►
From YouTube: SIG Cluster Lifecycle - Cluster API 21-08-25
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
Hi
everyone:
this
is
the
cluster
api
meeting
on
august
25th,
welcome
everyone
we'll
be
following
the
cncf
code
of
contact
code
of
conduct.
That
means
everyone
being
nice
to
each
other,
and
if
you
want
to
speak
up,
please
use
the
raise
hand
feature
in
zoom
chat.
A
Please
add
your
names
to
the
attending
list
and
then
we'll
and
we'll
get
started
with
psas
looks
like
cluster
api,
zero,
four,
two
and
zero:
three
twenty
three
have
been
released
and
imaginations,
and
stefan
you
have
the
next
item
on
pc.
B
Yeah
I
just
wanted
to
mention.
We
did
some
improvements
to
our
unit
tests
the
last
few
weeks
and
that's
that's
mainly
a
tip
for
someone
developing
on
the
main
repository
and
executing
new
tests
locally.
But
what
you
can
do
now
is
one
a
little
bit
up.
B
B
The
advantage
here
is
that
you
don't
have
to
wait
until
the
test
end
is
started
up
on
each
execution.
B
One
warning:
not
all
our
tests
right
now
are
repeatable,
so
it
could
be
that
that
a
few
tests
fail
on
the
second
one,
but
I'll
open
an
issue
and
those
are
only,
I
don't
know
100
or
something,
and
the
second
tip
is,
if
you
run
a
specific
test
in
a
package
which
has
fake
client
tests
and
test
engine
tests
as
well,
and
you
only
want
to
run
fake
line
tests,
then
you
can
set
the
environment
variable
copy,
disable
test
end
and
then
the
test
end
is
not
started.
B
A
Thank
you,
stefan.
Next,
let's
look
at
release
blocking
issues.
A
There
are
no
release
button
issues.
Proposals.
Are
there
any
updates
on
any
of
the
open
proposals
that
people
want
to
discuss.
A
Okay,
I
see
no
hands.
Let's
continue
discussion
topics
alex
you
have
the
first
one.
Oh,
I
see
a
hand
for.
C
Mike
yeah,
sorry,
they
moved
the
raise
hand
button
on
me.
I
couldn't
find
it.
I
just
wanted
to
say
that
I've
updated
the
scale
from
xero
proposal
and
I
would
really
love
any
extra
reviews
that
people
are
willing
to
put
on
it.
I
think
I've
captured
everything
we
talked
about.
So
hopefully
you
know
yeah.
Hopefully
it's
in
a
good
state
for
people
now.
A
Thanks
mike
yeah
for
anyone
who
is
up
for
reviewing
it,
please
go
ahead
and
take
a
look
at
that.
Thank
you
any
other
discussions
from
open
proposals.
A
Okay,
I
see
no
hands,
let's
continue,
yeah.
Let's
continue
discussion
topics.
First,
one
is
alex.
You
have
the
photo.
D
So
yesterday
we
went
to
the
sig
meeting
and
we
mentioned
so.
We
introduced
ourselves
being
myself
and
chiang
chang
both
work
at
apple
and
several
months
ago
we
started
an
effort
of
provisioning,
a
ephemeral
clusters
on
our
super
kubernetes
cluster,
using
keyword
provider
as
the
infrastructure,
interface
or
virtual
machines
interface,
and
we
have
looked
into
openshift
provider
for
keyboard.
A
Alex
sorry
for
interrupting,
but
we
can't
see
your
screen
if
you
are
sharing
okay,
let
me
see:
what's
going
on
there.
A
You
might
need
the
host
key
vince.
Do
you?
Can
you
share
the
host,
give
it
alex
so
that
he
can
share
his
screen.
E
D
Oh,
that's
my
system
preferences
just
one
second,
and
I'm
going
to
continue
talking
so
we've
looked
into
openshift
provider
for
keyboard
and
it
it
looks
outdated
and
it's
still
based
on
actuators
logic,
and
we
would
like
to
be
as
close
to
the
upstream
as
possible.
So
we
actually
developed
our
own
and
we
posed
the
question
of
open
source
in
it.
So
currently
we're
pending
a
legal
process
at
apple
for
the
open
source
and
the
repository,
but
we
do
have.
D
We
do
have
options,
so
we
are
talking
about
open
sourcing
and
under
apple
namespace
in
github
or
ideally,
we
would
like
to
open
it
to
the
community
and
open
source
it
under
under
kubernetes,
6
namespace.
So
here's
our
motivation.
We
have
a
proof
of
concept,
we
keep
it
very
light
and
very
open
source,
so
the
logic
and
the
repository
is
based
strictly
on
kubert
provider.
D
So
as
long
as
you
have
a
kubernetes
cluster
large
enough
to
serve
as
the
infrastructure
cluster
and
you
have
keyword
running
on
it,
so
you
can
create
virtual
machines
on
it.
You
can
use
our
provider
for
cluster
api
to
create
clusters.
Clusters
can
be
used
for
development.
In
our
case,
we
use
these
clusters
for
integration,
testing,
functional
testing
of
new
features
for
kubernetes
that
we
develop
internally.
D
So,
basically,
just
just
a
few
bullet
points
internally,
we
use
it
at
apple
already
and
I
would
like
to
put
it
under
kubernetes
6
namespace.
This
is
our
use
case
for
it
and
yeah
the
the
main
benefit
of
it.
We
can
actually
customize
the
images,
underline
images
and
also
customize
the
kubernetes
components,
and
that
we
bring
up
for
ephemeral
clusters
that
that
we
use
for
testing,
and
here
is
the
design.
So
it's
it's
gonna,
it's
conformant
to
the
latest
cluster
api
apis
and
we
basically
just
added
several.
D
We
basically
just
added
several
controllers
in
keyboard
infrastructure
provider.
So
we
have
keyboard
cluster
and
keyword,
machine
data
structures
and
we
have
new
controllers
that
handle
keyword,
clusters
and
keyboard
machines.
The
rest
of
the
components
is
unchanged
down
the
line.
We
will
probably
also
be
updating
or
contributing
to
a
control,
plane,
provider
and
bootstrap
provider,
but
at
this
time
our
effort
is
limited
to
to
queue
to
keyboard
infrastructure
provider.
D
I
should
mention
a
couple
issues.
We
need
to
resolve
before
we
go
upstream
so,
first
of
all,
obviously
we
need
to
get
legal
approval
at
apple
to
open
source.
It
and
our
main
technical
issues
at
this
time
include
the
testbed
environment.
So
the
large
supercluster
that
I
mentioned,
that
we
run
an
apple
that
we
use
for
infrastructure,
runs
on
bare
metal
and
it
works
great.
But
when
we
try
to
do
development
using
the
kind
or
mini
cube,
we
actually
came
across
shortcomings
for
either.
D
So
we
see
network
issues
with
kind
which
does
not
allow
us
to
create
virtual
machines
in
it
and
in
mini
cube,
it
just
seems
too
slow.
So,
basically,
when
you
create
a
mini
cube,
you
now
have
three
levels
of
recording
in
terms
of
virtualization
and
it
slows
things
down
enough
that
it
doesn't
really
work
for
us,
so
we
will
probably
need
to
touch
base
with
the
community
or
find
another
solution
what
to
use
for
testbed.
D
But
as
long
as
you
run
a
kubernetes
infrastructure,
cluster
environmental,
you
know
from
our
test:
it
works
fine
and
currently
we
are
working
on
the
open
source
solution
for
the
load
balancer.
So,
like
I
mentioned,
we
have
some
or
yeah.
We
do
have
some
apple
specific
implementation,
but
we
split
it
out
of
this
repository,
so
we
have
a
repository
for
keyboard,
which
is
strictly
open
source
and
we
have
another
repository
that
contains
apple,
specific
implementations
and
in
apple
specific
implementation.
D
We
use
internal
implementation
for
load,
balancer
and
internal
internal
implementation
for
data
for
infrastructure
cluster
cluster,
but
for
the
open
source
version
we
are
looking
into
in
will
to
use
for
load
balancer
the
solution.
So
this
is
pretty
much
it
that's
what
we
bring
into
the
table
today
and
what
we
would
like
to
get
from
this
group
is,
if
it's
possible,
if
there
is
interest
in
open
sourcing
it
and
what
would
be
a
good
way
for
us
to
proceed
with
this.
E
Thanks
this
is
this
looks
great
one
question
that
I
had
like
after
the
presentation
is,
that
is
the
cube.
If
we
were
to
put
it
on
the
six
like.
Is
that
gonna
be
usable
out
of
the
box
without
the
private
providers?
E
D
Yes,
so
the
main
objective
of
this
is
to
keep
it
is
to
keep
it
very
light
right.
So,
basically,
if
you
look
at
other
providers,
they're
dependent
on
vendors
and
third-party
infrastructure,
with
this,
you
can
basically
bring
up
the
kubernetes
cluster
deploy
keyboard
on
top
of
it,
and
you
can
use
as
long
as
you
deploy
cluster
api
management
cluster
with
keyword
provider,
you're
good
to
spin
up
ephemeral
clusters.
That's
actually
our
development
test
pad
like
there
is
nothing
else
needed
there.
D
Also,
we
can
include
the
code
for
image
building,
so
we
are
using
plain
sentence
images
and
we
will
add
that
code
to
the
repository,
so
you
can
customize
underline
guest
to
us
images
or
you
can
use
ours
that
will.
We
are
also
going
to
publish
to
the
community
yeah.
So
if
we've
looked
at
openshift,
that
is
that
that
was
our
first
try
but
that's
outdated
and
like
I
I
don't
know
how
well
it
is
maintained.
D
Doesn't
it
it's
still
based
on
actuator
logic,
but
we
are
trying
to
keep
this
one
up
to
date
with
the
with
the
latest
data
structures
and
apis.
I
I
hope
that
answers
your
question.
Vince,
if
not,
let
me
know.
F
Yeah,
so
I
have
multiple
questions
for
different
parties.
I
guess
the
first
one
is
for
red
hat
members
on
the
call
like
what
is
your?
What
is
your
position
about
the
open
sourcing
of
this
provider?
Do
you
agree
with
it?
Do
you
agree
with
apple
leading
the
charge,
given
the
existing
provider
that
you
have?
You
know
general
consent
about
this.
A
C
C
You
know
apple,
taking
the
lead
here,
especially
given
the
last
commits
on
that
repo,
but
I
I
will
reach
out
internally
to
see
who
was
responsible
for
that
just
to
see
if
we
could,
you
know,
maybe
connect
with
the
apple
folks
to
see.
If
you
know
there's
some
commonalities,
we
can.
We
can
work
on
there.
F
C
If
we
get
into
a
situation
here
where
the
the
owners
of
the
coop
repository
and
openshift
want
to
keep
developing
it,
then
yeah.
We
would
definitely
want
to
put
something
in
the
readme
that
that
points
people
towards
the
apple
one
for
cluster
api-
and
this
is
you
know,
just
kind
of
an
organic
failing
on
our
part,
keeping
those
repositories
named
cluster
api,
when
many
of
them
are
actually
specific
to
our
machine
api.
So
yeah,
like
you
know,
apologies
for
any
confusion
there,
but
I
will
reach
out
internally
to
figure
out
who's.
F
That
thanks
for
that,
that's
just
absolutely
fine.
I
think
you
could
just
think
the
apple
folks
in
a
in
a
thread
in
the
quest
rpi
slack
to
just
finalize
the
the
whole
position
on
your
side,
so
that
that's
inspires
to
me.
The
right
question
that
I
have
is
for
the
apple
function
like:
is
it
possible
for
you
to
open
source
this
provider
before
we
directly
move
it
to
the
kubernetes
6
repository.
G
So
I
can
answer
this
question
so
yeah.
So
currently
that
is
part
of
the
plan
and
we
after
we
go
through
the
legal
review
process,
we're
gonna
put.
We
can
open
opened
up
into
the
apple
namespace
like
a
github
apple
and
then
because,
like
we
also
need
a
temporary
namespace
for
like
like
get
it
easier
to
go
public
and
and
then
we
can
build
our
like
a
cis
and
to
wear
it
up
into
the
prong
testing
grid.
So
that's
that's
part
of
the
plan.
F
All
right,
so,
once
you
pass
lego,
I
guess
you
can
again
use
the
quasar
api
slack
channel
to
notify
people
for
review.
So
it's
not
clear
to
me
who,
from
the
coaster
guy's
side,
does
review
on
your
providers
like
do
we
have
like
this
is
a
question,
for
course,
regardless
against
who
who
is
reviewing
new
providers.
E
I
guess
we're
not
really
reviewing
new
providers
like
we
just
make
sure
to
help
like
along
the
way
and
yeah
old,
like,
for
example,
to
set
up
test
infra
and
things
like
that
as
it
needed.
But
it's
up
to
the
provider
folks
and
maintainers
to
actually
make
sure
that
the
provider
is
compatible
with
our
contracts.
F
E
Correct
and
actually
I
was
I
was
going
to
say,
like
we-
should
try
also
to
reach
out
that
cubic
books
if
they
are
interested
in
reviewing
or
contributing
because
they
might,
they
might
be.
Maybe
one
sits
in
under
the
kubernetes.
G
Yeah
sorry
for
championing,
so
since
we
before
we
start
this
efforts
around
like
a
february,
we
contact
the
kubert
community,
especially
the
the
maintainers
for
the
keyword,
open
shift,
keyword,
provider
maintainers
and
we
checked
whether
it
is
working
out
of
box
and
the
conformance
to
the
upstream
seems
it's
it.
Cannot.
It's
only
works
on
the
openshift
distributions.
So
that's.
G
Why,
like
we
start
this
effort,
and
but
but
we
do
agree
that
if
we
can
put
in
well
like
the
keyword
maintainers
and
then
they
can
give
us
help
on
the
current
keyword
specific
parts,
so
that
would
be
wonderful
and
we'll
reach
out
to
them
for
sure.
F
It
would
be
nice,
I
guess
I
have
a
final
question,
which
is
like
the
most
the
least
relevant
one,
and
what
do
the
course
republicans
think
about
the
cup
k
abbreviation?
Do
you
want
to
reserve
it
from
something?
Else?
Importantly,
is
it
okay
to
use
it
here.
E
And
I
would
call
it
cap
to
be
honest,
because
cap
k
is
like
super
close
to
cap
bk,
which
is
the
cube
dm
provider
inside
a
cluster
api
yeah.
I
think
we
need
to.
I
don't
know
if
we
still
maintain
that
glossary
somewhere,
but
we
should
probably
add
it
once
it's
in
other
things,.
E
H
F
So
so,
going
back
to
the
question
that
I
had
for
it
had
potentially,
if
you
gain
interest
into
contributing
to
this
provider,
I
would
like
that
we
have
a
easy
way
to
grant
red
hatters
approval
and
review
privileges
in
this
repository.
If
they
have
interest
because
they
have
the
now
they
worked
on
the
original
provider,
I'm
not
sure
if
that's
fair
or
not,
but
I
just
don't
want
to
block
people
out
of
being
maintainers
of
this
project
and
we
should
try
to
leverage
basically
an
easy
contributor
ladder
in
their.
F
A
A
Okay,
yeah:
let's
move
to
the
next
item:
sagar
you
have
the
next
one.
I
Yep
thanks
so
folks,
yeah
thanks
for
opening
the
issue
raj,
so
I
just
basically
created
this
issue
to
kind
of
like
as
a
net
to
catch
all
of
the
use
cases
that
folks
in
the
community
might
have
so
just
a
brief
introduction
of
what
this
issue
is
about.
So
we
were
looking
at
introducing
a
way
wherein
you
could.
You
could
basically
attach
lifecycle
hooks
to
the
capi
objects,
so
a
little
bit
more
about
it.
I
So
currently
we
have
the
two
machine
deletion
hooks
present,
wherein
we
add
annotations
to
the
machines
and
based
on
the
presence
of
those
annotations,
the
reconciliation
of
the
machine
pauses
until
those
annotations
are
removed.
So
that's
a
very
specific
implementation
for
the
machine
object
that
is
present.
I
We
were
thinking
if,
like
folks,
have
ever
come
across
any
use
cases
where
they
need
some
like
they
need
to
pass
the
reconciliation
of
like
other
cappy
objects,
namely
say
cluster
object
or
the
kcp
object
for
to
to
do
some
pre-processing
or
to
run
some
custom
logic
on
their
side
before
having
the
reconciliation
of
the
object
to
move
forward.
I
So
that's
basically
the
purpose
of
this
issue.
If,
if
y'all
think
that
there
are
some
use
cases
that
that
you
would
benefit
from,
I
would
definitely
want
to
hear
more
about
them.
That
would
help
us
think
of
a
generic
enough
solution
that
could
that
could
maybe
work
for
everyone.
I
guess
that
is
about
it.
I
I
would
like,
if
there
are
any
like
follow-up
follow-up
questions
after
the
meeting
itself,
I
mean,
like
you,
could
reach
me
out
on
slack
or
on
the
github
issue
itself
and
you
could
have.
We
could
have
a
discussion
and
I
can
share
my
thoughts
about
what
we've
been
thinking.
A
B
A
B
Okay,
so
then
I
take
this
one:
okay,
so
yeah.
What
was
the
topic?
So
the
topic
is
classic
class
practice.
So
I
start
with
some
context:
we're
currently
implementing
cluster
class
and
the
the
current
version
of
the
proposal
allows
to
specify
cluster
class.
Like
this,
you
have
all
your
reference
templates
and
those
templates
are
then
used
to
calculate
a
cluster
machine
deployment,
etc.
B
One
limitation-
and
that
was
something
which
was
in
future
goals-
is
that
those
templates
cannot
be
customized
in
any
way.
So
if
you
have
several
clusters,
which
are
using
the
same
cluster
class,
they
are
all
almost
the
same.
There
is
some
way
to
customize
them
a
little
bit.
So,
for
example,
you
can
change
the
version
or
you
can
set
some
labels,
but
that's
it
there's
nothing
else.
B
You
can
do
with
the
current
state
of
the
proposal
and
the
current
state
of
the
implementation
of
that
implications,
work
in
progress,
but
so
the
problem
with
that
is,
if
you
have,
for
example,
you
want
to
configure
an
heb
proxy.
It's
a
standard
example
right
now,
then
you
would
have
an
and
you
have
a
specific
url.
You
can
put
it
in
a
cluster
class,
but
then
every
cluster
would
have
to
use
the
same
proxy
or
use
another
cluster
class.
B
So
there's
not
a
lot
of
different
kind
of
things
you
can
do
with
one
classic
class
and
the
idea
of
classic
class
patches
or
transformations
is
to
make
that
more
dynamic
and
flexible
so
that
you
can
customize
your
cluster
as
yeah
as
an
end
user.
Basically,
so
that's
that's
the
overall
idea,
so
yeah,
that's
it
okay,
so
some
details.
So
the
idea
is
we
want
to
specify
how
classic
class
patches
or
transformations
work
as
the
problem
is
kind
of
huge.
B
We
split
that
up
a
little
bit,
so
we
split
it
in
basically
two
or
three
parts,
depending
on
how
you
split
it
and
the
first
one
is
we
try
to
define
variables
so
the
ideas
in
a
cluster
class.
You
have
variables
and
schema
schemas
of
variables
and
in
a
cluster
topology
you
can
provide
variables.
So
that's
kind
of
the
input
where
an
end
user
can
say.
Okay,
I
want
to
have
this
http
proxy.
B
So
that's
the
first
part
and
the
document
which
is
linked
in
the
meeting
notes
is
currently
only
about
variables
and
the
second
part
which
we
are
currently
working
on
and,
I
think
we'll
add
it
to
the
document.
The
next
one
or
two
weeks
at
the
latest
is
patches.
So
the
ideas
you
now
have
variables
and
you
want
to
use
the
variables
to
customize
the
cluster
class
or,
more
specifically,
you
want
to
use
them
to
customize
the
reference
templates
here.
B
So
that's
the
high
level
overview,
but
I
think
you
will
have
some
time,
so
I
would
dive
a
little
bit
into
the
current
variables
doc
and
how
that
would
work
and
yeah
yeah.
As
I
said,
the
idea
is
to
gather
some
early
feedback,
because
the
big
problem
and
the
variables
are
already
a
huge
problem
by
itself,
because
you
have
to
think
about
the
way
you
can
define
it
and
you're,
basically
using
crt
schemas.
B
For
that,
then,
you
have
to
think
about
how
you
how
to
swipe
yours
made,
might
change
over
time
or
how
you
can
remove
a
variable,
and
so
we
already
put
it
out
there
to
get
some
feedback
and
even
before
we
have
patches.
So
that's
some
important
thing
to
remember
when
reviewing
the
stock
that
we
will
add
patches
the
next
few
days
or
weeks,
okay,
so
the
document
is
written
in
a
way
that
it
describes
the
delta
to
the
current
cluster
class
proposal.
B
So
it's
a
little
bit
complicated,
but
I
think
it's
easier
than
converting
the
markdown
back
to
google
doc
and
then
trying
to
modify
it
somewhere
in
some
way.
So
what
is
the
delta
so
delta?
Is
we
we
added
a
new
goal
and
the
goal
is
now:
we
want
to
provide
a
mechanism
and
to
to
be
able
to
to
transform
the
cluster
class
for
a
specific
cluster
and
that's
why
we
of
course
stopped
according
future
future
goal.
We
add
a
new
one
in
the
first
iteration.
B
We
want
to
have
some
basic
implementation
and
ideas.
We
will
introduce
variables,
but
we
don't
want
to
start
with
the
full
feature
set.
That's
something
I
come
back
to
below,
but
the
rough
ideas.
We
will
start
with
some
basic
types.
We
don't
want
to
allow
nested
object
and
arrays
and
all
the
other
stuff
you
can
do
with
crds
in
the
first
iteration.
So
the
future
goal
is
to
extend
the
first
implementation
by
support
for
complex
types.
B
We
have
two
new
user
stories.
One
new
story
is
as
an
end
user,
which
creates
the
cluster.
I
want
to
be
able
to
customize
my
cluster.
So
that's
what
I
was
talking
about
before
and
the
other
use
story
is
from
the
point
of
view
of
a
cluster
class
offer
and
as
a
clustering
class
offer,
which
is,
I
guess,
in
the
open
source
community,
it
will
be
a
cluster
api,
aws
maintainer,
for
example.
B
As
an
author,
I
want
to
be
able
to
configure
in
the
cluster
class
how
end
user
can
later
customize
the
cluster.
So
I
want
to
define
the
schema
of
the
variables
how
the
patches
look
like
things
like
that,
so
I
I
think
I
skipped
the
destruct
stuff
for
now.
I
just
show
the
example
and
then
go
back
up
again
afterwards.
So
how
would
this
actually
look
like?
So
in
the
cluster
class?
We
add
a
new
variable
section
and
we
have
an
array
of
variables.
B
B
So
we
don't
have
to
reinvent
the
wheel
and
in
the
first
iteration
we
will
only
allow
basic
types.
The
basic
types
are
integer
numbers
numbers
are
integer
and
floats
strings
and
boolean,
and
we
will
have
some
validation,
so
we
can
do
things
like
minimum
maximum
min
length
same
feature
set
as
crds,
at
least
regarding
basic
types,
and
we
will
also
have
defaulting.
So
I
think
I
dropped
it
here.
B
I
had
it
before,
but
what
you
can
do
is
you
can
say
you
have
a
default
here
and
if
the
value
is
not
provided
by
the
end
user,
then
you
will
get
this
default
instead
of
instead
of
nothing
yeah.
So
that
would
be
how
a
cluster
class
looks
like
when
you
then
create
your
cluster.
We
will
have
a
new
variable
section
in
topology
and
there
you
can
set
the
variables
as
an
end
user.
B
B
A
J
So,
first,
first
of
all,
thanks,
stefan
for
all
this
work
and
the
research
on
open
api
schema
and
crd
and
validation.
I
think
that
having
variable
and
paving
the
path
for
path
and
transformation
is
super
important
for
cluster
class
and
they
see
a
huge
value
in
adopting
open,
va
schema,
because
this
will
allows
us
to
have
typed
variable
and
also
have
a
validation
on
top
of
it,
and
this
will
prevent
the
user
to
do
error
also
and
when
adopting
cluster
classes.
So
it
will
be
comple.
J
It
will
be
powerful
in
the
cluster
class
side,
but
it
will
be
really
simple
and
and
user
friendly
in
the
cluster
side
of
the
things.
So
thank
you
for
this.
F
I
may
have
missed
some
of
the
details
about
variables
because
I
was
distracted,
but
is
it
really
a
variable
but
to
the
extent
where
the
user
can
change
it.
B
Yes,
it
definitely
is
so
when
you
change
your
variable
in
your
cluster
topology,
we
will
reconcile
the
change.
B
It
depends
on
how
this
variable
is
used
in
patches.
So
if
you
change
a
variable
which
affects
let's
say
all
your
machine
deployments
in
a
cluster,
then
it
will
reconcile
all
your
machine
deployments
and
cluster
because
it
has
to
if
it's
yeah,
so
it
will
have
to
do
what
it
has
to
do.
Apart
from
that,
it
won't
be
expensive,
but
it
depends
on
the
change.
K
To
see
you
hey
thanks
define.
I
have
a
couple
of
questions.
First,
one
is:
do
you
have
an
example
in
the
dock
of
like
actual
usage
of
the
variables
in
templates
or
in
objects.
B
K
Other
question
was:
what
happens
if
you
try
to
change
a
variable
that
is
used
on
an
immutable
field.
B
That's
a
good
question,
so
without
further
validation
it
will
the
reconsolation
will
fail,
because
what
we're
currently
thinking
about
is
you
have
your
variables
and
patches
in
the
cluster
class
and
without
any
further
validation.
It
will
just
try
to
patch
your
objects,
your
infrastructure,
cluster
or
whatever,
and
then
reconciliation
would
fail.
K
B
K
Okay-
and
my
last
question
is
how
is
it
possible
to
have
like
a
variable
default
or
value
be
based
on
the
value
of
another
variable
like,
for
example?
Let's
say
I
want
my.
I
have
this
like,
like
virtual
network
name,
and
I
want
the
default
to
be
the
cluster
name
plus
some
string
is
that
possible.
B
J
Yeah,
I
have
a
comment
about
the
the
problem
about
a
variable
changing
an
immutable
field,
so
I
I
think
that
a
good
way
to
see
these
is
that,
let
me
say,
variable
patches,
provide
an
exp,
an
extension
point
that
allows
cluster
class
outward
to
change
things,
but
in
a
given
that
cluster
classity
at
the
end
should
should
be
agnostic
that
the
only
things
that
cluster
class
knows
is
the
contract.
J
Basically,
it
is
up
to
the
patches
outer
to
make
sure
that
a
variable
is
used
in
a
legitimate
way.
So
this
is
kind
of,
and
it
is
an
extensible
point
and
it
is
up
to
the
user
to
use
it
well.
J
We
will
do
some
validation
as
much
as
possible,
but
I
don't
think
that
we
can
validate
whatever
is
provider
specific.
It
will
be
hard
to
do
online.
Maybe
we
we
will
do
some
dry
one
and
we
will
think
about
it.
E
Yeah,
this
is
going
to
say,
like
a
drive
around
in
costa
rica
could
actually
do
like
work
like
if
the
web
books,
like
defaulting
validation,
web
books,
don't
have
any
side
effects,
but
so
for
cecil.
You
asked
for
an
example
of
what
it
looks
like
in
the
cluster
right.
If
everyone
correctly.
E
E
Oh
use,
okay,
I
see
yeah
yeah
on
the
interpolation
side
of
things
like.
I
think
the
patching
side
like
should
support
some
interpolation
of
variables,
like
with
like
some
sort
of
template
style.
Then
you
can
say
like
I
want
to
use
this.
For
example,
I
wanna-
I
don't
know
like.
Maybe
a
built-in
one
could
be
the
clustering
name
so
that
you
can
interpolate
it
with
other
availables
as
well
and
then
make
up
the
patch.
E
F
B
I'm
not
sure
maybe
someone
else
has
an
opinion
so
right
now
we
are
just
building
it
for
the
purpose
of
of
using
it
in
cluster
class
and
abstracting
away
everything
below
a
class
class.
B
So
I'm
not
sure
if
the,
if
it
makes
sense
to
have
similar
things
like
variables
and
patches
and
other
low-level
resources.
F
E
Yeah
I
mean
you
could
apply
to
lower
level
objects,
though
within
the
a
cluster
class.
I
don't
I
mean
if
you
have
any
examples
in
mind
like
I.
I
personally
wouldn't
want
to
apply
everywhere,
because
then
there
is
going
to
be
layers
of
templates,
which
are
like
really
hard
to
debug,
usually
something
to
consider.
K
So
see
yeah
I
was
just
gonna
say
actually
that's
a
good
question
and
also
makes
me
think
like
the
way
that
it's
defined
right
now.
I
guess
it's
all
part
of
topology,
but
the
variables
aren't
like
strictly
associated
with
the
class
right.
No,
I
get
never
mind.
I
guess
it's
part
of
topology.
So
that's
all.
A
A
L
Thank
you.
I'm
I'm
brad
wilson,
I'm
here
with
stephen
garf
and
shreya
we're
from
twilio.
We
just
started
work
very
very
recently
on
a
new
cluster
api
provider
that
so
far
we've
been
calling
cluster
api
provider
live,
but
that's
that's
kind
of
up
in
the
air.
We
opened
an
issue
to
talk
about
supporting
our
raw
cube,
adm
format
for
the
bootstrap
provider.
L
We
don't
have
it
on
the
agenda
or
anything
we
just
we
just
kind
of
joined
the
meeting
this
morning
at
the
last
minute,
but
we
know
that
there's
an
ignition
pr
that
we
know
there's
an
ignition
pr
that
is
being
worked
on.
Currently
we
thought
about
perhaps
requesting
a
new
format
where
we
just
have
a
raw
cube,
adm,
config,
so
yeah.
We
wanted
to
just
join
the
meeting
and
get
a
feel
for
the
community
and
then
see
if
we
could
maybe
start
some
conversations
about
that
at
some
point.
A
Go
ahead:
hey
thanks!
This
is
akshay,
I'm
from
nutanix.
We
also
have
started
on
the
cluster
api
very
recently,
so
just
wanted
to
get
to
know
people
in
the
cluster
apsec
and
how
community
functions
and
any
opportunities
for
us
to
contribute
and
help.
That's
why
I'm
joining
the
meeting
today?
This
is
my
father's
meeting.
A
Welcome
welcome
akshay,
welcome
to
cluster
api,
and,
if
you
want
to
contribute,
you
can
definitely
take
a
look
at
the
good
first
or
health
issues
in
cluster
api.
Maybe
that's
a
good
place
to
start
all
right.
Yeah
I'll,
definitely
take
a
look
thanks
thanks.
Everyone.
A
Okay
looks
like
that's
it
thanks
everyone
for
joining.
You
can
get
back
10
minutes
of
our
time.