►
From YouTube: kubeadm office hours 2020-05-13
A
B
B
B
Let's
see
the
arrow
of
what
this
broken
city
code,
rubella
is
blocked
by
flex
volumes.
Part
6
storage
knows
more
about
this
I.
Think.
If
you
jump
into
this
link,
you
can
follow
the
rabbit
hole
of
PRS
and
issues
that
they
explain.
The
problem
I
think
it's
very
important
for
us
that
we
manage
to
switch
to
not
using
bean
/
sleep
during
upgrade.
B
C
B
B
C
C
B
B
B
D
B
Basically,
wrapped
both
the
client,
dial
and
memberid
in
a
exploration
back
off,
so
they
both
free,
try
now.
But
we
are
only
doing
this
for
our
member
with
the
idea
that
we
have
to
backwards
the
change
to
a
version,
question
I,
guess
here
for
Fabricio:
can
we
avoid
the
back
port?
It
was
16.
Do
we
care
only
about
117
and
118.
B
C
Don't
see
these
so
the
they
are
fixing
the
problem
that
user
are
facing,
and
so
they
are
bug
fixes,
even
if
they
are
bug
fixes
for
use
case.
That
happens
only
under
particular
circumstances
that
are
slow,
etcd,
basically,
but
but
at
the
end,
ER
bug
fixing
so
I
don't
see
a
big
problem
being
bit
of
recording.
B
B
Okay
to
close
the
issue
after
the
Bucksport
server
reached,
this
is
good
to
go
I
think
so.
This
is
a
kept
that
I'm,
not
sure
people
saw
so
quality
eyes
of
communities
have
been
stuck
in
bit,
some
of
them
for
a
very
long
time.
People
just
abandoned
them
because
you
know
they
implement
some
tanning
product
in
upstream
communities
and
they
decide.
Okay,
we
have
our
products
now,
let's
about
the
lobster
work
completely
and
a
lot
of
AP
is
stuck
in
out
for
a
bit
upstream.
So
David
EADS
decided
to
make
a
slight
change
here.
B
So
basically,
this
is
like
you
know,
like
a
stick
from
the
core
maintainers
to
kick
out
ApS
that
are
still
and
I
believe
this
with
a
key
period
discussion
that
I
think
we
with
our
API.
So
we
have
to
I,
don't
know
what
is
that
there?
Maybe
we
should
be
slightly
braver,
introducing
incremental
API
such
as.
Do
you
want
me
the
tree,
even
if
it's
not
doing
that?
Many
changes
no
increments.
So
this
is
basically
why
I
brought
this
into
this
discussion.
B
C
Got
a
point:
I
came
today
to
be
honest,
I
agree
in
in
in
what
they
are
doing,
because
it
is
not
not
fair
that
someone
interesting
implements
have
something
and
then
about
Donder
the
feature
in
some
intermediate
stage.
We
are,
we
are
doing
the
some
things
we
propose
it,
something
similar
in
copy
for
the
experimental
feature
of
the
copy,
experimental
features
and
going
back
to
cubed
mean
API.
I
I
got
your
point
to
be
honest:
I
I'm,
I'm
kind
of.
C
Holding
on
this
topic
there,
sorry,
because
we
are,
depending
from
the
community
I
at
the
end,
is
really
dependent
on
the
component,
configure.
Fourth
on
the
idea
of
a
beam
component,
config
per
instance,
only
on
DD
of
supporting
upgrade
of
component
company
and
I
don't
want
to
push
the
already
I
and
then
being
forced
to
work
as
soon
as
this
trait
gave
to
ever
some
progress.
So
this
is
why
I'm
holding
on
on
coupon
mean
personally
I'm
trying
to
postpone
any
change
to
the
coupon
mini
API.
C
C
We
have
to
raise
this
point
in
the
seeker
meeting
and-
and
we
have
to
take
a
decision
at
sea
level
because
now
changing
the
KU
banben
API
is
not
a
matter
of
Cuban
mean
what
is
the
matter
of
Cuban
mean
and
the
ecosystem
of
tools
that
fails
on
on
top
of
common
mean.
So
if
we,
if
we
have
to
be
braver,
I
I
will
like
the
support
of
the
entire
Sigma
I.
E
Also
agree
with
what
Fabrizio
said
and
talked
to
that.
There
is
also
the
advanced
country
cap,
which
you
may
actually
proposed
these
days
and
I
see
that
the
departure
idea
there
can
also
benefit
from
integration
with
the
component
config
of
qadian.
So
probably,
we
should
wait
put
that
to
merchant
to
implement
the
piston
of
abortion
using
common.
Finally,
and
then
think
of
a
good
integration
with
component
in
Kasbah.
B
Yeah,
it's
really
hard
to
find
the
balance
and
the
appropriate
time
to
introduce
a
new
API
person.
At
this
point
we
are
waiting
on
too
many
factors
to
converge
to
the
picture,
so
we
just
have
to
find
the
right
time,
but
also
subject
to
note
here
is
that
release
cycles
are
becoming
longer.
I.
Think
signal
is
pushing
for
a
four
month
release
cycle.
B
B
So
if
we
don't
release
a
new
API
dispersion,
we
have
to
wait
four
months
for
a
new
version
to
be
released,
and
then,
in
this
period
we
have
to
take
whether
corporate
config
is
going
to
change
it.
Subway
or
if
you
know
the
other
tools
would
have
different
demands.
The
Questor
api
or
maybe
I,
don't
know
cube
spray
or
something.
B
B
C
B
B
Basically,
it's
based
completely
of
a
British
camp
about
customize
and
after
discussions
with
Tim
and
the
wider
seek
we
kind
of
decided
to
move
away
from
Chris
rights
and
try
with
and
all
patches,
but
now
the
question
is
like:
how
do
we
manage
their
own
patches
and
so
forth?
First
of
all,
the
naming
this
is
the
first
major
covet
I.
Maybe
we
should
go
with
experimental
patches.
E
B
Yeah
so
I
have
an
idea
here
we
can
introduce
the
the
feature
with
the
directory
and
continue
supporting
it,
but
also
as
part
of
the
alpha
to
beta
graduation.
Maybe
I,
don't
know.
If
they're
outlined
this
here,
we
can
also
integrate
the
same
functionality
in
the
cube.
Atm
API,
like
you,
said,
in
the
registration
options.
Maybe
we
can
have
the
same.
I
can
I
can
design
the
backend
here
in
such
a
way
that
it
can
work
from
a
directory
or
also
from
the
API.
E
B
E
So
the
idea
is
to
basically
use
the
the
files
that
are
spiked
in
in
the
order
that
they
came
action.
So,
for
example,
we
can
basically
just
say:
experiment,
bastard,
strategic
and
then
a
strategic
patch,
then
say
experiment
of
experimental
patch
days
and
then
say
another
like
Jason
patch,
then
add
another,
for
example.
E
E
B
E
B
So
I
was
thinking
about
creating
something
that
is
called
a
pod
set
at
the
you
know
at
the
back
inside
and
parse
it
can
handle
both
files
that
are
multi
dock
and
also
a
list
of
files.
Now
a
list
of
paths
set
and
then,
for
example,
in
the
Kuban
MAPI,
we
can
have
a
field
somewhere.
Eventually
that
is
called
patches
and
the
patches
is
going
to
be
a
list
of
patch
sets
that
have
a
type
and
some
go
up
of
data.
B
There
is
a
multi
doc
potentially
or
we
can,
you
know,
think
of
something
else,
but
the
order
in
the
path
set
is
also
going
to
become
trouble
there.
So
eventually
we
can
completely
remove
this
flying
as
well
and
say:
hey
you
can
use.
Only
that,
like
you
know,
like
the
couplet,
is
removal
always
Flags.
You
can
say
you
can
use
that
only
the
QPD
mapi
either
personal
folder
or
this
list
of
files.
So
this
is
like
a
very
very
like
overview,
long-term
idea
here.
E
E
E
B
Yeah
so
Kara
concern
here
that
if
we,
if
we
add
this,
we
are
adding
a
new,
very
policy
competitive
feature
to
the
old
one
uses
users
might
get
confused
and
also
we
are
having
a
disagreement.
Unity
of
the
customized
feature
so
I
personally
again,
going
back
to
the
topic
of
moving
forward.
I
think
with
medium
I
think
that
we
should,
potentially
you
know,
make
more
decisions.
Even
if
we
add
a
bunch
of
stuff
remove
a
bunch
of
stuff
they're
kind
of
protected
by
the
whole
experimenter
offered.
B
C
Iii
I'm,
okay
in
changing
my
concern,
is
to
properly
communicate
that
we
are
changing.
So
we
have
to
send
an
email
to
the
list.
We
have
to
write
in
the
channels.
We
have
to
prepare
an
action
required
for
the
part
of
you,
but
I
I'm,
okay
in
changing
and
experimental
features.
So
it's
just
a
matter
to
properly
communicate
it.
We
we
are
back
at
by
the
contract.
This
is
experimental
and
experimental
can
be,
can
be
changed
in
this
case.
The
adoption
is
fairly
low
as
far
as
I
know.
C
C
Would
like
to
add
a
flag:
I
have
no
no
big
preference,
which
the
name,
but
probably
we
should
write
in
the
proposal
that
the
freighter
will
be
replaced
by
a
config
options
when
it's
just--come.
Coming
to
my
mind,
a
question
with
regards
to
compatibility,
so
we
are
saying
that
we
cannot
change
an
API,
but
water
annik,
the
current
API
type,
but
what
about
adding
the
types
to
v1
fun
with
too
long
beta
yeah
so
still
consider
a
break
and
change,
adding
a
type
angel.
B
So
I
a
supported
time.
I
read
the
official
guide
about
writing
recipe
ice,
which
is
basically
what
the
kubernetes
api
and
Kopernik
configure
based
on.
They
are
following
the
principles
of
res
api.
The
problem
there
is
that
they,
like
you,
say
it's
fine
to
introduce
any
field.
It's
not
a
breaking
change.
B
Removing
a
futile
or
renaming
a
field
has
to
employ
a
new
version,
but
adding
fields
is
fine.
However,
I
I
really
don't
like
if
we
start
doing
that,
because
we
have
a
pretty
good
understanding
as
a
developers,
what
we
added
when
we
added
it,
and
also
it
tells
the
users
that
you
know
that
follow
advanced
it
users
that
follow
closely
the
quick
baby
MAPI.
B
They
know
when
was
this
introduced
if
we
introduce
a
new
field,
suddenly
like
I'm,
going
to
forget
when
we
added
that
and
also
the
users,
maybe
going
to
be
confused
when
we
added
that
as
well.
So
this
is
one
of
the
reasons
I
think
our
API,
you
know,
quote-unquote
set
in
stone
after
release
is
maybe
the
better
approach.
E
Yeah
also
that
actually
went
to
the
component
context
of
like
the
existing
components
and
basically
added
serializers,
and
we
actually
have
to
go
and
make
those
just
a
warning
sticks
in
your
visors
for
now,
because
people
are
quite
used
to
adding
new
fields
to
their
context,
and
basically
you
can
basically
end
up
with.
For
example,
we
wanted
one
version
that
is
not
compatible
with
the
older
component
version,
because
it
has
a
new
field
in
the
older
component.
E
C
C
And
so
my
point
is
I
understand:
what
are
you
seeing
I
understand
the
problem
strikes
already
here
and
so
and
so
forth,
but
it
seems
that
in
the
kubernetes
name
this
is
a
common
practice.
So,
in
light
of
luminaire
idea
that
we
have
to
keep
moving,
I
am
just
wondering
if
we
are
being
too
strict
here.
B
Yeah,
it's
a
it's
a
trade
of
moving
faster
by
adding
fields
or
sacrificing
this
moving
faster
aspect,
because
we
want
to
have
better
control
and
better
understanding
for
the
users.
What
is
added
and
where
it's
added
also
take
about
dogs.
We
have
to
update
dogs
now.
Suddenly
a
new
field
appears.
We
might
get
questions
more
questions
when
when
was
decided,
how
is
it
supported?
We
have
a
I
think
it's
going
to.
What
we
have
right
now
is
easier
for
us
to
coordinate
with
the
users.
C
C
Or
Justin,
so
we
have
an
agreement
at
sea
level
and
saying
the
end
state
somewhere
that
we
consider
fine
to
do
so,
and
we
suggest
to
document
this
change
because
because
it
does
therapy
I,
they
are
going
to
do
this
and
probably
other
in
needs.
They
are
doing
these
and
other
projects.
So
I
think
that
as
we
have,
we
have
a
kind
of
policy
at
sea
level,
Britain
somewhere
and
then
okay,
in
fact,
that
we
don't
have
to
abuse
that
we
have
to
use
with
judgment.
B
C
C
B
E
Probably
just
sort
of
extension
there
Jason
yeah
actually
would
work.
E
B
B
B
C
G
There,
okay,
yeah
I,
just
might
be
welcoming
a
yeast
case
around
how
cost
IP
I
could
consume
it
without
having
to
pass
in
the
files,
even
if
you
don't
do
it
in
the
first
pass
like
I,
see.
B
So
the
yeah,
the
first
pass
of
this,
the
experimental
alpha
stage,
will
require
a
folder.
If
we
move
to
a
new
you
most
of
the
cube,
a
DM
API
Costa
Rica
is
going
to
be
able
to.
But
this
means
that
question
today
has
to
move
to
support.
The
new
version
of
the
Cabana
API
branches
is
going
to
be
very
tricky,
I
think
I.
G
B
A
B
G
It
can
I
mean
we
can
right
now.
People
can
create
their
files,
pass
them
in
wire
cloud
in
it
as
extra
files
etc.
G
But
today,
from
a
casa,
API
perspective,
that
would
be
the
best
UX.
But
there's
some
discussion
about
Howard
Lincoln
seem
cloud
in
it
going
forwards
as
well
say:
yeah
I,
don't
know,
I,
think
I
think
I'm
just
make
sure
people
get
some
ice
on
it
from
the
cluster
API
community
and
see.
If
there's
comment.
B
Sheik
I
mean
I,
haven't
said:
I
think
that
I
didn't
send
a
message
to
the
mailing
list,
but
I
think
I
said
PR.
Yeah
I
picked
this.
This
is
probably
fleet.
Jason
saw
this
and
I
think
this,
but
you
know
all
responses.
I
I
really
think
we
should
proceed
this
way
for
this
cycle,
even
if,
in
this
alpha
state
the
way
it
is
with
a
folder,
maybe.
B
B
The
Cape
is
here
okay,
so
we
have
seventy
base,
so
the
open,
RC
people
have
been
bugging
us,
sorry
for
using
the
expression,
but
they
have
been
asking
for
us
to
fix,
open
RC
support.
We
don't
really
have
that
many
users,
first
of
all,
opener
C,
is
used
on
Alpine
correctly
is
broken
for
the
fact
that
cube
ADM
creates
sorry
Q
medium
starts
the
couplet.
An
alleged
crash
whoop,
then
writes
the
static
pots
to
disk
and
the
couplet
configuration
to
disk
and
the
crash
looping
system.
B
This
service
is
supposed
to
pick
the
so
this
can
start
the
cool
it
properly.
The
problem
with
that
is
that
opera
RC
apparently
does
not
support
automatic
results
of
services.
There's
no
way
to
tell
opener
C
to
even
I,
don't
know
how
to
programmatically
restart
the
service
within
the
spec
of
the
service.
You
can
only
execute
it
on
the
command
line
and
say
hey
restart
the
service
for
me.
So
we
had
a
discussion.
B
B
Exactly
so,
this
is
a
problem
with
repellency.
They
not
supporting
research
is
very
problematic,
something
that
I
also
propose
here
it.
So
if
the
first
point
Rossi
III,
don't
know
how
we
cannot
do
that
in
the
short
term.
I
guess
for
119
I,
don't
think
we
should
be
changing
the
spec
for
system
yeah
I.
Don't
think
that
that
we.
E
Should
be
doing
that
at
least
it
means
a
rejection
required
item
somewhere,
something
costs.
It
may
surprise
some.
So,
for
example,
in
our
case,
we
need
to
talk
face
here,
so
we
can
actually
just
enable
and
start
the
cupid
cells,
and
we
also
need
to
base
because
we
currently
only
start
or
restart
them.
We
also
need
to
enable.
B
E
B
B
E
E
While
we
sleep
at
this
somewhere,
but
not
in
the
cupboard
start,
the
Qi
start
phase
is
basically
the
last
phase
here.
Just
expand
this
unit
and
face
section
here.
The
original
source
I
think
you
can
see
that
that
keep
that
stuff
face
is
basically
the
last
thing
there.
It's
actually
a
few
things
after
the
wait
for
Qibla.
So
it's
key.
We
cannot
discover
things
well.
B
B
B
E
E
So
this
is
the
expected
behavior
currently,
and
the
other
thing
is
that
if
you
meet
any
Q
beta
incorporation
from
there
and
supply
just
some
component
Coptics,
it's
going
to
use
only
those
two
politics
and
override
ease
in
the
course.
It's
not
going
to
generate
new
config
so
generating
types,
but
it's
sort
of
strangely
working
code.
You
expect.
B
D
E
I
was
intentional
to
everything
that's
working
correctly.
With
the
same
so
as
long
as
cubm
types,
everything
is
going
to
stay
the
same,
but
if
you
meet
the
cuban
types
and
just
supply
component
come
fix
it
there.
It's
not
going
to
generate
a
billion
times
in
that
file.
Also
I
changed
a
little
bit
basically
integrated.
The
proposed
a
couple
of
alpha
commands
with
the
variant
config
command,
sub
T
they're,
still
under
the
name,
alpha
config,
but
won't
be
just
straight
away:
integrated
thing
to
2
billion
config
once
this
actually
reaches
bed.
E
B
E
Now
I
just
tackled
the
feedback
from
the
from
a
couple
of
weeks
ago,
always
coaches,
the
config
file
exchanged,
but
I
had
to
do
some
sort
of
like
a
weird
this
because
of
we
actually
using
it
and
just
started
a
few
users,
because
I
think
it
was
forbidden.
This
is
just
fast
and
it
might
be
a
good
idea
to
add
a
few
user
stories
here
and
also
change
the
names
alpha
commands.
E
B
Have
a
look
at
this
today
and
I'm
going
to
I?
Think
it's
pretty
good,
so
I
were
to
approve
it
and
let
the
edge
be
up
to
somebody
else
make
sure
it's
implementable.
We
can
potentially
balaji
next
week
before
Tuesday
I
guess.
The
same
applies
to
bite
yeah,
my
cap.
What
what
other
kids
nuke
on
frosty
I
don't
think.