►
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).
B
Okay,
good
morning,
everybody
it's
february
4th.
This
is
the
cluster
api
provider
for
azure
office
hours
and
meeting.
We
are
a
project
under
the
umbrella
of
sig
life
cycle
and
thus
for
a
kubernetes.
B
Group,
oh
my
sorry,
my
dusk
is
jumping
around
and,
let's
all
adhere
to
the
general
standards
of
conduct
basically
be
nice
to
each
other.
Let
everyone
speak
and
we
can
try
and
raise
hands
through
zoom,
although
we
don't
have
that
many
people
so
we'll
see
how
it
goes.
B
B
Well,
you
can,
you
can
always
type
something
in
the
discussion
chat
if
you
want,
or
if
you
get
your
mic
working
later,
definitely
just
interrupt.
B
Us,
okay,
I
guess
the
first
discussion
topic
david
justice:
you
want
to
talk
about
machine,
pull
machines,
next
search
max
unavailable.
D
Yeah
you'll
be
happy
to
looks
like
the
screen
is
not
updated.
Just
a
heads
up
there
are.
E
B
B
B
Sometimes,
if
I
click
optimize
for
video
clip
before
I
share,
and
it's
happier
yes,
oh,
what's
it
doing
to
share
your
audio,
please
install
the
zoom
audio
device.
No.
Thank
you.
F
Oh
finally,
okay
go
ahead.
I'm
just
saying
this
is
my
first
time
joining
the
meeting.
I
wanted
to
talk
about
the
setting
up
the
dev
development
environment
so
I'll.
Let
you
guys
carry
on
and
you
let
me
know
when
I
can
jump
in.
B
Yeah
we
can
just
do
that.
I
guess
open
discussion
topics.
Do
people
see
my
screen
now
or
should
somebody
else
share.
C
D
Awesome,
thank
you
so
much
man.
I
appreciate
it.
Okay,
so
just
want
to
give
a
quick
update.
I've
pushed
out
a
bunch
of
new
changes
for
machine
pool
machines,
max
surge
and
max
unavailable.
D
At
this
point
you
should
I
I
do
need
to
rebase
it,
since
we
had
some
pretty
large
changes
just
flow
in.
But
if
you
were
to
pull
down
that
branch
right
now,
you
would
be
able
to
use
async
machine
pools
with
machine
pull
instances,
vm
vmss
vms,
actually
being
broken
out
into
their
own
resources.
So
now
there's
machine
pools
and
machine
pool
machines.
D
They
they
work
as
as
they
currently
do
today
without
mac
surge
max
unavailable.
D
D
D
D
So
at
this
point
my
plan
is
to
take
the
coordinate
drain
functionality
from
machines
and
apply
it
to
machine
pool
machines
with
the
hope
that
eventually
we'll
bubble
that
back
up
into
cappy
and
have
a
solution
where
cabbie
is
responsible
for
coordinating
these
machines
and
and
kind
of
managing
that
portion
of
the
life
cycle.
D
But
I
think
for
expedience
we
can
get
it
in
place.
Working
today
show
how
it
works
and
then
use
that
as
a
template
to
take
back
into
cabbie
and
maybe
also
put
out
for
like
aws
and
and
others
and
see.
If
that
pattern
fits,
does
anybody
have
any
feedback
or
any.
A
Questions
cecile
yeah,
I
haven't
looked
at
or
I
haven't
reviewed
the
pr
yet
so
this
might
already
be
taken
care
of.
But
I'm
just
wondering
do
you
even
plans
to
keep
the
code
that
we're
planning
to
move
upstream
to
capi
in
the
future
kind
of
separate
and
isolate
it
so
that
it
doesn't
become
tightly
coupled
with
cap
z
code
in
the
future,
with
like
updates.
D
Yeah,
so
really
what
happens
or
where
this
will
get
plugged
in
or
at
least
right
now
is
a
machine
pool
machine.
When
we
find
out
that
this
machine
isn't
running
the
latest
model,
we
then
mark
it
for
delete.
We,
we
delete
it
in
the
kubernetes
api
marks
that
machine
for
delete
in
the
delete,
reconcile
we'll
go
in
and
we'll
see
should
we
drain.
D
This
should
be
according
to
this
and
then
we'll
start
a
long-running
operation
where
we'll
asynchronously,
cordon
and
drain
and
then
after
that's
done,
actually
delete
the
infrastructure
underneath
it
so
that
that
chord
and
drain
is
actually
just
really
kind
of
just
off
on
its
own
it'll,
probably
just
be
like
two
funks
and
then,
when
those
two
funks
go
away,
then
we'll
just
change
the
reconciler
a
little
bit.
D
So
yes,
it
should
be
really
focused
the
thing
that
the
selection
logic
is
the
thing
that
we
really
need
to
think
about
and
probably
need
a
second
pr
after
this
to
really
start
becoming
smarter
about
how
we
select
nodes
that
we're
going
to
destroy
so
right.
Now
I
kind
of
take
the
naive
approach
of
saying:
hey,
I'm
going
to
pick
out
anything,
that's
failed,
anything
that's
broken
and
we'll
clean
those
out
first
and
then
go
anything
with
an
older
model,
we'll
clean
those
up.
D
And
finally,
I
only
start
to
you
know:
remove
ready
nodes
like
goods,
successful
nodes
when
we
must
so,
I
think
we
need
a
order
of
grading
and
we
need
to
start
taking
into
account
disruption
budgets.
We
have
a
lot
of
stuff
and
hopefully
that's
not
something
that
we're
using
individually
like,
hopefully
other
people
can
use
that
too,
or
we
can
suss
it
out
from
other
projects.
Maybe
a
cluster,
auto
scaler,
or
something
like
that.
So
yeah.
B
Excuse
me
all
right
david,
you
got
the
next
topic.
D
Yeah,
so
part
of
machine
pools
is
the
backing
with
vmss
and
vmss.
Has
this
idea
of
a
model
and
the
model
means
like
what
image
reference?
Am
I
using
what?
What
does
my
virtual
machine
template?
Look
like
so
a
vmss,
a
virtual
machine
scale
set,
is
really
a
template,
plus
instances
that
it
built
out
with
this
template,
and
when
you
update
the
template,
it
shows
in
the
api
that
the
model
is
out
of
date,
that
it's
not
running
the
latest
model,
and
really.
D
This
is
how
we
tell
that
a
machine
needs
to
be
reformatted
or
you
know
we
need
to
replace
it
with
a
newer
version
of
this
model.
D
Now
the
order
of
the
reconciler
for
vmss
goes:
I'm
going
to
go,
create
or
update
the
vmss,
then
I'm
going
to
add
any
roles
that
are
needed
for
identities
and
then
I'm
going
to
add
vm
extensions.
So
the
creator
update
to
a
vm
extension,
is
actually
manipulating
the
vmss
model,
which
is
really
counter-intuitive
because
it
has
its
own
api
calls,
but
those
actually
connected
to
the
dss.
D
So
in
the
pr
I
actually
have
taken
the
vmss
extensions
and
I
I
added
them
to
the
put
for
vmss
itself.
So
the
model
is
now
consistent.
So
we
we
don't
update
the
model
a
second
time,
but
I
think
we
need
to
figure
out
how
we
use
that
vmss
extension
and
suss
out
the
status
so
that
we
can
differentiate
the
error
types
that
we're
getting
back
like
did
infrastructure.
Just
not
did
azure
fabric
fall
over
and
didn't
build
your
machine
or
did
the
did.
D
The
error
come
out
of
the
cloud
init
script
like
we
couldn't
join
to
the
cluster,
or
something
like
that.
So
there's
like
two
different
categories
of
error.
Cecile
could
probably
talk
a
little
bit
more
about
that.
Do
you
do
you
want
to
explain
the
vmss
extension
or
the
vm.
A
Extension
sure
yeah.
So
this
the
way
it's
set
up
works
great
with
the
vms
and
vm
extensions.
It
doesn't
work
so
well
with
the
nss.
A
I
think,
because
of
the
model
thing,
but
really
what
we're
trying
to
do
with
the
extension
is
take
leverage
the
sentinel
file,
which
was
recently
added
in
capi
as
soon
as
we
can
get
that
as
per
a
v1
alpha
4
and
then
use
that
as
like
a
bootstrap
failure,
signal
or
success
signal
and
the
vm
extension
will
just
run
a
sync,
a
simple
script
which
will
pass
in
to
it
as
part
of
the
capsi
vm
extension
properties,
and
it
will
just
run
a
simple
script
that
checks
for
the
presence
of
that
file
and
retries
for
a
given
amount
of
time.
A
And
then
after
we
time
out,
then
that
will
be
our
signal
that
you
know
we
can
find
the
files.
So
bootstrapping
was
not
successful,
so
we
can
mark
the
condition
for
bootstrapping
as
bootstrapping
is
false
or
bootstrapped
as
false,
and
so
the
idea
to
separate
them
was
really
to
separate
that
from
like
machine
creation
was
that,
first
of
all,
it
shouldn't
block
like
vm
creation,
because
it's
more
of
a
check
for
like
user
pers,
like
user
information,
it
doesn't
actually
do
anything
functional
and
then
the
other
thing
is.
A
A
So
I
think
it's
really
important
that
we
keep
those
errors
separate,
no
matter
what
we
do,
since
that
was
the
whole
point
of
having
the
extension,
but
I
think
it
makes
complete
sense
what
you
said
david
to
like
not
do
that
after
and
update
the
model
and
then
have
to
like
roll
the
roll
the
instances,
although
I'm
kind
of
surprised,
that's
the
behavior.
I
would
think
that,
since
an
extension
is
not
really
part
of
the
instance
itself,
it
would
just
add
it
to
each
instance
already
present.
Apparently
it
does
not
do
that.
A
So
that's
my
bad
for
not
checking
but
yeah.
I
think
it
makes
sense
if
we
just
do
it
as
part
of
the
creates,
so
that
it
adds
it
to
every
instance,
and
then
we
just
need
a
way
to
like
separate
the
error
we
get
back.
If
we
get
an
error.
D
Yeah,
I
think
we
can
still
use
that
api
to
get
the
extension
and
get
the
status
for
it,
and
it
also,
I
believe
it
shows
up
in
the
vmss
body
like
if
you
same
with
vm.
So
if
you
use
a
vm,
get
instance
view
and
you
have
the
expanded
vm,
I
think
you
actually
see
the
the
extensions
as
well.
A
D
Awesome,
maybe
maybe
we
could
get
some
time
to
pair
up
on
that
a
little
bit.
B
Questions
nope
all
right.
I
tried
to
write
down
what
you
guys
were
talking
about,
but
some
of
that
was
a
little
dense
for
me.
So
we'll
see
next
cecile
you
want
to
talk
about
the
patch
release.
A
Yeah,
so
we
have
two
bug
fixed
pr's
there
one
is
merged
and
one
is
still
open,
but
I
think
very
close
to
merging,
which
are
pretty
important,
bugs
like
one
is
for
the
spot
instances,
there's
a
failure
to
serialize
the
max
price
with
controller
runtime
and
the
other
one
is
for
the
managed,
v-net
scenario.
A
In
the
delete
case,
we
try
to
delete
the
nsg,
which
is
really
bad,
so
I
think
we
should
get
those
out
in
a
release
sooner
rather
than
later,
to
clarify
they're,
not
regressions
from
the
latest
release
they're
actually
like
missing
or
like
bugs
in
the
original
implementation
of
those
features.
So
it's
it's
not
like
a
recent
regression,
but
I
still
think
we
should
fix
them.
A
So
I'm
thinking
of
if
everyone's
okay,
with
that
I'm
just
doing
a
0-4-12
release
just
for
those
two
commits
cherry
picked
on
top
of
zero
four
eleven
as
soon
as
we
have
both
merged.
If
nothing
else
comes
up.
A
How
does
that
sound
good
yeah
that
was
good,
okay
and
then,
if
anything,
if
anyone
learns
of
anything
else,
that
is
important
like
an
important
bug
fix
or
like
if
we
find
something
in
the
release
that
we
should
deep
like
dive
deeply
into
like
investigate,
then
like
please,
like
let
us
know
so,
we
can
take
that
into
account.
H
No,
I'm
trying
to
add
this
pr,
I'm
trying
to
add
jobs
for
like
the
zero
port
branch,
and
one
of
them
keeps
giving
me
this
weird
error
and
I
don't
know
what
it
is
like
in
in
bazel
config,
some
error.
Thingy
like
I
can
yeah.
I
don't.
I
don't
know,
what's
wrong
with
it.
I
tried
a
couple
of
different
things.
I
keep
getting
there.
So
if
you
go
look
at
the
intra
basil,
job
failed,
you'll
see
what
failed
and
anyway.
D
D
Of
the
logarithms,
the
law.
H
It's
only
because
of
how
I
configured
the
jobs,
even
though
I
copied
them
from
like
I
copied
our
existing
configs
into
like
dash
release,
04.
files
and
like
change
to
make
it
run
on
the
release
branch
like
the
zero
four
branch,
but
just
like
one
job,
which
is
the
coverage
pr
coverage
and
the
pre-submit.
H
H
H
Okay,
I
don't
know
how
to
set
that
up,
though,
like
where
do
I
create
that.
A
Let's,
let's
pair
on
that
offline,
I
don't
think
we
need
necessarily
an
email
for
zero4
we're
just
using
that
as
pr.
A
H
Oh
yeah,
I
tried
taking
out
the
email,
a
complaint
that
doesn't
have
a
tab.
I
tried
putting
a
tab
but
didn't
like
that
so
anyway,
I
think
something
has
to
be
defined
somewhere
else
for
it
to
use
it,
but
I
just
don't
know
where
the
somewhere
else
is:
okay,
that's
it
for
the
next
one
is
also
mine.
Just
a
quick
update
the
view
on
alpha
four.
I
think
I've
done
most
of
the
things
just
waiting
for
like
cappy
to
have
a
nightly
build
which
the
pr
is
almost
merged.
H
H
H
A
True,
true,
okay,
that
was
a
bad
example.
What
about
like
the
basically
I'm
thinking
about
the
things
that
are
in
that
view
on
alpha
3,
to
be
one
off
of
four
doc?
No,
some
of
those
items
will
have
to
take
separately
right
like
implement
the
watch
label,
and
things
like
that.
H
B
F
So
I'm
just
following
the
basic
document
that
you
know
just
says:
hey
install,
go,
install
kind
all
that
stuff
and
then
getting
stuck
on
the
the
go
get
for
the
cluster
api
provider
for
some
reason.
There's
a
dependency
on
something
called
json
patch,
slash,
v2
and
I
think.
F
I
I
think
this
version
of
of
the
module
might
be
obsolete
or
something
when
I
put
in
a
like
a
redirect
in
the
go
mod,
it
seems
to
work.
I
put,
I
put
it
in
the
chat.
Let
me
can
I
show
you
my
screen,
so
you
could
see
what
I'm
seeing
a
host
disable
attendee
sharing.
A
Okay,
though
matt
you
might
have
to.
F
Can
you
guys
see
it
yeah?
So
just
I
mean
it's
basic,
just
I'm
just
running
the,
and
so
these
are
my
errors.
F
Yesterday,
I
I
put
in
a
redirect
in
the
go
mod
for
v3
instead
and
that
seemed
to
work,
but
I
don't
know
if
I'm
missing
something,
because
you
guys
are
working
on
this
every
day
or
I
don't
know
if
you
just
happen
to
have
all
this
stuff
already
set
up
so
you're,
not
seeing
you're,
not
setting
up
a
new
environment
so
you're
not
experiencing
that.
D
Directory
you
have
the
ghost
sum
there
and
that
ghost
song
is
going
to
say
the
exact
versions
to
take,
so
we
might
run
into
an
issue
when
we
go
to
update
this
at
some
point
in
the
future,
where
you
know
it
complains
and
we'll
have
to
navigate
that.
D
Yeah
give
give
cloning
a
try.
I
think
I
think
you'll
have
better
luck,
just
doing
a
clone
and
then
doing
a
make
binaries
or
inmate
tests
within
that
directory.
F
F
F
Yeah,
I
I
did
a
search
for
this
particular
era
in
this
package.
I
think
there's
an
issue
with
this
particular
package
and
that
version,
not
only
with
this
project,
doesn't
look
like
you
guys
are
using
it
directly,
but
maybe
some
dependency.
F
A
F
Yeah,
so
that's
what
I'm
trying
to
get
to
the
tilt
part
to
do
that
and
just
wanting
to
make
sure
I'm
actually
following
the
right
path
or
that
makes.
B
A
E
F
B
Oh,
it's
trying
to
share
sound
again
oops
there
we
go.
Oh
me
yeah,
so
david
had
an
idea.
I
don't
know
a
month
ago
and
I've
been
tankering
on
what
what
I'm
calling
azcappy,
which
is,
if
you
use
the
azure
command
line,
you
can
install
and
install
an
extension,
and
this
is
an
extension
that
the
idea
is
basically
to
make
using
cappy
easier,
using
cap
z,
easier
specifically.
B
Because
if
someone
just
wants
to
kick
the
tires
and
see
what
cappy
is
all
about
or
capsi,
it's
a
pretty
high
bar,
so
the
two
ideas
here
basically
and
now-
this
may
never
really
see
the
light
of
day
or
it
may
take
off.
We
really
don't
know
yet
I'm
just
kind
of
getting
it
ready.
Two
ideas
are
file
off
all
the
rough
edges
and
also
hold
the
user's
hand
as
much
as
possible
so
that
you
could
blender
into
a
command
line,
experience
and
say
effectively.
I
want
to
create
a
cluster
and
it'll
figure.
B
B
What
type
you
didn't
tell
me
what
kind
of
cluster
you
want,
so
I'm
going
to
ask
you
a
few
more
questions
and
then
we'll
enable
some
features
and
then
it'll
just
deploy
the
whole
thing
in
in
one
command
line
that
could
at
least
help
people
really
kick
the
tires,
and
also
maybe
generate
some
ideas
that,
hopefully
we
could,
you
know
bring
up
stream
because
anyway,
and
the
other
idea
is
right
now,
flavors
are
the
main
way
of
specifying
how
you
want
to
customize
your
cluster,
but
they're
kind
of
little
silos.
B
You
know
they
exist
on
their
own.
If
you
wanted
to
mix
and
match
flavors
say
you
want
to
have
ipv6,
plus
gpu
nodes
plus
system
assigned
identity.
That's
essentially
right
now
left
as
an
exercise
to
you,
the
user,
to
look
through
all
our
templates
figure
out
how
they
work
or
maybe
create
some
end
templates
and
then
compare
them
and
do
some
diffs
and
figure
out.
Oh
well,
if
I
put
this
here,
then
I
probably
end
up,
and
so
that's
so.
B
So
those
are
the
ideas
I
had.
I'm
actually
not
quite
ready
to
demo
it
yet,
but
I
just
want
to
let
people
know
that's
what
I've
been
working
on,
because
I
haven't
been
doing
much
core
stuff
for
a
while
and
I'm
hoping
to
just
open
source
this
see.
If
anyone
thinks
it's
interesting,
maybe
we
can
all
pile
on
it
and
make
it
better
or
or
generate
some
ideas
that
go
back
upstream,
because
I
also
I'm
sensitive
the
idea
that
this
is
kind
of
off
the
rails,
and
you
know
why
are
these?
B
I
think
the
reason
is
we
can
some
of
this
ends
up
being
very
azure
specific
and
so
wouldn't
make
sense
in
cluster
cuddle,
also
we're
having
a
hard
time
seeing
how
we
sort
of
break
out
of
the
flavors
approach
to
doing
templates,
and
so
this
is
a
way
to
just
do
it
in
a
completely
different
way
and
see
if
that
works
better
for
our
needs
all
right.
So
I
talked
a
lot.
If
anybody
has
any
questions,
go
ahead.
D
Hey
matt,
hey,
really
super
excited
to
see
this
stuff.
This
is
so
cool.
I
I
have
to
ask
how
I
I'm
not
very
familiar
with
azure,
cli
or
extensions.
F
D
B
You
can
just
say
a
z
extension
add,
is
a
built-in
command
it'll
go
off
on
the
internet,
download
a
python
bundle
that
adds
new
functionality
to
the
cli
we
can
and
and
pro
right
now
it's
all
just
in
my
repo,
so
you
would
say,
like
az
extension,
add
amborsma,
azcappy
and
it'll.
That's
that'll
be
enough
to
get
you
going.
B
The
very
next
step
once
we've
had
a
few
eyes
on
this
is
to
put
it
in
the
official
extensions
repo
and
then
it's
even
a
little
easier
to
install,
because
you
can
say
a
z,
extensions
list
and
cappy
would
show
up
as
one
of
the
extensions
you
could
install
and
you
can
even
you
know
if
you're
really
into
it.
If
you
had
a
site
that
needed
it,
you
can
do
things
like
when
we're
installing
az
always
install
these
extensions
stuff
like
that.
H
B
A
I
have
one
thing:
you
said
that
it's
you
know
why,
wouldn't
you
do
this
in
cluster
cuddle
to
me?
They're
two
different
things.
I
think
cluster
cattle
is
more
like
setting
up
your
management
cluster
and
managing
your
management
cluster,
like
upgrading
the
version
of
cappy
and
all
of
that,
and
I
think
what
you're
trying
to
do
is
not
at
all
trying
to
do
that,
it's
more
like
creating
workload
clusters
and
setting
up
all
the
identity.
That
goes
around
that
because
I
think
that's
the
big
part
right
now.
A
That's
like
extremely
manual
like
having
to
create
a
service
principle,
a
set
like
set
all
those
environment
variables
set
like
your
azure
location,
your
azure
vm
size,
like
all
that
stuff
that
we
can't
put
into
cluster
cattle
because
it's
like
azure
specific,
otherwise
they
would
already
be
there,
and
I
think,
like
there
is
value
in
having
that
separate
like
in
the
future.
A
You
know
like
eventually,
maybe
it
could
be
a
plug-in
or
something
for
cluster
cuddle
or
for
cute
kettle,
but
it's
definitely
not
going
to
be
in
the
core
cluster
title
anytime,
soon,
right
so
yeah.
It
makes
sense
to
have
that.
I
think.
B
Yeah
totally,
this
also
allows
us
a
place
to
become
opinionated
like,
like
you
said,
cluster
cuddle
and
cluster
api
is
just
about
the
infrastructure.
We
have
to
support
all
the
options
everything
that's
available,
but
in
this
tool
we
could
maybe
be
more
opinionated
and
have
the
defaults,
be
you
know,
use
machine
resource
set
or
whatever
we
think
is
a
better
way
to
go
so
yeah.
B
So
definitely
by
the
next
meeting,
I'll
have
we'll
all
have
hopefully
looked
at
this
and
I'll
give
you
the
url
anything
else
to
talk
about.
That
was
all
the
topics
we
had
listed.
H
Oh
just
analyze
a
quick
question
since
we're
going
to
do
a
new
release
like
the
alpha
4
and
it's
going
to
be
all
breaking
and
stuff.
Do
we
want
to
move
anything
out
of
the
experimental
stuff
into
like
not.
D
Well,
that's
that's
one
of
the
fun
parts
about
it
being
the
experimental
folder.
I
don't
feel
so
horrible
if
stuff
breaks,
but
I
I
do
feel
like
no
matter
where
it's
at.
I
would.
I
feel
terrible
when
we
have
api
breaking
changes
like
that's
just
sadness
right,
I
I
don't
know
who
any
other
any
strong
opinions
james
is
raising
his
head.
B
E
I
I
can
say
something
about
having
machine
pulls
in
experimental
or
not
so
at
giants
form
we
haven't
been
using
machine
pools
for
some
time,
mainly
on
the
api
level,
so
using
crds,
so
copy
machine
pools
and
cut
the
azure
machine
pools.
So
we
did
find
some
rough
edges
around
it.
More
than
a
few
times
latest
has
been
the
fix
for
spot
x
price
that
you
talked
about
now,
and
so
that
was
a
pr
from
adrian
vartin
from
our
team.
E
E
I
mean
just
the
feeling
I
got
so
nothing,
nothing
concrete.
D
I
would
commend
it.
Thank
you
so
much.
The
best
input
best
input
is
people
using
it
and
telling
us
why.
A
Valuable
james,
we
couldn't
hear
what
you
were
trying
to
say
earlier.
Is
your
microphone.
G
Now
yes,
yeah,
okay,
I
switched
I
was
driving
earlier
and
I
switched
to
my
computer.
I
yeah,
so
I
guess
my
audio
wasn't
working.
I
tried
jumping
in
once
or
twice.
I
had
a
sorry
to
bring
it
back,
but
I
had
a
quick
question
about
like
the
azzy
cappy
azure
cli
extension
yeah.
E
G
In
particular,
just
like
you
had
mentioned
like
there's
some
functionality,
that's
that's
up
that
won't
ever
kind
of
make
it
back
into
the
cluster
cuddle
command.
I
was
wondering
like
there's
that
new
feature
for
debugging,
that's
really
kind
of
nice
to
see
what
the
state
of
all
the
different
controllers
are
in.
I
was
wondering
if
you've,
given
any
thought
to
like
how
the
azure
cli
would
interact
with
something
like
that,
so
that
we
can
get
that
debugging
experience
that
that
was
added
to
azure
or
the
cluster
ccli.
B
Yeah
I
saw
it,
I
haven't
had
a
chance
to
look
at
it
yet
also.
The
described
cluster
seems
like
something
we
should
piggyback
on.
Rather
than
do
it
ourselves,
I'm
I
mean
we're
we're
using
cluster
cuddle
a
lot
as
a
helper
underneath
in
some
of
these
commands
and
where
we're
throwing
it
away
is
when
we
generate
the
templates
and
stuff,
but
I
don't
intend
to
reinvent
the
wheel.
Cluster
cuddle
does
a
lot
of
important
stuff,
so
yeah.
I
B
A
lot
of
it
is
shelling
out
yeah,
okay
to
other
commands,
support.
B
A
To
bring
up
to
go
back
to
the
experimental
thing,
there's
also
the
managed
clusters
which
aren't
experimental,
but
I
don't
think
we're
there
yet
in
terms
of
user
mileage
and
feature
like
support
to
get
it
out.
I
I
don't
think
anyone
here
has
been
using
it
or
maintaining
it,
but
so
we
should
probably
ask
those
people
but
yeah.
Just
my
perspective,
like
from
echoes
I've
heard.
E
C
B
Going
once
going
twice
sold
all
right
thanks
everybody
for
coming.
That
was
great,
we'll
see
you
again
in
the
same
place
in
two
weeks.