►
From YouTube: CAPZ 2020-04-03
A
B
C
E
So
what
we
do,
the
advanced
form
is
providing
managed
kubernetes
for
our
customers
and
we're
also
buying
at
a
national,
and
we
find
our
API
project
very
interesting.
So
I
would
just
like
to
catch
up
a
little
bit.
What
you
folks
are
up
to
and
what's
the
correct
set
of
things,
so
we
are
thinking
about
adopting
faster
API
in
what
way
and
how
much
and
how
we
are
not
exactly
sure
we're
just
still
investigating
the
idea.
So
we
will
be
probably
joining
the
meetings
in
the
future.
More
at
least
I
hope
I
will.
F
G
Okay,
should
I
talk
go
for
it.
This
is
my
second
meeting,
sorry
about
the
messy
background.
I
haven't
had
time
to
clean
because
I've
been
working
about
18
hours
a
day
lately
so
I'm
part
of
the
azure
monitor
team.
I,
don't
know
if
you
guys
know
what
a
virtual
monitor
is.
You'd
probably
do
that
said.
This
is
my
second
meeting
we're
getting
more
interested
in
possibly
moving
our
cluster
management
infrastructure
to
cluster
API,
because
we
have
our
current
cluster
management
and
deployment
infrastructure
on
an
RD
prototype.
G
I,
don't
know
if
you
guys
know
what
that
is.
I,
don't
know
if
I
can
talk
about
it
here,
but
we
got
our
infrastructure
paid
to
that
with
ace
working
on
it
and
we
would
rather
move
our
infrastructure
to
the
more
mature
version
or
hopefully
soon
to
be
more
mature
version.
Cluster
API
and
possibly
I
would
like
to
contribute
it
possible
because
I
have
some
knowledge
about
the
environments,
we're
deploying
into
eventually.
B
H
D
D
B
D
You
had
an
incident
I
didn't
give
I'm
and
she's
done
this
like
either
during
a
meeting
or
before
meeting,
and
it's
on
my
carpet
so
I
want
to
see
it
as
soon
as
possible,
but
yeah
thanks
guys
thanks
to
seal
for
doing
all
the
intros.
So
we
have
a
couple
things
on
our
discussion
list.
I
want
to
do
the
milestone
last,
since
that's
just
going
through
issues.
So,
let's
start
with
Azure
mist.
Imagine
machinable
discussion
with
david.
If
that's
okay
with
you
yeah
it'd,
be
super.
I
Cool
yeah,
so
we
have
actually
a
couple
really
interesting
things
out
there
for
Michigan
cool
and
managed
clustered
related
stuff
is
he's
not
here
today.
Now,
there's
there's
a
work
in
progress
for
managed
clusters.
No
I
would
definitely
check
that
out.
If
you
get
a
chance,
I
would
love
to
get
some
feedback.
There,
I'll
probably
be
taking
a
look
at
that
PR
a
little
bit
later
today,
but
also
a
PR
out
there
for
machining
pools.
It's
really
early
and
I
wanted
to
get
it
out
there
so
that
we
could
get
some
feedback.
I
Basically
it
takes.
It
builds
upon
the
Machine
cool
functionality
in
capi
and
ends
up
with
agile
machine
cool,
backed
by
VM
scale.
Sets
there's
still
work
to
be
done
in
the
networking,
there's
like
work
to
be
done
and
tests,
but
but
it
creates
a
experimental
folder
similar
to
you
the
way
that
was
done
in
Kathy
and
separates
out
the
functionality
that
way.
I
would
love
to
hear
pros
and
cons
of
how
that
works.
There's
there's
a
lot
of
code
refactoring
that
that
could
take
place
right
now.
I
This
is
basically
replication
of
code,
but
we
could
definitely
dry
up
a
lot
of
stuff
between
machine
and
machine
tool.
Yeah
just
just
generally
wanted
to
get
some
feedback
on
it.
I'm
going
to
continue
working
on
it
and
as
as
feedback
comes
in
definitely
start
to
incorporate
that
as
well.
But
yeah.
Let
me
know,
let's
see.
Oh
sorry,
I
thought
God
declare
was
chiming
in
on
that
anybody
have
any
questions
about
as
your
machine.
I
Available
to
set
flag
supported
through
that,
no
there
is
we're,
not
we're
not
taking
in
zones
right
now.
We
definitely
could
what
I?
What
I
wanted
to
do
is
I
wanted
to
have
the
minimal
feature
set
possible,
basically
for
VM
scale
sets
and
then
once
we
get
that
link
down
start
adding
features,
because
I
actually
I
think
we're
gonna
have
mutations
in
the
API
model.
I
I,
don't
know
that
the
API
model,
the
way
it
is
and
capi
right
now
will
continue
on
the
exact
same
trajectory,
because
it's
really
kind
of
weird
like
we
have
machine
pool
and
then
you
have
your
spec
and
then
you
have
your
template
and
then
you
have
another
spec,
and
then
you
have.
Finally,
your
infrastructure
reference
down
below
it
that
I,
don't
I,
don't
know
if
that
infrastructure
reference
is
really
really.
I
Where
do
you
want
it
to
be,
because
you
then
have
provider
IDs
all
the
way
on
the
top
of
machine
pool
and
that
that
list
of
provider
IDs
is
really
the
provider
IDs
for
the
the
VM
for
all
the
VMS
right?
So
really,
it
kind
of
seems,
like
you,
want
machine
cool
to
be
the
head,
the
provider
ID
of
actually
like
the
BMS
s
and
then
below
it,
have
like
or
have
a
list
of
provider
I
knees
that
are
the
VM
instances,
I
I,
don't
know.
I
I
I
just
have
a
feeling
that,
after
some
feedback
we're
gonna
end
up
in
a
situation
where
we
we
have
a
little
bit
of
API
manipulation.
Does
that
make
sense?
So
if
we
we
just
get
the
base
feature
set
first
and
then
we
start
to
talk
about
it.
A
little
bit.
I
think
we'll
start
to
sort
that
out.
Okay,
but
but
that
is
a
goal
that
is
definitely
cool.
I
A
I
I'll
try
to
get
it
to
working
State
today,
I,
it's
it's
closed,
like
the
service
call
should
be
good.
I
need
to
get
it
working
and
to
end
and
I'll
be
doing
that
today.
I
wanted
to
get
the
structure
out
there.
First
get
the
API
model
in
the
controller
start
talk
through
the
services
a
little
bit.
Okay,.
A
D
F
B
D
B
A
I
D
A
Just
some
clarifications
for
milestone
stuff
for
people
on
the
call
nerves
that
are
not
necessarily
maintained
errs
when
we
put
something
in
a
milestone,
that
kind
of
means
that
we,
as
the
maintainer
group,
are
committing
or
trying
to
commit
to
delivering
it
for
that
milestone,
not
necessarily
that
someone
else
can't
come
and
do
the
work
and
land
it
for
for
an
earlier
milestone
than
what
we
intend
right.
It's
just
more
about
the
the
work
commit
than
anything
else.
A
B
So
I
just
wanted
to
bring
this
up
here
as
well
brother
up
in
the
Cappy
meeting,
but
I
know
it's
two
different
sets
of
attendees.
There
is
an
issue
that
I
opened
in
the
cluster
API
repo.
That
is
basically
a
proposal
for
expanding
the
test
framework
to
run
conformance
tests
with
the
different
providers,
and
this
was
a
result
of
a
discussion
that
I
had
with
a
few
folks
last
week,
and
the
idea
is
really
that
we
realized
that
right
now,
there's
scripts
in
different
repos,
like
AWS,
has
a
component
scripts.
B
B
That's
not
released
yet
and
then
like
run
the
tests
on
it
and
yeah
and
there's
two
goals
from
this
I
guess
one
is
to
test
cluster
API
and
test
that
its
conformance
with
the
Cabrillo
conformance
tests
and
the
other
one
is
to
test
of
extreme
Korean
ease
and
to
use
cluster
API
as
a
tool
to
the
coroners
clusters
to
test
grenades
features
if
that
makes
sense.
So
please
take
a
look
and
if
you
have
any
comments
or
questions
about
that
first,
another
issue
or
that
you
know.
B
B
B
So
it
doesn't
exist
it
right
now.
This
is
not
opposed
all
right,
so,
okay,
so
I
just
urge
the
PR
in
cabs
ii
lasts
like
early
this
week,
which
adds
a
script
that
enables
running
conformance
tests
on
our
cabs
in
clustered
and
right
now,
it's
just
a
bash
script
and
it
works,
but
the
conformance
tests
are
currently
failing.
B
B
B
I
also
had
the
next
one
so
yeah.
I
just
wanted
to
say
that
Vince
and
I
did
a
deep
dive
for
rejects
that
I
you
this
week
and
in
case
people
are
looking
for
it.
It's
not
on
the
recording
of
the
live
stream,
because
relation
was
too
long
and
it
got
cut
out
at
12
hours.
So,
but
it
is,
it
was
recorded
and
I
understood
from
the
organizers.
B
D
D
Yeah,
like
I
realize
I'm
probably
did
insist
that
one
and
that
I
did
the
first
one
for
David
I,
just
kept
sorry
it
there
a
little
bit
there.
Okay,
some
of
the
details
aren't
there
but
okay.
The
last
thing
we
have
is
going
through
the
milestone:
the
next
milestone
for
cab
Z,
so
I'll
go
ahead
and
open
it
up
in
Germany
screen.
We
want
to
make
sure
everyone's
on
the
same
page.
D
D
D
I
think
the
date
was
like
the
release
for
the
next
milestone
copy
we
were
thinking
would
be
end
of
June
early
July,
so
that
gives
us
about
three
months
so
for
that
time
frame
this
kind
of
makes
sense,
there's
a
lot
of
stuff
if
things
do
fall
out,
we'll
just
put
them
in
to
me
next
and
hopefully
in
the
next
milestone
after
top
five.
So
let's
just
go
quickly
through
it.
D
If
you
see
something-
and
if
you
want
to
add
yourself
to
it,
feel
free,
if
you
see
something,
that's
missing
just
please
let
us
know
either
message
us
or
you
can
just
call
it
out
right
now
and
we
can
also
go
through
the
next
milestone
and
then
I'll
bring
up
the
project
board
and
we
can
do
a
little
bit
of
switching
for
the
smaller
issues
and
things
okay.
So
we
put
all
bugs
in
this
milestone,
so
any
bug
that
was
a
pain
or
issue
tracker
was
just
automatically
place
into
here.
D
There
is
documentation
issues,
not
all
of
them
are
up
yet
or
in
here,
but
in
terms
of
features.
A
diagnostic
script
for
log
collection
landed
here,
as
your
machine
pulls,
is
definitely
in
here
as
you're.
Seeing
I
have
a
makutsi
I'm,
also
in
here
at
a
tree,
support
for
cloud
provider,
components
re-evaluating
at
CD
and
data
desk.
D
If
anyone
wants
to
add
more
details,
support
failure,
domains,
refactor
using
defaulting
abouting
web
hooks
for
more
properties,
and
he
has
been
proposal
for
breaking
out
resource
reconciliation
as
respond
again,
a
multi
control
playing
ete.
This
is
a
lot
okay,
I'm,
just
gonna
talk
about
the
big
ones.
Do
a
lot
of
EDS
at
a
bug,
adding
ipv6
support,
adding
msi
support.
I
can't
wait
for
that.
One
accelerated
networking
for
the
emphasizes
that
support
it,
implementing
a
bass,
Jian
post.
They
say
that
right.
A
D
D
H
B
There's
a
lot
of
people
assigned
in
the
other
nine
stone.
If
you
see
some
that
aren't
assigned
Jill
pretty
pick
them
up,
but
also,
if
you
are
assigned
to
one
of
please
and
you
don't
think
you
can
commit
to
working
on
it
in
the
next
year,
can
you
please
unassigned
just
so
someone
else
can
pick
it
up.
And/Or
talk
to
us
by
the
way
yeah.
D
D
This
is
our
next
step,
so
this
was
stuff
that
we
definitely
want
to
do.
We
just
don't
think
it's
gonna
happen
in
the
next
three
months.
That
could,
of
course,
changes
some
picks
it
up
and
really
wants
to
go
towards
it.
So
supporting
as
your
client
requests
to
another
data
center
I
support
for
GP,
you
know
it's
somewhere
eating
me
support
for
a
cast
me,
a
big
one,
I'm
reducing
the
amount
of
cluster
Korea
time.
That's
more
of
just.
D
We
need
to
figure
out
a
metric
to
go
off
of
so
and
then
asking
adding
tracing
image,
fixed
support
for
Windows
worker
notes.
We
kind
of
place
windows
in
here,
but
yeah.
That's
our
next
milestone,
which
doesn't
have
an
official
Nelson
yeah
and
then
I'm
going
to
go
to
our
project
board,
and
we
can
see
there's
anything
that
we
need
to
clean
up
from
last
time
that
we.
A
D
D
I
Whenever
you
see
large
amounts
of
cards
like
that
spread
out
across
a
long
period
of
time,
it
just
leads
me
to
believe
that
we
are.
We
will
not
be
a
blast,
may
well,
because
it's
just
a
large
variance
right,
large
variance,
large
friend,
so
smaller,
smaller
timelines
and
less
cards
or
to
me.
Oh,
it's
easier
to
start
judging
progress
right
so.
D
Yeah
and
maybe,
as
we
like
as
another
week
or
two
passes,
we
can
kind
of
track
our
progress
I
do
think
some
things
are
going
to
be
falling
out
of
this
milestone,
but
maybe
there
is
a
shorter
way
that
we
can
start
tracking
this
stuff
up
till
like
two
weeks.
Instead,
let's
go
back
and
think
about
that
and
think
about
how
we
could
do
that
especially
project
and.
J
I
wonder
it
like,
and
this
might
be
in
the
V
in
the
next
milestone
is
like
features
that
we
know
we
want
is
just
like
we're
not
gonna
get
to
it
yet,
like
I,
do
see
a
lot
of
them
in
next,
like
GPU,
for
example,
but,
like
I,
think
someone
asked
about
zone
tray
like
right,
I,
wonder
if
we
should
add
them
here,
yeah
just
so
that
at
least
people
know
hey.
This
doesn't
work
yet
right
now,
oh.
D
A
A
F
F
H
D
Or
initiation
mode
yeah,
if
you
want
to
write
up
an
issue,
we
can
go
through
the
technical
details
and
kind
of
figure
out
where
this
fits
in
Azure
and
if
and
where
it
fits.
And
you
know
our
world
right
here
so
go
ahead.
Add
an
issue
in
all
the
details.
You
know:
okay,
any
link
and
yeah,
we'll
figure
it
out.
Thank
you.
B
Yeah
I
think
in
in
general
to
read
us
point.
There
are
a
lot
of
missing
issues
here
and
it's
just
going
to
be
a
matter
of
time
as
we
collect
like
different
feedback,
and
but
yes,
please,
please,
if
you
think
of
anything
that
you'd
like
to
see
be
done
in
the
project.
Even
if
you
don't
personally
have
time
to
work
on
it,
please
file
an
issue
and
we
can
track
those,
even
if
it's
not
for
right
now
or
for
the
next
milestone.
D
E
H
D
E
Explained
better
so
the
defenders
it
correctly
a
KS
engine
can
also
be
used,
basically
as
a
library
which
is
providing
sort
of
natural
resources
that
all
together
combined
like
can
make
communities
clusters.
So
there
are
like
many
many
resources
that
are
implemented
there
and
it
can
be
used
obviously
to
create
a
KS
clusters.
So,
but
is,
do
you
see
it
possible
that
cluster
API
in
the
implementation
for
a
sure,
instead
of
using
Azure
API
directly
and
like
creating
resources
one
by
one?
Whatever
is
the
correct
implementation?
This
is
possible
that
cluster
API
could
use
aks.
E
A
We
had
yeah
I,
think
I,
think
David
and
I
were
bouncing
back
and
forth
about
this.
This
was
kind
of
the
idea
that
maybe
there
would
be
a
library,
not
necessarily
cap,
Z
or
not
necessarily
a
KS
engine
that
would
abstract
the
different
components
right.
I
want
to
create
a
an
NSG
I
want
to
create
a
load.
Balancer
I
want
to
create
foo
right,
so
common
implementations
for
the
same
thing
that
wouldn't
necessarily
be
caps,
II,
wouldn't
necessarily
be
a
KS
engine,
wouldn't
necessarily
be
implemented
on
the
cloud
provider
level.
A
It
may
be
a
Microsoft
thing
and
maybe
a
kubernetes
thing.
It
may
be
a
nose
thing.
I
think
several
teams
internally
at
Microsoft
are
talking
about
this,
and
there
are
some
ideas
floating
around,
but
nothing
has
come
to
fruition.
Quite
yet
that
is
ready
to
be
shared
with
the
public.
But
yes,
it's
an
idea,
that's
floating!
A
What
I
did
ask
him
that
thread.
I
believe
was.
If
you
could
give
us
more
clarity
on
the
operator
that
you
said
you
y'all
were
delivering
an
operator.
If
you
can
give
some
more
clarity
on
the
operator
that
you're
delivering
and
maybe
where
our
feature
sets
intersect
right,
so
if
it's
maybe
something
that
makes
sense
to
build
within
cap,
see
our
build
in
this
new
thing
that
maybe
will
exist
one
day
it
get
getting
more
clarity
on
what
you're
trying
to
deliver.
What
would
help
us
there.
H
E
It's
using
our
API
directly,
so
we
are
using
operator
framework
that
was
developed
within
the
company
years
ago.
I'm,
not
sure
and
I
think
the
idea
the
internal
framework
was
developed
because
at
the
time
the
cube
builder
or
operator
framework
was
nothing.
So
it's
actually
probably
one
of
the
older
ones
out
there
and
now,
as
the
cluster
API,
is
gaining
traction
and
we
believe
we
also
believe
it's
the
future.
E
We
are
contemplating
we're
thinking
about
moving
to
Casa
API
so
and
just
thinking
how
we
could
have
a
good
migrate
or
could
we
adopt
cluster
API
interfaces
of
front-end
only
and
then
have
actually
our
implementation
with
our
operator
framework
behind
it
and
not
use
queue
builder
or
the
other
idea
is
maybe
to
fully
use
our
API
to
deploy
creatives
clusters.
So
that's
where
the
question
came
from
I
I.
A
E
A
A
B
Sort
of
related
I
pasted
an
issue
in
the
chat
I,
don't
know
if
you've
seen
this,
but
David
opened
up
a
proposal
to
break
out
the
clouds
infrastructure,
provisioning
from
cab,
Z
and
I
love
your
feedback
on
that,
since
it's
sort
of
related
to
what
Stephen
was
talking
about
with
having
a
common
provisioning
tool
instead
of
having
cab
z2.
It's
going
add
your
API
calls
for
provisioning
resources.
G
F
A
I
A
At
this
point,
we're
questioning
why
we
did
it
initially
and
I
think
we
did
it
initially
or
I.
Did
it
initially,
because
it
made
more
sense
at
the
time,
but
maybe
it
doesn't
anymore
and
honestly,
I
haven't
had
an
opportunity
recently
to
deep
dive
into
like
the
v1
alpha,
3
reef
actors
and
all
that
stuff.
So
it
may,
it
may
not
even
make
sense
anymore.
A
I
A
Given
have
we
are
we
saying
that
the
everything
that
the
compatibility
guarantee
for
closer
API
is
a
super
strong
one
for
the
feature
set
of
cluster
API
yeah
sure
as
well?
So
what
I
had
in
mind
was
potentially-
and
we
should
probably
discuss
this
as
well
with
the
cluster
API
folks,
what
if
we
do
not
do
zero
four
zero,
zero
five
zero
as
the
initial
Rev,
whatever
new
version
of
cluster
API
it's
coming
out,
and
instead
do
you
know
pre-release
those
right.
So
the
alpha
alpha
beta
RCS
up
to
the
point
that
we
actually
hit.
A
B
B
But
the
other
thing
is
I
actually
talked
to
Vince
about
this,
and
he
was
saying
that
we
don't
necessarily
need
to.
We
don't
need
to
stop
adding
features
after
we
released
the
API.
We
just
need
to
make
sure
that
whatever
we
add
is
additive
and
not
breaking
bad
competição.
If
we
did
add
a
new
thing,
it
would
have
to
be.
B
A
I'm
sort
of
in
in
favor
of
the
the
pre-release
option
right
and
maybe
we
track
along
the
weed
track
along
because
we've
been
doing
it.
The
alpha-beta
is
up
to
their
actual
release
right,
and
maybe
we
follow
that
for
a
bit
because
it
feels
like
it
feels
super
final.
Once
we
drop
that
dot,
zero
and
you're
like
okay
done,
and
we
immediately
start
thinking
about
the
next
next
milestone,
not
that
we
shouldn't,
but
it
kind
of
it
feels
like
it
says,
no,
no
new
features
and
a
patch
releases.
A
A
I
Can
we
just
put
a
PSA
up
and
just
say,
hey
until
you're
at
1-0,
we
will
like
just
state
our
claim,
like
simply
until
we're
at
1-0
in
point
releases,
not
Pat,
releases
minor
releases.
We
intend
to
run
up
in
patches
they're
up
until
like
we
hit
a
break
and
change
and
then
we're
gonna
will
increase
the
patch
or
not
the
patch,
to
minor
version.
A
So,
but
we
only
if
so
right
now
we're
only
increasing
so
one
I
do
think
we
have
a
provision
for
that
in
the
readme.
It
says
that
we
follow
blah
blah
blah.
We
follow
semantic
versioning
data,
yada,
yada
I-
think
maybe
it's
still
there
yeah
as
the
full
closed
as
a
versioning
for
this
project
is
tied
to
the
versioning
of
cluster
API
future
modifications,
so
this
policy
may
be
made
to
much
more
closely
aligned
with
other
providers
in
the
cluster
API
ecosystem.
So
that's
something.
A
But
yeah
I
don't
think
that
I
don't
think
that
we
should
have
the
option
of
because
right
now
we've
been
bumping
minors,
along
with
a
new
version
of
a
new
API
version
of
cluster
API
right.
So
we
have
to
decide
to
either
continue
doing
that
or
saying
that
it's
okay
to
bump
the
minor
version
arbitrarily
and
I.
Don't
think
that's
what
the
the
other
provider
implementations
are
doing
today.