►
From YouTube: kubeadm office hours 2020-01-22
A
A
A
Okay,
do
we
have
any
new
participants
to
the
call
I?
Don't
see
anybody
new,
okay,
some
quick
PSAs,
so
the
cap,
4
million
cube
a
DM-
saw
a
lot
of
discussion
yesterday
with
cig
release,
it's
quite
unfortunate
that
they
were
only
able
to
review
now,
because
it's
really
late
Tuesday
is
the
deadline
for
the
for
the
clips.
This
is
without
talking
about
the
exception
process,
because
we
also
have
this
process
where
we
can
delay
the
campaign.
A
Potentially
inaudible,
maybe
send
it
a
couple
of
weeks
later,
but
we,
the
kept
pretty
much,
has
something
like
around
170
70
comments.
At
this
point,
I
tried
to
answer
most
questions
by
Tim,
pepper
and
in
fact,
we
booked
a
private
meeting
today
to
discuss
some
of
that
because
it
we
ended
up
discussing
way
too
many
things
that
I
think
are
orthogonal
to
the
work
that
we
can
do
from
the
cap.
A
So
maybe
we
have
to
drop
the
proposal
that
Luca,
who
is
on
the
call
proposed
to
pre
tag,
releasing
the
cue
medium
repository,
because
the
tagging
coat
in
the
released
of
kubernetes
is
really
poorly
designed
them.
Maybe
we
are
going
to
only
use
the
quarry,
this
release
repository
to
release
comedian,
but
not
hang
it,
and
the
tagging
can
be
done
by
the
police
postman
job
that
I
discuss
in
the
clip.
A
A
little
details
like
that
and
also
Tim
pepper,
is
not
exactly
sure
whether
our
proposal
is
going
to
affect
cube,
karo
and
tools
that
also
move
in
the
future,
but
I'm
trying
to
explain
that
this
is
really
a
cube
ATM
proposal.
If
Kip
Grove
decided
to
do
something
else,
they
are
going
to
have
to
do
a
separate
cap
in
any
case,
so
yeah
I'm
going
to
have
this
meeting
later.
A
A
B
First
of
all,
if
I
can
add
you
in
any
way
to
push
the
cap,
let
me
know:
I
will
do
my
best
support
you.
To
be
honest,
I
tried
to
follow
all
the
discussion
indicator,
but
there
are
too
much
for
that
of
them,
and
but
if
you
want
the
ITU
in
the
in
the
meeting,
just
let
me
know
other
we
do
my
best
to
attend
or
not
hesitate
to
ask
for
app.
B
So
first
of
all,
first
point:
second
I
will
be
disappointed
if
we
are
going
to
postpone
this,
but
in
that
case,
I
think
that
discussion
will
be
between
should
be
between
and
cigarette
ease
and
the
sick
architecture,
because
we
as
a
cig,
we
are
doing
our
best.
Natalie
did
the
move,
but
we
can
do.
We
can
not
do
the
move
by
ourself.
A
Yeah
so
I
completely
agree
with
that.
Yesterday,
I
was
asked
by
signal.
Is
why
like?
Why
are
we
you've
given
us
an
ultimatum
to
approve
this
cap
and
I
said
that
we
are
not
even
the
ultimatum,
the
automate
almost
originally
created
by
Cigna
retexture,
who
told
us
hey,
please
give
us
a
timeline
when
chameleon
can
move
out
and
I.
We
gave
one
a
team
as
the
psycho
for
that
and
now
seek
release
a
lot
ready
and
not
happy
that
we
are
giving
them
ultimatum
so
you're
correct.
A
C
A
A
Okay,
so
in
this
repository
right
here
there
is
this
tool
called
Anago,
you
can
find
it
in
Google,
it's
a
big
pile
of
bash.
This
is
the
original
to
that
Google
created
when
the
kubernetes
release
process
started,
and
you
have
to
make
changes
in
this.
So
what
I
propose
to
seek
releases?
We
are
going
to
stuff
the
work
in
this
bash
bash
code
that
we
can
apply
the
change
to
build
qadian
from
here
and
I
already
looked
at
it.
I.
D
One
question
from
my
side:
so
I
agree
with
what
you
said
for
it.
What
I
wonder
is
if
this
is
a
situation
in
which
is
all
or
nothing,
because
maybe
we
can
try
to
get
Covidien
out
of
tree
for
118
but
release
from
KK,
and
while
we
test
all
the
machinery
from
being
QA,
the
amount
of
tree
I
mean
118
is
going
to
be,
as
always
only
with
the
difference
that
we
are
going
to
be
building
a
VM
outside
of
tree.
But
the
release
happens
from
KK
adding
from
118.
A
A
D
Yeah
but
listen
the
intermediate
I
mean
it
may
be
just
some
middle
ground,
because
if
it's
all-or-nothing
case
maybe
for
sake,
release
is
going
to
be
harder
because
they
are
under
pressure.
So
if
they
could,
you
know
we
could
trigger
this
experiment
without
releasing
from
there
just
triggering
the
experiment.
While
we
release
from
KK
for
118
and
we
have
everything
working
for
119
and
then
does
the
moment,
that's
where
we
actually
trigger
everything
and
we
actually
released
from
from
kubernetes
Kiev
atm.
It
was
just
an
idea.
Maybe
we
can
build
some
middle
ground.
Yes,.
A
So
I
don't
think
he
that
kept
exactly
covers
this.
At
this
point
it's
also
became
such
a
mess,
but
I
somebody
propose
that
we
can
copy.
We
can
copy
cube
ATM
to
the
cube,
ATM
repository
and
then
release
continue
releasing
from
KK
and
maybe
at
some
point
in
time
we
can
switch
the
flag
array
like
a
feature
gate.
We
can
switch
a
flag
or
environment
variable
somewhere
and
we
can
start
doing
cube
ATM
from
the
new
place,
and
this
is
like
the
fallback
strategy
for
the
whole
thing.
A
E
A
A
D
A
I
will
pick
folks
if
there's
anything
I
like
I
need
a
question
asked
by
like
also
by
this
group
in
general.
At
this
point,
we
are
entering
that
political
discussion
of
whether
signal
is
okay,
with
the
move
as
a
whole,
whether
they
can
help
with
PR
reviews,
and
things
like
that.
So
it's
not
really
on
this
group,
but
yes,.
B
Yeah
but
okay,
I
got
I,
got
to
the
point.
I
think
that
a
fair
comment
makes
sense.
So
he
even
if
we
don't
get
the
final
call
from
single
release,
I
think
that
this
cycle,
we
have
to
work
to
make
our
repository
ready
about
CIE
jobs
in
place
and
whatever,
even
if
the
rheinmetall
of
the
code
is
a
manual
or
or
periodic,
but
I
think
that
is
doable
because
we
can't
or
what
matches
in
the
in
the
hook,
admin
repo.
So
this
is
the
plan
B,
as
a
group
I
think
we
should.
B
A
Okay,
yes,
this
makes
sense.
We
can
create
a
list
of
tasks
that
we
can
perform.
We
can
even
have
the
peers
for
K
release
ready,
but
they
can
be
inactive
like
the
code
path
can
be
blocked,
disabled
and
at
some
point
we
can
enable
it.
So
we
can
do
a
lot
of
work
in
118
in
if
seek
release
are
happy.
We
can
just
flip
those
switches
updates
like
everything
in.
A
F
I
have
a
good
comment
here
and
I
think
that
we
should
probably
try
to
do
things
a
little
bit
differently.
We
should
probably
try
to
get
sick
release
and
cigar
textured
talking
so
because
we're
hitting
this
problem
because
we
are
the
first
piece
of
cake-
eight
I'd
be
actually
moving
out
of
cake.
Everybody
else
ain't
in
staging
and
they
don't
have
to
do
this
problem.
F
B
A
So
dims
already
committed
yesterday
on
our
discussion
in
seek
release.
He
said
that
we
can
proceed
with
whatever
state
the
Kapus
and
welcome
to
the
PRC.
If
something
is
not
good,
I
think
deems
is
trying
to
say
that
the
cap
is
a
process
that
can
be
iterative,
so
you
can
add
more
things
in
the
cap
in
the
future.
A
I
think
we
can
keep
it
implement
a
provisional
at
this
point
and
then
maybe
at
some
point
when
things
are
in
a
good
shape,
we
can
change
the
Cape
to
implementable,
but
in
the
meantime
we
can
work
on
the
stuff
roasty.
You
have
a
very
good
suggestion,
but
me
trying
to
book
a
meeting
between
these
six
is
like
almost
impossible.
They
are
very
busy.
Sick
release
are
on
fire,
currently
silica
texture
a
very
hard
to
reach,
so
I,
don't
think
I
should
be
trying
to
get
them
in
the
same
meeting
but
yeah.
This
is.
A
F
A
A
My
point
is
that
if
we
create
our
own
proposal,
we
can
move
out
and
maybe
at
some
point
see,
release
are
going
to
say:
hey
your
tools
are
not
correctly
designed,
let's
redesign
the
release
process
Kapadia.
My
point
is
that
we
can
have
this
our
one
way
of
moving
out
and,
if
cube,
cannot
want
something
else,
I'm
sure
they
are
going
to
want
something
else,
because
I
don't
know
how
they're
going
to
tag
their
commits.
A
A
All
right,
Anna
PSA,
is
that
this
is
related.
Donna
go
to
had
some
concurrency
bugs
in
bash
that
broke
a
lot
of
releases
of
different
artifacts,
karoku
blade
comedian.
I
think
this
is
given
a
good
outline
of
the
problem.
Basically
like
the
Debian
package
includes
a
binary
that
does
not
match
the
version.
So
the
package
is
like
you
see
it's
one,
sixteen
five,
but
the
contents.
The
binary
itself
is
a
different
version.
A
F
F
B
A
A
F
B
A
F
F
D
One
question
regarding
that
me:
we
can
ask
them
to
support
that
early
in
their
configuration,
so
we
have
time
from
there.
They
are
supporting
these
into
the
cubelet
configuration
until
they
start
duplicating
it,
so
they
if
they
implement
these,
the
earliest
things
into
the
config
and
they
duplicate
this
afterwards.
Maybe
we
have
time
to
do
with
change
and
to
test
the
change.
I
think.
A
D
A
D
A
F
A
Let's
move:
this
is
a
tricky
topic.
We
we
have
been
discussing
this
for
over
a
year
already
I
think,
but
it's
just
not
actionable
in
profits.
The
concurrent
agility
on
support,
so
somebody
from
VMware,
clean
and
ticket.
They
have
a
basically
testing
bed
where
they
are
testing
a
pretty
big
poster
with
density
AJ
things
like
that,
but
that
is
not
like
a
fairly
old
version,
so
we
had
a
big
discussion
here
and
apparently
we
ended
up
with
some
solutions
like
phablets.
We
proposed
a
way
to
prevent.
A
The
actual
problem
which
I
forgot
to
mention
is
that
there
is
basically
the
coke
or
adjoint
of
HCG
does
not
work
members
if
they
join.
At
the
same
time,
two
members
can
end
up
with
without
names
and
the
cube
ATM
code
is
going
to
fail,
marking,
adding
the
names
to
the
members
and
affably
show
proposes
solution.
Then
I
did
some
testing
I
also
like
proposed
an
iteration
of
his
solution
to
fix
the
problem
and
then
Fabrice
show
and
the
others
again
came
back.
Him
propose
yet
another
iteration
of
my
proposal.
A
E
B
The
change
is
more,
it
is
the
change
sector
that
we
see
now.
So
it
is
not
a
big
change
to
before,
but
is
touching
a
critical
part
of
cupboard
mean
so
I
am
debated
if
backporting,
the
till
the
115
I
I,
think
that
we
have
to
become
cultures,
so
we
have
to
deploy,
muster,
see
that
everything
works
possibly
to
set
up
a
parallel
joint
test
that
now
we
don't
ever.
B
Currently,
we
are
not
able
to
reproduce
it
for
two
reasons.
First
reason
is
that
they
we
use
kinder
for
tests
and
kind
of
does
not
support
parallel,
join
and
second,
is
that,
even
if
implemented
at
the
parallel
join
with
kind
of
I'm,
not
sure
we
can
reach
the
same
corner
case,
because
it
is
a
really
random.
So
it
really
depends
by
the
different
that
you
see
remember
what
time,
what
time
they
keep
to
start.
B
So
it
is
really
not
not
deterministic
to
reproduce
the
the
problem
so
I'm
not
opposed
to
backcourt
it,
but
I
I
want
I.
I
prefer
to
I
would
like
to
be
a
little
mm
in
this
in
a
safe
path,
so
they
may
be
the
first
to
discuss
with
team
and
see,
but
for
the
time
being,
for
the
moment,
I'm
I
just
want
the
change
in
master
and
see
if
this
does
not
create
a
regression.
B
B
So,
yes,
it
is
not
is
not
supported
and
also
they,
the
layer
mode
they
are
implementing
is
not
air
to
for
concurrent
join
because
basically,
their
support
one
one
learner
at
time.
So,
according
with
this
information,
I
I
think
that
as
a
cupboard
mean
which
I
where
we
are
consuming
a
fatty
CD,
what
we
can
do
is
to
try
to
implement
workarounds
and
like
this
one,
but
I
think
that
we
should
state
clearly
somewhere
that
we
are
not
supporting
the
concurrent
join
of
a
TCD
member
because
it
is
not
supported
by
TCD.
B
A
B
Okay
and
last
point,
I,
think
that
we
should
made
clear
is
that
this
this
the
problem
only
apply
for
the
concurrent
join
of
masters
and
and,
and
that
mean
that
you
do
in
it
and
in
it
is
already
a
separate
step.
Then
you
are
doing
you
want
to
do
the
concurrent
join
of
the
two
addition
master
foreign
sha
cluster.
So
it
is
not
a
big
import
for
the
user
because
it,
the
parallel
join
of
worker,
is
supported.
Is
it
only
applies
to
the
the
problem
of
perilous?
B
C
So
that's
very
interesting
because
we
are
using
history
to
be
we're
doing
parallel
join
you
actually
do
not
even
do
the
process
I
probably
to
describe,
but
we
just
run
B
and
we
have
the
jump
of
Eurasian
D
in
the
comedian
conflict.
So
it's
one
step
to
join
them
all
and
I've
created
at
least
ten
clusters
in
the
last
we
can
never
had
this
problem
and
we're
using
115,
so
I
think
is
a
very,
very,
very
edge
case.
C
A
D
C
C
B
B
A
C
A
Some
point
I
left
my
machine
to
create
in
destroyed
clusters,
and
maybe
it
created
100
clusters
or
so,
and
also
the
folks
from
VMware
I
think
they
say
that
this
is
one
in
a
thousand.
So
what
we
can
link
from
our
dogs?
We
can
link
to
the
HCD
dogs
that
say
that
you
should
add
a
new
member.
Only
after
the
previous
member
is
healthy,
which
is
you
know,
implying,
sequential,
join.
A
D
A
B
G
B
A
A
F
A
A
A
A
Like
I
caused
this
one
name
of
the
road
CA
certificate,
publisher,
comptroller,
it's
still
office,
not
clear
graduate
boost.
Our
talk
is
to
GA.
We
are
like
unofficially,
graduating
G
tokens
to
GA
because
they
essentially
a
secret
that
has
been
GA
for
a
long
time.
Our
tools
are
not
going
to
graduate
to
anything
because
we
are
going
to
keep
our
tools
in
custom
structures
inside
the
comedian.
A
Repository
/
code
base.
Put
stop
talking,
so
visually
are
now
like
GA,
simplified,
cube
ad
mm.
Oh
I
closed
this
one
because
it's
we
already
have
this.
We
can
only
improve
at
this
point.
Confirmation
of
the
version
based
on
get
tax,
for
instance,
si
si
compliance.
I
chose
this
because
I
don't
see
anything
action
upon
our
site.
If
somebody
wants
to,
you
know,
provide
more
detail
on
CIS
compliance,
they
should
state
all
the
details.
This
was
just
a
ticket.
Hey
please,
provide
CS
c.
Is
compliance
across
this
one?
A
G
A
A
B
H
A
A
B
H
B
H
Is
one
of
BR,
but
so
I'm
going
to
I
talked
to
Ralph
a
bit
about
this
I'm
I'm,
going
to
make
it
into
three
PRS,
I,
think
and
but
there
it
there
in
a
way
where
it's
like
you
have
this
one,
and
then
this
one
and
then
this
one.
So
what
I
think
it's
going
to
be
is
that
each
PR
is
going
to
be
a
commit,
and
so
this
PR
will
have
all
three
commits
in
it
and
each
one
will
reference
each
other
I
think
because
they
kind
of
build
off
of
each
other
yeah.
B
G
B
Really
quickly,
we
discuss
last
week
to
upgrade
to
the
cap
about
removal
of
castor
cattle.
Adding
there
are
knows
that
cluster
status
is
following
the
beta
deprecation
policy,
so
I
had
I
added
this
node
completed
as
addresses
some
other
pending
comments.
So
please
take
a
look
and
possibly
ng
M.
Yes,.