►
From YouTube: kubeadm office hours 2020-04-29
B
B
B
D
B
B
Think
that
probably
we
should
not
take
on
that
path,
simply
because
the
semantics
are
a
little
bit
different
as
I
actually
explained
somewhere
in
the
comments
here
like
the
topic,
semantics
is
basically
like
you
provided
qadian
configuration
and
the
in
it
case.
You
also
can
provide
I
think
in
a
great
case.
You
can
also
provide
component
context.
Those
will
get
loaded
the
file
and
they
will
get
uploaded
and
all
of
the
contents
of
that
file
will
do
all
it
and
used
in
this
new
plug
that
I
proposed
here.
B
It's
basically,
you
can
have
practically
everything
inside,
but
it
will
only
pick
the
component
computers
that
need
upgrading
that
will
basically
allow
you
to
use
the
defaulting
command
and
all
the
component
convicts
using
the
current
settings
of
the
cluster
and
then
just
pick
the
ones
that
are
that
they're
needed
and
you
don't
need
to
just
go
and
specify
each
group
separately.
In
addition
to
that,
you
can
also
specify
a
single
config
that
can
be
used
with,
say,
cube,
ATM
types
in
inside
and
component
context,
and
you
can
use
this
to
instantiate
new
quarters.
D
B
D
D
Because
at
the
end
we
have
to
manage
this
insta
specific
components
of
so
I'm
kind
of
trying
to
see
in
the
middle
of
this
confusion,
if
we
should
expect
hooba,
proxy
or
corporate
to
cut
early,
a
new
release
of
their
component
confit
before
the
instant
specific
things
or
after
a
big
question,
because
he
is
basically,
there
is
an
agreement
that
for
Cooper
Adam
for
Cooper
proxy.
The
next
component
County
will
support
into
specific
things.
Probably
we
should
put
these
on
nolza
and
wait
and
wait
for
understand
what
component
of
instant
specific,
really
easy.
B
I
get
your
point
so
I
need
to
check
on
doing
specific,
kept
in
its
latest
state,
so
but
I
was
initially
pushing
there
to
make
it
more
relaxed.
So
at
least
qadian
won't
need
to
be
changed
simply
because
we
actually
have
types
that
are
instance
specific
and
those
are
the
yet
configuration
in
configuration
types.
Those
are
strictly
install
specific
and
only
the
coastal
configuration
is
shaken.
A
What
question
that
we
have
to
ask
ourselves
for
the
future
is:
how
is
this
fork
going
to
relate
to
potential
upgrades
of
instance,
specific
configure?
Do
we
need
a
separate
fork?
How
do
we
manage
that?
Do
even
encourage
the
user.
You
know
do
even
preach
to
the
user,
a
message
that
their
instance
specific
configured
disk
is
outdated.
How
do
we
manage
that.
B
Probably
should
be
upgraded
by
specifying
like
either
the
same
file
or
allowing
for
multiple
instances
of
this
common
client
won't
be
specified,
or
at
least
again
that's
what
I'm
pushing
for
to
be
able
to
just
specify
instance
and
shared
specific
and
shared
context
tool
in
the
same
file,
but
basically
not
force
people
to
use
separate
files,
but
allow
for
them
to
use.
It
simple
like,
on
the
other
hand,.
B
Yeah
I
think
that
the
context
of
that
discussion
was
whether
she
should
use
instance
versus
shared
context
like
with
different
facts
or
not,
and
the
idea
here
is
the
the
new
introduces
a
slightly
more
different
semantics.
So
if
you
specify
a
file
with
same
five
different
component
configurations
inside
and
only
one
of
them
needs
upgrading,
it
becomes
configuration
that
needs
a
great.
A
B
B
Inconvenience
there
would
be
to
basically
users
would
be,
for
example,
the
default
command
and
specify
the
exact
component
of
the
groups
that
they
want,
put
them
to
be
generated
by
cubed,
M
stores
in
the
file,
and
then
this
file
to
be
consumed
with
config,
and
also
that
way
we
should
add
support
to
config.
If
no
cube
alien
types
are
supplied
with
it,
it
should
go
in
downloading
from
coaster,
and
currently
this
is
not
the
case,
and
we
expect
everything
to
be
in
a
pod.
A
Yeah,
so
if
we
do
what
to
duplicate
config
because
of
reasons
I,
think
that,
because
kind
of
sue
make
sense
on
all
the
users
pass
pass
configuration
for
a
great
I
just
think
that
it's
really
wrong
to
allow
them
to
modify
init
configuration
and
coaster
configuration
using
does
config
on
upgrade.
It's
like
it
can
break
things
pretty
bad,
but
for
corporate
config
Adams,
it
kind
of
makes
it's
better.
I
guess
I.
B
D
B
A
This
is
the
related
topic,
the
Cape
that
I
had
if
we
I
think
it's
appropriate
to
introduce
the
command
under
alpha
for
printing.
What
is
upgradeable
but
I
think
that
maybe
we
should
call
it
just
naming
cube.
Adm
Alpha
config
was
like
that,
because
we
I
think
there
has
to
be
a
match
between
what
is
alpha
and
what
graduates
you
know
in
terms
of
where
does
this
command
is
going
to
go?
Was
graduates
correctly?
We
have
comedian.
Config
command
should
just
graduate
this
over
Comanche.
There.
B
My
only
like
concern
here
is
that
we
should
specify
somewhere
that,
like
in
the
quant
naming
that
this
is
actually
about
component
convicts,
so
probably
config
space
component
space
and
then
bring
the
blades
and
stuff
like
that,
simply
because
they're,
like
yeah
exactly
like
here,
we
have
a
couple
of
slightly
different
semantics.
All
these
mumbo
Jumbo's
that
we
actually
do
is
to
basically
tackle
the
problem
where
we
actually
cannot
upgrade
ourselves
component
convicts,
and
yet
we
actually
have
here
a
c'mon.
That's
called
cubm
Copic
migrate.
B
A
Why
we
don't
we
so
yeah
I
think
that
comedian
config
migrate
should
error
out
if
you,
so,
if
you
pass
it
component,
configs
right.
If
you
pass
corporate
context
in
your
config
on
disk
and
pass
this
to
migrate,
shoulda
just
have
a
route
and
say:
hey:
we
don't
support
migration
of
corporate
coffee.
I,
don't
think.
A
D
D
A
A
D
Means
that
we
can
start
planning
what
what
are
they
impacting
in
incubate
mean
on
his
instance
Pacific
slay,
okay.
So
that
means
that,
for
instance,
we
want
to
join
to
acceptance
for
specific
types.
We
want
to
store
these
in
specific
types
somewhere
to
being
able
to
do
upgrades
and
so
on
and
so
forth.
So
I
think
that
it
is
probably
worth
I,
don't
ever
a
good
feeling
of
what
is
Amanda
material.
B
Here
we
can
then
probably
hold
on
the
you
exchanges,
but
at
least
what
we
should
do
is
probably
start
cycling,
the
config
Maps,
and
this
would
basically
the
earlier.
We
actually
do
this,
the
more.
We
are
going
to
be
prepared
with
existing
posters
that
up
sign
config
Maps
whenever
we
actually
have
to
deal
with
upgrading
automatically
generated
by
Q
Balian
component
to
fix.
D
A
D
A
B
Well,
the
thing
is
that
we
can
probably
get
away
with
cubelet
itself
and
probably
some
of
the
components
that
are
actually
back
by
static
odds,
simply
because
we
actually
need
cube
ATM
to
be
executed
for
that
commitment
to
be
seen
the
file.
That's
actually
on
this
came
from
assumed
by
the
people,
but
the
problem
is
that,
like
with
atoms,
for
example
with
cube
proxy-
and
we
actually
don't
have
current
any
versions
to
fix
or
anything
of,
this
sort
proxies
complete
map
name,
and
this
can
actually
cause
problems.
B
D
Everything
goes
right
and
but
I
I
don't
think
that
this
should
be
up
to
the
users
or
to
the
consumer,
because
there's
a
common
program-
and
this
is
a
problem
that
applies
mostly
on
our
carpet
environment,
where
you
can't
images
and
you
rely
on
the
images
that
are
naked
into
the
your
machines.
So
these
are
the
most
tricky
because
if
you
you
have
to
basically
to
being
able
to
match
the
version
of
the
machine
with
the
version
of
Courtney
in
Russia
than
me.
D
D
A
D
Component
configure
release
that
that
we
must
we
we
have
to
expect
it
will
be
a
version
which
is
not
instance
specific
or
a
version
that
is
and
isn't
mr.
speciality,
so
Cuba
proxy
pork
cutlet
are
planning
to
cut
a
release
in
the
old
way
or
they
are
waiting
for
the
new
cap
for
capping
for
cutting
a
new
revision
anyway.
This
is
if
they
are
waiting
this.
This
change
is
not
required
because,
basically,
we
have
to
to
make
two
changes
in
the
same.
B
Probably
we
should
ask
my
dolphin
whether
he
has
dislike
will
high
priority
on
his
agenda
to
basically
roll
down
on
you,
cubed
conflict
with
instance,
and
share
specific
kinds
in
it
for
the
queue
proxy
config
I
have
a
kept
merged
waiting
last
summer
about
it's
being
upgraded.
This
cap
is
currently
implemented
and,
like
really
old
and
that's
simple,
because
the
first
outfit
duration
was
kind
of
locked
and
I
decided
to
basically
wait
for
the
instances.
B
D
A
B
Yeah
I
think
at
least
I
make
you
outside
their
fields
that
are
currently
in
the
cubicle
crash.
Would
we
move
to
an
instant
specific
kind?
There
are
also
other
fields
that
are
not
in
the
code
configuration
routines.
That's
fixed
all
I
think
that
this
is
probably
the
thing
that
expect
get.
The
new
configuration
instance
and
shapes
can.
A
A
B
A
A
A
B
A
Did
the
table?
Is
this
yeah
the
question
he
was
doing
here
if
we
are
going
to
track
also
the
kinds
and
version
separately,
because
we
know
with
this
table?
We
are
already
following
the
group
and
the
version
and
we
don't
like,
should
we
treat
the
whole
corporate
conflict
kind
bundle
or
should
we
treat
individual
kinds
so.
B
Basically,
a
kind
by
if
a
component
of
it
has
a
single
kind
and
the
kind
and
motion
sort
of
like
much
out,
but
if
a
component
computer
actually
gets
several
kinds,
as
in
the
case
with
having
an
instance
and
share
specific
conflict,
then
the
idea
here
is
to
base
clear
up
those
kinds
and
just
aren't
the
muscle
sink
overdue.
So
if
that
companies
actually
upgraded-
and
it
can,
for
example,
change
the
kind
names
or
change
the
number
of
kinds
in
there,
we
don't
get
these
sort
of
these
objects.
B
This
was
basically
what
I
actually
was
fixing
with
the
first
yeah,
but
got
merged
in
a
cycle
or
two
ago,
which
was
basically
reversing
the
way
we
actually
look
at
component
config.
So
previously
we
looked
at
component
critics
basically
having
their
kind
and
right
now
we
are
looking
at
them
by
starting
from
the
API
group.
So
a
component
config,
a
components:
config
T
space,
key
identified
by
the
API
group.
B
A
A
A
A
Actually
it's
the
corporate
integration,
so
this
is
a
different
patient
I'm,
making
a
mistake:
okay,
well,
yeah,
I,
definitely
remember.
Look
is
saying
that
implementation
detail
should
be
here,
I'm
kind
of
leaning
towards,
like
you
are
saying,
to
move
this
at
this
point
to
the
cube
ADM
report,
except
that
we
like,
should
we
just
placed
in
a
top-level
here
and
also
how
do
we?
How
do
we
say
that
these
are
implementation
details
for
version
X
or
how
do
we
version
the
implementation
details?
B
A
A
B
A
A
B
A
C
A
This
is
a
little
CPR
mighty
564.
It
touches
cue
medium.
We
found
out
that
the
people
can
pass
arguments
to
a
component
such
as
they
pass
silver.
Let
me
give
you
an
example:
actually,
this
window
I
cannot
share
it,
so
I
go
to
top
it
in
the
browser.
I
guess
so.
If
they
did
something
like
that,
cube
API
server
does
does
version.
This
is
going
to
work,
but
if
they
do
something
like
this
without
the
flag
prefix,
this
is
not
going
to
cause
any
error.
A
Reality
just
want
to
pass
without
a
warning
without
an
error,
so
they
basically
people
pass
something
like
the
Unicode
dashes
here.
Instead
of
the
you
know,
duck
show
and
see
dashes
this
combat
is
not
going
to
work.
It
is
going
to
be
ignored,
so
people
can
pass
our
back
around
car
back
rules
and
stuff
like
that.
Basically,
we
found
that
this
yesterday
payments-
and
this
is
the
pair
to
fix
all
comparison
key
medium
I-
think
that
the
discussion
here
is
that
whether
we
should
just.
A
A
A
A
A
Instead
of
removing
them,
so
this
change
I
like
this
very
much
we
did.
We
did
this
for
the
resolve.
Confed
the
secret
driver,
but
I
requested
action
required
to
be
added
to
both
peers
problem
here
is
that
users
should
have
visibility
for
this
change
as
soon
as
possible,
because
if
you
know
the
couplet
removes
the
flags
in
120,
they
have
been
deprecated
for
eight
releases
or
something
like
that.
So
the
couplet
removes
the
flags
tomorrow.
We
have
to
we
shouldn't
get
in
issues,
hopefully
for
that
incubating
I.