►
From YouTube: SIG Cluster Lifecycle 2020-10-06
A
A
A
I
think
we
submitted
the
following
talks:
that
is
the
sig
intro
session
image
builder
request
api.
B
C
It's
it's
me
and
katie,
like
yeah,
we
got
the
message
from
nancy
and
I
didn't
sign
up
already
and
I
think
katie
as
well.
Okay,
so
great
solve
it.
Thank
you
for
yourself.
C
A
Okay,
do
we
have
any
questions
about
kubecon
n
a?
I
don't
just
say
that
I
can
present
solo
if
you
will
not
have
the
time
for
this.
A
Yeah
tips
suggested
that
we
should
have
some
pie
charts
with
the
user
feedback
from
the
survey.
I
don't
see
anything
else
that
we
can
change.
Quite
frankly,.
D
That
sounds
good.
I
mean
we
highlighted.
I
think
three
sub
projects,
cluster
api
kubereum,
something
else
it
would
be.
I
mean
we
could
probably
highlight
another
one,
but
yeah
there's
not
going
to
be
if
the
format
remains
basically
the
same.
I
don't.
I
don't
think
we're
going
to
make
too
many
changes
to
it,
which
is
a
bit
disappointing
because
eventually,
we'll
run
out
of
people
that
haven't
heard
the
message
but
yeah
it
makes.
D
A
I
guess
we
can
highlight
image
builder,
even
if
we
have
a
talk
for
it.
Last
time
we
highlighted
corporate
configuration.
We
can
also
highlight
it
again
because
it's
understaffed
my
general
idea
is
that
we
should
have
more
visibility
for
the
alpha
projects.
A
So
yeah
it's
stupid
is
decided.
I
can
start
working
on
the
draft.
Eventually
let
you
know
to
have
a
look
and
we
can
decide
how
we,
if
you
want
to
participate
in
the
recording.
A
I
I
don't
know
if
the
platform
is
decided
yet
so
there's
some
lack
of
clarity
around
the
platform
they
wish
to
use
with
we're.
Definitely
changing
platforms.
This
time.
A
I
think
we
can
follow
up
in
the
next
sick
meeting
on
this
topic.
I
think
it's
not
clear
when
the
deadline
for
the
presentation
is
presentations
is
last
time
we
had
a
deadline
before
we
uploaded.
C
C
A
C
A
Record
the
video
presentation
until
october
22nd.
C
A
C
A
All
right,
this
means
that
we
don't
have
that
much
time.
Okay,
so
this
is
like,
like
two
weeks
from
now,
pretty
much,
which
means
that
I'm
going
to
start
working
on
the
slides
for
the
talk
I
participate
in
and
see
like
if
bhcc
can
record
the
vote
or
how
we're
going
to
proceed.
But
I
encourage
everybody
else
to
proceed
with
recording
as
well
and
preparing
the
slide.
E
Can
you
hear
me
yes,
yeah
prakash
here
I
thought
brian
was
working
on
image
builder
a
couple
of
times.
I
have
attended
this
meeting.
I
would
like
to
know
what
exactly
will
include
like.
I
have
been
dealing
with
image
builders
in
other
projects
like
airship,
and
we
are
trying
to
figure
out
exactly
what
for
kubernetes.
We
should
include
for
cube
admin.
E
A
So
is
this
a
question
what
we're
going
to
present
at
cubecom
yep?
Basically,
the
presentation
is
going
to
be
led
by
I
believe
mosh.
Can
you
see
my.
A
So
basically,
this
person
is,
you
can
pick
them
on
slack
and
they
can
give
you
some
details
about
what
they
are
going
to
include
in
the
presentation
or
I
personally
do
not
know
yet
what
the
contents
are
going
to
be,
but
you
should
definitely
conduct
mosh
about
it.
He's
one
of
the
people
presenting
there.
A
All
right,
any
more
psas
or
group.
A
A
A
The
list
is
quite
big,
but
basically
we
found
out
that
there's
a
lot
of
stale
code
in
cuba
in
some
of
the
alpha
features
that
we
have,
and
also
a
lot
of
features
were
already
used
by
a
lot
of
people,
so
we
decided
to
graduate
in
from
alpha
to
apparent
top-level
command
such
as,
for
instance,
kubernetes.
Alpha
cert
was
already
used
in
a
number
of
places
and
by
our
consumers.
Apparently,
so
we
decided
to
graduate
to
occupation
search
parent
command.
A
We
are
also
deprecating
self-hosting
support,
based
on
like
no
like
feedback
that
we
don't
have
that
many
users,
or
maybe
no
users
at
all-
and
I
just
have
a
look
at
these
tickets
that
I
added
in
the
list
here.
A
Also
helped
with
clearing
some
of
the
language
there
and
basically
we
transformed
I
transformed
the
draft
into
a
cap
this
weekend
and
we
were
able
to
budget
this
implementable,
and
I
want
to
be
very
clear
here
that
it's
implementable
in
terms
of
one
of
the
stages
it's
going
to
have
multiple
stages
and
the
first
one
is
basically
adding
the
control
plane
label
in
parallel
to
the
bus
label
and
also
announcing
to
users
that
they
should
tolerate
a
new
taint
that
is
controlled,
maintained.
Preemptively.
A
A
Secondary
stage
is
going
to
be
like
roughly
one
year
after
the
primary
stage,
but
this
depends
on
feedback
as
well.
So
if
users
are
complaining,
we
are
not
going
to
execute
after
four
releases
or
one
year,
and
we
may
delay
this
even
further
and
after
some
feedback
from
jordan,
we
decided
that
we
should
not
do
too
much
in
the
second
secondary
stage.
A
We
should
also
include
a
third
and
a
fourth
stage,
which
is
primarily
primarily
driven
by
the
fact
that
cubadiam
also
supports
immutable
upgrades
when
new
nodes
join
the
closer
there's,
some
complexities
around
which
labels
and
dates
are
tolerated
by
workloads
and
so
on.
A
This
is
also
out
loud
in
detail
below,
but
basically
the
summaries
here
somebody
here
is
that
this
is
going
to
take
20
months
at
minimum.
So
it's
going
to
take
a
while.
I
have
not
seen
replies
from
the
or
working
group
naming
on
this
topic.
A
I'm
pretty
sure
they're
going
to
be
quite
surprised
that
this
is
going
to
take
so
much
time.
So
I'm
going
to
keep
this
group
updated
in
terms
of
what
their
reply
is
going
to
be.
I
think
that
they
expected
something
that
is
much
shorter
than
that,
but
yeah.
This
is
the
complexity
of
changing
a
ga
feature
used
by
a
lot
of
people.
A
D
A
A
Depends
so
yeah
this
is
the
cap
I'm
going
to
start,
I
mean
I
don't
have
anything
to
start.
I
just
I'm
going
to
say
the
small
pr.
That
is
the
first
phase.
The
most
important
part
of
the
pr
is
the
action
required
release
load
and
after
that
we
also
have
to
announce
it
to
a
lot
of
people.
I
guess
I'm
going
to
try
to
contact
contribex
to
ask
them
what
is
the
best
way
to
communicate
to
the
wider
public,
possibly
twitter.
E
Item
so
are
we
with
respect
to
trains?
E
Are
we,
including,
is
when
we
add
a
new
node?
Those
stains
have
to
be
reorganized
or
something
or
have
to
be
tested
in
the
label
or
what
I
didn't
get
it.
What
exactly.
A
You
should
definitely
read
the
camp.
The
the
processes
explained
there.
Okay
I'll
do
that.
Thank
you
yeah,
because
currently
we
applied
a
taint
that
is
called
master.
It
has
a
key
that
is
includes
the
word
master.
We
have
to
change
that,
but
if
there
are
deployments
that
have
to
tolerate
this
particular
master
taint,
there's
a
complication.
There
is
a
process
that
we
have
to
apply.
A
E
E
D
Just
something
I
put
in
chat,
I
was
just
reading
the
last
wg
naming
minutes
as
far
as
I
can
tell
they
are.
They
are
not
yet
they
are
not
yet
in
agreement
as
to
how
they
are
going
to
decide
for
the
new
names.
A
But
I
am
confused
because
the
work
already
is
happening.
We
have
people
like
aaron
and
others
already
approving
peers.
D
D
This
one
is
user
with
a
16-month
period
of
of
like
safe
change
or
more
like
if
we
rename
a
directory
or
rename
a
class
or
identifier
like
big
deal
right,
but
like
the
renaming,
the
things
that
you're
user
facing
are
are
more
significant
so
like
it
seems
a
valid
strategy
to
go
through
and
change
the
the
things
which
are
easy
to
change
like
we
can
change
them
twice
this
one.
I
don't
know,
I
don't
know
how
often
they
meet
either.
So
it
sounds
like
I
should
definitely
attend
the
next
meeting.
D
Maybe
we
can
maybe
we
can
achieve
consensus
only
on
the
the
label
and
taint
and
not
the
general,
because
it
sounds
like
they.
They
they
do
have
consensus
on
the
naming
of
the
master
branch
or
the
branch
formerly
known
as
master.
A
A
We
have
rose,
one
of
the
hardbacks
in
the
system
are
backs.
The
built-in
airbags
is
system
masters,
so
this
is
basically
the
explains
an
administrator.
So
that's
one
of
the
meanings,
the
other
meaning
is
a
master,
can
mean
a
control,
plane
node
as
well,
and
the
third
meaning
is
an
instance
of
the
api
server
itself.
A
So
you're
communicating
to
the
master,
you
can
say
you
can
replace
the
same
sentence
with
communicating
to
a
control,
plane
instance
of
the
api
server,
and
we
saw
this
in
a
couple
of
places
in
kubernetes
in
the
kubernetes
source
code.
A
So
if
a
node
has
a
role,
there
is
master,
we
thought
that
the
the
best
rename
here
is
this
node
is
our
control,
plane,
node,
so
that
that
is
basically
the
ratio
now
for
us.
D
C
A
Yeah
so
people
like
steven
oguso,
said
jordan
already
participated
there,
so
I
I'm
pretty
sure
a
lot
of
people
saw
the
kept
draft
and
control
plane
was
already
there.
Nobody
objected
the
cap
I
just
like.
Earlier
today
I
sent
an
email
to
the
working
group
mailing
list
with
the
latest
kept
iteration,
so
everybody
should
have
seen
this
already.
A
A
But
I
agree
that,
for
the
wider
rename
of
master,
we
should
think
about
what
is
the?
What
is
the
actual?
What
is
master
referring
to
in
this
particular
case,
because
it
has
multiple
meanings.
A
A
E
A
A
Change,
yeah:
okay,
any
more
comments,
so
the
the
rename.
B
Thank
you.
Can
you
hear
me?
Yes,
so
follow
up
from
last
update
about
kubernetes,
we
are
starting
to
collect
idea
about
the
what
could
be
a
roadmap
for
kubernetes.
There
is
an
umbrella
issue
and
the
operator
issue
is
linking
other
issue
where
we
are
discussing
the
the
possible
option
and
just
a
quick
explanation
about
the
the
process.
So
the
better
issue
is
just
the
collection.
B
We
are
not
defining
the
priorities
or
accepting
an
idea,
and
at
this
stage
we
are
just
collecting
information
and
as
soon
as
one
idea
is
mature
enough,
basically,
we
will
start
drafting
a
roadmap,
md
document
that
we
are
going
to
add
in
the
kubernetes
repo
and
from
that
we
will
start
with
with
cap
and
implementation
planning.
A
Yeah,
as
as
I
discussed
before,
butcher,
I
really
like
this
idea,
because
every
project
should
have
a
roadmap.
A
It
depends
like
when
you're
adding
new
features
like
what
are
your
plans
for
the
future
like
two
years
from
now.
Maybe
everything
should
be
documented,
because
this
way
can
also
attract
contributors.
Maybe
a
feature
that
is
delayed
later
later
in
time
can
attract
somebody
who
can
contribute
in
this
particular
cycle,
but
if
they
don't
know
that
you
even
want
this
feature.
If
this
is
not
even
on
the
roadmap,
it's
going
to
be
more
difficult
for
them.
To
discover
that
you
know
help
is
wanted.
B
For
british
request,
api
yeah,
quick,
quick
update
about
cluster
api,
so
v03.10
is
out,
and
I
this
should
be
the
the
the
latest
big
release
of
the
the
v03
series,
the
v103
series,
starting
from
now
with
zero,
the
the
v1
alpha
3
series.
We
got
patch
monthly
patch
release
if
necessary,
if
necessary.
B
While
on
the
main
branch
we
already
started
work
for,
we
won
alpha
four,
mostly,
we
are
doing
cleanups
preparation
like
upgrading
kubernetes
version
dev
and
all
the
dependencies
and
cuba
builder,
while
the
community
is,
is
starting
to
to
work
on
the
on
the
cab
and
and
the
the
the
rough
planning
is,
is
to
dedicate
the
the
month
of
october
to
discuss
all
the
kind
and-
and
so
we
can
start
beginning
of
november
middle
of
november
implementation.
A
I
have
a
question
a
hoping
that
this
is
not
very,
very
provocative
to
the
quest
api
planning.
What
do
you
think
will
happen
if
in
121,
cuba,
dm
removes,
if
you
want
beta
1
support?
B
Yeah,
this
is
an
interesting
topic.
We
have
an
issue.
This
is
part
of
of
our
robot
of
the
discussion.
My
gut
feeling
is
that.
B
In
in
in
the
channels
so
so
and
also
on
the
issue,
so
we
we
are
on
it,
but
I
I
don't
have
a
a
good
answer
for
you
now.
A
Okay,
do
you
think
that
view
on
alpha
4
is
again
going
to
embed
a
certain
version
of
the
kubernete
and
config,
or
should
it
just
treat
it
as
a
like
a
bob
of
data.
B
B
No,
I
don't
get
the
final
answer,
but
if
I,
the
the
the
latest
discussion
on
this
topic,
were
moving
in
another
direction,
the
idea
was
to
have
an
api
in
cluster
api,
which
define
basically
the
option
that
we
want
to
forward
to
kubernetes
and
this
api.
The
need
for
these
apis
is
is
for
mainly
for
two
reasons.
One
is
for
the
coupling
the
the
process.
The
two
project
and
second,
is
because
by
exposing
the
kubernetes
api
directly,
there
are
some
wireless.
B
A
So
yeah
I
see
this
is
like
this
is
on
this
idea.
I
guess
he.
This
is
what
he
proposed
at
some
point
to
abstract
away.
No
one
problem
with
that
is
that
if
users
want
more
and
more
features,
you
have
to
iterate
on
v1,
alpha
4
to
add
new
fields
to
the
abstracted
layer.
B
A
Yeah
any
more
questions,
for
course,
api,
a
prakash.
E
E
E
B
Could
you
kindly
explain
me
better?
What
do
you
mean
with
an
api
that
requires
at
the
mean?
Yes,
so
let
me
just
complete
so
in
my
opinion,
from
my
understanding,
an
api
just
define
a
data
model.
B
B
E
E
B
B
A
All
right,
let's
move
to
the
next
project,
just
seeing
cops.
D
Yes,
so
I
think
two
items
of
cross
project
interest
I'll
do
the
second
one
first,
the
second
one
is
that
we
are
working
on
adding
arm
support
or
a
contributor
cyprian
hackman
is
working
on
adding
arm
support.
One
of
the
blockers,
one
of
the
surprising
blockers,
I
guess,
is
that
we
need
to
find
a
reasonable
base
container
image
for
the
non-distrolus
case,
and
essentially
we
don't
really
have
one
anymore.
D
We
were
using
one
from
kubernetes,
which
is
no
longer
we're
not
on
getting
security
updates.
We
are
probably
converging
so
this
is.
We
want,
for
example,
mount
fsck
and
make
mkfs
make
file
system
in
that
container.
Those,
I
think,
are
the
three.
I
think
we're
likely
to
use
the
official
official,
the
one
that
docker
maintains
for
debian
10.
D
It
seems
like
a
reasonable
image
and
has
a
script
that
we
could
use
to
reproduce
it
we're
just
trying
to
make
sure
that,
but
not
sure
what
other
people
are
doing,
but
that's
sort
of
where
we're
thinking
about
for
the
it
was
a
surprising
gap.
I
think
I
don't
know
whether
other
people
have
the
same
requirement
for
a
non-trivial
base
container
image.
D
The
other
item
was
that
we
are
participating
in
oktoberfest,
cops
started
it,
I
guess,
but
like
actually,
it
seems
to
be
a
pretty
cross
project
participation
across
a
bunch
of
sick
cluster
lifecycle
ones.
Thank
you
to
lumiere
and
others
that
attended
a
couple
of
these
things.
It's
actually
it's
highlighting
to
me
the
that
there
are
cross-project
technical
discussions
that
probably
need
to
happen.
D
That
don't
necessarily
happen
here,
because
it's
more
of
a
structural
discussion
like
update
and
like
a
roll-up
discussion
and
obviously
don't
happen
in
the
individual
silo
meetings.
So
I'm
going
to
try
to
get
better
at
initiating
those
and
reaching
out
to
stakeholders,
but
sometimes
I
don't
even
know
who
those
people
are
that
have
overlapping
interests.
A
Yeah,
hopefully,
these
events
like
that
can
like
you're
saying
that
they
can
surface
some
of
the
shared
problems,
apparently
there's
also.
There
was
also
a
lot
of
spam
around
oktoberfest,
at
least
on
the
website
repository,
but
I
think
they
were
able
to
calm
the
situation
down.
D
Yes,
there
was,
there
was
apparently
spam.
The
repos
now
have
to
act
up,
the
reaper
doesn't
have
to
opt
in,
so
cops
has
opted
in
whether
that
of
course
means
that
we
now
are
gonna
be
a
target
for
the
spam
has
yet
to
be
seen.
We
have
not
yet
received
that
deluge
of
spam,
but
yeah
they've
they've
changed
the
rules
to
be
to
accommodate
some
of
the
spam
problems.
If
you're
reported
as
a
spammer
a
spam
pro,
you
are
ineligible
for
your
quarter
of
a
free
t-shirt.
D
So
hopefully
that
will
address
some
of
the
the
challenges.
A
Yeah,
hopefully
I
wanted
to
also
ask
about
arm.
Do
you
have
some
sort
of
a
communication
shadow
with
some
of
the
arm,
employees
that
are
work
lurking
inside
the
kubernetes
project
because
we
saw
them
at
signaled?
They
may
have
a
recommendation
for
the
distro
here.
D
That
would
be
great
yeah,
I
don't
know
if
it
is,
it
would
be
great
to
connect
with
them.
I
don't
we
are
not
in
contact
with
them.
I
don't
know
whether
it
is
a
we
don't
want
to
optimize
specifically
for
arm
it's
more
like.
We
want
to
find
a
an
image
that
works
for
us
and
on
all
the
architectures,
but
it
would
regardless
be
great
to
connect
with
them.
I
don't
know
where
do
they?
Where
have
you
seen
the
the
army
of
armors.
A
Yeah,
I've
seen
them
at
the
sick,
note
and
cigar
architecture
meetings,
but
it
might
be
a
bit
difficult
for
me
to
find
names.
A
A
You
can
maybe
find
arm
topics
yeah,
that's
great!
Thank
you.
E
Yep
prakash
here
tina
sue
is
the
one
who
participates
in
the.
A
Yes,
can
you
please
write
the
name.
A
Cops
what's
midi
cube.
F
Yeah,
so
we
are
releasing
a
beta
for
114
this
week.
It's
gonna
fix
a
lot
of
networking
issues.
We've
had
of
the
docker
driver
by
adding
a
dedicated
network
for
each
cluster,
as
opposed
to
just
reusing
the
docker
bridge
network,
we're
also
adding
a
stable
add-on
for
our
automated
gcp
authentication
for
your
kubernetes
applications.
F
So
you
can
just
enable
an
add-on
and
as
long
as
you
have
google
cloud
platform
credentials
in
your
environment,
we'll
find
them
and
throw
them
in
your
kubernetes
ammo
automatically
we're
also
adding
multi-arch
support
for
our
storage
provisioner
so
that
we
can.
You
can
actually
store
things
in
your
in
your
cluster
if
you're
not
on
amd
64..
F
So
that's
what's
going
to
be
in
the
next
release
and
the
other
thing
is
so
right
now
we
have
a
lot
of
users
in
china
and
they
can't
access
any
google
anything.
So
every
time
we
release
something
to
gcs,
whether
it's
like
our
tar
balls
or
our,
whether
it's
our
tar
balls
or
our
binaries
or
images
or
whatever
they
have
been
mirrored
to
alibaba
sort
of
by
users
on
a
on
a
best
effort
sort
of
basis.
F
But
we
kind
of
want
to
take
over
that
so
that
it's
available
to
all
users.
So
we
just,
I
need
to
be
able
to
know
what
to
how
to
go
about
getting
an
alibaba
account
for
us,
as
on.
A
Well,
sql
provider
definitely
have
us
alibaba
cloud
super
project,
so
maybe
that's
one
of
the
ways
to
find
someone
that
can
help
these
are.
Do
you
want
me
to
give
you
the
link?
Yeah
that'd,
be
great.
A
A
Who
is
going
to
pay
for
the
accounts
that
that
I've
seen
discussions
that
this
is
something
that
the
cncf
can
sponsor?
I
know
that
igor
varecki
is
in
charge
of
some
of
that,
and
you
can
also
ask
contributes,
because
igor
is
that
that
is
where
igor
hangs
out
right
so
yeah.
I
think
that
maybe
the
the
process
should
be
like
ask
three
backs
first
like
can
we
get
sponsorship
for
this
and
then
us
call
provider
like
hey?
Can
you
actually
help
us
with
the
setup
of
the
accounts
for
them?
A
A
Cube
hcdm.
G
Hey
yeah,
oh,
can
you
hear
me
sorry?
Yes,
okay,
great
yeah,
yeah,
the
there's,
just
a
minor
update
that
tdm
is
updated
to
default
to
scd-349,
which
is
the
the
version
in
in
119..
G
A
Thanks
daniel,
what
what
do
you
think
about
adding
potentially
hcadm
as
part
of
events
like
such
as
the
oktoberfest
to
get
more
folks
to
help
and
solve
the
pending
changes?
Yeah.
G
I
sorry
I
I
have
been
yeah,
I
I
know
kind
of
a
little
about
hacktoberfest
but
yeah.
Maybe
I
could
ask
ask
justin,
it
sounds
like
cops
is
taking
advantage
of
that.
So
maybe
a
cdm
could
as
well
yeah.
D
Absolutely
I
mean
I'd,
say:
there's
very
little
structure
I
tried
registering
for
the
official.
They
were
partnered
with
some
company
that
wanted
to
like
to
post
some
sort
of
the
official
calendar
in
order
to
do
that.
You
sort
of
had
to
jump
through
all
these
hoops,
and
I
gave
up
on
those
hoops
because
they
were
a
little
bit
silly
like
trying
to
find
the
city
in
which
your
virtual
event
was
going
to
be
hosted
was.
D
It
was
an
odd
question
that
was
required
things
like
that,
and
then
we
just
kept
going
round
and
round
and
we
missed
the
deadline.
So
I
was
like
well
that's
the
end
of
that.
I
think
doing
outreach
and
doing
meetings
and
advertising
them.
I
think
is,
is
a
a
good
idea
and
hector
references.
A
nice
excuse.
I
don't
think
you
know
if
it
wasn't
actually
professionally
grouped
with
something
else,
there's
not
a
lot
from
octoberfest
other
than
the
self-momentum.
A
A
So
as
long
as
you
have
a
good
list
of
basically
changes
that
you
want
students
to
make,
there's
a
high
chance
that
somebody
can
get
attached
to
contributing.
So
I
would
say
it's
a
it's.
A
some
of
these
events
are
very
good
idea.