►
From YouTube: 2017-06-06 17.01.44 SIG-cluster-lifecycle 166836624
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
The
first
two
were
just
from
a
topic
that
I
discussed
how
to
catch
up
at
Lucas
earlier
Lucas
is
up
and
running
with
us
again
and
working
on
various
things.
I
think
he's
going
to
join
the
meeting
in
a
minute
and
the
first
two
were
just
to
flag
that
there
are
a
couple
of
issues
that
are
currently
stopping
qk
m
@
CI
going
green
and
they
linked
to
in
the
notes.
So,
if
necessary,
I
think
the
first
one
has
already
been
approved,
but
the
great
gifts
and
input
on
on
the
second
one.
C
Thankfully-
and
at
this
point
it's
just
a
skeleton
overview
of
your
the
process
for
upgrading
mix
mac-
consumer-friendly
it's
not
very
user
friendly
at
this
point.
Just
like
do
this
justice,
we
assume
you
know
what
you're
doing
to
that.
Other
developers
could
help
us
test.
The
upgrade
process.
Lucas
and
I
have
got
in
a
few
PRS
to
make
cubeb
and
in
its
idempotence.
C
You
can
actually
use
it
for
free
and
it
sounds
like
one
of
Luke's
as
PRS
just
got
merged
in,
but
I'm
going
to
try
to
create
candidates,
Deb's
and
rpms,
and
link
them
from
the
guide,
so
people
just
download
them
before
the
next.
Your
candidate
releases
cut
and
start
testing
the
instructions
earlier
yeah.
B
I
have
one
Tim
will
be
off
to
be
most
I.
Think
you
have
one
Baku,
then
we
also
need
a
control
plane
from
late
yesterday,
something
for
it
to
work.
My
cuts
and
changes
on
the
control
plane.
Side.
Api
server
are
back
rules
that
are
required
for
for
cubed
M,
the
CSR
approval
so
to
work,
that's
not
it.
Okay,
the
next
beta
should
just
work
and
when
my
node
authorizer
PR
gets
in
cubed,
the
Kaveri
MCI
should
should
thank.
We.
A
Good
having
green,
CI
silently
healthier
project,
that's
great
and
cool.
So
thank
you.
Joe
people
Lucas
for
that,
but
if
they
Jacob
are
you
looking
for
feedback?
Are
you
looking
for
testers
on
those
docks
changes
at
the
moment
versus
actually
trying
to
get
docks
change?
Merch,
or
do
you
think
we
should
get
the
docks,
change,
merch
and
then
improve
it
right.
C
Yeah
so
at
this
point
it
was
not
written
very
it's
not
hot
at
all,
so
it
was
mainly
to
get
other
testers
touching
it.
But
it's
a
little
hard
for
people
tested
right
now,
because
we
don't
have
Deb's
and
rpms
cuts.
Has
there
have
been
changes
since
last
beta,
so
it
might
be
worthwhile
to
just
wait
until
we
cut
the
next
vid
I'll
test.
I'll
check
the
schedule
to
see
how
far
out
that
scheduled
me
and
if
anyone
wants
to
build
them
locally,
you
can
do
that.
C
D
E
And
Simon
I
look
like
I,
don't
know
on
the
first
one
Lucas
was
asking
for
a
feedback
on
the
PRS
ii
PR
donned.
The
release
manager
for
1:7
said
that
we
either
need
to
go
through
the
exception
process.
If
we
think
that
is
a
feature
or
we
need
to
qualify
it
as
not
a
feature
for
the
one
7
release
branch,
so
we
should
decide
what
that
is
the
exception
process.
Nominations
are
due
book
by
tomorrow,
so
if
we
do
think
it
needs
to
go
through
the
exception
processing
to
do
that,
ASAP
yeah.
B
I
was
going
to
comment
on
that,
oh
well,
it
was
the
original
PR
is
somewhere
in
the
queue
as
well
and
Mike
might
get
that
one.
However,
it
was
he.
He
just
said
the
the
no
does
authorized
arguments
unconditionally
like
so
it
broke
one
six
clusters
or
it
would
have
broken
one
successor
so
I
I,
often
to
just
move
that
comment
from
from
the
PR
that
was
LG.
Tm
did
time
for
code
freeze.
Let
me
see
if
I
can
find
the
right
number
yeah.
F
So
exactly
where
are
so?
The
argument
here
would
be
that
we
actually
prevented
something
to
go
in
because
it
would
have
been
buggy
so
that
we
could
actually
pre
fix
it
before
it
went
in.
If
we
had
been,
you
know,
if
we
just
slam
that
in
then
we
could
have
said
well
now
we're
doing
a
bug
fix
damn
trying
to
be
more
careful,
and
so
therefore
it's
slowed
down
actually
getting
nothing
in.
Is
that
kind
of
argument
here,
yeah
exactly
and
so.
B
It
was
that
the
code
was
was
LG
TM
as
if
before
the
code
freeze,
but
then
I
found
a
bug
that
I
said:
I
could
fix
before
we
merged
it
within
the
cube,
ATM
part
of
it,
and
then
I
sent
a
new
another
PR
that
used
to
be
much
now
and
so
so
yeah.
That's
that
the
argument,
and
also
it
is
kind
of
a
bug
fixable.
Without
this,
the
the
cube
cubed
M
C,
is
red,
and
with
this
it
will
probably
go
clean.
So.
F
B
F
May
be
worthwhile
for
somebody
to
sort
of
just
ask
dawn.
You
know
which
way
she
prefers
to
go
and
sort
of
explain
the
situation
to
her
and
see
if
she
would
rather
have
it
be
an
exception
process
or
or
just
treated,
as
you
know,
a
feature
that
was
LG
TM
before,
but
in
a
different
PR
yeah.
F
B
F
A
Good
cool,
so
the
next
generation
is
from
Simon
yeah
yeah.
G
G
So
one
question,
or
maybe
more
than
one
question.
The
first
question
we
have
is
like
is
the
custom
name
suitable
to
be
populated
with
something
that
we
can
identify.
The
pastor
was
I
was
referred
to
this
state.
They
say
these
parties
that
writes
it
for
discussion
like
this
yeah
awesome,
so
I,
don't
know
the
custom
name,
I'm,
not
sure
if
you
guys
know
which
one
I'm
talking
about
I
can
find
the
link.
This
is
just.
G
Okay,
so
some
the
comment
does
on
the
in
the
code.
It
says
the
field
cannot
be,
you
will
be
in
or
it
was
set
in
Korean
update,
so
it
sort
of
make
it
invisible
it
yeah
make
it
immutable.
I
could
be
wrong
and
it
looks
like
it's
not
being
used
right
now
and
I
kinda
track
back
to
the
TR
that
that
put
this
fill
in
and
our
original
paralysis
is
either
to
to
identify
federated
cluster
or
just
individual
cluster,
so
just
kind
of
bored,
the
usage
and
I
honestly,
not
too
sure
like.
F
Think
there's
two
issues
here.
So
the
first
issue
is:
how
do
you
actually
set
the
cluster
ID
itself
and
where
is
that
discoverable
across
the
cluster?
And
the
second
issue
is
that
do
we
change
the
API
machinery
such
that
that
cluster
ID
gets
stamped
into
every
resource,
similar
to
say
the
UID
of
the
resource
itself
or
the
self
link,
or
you
know
even
namespace
and
resource
version,
because
these
are
all
part
of
that
object,
meta
that
users
aren't
allowed
to
change
into
the
system.
F
Changes
right,
so
I
think
if
we
come
up
with
a
unique
ID
for
a
cluster,
stamping
that
into
the
into
the
so
the
cluster
name
is
probably
okay,
I
think
it's
it's
a
little
bit
wacky,
because
there's
a
lot
of
custom
code
that
folks
will
write
to
actually
take
resources
out
of
the
out
of
the
the
API
server
into
a
strip
club
for
specific
stuff
from
them,
and
this
is
going
to
be
one
of
those
that
they're
going
to
have
to
change,
to
be
able
to
strip
and
so
like.
F
F
So
if
I
do
keep,
control
get
and
I
do
an
export,
then
that
it
should
actually
ideally
go
ahead
and
like
strip
this
cluster
name
from
resources
as
I
get
them,
and
so
that's
the
type
of
thing
that
we'd
want
to
make
sure
it
gets
fixed
up
as
part
of
this
and
the
API
machinery
folks
would
probably
have
a
much
better
idea
of
wearing
one
and
how
some
of
this
stuff
happens.
That
being
said,
how
do
we
actually
go
through
and
set?
F
F
F
You
know
system
in
book,
config,
nap,
I
think
the
idea
was
that
if
we
get
folks
to
agree
eventually
that
becomes
more
standard
in
the
cluster
itself,
updates
that
with
information
about
the
cluster
as
things
change,
so
so
I
think
in
terms
of
setting
that
you
know
a
new
component
config
option
for
the
for
the
API
server
is
not
crazy
in
terms
of
publishing
that
to
anybody
who
wants
to
walk
up
doing
that
in
the
Q
public
cluster
info
is
not
crazy
and
then,
in
terms
of
setting
this
on
every
object.
F
E
Those
in
terms
of
setting
it
on
the
config
map,
the
one
of
the
arguments
was
it
shouldn't,
be
mutable
and
if
it's
Nick
and
it
can
only
config
for
your
guide
server,
this
controller
manager.
That
makes
that
implies
that
it's
mutable
right,
because
it's
easy
for
someone
to
change
that
config
map
and
then
update
that
I.
Think
I
was
at
the
Clement
series.
I
Yeah,
okay,
one
one
narrow
solution
is
to
look
at
like
the
UID
of
the
coop
system,
namespace,
which
is
in
practice
immutable,
because
once
you
delete
that
you're
basically
asked
but
I'm,
not
crazy,
yeah.
The
I
think
that
was
really
landed
on
our
plate,
because
we
like
have
a
bunch
of
other
things
like
I
saw.
Tim
was
asking
about
ciders
and
I.
Think
so.
I
think,
like
the
the
notion
of
the
config
map,
holding
things
is
a
good
one.
It
might
even
mention
out
there.
I
Config
map
right,
I,
think
something
so
the
UID
will
be
immutable
on
something
and
we
could
do
cube
system
or
we
can
say
it's
a
config
map
which
might
one
day
be
an
API
object.
If
we,
if
we
don't,
have
it
be
the
namespace,
then
if
we
switch
it
from
a
country,
mapping
API
object
table,
it
will
change.
F
In
terms
of
like
at
the
config
map
and
like
I
mean
yeah,
there's
all
sorts
of
things
you
can
do
by
walking
up
to
a
UNIX
system
and
like
deleting
like
the
kernel
image,
and
then
the
machine
no
longer
boots
right.
Just
because
you
can
do
something
to
screw
up
a
machine
doesn't
mean
I
mean
we
can
set,
the
appropriate
are
back
roll
so
that
we
come
to
that
you're
going
to
end
up
with
a
horse
machine
I
mean,
like
you
know.
F
F
G
Like
one
one
advantage
of
using
the
config
map
is
some
they
are
going
to
cluster
ID.
That
will
be
other
cluster
information
or
other
Conficker,
because
the
cluster
configuration
information
that
people
might
want
to
stick
in
there
doing
it
will
cause
the
name
kind
of
pop
it
in
that.
So
that's
an
argument
so.
F
I
would
say:
I
like
Justin's
idea
of
taking
the
you
idea
of
the
cube
system.
The
only
problem
with
that
is
you
don't
know
what
that's
going
to
be
until
after
the
cluster
is
up
and
running,
so
there's
going
to
be
a
time
period
where
the
cluster
is
kind
of
running.
But
but
you
don't
like
you
know,
you
don't
know
from
the
outside
when
you're
starting
the
cluster.
What
that
UID
is
and
so
I
think
having
like
you
know,
having
a
way
such
that
you
can.
F
Actually
you
know
have
that
UID
said
very,
very,
very
early
I
think
might
be
something
that
that
would
be
a
good
good
benefit
here,
so
yeah
so
like
I
said
setting
it
publishing
it
and
then
and
then
setting
it
in
the
object
meta.
Those
are
all
I,
think
separable
issues.
The
object
meta
thing
seems
like
it's,
you
know
very
separable
by
giving
to
do
the
cluster
name
thing
and
every
object.
F
F
I
F
True
I
mean
one
of
the
things
that
I
think
you'll
use
it
for
is
to
actually
you
know,
identify
a
cluster
in
some
sort
of
future
version
of
cube,
config
right
just
to
make
sure
that
you're
talking
to
the
cluster
that
you
think
you're
talking
to.
F
So
that's
that's
going
to
be
one
thing.
I
can
imagine
that
that,
as
we're
writing
out
structured
logs,
we
might
want
to
append
it
to
structured
logs
for
pretty
much
everything
that's
going
on.
So
we
can
actually
mix
structured
logs
from
multiple
clusters
into
one
analysis
system
and
be
able
to
sort
by
cluster.
Wait.
Yeah.
I
I
I
The
question
is
in
this
case,
like
if
I
backup
and
restore
a
cluster
on
a
different
IP
different
from
our
private
key
everything
different
but
I
put
the
same
things
on
it
and
I
feel
like
it's
the
same
like.
Should
that
be
the
same
cluster
or
not
from
your
point
of
view
in
in
so
I
forget
to
in
the
with
the
perfect
answer
service
hemlock,
one.
I
G
That
brings
and
the
other
question.
So
our
requirement
is
right.
We
would
like
the
operator
not
to
have
to
manually
set
the
UID
it's
an
option
for
them
to
do
it,
but
like
in
the
case
that
one
is
not
in
set,
then
the
system
you
can
probably
generate
one
and
then
you
want
that
to
be
unique.
So,
like
the
first,
let's
say
he
gets
a
comeback
with
channel
1
and
we
need
that
to
be
stored
somewhere.
G
F
It
really
okay,
I
didn't
go
I'm
course.
A
good
topic
will
say:
Windows
has
this
idea
of
like
a
machine
ID
that's
saved
in
at
a
pretty
fundamental
level
and
to
reset
that
you
essentially
have
to
sysprep
the
machine
which
makes
creating
Windows
images
on
cloud
the
total
pain
in
the
ass,
oh
yeah,
so
the
shade
is
trial.
I.
Think
about
sort
of
like
analogous
things
that
that
that
relate
to
this
yeah.
A
F
Okay,
so
I
here's
here's
my
idea
in
terms
of
what
what
should
happen
here.
This
get
sets
into
a
config
map.
We
have
some
controller
in
the
API
server
that
looks
for
this
if
it
sees
that
it
changes,
it
just
starts
crashing
with
too
loud
message,
because
you're
not
supposed
to
change
the
thing
right
and-
and
so
in
terms
of
where
this
gets
stored,
that's
the
canonical
place
that
it
gets
stored
and
in
what?
If
an
API
server
boots
up
and
there's
not
one
of
those
config
maps,
it'll
go
ahead
and
create
one.
F
E
I
will
say
that
a
lot
of
the
things
we're
discussing
have
already
sort
of
been
proposed
and
hashed
out
on
the
pursue,
so
we
might
want
to
bury
the
issue.
Have
people
read
that
issue
and
put
their
comments
there?
So
we
can
also
loop
in
the
other
people
that
are
already
discussing
these
things,
sort
of
try
to
act
out
like
now
in
terms
of
the.
I
F
F
E
F
Okay,
shall
we
just
maybe
put
a
pin
in
this
and
have
folks
who
are
interested
in
commenting
to
bits
on
the
issue?
Is
that
kind
of
where
we're
at
I
guess.
E
E
G
I
I
make
a
suggestion.
There
is
a.
There
is
a
short-term
thing,
which
is,
we
can
say
today
use
the
UI
Dion
cube
system
that
will
give
you
your
unique
ID
and
then,
when
we
build
something
that
so
there
properly,
we
will
seed
it
from
the
UID
on
cube
system.
If
there
isn't
I'm
not
already
yeah,
that's
what
I
was
going
to
recommend
yeah.
F
Worry
about
ordering
there
I
got
to
be
honest.
It
kind
of
scares
me
a
little
bit
because
there's
going
to
be
this,
there's
going
to
be
like
this
time
when
there's
no
config
map,
but
we're
going
to
want
to
create
objects,
and
if
we
do
something
like
the
cluster
name
as
part
of
the
objects
made
of
meta,
then
how
do
we
create
cube
system?
If
we
actually
don't
have
that,
and
we
need
to
actually
set
that
into
cube
system
I,
don't
think
we
should
shove
it
on
every
single
object.
I
think
we
should
do
that.
F
The
idea
is
that
if
you
go
through
and
if
you
get
something
from
a
cluster,
you
want
part
of
that
part
of
like
that.
Meta
of
that
object
is
which
cluster
is
that
thing
in
right,
and
so
that's
to
the
cluster
name.
The
cluster
name
thing
that's
in
object,
meta,
I
believe
the
intent
there
is
at
least,
if
I'm
reading
the
comment
right.
E
F
A
A
B
The
that
one
and
basically
you
bet
am
in
one
seven-
will
always
use
the
no
doubt
Rosa
and
the
node
admission
plugin
for
for
security
and
when
upgrading
this
will
mean
that
the
system,
currently
there
there's
a
cluster
role
binding
that
bind
this
M
nodes
group
to
the
system,
node
plus
the
row,
which
means
that
every
third
in
the
system,
node
nodes
organization,
will
have
read
access
to
have
a
secret
and
so
forth,
and
the
node
authorized.
And
now
the
mission
plug-in
doesn't
have
any
effect.
B
B
D
So,
just
as
a
point
of
order,
because
I
saw
you
put
in
those
PRS
and
I
understand
the
reasoning
and
logic
behind
it,
but
I
was
blind
to
it
throughout
the
release
cycle
for
1:7,
because
we
didn't
have
like
labeling
specified
for
issues
we
wanted
to
address
and
Covidien
for
1:7,
so
I
knew
that
this
existed.
I
did
not
know
because
there's
so
many
things
going
on
so
many
issues
to
track
and
follow
that
we
were
going
to
try
and
tackle
this
as
part
of
this
release
and
the
default
would
be
updated.
D
Was
there
any
roots?
Just
so
I
have
like
history
here
it
was
there
any
roots
issue
for
Kubb
ADM.
That
pointed
back
to
one
seven
milestone,
or
something
like
that,
so
we're
just
for
my
own
edification.
So
as
for
the
next
release
cycle,
as
we
start
to
do
these
things
and
start
to
get
more
structured
with
how
we
target
items
for
release,
there's
there's
traceability
for
the
reset,
because
we
do
this
as
part
of
sig
scheduling.
We
always
have
like
these
things
are
part
of
the
release.
There's
no
ambiguity
right.
B
Be
honest:
nothing
has
been
tracked
in
the
repo
to
really
kind
of
for
for
1000
I.
Did
this
triage
or
I
labeled
the
right
issues
with
one
six
that
we're
going
in
and
like
set
the
milestone
everything
so
so
we
had
a
clear,
clear,
historic
view
of
how
things
things
turned
out
as
far
as
far
as
I
know
that
wasn't
done
pretty
much
at
all
in
in
the
end
here,
I
saw
when
you
got
right
access.
You
start
doing
this
and
that's
great,
but
before
that
nothing
really
I
think.
D
D
I
think
there's
a
bleeds
into
1/8
talking
about
valid
stuff,
I
want
to
start
making
sure
that
we
we
tackle
and
label
things
for
1/8
just
so
we
have
a
known
list
of
features
that
we
want
to
work
on
and
we
can
actually
Feder
eight
and
lay
out
our
delegate,
probably
delegate
folks
who
are
on
the
call
who
want
to
contribute
some
of
these
new
items
to
so
that
way,
it's
a
little
more
structured.
We
have
executors
I.
F
D
F
That's
you
know,
and
the
note
authorizer
stuff
was
mostly
driven
by
Jordan
of
the
sig
off
and
I.
Think
it's
just
it's
more
of
an
opportunistic
thing
to
turn
it
on
for
kube
admin.
Since
cube
admin
has,
you
know,
been
sort
of
bleeding
edge,
was
our
backup
until
now,
and
so
yeah
so
I
think
it's
definitely
agree
with
you
that
moving
forward
into
one
8's
to
get
you
know
more
focus
on
I'm,
moving
cave
admin
forward
and
and
other
stuff
in
this
league.
For
we,
you
should
definitely
dragon
that.
B
A
E
Short
I
didn't
want
to
mention
before
we
start
going
into
the
dark,
because
I
noticed
that
the
dark
has
lots
of
items
in
it
and
lots
of
people's
names
next
to
them
that
one
of
the
things
I
felt
that
didn't
go
very
well
with
one
seven.
His
we
signed
up
for
sort
of
too
many
things
and
didn't
put
enough
people
all
the
things
we
signed
up
for,
and
we
need
to
keep
in
mind
that
signing
up
for
making
something
doesn't
just
mean
writing
a
little
bit
of
code.
E
E
Have
every
person
pick
up
one
thing
and
not
quite
get
it
done
so
I
would
encourage
us
to
try
to
be
very
circumspect
in
terms
of
what
we
think
you're
going
to
shoot
for
and
to
sort
of
under
promise
and
over
deliver
and
like
make
sure
we
actually
nail
everything
is
set
up
for
other
than
focus.
I
have
a
whole
bunch
of
stuff
having
you
know,
other
SIG's
or
other
people
see
our
document
and
think
wow.
B
The
the
things
I
put
behind
that
and
then
next
to
I'm,
I'm
gonna
be
really
focused.
I
live
in
those
at
least
the.
D
One
thing
I
don't
see
in
a
doc
as
I
look
through
it
just
I,
just
scanned
it
real
quick
is
the.
There
are
a
fair
number
of
issues
coming
in
to
have
somebody
who
can
triage
the
issues.
You
know
we
talked
about
this
kind
of
at
the
developer.
The
Leadership
Summit
was
too
somebody
who
could
potentially
triage
the
inbound
requests
and
potentially
address
them.
D
So
that
way
other
folks
can
focus
on
executing
to
because
there
is
a
fair
number
of
bugs
that
actually
exist,
some
of
which
could
be
considered
as
bugs
whether
they're
just
minor
configuration,
knobs
or
whatnot
I
think
that's
to
be
determined,
but
there's
nowhere
on
the
list
is
like
getting
the
backlog
of
rubidium.
So
it's
like
level
or
down
from
this
next
review
cycle
is
right.
Now
it's
growing
this
way.
That's.
F
We
also
need
to
be
careful
to
make
sure
that
we're
looking
beyond
just
cube
admin
for
this
stuff,
I
think
a
lot
of
the
scenarios
are
driven
by
cube
admin,
but
these
parts
have
been
dependent
utility,
and
so
you
know,
there's
going
to
be
other
other
issues
that
people
see
through
use
cases
that
are
not
coming
through
to
Badman
that
well,
hopefully,
they'll
see
it
as
they
start
using
these
different
facilities
in
different
ways.
So
well.
B
H
E
That
that
intro,
let's
walk
to
the
dock,
only
about
20
minutes
left.
The
first
thing
is
getting
bootstrap
tokens
up
to
beta
I,
don't
know
if
people
had
sort
of
self
sign
up
for
these
things
or
if
no
something
went
through
and
assigned
other
people,
but
I
would
also
like
to
make
sure
that
people
whose
names
nurses
are
actually
committing
to
getting
them
done
with
all
of
the
requisite
items
that
I
had
mentioned
this
beginning.
So,
yes,.
B
E
B
E
F
Part
of
the
cube
admin
experience
for
the
general
Golz
actually
give
admin
to
GA,
because
there
was
a
lot
of
folks
who
just
won't
use
it
unless
extremely
ahead
and
so
making
sure
that
we're
willing
to
export
this
part
of
this
protocol
process.
That
was
much
better.
Thank
you.
Now
you
deliver
now
how.
I
I'm
also
going
to
continue
working
on
this
I
have
been
distracted
by
other
things
like
tests
right,
but
this
is
probably
gonna
have
to
involve
a
new
API
type
in
core
and
it's
always
going
to
be
a
big
design
up
front
type
effort,
even
if
it
is
going
to
be
alpha
and
it's
going
to
be
super
minimal,
but
I
believe
it
is
important,
because
Adams
continued
to
fail
for
some
of
it
is
so
easy.
Adams
continues
to
fail
is
very
surprising
ways,
for
example
in
it,
turns
out
NH
a
clusters.
I
F
Iii
want
to
put
a
point
in
this
also
I
think
you
know
as
we
look
towards
taking
the
API
server,
the
controller
manager,
the
scheduler
and
breaking
goes
onto
smaller
pieces.
I,
don't
know
if
it's
going
to
end
up
like
Brian's
layer-cake
robots,
but
we
do
it's
very
clear
that
we're
going
to
we're
going
to
see
the
control
plane
actually
include
more
components.
F
You
know
beyond
what
we
have
now
and
right
now,
adding
a
new
binary
to
that
control
plane
with
a
total
crapshoot,
because
there's
just
so
much
variability
in
terms
of
how
things
get
launching
in
right.
So
if
we
can
model
those
that
on
and
consider
creating
way
to
do
that,
then
it
enables
us
to
actually
have
more
systems
processes
in
some
ways
that
we
can
actually
launch
later
and.
F
E
E
I
had
a
couple
other
notes:
there,
too,
we
either
consistent,
add
on
experience
as
part
of
the
core
kubernetes
right.
That's
part
of
the
layered
architecture,
doc.
We
want
that
to
be
the
same
everywhere
and
that
also,
like
you
said,
Joe
makes
it
simpler
to
do
things
like
add
the
the
cloud
controller
manager.
You
know
if
we
took
the
cloud
controller
manager
as
sort
of
an
add-on
with
component
config,
then
that's
really
easy
to
bootstrap
in
a
specific
way.
Yeah.
F
D
I
think
isn't
this
predicated
on
component
of
fig,
if
we're
doing
like
a
strict
bag
ordering
in
order
for
us
to
do
some
like
if
we
mean
the
ideal
world,
where
all
the
atoms
are
controlled
by
some
control,
some
type
of
controller
right
as
part
of
this
routine?
Doesn't
that
mean
that
everything
needs
to
be
so
all
the
state
needs
to
be
stored
in
it?
C
d-four,
self-hosted,
ish
pieces,
I.
I
Certainly,
when
I
was
working
on
this,
a
big
question
was
like
what
what
is
an
add-on
or
what
is
what
is
one
of
these
going
to
look
like
so
that
we
know
like
doing
a
temp
fighting,
for
example,
and,
as
you
say
like,
it
seems
that
the
way
we
are
all
headed
is
to
say
we
don't
want
templating.
We
want
simple
manifests
where
this
variability
is
sort
of
a
component
configure.
E
Component
company,
and
so
I'm
Adam,
as
some
atoms
are
already
add-ons,
are
already
using
component
config,
so
queue
proxy
already
uses
it
in
1.7
when
installed
with
cube
admin,
I'm
trying
to
push
the
networking
or
the
some
of
the
other
SIG's
to
move
their
things
over
to
coding.
Also
like
fluent
D
U
is
starting
to
use
component
config.
Somebody
I
think
it
was
you
Tim
said:
you'd
move
the
scheduler
over
to
use
it
and
1.8.
Well,.
D
E
F
So
I
think
maybe
one
way
to
look
at
this
is
that
maybe
in
one
age
this
is
something
that's
an
alpha
and
we
use
it
optionally
or
current
add-on
like
DNS
and
maybe
the
proxy
and
then
as
it
gets
proven
out
and
bit
gets
used
for
more
system
stuff
will
build
confidence
in
it
and
that's
sort
of
one
of
the
gates
to
going
towards
beta
and
beyond.
If
we.
D
Do
the
if
we
do
what
other
SIG's
do
and
we
say:
something's
our
p0,
p1
and
p2?
This
would
fall
in
eight,
the
category
of
p2.
If
it's
telephone-
and
you
know
you-
can
assign
priority
relative
to
the
person
who's
executing
on
it
right
since
you're
yeah,
it's
your
precious
pet,
but
as
the
sig
it'd
be
nice
to
like,
say
like
we
are
going
to
deliver
these
p0
items.
We
are
going
to
potentially
deliver
some
of
these
p2
items.
I.
F
A
So
I
agree
that
it
makes
sense
to
have
priorities.
Do
we
want
to
try
and
put
priorities
next
to
any
of
the
items
that
we've
gone
through
so
far.
E
E
B
F
B
F
We're
going
to
do
server
so
yukia
client-side,
plush
config
file.
What's
right,
right,
yeah
compliant
call,
you
know,
call
component,
config
I
mean
I,
think
there's
some
debate
on
whether
component
config
should
be
using
the
API
types
like
like
it
is
I
seem
to
remember
hearing
that
and
all
the
latest
work
on
that
I.
A
E
D
Would
also
quite
a
mark
I,
don't
know
whether
or
not
it's
actually
t0
across
all
the
components.
But
it's
a
p0
item
that
component
and
figs
are
going
to
be
required.
Component
config,
slash,
config
maps
are
going
to
be
required
for
some
items.
Eventually,
when
you
start
to
get
other
features
on
top
of
it.
Great
yeah.
E
So
the
specific
component
config
to
beta
is
not
actually
that
component
a
bigger
component
a
bit
config
is
a
design
pattern
that
we
have
using
config
Maps,
so
that
there's
actually
an
API
group
there.
We
should
delete
the
component
config
alpha
API
group,
publicize
the
dock
that
said
micro
and
start
having
other
components
using
that's
sort
of
this
this
section,
and
then
this
the
top
part
is,
is
a
delete
right.
This
is
a
structure.
A
F
D
B
D
Is
kind
of
a
circle
here
right
like
there's
the
static,
manifest
step
but
they're
the.
If
you
wanted
to
have
a
self
hosted,
control
plane
right
for
high
available
I
mean
you
get
you
get
the
controller
manager
and
scheduler
pretty
much
for
free.
Once
you
start
doing
self
hosting
your
check,
pointing
right,
you
just
say,
and
you
you
lock
on
the
component
or
the
config
map
right.
D
Server
stuff,
I,
don't
know
how
far
we're
going
to
tread
down
this
room
right,
like
I,
would
say
week
start
putting
some
of
the
H
a
step
is
p1
or
it
depends
on
the
priority
of
how
we
want
to
do
this,
like
it's
a
circle
of
features
right,
we're
all
dependent
upon
each
other.
We.
A
A
A
story
is
like:
how
do
the
cubelets
find
the
API
servers,
for
example,
and
keep
those
lists
up-to-date
over
time
and
I
would
be
happy
with
saying
that
we
don't
the
we
want
to
solve
that
eventually,
but
we
don't
have
to
solve
that
in
time
for
1/8
we
can
find
saying
we
don't
want
to
solve
that.
Therefore
1/8
you.
F
Self-Hosted
in
particular,
nodes
right.
That
is
a
step
in
the
right
direction,
if
not
where
we
want
to
end
up,
but
that
at
least
gets
us
progress
so
that
we're
no
longer
writing
stuff
to
disk.
It's
actually
like,
if
you
lose
a
machine,
you're
still
have
to
go
through
and
market
sucks,
but
it's
at
least
moving
us
in
the
right
direction
towards
being
able.
Then
we
can
solve
this.
The
dynamic
discovery,
problem
and
updating
stuff
as
a
separable
issue
from
that,
okay.
F
D
E
Missing
all
right,
I
realized
the
dock,
so
that
the
peasy
rows
are
at
the
top,
and
then
it
goes
down
from
there,
with
the
one
exception
that
the
certificate
rotation
to
beta
and
Phillips
did
stuff
to
GA
does
not
yet
have
a
priority.
Someone
would
like
to
suggest
a
priority
for
that.
We
can
stick
it
in
the
right
spot
and
then
take
a
quick
pass
through
the
priorities
and
make
sure
they
look
like
I.
Think.
D
A
A
F
We
also
need
to
identify
really
which
of
these
is
going
to
require
sort
of
the
feature.
Effort
process
stuff
like
flipping
on
the
bootstrap
signer
by
default,
and
the
authorizer
is
going
to
be
the
type
of
thing
where
we
want
to
make
sure
that
gets
a
lot
of
eyes
and
discussion.
Cuz,
there's
going
to
be
yeah.
F
E
The
other
thing
we
to
identify
is
places
where
we
have
dependencies
on
other
SIG's,
so
in
particular,
like
the
self
hosting
depends
on
the
on
cig
node.
We
have
gotten
sort
of
verbal
agreement
from
dawn,
but
we
need
to
make
that
more
explicit
make
sure
it
gets
part
of
their
planning.
But
the
court
say.
E
A
Excellent
songs
is
well
I
suggest
that
we
leave
it
there
for
then,
and
we
come
back
into
another
Passover,
this
talk
of
beginning
of
the
next
meeting,
but
it
sounds
pretty
good
to
me.
Thank
you.
Everyone
for,
for
the
effort
of
input,
any
last-minute
comments
or
other
items,
I
think.
A
Right,
thank
you
for
reminding
me
so
I
just
suggested
of
the
back
of
a
conversation
that
several
conversations
we
had
at
the
Leadership
Summit
in
the
Bay
Area
last
week.
Last
Friday,
the
just.
Let
me
get
a
group
of
people
together
to
talk
about
the
specific
implementation
plans
in
a
particular
Lucas
and
Tim
to
talk
through
their
designs
and
so
I've
suggested.
A
meeting
on
Friday
at
10:00
a.m.
Pacific,
which
few
people
have
except
already
I,
want
to
make
sure
that
that
is
seen
as
a
public
meeting.
A
E
I
guess
I'll
have
one
other
thing
that
came
out
of
the
the
leadership
meeting
is
that
some
other
things
that
are
very
effective
have
sort
of
they're,
both
their
their
weekly
or
bi-weekly.
Sort
of
standing
me
like
this,
but
also
breakout
meetings
for
more
focused
topics,
and
so
this
is
sort
of
the
first
effort
to
have
that
sort
of
one
of
those
breakout
meetings,
and
so
it
just
becomes
more
common.
E
A
A
So
if
you're
on
this
call
and
you're
thinking
like
oh
I'd,
like
to
contribute,
but
I
haven't
spoken
up
yet
and
then
we
are,
we
would
love
to
encourage
you
to
get
involved
and
we'll
be
I,
think
it
makes
sense
to
start
wanna,
do
office
hour
sessions
to
to
encourage
contributions
and,
and
so
on,
I
think
we're
over
time.
So
thanks
everyone
and
and
see
you
all
next
week,
Yeah
right
it
points
bye,
bye,.