►
From YouTube: Kubernetes Cluster API Azure 2019-10-24
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
Instead,
hello,
everyone,
this
is
October
24th,
it
is
a
it
is
the
cluster
API
Azure
meeting.
This
is
a
meeting
that
is
recorded
and
available
on
the
internet.
So
please
be
mindful
of
what
you
say
do
please
be
sure
to
adhere
to
the
kubernetes
code
of
conduct
and
just
in
general,
be
excellent
people.
So
we
don't
have
too
much
on
the
agenda,
so
I
figured.
What
we
would
do
is
maybe
roll
through
some
of
the
things
that
we
have
been
working
on.
B
C
A
C
The
note
and
right
now,
I'm,
trying
to
I,
have
my
own
Fork
and
making
things
work
whenever
I
have
an
idea
of
what's
required.
Oh
I'll,
be
more
engage
more
with
you,
guys,
okay,.
D
A
E
So
sorry
I
was
not
able
to
add
the
document.
I
don't
have
the
XS
right
for
that,
so
I'm
from
Maine,
where
so
I'm
working
also
like
new
project
for
employee
carbonating
on
the
cluster,
so
I'm
very
interested
in
the
topic.
That's
why
I'm
joining
today
to
see
like
to
get
more
information
about
the
project
and
to
see
what
I
kept
it
from
there
awesome.
F
Think
I
can
go
so
my
name
is
Cali
a
super
Mayan
I
work
at
Microsoft
and
I'm
on
the
container
networking
team.
So
mostly
what
I'm
looking
into
is
providing
Windows
support
for
a
cluster
API
and
Azure,
because
I
did
the
wit,
cube
ADM
work
in
Windows
to
bootstrap,
Windows
nodes
and
I
own
the
script.
That's
in
SiC
windows
tools
right
now,
which
is
the
official
kubernetes
supported
way
of
bootstrapping
a
Windows
node.
So
I
just
want
to
integrate
that
here.
Awesome.
G
A
A
A
Alright,
well
so
I
guess:
I
should
introduce
us
a
few
of
us
I
Steven,
Augustus
I'm,
a
senior
cloud
native
architect
at
VMware
I
work
on
one
of
the
admins
maintainer
zuv
cluster
API
Azure.
Also,
the
product
management
chair
and
sig
release
share
for
the
kubernetes
project,
so
hello,
everyone
we've
also
got
Cecile
Wan,
Li,
Nick
and
David
ooh,
so
Nick.
Is
that
way,
fair,
Cecile
and
everyone
else
I
mentioned
are
at
Microsoft.
A
A
B
A
A
A
Matching
manifests
okay,
so
this
is
possible.
Go
recently
had
an
update
from
well
well
from
one
ten
to
one
eleven
and
then
one
twelve,
a
set
of
incremental
future
fixes,
as
well
as
some
CV
addressing
some
CDs.
So
what
likely
happened?
Was
we
tried
to
bump
this
before
before
the
images
were
available?
The
images
lagged
behind
the
actual
release.
So
that's
probably
what
that
was.
I
I
A
I
A
So
that's
before
it
makes
it
into
master,
but
this
is
like
the
primary
board.
If
we
got
to
a
point
where
we
had
so
many
tests
like,
we
were
doing
lots
of
periodic
and
post
submit
stuff,
then
maybe
we'd
consider
doing
separate
tabs,
but
I
think
that
you
know
this
is
a
small
enough,
a
small
enough
set
of
tests
to
watch
on
a
single
group.
A
To
give
you
an
example
of
something:
that's
you
know
this
may
be
more
involved
and
sig
release
master
blocking
has
a
few
more
tests.
As
we
look
at
like
sig
release
116
all
you
can
see
that
there
are
quite
a
few
tests
right
and
that
makes
sense
breaking
some
of
these
out.
These
are
not
our
fault
by
the
way,
so
ignore
ignore
the
GCE
GK
test,
but
you
can
see
we're
not
dealing
with
a
lot
right
now
and
then
the
the
Basel
ones
are
essentially,
they
will
disappear.
A
They'll
fall
off
this
board
eventually,
because
one
we've
removed
basil
from
from
the
current
master
branch,
the
branches
that
are
using
bait,
the
branch
that's
using
basil,
the
release,
zero
to
branch
now
uses
the
PR
test.
Pr
test
and
PR
build
results,
so
those
will
report
into
to
view
those
sections
in
this
bed.
So
these
will
all
of
these
based
on
ones,
will
eventually
fall
off
the
board
cool.
A
Cool
so
we're
gonna,
do
a
quick
project
board
review.
I
know
that
we
recently
had
a
few
more
things
pop
up
on
the
project
board
or
pop
up
on
the
repo
that
I
want
to
make
sure
we
capture
in
the
project
board.
So
what
I
usually
do
here
is
just
do
a
add
cards
right
and
we
can
see
that
we've
got
a
few
things
that
are
not
not
captured
on
the
project
board
and
I'm,
going
to
open
those
all
up,
because
I
want
to
close
a
few
of
these.
I
Context
reopened
a
bunch
of
these,
so
she's
gonna
be
onboarding
and
coming
to
help
us
p.m.
some
of
this
stuff,
and
so
we
were
just
trying
to
like
go
through
the
repo
and
add
some
good
issues
for
things
that
we'll
need
to
do
in
the
future.
But
these
might
need
more
details.
Yeah.
A
A
A
A
So
here's
the
thing:
the
flow
for
spinning
up
a
cluster,
a
control
plane,
the
first
control
plane
node,
has
to
come
up
first
right,
the
first
control
plane
node
has
the
cube,
ADM
and
knit
right.
So
the
cluster
does
not
exist
until
that
comes
up
right
and
then,
whatever
order.
Whatever
order,
the
next
two
control
planes
come
up
in
is
not
really
relevant.
I.
J
Have
a
habit,
I
have
a
potential
trick
that
we
can
do
here.
That
would
enable
it
would
allow
us
to
kick
off
all
the
all
the
provisioning
at
once.
Okay
and
then
still
serialize
the
control
first
control
plane,
but
the
BMS
will
be
ready
to
go,
and
so
the
gist
of
it
is
that
you
don't
use
the
bootstrap
data
right
away.
We
have
some
pre
bootstrap
data
and
what
that
amounts
to
is
that
is
talking
to
the
metadata
endpoint
and
periodically
asking
it.
You
know:
do
we
have
updated
custom
data,
and
so
what
would
happen?
J
J
F
A
I
We
opened
this
second
one
on
purpose
yeah,
oh
the
first
one.
The
reason
is
that
the
first
one
has
like
very
detail
like
basically
implementation
details
on
how
it
should
be
done,
which
we
think
is
maybe
not
relevant
anymore
solution
is
dated
back
in
March,
so
we'd
like
to
start
with
a
proposal
of
how
it
should
be
done,
and
so
it
was
kind
of
suit
outlined
that,
rather
than
you
could
say,
like
Windows
support
like
as
if
your
project
so.
I
F
A
F
A
F
F
A
A
D
J
F
A
F
A
F
B
A
A
A
J
Yeah
I
can
just
give
a
brief
rundown.
So
so,
after
hat
as
your
aad
as
well,
I
guess
two
categories
of
identities.
One
is
service
principle,
the
other
is
managed
identity,
so
manage
identity,
actually
has
two
types
of
managed
identities
and
and
and
that's
user
assigned
manager
at
Indian
system
assign
manage
identity.
J
The
problem
there
is
that
with
systems
assign
management-
and
those
are
each
individual,
each
of
each
VM
has
its
own
distinct.
Credential.
That
needs
to
be
assigned
a
particular
set
of
roles,
and
so,
for
so
user
assigned
is,
is
more
convenient
for
the
kubernetes
use
case,
because,
let's
say,
for
example,
you
create
one
user
assign
manage
at
any
for
the
control
plane
and
that
buying
a
role
to.
H
A
H
A
H
I
see
Till's
going
away
yeah.
H
A
J
A
J
A
There
are
more
things
that
we
need
to
care
about:
I
guess
so
for
control,
planes,
I
mean
cloud
provider,
interaction,
control,
planes,
managing
load,
balancers
things
of
that
sort.
We
would
still
want
some
sort
of
identity.
Definitely
what
about
workers
workers,
not
so
much
I,
guess
outside
of
CSI
I
guess
outside
of
the
entry
required
storage
stuff,
once
I
guess,
once
we're
fully
out
of
tree
on
the
cloud
provider
Azure
side,
then
that
story
starts
to
change,
but
we
probably
still
need
to
have
some
sort
of
identity
there.
A
J
A
Mean
there
are
also
there
are
also
tools
and
things
that
we
have
not.
You
know
inventoried
that
may
be
may
be
dependent
on
service
principle,
so
we
we
don't
want
to.
We
don't
want
to
wholesale,
remove
that
stuff
cool
all
right,
so
that
is
yeah.
I
definitely
feel
like
that's
a
that's
a
new
milestone
thing
and
should
be
simple
to
a
couple,
but
we'll
see
all
right:
okay,
zoom,
it's
passing
around
hosts!
That's
awesome
all
right!
So
this
one
concerns
me.
K
For
debugging
purposes,
it's
really
nice
to
be
able
to
see
trace.
You
know,
what's
going
on
in
your
code
and
having
metrics
output
from
the
code
as
well.
So
if
started
early
enough
instrumenting
the
code
with,
like
you
know,
open
telemetry
or
open
tracing,
it's
it's
actually
pretty
clearly
able
to
get
those
basic
parts
in
in
place.
A
A
J
J
You
know
with
that
with
that
additional
tracing
that
is
more
granular
than
just
you
know,
like
hey
V
net
with
I
mean
we
know
which
mean
it's
gonna.
It
is
but
or
hey
a
vienna's
provision,
but
we
really
go
to
context
like
oh
hey
control.
Plane
is
taking
this
long
versus
regular
worker
node
is
taking
this
long
yeah.
A
So
so
one
thing
I
would
caution.
There
is
to
not
wire
too
much
functionality
that
only
exists
in
the
azure
side.
Again,
like
you
mentioned
that
the
the
tracing
stuff
is
available,
I
know
the
tracing
stuff
is
available
because
I
we
run
into
dependency
issues
with
the
tracing
stuff.
That
was
only
recently
fixed,
but
we
need
to
make
sure
that
we're
providing
functionality,
that's
consistent
across
providers
as
well,
so
I
will
I
will
call
this
out
as
important
but
important
long
term
and
milestone
next.
Unless
someone
feels
super
passionately
about
it,
yeah
talget.
K
I
A
I
A
We
need
to
do
something
very
similar
to
what
we
do
for
I
believe
we
need
to
do
something
very
similar
for
what
we
do.
As
for
the
availability
zones
right,
so
we
need
to
see
if
the
type
supports
enabled
supports
accelerated
networking,
and
only
if
it
supports
it,
then
considering
passing
that
into
the
the
VM
spec
yeah.
I
So
in
cases
per
diem
size,
and
so
we
can
which
emphasizes
support
it,
because
it's
it's
by
like
number
of
CPUs,
but
unfortunately
there
isn't
a
straightforward
API
that
we
can,
just
like
you
know,
call
and
get
the
list
of
supported,
vm
spec,
or
at
least
I
haven't
found
one.
That's
the
biggest
difficulty
with
this
issue
is
finding
how
to
enable
it
only
first
support
it
games
that.
I
A
A
This
might
not
be
a
type
that
has
it,
but
accelerated
networking
enabled
false
right.
So
we
can
say:
okay,
here's
the
tricky
part
I
think
there
is
an
API
to
do
this.
It's
not
exposed
in
our
SDK
David.
Do
we
have
updates
on
that
or
I
should
just
read
the
issue?
Okay,
you
know
what
dates
are
their
internal
updates
regarding
that.
K
K
H
A
All
right,
yeah,
so
I
feel
like
we
can.
We
can
use
the
same
pattern
that
we
were
using
for
for
availability
zones
to
do
this.
What
we
do
need
to
do
is
I
know
that
you
said
the
availability
zones
thing
is
fixed.
I
would
like
for
us
to
still
wire
up
the
code
to
check
the
the
provider
capabilities
on
the
VM
type
before
removing
that
kind
of
hard
coded
truck
that
I
have
in
there
that
hard-coded
slice
I
have
in
there
for
the
VM
types.
A
A
A
J
A
So,
like
south
central
us
I
think
was
the
example
was
a
region
where
availability
zones
are
being
turned
up
so
yeah,
so
it
responds
that
availability
zones
are
available,
but
there's
a
secondary
API.
That
tells
you,
if
they're
actually
available
for
that
VM
type
at
that
time.
So
I.
B
K
H
A
H
A
I
Yeah
I'm
still
not
sure
about
how
we
should
do
this
notice
yeah.
Why
I've
made
that
comment
towards
so
right
now
what
kappahd
is
doing
is
they're
just
taking
whatever
latest
public
image
exists
for
that
kubernetes
version
so
and
they're
not
pinning
to
a
specific
version
of
the
image.
I,
don't
know.
If
we
should
follow
this
English,
you
have
a
version
that
we
update
in
the
repo.
The
like
there's
pros
and
cons.
Words
are
the
main
thing
I
was
thinking
up
is
if
you
learn
latest.
I
I
A
A
I
There
is
a
preview
stage
in
image
replication
process
in
which
it's
only
available.
Theoretically
to
a
few
subscriptions
when
I
publish
the
first
initial
I
wasn't
able
to
access
it
from
my
previous
subscription,
so
I'm
working
with
the
marketplace
team
on
that
to
see
if
it's
supposed
to
be
working
or
if
I
was
doing
something
wrong
or
what
but
yeah.
If
that
was
the
case,
then
we
could
easily
test
it
before
we
hit
the
light,
which
would
be
I,
think
that's
what
it's
meant
to
be
used
for.
Okay,.
A
Cool,
so
I
think
that
I
mean
something.
To
also
consider
here
is
that
we
claim
support
for
multiple
versions
of
kubernetes,
so
we
can't
hard
code
I
can't
hard
code
anything
there.
We
have
to
be
able
to
do
some
sort
of
latest
discovery,
so
we
would
key
in
on
yeah
exactly
what
you
said
key
and
on
the
latest
image
for
that
case.
Well,.
I
I
A
So
I
will
leave
that
to
you
and
all
right.
So
we
have
a
few
minutes
left
I
think
it's
sorry
go
for
it.
I
would.
I
I
J
I
A
Not
only
not
when
we
do
the
first
release,
but
you
know
I,
guess:
I
I
think
you
know
I
think
feet
when
alpha
2
is
just
going
to
be
our
stopgap
again
like
we
stopped
development
on
V
1
alpha
1,
we
should
mm-hmm
yeah.
We
should
we
should
plan
what
we
what
we
definitely
want
to
land
in
this
one.
So,
let's
put
it
I'm
gonna
leave
the
milestone
off
for
now
or
do
you
want
to
start
with
next
and
see
where
it
goes?
Yeah.
I
A
Any
next
few
days,
yeah,
we
were
waiting
for
that
image.
Stuff
Landon
looks
like
it
has
I've
been
I
was
out
for
a
few
days
for
coffee,
six
nests
sickness,
but
yeah.
We
should.
We
should
Circle
up
and
figure
out
what
we
want
to
do
there
I
yeah,
I,
think
I.
Think
the
the
image
stuff
was,
the
image
image,
stuff
and
documentation
which
I'm
I'm
gonna
be
working
on
is
is
the
last
piece.
A
J
A
B
A
Sleep
dues,
I'm
kind
of
happy
with
the
dues
that
are
in
there
to
be
honest,
but
I
can
say
that
on
the
issue.
A
H
H
I
I
was
just
gonna,
say:
David
and
I
still
have
one
last
thing.
We
need
to
do
regarding
the
image
before
we
release,
ideally
I.
Think
because,
basically
right
now,
our
publisher
is
not
a
first
party
publisher,
so
it
means
it's
a
bit
complicated.
There
basically
means
that
every
user
that
uses
the
image
for
the
first
time
will
have
to
accept
the
legal
terms,
and
we
have
worked
with
the
team.
That
was
the
publishing
to
get
us
an
exception
for
that,
so
we're
first
party,
so
that
our
users
don't
have
to
do
that.
All.
I
A
Cool
just
making
sure
so
anyone
who's
on
the
call.
Please
be
aware
that
you
can
use
the
image
or
you
can
build
your
own
image
via
the
the
image
builder
or
you
can.
No,
you
can't
go
that
route
anymore,
but
yeah.
You
can
use
the
image
or
you
can.
You
can
build
your
own
image.
We
have.
We
have
tools
for
that.
A
I
H
All
right
so
less
time,
last
time,
I
remember
us
talking
us
talking
about
getting
to
the
place
where
we're
ready
for
more,
like
freeform
contributions.
How
are
we
feeling
about
that?
I
know
that
we're
waiting
a
little
bit
I
think
maybe
viewing
off
of
three
is
kind
of
what
we're
landing
on,
but
I
just
wanted
to
bring
that
up
again.
Yeah.
A
No
I
I
feel
like
V
1,
alpha
2.
Actually,
people
can
get
started.
I
had
a
I
had
a
great
discussion
with
one
of
our
customers
today
about
cluster
API,
yeah
sure
and
basically
I
mean
basically
we
need.
We
need
customers
using
this
right.
We
need
for
varying
definitions
of
the
word
customer
right.
We
need
people
consuming
it
that
are
not
service
providers
right,
so
we
can
have
a
true
picture
of
the
world.
A
A
A
H
A
We
can
we
can
start
opening
this
up,
so
I
want
to
make
sure
that
Doc's
are
super
tight
for
development.
We
send
a
blast
wide.
We
can
do
it
to
ke
dev
and
a
cloud
provider
and
cluster
lifecycle
and
Twitter
and
LinkedIn,
and
all
that
good
stuff
and
get
people
get
people
active,
but
I
want
to
make
sure
that
the
people
who
get
active
are
not
filing
issues
like
hey
I
was
going
through
development,
Docs
and
I.