►
From YouTube: kubeadm office hours 2020-04-15
A
A
Ok,
thank
you
for
joining
I
wanted
to
share
with
you
a
couple
of
documents,
actually
one
document
that
we
have
already
for
this,
and
it
should
be
here
so
the
document
is
outdated,
but
we
have
a
couple
of
up-to-date
videos.
So
first
you
have
to
read
this
document.
If
you
want
to
get
familiar
with
how
we
work
on
the
project
and
a
at
the
bottom,
we
have
a
couple
of
videos
that
more
recent
you
know
2019
2020.
B
A
A
A
A
So
the
release,
119
branch
was
going
to
be
created
somewhere
in
the
schedule,
and
cold
freeze
came
a
synchronously
from
that,
but
what
they
are
doing
now
is
they
want
to
create
the
branch
and
establish
cold
freeze
at
the
same
time,
which
means
that
all
book
fixes
that
have
to
land
after
code
freeze
have
to
be
pretty
much
back
ported
at
this
point.
Up
until
now,
we
use
the
fast
word
mechanic
which
seek
release.
Basically,
they
want
to
remove
because
it's
burden
for
the
releasing.
A
A
D
Yep
I
think
well,
he
made
a
couple
of
remarks,
most
of
which
I
just
integrated
into
the
document.
There's
one
thing
where
I
was
not
totally
sure
whether
that
was
a
good
idea.
That
is
actually
yes,
but
what
you
just
see
here.
So
it's
more
about
the
structure
I'm
happy
with
either
way
of
handling
this.
So
if
you
want
to
take
another
look
and
see
whether
you're
fine
with
that
or
you
want,
the
structure,
slightly
changed
I
think
it
was
about
where
to
put
the
disclaimer
and
where
to
put
the
introductional
text.
A
D
D
Of
contents
here,
how
is
he
all
right
in
that
case,
I
probably
didn't
understand
that
correctly.
Well,
I
mean
yes,
we
could
put
the
table
of
contents
above
the
above
the
overview
section
title.
That
is
fine
for
me,
it's
really
okay.
So
if
you
agree,
we
should
do
it
like
that.
Then
I
can
just
quickly
change
that
and
then
we
should
be
done.
D
A
A
D
A
All
right,
I
wanted
to
I
wanted
us
to
do
some
PR
initial
triage,
because
we
haven't
done
this
in
a
long
time
and
we
have
some
pending
peers
that
we
should
have
a
look
at.
So
let
me
open
the
peers.
E
A
The
hack
has
to
move
them
to
the
config.
Okay
I've
seen
this
PR
before
I'm,
actually
taking
about
sending
this
myself
a
psycho
because
I'm
seeing
some
movement
in
the
cupola,
the
flags
and
we
don't
wanna
end
up
in
a
state
where
we
release
key
medium
weights,
not
supporting
the
ladies
college,
so
I,
you
know,
I
might
take
this.
This
cycle.
E
Yeah
I
think
that
it's
in
a
multiple
state
and
as
soon
as
we
merge
it,
the
better
simple
because
they
aren't
going
to
be
much
more
coastal
score
going
to
cap,
their
can
fix
tact
with
Jackson,
and
this
will
basically
allow
in
the
future
like
a
lot
less
interaction
with
the
user
directly.
So,
basically,
a
user
would
not
be
required
to
go
and
change
manually,
cube
alien
generated,
config.
E
E
Probably
our
documentation
would
be
nice
but
I'm
not
exactly
sure,
probably
and
in
some
sort
of
implementation
details
page
but
on
the
website,
because
this
is
like
pretty
like
this
is
quite
low
level
and
I'm.
Thinking
on
providing
some
Alpha
commands
at
some
point
to
be
able
to
deal
with
this
in
a
more
automated
way.
E
So
currently,
the
idea
is
to
build
upon
this
with
the
great
farm
table,
which
is
the
ER
which
we
originally
began
to
look
at,
and
this
PR
is
simply
adding
a
table
to
upgrade
one
output
that
is
going
to
print
out
the
component
conviction
whether
they
are
tacked
and
whether
they
be
upgraded
or
not,
and
probably
in
the
future.
There
is
going
to
be
something
like
additional
months
to
just
dump
these
complex
that
mean
upgrading.
E
F
E
A
Okay,
I
think
we
should
add
our
editing
here.
In
fact,
we
should
constantly
consolidate
this
page
somewhere
in
here
so
I.
It's
I
think
it's
okay
to
just
create
a
new
document
here
and
maybe
later
we
can
refactor
things
and
maybe
remove
this
or
rather
not
remove
but
create
a
new
one
that
is
up
to
date,
because
some
of
these
are
very
old.
A
A
A
E
E
A
A
E
I
think
that
the
last
time
we
actually
discussed
this
one,
we
sort
of
reached
an
agreement
to
get
proposal
for
me
from
me
and
I.
Send
it
so
I'm
I
think
I'm
still
waiting
on
Rafael
see
review
there
and
I
think
I
think
that
he
mentioned
he's
going
to
close
this
PR
at
some
point.
But
probably
he
hasn't
got
change.
A
A
A
A
This
is
a
nice
change,
but
it's
brought
to
the
reviewers,
so
the
sorry
approvers
stop
calling
beansleep
this
needs.
This
pyrrha
needs
a
body
wear
and
also
proposal
how
we
are
going
to
approach
this
I
thought
it
should
discuss
now
it's
low
priority,
because
this
promise
is
still
not
happening,
but
we
have
to
look
at
it
at
some
point.
I.
A
E
A
I'm
actually
considering
garden
I
entered
this.
For
that,
because
kinda
supports
this
flexibility,
I
think
yeah
we
can.
We
can
skip
the
phases.
We
can
skip
all
those
completely
with
flags
for
now.
Something
that
we
see
me
for
this
we
have
seen
is
that
from
time
to
time
we
get
weird
is
forbidden,
error
types
because
so
far
back
and
it's
we
don't
have
an
explanation.
It's
very
weird
is
happening
when,
right
after
you,
when
you
upgrade
a
couplet
on
a
note
on
the
worker,
knows
or
something
like
that,
but
this
is
about
control
play
mode.
A
B
A
A
A
A
A
E
A
A
A
F
A
A
A
A
A
A
How
I
can
issue
an
KQ
by
DM
that
we
should
remove
our
use?
Api
can
flag
completely
from
certificate
renewal
and
I'm,
not
sure
if
the
certificate
API
is
going
to
graduate
to
the
one-day
cycle,
but
I
just
sent
the
PR
in
case
just
in
case
it
decides
to
graduate
the
change
is
prematurely
moving.
All
the
related
code
to
this
flag.
A
A
A
A
And
we
are
dull
with
the
peers.
We
can
have
a
look
at
the
issues,
how
much
time
we
have
20
minutes.
I
wanted
us
to
discuss
quickly
some
of
the
top
priorities
items
for
this
cycle
to
get
some
clarity
on
you
know,
decide
what
we
want
to
do
with
the
retries
and
stuff
like
that,
because
it
wasn't
very
clear
to
me:
I
wanted
to
help
it
one
of
the
PRS,
but
so,
let's
start
from
the
top,
and
they
then
go
through
these
that
at
the
critical
ones.
A
E
E
A
E
Yeah,
the
current
couplet
configuration
kind
in
Cupid's.
We
want
that.
One
compact
is
really
meant
to
be
shared
one
and
specifically,
it
doesn't
include
any
instance
options,
or
at
least
not
obvious
ones.
So
the
idea
here
is
that
we
actually
get
this
kept
merged,
and
once
this
is
in
place
and
tional
kind
can
be
other
talk,
that
is
going
to
house
instance
specific
options
and
there
is
probably
going
to
be
some
overlap
so
between
the
options,
and
yesterday
we
kind
of
agreed
on.
F
E
We
haven't
talked
in
any
specific
terms
with
like
specifically
about
the
operation.
Currently,
we
are
talking
about
getting
just
sort
of
an
abstract
cap
that
is
going
to
be
used
as
a
guideline
on
any
such
config.
So,
for
example,
the
cube
the
cube
every
component
has
a
component
compared
or
cap.
Such
is
probably
going
to
fade
same
issue
having
two
separate
instance
and
shared
specific
context,
so
currently
we're
at
this
stage
and
once
the
carriages
will
probably
think
about
specific
implementations.
F
A
Wanted
to
close
this
ticket,
but
I
realize
that
we
have
to
make
changes
anyway,
because
we
have
a
constant
that
we
use
the
sealer
has
to
be
renamed
to
default
collateral
directory,
and
we
have
to
stop
your
hard
coding
logic
to
it.
If
we
make
it
an
interesting
problem
here
is
that,
okay,
if
the
user
position
is
the
specific
coolant
configuration
to
convey,
am
either
doing
in
it
or
join,
do
we
read
the
value
from
the
couplet
config
and
then
perform
logic
incubator,
for
instance?
A
E
A
F
This
interesting
push,
but
I
think
that
this
is
a
generic
question,
so
a
mother
I
would
like
to
generalize.
The
problem
could
mean
in
order
to
give
you
a
working
cluster,
he
it
is
doing
some
configuration
at
cabela's
level
and
basically
to
all
the
component
levels.
So,
no
matter,
if
the
the
component
confid
story
we
ship
out,
we
have
to
read
the
conflict.
The
user
provide
and
add
could
mean
specific
conflicts
may
be
that
that
we
make
the
same
approach.
F
We
prevent
the
same
approach
that
we
are
doing
for
flex
today
and
basically
for
flex
the
if
a
user
is
specifying
the
bad
way
could
mean
don't
touch
this
bad
way
so,
but
at
the
end
we
have
to
integrate,
merge
or
patch
for
little
limit.
As
you
want,
the
configuration
for
for
cooperate
and
other
compounds.
A
F
This
one
is
the
easiest
one
to
get
around
so
here.
Basically,
it
is
something
that
we
have
to
check.
So
you
we
have
to
run
a
coupon,
mean
join
counter
plane
and
then
basically
arrive
on
the
face
in
order
to
see
if
the
face
with
it
can
be
executed
more
times.
Okay,
I'm,
pretty
sure
that
most
of
the
faces
are
already
already
in
the
independent,
because
it
what
happen
if
you
regenerate
the
manifest
nothing.
What
would
happen
if
you
were
generate
research?
F
A
F
A
A
F
A
B
F
Far
as
I
know,
the
back
of
basically
the
back
of
unless
you
put
the
back
of
our
you're,
not
you
are
not
defining
at
amount.
We
are
just
defining
the
staple
for
for
the
next
weight.
So
if
an
operation
is
immediate
at
the
Danny
wait,
if
an
operation
takes
time,
then
then
you
wait
for
the
same
amount
of
time.
F
B
E
A
F
F
So,
basically
the
what
are
you
are
printing
is
the
sum
of
your
weight,
assuming
that
your
operation
is
take
so
not
not
consider
all
the
time.
Okay,
but
if,
if
your
operation
takes
time
the
overall
operation
length
such
thumb
up,
okay
in
this
case
can
be
unfortunate.
If
every
loop
will
get
could
be
CD
a
failing
to
answer.
442Nd.
F
The
I
think
that
I
I
think
that
if
so,
the
goal
of
their
try
is
to
wrap
every
operation
that
can
fail
and
execute
them.
Okay,
so
the
connection
can
fail.
So
what
we
can
do
is
to
put
the
connection
into
there,
try
and
know
where
they
the
time
out.
So
if,
if
activity
is
not
answering
okay,
then
we
try
to
we
retry
in
few
seconds.
We
don't
know
five
seconds
dial,
timeout
I
guess
we
have.