►
From YouTube: SIG Cluster Lifecycle - Cluster API 21-04-14
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
And
go
for
it
all
right!
Welcome!
Everyone
today
is
wednesday,
the
14th
of
april
2021-
and
this
is
the
kubernetes
sig
cluster
api
meeting.
We
are
following
the
kubernetes
community
guidelines
for
this,
so
please
treat
others
as
you
might
expect
to
be
treated,
and
please
raise
your
hand
if
you
would
like
to
speak.
B
I
did
so
yeah
we're
releasing
today
zero
316,
you
should
probably
we've
got
the
release
like
in
a
couple
of
hours.
Zero
316
is
gonna
include
an
important
bug
fix
for
kcp,
which
is
currently
preventing
upgrades
because
the
so
the
fix
was
to
allow
like
a
field
to
be
mutated.
B
It
was
being
defaulted
and
it
was
before
nil
or
empty
everywhere
correctly
and
that's
being
defaulted
to
something
else
which
it's
causing
the
the
admission
to
say
like
no,
you
can't
the
the
vibration
to
say,
like
you
know,
you
can't
you
can
mutate
it.
So
thanks
for
everyone
that
actually
like
added,
also
tests
for
this
and
yeah
and
the
one
that
contributed
for
the
fix.
B
The
other
thing
is,
I
think,
fabrica
has
also
like
another
pr
open
that
fixes
like
an
issue
with
core
dns
for
121
and
that
should
be
merged
today,
as
well
before
the
release.
That's
actually
blocking
right
now
and
yeah
we'll
wait
for
that
before
doing
the
actual
release.
A
Okay,
great,
I
guess,
if
there's
no
questions
or
anything
about
that
psa,
I'm
not
seeing
any
hands
up.
The
next
thing
will
be
brian's
demo
on
distributed
tracing
of
cappy,
so
I
think
just
need
to
get
host
to
brian
and
I
will
stop
sharing
and
I
guess
the
other
thing
you
need
to
make.
Sure
too
is
we
keep
the
recording
going
too
right.
D
I'll
I'll
start
in
this
window,
so
can
you
see
some
yaml
and
the
command
line
yeah?
I
can
see
it
okay,
so
this
is.
This
is
what
I'm
gonna
run
for
my
demo.
I
have
just
set
up
a
kind
cluster
and
I've
run
kind
load.
The
images
and
I've
run
cluster
cuttle
in
it.
D
So
some
of
you
will
have
seen
a
demo
a
few
months
back
where
I
I
added
distributed
tracing
code
into
various
of
the
cluster
api
components
and
and
that
that
worked,
but
it's
it
turned
into
a
lot
of
a
lot
of
kind
of
time
to
to
write,
kept
and
figure
out
how
to
change
all
the
bits
and
there's
a
whole
standardization
piece
kind
of
in
the
way
of
that
so
along
the
road.
D
D
So
so
I
guess
a
lot
of
you
have
seen
kubernetes
events.
You
know
you
could
type
cucuttle
get
events
and
well,
in
fact,
let's,
let's
do
that
now.
D
So
this
is
a
crazy
idea
which
I'm
now
going
to
demo.
So
I
have
let
me
just
check
yep.
I
have
my
program
running
in
the
background
which
is
listening
to
events,
and
I
also
have
a
jaeger
distributed
tracing
system
running,
so
I'm
just
going
to
go
ahead
and.
D
And
so
this
is
a
cluster
cuttle
created
yaml
with
with
one
control,
plane,
node
and
one
worker
node,
so
it
started
doing
things,
and
so
I'm
gonna
switch
over
oh
yeah.
I
have
to
stop
sharing
and
switch
and
then
start
again
with
my
jaeger
window.
D
I
hope
I
got
the
right
one,
so
this
is
jaeger.
This
is
jupiter
tracing
system,
and
in
here
I
have
so
it's
been
picking
up
traces,
a
lot.
These
ones
have
kind
of
funky
names,
which
is
is
kind
of
a
bug,
but
we
have
the
various
parts
of
kubernetes
and
and
cluster
api
and
so
on
emitting
traces.
So
if
I
I've
been
running
this
demo
a
few
times
just
to
make
sure
it
works
okay,
so
the
latest
one
is
a
few
minutes
ago.
D
Okay,
two
minutes
ago
seems
fair
enough.
So
what
do
we
have
here?
D
So
basically,
okay,
there's
two
things
going
on
the
events
that
are
emitted
and
then
there
weren't
really
enough
events
to
make
the
screen
look
interesting.
So
I
added
conditions
so,
basically,
every
time
a
condition
transition
happens
and
every
time
an
event
is
emitted.
We
get
a.
We
get
a
line
out
here
and
it's
it's
very
work
in
progress.
This
this
stuff
shouldn't
be
kind
of
marching
down
the
screen
like
that.
But
we
do.
D
We
do
get
a
kind
of
a
picture
of
what's
going
on
inside,
oh
yeah,
we
can
click
on
these
things.
I
just
wanted
to
drag
that
over.
D
There
we
go,
we
do
get
a
kind
of
a
picture
of
what's
going
on
inside
cluster
api.
In
this
case.
It's
it's
set
up
two
nodes
and
we
can.
We
can
click
on
these
things
and
we
can
so.
This
is
where
I've
extracted.
D
This
was
a
condition
change.
It
changed
to
true.
At
this
time
we
maybe
find
that
oops.
Is
that
an
event?
No
that's
another!
That's
another
condition
transition
anyway,
so
the
so.
This
is
all
done
by
a
program
called
k-span
which
is
in
the
weave
works,
experiments
repo,
no
changes
to
no
well,
you
know
no
caps
required.
D
No
there's
one
little
change
to
the
source
code
of
cluster
api
to
make
this
look
a
bit
nicer,
which
is
all
these
names
like
machine
controller
and
cluster
controller
and
machine,
no
cube,
adm
control,
plane,
controller
unmodified.
D
All
those
things
come
out
as
as
the
single
word
manager,
which
I
was
I
thought
was
dull,
so
I've
I've
got
a
I've,
got
a
pr.
Well,
there's
a
couple
of
pr's
and
a
bit
of
discussion
going
on,
but
but
basically
setting
the
so-called
field
owner
to
machine
controller
or
whatever
means
I
can
pick
it
up
here
and
make
the
screen
more
interesting.
A
E
Thank
you
and
brian.
This
is
very
cool
huge
fan
of
using
tracing,
particularly
in
debugging
scenarios,
but
I
I
so
I've
always
thought
of
spans
as
a
space
of
time.
So
things
happen
across
this.
This
this
time,
boundary
and
often
spans
are
used
to
say,
here's
a
function
and
here's
the
beginning
and
end
of
a
function,
and
then
spans
are
usually
decorated
with
like
tags
to
signify
different
event,
changes
or
perhaps
errors,
did
you
evaluate
using
tags
rather
than
using
spans
themselves.
E
Yes,
so
your
events
come
off
as
spans,
but
you
can
also
see
that
the
spans
are
like
zero
microseconds
right,
so
it
kind
of
it.
I
don't
want
to
say
glutes
it.
It
makes
the
the
actual
function
structure
very
noisy,
whereas
I
would
have
expected
a
span
where
a
function
had
occurred
and
then
have
maybe
tags
on
there
of
detail
to
yeah.
D
You're
using
this
word
tags
and
you're,
using
it
in
a
different
way
to
what
I
understand
so
so
maybe
you
just
come
from
a
different
technology.
These.
These
are
the
tags
in
in
jager,
in
open
tracing
and
in
open
telemetry.
E
D
Yeah,
I
think
you're
talking
about
where
you
have
effectively
log
lines
within
a
span
and
open
tracing
calls.
Those
events.
E
D
No,
they
would
show
up
as
log
lines,
but
so
so
yes,
I
have
definitely
considered
that,
and
so
basically,
the
the
leaf
level
of
what's
on
the
display
here
should
probably
turn
into
those
log
lines
within
the
next
level.
Up
span
which
open
tracing
calls
events,
and
I
forget
what
open
telemetry
calls
them.
D
D
The
tags
are
things
you
can
search
on
so
so
any
any
of
these
things
are
are
searchable
within
the
within
the
tracing
system,
and
that
might
be,
you
know
quite
useful
to
to
try
and
narrow
down.
Let
me
just
show
you
one
other
thing
which
I
think
I
can
show
you.
So
I
the
the
program,
is
not
not
at
all
specific
to
cluster
api
so
for,
if
I
ask
it
to
show
me
what
what
went
on
in
cubelet
same
thing
happened
right.
D
So
there's
a
bunch
of
events
come
out
when
you,
when
you
apply
a
deployment
or
anything
like
that,
and
we
do
exactly
the
same
thing
turn
all
the
events
and
all
the
condition
transitions
into
spans.
In
this
case,
which
which
should
really
be
collapsed
up
into
into
log
lines,
but
we
we
can
pull
out
what
errors,
basically
things
things
where
the
the
event
has
been
raised
at
a
at
a
warning,
state
and
and
those
get
highlighted
on
on
the
ui.
D
That
I
mean
that's
the
way
that
jaeger
does
it,
but
that
that
again
is
very
much
how
I
I
see
this.
This
kind
of
tool
could
be
used
like
oh
something's,
gone
wrong,
because
it's
it's
very
much.
My
experience
that,
once
you
know
what
went
wrong,
you
can
go
find
the
event
that
would
have
told
you
about
it.
D
D
A
Sorry
so
I
I
just
wanted
to
interrupt
you
for
a
second
you
seen
ed
raised
his
hand.
F
So
one
question
so
when
so,
specifically
on
events,
do
you
see
anything
that
could
be
improved
at
the
api
level,
something
something
that
you
would
see
added,
because
the
current
api
for
events
is
the
one
beta
one
and
folks
were
we
were
looking
at
upgrading
it
and
moving
it
to
ga.
So
any
changes
that
you
think
are
are
welcome
or
necessary,
might
need
to
go
in
before
that.
D
So
so,
if
we,
if
you
and
I
talk
about
a
change
to
kubernetes
it,
it
takes
like
months
to
agree
what
to
do
and
takes
a
year
for
that
version
of
the
code
to
get
out
to
most
of
my
customers
and-
and
you
know
life's
too
short,
so
the
the
the
fact
that
we
can
produce
this
display.
D
You
know
with
with
kubernetes
1.16,
you
know
whatever
that's
the
oldest
one.
I've
tried
it
with.
That
is
very
important
to
what
I'm
trying
to
show
you
back
at
your
question.
So
the
event
code,
api
doesn't
take
a
context,
a
go
context,
so
so
it
cannot.
It
can't
kind
of
participate
in
in
the
regular.
D
So
let's
take
a
step
back
in
the
regular
distributed
tracing
in
in
go.
You
use
the
context
to
kind
of
flow
through
who
called
what,
and
so
I
definitely
anticipate
that
if
we
do
have
some
real
spans
coming
out
the
code
that
we
can
interleave
them
on
the
screen
here.
The
the
one
place
that
would
be
very
hard
is
is
events
because
the
the
ap,
the
code
api
doesn't
take
context
so
adding
a
context
to
the
code
api,
adding
a
tracing
context
encoding.
D
You
know
basically
a
hex
string
to
the
object
and
that's
the
subject
of
a
cap
about
how
to
pass
those
contexts
which
should
apply
the
same
to
to
any
kubernetes
object.
D
Yeah,
definitely
definitely
that
on
a
much
more
trivial
level,
events
are
generally
time
stamped
with
with
in
seconds,
which
again
makes
this
display
really
boring,
and
I
have
a
gross
hack
in
there
to
you
see
you
see
times
that
are
not
whole
seconds,
and
and
that's
because
I
have
no
self-respect
and
it's
you
can
find
it.
You
can
raise
a
bug
in
the
repo
and
tell
me
tell
me
I
did
a
bad
thing.
A
So
I
guess
if
there
was
nothing
further
from
your
scene,
I
had
a.
I
had
a
couple
questions
as
well
and,
and
you
kind
of
answered
one
of
them
brian,
you
said
this
is
this
is
kind
of
a
generic
solution
and-
and
please
forgive
my-
I
have
a
rudimentary
knowledge
of
jaeger,
but
I'm
certainly
not
an
expert
so
like
please
forgive
my
ignorance
here,
but
the
way
that
you're
processing
the
events
are
you
able
to
like
process
on
the
content
of
the
event
as
well
to
differentiate
how
they
appear
in
jager.
D
Right
so
so
99
of
it
is,
is
just
like
making
up
strings
like
we.
We
see
an
event,
you
know
we
we
we
see
an
event,
so
that
so
events
have
a
have
a
involved
object.
D
So
if
the
evol
involved
object
is
a
deployment,
then
that
goes
there
in
the
string
and
then
the
event
has
a
reason.
So
that
string
goes
there.
You
know
it,
or
that
was
a
condition
that
you
know
what
I
mean,
I'm
just
I'm
just
string
concatenating.
For
the
most
part,
it
does
not
look
at
what
kind
of
event
happened.
D
Yes,
obviously
we
we
could
apply
logic
to
make
things
appear
differently.
I
I've
done
a
bit
of
tinkering
in
in
the
jager
ui,
I'm
not
much,
I'm
not
much
of
a
of
a
javascript
programmer,
but.
D
D
But
yeah
I'm
not
I'm
not
really
a
front-end
developer.
It's
too
hard.
A
Okay
cool,
I
mean
this
is
interesting
to
me.
I
was
just
kind
of
curious
about
like
more
general
applications
and
how
we
could
use
it
in
other
places,
but
I
think
I
think
I
think
I
get
what
you're
saying
it
seems
like
this
is
pretty
applicable,
like
you
know,
across
all
of
kubernetes,
so
that's
really
cool.
D
Yeah,
so
I
have
a
talk
at
kubecon
in
a
couple
of
weeks
on
the
the
kind
of
general
front
and
and
in
that
one
I
don't
talk
about
cluster
api,
so
I
I
thought
I
would
bring
it
here.
D
Well,
a
as
a
follow-up
to
my
to
my
last
presentation
about
tracing
and
be
because
it
is
it
is
now
you
know
the
the
display
is
is
a
bit
off,
but
it's
not
you
know,
I
think
it
shows
promise
definitely
on
on
cluster
api
and
and
that
was
kind
of
where
I
started
this
journey.
D
D
So
there's,
I
think,
there's
a
whole
initiative
to
move
to
structured
logging
and
if,
if
we
can
put
the
context
there,
then
we
can
actually
pull
that
in
as
well
in
into
a
kind
of
unified
view
very
much
from
the
point
of
view
of
troubleshooting.
I
mean
I
don't
really
know
what
else.
Well,
I
have
ideas
like
you
know.
D
A
It
sounds
cool
we've
had
a
couple
more
hands
pop
up
as
well.
Okay,
ashish
is
next
and
then
david.
G
Thanks,
michael
so
I
saw
that
you
were
referring
to
conditions
and
then
you
were
also
referring
to
events,
and
I
also
remember
you
mentioning
you
have
a
program
running.
Is
your
program
translating
condition
changes
to
events.
D
Yeah,
so
actually
that's
another
point
about
about
spans,
like
the
point
about
spans
should
have
a
beginning
and
an
end,
so
in
in
theory,
conditions
have
a
kind
of
a
beginning
and
end.
You
know
they
go
false
and
then
they
go
true
or
except
they
don't.
You
know
the
the
you
you
can
effectively
only
see
the
true
transition.
D
I
guess
because
of
timing,
or
I
don't
know
exactly
why,
but
that
that
was
kind
of
what
an
idea
I
had
at
one
point
was
that
you
have
the
you:
have
the
false
transition
and
the
true
transition,
and
that's
your
span.
Well,
that's
actually,
yeah.
That's
exactly
what
I
tried
to
write.
It
just
turns
out.
It
doesn't
work.
E
All
right
david,
just
one
tidbit,
it
would
be
really
interesting
to
be
able
to
see
so
in
cluster
in
capsi.
We
are
doing
spans
all
the
way
through
the
controllers.
E
A
Yeah,
I
guess
thanks
brian,
that
was
that
was
awesome
and
I
think
I
found
it
to
be
pretty
inspiring.
So
I
I
appreciate
that
we
have
no
other
topics
here
so
and
no
one's
added
anything
during
the
demo.
So
I
guess
we'll
take
a
bunch
of
time
back
unless
anyone
has
anything
they'd
like
to
add.
H
Hey
michael,
I
think
I
added
something,
but
it
didn't
refresh.
H
Thanks,
I
just
wanted
to
maybe
quickly
go
over
the
open
proposals
and
checking
status
with
each
owner,
if
that's
okay,
at
least
the
ones
who
are
here.
So
if
you
can
like
scroll
up
to
the
top,
where
the
proposals
are
since
we're
nearing
well
we're
yeah
we're
one
week
away
from
the
april
21st
date
that
we
agreed
on
so
just
wanted
to
check
if
any
of
those
are
at
risk.
Right
now
is
anyone
here
who
can
give
a
quick
update
on
the
operator?
Oh,
that's
interesting!
H
A
So
I
think
alex
would
have
been
someone
to
talk
about
this,
but
he's
not
here.
Fabrizio
has
his
hand
up
go
ahead
for
breezy.
I
Yeah,
echoing
was
stefan,
was
telling
us
that
the
proposal
is
merged
and
the
work
started
with
the
agreement
to
implement
what
is
in
the
proposal.
Even
if
the
author
of
the
proposal
is
not
working,
which
is,
foreign,
is
not
working
anymore,
implemented,
alex
and
stefano
are
doing
prs,
but
my
humble
opinion
is
that
there
is
still
much
work
to
do
so.
I'm
not
sure
this
will
make
everyone
alpha.
4.
A
I
A
Branch,
you
seen,
you
got
your
hand
up
yeah,
so.
F
F
H
So
I
I
think
yeah
I
just
want
to
go
through
the
list
quickly,
first
and
see
like
just
focus
on
like
proposals
merging
just
because
that's
we're
a
week
away
and
that's
what
we
had
talked
about
and
then
I
think
maybe
we
should
talk
about
implementation
and
like
like
future
inclusion
in
patch
releases
separately
after,
like
as
a
as
the
next
discussion
topic,
is
that
okay.
A
Yeah,
I
think
that
sounds
good,
so
so
why
don't
we
we're
just
gonna
focus
on
going
through
these
kind
of
quickly
now
to
see
what
their
status
is
and
then,
if
folks
would
like
to
talk
about
any
of
these
specifically
in
terms
of
implementation,
maybe
add
them
to
the
doc
or
something
we'll
come
back
to
that,
so
yeah
I'd
say:
take
it
away.
So
still.
H
Okay
thanks
so
okay,
first
one
is
the
proposals
merge,
so
we're
good
kcp
skeleton
also
merged
load,
bouncer
provider.
I
think
that
one
jason
is
here,
but
he
said
he
might
have
to
step
away.
I
don't
know
if
he's
still
around.
J
Yeah,
so
on
this
one
there
was
a
lot
of
great
feedback.
There
was
still
a
couple
of
things
that
I
needed
to
incorporate
from
the
google
doc
and
I
plan
on
wrapping
that
up
between
this
afternoon
and
the
next
couple
of
days.
J
So
I
expect
to
be
ready
for
another
round
of
reviews
by
the
end
of
the
week.
For
this
one.
H
J
A
K
Yeah,
just
while
we're
on
the
topic
of
load
balancers.
Obviously
I
started
that
discussion
a
few
weeks
back
around
the
topic
of
multiple
load
balancers
for
the
control
plane,
I've
been
trying
to
sort
of
keep
a
step
back
and
and
sort
of
like
let
that
live
at
the
moment.
Just
so,
we
could
get
some
of
these,
like
main
low
balance
provider
stuff
find
out,
I'm
guessing.
K
K
J
J
You
know
kind
of
taking
it
like
even
further
if
there
is
a
general
consensus
that
that's
that
we
do
want
to
try
to
merge
those
I'm
happy
to
work
with
joel
to
to
try
to
get
those
in.
I
just
don't
know
if
that's
just
like
a
step
too
far
with
this
kind
of
proposal-
I
I
don't
know
if
vince
or
cecile
have
thoughts
there.
K
Yeah,
I
I
don't
have
any
strict
like
timelines
on
this.
It
totally
could
be
a
follow-up.
It
was
just
yeah
not
entirely
sure
what
the
timeline
cycles
are
so
like.
If
that
means
it's
not
getting
done
for
another
six
months,
that
would
probably
be
a
bit
of
a
pain
so
yeah
it
was
just
if
yeah
I
totally
get
what
jason's
saying
I
agree,
don't
want
to
muddy
that
proposal.
K
It's
big
enough
already,
let's
get
it
through,
as
is
it
would
just
be
a
case
of
like
I
wonder
if
we
could
add
to
this
after
the
20th,
I'm
guessing,
probably
not
so
yeah.
When
would
the
next
cycle
be
sort
of
looking
at.
B
H
I
was
just
gonna
say
I
think
it's
a
vicious
circle,
the
more
we
add
stuff
at
the
end
of
cycles,
the
longer
the
cycles
are
going
to
be,
and
probably
adding
this
in
at
this
point
is
probably
going
to
delay
everything
else.
That's
been
waiting
for
this
release,
so
I
don't.
I
think
we
just
really
need
to
like
be
a
bit
more
strict
about
that
and
start
reducing
how
much
we
put
in
like
releases
so
that
it's
not
so
hard
to
get
the
release
out
the
door.
H
I
hear
you
that's
a
pain
that
right
now
you
think
you
know
if
we
don't
do
it
now,
it's
gonna
be
in
six
months,
but
there's
also
things
that
I
have
been
waiting
since
six
months
ago
that
are
still
not
out
so
yeah.
That's
my
view
on
it.
K
Yeah,
okay,
I'll
I'll,
try
and
focus
on
helping
jason
on
his
proposal,
and
then
we
can
follow
up
with
it
in
a
few
months
time,
then,
when
we're
ready
to
thanks.
H
K
Yeah
I've
sort
of
been
working
on
this
with
alberto.
As
far
as
we're
aware,
it's
pretty
good
to
go,
I
don't
think
there
was
any
outstanding
feedback.
It's
got
a
bunch
of
the
good
to
me
labels
already.
K
I
think
there
were
just
a
couple
of
people
who
have
been
reviewing,
who
haven't
had
a
chance
to
catch
up
on
it.
So
I
think
we
are
just
waiting
on.
I
think
vince
and
you
seen
had
left
some
feedback
before
so
not
sure
whether
they've
got
time
to
go
through
it
again
or
whether
we
can
get
it
merged
without
but
yeah.
It
seems
like
it's
very
close
to
me.
B
I
think
they
didn't
get
an
answer
to
this.
Like
are
we?
How,
when
are
we
gonna,
do
like
everything
externally
managed.
K
I
did
do
a
a
comment
and
an
evaluation
sort
of
that
and
I
left
that
on
the
thread.
Although
that
was
a
few
weeks
back,
and
I
can't
remember
what
the
conclusion
was,
I
it
was
vaguely
something
like.
I
don't
think
we
need
to
worry
about
this
right
now.
I
didn't
think
there
was
too
much
conflict,
but
I
can
double
check
that
later
this
week
and
get
an
answer.
A
F
Yeah,
so
I
I
did
the
first
pass
on
it.
Overall,
it
looks
good
to
me.
I'm
gonna
do
another
pass,
probably
tomorrow,
but
I
feel
like
some
of
these
proposals
that
that
don't
have
that
many
outstanding
comments
might
be
eligible
for
lazy,
lazy
consensus
by
the
end
of
the
deadline.
A
Okay,
I
guess
we
can
joel
or
I
can
poke
alberto
too
and
tell
him
to
come
back
around
here
and
add
something
about
the
load
balancers
as
well.
H
F
So
for
node
attestation,
the
bulk
of
the
comments
were
addressed
when
dna.
I
don't
think
that
there's
much
left
there's
a
few
minor
comments,
maybe
that
week
that
we
need
to
fix.
But
overall
I
I
started
poc
and
then
prototyping
the
controller.
So
I
I
still
get.
I
still
have
to
push
the
branch
by
the
end
of
the
week,
but
overall
it's
it's
looking
good.
Just
a
few
things
to
fix.
F
Yeah,
so
far
so
far,
it
looks
good.
The
only
piece
that
is
missing
is
that
we
that
we
need
to
add
before
merging
is
specifically
regarding
backward
compatibility
and
not
backward
compatibility,
but
compatibility
between
the
attestation
data
format
that
is
written
by
the
cli
and
the
one
that
is,
and
the
controller
that
reads
this
attestation
data.
F
The
reason.
The
reason
of
why
we
need
to
specify
that
is,
there
might
be
cases
where
you
have
a
management
cluster
running
an
old
version
of
cluster
api
that
has
the
attestation
enabled,
but
still
you
you
could
have
workload
clusters
that
have
machine
images
that
have
the
latest
cli
that
has
some
breaking
change
or
edition.
So
there's
a
lot.
We
have
a
logic,
basically
where
we
negotiate
the
api
versions
and
the
con
the
format
of
the
adjustation
that
we
just
need
to
specify
here.
H
Cool
thanks:
should
we
move
on
to
the
next
one,
all
right,
so
yeah
cluster
class
cluster
topologies?
I
guess
events,
that's
a
pretty
new
one.
Are
we
targeting
a
proposal
by
next
week
for
that
one.
B
This
is
not.
This
is
not
going
to
be
breaking
and
sagar
is
actually
working
on
the
proposal
on
md
we're
gonna,
probably
meet
1-1
tomorrow
to
just
discuss
like
a
few
goals
and
then
publish
the
the
hackmd
like
early
next
week.
B
That
seems
like
there's
a
lot
of
interest
so
like
we're
going
to
get
the
ball
rolling
like
relatively
soon
but
given
this
is
not
it's
just
an
addition
like
we
don't
have
to
like
rush
it
in.
H
Whatnot
all
right
spot
instances
is
what's
the
status
on
this.
Is
it
still
happening
for
zero?
Four,
I
thought
that
someone
had
updated
the
status
in
the
email
thread.
We
should
probably
update
it
in
this
stock.
K
Yeah
I
was
just
gonna
because
I
spoke
to
alex
about
this
semi
recently
I
I
thought
we
were
gonna
punt
on
this
for
zero
four,
but
I
don't
know
about
that.
Update
that's
still
mentioned,
so
maybe
my
information
is
out
of
date.
H
I
will
take
an
action
item
to
follow
up
and
check
and
then,
if
we
did
decide
to
punt
I'll,
just
cross
it
out
from
that
list,
all
right
sounds
good.
A
Yeah,
that's
mine,
so
there
were
a
couple
questions
back
I
mean
I.
I
got
some
response
on
this,
the
only
you
know
I've
been
working
through
a
poc
for
the
actual
implementation
of
this.
You
know,
maybe
maybe
a
few
more
reviews
would
help
out.
There's
one
outstanding
question
I
have
and
it's
kind
of
about
the
implementation
as
I'm
building
the
poc.
A
You
know
I've
run
into
a
few
things
that
was
just
making
me
think.
Maybe
the
way
the
approach
I
took
here
is
not
the
best
approach.
So
if
people
are
interested
in
this,
I
would
I
would
appreciate
a
few
more
reviews
on
it,
but
I
think
we're
pretty
close
to
like
there's
one
decision.
I
think
we
need
to
make
and
if
we,
if
we
can
make
that
decision,
we
could
probably
merge
this
proposal
next.
H
Week
sounds
good
yeah,
so
in
general,
for
everyone
who's
like
driving
a
proposal.
Remember
that
in
order
to
get
it
merged
by
next
week.
That
means
the
reviewers
need
a
bit
of
time
before
the
deadline
to
review
it.
H
So
if
you
can
try
to
have
like
any
like
major
updates
that
need
to
be
made
out
by
like
end
of
the
week,
that
would
be
great
that
we
give
it
ourselves
like
a
bit
of
time
to
actually
go
and
review
the
updates,
and
then
I
think,
for
anything,
that's
in
a
good
place
where
there
hasn't
really
been
any
movement.
Setting
up
a
lazy
consensus,
as
the
essence
suggested,
might
be
a
good
idea.
A
A
A
So
cecile
was
there
anything
else
you
wanted
to
get
into
on
this
topic.
H
F
Regarding
timelines
for
proposals
right
for
auditions,
if
I
recall
correctly
yeah
so
like
my
question
was:
do
we
have,
since
we
were
saying
that
basically
we'd
move
away
from
shipping
features
in
patch
versions?
Do
we
have
like
a
policy?
That's
or
should
we
define
a
policy?
Perhaps
that
says:
okay,
we
ship
features
for
the
first
iteration
up
to
n
patch
version,
and
then
it's
only.
F
It's
only
like
bug
fixes,
because
I
see
here,
a
lot
of
people
will
be
fair
for
folks
that
need
those
to.
A
H405,
so
your
mic
broke
up
a
little
bit
during
that
you've
seen,
but
I
think
the
the
general
idea
is:
do
we
have
something
written
down
about
kind
of
planning
for
when
features
are
released
and
kind
of
the
timelines
around?
That
was
that
was
that
kind
of
the
heart
of
the
question.
F
Yeah,
the
basically
the
tldr
is,
should
we
allow
features
for
patch
versions
and,
if
so,
up,
until
which
number
of
patch
versions.
A
I
I
think
that
the
key
here
is
to
define
the
the
roadmap
for
the
major
version
so
that,
let's
give
a
perspective
on
what
is
the,
let
me
say,
the
the
leg
between
current
version,
the
next
one
and
and
gives
us
arguments
to
discuss
if
we
should
put
something
into
the
future
version
or
not,
because
you
know
when
the
next
training
will
start,
we
move
on
so,
for
instance,
now
we
we
know
that
to
be
one
alpha.
I
I
F
Sense,
I
I
guess
it
does
like
the
basically
it
boils
down
to.
We
need
to
agree
on
something
and
document
it
to
at
least
set
expectations
for
developers
and
consumers,
but
I
agree
that
worthwhile,
like
once
we
once
we
have
a
major
or
a
beta
release,
we
should
be,
I
guess,
more
prosperous,
more
prospective
on
on
features
landing
on
patch.
A
Releases
fabrizio
did,
did
you
still
have
somebody
out
there.
B
No
worries
go
ahead:
vince
yeah,
I
was
gonna
say
like
there's
that
discussion
that
we
open
it
might
be
good
to
start
a
thread
and
then
start
replying
with
preferences
underneath
it,
and
maybe
we
can
follow
up
with
a
like
update
to
the
contributing
guide
and
our
policy
to
do
this
later.
B
Though
vince
no,
no,
I
think
github
under
discussions,
there's
the
beta1
v1
1.0.
The
discussions
are
kind
of
nice
because
you
could
start
a
thread
under
that
discussion
and
then
you
can
all
reply
under
that
thread.
A
A
Me,
okay,
I
guess,
is
there
any
other
topics
or
conversations
people
would
like
to
have
before
we
wrap
up.
A
Okay,
I'm
going
to
assume
not
then
thank
you,
everyone
for
attending
and
I
guess
we'll
see
you
all
next.