►
From YouTube: kubeadm office hours 2020-02-12
A
A
We
have
a
survey
point
for
Keep,
Calm
and
I
wanted
to
maybe
get
some
feedback
from
the
group
in
terms
of
what
topics
you
want
to
cover
in
the
new
survey.
So
this
document
here
is
provides
the
initial
guideline
for
this
new
format
of
the
same
way,
the
TLDR
is
that
we're
going
to
mix
orders
for
projects
of
sequence,
lifecycle
that
want
to
send
questions
to
the
survey
in
the
same
form
for
cubanÃa,
specifically
I,
guess
we
have
to
think
about
unique
Canadian
questions
from
maybe
questions
that
we
want
to
recycle
from
the
own
survey.
A
The
format
is
going
to
be
such
that
we
also
have
a
general
question
section
and
basically,
in
here
we're
going
to
list
some
of
the
questions
that
are
important
to
the
whole
list
of
projects
and
I.
Think
the
coaster
API
already
have
some
questions
of
the
pretty
generic,
so
I
think
we
have
to
revisit
with
them.
What
is
really
unique
for
the
cost
for
the
coastal
pie
project?
We
have
the
same
for
cube
atm.
B
A
B
Yes,
I
had
some
question
about
I
availability
about
aircraft
pad
and
about
conformance
test,
because
this
was
one
of
the
question
the
former
questioner
and
also,
if
you
scroll
down
I,
guess
that
I
added
some
other
question
about
the
operative
system
about
mix
architecture.
Note
about
Sienna
se
RI
about
how.
B
Okay,
this
section
is,
is
a
section
about
specific
question
for
cupid
mean,
and
the
question
is:
are
there
either
from
the
former
survey
that
we
did
I
just
reproduced
some
some
some
question
in
order
to
try
to
get
basically
a
feedback
from
the
user
in
terms
of
what
is
important
from
for
there
for
the
user?
So
I
asked
the
user
plays
sort
by
importance
that
the
coracoid
Amin
features.
B
B
So
if
they're
using
Google
mean
they
are
using
other
nodes
or
if
they
go
for
remote
ability
and
then
I
I
try
to
understand
what
is
what
is
the
they
need?
What
are
the
need
for
the
user
for
the
change
the
cast
are
stalling
because,
in
my
opinion,
this
is
the
array
of
could
of
could
mean
that
that
we
have
for
dress
and
I
would
like
to
understand
what
are
the
priorities
for
the
users?
A
A
A
A
B
A
Yeah,
then
again,
we
cannot
really
cut
to
a
fixed
number.
We
have
to
evaluate
later.
If
we
just
have
too
many
questions,
I
think
the
Costa
Rica
currently
don't
have
that
many
questions
most
of
their
questions
were
fairly
generic
for
Cuba
DM
I
can
think
of
some
better
for
cuz
I
play.
For
instance,
we
can
have
the
same
like
we
can
have
the
same
like
like
this,
like
what
is
the
UX?
What
is
your
UX
experience
with
some
of
the
tools
that
in
the
coastal
epi
repository?
A
A
I'm
also
interested
in
the
window
support
if
this
is
going
to
the
percentage
of
people
are
going
to
increase.
Maybe
a
publicity
thing.
This
is,
for
the
generic
section
rate,
your
experience,
so
this
is
something
that
maybe
we
should
leave
out,
because
multiple
groups
of
kubernetes
decided
to
do
a
survey.
So
we
have
a
OTS
survey.
We
have
a
production
readiness
survey,
I
mean
now
we
are
going
to
have
a
coastal,
a
psycho
survey
and
all
of
them
ask
him
this
question
like
the
OTS.
They
want
the
audience.
A
I
think
we
should
leave
this
to
the
audience
group
if
because
they
want
to,
but
there
was
also
the
proposal
to
merge
the
service,
so
the
OTS
group
can
send
us
something
like
five
questions
and
I
said
immediately
like
hey.
This
is
not
really
appropriate
because
we're
going
to
talk
about
deployers
in
our
survey
and
but
still
if
they
want
to
add
some
LTS
questions,
maybe
we
can
try
to
approve
their
proposal.
C
So
I'll
also
try
to
to
see
the
existing
questions
and,
we'll
probably
add
one
or
two
more
or
contribute
to
the
existing
ones.
Mainly
I
want
to
see
some
questions
around.
What
do
people
think
on
based
on
feature
basis
of
qadian?
So
we
can.
We
can
have
more
accurate,
more
accurate
predictions
and
more
accurate
positioning
of
priorities
for
the
next
few
cycles.
A
A
D
One
thing:
I
don't
be
on
the
questions
that
I
think
the
breacher
was
writing
down
I
if
I
remember
correctly,
there
was
one
of
one
about
sorting
for
importance
of
the
component.
Is
that
right?
Something
like
that
and
to
me
that
sounded
like
a
question
that
can
be
interpreted
in
multiple
ways
right?
What
doesn't
mean
importance
for
different
people
can
mean
different
things,
so
I
think
we
should
be
careful
when
we
phrase
the
question.
D
D
C
D
D
To
five
what
you
just
say
that
which
is
interesting
because
I
that
wasn't
my
interpretation
of
the
of
the
question
right,
you
just
say
that
you
want
to
know
which
feature
we
should
put
more
effort
on
if
I
could
find
correct
right,
so
in
a
actually
imagery,
be
the
opposite
of
what
people
are
gonna
vote,
because
they
may
find
more
useful
or
more
important
feature,
but
they
may
think
that
a
feature
needs
more
attention.
Another
feature
needs
more
attention.
If
I
mean
no.
B
C
D
B
B
B
D
A
D
It's
just
the
last
comment:
there
I
want
to
be
sure
that
I'm
going
the
right
direction
before
starting
even
implementing
it.
This
is
about
the
cubm
config
image,
two
images
pool
and
the
the
person
that
opened
this
issue
was
reporting
there
in
this
situation,
when
the
config
is
already
in
the
cluster
and
you're
using
an
external
ABCD,
the
the
command
is
gonna
pull
down
the
etcd
images,
even
if
it's
not
needed,
because
you
are
using
external,
its
P
and
that's
because
apparently
the
command
right
now
is
no
reading.
D
I
mean
consume
that
we're
gonna
use
this
logic
in
two
commands
already
impossible
in
more
than
nothing
in
the
future.
There
is
a
plan
to
enable
more
commands
to
read
the
because
the
comfy
I
mean
we're
ready,
we're
gonna.
If
we
decide
to
go
that
way
and
I
think
we
decided
for
that.
We're
gonna
use
this
function
already
or
the
same
logic
that
same
two
places
already.
E
D
Also,
so
the
what
what
this
does
it
tries
to
see
if
you
are
passing
a
config
if
you're
not
passing
a
company's
gonna,
try
to
load
the
configuration
from
the
cluster
and
if
the
completion
of
the
cluster
is
not
there,
so
you
get
an
error.
It's
gonna
be
fold
back
into
the
normal
behavior.
That
is
loading
the
effect
called
loading,
the
inner
configuration
for
plaster
and
or
load
they
default
in
configuration.
D
C
D
There
was
a
comment
in
the
original
PR
for
this
change
here,
and
the
comment
was
about.
These
can
actually
bring,
as
we
said,
a
fete
in
case.
You
cannot
reach
the
cluster
if
you
guys
ever
to
fetch
the
convict
right.
If
you're
gonna
fall
in
today
did
the
path
of
loading
a
default
comfy.
So
in
case
you
tried
to
do
these
and
that's
they
did
I
think
they
were
skits
had
it
fed
that
you
can
happen.
In
this
case,
you
try
to
fetch
the
confusion
from
the
pastor.
D
You
feel
for
any
other
reason,
except
the
one
that
dig
the
cluster
is
not
there
right
and
you
end
up
loading
to
the
full
contact.
We
know
there
is
a
caching
of
doing
that,
but
since
we're
only
talking
about
the
image,
Pokemon
I
mean
it
all
these
gonna,
it's
probably
gonna,
mean
only
that
you're
gonna
put
more
images
than
what
you
think
it's
gonna
be
required.
So
I
agree
indeed,
on
the
each
comment:
oh
yeah.
A
My
main
objection
to
this
future
request
was
that
it
applies
only
to
the
cases
when
users
have
already
created
a
question
and
originally
the
command
for
comedian.
Config
images
pull
and
list
was
designed
for
impact
scenarios,
so
you
don't
have
a
cluster,
you
don't
have
a
question
it,
and
now
we
have
a
cluster.
B
I
I
think
that
we
see,
since
the
discussion
about
this
config
with
that
dysfunction
that
basically
works
at
the
best
effort,
is
really
specific
to
the
comm
to
the
command
I'm,
plus
one
for
duplicating
dislodging
and
into
into
two
common
and
giving
an
explanation
why
we
are
doing
this,
which
is
best
at
four
to
instead
of
moving
the
function
in
a
utility.
And
then
it
is
difficult
to
understand
why
we
are
working
and
at
the
state
for
constant
support.
But
I
arrived
no
strong
opinion.
D
In
that
case,
we
put
the
link
of
the
two
function
in
their
respective
fallacy.
This
is
people
give
it
here,
and
this
is
the
beginning,
the
other
places,
because
there
is
there
is
there,
someone
is
gonna,
apply
the
detached
a
patch
to
that
function
in
the
future
is
gonna
forget
about
the
other
one
or
the
other
way
around.
D
A
A
This
is
going
to
tell
queue
medium,
config
images
list
to
use
the
default
version
for
the
images
I'm,
not
the
version
distorting
the
question,
so
it's
using
the
version
of
the
Canadian
binary
and
not
the
version.
The
question
I
think
we
should
preserve
this
functionality
because
it's
actually
a
breaking
change.
If
we
change
the
assumption
here
so
then
it
becomes
a
question
like
wire
with
applying
the
other.
So
the
similar
assumption
to
the
config
for
pool
it's
I
was
really
against
the
original
feature.
Request
and
I
I.
A
A
D
And-
and
there
is
actually
a
comment
from
the
person
of
invention-
you
were
already
discussing
that
in
the
itself.
Where
is
actually
saying
that
is
more
important:
that
image
all
is
consistent
with
the
upgrade
and
the
other
command
setting
day
with
image,
then
with
image
list.
But
it
was
also
saying
that
that's
just
an
opinion
and
I
understand
your
point.
So
the.
A
D
A
A
A
You
can
always
pass
the
config
and
in
this
case,
cube
ADM
is
not
going
to
pull
the
HDD
image
because
you
tell
it:
hey:
I,
have
an
external
HDD
I,
don't
care
about
the
HDD
image,
but
the
user
wants
to
Modbus
config
and
use
that
configure
the
question.
I
am
really
starting
to
to
develop
an
objection
against
this
feature.
Request
I.
D
Can
see
the
I
can
see
the
use
case
if
you
managing
multiple
cluster,
you
may
not
know
which
cluster
has
the
external
need
be
or
Internet
is
be
so
having
the
command
to
do.
The
right
things
is
a
nice
bonus.
Let's
say
I
also
see
your
your
your
reasoning
about
the
list
and
the
the
food
that
should
behave
consistently
with
each
other
and
I
also
see
that
changing
this.
You
have
a
breaking
change
in
the
list
because
you
get
a
different
behaviour.
The
clusters
is
a
it's
very
slow
and
rejection.
B
B
A
A
A
A
A
D
Three
convict
is
here:
a
three
three
possible
source
of
configuration.
Writer
is
the
default
one
there's
the
County,
and
now
there
is
the
classic
County,
so
I
sort
of
agree
with
Obama.
Here
that
we
may
require
something.
That
is
a
little
bit
more
explicit
to
be
clear
on
the
behavior
of
the
command,
where
you
say
even
the
order
of
precedence.
How
do
you
want
to
apply
this
right
or
something
to
say,
I
want
to
use
the
internal
classic,
how
to
give
possible
I,
don't
know
something
like
that.
B
A
D
But
it's
best
airport,
so
that's
also
another
behavior
right.
So
if
you
cannot
reach
the
API
server,
you're
gonna
go
in
default,
so
even
there
may
be
tricky.
Is
that
with
an
extra
with
an
extra
parameter,
you
may
say:
I
want
the
one
in
Class
C.
If
it's
not
there,
you
either
fail
or
then
you
fall
back
on
something
so
I,
don't
know
what.
A
B
B
C
Known
as
far
as
we
don't
have
any
upgraded
component
config,
yet
we
won't
expect
up,
probably
well
the
thing
that
is
probably
going
to
cause
some
concerns
in
the
somewhat
near
to
meet
future
is
if
we
want
offer
two
of
queue
proxy
appears,
but
so
far
we
are
only
dealing
with
cubelets
in
queue.
Proxies
component,
configs
and
I
think
that
cubelets
probably
is
not
going
to
get
into
beta
2.
C
But
that
also
depends
into
how
the
instance
versus
shared
component
config
stuff
goes
because
I
think
that
Mike
is
trying
to
get
the
instance
specific
versus
shared
specific,
kept
mulch
thing
to
implement
to
miss
the
specific
component
config
for
the
cubelet.
Simply
because
there
are
large
number
of
common
fine
parameters
that
he
wants
to
put
in
a
component.
Config
and
those
are
instant,
specific.
B
C
We
need
to
check
with
API
machinery
policies,
but
I
think
that
probably
this
is
going
to
introduce
probably
a
new
kind
to
the
cubelets
component
config,
which
is
instance,
specific
kind,
and
this
is
probably
going
to
lead
into
a
new
version
being
released
for
it.
So
I
don't
think
that
you
can
retrofit
a
second
kind,
even
though
it
doesn't
collide
with
the
previous
configuration.
B
C
Anyone
in
the
complete
community
knows
right
now,
probably
mike
has
some
plan
I'm,
not
sure
whether
whether
they
are
actually
going
to
release
this
conviction
or
if
they
are
trying
to
avoid
releasing
it
by
simply
adding
another
kind
to
the
existing
bet
one
so,
but
there
is.
This
is
something
that
can
possibly
surprise
us
in
the
next
one
or
two
cycles,
but
certainly
in
this
one,
the.
B
Using
API
machinery
instead
of
flaca,
so
we
have
the
whole,
the
goodness
of
ABI
machinery,
but
we
are
not
using
a
trait,
so
we
are
really
badly
failing
in
the
promise,
but
I
got
I
got
your
point
that
we
should
not
expect
program
for
this
cycle.
So
we
are
just
let
me
say
getting
reading
could
mean,
but.
C
This
was
so
so
part
of
my
initial
version
of
the
cap
for
graduating
proxies
component
config
to
better
in
the
current
version.
That
is
much.
It
isn't
currently.
I
have
some
preliminary
version
of
alpha
2
for
you
proxy,
but
I'm
deliberately
not
publishing
it
simply
because
it's
not
a
major
upgrade
over
alpha
one.
It's
just
several
fields
being
shuffled,
and
this
is
probably
going
to
cause
users
bigger
issues,
for
example,
having
to
do
manual
upgrades
than
actually
giving
them
a
benefit
of
a
slightly
better
structure.
A
A
There
is
not
going
to
be
any
way
to
convert
them
to
the
next
version.
So
if
have
couplet,
do
you
want
be
the
one
that
is
currently
a
monolith
if
you
split
it
in
the
future
into
multiple
parts,
as
a
user
of
the
public
types
you're
not
going
to
be
able
to
up
convert
to
this
llamo
document
with
multiple
parts,
so
I'd
have
to
do
it
manually.
This
is
really
a
problem
for
query.
Pia
is
problem
for
cube
ATM
for
all
consumers
of
the
couplet.
A
D
C
But
there
is
also
this
is
one
side
of
the
coin.
The
other
side
is
that
most
of
the
users
don't
actually
specify
their
own
component
configs.
They
I
actually
look
into
cube
ATM
to
actually
generate
the
component
config
and
generate
an
opinionated
component
of
it
to
be
able
to
run
this
component.
So
what
I'm
proposing
and
what
my
PR,
but
that
you
actually
both
oak
in
the
past
few
days,
is
that
we
should
probably
sign
such
convicts,
which
we
actually
generated
by
a
cube
ATM.
C
This
is
probably
going
to
affect
just
users
that
supplied
some
simple
overrides
of
one
or
two
fields
in
the
component.
Config
and
those
are
the
users
that
are
going
to
be
impart
the
most
simply
because
they
just
edited
something
that
the
cube
am
doesn't
expose
an
option
and
now
they're
forced
to
go
in
and
poke
into
the
some
config
map
somewhere
to
edit
it
again
and
pick
up,
convert
it
to
a
newer
version.
Take.
A
B
B
The
problem
is
big,
because
not
them
I
consider
that
today,
the
majority
of
the
user,
but
me
are
shifting
from
being
a
personally
individual
user
from
being
a
girl
level
to
like
trust,
API
or
kind
or
me.
Okay
and
I'm.
Pretty
sure
that
all
of
these
tools
that
leverage
on
Cooper
in
our
using
our
customizing,
the
component
con.
C
Yeah
and
I
think
that
in
that
case,
they
should
probably
took
cube
ATM
stance
on
component
config,
so
they
should
probably
stump
somehow
the
convinced
that
they
actually
generated
and
basically
just
check
that
if
the
stamp
which
is
valid,
then
then
they're
free
to
throw
away
that
config,
because
they,
it
is
the
config
that
they
actually
generated
or
if
they
have
some
means
by
which
users
can
actually
supply
and
override
config
taken.
They
can
ask
the
user.
A
Again,
I
was
also
wondering
about
the
generic
question
of
how
to
create
the
forcing
function
for
a
pair
machinery
to
approve
our
proposal
for
exposing
conversion,
and
the
survey
can
provide
a
question
in
the
lines
of
hey.
Do
you
want
kubernetes
to
convert
your
llamo
files
and
with
your
supervision,
of
course,
and
if
we
get
a
lot
of
responses
to
the
survey
equation,
we
can
show
this
API
machinery.
They
have
to
approve
it
because
I
don't
think
they
currently
respect
our
opinion,
because
we
don't
have
the
data.
That's
my
position
on
the
problem.
C
A
Once
imagine
in
the
distant
future,
they'd
be
a
server
has
corporate
config
that
is
graduated
to
v1.
This
means
that
it
may
have
passed
multiple
stages
of
alpha
beta.
If
you
ask
the
users
to
modify
all
those
flags
manually,
they're
going
to
have
a
problem,
they
might
lose
options,
they
might
miss,
configure
their
API
server.
I.
Think
that
the
this
compulsion
is
something
that
the
a
machine
is
not
seeing
it
I
agree.
A
So
make
sense,
we
have
a
discussion
with
everyone
in
the
sequence,
lifecycle,
meeting,
I,
guess,
yeah,
it's
a
it's
a
big
problem
and
my
I
don't
want
to
sound
pessimistic,
but
this
is
a
problem
that
other
projects
have
not
found.
The
solution
for
supper
just
wanted
to
to
give
you
a
warning
that
this
is
not
an
easy
problem
to
solve.
No.