►
From YouTube: 20201202 Cluster API Office Hours
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
Hello,
everyone
and
welcome
to
the
cluster
api
office
hours
meeting
today
is
december,
2nd
and
let's
get
right
into
it.
If,
for
folks
that
are
new-
or
this
is
your
first
time
at
this
meeting-
welcome,
we
do
have
meeting
etiquette
if
you
would
like
to
speak
up
like
please
raise
your
hand
in
the
participant
list
and
I'm
gonna
post
a
link
to
the
agenda
for
today.
A
If
the
chat
shows
up
feel
free
to
add
any
psas
or
topics
that
you
would
like
to
bring
up
all
right,
so
we
have
one
psa
since
you'll
see
you
typing
it
up
real
fast,
so
take
it
away.
B
Thanks
hey,
I
just
wanted
to
mention
that,
there's
a
bug
that
we
found
in
the
recent
days
in
kcp
upgrade
where
basically
there's
a
race
condition
and
the
lcd
member
gets
deleted
as
it
joins.
This
affects
zero.
Three
eleven,
which
is
the
latest
cappy
release
and
the
bug
has
the
bug
fix,
has
merged
thanks,
vince
and
everyone
else
who
helped
investigate
really
quickly,
and
I
think
the
plan
is
to
release
that
as
part
of
zero.
Three.
C
B
A
B
So
it
affects
all
releases,
but
it
became
more
frequent
in
the
zero
through
11
release
because
the
order
was
changed
and
so
it
make
it
became
way
more
likely
that
that
race
condition
would
happen.
So
it's
possible
that
you
run
into
it
earlier,
but
it's
more
likely
in
zero.
Three
eleven.
A
And
to
give
just
a
little
bit
more
context,
what
was
happening
was
that
like
a
node
would
come
up
and
then
kcp
would
like
match
the
nodes
with
the
the
control
plane
label
assigned
to
it
by
querying
the
cluster,
but
the
new
node
didn't
have
the
label
yet
because
cubitium
didn't
finish,
joining
the
control,
plane
and
kcp
would
like
remove
that
number.
A
So
now
we
switch
to
have
to
query
like
our
own
view
of
the
world,
so
like
the
machines
in
the
cluster,
they
have
a
node
graph
attached
to
it
and
then
node
is
equivalent.
Now
that
we
know
it's
a
control,
plane
node,
because
kcp
created
it
and
we
compare
with
that
list
now
before
doing
this
reconciliation
tv,
we
actually
do
check
that
all
the
machines
under
kcp
ownership
have
on
a
node
ref,
so
we're
safe
like
if
some
of
them
don't
have
it.
Yet
we
will
wait
for
them
before
proceeding.
A
In
terms
of
releasing,
I
was
hoping
probably
next
week
to
actually
release
zero
through
12,
given
that,
like
then,
that's
probably
gonna,
be
you
know,
the
last
two
weeks
of
december
are
usually
kind
of
a
wash
and
we
should
give
a
stable
release
out
there
for
folks
that
want
to
try
it
out
so
yeah
like
if
we
can
get
like
the
various
bug,
fixes
or
features
or
whatever
else
for
zero
through
12
lined
up
this
week,
then
yeah
like
late
next
week,
we
can
probably
release
maybe
on
the
10th.
A
All
right,
let's
move
on,
I
have
the
other
discussion
topic
here.
I
like
to
propose
to
relax
our
backboarding
guidelines
for
the
current
stable
release.
So
a
little
bit
of
background
like
why
this
change
is
being
proposed,
is
right.
Now
we're
working
on
alpha
four
on
the
mainline
branch
by
zero.
Three
stays
like
our
current,
stable
release,
and
initially
we
thought
like
well
we're
only
gonna.
A
Do
bug
fixes
critical
bug
fixes
for
zero
three
year
obese,
but
then
we
kind
of
realized
that
that's
kind
of
like
a
little
bit
aggressive.
Given
that
you
know
we
need
to
keep
supporting
the
current
release
that
is
currently
stable
and
folks
who
might
be
using
it
in
production.
A
So
what
what
this
pr
is
doing
is
amending
our
contributed
guidelines
and
pretty
much
saying
that,
like
it's
okay
like
if
we
want
to
backboard
even
small
features-
and
I
mean
like
features
of
like
a
hard
to
kind
of
size
up
but
like
the
thing
is
like
there-
there
must
be
no
breaking
changes.
That's
like
the
baseline
like
to
get
something
back
forwarded.
So
if
there
are
no
baking
changes,
then
we
can
decide
to
backboard
something.
A
The
only
requirement
is
that,
like
a
day,
one
of
the
maintainer
has
to
approve
that
change
and
the
backboard
that
and
that's
yeah,
the
only
requirement
that
we
have
and
no
major
refactors
thanks
to
ceo.
Yes,
we
don't
want
to
refactor
code.
We
want
to
keep
the
code
stable.
The
refactor
should
come
in
the
next.
A
Versions,
does
that
sound
good
to
everyone?
Does
anyone
have
any
concerns?
Probably
the
majority
of
us
would
be
happy
about
this,
but
you.
A
Know
all
right,
I
see
a
lot
of
plus
ones,
so
we
can
probably
well,
I
will
mend
these
and
then
we
will
merge
it
and
awesome.
Let's
keep
going.
I
think
vincent
is
next.
Go
ahead.
E
Hey
and
I'd
do
this
on
the
agenda
just
kind
of
see.
You
know
on
a
comment
based
on
having
a
conversation
around
there's,
a
few
ways
of
tackling
alternates
to
cloud
emit,
one
of
them
being
ignition,
and
there
was
an
issue
open
by
nadir
of
ways
to
handle
alternate
ways
like
ignition
or
talos
or
otherwise.
Part
of
that
relates
to
secrets
handling,
so
the
first
linked
37
61
is
kind
of
more
the
design
discussion.
There's
not
a
lot
as
far
as
the
design
proposal
goes
there.
E
I
guess
part
of
my
question
is
like
how
best
to
help
move
this
forward.
We
are
for
flat
car
working
on
effectively
kind
of
a
couple
of
different
approaches.
We
had
started
down
some
of
the
infrastructure
providers,
but
reworking
cloud,
and
it
was
had
some
issues,
so
we
actually
have
had
started
working
on
a
pr
that
a
work
in
progress,
pr
which
is
the
next
link
there
to
bring
in
an
ignition
bootstrap
provider,
and
there
are
some
the
biggest
hangout
there.
E
Besides
rebasing,
it
is
around
how
to
handle
secrets
management
so
we're
currently
working
on
now
a
rebase
on
the
v1
alpha
4.
So
we
have
a
local
work
in
progress
branch
that
we'll
probably
close
this
one
out
and
open
up
a
new
one
or
force
push
it
or
whatever
for
v1
alpha
4
and
try
and
address
the
secrets
management.
E
But
I'm
just
seeing
like
a
few
of
these
things
are
kind
of
circling
around
of
we
could
work
around
it
in
some
different
ways,
but
this
kind
of
design
proposal
decision
seems
like
something
that
would
later
refactor
it
again
so
kind
of
just
looking
for
next
steps.
Either
we
work
around
it
for
now
in
kind
of
a
new
bootstrap
provider
and
each
of
the
each
individual
infrastructure
provider,
or
we
also
move
ahead.
F
Hey
vincent,
so
I
think
we
are
definitely
interested
in
doing
this
or
coming
up
with
some
sort
of
design.
So
probably
the
best
sort
of
generic
advice
I
could
give
is
that
we
find
a
group
of
folks
who
can
set
up
some
regular
times
to
chat,
could
be
on.
Zoom
could
be
async
and
come
up
with
some
design
proposals,
and
maybe
some
pocs.
F
I
don't
know
that
there's
really
too
much
else
that
we
can
do,
because
we
need
to
be
talking
to
figure
out
what
would
make
the
most
sense.
So
I
think,
like
I
know,
nadir
has
been
doing
a
lot
of
thinking
about
how
composition
would
be
useful
and
I
also
know
nadir
is
in
a
lot
of
zooms,
so
I'm
sure
there's
other
folks
as
well.
So
we
comment
on
the
issue.
We
could
find
a
time
to
chat
to
do
a
little
kickoff
there's
a
few
different
ways.
G
Yeah
admit
I
haven't
got
much
further
than
writing
a
couple
of
use
cases.
I
think
I
so
that's
simply
a
link
with
it's
just
got
the
temp
template
in
so
we
can
start
documenting
use
cases
and
requirements
yeah,
but
I
think
it
makes
sense
for
someone
else
to
take
the
lead
on
it,
but
I
don't,
I
don't
think
I've
got
sufficient
time
to
dedicate
to
it.
So
I'd
be
quite
happy
for
vincent
to
take
it
off.
E
Yeah
so,
and
so
in
that
kind
of
a
working
group,
would
it
be,
I
guess,
just
all
the
considerations
needed
for
composition.
E
I
guess
for
a
working
group
like
that
how
to
outline
just
what
would
success
criteria
or
what
all
things
to
be
considered.
I
figured
like
that's
that's
something
that
a
few
people
would
want
to
have
their
voice
in,
even
if
they're
not
doing
the
work
of
their
working
group
like
what
all
things
need
to
be
considered
here,
if
it's
basically
seeing
what
does
and
doesn't
work
well
and
how
it
ultimately
applies
across
all
the
different
infrastructure
providers.
A
Yeah
so
in
general,
like
for
a
proposal
of
this
caliber,
because
this
touches
like
quite
like
a
few
different
areas
of
the
whole
system-
and
we
might
want
to
definitely
like
to
reach
out
to
each
infrastructure
provider,
maybe
maintainer,
to
see
like
what
they
can
offer
in
terms
of
like
expertise
and
like
maybe
also
you
know,
code
and
things
like
that.
A
The
first
thing
that
I
would
do
is
like
definitely
to
open,
like
a
google
doc,
with
the
kept
template
and
definitely
focus
on
the
goals
and
on
goals
first
and
the
use
cases.
So
those
are
like
how
we
kind
of
define
them
the
scope
of
the
proposals.
A
To
be
honest,
like
this
might
be
split
into
a
couple
because
of
like
how
big
it
is.
But
in
you
know
it
might
be
the
like,
where
it's
stretched
forward
enough,
that,
like
it,
could
go
into
one
as
well,
it
all
depends.
I
can
help
along
the
way.
The
this
is
something
that
we
have
kind
of
had
to
build
for
a
while,
but
there's
things
that
touch
the
infrastructure
providers,
which
means
like
we
need
composable
extensions,
which
we
already
have
like
kind
of
previous
art
here.
A
Of
course,
for
success
I
would
say
like
definitely
the
google
doc
will
will
get
it
like
going
around,
and
then
people
were
voiced
and
usually
would
move
to
a
pr
there's
like
another
round
of
reviews
in
the
pr
it's
usually
a
little
bit
deeper
and
then
yeah.
We
merge
it
like
that.
That's
usually
this
the
success
here.
It's
like
to
actually
get
a
proposal.
G
And
you
yeah,
I
I
also
wanted
to
answer
on
the
sort
of
short
term
option.
So
I
I
am
I
I
don't
want
to
say
anything
in
stone,
but
we
have
done
some
enabling
work
in
class
api
provider
aws
for
you
to
you,
build
an
s3
sequence
back
end,
so
you
could
end
up
with
the
similar
bootstrap
providing
mechanism
that
we
have
today
that
we
use
for
cloud
in
it
as
in.
G
E
A
All
right,
let's
keep
moving.
H
Yes,
that's.
That's
me.
H
Okay,
okay,
so
so
I'm
from
red
hat,
I'm
working
on
openshift
within
red
hat,
and
we
are
in
the
process
of
enabling
windows
workers
on
openshift
clusters,
and
as
part
of
that,
we
have
had
to
make
some
changes
to
the
the
cluster
api
provider
apis,
and
so
we
opened
a
pr
daniel
from
red
hat
opened
it,
and
I
think
we
had
some
discussions
on
segwindows
also
about
it,
and
then
you
know
cecile
requested
that
we
show
up
at
this
meeting
and
sort
of
discuss
this
a
little
further.
H
So
to
give
you
a
quick
introduction,
the
way
we
are
doing
things
in
open
shift
for
adding
windows
workers
is
a
using
an
operator
model.
The
operator
watches
for
certain
objects
to
be
added
to
the
cluster.
H
If
we
find
that
that
object,
which
is
which
we
call
a
machine,
is
a
windows
vm,
we,
the
operator,
takes
care
of
preparing
the
node,
but
as
part
of
that,
we
expect
certain
things
to
be
present.
On
the
vm,
for
example,
the
vm
has
you
know
we
want
to
configure
the
vm
with
ssh
capabilities
and
that's
where
we
ran
into
trouble
across
cloud
providers
in
aws.
We
used
cloud
in
it
with
azure.
H
We
had
to
do
something
different
where
we
sort
of
used
it
use
the
unattended
xml
to
get
this
going
and
again
with
vsphere.
We
had
to
do
something
again,
which
is
completely
different,
so
we
went
through
the
proposal
from
from
james
and
it
seems
at
a
very
high
level
the
suggestion
there
is
to
use
cloud
base
in
it
across
the
board
for
all
providers,
and
there
seems
to
be
a
way
from
the
api.
H
The
cloud
provider
api
point
that
the
user
data
that
is
needed
would
then
be
moved
into
the
vm
in
some
way
and
then
you'll
use
cloud
base
in
it
to
run
those
commands
or
whatever
set
of
instructions
that
are
needed
for
that
vm
based
on
whoever
is
making
the
call
so
so
james
you
should.
Let
me
know
if
I'm
not
on
the
right
track
here.
Is
that
I'm
I'm
thinking?
That's
the
approach.
I
Yep,
that's
that's
about
it!
We're
just
using
cloud-based
init
in
place
with
the
qbdm
bootstrapper.
H
Okay,
so
we
don't
have
any
what
you
call
we're,
not
adverse
to
this
approach.
We
are
also.
This
would
actually
solve
a
problem
that
we
face
right
like
we
don't
have
something
like
cloud
based
in
it,
which
is
consistent
across
cloud
providers.
The
only
questions
that
came
to
our
mind
at
this
point
was
this
implies
that
you
need
to
have
the
cloud-based
package
installed,
pre-installed
or
pre-baked
into
the
windows,
images
on
cloud
providers
right
like
aws
and-
and
you
know,
azure
gcp.
H
I
So
the
way
that
we're
setting
it
up
currently
is
using
image
builder.
So
there's
a
couple:
pr's
open
for
image
builder,
the
azure
one
merged
a
couple
weeks
ago
and
there's
one
open
for
aws
amis
as
well
as
I
think
one
for
vser.
I
Yeah
I
mean
we
published
some
for
the
marketplace
just
like
we
do
with
the
ubuntu
images.
So
I
think
it's
I.
I
think
this
is
kind
of
their
approach
that
someone
correct
me
here,
but
I
think
cluster,
a
cluster
api
in
general
doesn't
really
concern
themselves
with
the
vm,
that's
being
provisioned
if
they
just
expect
some
certain
things
to
be
on
that
component,
and
so
I
think
most
of
them
are
it's
the
ideas
that
you
would
use
image
builder
to
customize
the
image
specifically
for
your
use
case.
H
Okay,
so
we
it
once
we
move
to
this
model,
that's
the
requirement
we
should
sort
of
express
to
you
know.
Customers
who
want
to
use
use
windows,
vms
or
windows
workers
in
in
kubernetes
or
slash
openshift
clusters
is,
is,
I
guess,
the
approach
that
we
are
going
towards
right.
H
Okay,
sounds
I
don't
have
any
questions
further
at
the
moment
from
my
side,
if
anybody
else
on
on
the
red
hat
side
have
anything
more
to
add,
please
go
ahead.
J
So
hi
everybody
christian
glenvik
here
yeah,
just
maybe
to
add
to
that.
We
do
things
on
aws
and
azure
quite
similarly
currently,
and
the
only
real
difference
is
that
the
the
bootstrap
data
which
is
written
to
disk
as
the
custom
data
which
we
use
to
set
up
ssh
and
then
with
that
configure
the
machine.
J
That
custom
data
script
is
on
aws.
It's
auto
run
on
startup
on
azure.
It
isn't
and
that's
kind
of
the
problem
we've
hit-
that
we
had
to
add
some
unintended
xml
to
login
and
run
that
script
on
the
first
boot.
Essentially,
so
I'm
not
sure
why
there's
this
discrepancy
between
this
is
for
windows.
Images,
of
course
why,
on
aws
it
runs
that
automatically
on
this
or
it
doesn't.
If
it
also
did
that
on
azure.
J
We
wouldn't
really
have
a
problem
here,
because
we
really
just
rely
on
ssh
access
to
configure
the
machine
thing
with
sftp
and
ssh.
G
Oh
sorry,
I
just
wanna
ask
clarifying
questions
so
openshift,
so
the
we're
using
cloud
base
in
it.
Our
intention
is
to
use
cloud-based
internet
because
most
of
us
are
using
the
cube,
adm
bit
strap
method.
So
we
have
the
qba
and
b
controller
that
is
creating
cloud
in
it
configuration
it's
kind
of
related
to
the
previous
topic
as
well,
but
right
now
we're
creating
cloud
init
configuration
data
and
to
make
that
work
seamlessly
in
windows.
G
We
expect
cloud
base
in
it
to
be
present
for
openshift
you're,
not
using
the
that
bootstrap
operator,
and
therefore
you
should
be
able
to
present
arbitrary
user
data
to
which
the
infrastructure
provider
will
just
inject
in
as
user
data
online
on
that
particular
cloud.
G
J
Yeah
we
do.
We
do
get
that
custom
data
that
bootstrap
data
onto
the
machine
without
problems.
It's
just
the
matter
of
that
machine.
That
image
is
just
a.
I
don't
know
where
it
comes
from
it's
from
microsoft.
It's
an
official
image.
It
doesn't
have
all
of
those.
You
know
it's
not
pre-baked
with
with
that
cloud
base
in
it
in
it,
so
we
needed
to
then
execute
that
custom
data
and
aws
does
that
automatically
if
it
exists
and
azure
doesn't
do
it.
J
Yeah-
and
I
I'm
not
sure,
but
I
doubt
that
we
had
retired-
are
going
to
move
to
the
cube
adm
bootstrapper
model
anytime
too
soon,
because
we
don't
use
it
anywhere
so
far
and
yeah
we've
been
able
to
work
around
it
quite
well,
and
it's
just
yeah.
I
don't
know
if
ezra
were
to
change
to
change
it
so
that
the
custom
data
if
present
were
to
execute,
were
to
be
executed
automatically.
Then
that
would
essentially
solve
all
the
problems
we
have
with
it.
J
A
I
have
four
people
with
chris
and
dear
cecil
and
david.
K
Yeah
so
my
first
question:
well
I'll
just
keep
it
short,
the
I'm
looking
through
the
cluster
api
provider
docs
for
azure,
and
when
you
go
down
to
the
point
where
it
talks
about
using
a
custom
image,
it
leads
you
to
a
page
that
doesn't
exist
so
before
we
sit
there
and
say
it's
on
the
customer's
responsibility
to
build
these
custom
images
for
cluster
api
to
work.
There
are
some
shortcomings
in
the
environment
for
supporting
the
customer.
K
To
actually
do
that,
and
I
just
wanted
to
confirm-
is
somebody
going
to
actually
make
a
reference
image
for
windows
because
nobody's
really
seems
to
have
owned
that
in
the
discussion
so
far,.
I
Yeah
so
yeah,
so
we
merged
a
pr
for
the
custom
image
reference
image.
Last
two
weeks
ago,
I
can
link
to
the
to
the
image
there
in
the
chat
and
we
have
two
other
ones
being
built.
On
top
of
that,
I
think
we
have.
I
In
in
azure,
they'll
be
published
to
the
marketplace
and
they'll
provide
just
like
we
do
with
the
linux
images,
we'll
be
providing
a
default
that
will
pull
the
latest
image
and
it'll
automatically
get
deployed.
So
I'm
not
super
familiar
with
aws
or
or
cat
v,
but
maybe
I
think
I
saw
gab
on
the
call
he
might
be
able
to
speak
to
that.
A
I'll
raise
my
hand,
never
mind
all
right
so
dear
and
then
david.
G
Yeah,
if
I
suspect
concretely,
what
you
might
want
to
do
is
so
this
pr
that
you've
got
open
in
capsi
with
the
creating
this
unattend
xml
file.
I
think
you
want
to
move
that
out
of
cap
c
into
your
controller
and
which
is
then
creating
the
vm
with
the
references
secret.
G
So
I
think
maybe
we
just
need
to
figure
out
where
this,
where
this
unattended
xml
goes,
but
you
make
sense
what
you're
trying
to
do.
I
think
you
just
need
to
move
it
to
the
right
layer.
J
L
David,
thank
you.
So
we
do
a
lot
of
this
publishing
stuff
in
in
azure
for
these
images
and
it's
not
as
as
simple
as
ami
and
in
aws
and
that's
unfortunate,
but
we
are
also
willing
to
help
if,
if
you
know
creating
custom
images
based
on
some
of
the
stuff
we
have
in
image
builder
is,
is
a
path
you'd
like
to
go
down.
We
can
definitely
help
get
those
images
baked
and
help
you
down
the
path
of
publishing
those.
If
that's
something
you're
interested
in
as
well.
J
Yeah
that'd
be
great,
so
thank
you.
First
of
all,
so
yeah
we
would
rather
not
have
have
our
customers
build
custom
images
in
the
first
place,
so
yeah
ideal.
Ideally
we
could
kind
of
get
around
that
and
just
use
either
unattended
config
or
get
an
image
that
has
a
cloud
base
in
it
in
it
already.
B
Thanks
yeah,
I
think
most
of
what
I
wanted
to
say
has
been
said,
but
I
just
wanted
to
say
like
yeah
we're
happy
to
collaborate
like
definitely
want
to
get
you
unblocked
and
get
your
solution
working.
We
just
want
to
make
sure
that
it
works
for
both
users
that
want
to
use
cubanium
and
users
that
don't
so
it
shouldn't
be
one
or
the
other,
and
also
like
my
main
concern
with
the
pr
as
it
stands
right
now,
is
it's
not
adding
much
context
for
users
who
might
land
on
this
and
think?
B
Oh
now,
I
can
use
capzi
to
use
windows
vms
and
it
doesn't
really
have
like
any
documentation
or
how
to
do
that,
and
it
requires
like
external
bootstrappers,
which
aren't
documented
anywhere
in
capsi.
So
I
think
we
should
just
like
clarify
this
a
little
bit
either
through
a
proposal
or
through
a
doc,
but
just
want
to
make
sure
that
users
aren't
confused
about
whether
cazy
does
or
doesn't
support
windows.
J
Yeah
absolutely
right
now
the
pr
doesn't
do
much
it
just
kind
of
gives
you
the
the
ability
to
to
provision
a
windows
machine
up
as
a
as
a
node
in
kubernetes
or
anything.
It's
just
gonna
sit
there
and
not
do
anything
really
yeah.
I
agree,
we'll
we'll
add
more
documentation
to
it.
M
Yeah,
I
just
wanted
to
revisit
that.
It
sounds
like
we
have
a
like
sounds
like
we're
going
to
have
upstream
images.
I
just
was
wondering
we
is
there
any.
Should
we
be
helping
somewhere
with
that
or
something
because
internally
we
need
to
start
testing
these
images,
and
I
know
we've
got
a
cap
v
image.
That's
gonna
be
able
to
go
in
hopefully
soon
and
we've
got
a
cap,
a
one.
I
know
gabs
working
on.
So
on
my
end,
I
was
just
gonna.
M
I
don't
I
don't
know
what
the
frequency
of
cap
of
cap
a
and
cap
v
image
publishing
is,
but
are
the
windows
published
images
gonna
be
at
the
same
frequency
as
upstream,
because
as
upstream
linux
images,
because
I
guess
those
are
happening
all
the
time
right,
like
the
publishing
of
the
linux
images
right.
A
I
mean
someone
has
to
do
with
an
image
builder.
First,
before
it
gets
the
jobs,
get
kicked
off
right
and
appear
yes,
okay,
yeah!
So
like
it's,
someone
has
to
kick
that
off
that
job,
like
it's
not
going
to
be
like
auto
magic.
A
Yeah,
so
I
think
that's
going
to
be
the
exact
same
for
that's
what
I
would
expect
from
the
windows
machines
right,
yeah,
okay,
so
I
have
daniel.
C
Yep,
so
I
just
wanna
jot,
I'm
sure
if
you
can
hear
me
the
point
about
this
pr,
which
I
posted
recently
as
the
windows
team
from
our
machine
api
requested.
So
the
code
there,
as
you
probably
mentioned,
is
just
sitting
around
and
not
doing
nothing
just
patience.
Windows
vmware
allows
this
to
do.
I
guess
the
counterpart
appear
from
microsoft.
Does
this
so
I'm
pretty
happy
to
close
it?
C
We
could
help
out
from
machine
by
perspective,
writing
proposal
and
communication
with
windows,
folks
about
how
they
want
to
consume
this
data
so
far
in
machine
api.
The
reason
why
it
looks
like
it
looks
right
now
is
because
we
provide
the
bootstrapping
data
in,
I
guess,
using
data
secret,
which
is
getting
into
custom
data
on
the
virtual
machine
specification
and
that's
all
we
do,
and
this
successfully
allows
us
to
patient
windows
vms
so
far,
yeah
just
wanted
some
input
on
this
as
well.
A
So
at
this
point
I
think
like
we
need
to
clarify
some
bits
so
image
builder,
there's
gonna,
be
support
for
windows,
publish
machine
like,
but
we
I
think
the
microsoft
folks
want
to
clarify.
This
is
not
my
like.
Microsoft
is
publishing.
These
images
is,
you
know,
based
on
the
kubernetes
image
builder
process,
and
these
are
reference
only
images,
but
customers
in
production
should
use
these
references
to
build
their
own
custom
image.
A
David
says:
you'll,
like
a
mountain
drive
track
here:
okay,
but
red
hat
folks
like
I
need
to
install
some
things
because
they
won't
use,
they
won't
be
able
to
use
the
cubidian
bootstrapper.
Is
that
correct?
Okay?
So
I
think,
like
your
suggestion
from
before,
like
to
actually
either
like,
create
a
different
bootstrapper
or
add
it
to
an
existing
bootstrapper
that
you're
using
it's
probably
the
correct
one.
There
was
a
question
before
it's
like:
are
these
images
going
to
be
published?
A
The
answer
is
yes,
but
they
should
not
be
used
as
like
kind
of
like
the
final
image
that
you
should
use
in
a
production,
environment
and.
A
Is
because
like
if
a
cva
happens,
we
don't
want
to
be
held
responsible
and
because
you
know
this
is
a
community
driven
project.
So
we
don't
have
like
automated
checks
for
these
base
images.
So
we
want
customers
to
kind
of
use
them
or
like
build
their
own,
but
you
know
we
want
to
provide
a
reference
for
them.
A
So
next
steps.
I
think
it's
kind
of
like
to
align,
maybe
in
the
capsi
office
hours
meeting
in
for
the
windows
support
and,
I
think,
there's
like
probably
a
couple
of
auction
items.
So
if
you
find
links
in
the
in
the
docs
that
don't
work,
please
open
an
issue
so
that
we
can
fix
it
or
open
up
here
to
fix
it.
That
would
also
be
great
if
there
is
like
a
vr
like
this
with
no
documentation.
A
Please
provide
use
cases
before
maybe
opening
like
a
pr
like
this,
because
otherwise
like
it
could
create
confusion
in
the
community
and
for
next
steps
like
we
need
to
kind
of
align
between
the
cube
idiom
bootstrapper
and
this
other
bootstrap.
That
red
hat
has
that
it's
using
for
openshift,
but
so
that
both
can
use
the
same
reference
image.
J
A
Thank
you
all
so.
Who's.
O
P
Sorry
about
that
yeah
I
wanted
to
give
a
shout
out,
in
particular
to
carlos
panado
who's,
been
reviving
the
cluster
api
provider
gcp
and
thanks
to
his
work
and
the
other
contributors
who
have
had
outstanding
prs
that
have
recently
merged
we're
getting
ready
to
cut
the
v030
release
tomorrow.
For
cap
g
that'll
be
targeting
cluster
api
v1
alpha
3,
and
then
we
can
start
work
on
v1
alpha
4.
Shortly
after,
if
you're
interested
in
helping
out
with
the
gcp
provider,
we
are
now
actually
moving
along.
A
Awesome
great
to
hear
and
double
shout
out
to
to
carlos
and
and
all
the
others.
Thanks
for
working
on
the
gcp
provider,
that's
a
long
time
to
what
they
were
released.
C
A
There
is
also
a
slack
channel
dedicated
to
gcp,
so
it's
called
tripia
gcp
on
the
kubernetes
lag.
So
if
you're
interested
in
contributing
feel
free
to
reach
out.
A
There
all
right
cigar
there
is
the
get
up
issue
about
cappy
secret
type,.
O
Oh
yes,
hey
good
morning,
guys
or
good
afternoon
good
evening,
so
this
was
this
a
small
issue
that
we
picked
up,
which
was
around
setting
the
type
field
for
secrets
which
are
generated
by
cappy
like
I
don't
want
to
discuss
more
about
this
issue
on
this
call.
I
just
wanted
to
point
out
that
there's
this
issue
going
on
and
there
are
some
comments
at
the
bottom
and
we
would
like
to
have
some
input
from
like
everyone
in
the
community
around
how
to
solve
this
issue.
O
There
are
a
couple
of
methods.
There
are
a
couple
of
ways
that
I've
laid
down
in
in
this
issue
so
like
if
you'll
get
some
time,
please
go
ahead
and
look
at
the
issue
and
post
your
thoughts
or
comments
around.
A
It
thank
you
just
just
to
give
some
context
works.
I
don't
know,
like
secrets,
have
a
type
field
that
you
can
use
to
kind
of
verify
that
you
have
created
that
type
or
that
you
can
read
it
rather,
and
the
type
field
is
immutable,
so
it
can
only
be
set
at
creation
time.
A
So
in
closer
api
we
started
creating
new
secrets,
for
I
think
right
now
only
the
bootstrap
secrets
with
that
type
field,
and
then
we
require
that
type
field
to
be
set
to
a
specific
string
and
then
cluster
api
will
only
read
that
secret
and
inject
it
into
the
machine
user
data
if
the
type
matches-
and
this
is
mostly
like
to
avoid-
like
stealing
other
secrets
that
you
shouldn't
read
the
the
problem
is
like
I
said
there
mentions
like
that-
the
type
cannot
be
changed.
A
A
This
is
like
we
need
like
a
kind
of
like
a
temporary
secrets
to
because,
like
we
want
to
make
sure
that,
like
this,
these
operations
are
safe.
So
there
is
some
there's
a
comment
here
that
kind
of
summarizes
like
a
few
ways,
if
you
have
any
ideas
or
concerns
how
to
do
this,
like
feel
free
to
comment
on
the
issue.
D
James
yeah,
I
just
had
a
question
about
that
last
issue
about
the
secret
type
thing:
is
it
really
a
problem
in
practice
when
we,
when
there
was
not
a
custom
type,
that
the
wrong
secrets
were
being
grabbed
like
accidentally
or
somebody
was
tampering
with
secrets
that
they
shouldn't
have
kind
of
wondering
how
big
of
a
problem
this
is
in
practice?
You
know
versus
like
an
over-engineering
effort.
A
This
is
more
of
a
best
practice
I
want
to
say
like
we
have
not
had.
I
don't.
I
don't
think
I
heard
like
anybody
like
actually
saying
like
someone
has
done
this,
but
it
is
an
escalation
path
from
close
ship
because
plus
should
be.
I
can
read
secret,
like
all
the
secrets
in
all
namespaces
so
technically,
like
you,
could
kind
of
steal
secrets
and
put
them
in
the
user
data
if
they
don't
have
the
exact
type
field,
which
is
not
ideal,
and
you
know
like
you,
could
cause
some
trouble.
A
So
it's
more
like
of
a
yeah.
I
don't
know
how
to
play
it
but
yeah.
It's
it's
definitely
a
suggestion
and
we
should
probably
do
it.
D
Okay,
so
it's
just
it's
in
the
name
of
security:
it
wasn't
something
that
we
observed
in
the
wild,
but
it's
a
best
practice.
Correct
yeah
got
it
thanks.
A
I'll
stop
sharing.
Thank
you
all
for
joining,
see
you
all
next
week
bye.
Thank
you
see
you
next
week.