►
From YouTube: kubeadm office hours
Description
No description was provided for this meeting.
If this is YOUR meeting, an easy way to fix this is to add a description to your video, wherever mtngs.io found it (probably YouTube).
A
Hello
today
is
October
11th
2018.
This
is
the
convenient
office
hours
and
the
first
thing
we
had
on
the
agenda
was
to
talk
about
the
config
changes
to
get
it
to
be
1
beta
1,
or
to
talk
about
getting
kuba
diem
in
general
to
GA
for
BT.
Oh
did
you
want
to
walk
through,
maybe
like
the
original
documents
and
the
one
that
okay.
B
B
B
Okay,
so
at
the
beginning
of
the
document
today,
I
wrote
some
notes
about
the
pending
changes
that
we
have
to
do.
I
skip
the
first,
the
first
one,
because
there
is
a
PR
PR
already
out,
but
the
people
that
need
the
PR
today
is
not
here.
So
we
move
on
improve
control,
planar
configuration
by
using
substructure.
B
B
B
There
is
also
some
weird
things,
for
instance,
for
those
volumes
and
we
use
the
brighter
ball
instead
of
read-only,
which
is
more
or
less
the
standard
couldn't
mean,
but
those
are
minor
things.
So
the
original
idea
from
look
that
there
was
discusses
for
Lucas
is
we
have
a
rearrange
those
fields
in
the
into
three
sub
structure.
A
Which
are
the
three
components?
So
what
if?
Why
don't?
I
share
my
screen
by
your
date
while
you're
talking
and
we
can
actually
walk
through
the
actual
logistics
we
have
like
the
screen
share
along
with
it,
so
somebody
can
understand
what's
going
on.
Hopefully
this
is
the
right
screen,
yep
alright,
so
let
me
also
get
the
coup
BTN
back
in
here.
Alright,
so
the
first
one
you
said
was
improved
control
using
constructs,
and
that
was
also
this
portion
here.
So
when
you're,
when
you're
concretely
looking
at
the
types
file
and.
B
B
A
B
B
A
A
B
A
A
B
B
If
you,
maybe
it
is
easier.
If
you
look
at
the
source
code
and
it'll
search
for
joint
configuration,
you
see
you
have
all
the
discovery
attributes
there.
Will
someone
be
longest
for
to
the
to
discovery,
find
Mordor
someone
underscore
to
discourage
talking
mode,
but
basically
you
you
don't
know
which
one
you
do
you
were
to
to
assign
according
to
the
model,
so
it
is
not
clear
again
in
a
problem
of
the
ui/ux
and
in
the
original
design
of
the
API.
A
B
I
I
agree:
okay,
so
not
not
even
not
difficult
to
work
to
execute
button.
B
Okay,
okay,
the
the
ad
add-on
configuration
instead
II
is
something
that
that
we
have
to
do,
because
basically
it
is
a
problem
of
U
of
UX,
because
now
we
have
the
feature
flag.
Core
be
honest:
that
is
not
nice,
but
also
a
problem
of
missing
feature.
So
today,
I
try
to
I
went
through
this
issue
and
they
ended
in
the
document
of
the
API
I
summarized
the
main
question
or
pending
point
that
I
discussed
the
year
at
least
from
from
I,
so
Saudi
hookah
at
the
first
shop
do.
A
What
I
am
a
little
bit
remiss
I
get
the
I
get
the
idea
right
of
putting
in
here,
because
that
actually
abstracts
away
some
of
the
dependency
inside
of
the
code
itself
and
gives
us
out
of
some
of
the
hairy
problems.
We
have
it,
especially
at
the
end
of
cycle,
but
Google
has
been
trying
to
push
this
notion
of
a
bundle
which
is
a
well-defined
group
or
set
of
add-ons
that
you
can
apply
at
the
end
or
a
channel
so
to
speak.
So
that
way
you
could
vet
the
ability
to
yeah.
B
Honestly
I
went
to
the
project,
but
there
is
no
documentation
so,
but
I
got
the
idea
of
channel
or
bundles,
but
I
think
that
that
we
have
two
kind
of
problems.
So
the
first
problem
is
that
ow
included
at
the
coop
a
admin
and
could
mean
config.
We
define
what
we
want
and
at
which
degree
of
configurability
we
give
to
the
user
and
the
second.
B
Have
to
let
the
user
to
choose
which
DNS
they
want.
You
know
in
a
in
a
more
clear
way,
then,
that
there
is
the
we
get
into
the
problem
of
what
degree
of
configurability
we
let
the
user
ever
on
the
unsub
dawn.
So
considering
that
some
configured
ability
will
be
driven
by
the
component
config,
but
currently
the
currently
the
component,
the
component
coffee
applied
only
to
the
proxy.
A
Well,
yeah,
you
need
to
you
need
to
be
able
to
provide
the
that
the
problem
is
the
overrides,
apply
to
all
the
images
and
not
just
like
one
or
two.
If
you
want
to
be
able
to
customize
the
repositories
for
your
individual
add-ons,
you
know
that
you
want
to
apply.
It
becomes
it's
not
possible
today,
it's
again,
this
is
this
falls.
Underneath
e
I
think
this
falls
in
ether
nice
to
have
not
the
super
critical
like
we
can't.
A
Some
of
these
pieces
do
need
to
get
fixed,
especially
if
you're
doing
air
graft
installations
to
be
able
to
provide
the
overrides
that
pass
all
the
way
through
to
the
individual
components,
but
they're,
not
providing
many
registries
or
repos
is
not
a
I,
would
say
a
hard
requirement
to
get
to
GA.
It's
a
nice
to
have
yes,.
B
B
A
You
could
do
a
generate
if
people
wanted
to
override
and
it
generate
would
choose
all
the
defaults
and
specify
them
in
their
list,
and
a
person
can
customize
right
like
we
have.
The
print
defaults
option
could
config
print
defaults
right
then
print
the
fall,
so
basically,
throughout
a
configuration
file
that
has
all
the
details.
The
one
problem
we
don't
have
is
being
able
to
specify
the
manifest
or
the
the
registry
location.
B
B
A
If
they
wanted
to
do
it,
something
like
that
they
could
just
you
know
they
could
just
take
the
output
of
what
could've
idiom
specifies
and
read
those
manifests
into
their
own
settings
and
customize
them
appropriately
right.
There's
nothing
that
prevents
a
person
from
doing
this
ad
hoc
and
I
think
that
level
of
configuration
is
not
something
we
want
to
support,
because
it's
too
flexible
right.
Yes,.
B
C
A
From
a
from
an
Aaron's
perspective,
having
an
explicit
list
that
allows
you
to
customize
is
a
good
thing,
because
people
and
be
able
to
generate
it,
is
also
a
good
thing.
I
do
I,
do
wonder
what
the
overlap
is
with
bundles,
because
I
have
not
dug
into
bundles,
so
I
do
think
it
could
be
potentially
premature
right.
A
So
if
we
were
to
optimize
and
do
this
work
for
having
explicit
list,
it
seems
like
this
also
solves
the
same
problem
at
this
point,
but
I'm
I
again,
I
have
not
dug
into
this
in
detail.
Yet
so
I,
don't
know
where
the
overlap
is
I
would
I
would
put
this
as
needs
to
evaluate
need
to
evaluate
overlap
with.
A
B
B
B
B
B
A
So
I,
don't
I.
Don't
think
that
it's
requirement
for
for
getting
us
to
beta
I
do
think
that
it
eventually
in
the
long
haul.
It
would
be
a
nice-to-have
thing,
but
I
think
most
people
would
try
to
trust
the
defaults,
especially
as
Kubb
DNS
is
meant
to
be
ripped.
You
know
deprecated,
and
we
should
only
go
with
core
DNS
going
forwards.
A
B
Because
then
the
problem
gets
more
complex
a
because
there
was
some
issue:
I
can
leak
them
too.
So
the
people
in
Stanley
accordion
s
customizing
the
config
map
directly
into
kubernetes
and
then
using
could
add
me
in
a
for
doing
a
break
upgrade
and
cubed
min
was
not
aware
of
the
customization
of
the
company
map.
A
Okay,
so
I
still
think
it's
a
nice
to
have
I,
don't
think
it's
for
most
people.
They
are
not
going
to
pay
attention,
especially
as
senior
defaults
get
chosen
over
time.
I
do
think
that
there
there
will
be
people
that
will
want
to
customize
and
tweak,
especially
as
they
scale
out
of
their
deployments.
Okay,.
B
Iii
can
agree
so
there
are
only
some
people
that
really
want
one
issue
about
talking
about
these,
but
there
is
another
problem
which
is
a
related
to
a
dawn,
and
this
problem
I,
if
I
remember,
were
authorized
by
a
JSON,
and
these
are
related
to
the
not
specific
customization.
In
that
case,
if
I
remember
where
it
was
not
a
specific
assimilation
for
Cooper
proxy
and
I.
Remember
that
that
we
were
talking
about
I
need
container
before
changing
configuration.
Oh
that's.
B
C
A
B
B
B
A
B
B
A
B
A
True,
so
it's
not
it's,
it's
not
a
requirement.
It
makes
our
lives
easier
for
managing
this
stuff.
But
it's
not
it's
not
a
requirement.
Well,
it
depends.
If
somebody
uses
it
from
the
API,
then
it
becomes
a
pain
in
the
Kiester
because
we
have
to
manage
it
going
forward.
So
I
guess
you
do.
We
do
have
to
excise
it,
because
otherwise,
if
it's
in
the
API
so
look
and
use
it,
and
we
have
to
support
it
right
in
the
fashion
that
it's
in
so
it'd
be
like
v1
beta
1
that
bootstrap
tokens
dot
whatever
well.
A
B
B
Okay,
okay,
then,
the
the
last
the
last
tool
changes
which
are
related
to
the
PIR
one
is
to
implement,
take
the
TCD
yuckies
so
basically
to
have
a
cubed,
minus
b,
now
spin
up
a
new
local
haiti
CD
and
when
joining
us
a
second
contour
planing
instance,
and
it
is
it.
This
was
listed
in
the
in
the
api
changer,
because
we
were
acting
that
that
some
kind
of
impact
on
on
the
API
as
well.
I.
A
C
A
B
C
A
A
You
could
you
could
have
another
option
for
join
that
you
could
add
at
a
later
date.
It
doesn't
preclude
you
from
doing
this
great
words,
so
that
seems
like
like
a
legitimate
path.
Words
this
this
is
the
default.
For
now
we
can
always
add
the
other
option
in
the
next
iteration,
but
I
would
definitely
open
an
issue
for
adding
this
option.
Do
we
do
we
want
to
support
this
as
a
because
right
now?
It's
not
even
it's
like
an
alpha.
Ish
thing,
it's
experimental
right
like
we
could.
C
So
what
use
cases
do
we
have
for
scaling,
stat
control,
plane?
And
then
you
know
scaling
address
additional
control,
plane
instances
off
of
that
versus
kind
of
the
complete
external
@cd
workflow,
where
you
have
your
external,
let
CD
and
then
you
scale
up.
You
know
the
the
control
plane,
instances
independently.
A
C
A
B
Sorry
for
their
and
I
need
to
search
for
desisting
at
the
city
instance,
and
basically
now
what
what
I'm
doing
for
searching
the
existing
ethnicity
and
point
I'm
querying
the
kubernetes.
Please
give
me
all
the
components
that
have
such
labels
and
I
find
all
the
thoughts
that
are
running
at
ECD
in
the
system
explains.
It
is
fine,
but
then
I
have
to
grab
grab
out
from
this
folder.
The
endpoint
address
and
I
I
have
to
two
ways
to
do
this.
B
B
A
B
A
You
learn
it
from
the
circular
might
be
from
the
eyepiece
and
it'd
be
better
to
use
the
pod
IP,
but
in
that
particular
case,
but
we
could
always
embed
metadata
inside
of
the
because,
if
you're
searching
in
the
system
namespace
for
the
pod,
nothing
prevents
you
from
adding
metadata
either
as
annotation
or
as
labels
to
make
it
easier
to
parse.
Then
then
it
would
be
to
try
to
just
infer
from
the
data.
That's
already
there
as
part
of
the
manifest
I.
A
B
A
B
A
B
B
A
C
You
know
I
think
we
need
to
make
sure
that
if
we
are
discovering
the
endpoints
that
we
don't
just
necessarily
rely
on
just
data
from
the
config,
but
we
can
use
that
data
from
the
config
to
do
the
introspection
to
get
it,
because
once
we
once
we
have
kind
of
the
credentials
that
we
need
we
connect.
We
should
be
able
to
just
query
at
CD
itself
to
get
the
valid
list
of
current
endpoints
yeah.
A
C
A
There's
like
a
bunch
of
other
PRS
that
are
kind
of
flying
around
too
so
I
think
this
would
be
it'd,
be
nice
to
promote
to
beta,
but
I,
don't
think
it's
required
to
promote
to
beta
I
think
we
could
leave
it
as
experimental
and
people
would
still
use
it.
I
know
we
would
we
would
beat
on
it
and
then,
once
once
we've
shaken
the
orange
tree,
we
can
absolutely
promote
this
feature
to
beta.
A
A
C
A
B
A
A
A
B
Ok,
I
started
work,
working
with
Lou
Hamer
on
the
phases,
yes,
and
and
basically
we
divided
the
work
in
this
way.
I
implemented
the
1
PR
that
that
is
that
I
sent
some
hours
ago
that
introduced
that
they
face
the
de
film,
the
famous
face
runner,
yet
that
we
will
take
charge
of
running
phases
in
generating
subcommander
for
each
phase
and
so
on.
B
So
the
end,
I'm
senticles,
so
another
PR
for
graduating
the
first
phase
just
web
annex
an
example,
the
tent
and
then
people
kept
happy
exactly
and
lyubimov
star.
It
is
not
the
era.
Every
second
same
started
to
try
to
define
the
list
of.
There
are
two
a
list
of
phases
that
we
are
going
to
permit
and
we
open
a
PR
for
discussing
T's.
It
is
the
status
of
today
so
before
a
meaning
to
to
PR
in
flight
that
that
start
of
the
MU
or
and
an
issue
that
trying
to
define
the
plane
up
to
completion.