►
From YouTube: Kubernetes SIG Cluster Lifecycle 20171219
Description
Meeting Notes: https://docs.google.com/document/d/17J496IR2tXKw7k97fxwz2KUWOf9rpBD3pIEsmDiJQSw/edit#heading=h.n1as6vm6zx8z
Highlights:
- How to run kubeadm e2e tests
- Path forward on https://github.com/kubernetes/kubernetes/pull/56956
- KubeCon retrospective
- KEPs for sig cluster lifecycle
- Change the Wed morning meeting to office hours for kubeadm
- Status of kubeadm adoption working group (fold into office hours)
- 1.9 planning retrospective
A
B
It
looks
like
there's
an
issue
that
X
and
a
pull
request
related
to
some
of
the
documentation
around
how
to
hack
that
together
yesterday,
I
played
with
Moran
toises
work
using
the
darker
and
darker
clusters,
and
that
seems
to
work
okay,
but
I'm
struggling
and
so
I'm
curious.
How
other
people
do
this
and
what
we
can
do
to
help
document
it
it
better.
Ok,.
A
So
the
short
answer:
I'll,
give
Jacob
Beecham
a
couple
of
months
back,
did
like
a
three-part
tutorial
on
how
our
ete
infrastructure
is
configured.
It's
like
how
you
can
configure
you
know:
prowl
and
the
different
sort
of
runners
and
the
pull
request,
builders
and
so
forth,
and
to
run
our
tests
either
on
a
schedule
or
elsewise
and
I.
Think
as
part
of
that
he
talked
about
how
to
run
the
tests
themselves.
A
Obviously,
as
a
Googler
who
has
you
know
easy
access
to
GCP,
I
think
those
breed
GCP
specific
and
that's
how
we
tend
to
tests
because
we
aren't
getting
a
bill
for
that
directly.
So
it's
kind
of
it's
pretty
convenient
for
us,
so
I
think
there
are.
Are
some
instructions
on
how
to
use
communities
anywhere
to
launch
cube
admin
clusters
to
do
end-to-end
testing
the
same
way
that
we
do
it
on
pull
requests?
B
B
A
I
saw
there
was
a
pull
request
a
couple
of
days
ago
to
make
some
of
our
use
less
flaky,
which
is
also
another
issue
like
a
lot
of
times.
If
you
write
a
test
and
locally,
it
works
just
fine,
and
then
you
check
it
in
and
it
doesn't
work
100%
of
the
time,
which
is
then
source
of
file
issues
against
this,
which
is
frustrating
so
if
there,
if
there
does
seem
like
it,
doesn't
work
consistently,
then
we
should
try
to
think
about
how
we
can
write
the
test
so
that
it
will
work
consistently.
A
Cool
all
right,
so
I
would
search
for
those
videos,
I
believe
that
Jacob
recorded
them
and
posted
them.
If
anybody
happens
to
have
a
link
or
can
find
one
farther
down
in
the
lifecycle
meeting
notes
that
we
can
share
with
Lee
that'd,
be
really
great
I,
don't
know
if
they
made
it
on
to
the
YouTube
playlist
for
our
sig,
but
we
do
have
a
playlist
where
all
of
the
stay
recordings
get
posted,
and
so
that
might
be
another
place
to
look
for
them.
A
C
C
The
quick,
easy
way
to
fix
it
is
to
give
up
a
warning,
but
I'm
not
sure
this
is
the
proper
way,
because,
if
I
give
a
warning
but
but
I
have
used
the
day,
P
the
user
didn't
get.
What
he
asked
for.
So
I
think
that
we
should
consider
at
least
two
other
TANF
yields
which
are
blocking
so
don't
accept
the
DNS
name
in
the
validation,
validation
or
end
on
the
net
name
properly,
which
can
be
useful
for
achievement,
and
so
I
think
that,
in
my
opinion,
we
should
go
for
free.
C
A
C
A
D
Whoops
regarding
ipv6
support
this,
whether
we
should
have
it
like
this
implicitly
after
going
to
be
it
as
well
or
if
we
should
have
some
specific,
enable
ipv6
mode
right
now.
We
just
tell
it
implicitly
by
seeing
the
advertised
address
with
IPS
or,
but
we
might,
as
it
touches
a
lot
of
different
places
like
everything
from
a
few
proxy
to
cube
DNS.
So
it's
probably
something
that
could
make
sense
in
a
cluster
API
because
of
classical
role-playing
or
like
that
has
the
whole
you
yeah.
A
And
I
think
the
other
thing
with
ipv6,
as
my
understanding
is
right
now
it's
it's
an
all-or-nothing
right.
We
have
duals
tech
support
or
you
can
have
a
mixed
cluster.
So
if
you
do
want
it
like
advertise,
your
API
server
over
maybe
ipv6
and
ipv4
right,
where
it
could
maybe
find
both
and
maybe
externally
or
you're,
addressing
the
ipv6
that
could
end
up,
causing
your
entire
cluster
over
to
ipv6,
which
might
not
be
exactly
what
you
want.
I,
don't
think
keying
off
that
to
make
the
global
decision
makes
a
lot
of
sense.
So
just
reading.
B
Through
the
flags
there's
a
distinction
between
the
bind
address,
IP
and
the
advertiser
address,
IP
things
like
advertised
addresses
used
to
tell
members
of
the
cluster
how
to
reach
or
how
to
route
to
the
API
server.
This
is
by
default
equal
to
the
bind
address
IP,
which
is
a
separate
flag.
That
is
the
definition
of
what
interfaces
to
listen
on.
So,
for
instance,
you
could
set
bind
address
IP
to
zeros,
but
then
like
specify
what
to
advertise
to
the
rest
of
the
cluster,
and
we
might
want
to
maybe
separate
the
configuration
of
these.
A
If
that
also
implies
that
we
expect
cluster
components
and
end
users
to
reach
the
cluster
over,
this,
it
might
be
right,
like
I,
could
have
met
in
a
scenario
where,
if
you
have
private
IPS
for
all
of
your
cluster
components,
so
your
two
proxies
and
your
cubelets
are
all
talking
over
ten
dot
address.
But
you
then,
additionally
want
to
expose
an
external
IP
for
cue
pedal.
A
D
A
A
Okay,
so
I
think
we're
okay
going
forward
number
two
for
now.
I
think
the
other
good
thing
about
number
two
is:
we
can
always
extend
the
functionality
later
if
we'd
like
to
expand
it
to
number
three,
but
this
at
least
gives
a
fixin
for
now
that
will
block
people
from
shooting
themselves
in
the
foot.
C
D
A
So
last
week,
I
put
on
the
agenda
to
have
sort
of
a
retrospective
about
Q,
Khan
and
I.
Wasn't
able
to
attend
for
that
part
of
the
meeting.
I
think
there
was
a
little
bit
of
discussion
about
you
know,
focusing
on
membership
and
getting
more
people
involved.
I
didn't
know
if
there
were
any
specific
takeaways
that
people
had
that
we
should
talk
about.
A
As
a
group
I
know,
I
filled
out
my
coupon
survey
last
night
and
put
some
notes
in
there
about
how
fear
to
me
felt
a
lot
less
focused
on
SIG's
and
sort
of
building
sig
communities,
and
that
was
something
that
I
missed
from
the
year
before,
where
I
felt
like
we
had
a
lot
more
sort
of
time
as
a
sig
to
sit
down
and
do
sig
planning
and
talk
about
future
direction,
and
that
might
be
just
because
last
year
we
had
had
a
whole
day
to
sit
down.
Just
as
I
said
can
do
that.
B
Things
I've
really
liked
from
the
OpenStack,
so
much
I
pretended
is
that
they
have
like
sub
team
onboarding
days
as
like
a
separate
part
of
the
conference
to
do
that
kind
of
planning
and
I
guess
a
new
spin
on
boarding
Klepper
from
the
conference.
So
there's
like
days
dedicated
to
that,
which
is
something
you
may
want
to
consider
it
given
our
size
that,
just
might
you
sense.
A
Yeah
I
think
that's,
maybe
what
I'm
thinking
of
because
last
year,
when
it
was
installed,
we
had
a
full
day
before
the
conference
started.
It
was
just
a
cluster
lifecycle
where
you
know
we
sat
in
a
room
for
five
or
six
hours
and
we
did
a
bunch
of
whiteboarding
and
talked
about
planning
and
and
that's
sort
of
along
the
lines
of
what
you're
saying,
and
that
was
really
nice.
A
D
Yeah
I
mean
I
raised
the
issue
on
the
mailing
list
some
time
ago,
but
I
think
the
conclusion
was
that
people
weren't
in
town
by
like
Monday
morning,
not
enough
at
least
so,
but
but
generally
planning
for
a
whole
day
like
from
the
cube
con,
organizes
point
of
view.
It
could
be
a
good
thing.
So,
like
okay,
we
have
the
official.
The
conference
is
three
days
and
then
we
have
one
day,
maybe
for
the
contributor
or
Leadership
Summit.
D
We
have
one
day
for
onboarding
and
it's
of
course
optional
certain,
but
at
least
informally
for
a
lot
of
things,
not
just
the
class
lifecycle,
the
one
that
happens
to
have
an
office
nearby
I
mean
we
were
in
Seattle
last
time.
So
then
Robert
could
host
us
in
the
Google
office,
but
it's
well.
It
depends
on
the
sig
then
again,
who
has
access
to
space
and
where
so,
maybe
we
could
race
it
like
it's
an
organized
the
level
that
it
could
make
sense
to
have
some
kind
of
informal
meet
up
for
more
things.
Yeah.
A
I
guess
the
way,
so
we
did
the
contributor
summit
the
day
before
the
conference.
If
we
push
that
forward
one
day
and
assume
that
most
people
going
to
contributor
summit
were
also
going
to
be
involved
in
sort
of
the
sig
don
boarding
day,
we
did
the
sig
onboarding
day
right
before
the
conference
started.
That
would
allow
people
that
were
not
part
of
the
contributor
summit
to
just
come
one
day
earlier
for
the
conference
and
be
sort
of
part
of
the
sig
process.
I
think
that
schedule
might
make
a
lot
of
sense
right.
A
Any
other
wrap-up
items
from
from
Kubek
on
I'll
point
out
that
I
met
a
lot
of
people
in
person
which
was
awesome,
I,
don't
think
I
intended
very
many
talks
this
year,
but
it
was
great
to
see
a
lot
of
people
there
in
person,
I
was
great
to
meet
a
lot
of
new
people.
It
was
great
to
see
a
lot
of
the
interests
that
we
had
in
our
sig
I
added,
both
Lucas
and
my
talks
to
our
YouTube
playlist.
First
enclosure
lifecycle
last
night.
A
Unfortunately,
the
sig
update
was
not
recorded,
so
we
had
like
an
overflowing
room.
Full
of
people
leave
us
a
really
small
space
for
that
which
was
also
kind
of
unfortunate,
and
they
didn't
record
it.
So
we
can't
post
that,
but
it
was
the
slides
were
available
if
people
are
interested
in
and
the
topics
that
were
presented
there,
I.
F
Just
mentioned
it
something
the
open
sex
I
mean
would
do
was
they
would
have
a
developer
track
and
a
that's
what
a
non
developer
track.
Basically,
because
it
exactly
just
said
where
the
developers
would
typically
not
go
to
most
of
the
talks
anyway.
So
then
you
have,
you
know
three
days
for
scheduling
whatever
we
want
to
schedule.
F
G
Thing
I'd
love
to
get
folks
feedback
on
and
I
was
talking
to
some
of
the
CNCs
folks
is
that
there
were
about
four
thousand
people
at
cube.
Con
Austin,
the
Seattle
Convention
Center,
holds
up
to
eight
thousand
as
cube
con
grows
and
as
the
audience
changes
you
know,
is
it
gonna
actually
be
the
right
place
for
folks
to
get
together
and
meet
you
know.
G
Do
you
think
that
cube
con
is
the
right
place
for
that,
or
should
we
try
and
find
other
ways
to
get
people
together
for
face
to
face
and
having
a
more
of
a
like
one
of
the
things
that
I
think
might
be
interesting
is
to
have
a
very
open.
You
know
kubernetes
community
conference,
which
is
focused
at
people,
actually
helping
build
the
thing
and
then
have
a
more
outward
facing
business
case
cube
con,
but
I
mean
it's
not
my
decision,
but
I'd
love
to
get
feedback
from
folks.
E
I
think
I
think
that
there
are
like
issues
with
the
fact
that
when
we
try
to
have
like
so
you
get
to
offset-
let's
say
you
know-
let's
say
there-
are
some
sort
of
budget
really
difficult
strengths
inside
states
like
because,
because
like
these
are
the
rooms
that
we
have
that
these
are
the
ones
we
can
use
and
it's
kind
of
hard.
So
you
can
were
to
consider
you
know
the
the
contributor
summit
as
a
separate
event,
which
is
which
is
you
know,
doesn't
have
to
happen
that
the
same
Convention
Center
and
doesn't
happen.
E
G
Obviously,
there
are
some
downsides
to
that.
I
know
for
me
personally:
I'm
torn
because
I
have
responsibilities
both
on
sort
of
the
business
side
and
then
also
you
know,
responsibilities
to
the
community
and
I
think
there's
a
lot
of
other
folks
that
were
in
the
same
boat
and
so
I
wasn't
able
to
go
to
as
many
of
the
community
talks
or
you
know,
I
was
luckily
able
to
attend
a
lot
of
the
contributor
summit,
but
I
know
that,
like
there
was
a
CNC
F
governing
board
that
day
right.
A
That
that
sort
of
perspective
Joe
would
what
Justin
proposed
to
be
intractable.
For
someone
like
you,
where,
if
we
had
multiple
tracks
during
Yukon
so
like
in
Seattle,
if
we
have
8,000
attendees-
and
we
say
you
know-
maybe
300
of
those
people
are
on
sort
of
this
developer
track.
Where
we're
having
lots
of
Signet
ups
we're
having
sort
of
breakout
sessions,
we
have
maybe
more
inwardly,
focused
talks
even
and
then
we
have
a
bunch
of
sort
of
outwardly
facing
things
for
the
other
7,500
attendees.
G
Would
be
paid
or
multi
would
be,
it'd
be
really
hard
for
me,
at
least
personally
because,
like
you
know,
I
spent
a
lot
of
time
doing
like
podcasts
and,
like
you
know,
meeting
with
other
companies
because
they
were
there
or
you
know,
meeting
with
candidates.
Try
to
recruit
so
there's
I
mean
there's.
You
know,
I
think
you
know
I'm
just
hopefully
like
torn
in
every
direction
at
these
things,
and
so
the
community
stuff
tends
to
fall
by
the
wayside.
Unfortunately,
yeah.
G
D
Yeah,
the
the
thing
also
is
like
to
some
extent.
We
already
had
this
like
split
myself
and
didn't
even
go
to
a
normal
talk,
I
think,
maybe
one
and
then
focused
most
of
my
time
on
like
meeting
people
from
the
sig
and
things
like
that,
and
also
talking
that
is
also
like
for
me.
It
is
non-trivial,
well
go
to
Austin
like
just
like
that
right
and
for
many
people.
D
So
if
we,
if
we
do
split
the
conference's
or
whatever,
then
you
also
need
to
consider
all
the
things
with
like
how
do
we
get
people
there
that
wouldn't
be
able
to
come
otherwise,
now
like
I'm
speaking
at
cube
con,
like
also
the
business
conference,
so
that
gave
me
a
ticket
contributor
summit
as
well.
That's
yeah!
It's.
G
Is
you
know
yeah
when
folks
are
coming
from
far
away?
They
don't
want
to
have
to
travel
place.
I,
don't
know
I,
just
like
I
I,
don't
know
how
to
fit
it
all
into
one
week
without
you
know
where
there's
both
sides
of
the
coin,
but
they
inward-facing
stuff,
but
also
the
the
outward-facing
stuff
so
yeah
Jason
mention
it
I.
Don't.
A
Have
a
silver
bullet
here
you
can
mention
in
chat
that
give
me
Andrews.
Just
access
to
engineers
is
good
value
too,
but
I
worry
that
if
we,
if
we
split
things
completely
separately,
many
of
the
engineers
aren't
going
to
go
to
the
you
know
outwardly
facing
one
less
they're
giving
talks
and
we're
gonna,
you
know,
send
all
those
people
to
the
community
event
because
travel
budgets
and
so
forth.
A
You
know
you're
gonna
have
to
pick
maybe
one
or
the
other
to
go
to
and
that's
the
other
value
of
having
them
together
as
it
sort
of
does
get
a
lot
of
engineers
together
in
the
same
place.
So
I
don't
know
if
there's
a
great
answer,
but
you
definitely
are
some
George
and
try
to
figure
out
how
to
I
mean
I,
maybe
they're
not
in
charge
I
guess
they're.
Maybe
there
are
sort
of
community
representatives
to
cube
con
well.
G
I
mean
I
can't
tell
you
that
the
contributor,
some
at
this
time,
landed
hard,
I,
don't
think
I
think
it
makes
sense
for
us
to
take
it
seriously
and
plan
it
further
in
the
future.
Make
sure
people
know
the
dates,
get
the
attendance
list
sort
of
figured
out
in
terms
of
what
space
we
can
do,
how
many
people
can
actually
attend
the
way
the
attendance
list
was
chosen.
This
time
around
was
not
I,
wasn't
happy
with
it.
G
G
I
I
don't
know,
Caleb
I
mean
I
was
talking
to
George
and
Parris,
and
you
know
there
was
ranked
in
terms
of
what
was
going
on,
who
was
invited
well
and
it
did
very,
very
late.
There
were
things
like
you
know
the
presentations
and
who
was
presenting
in
the
sig
updates
that
weren't
widely
vetted.
So
there
were
surprises
across
that
I
think
there
was
definitely
refer
a
lot
more
planning
going
into
going
into
the
into
this
conference.
That's
that's,
you
know,
that's
my
take
on
it.
I
don't
know.
If,
if
others
would
agree,
I
mean.
B
B
D
F
I'd
love
to
try
like
I,
think
one
of
things
we
can
do
is
we
can
try
things
right
like
we
have
all
these
safes.
We
can
try
different
things
in
different
SIG's
I'd
love
to
try
like
a
24-hour
online
hackathon
type
thing
right
or
online
meet
a
thon
right,
we're
with
24
hours.
Most
people
are
not
going
to
attend
or
24
hours.
We
expect
so
there's
no,
like
timezone
preference,
there's
no
travel
requirement.
F
It's
like
pop
in
hop
out
do
what
you
want
break
off,
that
sort
of
thing,
I
think
would
be
cool
and
obviously
it's
a
lot
of
work
to
like
set
up
the
structure
for
that.
But
we
can
I
think
it'd
be
fun,
also
to
try
that
and
see
what
what
happens-
and
it
might
be
a
great
way
to
get
more
involvement
with
people
that,
for
whatever
reason,
wouldn't
otherwise
join.
A
Alright,
the
interest
of
time
we'll
move
on
to
the
next
topic
again.
I
think
I
have
the
rest
of
the
agenda
here.
So
if
so
many
has
something
they
want
to
talk
about.
Please
add
it
in
I.
Don't
think
this
was
covered
last
week,
but
we
started
toward
the
kept
process.
So
this
is
the
kubernetes
enhancement
proposal
process
with
Jace,
maybe
a
month
ago,
or
so
we
went
through
and
filled
out
one
as
a
group
during
the
cluster
API
breakout
meeting.
A
We
filled
out
another
one
for
that
effort,
but
we
probably
need
more
than
sort
of
two
caps
to
be
tracking.
It's
a
sig,
if
you
think
about
the
the
scope
that
they're
supposed
to
have
there's
a
lot
of
scope
for
this
thing
and
if
we
want
to
break
that
down
to
somewhat
manageable
chunks
and
talk
about
what
we're
going
to
do
for
1.10
I
think
we
need
to
put
some
more
effort
in
there.
A
So
we
talked
during
the
meeting
about
the
fact
that
doing
those
as
a
group
was,
you
know,
somewhat
slow
and
laborious
and
we
should
take
it.
Offline.
I
haven't
had
time
to
do
it.
Offline
and
I
don't
think
anybody
else
has
either,
but
we
should
maybe
try
to
set
the
goal
of
you
know:
I
guess.
The
next
meeting
is
January,
2nd,
there's
a
lot
of
holidays
between
now
and
then,
but
we
are
sort
of
already
into
the
1.10
development
cycle.
So
I
think
we
should
set
the
goal
of
having
January
2nd
to
be
a
day.
A
We
can
review
caps
and
hopefully
people
can
find
some
time
between
now
and
then
to
write
some
thruster
review.
If
you
don't
have
any,
we
won't
have
anything
to
review,
but
ideally
we'd
have
some
to
go
through
and
use
that
as
sort
of
the
driving
for
what
we
expect
to
finish
during
1.10
I
know,
people
have
already
started
sending
PRS.
You
know
to
update
cube
admin
through.
A
G
Now
are
you
thinking
Robbie
that
this
is
just
for
cluster
lifecycle
internal
facing
stuff?
So
it's
just
a
way
to
organize
efforts
in
the
state,
because
I
think
there's
this
dual
purpose
for
caps
for
both
organizing
stuff
happening
within
a
sig,
but
then
also
as
a
tool
to
actually
communicate
and
organize
cross
sig
to,
but
most
of
the
stuff
is
really
just
within
a
state
within
the
state
right.
Well,.
A
A
Will
the
past
that
feature
actually
has
right?
We've
been
working
with?
No,
it's
like
working
really
hard
with
the
no
team
to
get
checkpointing
working.
We've
been
talking
a
lot
to
see
gaps
to
get
the
the
day's
a
surge
update
going
which
didn't
get
off
the
ground,
but
it
did
actually
have
dependencies
on
other
things.
A
In
that
specific
case,
I
think
the
goal
with
the
caps
was
was
both
to
try
out
the
new
process
and
see
how
it
worked
for
us
and
try
to
replace
our
old
process
of
having
a
Google,
Talk
and
just
kind
of
chatting
about
it
during
a
meeting
and
not
ever
really
going
back
and
updating
that
throughout.
Throughout
the
you
know,
the
release
cycle,
and
so
I
guess
I'm,
hoping
that
we
can
use
it
kept
both
for
inwardly
and
aspects
externally
facing
features
and
maybe
do
a
better
job
keeping
them
up-to-date
throughout.
G
Because
I'm
working
on
reorganizing
some
of
the
I'm
creating
a
kept
directory,
we
talked
about
this
as
Sega
architecture
this
week,
I'm
creating
a
kept
directory
at
the
root
of
the
community.
Repo
and
I'm
gonna
try
and
build
some
some
low
grade
tooling
around
that,
so
that
we
can.
We
can
start
creating
some
value
out
of
that
process
that
stretches
across
things.
So
awesome.
A
Yeah
and
I
think
that's
the
that's
the
main
benefit
right
for
what
we
were
doing
before.
This
is
a
little
bit
of
extra
process
compared
to
what
we
had
before,
but
I
think
that
if
that
extra
process
is
standardized
across
SIG's
and
we
can
use
it
to
track,
multi
release
feel
features
with
some
tooling
that
that's
gonna
actually
give
us
more
bang
for
the
buck
than
what
we
had
before
all
right.
Well,
any
feedback.
There
is
appreciated.
A
The
other
thing
if
people
want
to
contribute,
but
they
don't
feel
like
they
can
maybe
write
a
cap,
is
they
can
go
through
the
doc
of
ones
that
other
people
are
writing
and
turn
those
into
PRS.
Alright,
that
would
be
another
way
to
help
is
to
just
get
those
into
prz
and
you
know,
send
them
off
towards
Joe,
maybe
in
to
make
sure
they're
in
the
right
format
to
get
them.
The
emergent
unity,
repel,
yeah.
D
A
So
next
was
the
the
Wednesday
morning
so
Wednesday
morning
we
have
breakout,
meaning
it's
at
the
same
time
as
this
meeting,
but
a
day
later,
that
to
date
has
been
largely
focused
on
cube
admin
and,
in
particular,
building
self
hosting
and
building
H
a
and
doing
sort
of
deeper
dives
into
the
cube
admin,
design
and
direction
during
q.
Con
Lucas
have
proposed
that
may
we
want
to
switch
that
into
office
hours
and
instead
of
doing
deep,
dives
into
design
of
cube
admin.
A
F
F
E
I've
been
helping
Joe
with
the
officers
general
radius,
like
so
the
the
1
up
to
theta,
n
0,
K
1
and
that's
like
I,
think
I
every
month.
Right
now
on
you
know,
a
lot
of
people
turned
out
that
people,
so
yeah
I
think
there's
this
great
value
for
the
users
to
have
this
no
dedicated
slot
to
where,
where
they
can
ask
their
the
burning
the
question,
and
obviously
this
context
says
some
of
this
would
be
more
contribute,
oriented
and
there's
so
much
and
user
commended
that
questions,
but
like
they
won't
fix.
D
Now
I
mean
I
think
this
would
be
mostly
like
rebrand
or
a
better
repurposing
as
well,
but
also
like
sending
a
clear
message
to
our
contributors.
Maybe
some
users
as
well.
If
they
have
questions
about
like
whatever,
but
but
mostly
like,
say
it
use
a
word
office
hours
as
some
more.
That
is
a
more
standardized
across
kubernetes
term
than
like
implementation
meeting
right.
Yes,
well
so
to
someone
that
is
new
to
the
community,
there
might
be
a
huge
difference,
although
try
to
cover
basically
the
same
things,
but
but
yeah.
D
Also
emphasizing
more
that
you
have
a
good
chance
to
come.
Bring
your
questions
here
about
how
to
like
get
this
PR
merged
or
whatever,
like
making
a
lower
the
barrier
to
actually
like
pinging
us
again
like,
for
example,
a
reviewer
hasn't
answered
your
PR
in
a
month
cuz
it
was
like
you've
got
lost
in
the
github
flood.
D
A
C
A
And
I'll
change
the
calendar
invite
also,
which
also
brings
up
the
question
I.
Think
Justin,
you
have
a
couple
of
sort
of
sig
ish
calendar
invites
as
well
like
the
cube
admin,
adoption
working
group
and
the
cop's
office
hours.
Do
you
want
to
sort
of
promote
those
onto
the
sig
calendar,
or
are
you
happy
having
those
separate
I?
Will.
F
A
I
think
Luke
I
think
I
gave
Lucas
right
access
to
it.
Luke
Marcin
created
the
calendar
and
gave
me
right
access,
so
I
can
I
can
probably
add
you
to
that.
All
that
might
be
the
easiest
way
to
do
it
and
you
might
have
to
I,
don't
know
if
you
can
just
move
it
over
if
you
have
to
create
a
new
event
or
whatever
the
way
shouldn't
be
to.
D
Do
yeah
also
another
thing
which
is
that
the
the
cubed
amateur
adoption
working
group
has
been
suspended
for
a
month,
or
so
maybe
a
month
and
a
half
I
won't
personally
have
time
to
drive
a
third
meeting
or
so
well.
Now,
now,
from
the
from
the
next
year,
I
won't
tell
time
to
drive
any
meetings,
and
fortunately,
but
that
is
one
extra
I,
can't
at
least
even
a
tenth.
So
do
we
I
mean
it's
fair
to
say
that
we
just
will
wrap
it
inside
of
the
cube
bottom
office
hours
or
something.
F
Like
that
I
think
we
identified
a
bunch
of
things
that
would
be
helpful,
but
I
think
the
big
thing
that
would
be
most
helpful
for
debating
adoption
as
if
kubernetes
ships
difficult
to
install
a
feature
like,
in
other
words,
if
there's
a
big
carrot
using
cube
idiom
for
something,
then
everyone
will
be
like.
Oh,
we
want
to
use
qadian
because
it
gets
us
this
for
free
right.
That
will
be
that
that
will
be
the
actual,
forcing
function
for
using
to
video.
More
than
thou
shalt
use.
Covidien
I
think.
D
And
I
think
that
gradually
becomes
the
case.
I
guess
as
as
we
end
up
the
the
features,
for
example
like
component
config,
as
we
push
things
towards
that,
and
also
like
dynamic
cubelets
config,
which
hopefully
makes
me
than
110,
which
is
gonna,
be
a
huge
improvement.
But
has
some
effort
actually
actually
say
setup
and
configure
so
eventually
so.
F
A
A
A
Okay,
well,
let's,
let's
leave
it
where
it
is
for
now
we
can
send
out
something
to
the
email
list
and
see
if
it
makes
more
sense
to
scooch
it
over
I.
Don't
know,
I
mean
either.
One
is
gonna
overlap
with
other
SIG's
like
if
people
are
gonna
try
to
come
to
our
office.
Hours
drive
from
other
SIG's.
Both
slots
are
gonna,
have
some
overlap
and
make
some
people
unavailable.
I,
don't
know
which
one
will
be
worse.
A
You
know,
especially
as
Joe
mentioned
part
of
the
purpose
of
keps.
Is
that
that
other
SIG's
can
look
at
what
you're
working
on
and
plan
around
your
timelines,
and
so,
if
you
say
we're,
gonna
do
X
by
1.10
and
really
that
happens
in
1.12,
because
you
know
things
are
much
slower.
That
may
actually
affect
this
schedule
for
other
teams,
because
if
they're,
assuming
you're
gonna
get
something
done,
do
you
know
that
it's
gonna
slip
their
schedule
too?
A
So
you
know
I
think
that
you
know
we
just
need
to
be
able
as
realistic
as
we
can
about
what
we're
gonna
be
able
get
done
during
each
release
cycle.
Make
sure
that
you
know
when
we
sign
up
for
things,
especially
if
they're
externally
facing
to
other
developers
that
we
are
realistic
with
our
timeframes.
So,
looking
at
the
planning
doc,
we
had
a
p0
for
getting
testing
done
for
cube.
Admin
seems
like
it's
in
place,
which
was
a
implementing
automatic,
upgrade
testing
I,
believe
we
have
automatic
up
retesting
now,
yeah.
D
We
do
one
one
thing
that
we
yet
want
to
add
this
automated
downward
testing
to
eventually
move
the
dependencies
out
of
Google
with
their
own
like
bash
scripts.
That
I
don't
know
personally
at
least
how
they
work,
but
in
icing
rebasing
that
on
top
of
Cuba-
and
it's
a
lot
better
because
you
could
just
do
right,
I
mean
I
ran
a
lot
of
downgrade
testing
on
my
laptop.
So
it
was
just
a
cube.
D
A
D
D
But
they
don't
ship
within
it
and
join
thing.
That
might
also
be
an
option,
and
you
know
so
what
we
discussed
was.
Basically
we
want
to
go
to
GA
as
soon
as
possible
in
one
the
next
year
to
unblock
people
that
we
met
during
coop
con.
That
said,
oh
I
think
Cuban
is
great.
It
fits
my
what
I'm
trying
to
do
exactly,
but
I
can't
use
it.
D
My
company
cuz,
it
says
data,
and
that
is
my
my
company's
policies
like
YC
and
well
I
can't
so
that
is
gonna,
be
the
mole
important
dating
then
shipping,
something
that
is
that
would
be
and
inspired
by
swarm
within
it
and
join
for
masters
as
well
is
something
that
we
should
take
up
to
debate.
Is
that
something
with
kubernetes
architecture
that
is
actually
achievable,
Chris,
so
I
did
an
interview
with
the
new
stack
during
kilkanin.
We
talked
about
this
there.
There
are
so
many
aspects
of
a
company's
clusters
that
has
to
be
AJ.
D
D
D
A
Yeah
I
totally
agree.
I
think
it
would
be
useful
to
write
like
a
or
maybe
rewrite
like
a
cube,
Adam
scope
argument
about
what
we
are
building
and
sort
of
where
it
ends.
I
think
you
know
I've
seen
a
lot
of
you
know:
email,
threads
and
dark
comments.
Basically
saying
you
know,
there's
no
way
you
can
declare
it
GA
until
it
has
AJ
for
various
reasons,
and
we
need
to.
If
we
don't
agree
with
that
sentiment,
we
need
to
write,
write
it
down
and
have
a
good
rationale
for
that.
A
That
kind
of
dovetails
and
the
next
one
was
was
AJ,
which
I
think
Luke
has
covered
pretty.
Well
there,
it's
it's
not
in
1/9
we
in
the
design
phase.
For
that,
after
that
was
switched
using
dynamic
cubic
configuration,
I.
Think
from
our
side.
We
did
everything
it
just
it
didn't
make
it
out
of
alpha
ink,
you
Burnett
ease
itself
and
so
there's.
We
can't
turn
it
on
by
default,
because
it's
not
going
to
be
available
in
all
clusters.
A
Next
was
getting
a
cube
admin
sort
of
phases,
API
out
of
alpha
or
in
into
a
new
brevin
alpha,
and
so
that
didn't
make
it
in
1:9
next
was
the
cluster
API
an
alpha
version
of
that
we
have
an
alpha
version
of
that.
That's
what
we
presented
at
cube
con
I,
do
want
to
get
the
comments
on
the
PRS
resolved
and
get
the
sort
of
official
PRS
merge.
A
So
we
can
move
forward
and
we
talked
last
week
one
of
the
dilute
C
guys
is
going
to
start
working
on
iteration
of
the
API
using
machine
sets
on
top
of
machines.
So
we
should
expect
that
pretty
soon
next
was
improving
test
coverage
for
for
cube
ATM.
So
in
particular,
people
have
requested
in
the
past
that
we
test
with
two
or
more
CNI
providers,
which
I
believe
we're
now
doing,
isn't
right.
Yeah.
D
A
Cool
I
don't
think
we
didn
t
work
on
adding
other
specificity
tests,
as
we're
mentioned
here,
but
limited
to
being
a
call
before
you
join
Lucas
that,
as
he
was
working
on
that
CV
stuff,
you
wanted
to
write
ETA
test
for
that.
So
I
think
as
we
do
from
new
features
in
incubating.
We,
which
either
new
unit
integration
or
ET
tests
as
a
side
effect
of
having
good
features,
be
added.
Yeah.
D
And
also
the
unit
test
coverage
has
been
proved.
Oh,
the
cost
of
the
cycle
was
something
yeah,
so
we
talked
about
in
the
conformance
meeting.
That's
something
that
kind
of
belongs
to
signal's
life
cycle
is
to
write
some
kind
of
IDI
test
that
make
sure
the
cubelet
api
Fonzie
secured,
for
example,
that
is
kind
of
trivial
to
write
because
you
can
proxy
through
through
the
api
server
or
get
that
information
from
the
api
server
and
just
curl.
D
A
Yeah
there's
some
great
talks
about
that
clearance.
Security
at
cube
con
and
there's
also
a
product
called
cube
bench
which
attempts
to
go
through
and
sort
of
audit
clusters
and
tell
if
they're,
secure,
I,
don't
know.
If
there's
a
way
we
could
integrate
some
of
those
checks
and
this
on
a
buoy
as
well
and
say,
like
your
conformance
and
secure,
which
would
be
a
big
part
of
performance,
is
that
you're
like
had
meet
some
in
empower
up
security
I,
think
it's
implement
differently
from
sauna.
Buoy
want
to
jump
in
yeah.
G
I
think
there
is
an
adapter
to
run
the
cue
bench,
stuff,
understandably
I
think
in
terms
of
conformance
I
think
it
it
makes
sense
to
think
about.
Do
we
want
to
actually
be
able
to
attest
to
this,
but
there
are
definitely
different
views
in
terms
of
how
secure
clusters
need
to
be
in
its
environment
specific,
and
you
know
so
it's
it's
not
a
it's,
not
as
clear
picture
as
I
think
it
probably
shouldn't
be
in
terms
of
being
able
to
say
that
and
there's
there's
a
lot
of
history
there
too,
and
is.
F
G
D
A
Alright,
so
next
was
improving
that
cube
admin,
release
process,
I'm,
not
sure
I,
think
Tim
had
proposed
some
some
changes
here
to
make
it
better.
I
know
that
the
as
part
of
building
a
released
the
release
team
is
supposed
to
be
building
cubed
packages,
although
I
think
for
1.9
Jacob
just
went
ahead
and
built
those
for
expediency
since
we're
doing
it
on
our
Friday
afternoon.
A
But
but
I
do
believe
that
the
there
is
an
improvement
in
the
fact
that
we're
not
reliant
on
siculus
or
life
cycle
to
build
those
as
part
of
extraneous
release,
but
I
think
Tim
had
some
more
specific
improvements
that
he
wanted
to
do.
Jacob
to
go
back
and
rebuild
all
of
our
packages
to
fix
our
CNI
dependencies,
so
I
guess
we
can
consider
that
an
improvement
before
we
were
saying
C
and
I
version.
A
You
know
five
zero
point,
five
point
one
or
greater,
but
if
you
actually
run
six
on
old
version
of
kubernetes,
it
doesn't
work,
and
so
he
bumped
the
package
version
and
made
it
an
equals
instead
of
a
less
than
or
equals,
which
required,
rebuilding
and
pushing
something
like
300
packages.
Before
we
cut
1/9
so.
D
The
reason
we're
black
here
or
we
didn't
make
any
progress-
is
that
we're
basically
waiting
for
basil
to
well
come
and
save
us
I,
don't
know.
The
thing
we
have
now
is
a
homegrown
some
homegrown
scripts,
and
we
want
to
standardize
that
wait
basil
and
that
will
also
make
it
much
easier
to
do
like
check
out
a
comet
run
basil
build.
It
will
give
you
binaries
will
give
you
Deb's
product
emerges
and
then
you
could
push
that
to
whatever
GCS
or
cloud
storage
or
you
could
have
it
local
and
just
run
EDA
tests.
D
So
that
is,
and
as
Fabrizio
also
pointed
out,
it's
something
that
a
lot
of
contributors
struggle
with,
because
it's
not
at
all
obvious
it'sit's,
where
we,
where
wacky
so
yeah
we're
waiting
for
basil
to
add
multi-touch
support,
I,
think
it's
in
at
head
there.
Since
last
time,
I
was
talking
to
Jeff,
but
we're
still
waiting
for
committing
us
to
adopt
such
a
released
version
for
for
things
for
basil
and
yeah
cool.
A
D
Yeah
we
also
I
went,
went
through
all
of
the
cubed
I'm
documenting
and
try
to
for
exam
some
of
the
feedback
that
I
got
was
there
isn't
even
a
definition
on
what
cube
enemies
in
our
Docs,
and
that
is
true,
because
we've
updated
the
dark
gradually
but
not
like
seeing
to
the
full
picture,
and
there
was
kind
of
'some.
There
was
a
lot
of
text
that
was
written
in
1/4
that
was
still
present,
so
yeah.
D
Yeah
and
no
component
config,
or
we
did
make
some
progress.
For
example,
although
the
cube
cubelets
component
configuration
didn't
make
beta,
there
were
some
significant
progress
in
terms
of
reshuffling
things
internally
to
make
it
even
possible.
The
same
thing
have
been
done
for
for
the
proxy
and
the
cube
scheduler,
so
we're
nearly
good
to
go
there.
There
is
some
open
debate
still
like.
A
We
found
of
Q
Khan
that
sig
API
machinery
has
no
intention
of
doing
this
anytime
soon
for
the
API
server.
So
if
we
want
to
make
that
part
of
the
cluster
API
or
part
of
the
cube
admin
API,
then
we
need
to
get
say
architecture
to
tell
them
that
they
have
to
do
it.
So
we
should
think
about
escalating
and
if
we
believe
it's
important
and
come
with
a
strong
rationale
of
why
it's
more
important,
the
other
things
on
my
backlog.
Do
you
think
the
list
that
we
have
here
is
too
long.
A
A
D
And
also
the
control
manager
might
be
achievable
with
the
cloud
provider
refactoring
efforts
as
our
cloud
node,
a
cloud
controller
manager
needs
to
have
component
config
we're
basically
making
that
a
beader
requirement.
You
know
that
for
us
to
push
harder
on
the
core,
Control
Manager
as
well,
and
also
it
will
help
a
lot
when
running
your
out
of
three
cloud
providers.
A
All
right
so
we're
just
about
out
of
time,
so
I
think
we'll
just
press
two
things
in
the
dark,
because
they're
all
lower
priority
in
program
we're
talking
about
and
then
I'm
punting
the
last
agenda
item
to
next
time
and
speaking
of
next
time.
I
don't
know
if
anybody
is
planning
on
being
here
next
week,
I'm,
certainly
not
gonna,
be
here
next
week,
so
I
propose.
We
cancel
that
meeting
and
meet
again
on
January,
2nd
yeah
sounds
sounds.