►
From YouTube: kubeadm office hours 2019-07-31
A
A
A
Rossi
proposed
that
maybe
we
should
have
some
automation,
I'm,
also
thinking
about
having
an
end-to-end
test
for
this.
By
having
an
internal
test,
we
can
verify
that
the
constants
file
action
matches
the
master
branch.
We
need
this
for
master.
Only.
We
don't
have
to
have
entered
this
from
the
rest
of
the
branches,
but
I'm
thinking,
I
started.
Thinking
about
the
intent
is
for
this
and
I
mean
it
can
be
periodic.
Even
you
don't
have
to
block
PR
service,
but
it's
really
it's
very
nice
graph,
because
we
forgot
to
update
this
concept
and
it's
not
great.
B
A
C
D
Would
also
have
concerns
with
further
coupling
to
the
kubernetes
build
process
when
the
idea
is
at
some
point
to
rip
these
things
away
from
the
mono
repo
yeah.
That's
a
problem:
do
we
have
a
I
mean
there?
There
are
several
manual
things
that
we
need
to
change.
Do
we
have
a
complete
list
of
these
that
we,
like
a
checklist,
that
we
have
for
Lisa
I.
B
D
Think
having
the
versions
in
the
constants
file,
because
it's
actually
code,
it's
not
deriving
it,
which
means
that,
if
I'm
messing
with
something
version
related
as
a
developer
on
Covidien,
which
I've
had
to
do
in
the
past,
that
I
have
full
control
over
what
versions.
It's
trying
to
work
with,
like
I've
used
this
especially
to
determine
if
something
was
regression
before
by
like
taking
the
113,
build
and
trying
to
run
it
on
111
112.
D
A
Let's,
let's,
let's
probably
to
look
up
this
as
an
action
item
or
we
can
continue
the
description
here
because
we
have
all
of
agenda
but
definitely
something
that
you
have
to
keep
in
mind
every
cycle:
I
I'm,
sorry
I
forgot
so
yeah,
let's
move
to
the
next
one!
So
Blitzer!
Why
don't
you
talk
about
the
G
repeat?.
B
B
Tokenism
that
could
mean
stop
to
use
this
file
and
why
I
sent
the
PAPR
for
the
rating.
This
file
is
because
now
we
are
allowing
user
to
pass
Kadesh
credential
for
the
TLS
booster
token.
So
that
means
that
there
is
the
potential
risk
that
a
user,
Passa
I
privilege
credential,
like
custard,
mean,
and
this
credential
remain
stored
in
the
booster
cuckoo
bird
country.
So
I
think
that
it
is
saying
to
the
rate
of
this
file
and
the
PR
code
method,
and
now
we
have
to
the
choose
two
things
one.
B
A
A
If
somebody
is
using
the
bootstrap,
the
the
couplet
boosts,
are
found
for
some
reason:
I
don't
know
it's
probably
wrong
to
use
it,
but
maybe
somebody
is
doing
something
with
this
file
and
we
have
to
keep
in
mind
that
we
are
GA
and
such
by
parts
with
an
action
item
and
fairly
uncommon
in
coverages
in
general.
That
was
my
comment
on
this
I.
D
E
B
B
A
If
somebody
really
needs
this
patch,
they
can
apply
it
and
use
it
in
their
custom,
builds
on
115.
So
it's
fine
and
the
workaround
is
super
simple,
but
I
agree.
The
token
is
the
only
secret
that
they're
going
to
see
as
a
leftover
in
the
bootstrap,
the
couplet
will
suffer
because
the
elevator
is
not
working
correctly.
Okay,
the
next
one,
the
customize
POC.
B
D
A
A
B
B
Yeah
robbery
I
am
going
because
the
the
Cora
disparities
later
the
let
us
commit
that
you
see
but
and
then
there
is
a
lot
of
changes
in
anodic
the
flag.
You
know
the
faces
and
know
the
Wardrobe,
so
I
can
send
the
car
of
the
change
as
a
single
PR
and
then
change
in
a
separated
PR
send
the
changes
to
the
uux
deck.
B
A
A
Yesterday
we
had
a
discussion,
the
sequence
life
cycle
meeting
about
best
practices
for
upgrades
and
this
chiral
able
to
keep
adium,
but
the
topic
is
very
thorny
because
it
touches
a
recommended
upgrade
skewing
communities
and
also
the
recommended
upgrade
order
of
components.
So
I
created
basically
did
a
brain
dump
here
after
him
and
just
any
question
that
I
wrote
an
issue
and
it
covers
a
lot
of
basically
detail
on
what
what
type
of
topology
we
are
currently
imposing
with
the
documents
that
are
in
place
like
the
recommended
upgrade
process,
Jordan
commented
3
comments.
A
E
A
So
I
like
the
idea,
somebody
who,
however,
has
to
take
action
on,
writes
in
a
document,
because
we
have
to
start
formulating
the
idea
into
a
Google
dot,
possibly
then
create
the
kit.
So
do
you
think
that
we
should
start
now
I,
don't
think
they're
going
to
have
the
time?
Maybe
we
should
start
make
cycle
yeah.
E
Well,
I:
don't
think
this
is
a
high
priority
for
now.
I
think.
The
reason
for
the
issue
in
the
PSA
is
just
that
if
you
have
specific
things
that
you
would
like
to
see,
please
add
them
there
and
for
broader
consumption
and
visibility
and
then
in
next
we
can
prioritize
it
for
next
cycle
and
talk
about
it.
A
A
G
A
A
lot
this
is
awesome.
We
we
have
I'm
looking
at
the
checklist
so
for
this
cycle.
This
is
this,
is
the
sorry
is
Mike
internet
okayed.
You
can
hear
me
yes,
okay,
so
this
is
the
depending
pair
here.
The
remaining
action
is
going
to
be
end-to-end
tests
for
the
coastal
city
changes.
Do
you
think
that
we
should
this
also
question
for
team?
But
do
you
guys
think
that
it
should
be
a
good
idea
to
other
section
in
our
dogs
for
creating
a
control
plane,
a
single
control,
plane
cluster?
E
Normally
I
generally
do
not
like
the
the
practice
that
we've
tried
to
employ,
but
never
really
have
is
that
we
could
have
kuba
DM
Docs
in
the
committee
M
site.
That
outlines
how
to
enable
these
things,
but
we
should
not
be
publicizing,
and
this
has
even
existed
in
the
main
box
saying
where
we
talked
about
it
with
them.
This
happened
a
long
time
ago.
E
We
should
not
be
publicizing
alpha
grade
features
in
the
main
documents
where
possible,
because
there's
no
guarantee
of
support,
there's
no
contract,
so
the
and
I
know
like
I
removed
a
bunch
of
stuff
and
somebody
added
a
bunch
of
stuff
back
I
think
was
Lucas,
but
the
about
the
state
of
features-
and
we
generally
should
not
be
doing
that
for
anything.
Alpha
beta
is
okay,
that's
a
different
story
because
there's
some
guarantee
of
support,
but
alpha.
We
should
not
be
doing
that.
A
But
by
the
way,
the
sick
doctor
have
a
very
weird
process
of
merging
changes,
so
if
I
remove,
for
instance,
if
I
remove
this
table
and
appear
during
repays,
it
might
might
reappear
because
of
a
human
mistake.
It's
very
weird
I
have
complaints
about
the
way
they
are
doing
things.
So
do
you
think
that
we
should
remove
this
table
completely
at
this
point
or.
E
I'd
be
fine,
removing
this
table
completely
I,
don't
I,
don't
feel
that
the
people
come
into
the
committee.
Mdoc
side
will
be
interested
in
knowing
the
state
of
certain
features
they
care
about
them
and
we
have
separate
docs
for
them
right
like.
If
you
want
to
know
more,
you
can
dig
it.
You
can
dig
deeper.
E
This
is
like
this
weird
overlap
between
what
is
a
developer
idea
and
what
is
a
publicly
facing
thing
like
we
should
have
Docs
for
our
GA
things
and
make
sure
that
they're
publicly
visible
just
by
default,
but
we
should
not
be
publicizing
things
that
require
feature
knobs
on
our
alpha
level
on
the
main
documentation
site
other
than
people
can
experiment
and
they
can
discover
things
on
their
own.
But
it's
not
it's
ass
in
a
bad
state.
Great,
ok,.
A
I'm
going
to
create
a
ticket
to
ask
somebody
to
help
removing
disabled
I
did
I
also
agree
that
we
should
not
this.
This
is
development
developer.
Specifics
like
we
should
not
be
doing
this
in
the
main
documentation.
So
again,
going
back
to
the
topic
of
door
stack
I,
guess
we
should
not
document
dosac
for
now.
Maybe
if
we
get
but
there's
a
question
here,
do
you
think
that
we
we
might
get
a
lot
of
user
questions
in
our
issue?
Tracker?
If
you
don't
have
these
dogs,
because.
G
In
my
yeah,
in
my
opinion,
I
think
people
who
want
to
use
Google
stack
will
really
not
need
to
know
what
they're
doing
it's
not
like
a
trivial
feature.
There's
a
lot
of
stuff
outside
of
to
baby
and
I
need
to
be
set
up.
I
would
imagine,
at
least
for
the
Alpha
part,
that
they'd
be
smart
enough
to
sort
of
look
at
the
PRS
and
make
sense
of
things.
G
A
G
Maybe
when
it
hits
paid
I
just
for
now,
I
think
we
can
just
leave
it,
but
once
it
hits
data,
probably
early
next
year
or
mid
next
year,
we
can
definitely
document
that
okay,
yeah
again
my
opinion,
would
be
again.
People
who
really
want
to
mess
around
with
dual-stack
need
to
know
what
they're
doing
and
yeah
they
might
bug
us
with
an
issue
or
something
but
I
think
they
should
be.
You
know
clever
enough
that
go
read
the
PRS.
G
D
A
D
Cool
so
there's
this
there's
a
component
config
style,
API,
I
created
and
that's
in
the
I
suppose
I
could
link
this
patch
later.
This
is
in
the
add-on,
installers
repo
under
the
Installer
folder
and
then
here
under
demo,
there's
some
examples
of
just
this
add-on.
Installer
configuration
there's
only
two
fields
in
this
API.
At
the
moment,
one
of
them
is
a
list
of
customized
package
drafts.
So
the
approach
that
we
chatted
about
in
the
add-ons
group
has
been
to
use
customized
packages
to
ultimately
get
whatever
is
being
used
to
manage
the
add-on
into
the
cluster.
D
Then
it
gets
iterated
through
in
serial
or,
however,
the
consuming
user
of
the
library
decides
to
install
this
list
of
things.
That's
done
with
a
cube,
CTL
exec,
currently
or
rather
in
an
exec
to
gup-c
tail.
So
this
is
the
library
it
just
exposes
functions
for
making
sure
that
goop
CTL
is
on
the
path
and
all
of
that,
and
then
it's
got
a
plural
and
a
singular
version
of
installing
the
add-on
and
just
the
plural
version
goes
through
that
list.
Serially,
the
runtime
object
holds
the
reference
to
the
add-on.
D
Installer
configuration
that's
expected
to
be
parsed
and
read
in
by
the
consumer,
and
then
it
also
gets
passed
in
a
standard
out
in
the
standard
error
so
that
you
can
control
logging
a
little
bit.
We
could
put
a
logger
interface
here
instead,
but
I,
don't
know
what
people
want
for
this,
and
this
worked
pretty
well,
so
nothing.
Nothing
here
is
breaking
kind
of
the
encapsulation
of
output,
that's
being
represented
by
the
runtime,
and
the
runtime
is
controlled
by
the
caller.
And
then
this
command
set
is
an
internal
struct.
D
B
D
D
Installed
yeah,
so
coop
Carol
doesn't
have
a
component
config.
A
B
D
D
B
D
B
D
B
B
I
wish
on
the
issue:
okay,
the
Dana
ever
another
question
about,
so
you
are
amazing
that
they
you're,
assuming
that
there
will
be
some
place,
a
guitar,
a
ball
or
something
or
a
locker,
where
the
user,
where
there
are
there,
dawns,
I'm
thinking,
especially
to
proxy
end
and
incor
B&S.
So
they
don't
like
the
mini
studying
right
now
and
how
is
this
supposed
to
work
in
in
our
gap
in
environment?
D
B
E
D
Yes,
the
those
would
be
for
the
image
repositories
or
the
image
registries
right,
so
the
other
bit
of
this
is
I.
We've
chatted
in
the
group
about
extending
customized
to
support
using
docker
and
OCI
images
for
packaging,
so
that
would
that
would
be
in
line
with
all
of
our
other
artifacts.
You
know
in
an
air
gap.
D
A
So
I
really
like
this.
This
is
a
great
proof
of
concept,
but
if
we
want
to
implement
the
cube
ADM,
maybe
it's
time
to
start
writing
a
Google
Doc
and
a
cap,
that's
you
know
we
have
to
have
a
replacement
for
a
proxy
in
Cordilleras
in
a
way
that
it
support
air
gapped.
Everything
is
proposed.
The
order
of
applying
the
the
the
kubernetes
objects
is
similar,
and
then
we
have
to
apply,
for
instance,
using
the
same
mechanic,
the
network
plug-in.
So
there
has
to
be
some
sort
of
key.
A
D
A
E
E
D
E
D
D
A
He
has
been
helping
us
what's
with
the
hard-coded
coordination
cube
ATM,
but
I
really
think
that
we
cannot
get
this
as
a
replacement,
because
the
sheer
amount
of
decision-making
and
work
on
the
operator
for
coordinates
to
process
happen
in
116
yeah.
So
if
you're
not
comfortable
with
a
cap,
you
can
write
the
doc
we
can.
You
know
we
can
evolve
the
doc
over
time.
We
can
have
a
discussion,
but
if
you
don't
want
to
go
for
the
cap
directly,
that
that
is.
D
B
Two
other
question:
two
things
that
I
didn't
understood
looking
at
the
POC
is:
if
there
is
a
way
for
Cooper
to
mean
to
past
configuration
like
the
component
config,
for
instance,
the
proxy
over
the
mini
receiving
input
from
the
user,
the
Cooper
proxy
configuration,
and
then
he
test
pass.
This
configuration
to
the
to
the
proxy.
D
So
if
you
wanted
to
override
something,
you
can
just
extend
the
customize
with
your
own
layer,
and
you
can
even
store
that
just
on
your
local
system,
it
could
be
from
a
git
repo.
It
could
come
once
we
expend
customized
with
the
image
support.
It
could
come
from
a
docker
image,
but
basically
it's
it's
just
using
customized
in
probably
a
similar
manner
to
extending
the
static
pod
manifests.
D
Does
that
make
sense
like
basically
I
could
go
into
the
add-on,
installer
configuration
and
then,
instead
of
having
this
right,
then
I
could
comment
that
out
and
then,
instead
of
you
know,
having
could
be
in
any
sig.
So
I
could
put
my
own
repo
here
or
I
could
use
the
local
file
system
could
even
just
yes.
D
It's
not
that
I
mean
right
now.
Could
cuddle
does
that
for
you,
so
you
just
if
you
want
to
provide
a
proper
customized
folder
on
the
local
file
system,
then
this
library
is
just
going
to
call
coop
cuddle
with
that
folder
and
if
you
are
using
a
git
repository
instead,
then
it's
going
to
fetch
it
from
a
git
repository
unpack
it
temporarily
and
then.
B
Yes,
well
I'm
studying
kupah
demeanor
on
a
machine,
so
it
doesn't
mean
that
the
process
for
his
public
proxy
is
good
means.
She'll,
go
from
sorry
positive,
fetch
today,
the
day
the
resource,
all
the
fetches
are
the
ephraim
fetches
and
then
shell
out
the
adonis,
tala
cook
or
cool
or
cook
with
cattle
and
East
Asia.
It
could
mean
it
is
doing
the
same
work
of
this
talent
and.
D
D
I
mean
basically
right
now,
all
of
our
add-ons
are
hard-coded
into
queue
bottom,
and
we
want
to
be
able
to
package
and
distribute
those
in
a
way
where
I
think
tools
like
cops,
ETS,
cuddle
and
covet
I'm
like
have
a
way
to
pull
them
in
from
a
central
place
and
for
people
to
also
not
just
have
a
central
place
for
like
core
add-ons,
but
to
then
be
able
to
publish
their
own
add-ons,
whether
that
be
a
network
provider.
Or
you
know
a
service
measure
or
whatever
yeah.
E
I
need
to
do
to
be
honest:
I
need
to
catch
up
on
the
current
state
of
patents
and
I.
Think
we're
kind
of
spinning
around
the
drain
a
little
bit
here.
I
think
we
need
to
do
a
little
bit
of
homework
and
we
need
some
more
documentation
before
we
kind
of
do
any
more
hot
takes
to
make
sure
that
we
are
informed
and
then
from
there
maybe
loop
back
on
the
dock
and
follow
up.
D
I
I
just
really
wanted
to
take
the
time
in
this
meeting
to
just
walk
through
these
ideas
here.
What
kind
of
questions
you
have
and
then
ultimately,
just
show
that
what
we've
built
right
now
is
simple,
like
that,
that's
the
command
running
and
the
first
one
is
running
from
a
local
file
system.
This
is
doing
a
network
fetch
and
creating
a
pod
from
one
of
the
github
examples
from
customize
and
it
takes
in
a
config
which
is
a
component
config
yeah.
D
So
that's
that's
basically
what's
currently
built
and
it
execs
out
too
Coop's
ETL,
you
can
manage
the
commands
using
the
library
currently
I
do
think
that
the
coop
config
should
probably
be
passed
in,
that's
very
basic,
so
we
can
do
that
and
then
other
than
that
that
sounds
yeah.
Any
other
feedback
I'll
put
together
some
more
docs
and
stuff.
It's
really
easy
to
pull
and
build
it.
I
does
have
a
readme
and
all
that
so
woods,
yeah
and.
A
A
A
A
A
A
A
Shouting
Gao,
he
has
been
a
contributor
of
the
commedia
before
he
appeared
today
with
a
something
that
is
very
peculiar
in
the
upgrade
process
we
have.
There
is
a
certain
condition
where
I
I
don't
understand
why
it
happened.
We
can
end
up
in
a
situation
where
we
try
to
delete
the
pre-poo
demon
said
that
no
longer
exists
and
the
question
is:
did
we
delete
it
before
that?
Why
is
this
even
happening
and
I
commented
that
I
saw
this
occurring
in
a
couple
of
different
tickets,
incubated
in
the
chameleon
repo
and
I'm
asking
like
hey?
A
We
need
to
understand
why
this
is
actually
happening
instead
of
doing
what
he
is
trying
to
do,
and
that
is
to
ignore
the
not
found
error.
I.
Think
I
am
proposing
the
right
thing
here,
but
the
question
is
like
I
need
to
I
need
to
get
an
answer
for
that.
I
need
to
get
an
answer
like.
Why
is
this
happening,
and
maybe
we
should
lock
a
ticket
and
then
I'm,
maybe
I
can
ask
shouting.
A
B
A
A
A
Okay,
this
is
we
already
talked
about
this.
We
forgot
to
pump
these
versions.
It's
not
a
massive
problem.
The
cherry-picks
for
row,
14
or
15
are
still
up
for
file
discovery,
which
you
didn't
get
a
response.
That
Ardi
is
going
to
make
the
next
release
of
115
this
particular
one,
or
is
this
good?
Let.
A
A
So
what
is
doing
is
basically
automatically
applying
API
version
and
kind
through
pipes,
and
this
is
a
major
change
and
if
somebody
wants
me
to
ping
him
message
me
on
slack
I.
Think
I
already
mentioned
Ross
GM
Vince
on
this,
but
this
is
labeled
Kapadia,
because
it's
also
touching
our
I
believe
our
schema.