Where
is
the
cuban
shoe
cool
I'm
going
to
link
to
this
cube,
adm
issue
and
then
I'll
fill
up
more
clustering
per
capstick
specific
details
async,
so
you
guys
are
watching
and
so
I'll
submit
this
issue
incomplete
and
I
will
tag
it
with
milestone.
A
So
we
definitely
want
to
fix
this
before
1.4
or
do
it
cecil.
Is
this
something
that
we
want
to
track?
Would
you
consider
this?
I
can
extend
it's
a
bug,
so
I
think
we
want
to
track
this.
A
I
don't
know
how
many
users
are
running
latest
in
production,
but
it's
still
well,
it's
a
test,
bug
yeah
and
we,
I
also
don't
think
we
fully
understand
the
scope.
So
until
we
do
it's
fair
enough
to
just
call
it
a
bug,
cool,
so
I'll
tag
this
with
the
milestone
and
then
I
will
edit
the
pr
that
I'm
working
on
to
reference
this.
A
A
Let's
go
to
the
issue:
do
you
anticipate
landing
this
in
the
next
week
or
two.
C
I
think
it
was
pretty
close.
I
just
need
to
review
from
what
I
saw.
A
C
A
Thank
you.
We
talked
about
this.
This
is
a
sort
of
epic
pr
that
will
not
merge
anytime
super
soon
make
public
ips
reconciles.
We
basically
talk
about
that.
A
A
A
C
A
C
Are
we
testing
windows
external
cloud
provider
with
the
helm
chart
in
our
external
cloud
provider
test
that.
A
A
Sorry
template
because
we're
not
using
machine
pools.
So
actually,
let's
look
for
our
kinds.
We've
got
four
matches,
so
that's
a
good
sign,
there's
two
pools,
so
we've
got
our
control
plane
machine!
Oh
no!
It's
like!
I
am
wrong.
At
least
this
test
is
only
linux
if
I'm
going
too
fast.
Let
me
know
I'm
super
familiar
with
this,
so
I
may
be
going
too
fast.
B
Yeah,
it's
gonna
say
you
have
to
be
enabling
those
so.
C
A
A
I
think
this
is
probably
happening
in
ci
entry
point,
sorry
to
folks
who
are
have
no
idea.
What
I'm
doing
this
is
like
super
in
the
weeds
of
stuff
that
I'm
familiar
with
recently,
but,
as
cecile
said,
we
there's
a
lot
of
different
test
sort
of
vectors
and
the
the
that
template
with
the
ci-version
suffix
indicates
that
a
template
is
designed
to
accommodate
sort
of
custom,
ci
built
bits
of
kubernetes
and
of
and
potentially
cloud
provider.
A
A
When
we
choose
this
template
based
on
the
passing
in
of
a
ci
version
and
then,
if
they're,
combining
ci
version,
which
is
a
variable
that
represents
a
kubernetes
ci
built
set
of
bits
if
they
combine,
if
you
combine
that
with
this,
then
we're
going
to
do
a
bash
substitution
that
substitutes
pro
for
pro
external
cloud
writer,
and
so
the
template
is
going
to
eventually
explode
to
the
one.
We
were
looking
at
prior
external
cloud,
credo-ci
version,
so
it
would
take
a
little
bit
of
detective
work
to
identify
exactly
which
tests
are
running
this.
A
But
I
think
seal
is
right
that
these
are
the
tests
used
by
the
club
fighter,
azure
maintainers,
to
get
signal
on
the
cloud,
controller
manager
and
cloud
node
manager
bits
as
they
develop
them.
So
they
will
use
that
template
to
build
a
capsid
cluster
running.
Both
linux
and
windows
pools
running
out
of
tree
cloud
fighter.
C
No
question,
but
this
sounds
good
and
then
yeah
I
think
maybe
we
should
just
comment
on
the
issue
saying
we
should
be
testing.
You
know:
windows
with
external
cloud
provider
and
cloud
providers,
so.
C
B
A
But
I
yeah,
I
agree
with
james
that
there
is
a
possible
race
condition
here,
which
I
would
think
that
kubernetes
itself
should
take
care
of
yeah,
so
that
would
be
concerning
if
that
race
condition
was
terminal.
A
The
race
condition
being
like
the
cloud
node
manager
stuff
arrives
before
there's
scheduled
networking,
and
so
there's
no
node
to
schedule
it
onto
that
should
eventually
reconcile,
as
the
cni
bits
aren't
actually
installed
by
the
q
proxy.
Weird
thing
we
do
in
windows
and
then
sorry
to
call
it
weird.
A
It's
confusing
to
me
once
that's
done,
then
the
node
should
turn
ready
and
then
the
cloud
node
manager
bits
should
be
scheduled
successfully.
A
C
C
A
Yeah,
I
mean,
I
think,
if
we
have
our
github
foo
nicely
dialed
in
doing
these,
slash
milestones,
slash
whatever
can
be
really
helpful
to
the
community,
because
we're
sending
messages
about
the
way
that
we're
responding
in
real
time,
the
prioritization
et
cetera,
et
cetera,
so
that
we
should
definitely
do
this
regularly
every
two
weeks,
or
so,
I
think
so,
yeah.
If
we
landed
on
office
hours
where
it
was
all
contributors
we
probably
would
not
want
to
take.
We
wouldn't
have
the
time
to
take
this,
so
we
need
to
schedule
some
alternate
time.
D
So
somewhat
related
to
that
we
should,
we
should
see
if
we
can
scare
up.
We
had
a
plan,
ideally
of
trying
to
have
a
demo
or
something
interesting
to
make
sure
people
want
to
come
to
the
meeting.
Maybe
we
should
plan
on
or
at
least
see
if
we
have
any
ideas
for
something
we
could
do
two
weeks
from
now
by
the
way
we
have
a
demo
or
something
that
we
could
put
a
headline
on
the
meeting.
D
Think
of
something
we
could
do,
I
could
I
could
maybe
demo
azure
machine
pools
by
then,
but
it's
not
working
yet
I
mean
machine
pull
machines.
A
If
we,
if
we
get
the
the
aks
windows,
node
type,
merge,
we
can
demo
building
an
aks
cluster
with
windows
nodes
on
it.
That's
not
possible
right
now.
Did
somebody.
A
Out
yeah
yeah,
the
contributor
okay
yeah,
just.
B
A
B
Oh
cool,
was
it
luther.
B
Again,
he's
active
in
sig
windows.
He
was
from,
he
was
at
a
rancher
and
just
moved
to
another
company.
So.
A
Cool
cool
yep.
C
And
I
think
an
aks
cluster
demo
would
be
really
cool
and
might
be
even
better
if
we
can
get
one
of
the
actual
users
to
you
know,
show
it
off
or
maybe.
E
C
Do
like
a
pair
like
pair
on
it,
like
you
know,
one
of
us
plus
someone
who's
using
it.
A
C
Another
topic
that
I
could
show
at
some
point:
it's
nothing
new,
but
just
it's
a
question
that
comes
up
a
lot
is
just
the
like:
all
the
networking
setup
and
all
the
stuff
you
can
do
with
like
bringing
your
own
v-net
and
bringing
your
public
id
and
how
like
cabzi
handles.
C
All
of
that
like
how
does
it
know
which
life
cycle
like
which
resource
life
cycle
it
owns
versus
things
that
you
like
infrastructure,
that
you
bring
pre-existing
and
just
like
how
all
that
stuff
is
configured,
but
that's
more
like
a
you
know
later,
you
know
if
we
want
some
additional
demo
and
we
as
a
backup.
A
A
You
know
as
a
pr
viewer.
Sometimes
you
get
these
very
interesting
scenarios,
but
you're
not
quite
sure
why
these
scenarios
exist
and
how
they
relate
to
other
alternate
scenarios.
So
that
would
be
really
good.
I,
what
I
would
love
to
see
honestly
is
an
hour
long
demo
like
a
long
demo,
where
you
create
a
cluster
and
then
you
just
completely
change
it
like
30
times
and
watch
the
transformation
of
the
cluster
over
time.
A
This,
maybe
we
should
do
like
a
private
demo
to
uncover
gremlins
and,
like
you
know,
maybe
at
the
end
of
that
session
comes
like
six
bugs,
but
that's.
I
would
like
to
see
more
demos
around
that
for
the
community
for
folks
to
really
understand,
especially,
I
was
talking
with
you
still
offline
about
this.
A
lot
of
folks.
They
get
their
cluster
running.
A
They've
got
all
their
production
workloads
and
they
want
to
keep
that
cluster
around
for
as
long
as
possible
and
so
being
able
to
demonstrate
the
sort
of
mutation
capabilities
of
cluster
api
and
capacity,
I
think,
would
be
very
compelling.
C
Yeah,
I
think
that's
that's
a
great
like
introduction,
no,
not
introduction,
but
like
folks
who
are
not
familiar
with
cluster
api
demo.
Like
oh
here's
what's
appealing,
and
you
know
how
like
what
it
brings,
I
feel
like
regular
contributors
and
users
of
cabzi
are
already
past
that
point
where
they
think
they
know
that,
like
that's
what,
but
that
like
that
would
be,
I
think,
that's
a
really
good
one
to
bring
to
a
larger
audience,
not
just
the
office.
Hours
is
what
I
mean.
A
Cool,
I'm
not
sure
that
I
agree
100
with
that,
like,
I
think,
say
someone
who
built
their
cluster
with
the
cluster
resource
set
calico
and
it's
been
sitting
there
for
nine
months
and
demoing
how
you
can
rip
that
apart
and
put
on
a
helm
chart
using
the
official
calico
like
this
is
just
an
example.
We
I
can't
do
this
right
in
a
second,
how
you
could
basically
upgrade
their
cni.
A
A
Because
a
lot
of
these
things
are
exposed
exactly
that's
exactly
what
I'm
talking
about
like
if
I'm
a
naive
cluster
api
user-
and
I
see
okay,
cool
I've
got
these
abstractions.
I've
got
these
like
declarative
specs.
What
happens
if
I
change
this
this
configuration
from
this
to
this?
I
would
expect
cluster
api
to
magically.