►
From YouTube: kubeadm office hours 2019-10-16
A
A
A
A
So
basically
a
version
exists
at
at
an
end
point,
but
if
we
try
to
get
this
version,
for
instance,
the
cube
ATM
binary
is
still
missing.
Him,
like
two
minutes
later,
is
being
applauded
him
it's
there,
but
this
is
a
really
a
problem
for
sig
release.
For
brief,
she
attempted
to
send
a
fix
for
this,
but
the
problem
is
that
I
think
we
have
to
extend
the
wait
period
to
more
and
I.
Don't
even
know
like
what
is
the
appropriate
value
for
this,
maybe
at
least
two
minutes.
A
A
I
think
it
was
40
seconds
in
the
back
of
maybe
okay,
all
right,
I'm
going
to
set
the
piano
for
this
later.
The
other
PSA
is
moving
ownership
from
Seaquest
lifecycle
to
seek
code
provided
GCP
of
some
of
our
tests.
So
basically,
this
again
about
testing
the
story.
There
is
that
if
you
look
at
some
of
the
tests,
for
instance
in
test
grid,
we
like
here,
for
instance,
you
get.
A
These
are
tests
executed
in
a
job
in
the
tape
job,
but
in
some
cases
we
have
like
here
it
says
sequester
lifecycle,
but
the
actual
test
is
working
on
GCP
infrastructure,
it's
really
testing
aspects
of
gke
and
it's
not
really
owned
by
sequence.
Life
cycle.
At
this
point.
This
is
a
legacy
from
the
olden
days
and
Tim.
Sinclair
is
going
to
agree
with
me
that
we
really
wanted
to
move
this
away
from
the
sick.
So
this
is
one
of
the
PRS
that
is
doing
that
it's
pretty
much
waiting
on.
A
We
had
a
big
discussion
here
because
you
see
when
you
change
this
little
tank
in
the
test
it
affects
like
who
is
going
to
be
the
sick,
that
the
release
team
is
going
to
poke
about
the
problem,
and
you
know
every
release.
We
get
a
poke,
hey
can
sequence,
laps,
I
could
look
at
this,
but
only
Justin,
who
is
course
for
Google
can
investigate
because
some
of
these
tests
really
work
on
Google
specific
bits.
A
So
by
doing
this
move
we
are
now
transitioning
the
ownership
to
the
GC
Pequod's
provider,
and
this
is
one
of
the
beers
already
sent
a
couple
of
different
ones.
There
was
a
bike
sharing
on
slack
like
what
is
appropriate
thing
to
do,
but
just
wanted
to
give
this
PSA
that,
after
these
peer
merges
on
lazy
consensus
later
this
week,
the
release
team
is
not
going
to
focus
anymore
about
this
problem,
which
he's
good.
Thank
you
for
this
work
problem.
A
A
So
this
works
fine,
except
that
it
broke
when
you
use
a
couplet
in
a
cloud
provider
and
I
already
spoke
to
the
owners
of
this
particular
area
of
the
couplet
and
some
of
their
like
under
under
Kim.
Does
that
I
spoke
to
him
like
hey?
Do
you
know
why
this
is
happening?
Is
he
told
me
that
they
did
like
the
rest
of
the
fields
that
are
important
are
being
updated,
but
only
this
provider
ID
field
is
not
being
so.
A
A
C
B
I
discussed
this
with
Lumiere
they'd,
be
the
main
TL,
TL,
DR
and
abyss
is
that
creating
the
node
we
get
into
some
weather
situation
so
the
days
to
revert
and
go
back
to
the
previous
version
of
in
it
and
add
a
new
phase
that
with
me
after
at
the
end
of
the
init
process,
that
that
will
make
the
the
first
control
train
ready
for
certificate
rotation.
So
this
is
not
ideal,
but
probably
it
is
the
simplest
thing
to
do,
and.
B
C
Instead
of
waiting
for
the
provider
ID
to
be
somehow
created
by
the
underlying
provider
in
Class
VI,
because
I
as
far
as
I
remember,
I,
think
this
provider
idea
is
created
by
because
the
EPI
art
waiters
or
something
like
that.
So
maybe
we
can
check
that
because,
once
that
we
had
this
face,
get
rid
of
it.
If
we
find
a
way
to
create
a
note
from
cube
ABM.
If
we
fix
this
issue
in
cluster
epi,
if
it's.
A
A
This
problem
should
come
from
the
couplet,
it's
not
from
Casa
API,
it's
from
actually
the
integration
with
the
legacy
quad
providers.
There
is
this
embedded
logic
for
quad
providers
in
the
couplet,
but
the
it
really
doesn't
make
any
sense.
So
the
approach
for
created
a
lot
was
bit
object
is
risky
because,
first
of
all,
nobody
like
understands
this
colt
really
except
for
Andrew,
but
even
he
couldn't
find
why
this
has
happened.
So
if,
for
instance,
this
is
broken
on
GCP
right
now,
but
it's
if
it's
also
broken
on
I,
don't
know
Asia
or
iws.
A
B
B
A
There's
a
lot
there's
another
problem
here
that
if
we
I
discovered
that
if
we
use
this
method,
for
you
know
the
writing,
the
boots
of
Kubrick
off
I
realized
that
certificate
relation
does
not
work
at
all.
Sorry,
disabling
certificate
rotation
is
for
possible.
So
it's
very
weird
I
need
to
Deacon
the
correct
again.
A
A
F
A
A
A
Okay,
so
this
is
a
something
that
seek
out
or
introducing
kubernetes
and
the
kept
pretty
much
managed
to
yesterday.
Sorry,
the
yeah,
the
pier
for
a
cape,
maliciously,
it's
a
complicated
change.
The
TLDR
is
that
now
we
have
to
patch
our
namespace
the
keep
system
namespace,
if
you
want
to
keep
the
labels
we
have
to
control,
incorporates,
is
just
tier
control,
plane
and
also
component
like
these
two.
A
We
now
have
to
apply
some
logic
to
handle
this
on
the
namespace
level.
Otherwise,
the
node
restriction
controller
that
we
enable
is
going
to
block
the
creation
of
mirror
pots
for
the
static
pots
that
we
use
for
the
control
plane
and
you
can
find
more
details
in
the
Kip
I
already
have
a
peer
for
this.
By
the
way
it's
here,
you
can
have
a
look
at
that.
A
A
So
we
have
to
pick
a
location,
maybe
again
introduce
a
new
face
which
is
called
like
whitelist
namespace,
something
like
that.
The
change
is
actually
doing
is
applying
this.
Finally,
sorry,
creating
an
annotation
with
this
key
in
the
keep
system
in
space
having
this
value
for
the
annotation
and
pretty
much
patches,
the
namespace,
a
sock,
a
question
I
have
here
that
I
guess
I
should
ask
the
authors
of
the
cap
is
like
where,
where
should
we
apply
this?
A
A
D
A
Problem
with
the
mirror
pots
is
that
they
found
some
security
issues
with
them
by
creating
a
static
pot.
This
is
all
indicate
by
the
way,
by
creating
a
static
pot
on
a
node
that
is
already
compromised
by
an
attacker.
You
can
basically
create
a
mirror
pot
in
the
cluster
that
has
privileges
to
do
disruptive
actions,
so
by
whitelisting
a
set
of
pots
you
can
prevent
that
and
the
mechanic
is
going
to
be
applied
as
alpha
117
feature
gate
disabled
by
default
in
119.
A
The
future
gate
is
going
to
be
on
by
default
and
in
120
is
going
to
be
GA,
which
means
the
feature
is
going
to
be
always
enabled
so
I'm.
This
is
a
long
term,
but
I
just
wanted
to
or
get
the
discussion
going
as
soon
as
possible,
like
the
site
in
which
face.
We
want
to
apply
this
whitelist
like
what
do
you
think
about
another
face?.
B
A
Maybe
the
face
is
going
to
be
the
same,
but
we're
going
to
remove
the
logic
from
it.
So
it's
a
question
whether
okay,
we
can
introduce
these
new
two
phases
and
remove
one
of
them
at
some
point
in
time
which
yeah
is
going
to
need
an
actually
qualities,
wait
to
break
the
user
bar
script
that
fragment
phases.
A
D
A
A
So
I
long
term
I
have
a
problem
with
the
phases
a
bit
like
when
I
spoke
to
her
yesterday,
like
I,
explained
to
him
hey,
we
have
these
phases
that
patch
stuff
in
the
node
object.
He
said
that
like.
Why
are
you
doing
this
serum
and
I?
Try
to
explain
that
we
have
CRI
sockets
and
also
we
have
the
like.
What
was
the
other
one
things
we
punched
the
tins
and,
first
of
all,
the
couple
authority
supports
pains
from
the
execution
point,
so
we
don't
have
to
punch
this
later
at
all.
A
B
Think
that
okay,
we
have
a
different
set
of
problem
in
the
on
the
table,
so
one
is
what
we
are
doing
today
and
if
we
can
do
it
better
and
when
I
rotate
Sierra
I
was
added,
there
was
no
way
to
get
this
information
out
of
the
KU
ballot.
Maybe
today
is
possible,
so
if
today
is
possible
and
we
win
open
an
issue
and
we
try
to
fix
it
for
for
next
date.
The
second
point
is
the
second
question,
which
was
the
original
question
from
Ross.
What
was
why
not
conflate
the
action
on
Q
ballot?
B
Unmarked
control
pain.
He
in
my
opinion,
we
should
not
conflate
that
are
not
related.
Mark
control.
Pain
does
not
means,
is
not
a
related
to
covalent.
This
phase
already
as
a
as
a
semantics,
I
think
that
it
is
cleaner
for
the
user
to
add
a
new
phrase.
For
instance,
I
am
making
an
example
when,
when
I'm
doing
he,
if
I'm
creating
a
cluster-
and
it
is
a
single
node,
cluster
I
will
be
tempted
to
skip
a
marker
control
rain.
B
A
A
We
already
have
complains
that
our
upward
config
phase
is
using
is
doing
more
that
the
needed,
like
it's
annotating,
the
sierras,
okay,
the
strobe
and,
like
a
user
said
like
hey
I,
don't
want
you
to
remote
eat
my
CNI
circuit
when
I
execute
only
the
port
config
phase.
So
ideally
it
should
have
been
a
separate
phase.
The
annotated,
zero,
so
I
will.
B
A
E
G
Component
config
cap,
so
the
enhancement
freeze,
has
passed,
but
we
can
still
file
an
exception
request
if
everybody
is
okay
with
like
merging
this.
So
far,
I
have
seen
only
lunar
and
Tim,
and
today
me
and
probability
actually
spent
quite
some
time
discussing
this
gap,
and
one
possibility
for
this
gap
is
to
extract
the
customized
portion
of
it
and
make
it
part,
possibly
with
some
modifications
of
the
advanced
customizations
cap
and
also
actually
not
implementing
this.
G
Not
getting
this
cap
merged
for
this
cycle
does
not
mean
that
we
are
not
on
a
good
track
to
move
out
of
three,
at
least
in
my
opinion,
so
removing
the
dependencies
on
internal
types
is
going
to
just
move,
defaulting
and
validation
right
now
and
I.
Don't
think
that
this
is
quite
a
disruptive
you
exchange.
A
A
B
B
A
A
H
You
mean
it's
a
path
that
is
supported.
It's
not
like
it's
not
recommended
by
by
the
maintainer
or,
for
example,
capi
machinery.
I
recall
us
discussing
this.
Basically
rendering
kubernetes
kubernetes
is
not
supported,
even
though
for
some
other
parts
on
the
project,
we
are
doing
it,
for
example,
for
the
CCM
yeah,
so
if
possible,
we
should
avoid
doing
this.
Yes,.
A
H
A
E
G
G
B
B
B
B
B
The
current
Europe
UX
for
the
simplest
case,
so
covered
mean
in
it
and
I'm
a
simple
user
I
want
to
go
to
the
fold.
Basically,
today,
I
can
do
everything
to
reflects
the
most
common
use
case
today
is
that
the
people
pass
some
config
files,
and
then
there
is
the
advanced
use.
Cases
supported
today,
which
is
the
user
pass,
also
a
folder
with
static
code
manifest.
This
is
the
current
situation.
Okay,
from
the
current
situation,
we
are
already
moving
to
a
new
situation,
which
is
the
disk
in
the
middle
that.
B
B
B
B
B
What
are
the
effect
of
this
cap
on
on
the
user
and
and
and
that's
my
my
TL
DR-
that
things
are
changing,
but
the
dopamine
women
should
be
should
remain.
Really
simple,
for
the
simplest
case
should
should
remain.
Result
may
be
reasonably
simple
for
comfort
for
common
use
cases
and
then
I'm
open
to
to
make
advanced
a
use
case.
More
complicated,
I,
don't
know
if
it
make
sense
this
at
this
round.
Yes,.
A
B
A
Exactly
inside
the
Cape
I
didn't
see
some
of
the
interactions
between
Cuba
game
in
the
other
installer,
in
terms
of
like,
for
instance,
I'm
going.
To
give
you
an
example
like
hey
I
want
to
upgrade
kubernetes
my
qualities
question
using
cube
ADM
from
one
version
to
another,
like
I,
have
a
version
that
I
want
to
use.
Let's
pass
this
version
to
another
add-on
installer
and
get
the
latest
like,
for
instance,
core
DNS
in
keep
proxy
components
that
I
want
to
install
I,
didn't
see.
A
B
B
G
B
So
there
is
a,
let
me
say
something
is
not
automatic
alight
like
today,
but
it
is
really
a
corner
case
when
define
and
worship
it,
and
it
is
acceptable
for
me
and
I
would
like
to
make
the
same
schema
for
there.
Don't
stutter
the
same
free
schema
for
the
Dominus
Allah,
trying
to
figure
out
who
or
what
are
the
effect
on
on
the
user's.
A
There
has
to
be
awake
to
the
way
to
pipe
conversion
for
a
certain
atom.
So
if
you
have
a
stale,
Q
proxy
configuration
on
disk-
and
you
say-
hey
I
want
to
upgrade
cube
ATM
and
its
add-ons,
but
I'm.
Please
use
my
configuration
because
I
or
I
have
important
fields
in
this
proxy
config.
If
you
pass
this
old
Q
proxy
config
to
the
new
add-on
version
in
the
add-on,
installer
is
going
to
fail,
so
the
other
one
is
static.
Has
to
pipe
the
message
back
in,
say:
hey!
A
G
B
B
E
A
A
E
A
E
A
F
Yeah,
this
is
what
you
met,
because
you
mentioned
to
join
phases
and
I
remembered
it
once
I
had
to
open,
and
this
is
something
we
should
probably
focus
on-
I
mean
that's
what
I
think,
although
I
don't
have
anything.
Imagine
died
yet
because
it
was
some
time
yeah
I
just
joined
sorry
for
not
attending
lately,
oh
I
need
to
I
need
to
in
to
bring
myself
up
to
do.
I
see
you.
A
F
A
Okay,
at
this
point,
we
are
implementing
the
blade
phases.
They
are
slightly
more
antinodes,
I
would
say,
but
we
are
probably
going
to
get
it
working
if
we
are
have
concerns
Matic
by
the
way
we
have
to
probably
not
implement
the
PR,
because
you
know,
if
you
have
concerns,
we
not
going
to
be
able
to
merge
this
so
yeah.
We
have
to
think
about
this
pretty
well
all
right.
Anybody
have
any
general
topics.
I
wanted
to
do
the
quick
edge
triage.
We
have
all
ten
minutes,
I.
Think.
A
A
A
A
E
A
Well,
so
everybody
who
has
the
time
please
look
for
this
beer
and
review
it
I'm
going
to
do
it
myself,
probably
like
before
much
to
have
a
look
as
well.
We
have
another
peer
that
is
from
Santa
qe4,
virtually
every
DMV,
this
PR.
This
is
for
those
who
check
the
see.
I
see
a
explanation
when
we
do
the
you
know,
the
check,
expiration
commands
I
think
it's
probably
not
passing
some
of
the
Google
just
the
diseases-
yes
that'd,
be
quite
it
I
better
set
here.
It
is
this
one
is
good
to
go.
A
A
D
A
F
F
A
All
right,
so
so,
actually
this
is
pretty
moving
it
from
Cuba.
So
keep
up
is
a
separate
bootstrapper.
It
lives
inside
the
kubernetes
kubernetes
repository
and
it
it's
basically
shouldn't
touch
a
medium
at
all,
so
I
think
the
PR
just
you
know
the
user
has
created
a
little
bit
of
a
mess
with
the
commits
and
it
touched
qadian,
but
now
I
will
remove
the
label.
A
H
It's
basically,
we
had
an
issue
with
a
function
that
determines
the
the
number
of
IP
is
available
in
a
subnet.
So,
while
moving
it
to
the
you
chills
package,
we
had
some
issues
with
the
some
boundaries
check
now
that
they
are
fixed
and
approved
by
Jordan
I
bumped,
again,
the
you
tools
were
seen
in
Cuban,
Aires
communities
and
I
will
be
able
to
update
this
PR.
H
A
Thanks
is
that
okay
to
did
this
is
a
valid
if
change
I
have
to
modify
this
PR.
What
else
we
have
this
is
just
changing.
An
API
field
is
not
approved
by
the
couplet
folks.
This
is
another
PR
on
me.
I
have
to
basically
address
later
and
we
don't
have
any
more
time.
Okay,
so
see
you
again
next
week,
folks,
we
have
to
leave
the
room
for
the
coaster
API
all
right,
bye-bye.