►
From YouTube: kubeadm office hours 2019-10-09
A
A
A
C
A
A
Anything
else
about
the
release.
We
can
revisit
this
if
needed
psays.
So
this
was
a
book
we
found
yesterday.
Somebody
sorry
a
couple
of
days
ago.
Somebody
reported
this
and
we
discovered
that
n
CD
is
broken
with
ipv6.
We
investigated
the
Google
folks
that
maintained
the
HCV
clients.
An
image
in
kubernetes
such
qualities
got
activate
it
immediately
to
send
some
PR.
So
we
have
to
back
port
the
fix
for
this.
This
only
affects
ipv6
Questers.
A
Another
one
is
that
I
got
contacted
by
sick
windows
yesterday
on
private
and
I
was
told
that
the
Microsoft
developer
that
wanted
to
work
on
the
cube,
ATM,
/
windows,
feature
to
get
into
beta
and
117
is
not
going
to
be
able
to
work.
Basically,
they
allocated
resources
to
support
container
D
for
kubernetes
and
Windows
deployments
so
I
was
asked,
can
I
work
on
that
and
my
response
was
that
I
don't
know
PowerShell,
so
I
can
only
coordinate
the
work,
but
somebody
else
has
to
be
assigned
to
do
the
PowerShell
changes.
A
A
A
Similar
for
glowstick,
but
we
for
those
that
we
already
have
tests.
I,
guess,
ok
in
terms
of
the
upgrade
I
think
the
upgrade
is
going
to
be
pretty
simple.
I
can
reproduce
upgrades,
probably
in
my
setup,
but
the
thing
is
that
we
have
to
support
upgrades
from
this
complicated,
install
script
written
in
PowerShell
and
basically
somebody
has
to
implement
the
logic
in
PowerShell.
I
I.
Don't
have
the
time
to
start
letting
PowerShell
right
now,
so
I
think
we
can
get
something
working
in
terms
of
upgrades,
but
the
testing
picture
is
very
unclear.
A
A
A
E
F
E
A
Okay,
so
this
is
basically
send
a
cup
of
beers.
They
saw
some
opposition
I'm,
proposing
that
we
start
duplicating
the
config
flag
for
upgrade
for
bridge
was
basically
said
that
this
is
maybe
too
early,
because,
though
we
don't
have
a
good
work
route
but
I'm.
Seeing
the
occasional
user
report
in
the
qpm
report,
where
people
are
breaking
things
using
graphic,
so
I
was
wondering
if
it's
time
to
duplicate
it
and
empower
network
or
the
operator,
and
basically
like
really
claim
that
this
should
not
be
used.
G
G
G
A
H
So
one
thing
from
my
side:
I'm
sorry
I-
have
workers
over
here.
So
maybe
it's
a
bit
noisy
I
will
be
fast.
So
when
one
possible
thing
will
be
to
identify
what
kind
of
us
are,
those
things
can
be
done
with
config,
and
maybe
until
we
have
the
cube
ATM
operator,
we
can
probably
check
if
this
config,
these
new
config,
comparing
it
to
the
one
that
we
have
in
the
cluster,
is
thingy
in
something
that
is
going
to
possibly
break
the
cluster.
A
Yeah,
this
is
one
option.
Yes,
I
like
this.
What
problem
with
this
does
config
feature
that
we
have
is
that
nobody's
all
in
it,
so
I
see
the
reports
in
the
Kuban
DM
repo,
but
nobody
is
replying
to
the
users
and
also
what
your
suggestion
is
going
to
introduce
some
boilerplate
that
we
have
to
maintain
as
well.
So
if
we
are,
if
my
proposal
here
to
not
be
you
know,
by
the
way,
we
can
duplicate
this
and
keep
it
for
I,
don't
know
seven
releases,
it's
fine,
but
if
we
don't
duplicate
it,
I'm
really
I'm.
G
So
I
think
that,
basically
having
something
to
check
whether
the
new
config
is
like
safe
enough
is
probably
not
going
to
be
reliable,
simply
because
users
can
actually
change
the
extra
arguments
of
some
of
the
like
static
packed
stuff,
and
this
can
actually
be
like
changed
in
a
way
that
we
cannot
foresee
being
like
dangerous.
So
I
think
that
probably
the
best
way
to
go
is
to
basically
just
market
us
dangerous.
G
Here,
basically,
there
is
an
if-else
section
in
site
upgrade
which
basically
checks
whether
the
config
buff
is
filled
in
and
it
will
load
the
config
from
the
file,
or
else
it
will
actually
go
to
the
cluster
and
fetched
from
there.
So
whenever
we
go
loading
from
the
file,
we
should
just
Kay
walk
warning.
Yeah.
A
G
B
B
A
A
B
A
Right,
let's
move
to
the
next
one,
because
we
have
a
number
of
topics
soon:
okay,
so
this
is
another
one
another
sketch
one
is
that,
should
we
overwrite
the
API
server
out
modes
and
basically
the
user
already
sent
a
PR
I
think
we,
like
we
had
a
little
bit
of
agreement
here
between
our
file
me
that
the
PR
looks
good.
A
roster,
however,
commented
that
we
should
probably
expose
this
in
the
config.
A
So
what
I
want
two
people
to
vote
on
this
so
that
we
can
get
it
moving
my
parcel?
What
is
that
we
should
do
complete
overwrite
if
the
user
passes
authorization
modes
are
some
extra
argument,
so
this
is
going
over
it
completely
the
our
defaults,
even
if
the
Questor
might
not
work,
we
probably
should
support
this
over
it.
I
mean
there
are
many
different
ways
to
break
a
question
so
yeah.
D
A
Okay,
so
this
one
additive
GTM,
so
this
is
about
the
structure
that
output
at
send
the
PR
a
mean
rhasta,
had
pretty
thorough
review
on
the
contents
of
the
peer
I
think
it's
a
good
shape
and
it
has
been
there
for
a
very
long
time
already.
I
was
thinking
that
maybe
we
can
merge
this
to
start
supporting.
You
know
the
first
command
is
talking
list,
a
structured
output.
If
there
are
problems
we
can
iterate
on
the
underlying
implementation,
if
needed.
Also
ed
said
that
he
might
add
new
commands
to
support
with
structured
output.
B
G
E
Just
like
curiosity,
we
haven't
done
this
in
a
long
time,
especially
for
stuff
like
this,
where
it's
kind
of
user
facing
type
of
behavior,
and
this
is
a
short
cycle
and
you
don't
have
many
major
committed
items.
It
might
be
beneficial
for
us
to
do
like
a
colored,
walkthrough
/
audit
near
like
mid
cycle
to
see
where
we're
at
just
to
get
like
a
an
overview
of
you
know.
This
is
where
we
are
today.
E
We
had
a
lot
of
stuff
that
we
talked
about
in
the
past
and
just
also
maybe
bring
in
new
people
into
the
fold
to
publicize
it.
So
like
we're
gonna,
do
a
code
walkthrough
and
bought
it
from
the
user
experience
standpoint
it
sort
of
outlines
some
of
the
state
of
where
we're
at
and
maybe
outlines
some
things
that
become
involved
in.
A
Its
I
wanted
to
see
what
the
mentoring
worked,
the
two
components
standard
the
currently
doing,
and
maybe
we
can
leverage
the
same
process
as
them.
We
definitely
lack
a
lot
of
contributors,
I
guess
it's
mostly
again
the
problem
with
complexity
and
people.
Just
you
know
they
don't
have
a
job
to
work
on.
E
So
they
kind
of
focus
on
the
shiny,
even
there's
still
a
ton
of
work
to
do
I
think
under
publicizing
that
there's
still
a
lot
of
work
to
do
and
that
books,
you
know
you
can
make
a
lot
of
progress
and
contributions
to
community
is
something
that
I
think
I
need
to
figure
out
how
to
do
this
at
a
macro
scale
across
communities
proper-
and
this
is
one
example
of
it
right.
Yeah.
D
B
E
A
Personally,
clear
can
basically
take
an
action
item
to
create
some
video
recordings,
possibly
also
via
like
zoom
meetings,
where
I
can
announce.
Okay,
if
you
want
to
join
the
zoom
I
think
I
want
to
show
you
what
what
is
the
structure
of
cube
a
DM
explain
some
of
the
maybe.
If
we
have
a
road
map,
I
can
also
talk
about.
The
road
map
show
what
issues
some
of
the
issues
we
have
in
the
bug
tracker,
incubate
iam.
A
C
I
thought
just
for
the
new
contributors.
One
thing
that
would
be
kind
of
nice
is
a
set
of
tickets.
That
people
who
have
contributed
don't
touch
and
don't
do
even
if
they're
easy
and
could
be
knocked
out
so
that
new
contributors
can
do
them.
Sometimes
it's
hard
for
I
still
struggle
finding
a
thing
to
actually
do
because
I
don't
know
the
project
so
like
unless,
if
it's
a
ticket,
that's
out
there.
It's
it's
hard
for
me
to
know,
and
some
of
the
easy
ones
get
really
fast
and
just
are
done
because
they're
easy
so.
A
Yes,
in
qadian,
there
seems
to
be
no
middle
ground
from
what
I've
seen
in
the
more
issue
tracker.
Either
things
are
super
simple
and
like
clean
ups.
People
take
them
right
away
or
something
is
very
complicated
in
his
discussion,
so
I'm
not
seeing
the
middle
middle
ground.
To
be
honest,
but
you
know,
coordination
in
the
issue.
Tracker
is
also
super
difficult
for
both
the
maintenance
and
both
the
people
who
want
to
contribute
so
I
don't
have
a
solution.
A
A
Mean
yeah
if
we
can,
if
we
ever
end
up
with
a
good
decision
proposal,
sorry
because
it's
not
easy
to
manage
I
guess
next
week
we
can
talk
about
new
contributors
in
this
stuff,
so
we
should
continue
all
right.
The
next
item
is
for
our
ostium
team,
basically
team.
This
requested
discussion
for
a
component,
coffee,
Kip,
so
I
added
this
to
the
agenda.
A
E
E
Select
the
first
comment:
there's
a
couple
of
pieces
here
that
it's
it's
funky
says
defaulting
that
configs
that
are
generated
by
a
cuvette
diem
or
supplied
by
users
populates
additional
fields
of
the
default
values.
This
makes
it
difficult
to
determine
retrospect
what
fields
are
set
in
purpose
with
their
default
values
and
what
fields
are
simply
automatically
filtering
and
faulting
code,
and
my
first
comment.
G
H
G
Meet
empty
or
actually
any
fields
that
are
made
like
marked
as
optional,
so
basically
defaulting
code
is
going
to
pre-fill
with
a
lot
of
values
and
it
kind
of
gets
difficult
to
distinguish
between
what
values
were
actually
filled
in
by
the
user.
What
values
for
actually
fill
in
by
cube,
ATM
and
what
values
were
actually
just
defaulted?
G
E
We
should
have
seen
defaults
a
little
bit
empty
on
the
component
context
for
the
system.
So
that
way
you
only
specified
the
overrides
that
are
required
and
there
has
to
be
user
specified.
We
don't
do
any
defaulting,
I
think
that's
part
of
what
respect
was
about
and
we,
but
we
do
maintain
some
defaults
in
committee
M,
but
those
defaults
should
be
pushed
in
I.
Think
that's!
The
Steven
I
was
trying
to
make.
G
Here
suppose
that
we
can
actually
try
to
push
some
of
the
defaults,
but
I
am
not
sure
whether
we
can
always
push
for
every
component,
all
of
the
defaults
into
that
component.
So
and
the
fact
that
we
actually
have
overrides,
which
we
in
our
code
base.
We
state
that
their
defaults,
but
they
are
not
default,
they're,
simply
overrides
which
we
specify
over
the
components
defaults
and
we
actually
go
in
and
default.
The
components
configure
with
it's
defaulting
code
and
it
grows
in
size.
E
B
E
B
B
B
A
G
A
E
A
Yeah,
this
change
is
also
the
order
for
in
it
about
it.
Overall
I
think
it's
doable.
G
Yeah
I
think
that
validation
should
be
one
of
the
last
steps
here,
at
least
like
from
safety
point
of
view.
So
if
we
actually
load
them
validate
the
stuff,
it's
probably
going
to
be
like
well
at
least
usually
validate
the
component
config
stuff.
We
can
validate
that
it's
okay
at
loading
time
and
it's
going
to
be.
This
is
going
to
be
okay
and
it's
safe.
G
But
probably
the
more
crucial
thing
here
is
to
basically
validate
it
at
storing
time,
so
we
can
actually
do
something
but
you'd
like
with
a
bulk
or
something
inside
cue,
medium
and
basically
corrupt
the
config
and
make
it
like
not
valid.
So
it's
best,
probably
shift
validation
to
just
before,
storing
and
if
possible.
We
can
also
validate
unloading.
B
Dependencies
in
the
middle
of
all
Dow
will
mean
more
flow,
and
this
will
make
how
they
could
be
more
for
or
less
stable,
less
country
and
have
a
lot
of
possible
error.
So
my
olessa
is
to
do
all
the
validation
in
the
provide
phase
and
then
has
one
when
we
know
that
a
component
config
is
good
start
for
the
deploying
without
interacting
with
with
the
component
container
or
binary.
G
E
H
G
Have
done
kept
I,
don't
see
any
reviews
but
I
submitted
to
the
beginning
of
the
week,
which
basically
adds
a
standard
command-line
interface
to
create
his
components
that
Mike
white
component
convict
and
wish
to
supply
an
interface
to
validate
that.
So
it's
not
mandatory.
But
the
idea
is
to
consume
this
interface,
either
fool
the
binary
directly
or
for
the
container
image
in
a
like
easy
standard
and
accessible
way,
and
we
should
probably
use
that
one
if
the
cap
gets
actually
merged
and
implementation
start
to
arrive.
A
Thank
you.
This
is
the
cap.
One
problem
with
that
is,
as
you
know,
it's
probably
going
to
see.
Some
opposition
is
going
to
take
a
lot
of
time
to
implement
the
validate
or
the
validate
command
or
whatever,
and
we
want
to
stop
validating
inside
cube
ATM,
which
means
that
there
might
be
a
period
of
time
where
we
don't
have
any
validation.
We
just
are
going
to
pass
the
user
value,
the
user
config
directory
to
the
components.
I.
Think
it's
fine,
even
if
we
just
say
hey,
we
don't
have
validation.
G
Yeah,
that's
exactly
what
is
suggested
by
the
component
config
cap.
So
in
the
beginning,
it's
going
to
be
like
just
some
pink
warnings.
If
the
standard
CLI
for
company
config
validation
is
merged,
I'll
probably
try
to
push
implementation
for
Q
proxy,
at
least
and
I
also
chat
with
Michael
token
for
cubelet
implementation.
So
to
get
these
to
us
up
for
consumption
by
a
cube
ATM-
and
this
is
probably
yeah-
this
is
probably
going
to
take
a
cycle.
G
A
A
So
basically,
the
user
sent
a
PR
because
they
saw
that
if
you
pass
only
a
belief,
the
certificate
key
without
control
plane,
they
are
not
seeing
any
indication
any
warnings
that
control
plane
is
not
set
according
to
them.
This
should
either
print
a
warning
or
an
error
row
still
you
want
to
discuss
your
proposal
to
always
feel
the
control
plane
if
certificate
Keys
passed
here.
G
Well,
the
idea
here
is
to
just
if
we
turn
a
face
manually
like
not
via
the
standard
join
workflow,
but
just
invoke
the
face
manually
and
we
omit
control
plane
and
the
face
is
actually
control
plane
face.
We
should
either
like
presume
that
control
plane
is
actually
present
and
default.
The
joint
configuration
according
to
this
way,
just
don't
pay
like
huge
warning
stating
to
the
user
that
the
user
wanted
to
execute
manually
control
plain
face,
but
the
mask
control
plain
option
was
not
used
or
no
control
plate.
Key
was
found
inside
of
the
tongue.
Configuration.
A
A
G
B
B
A
A
Instead
of
a
warning
and
error,
thus
that's
also
a
pretty
good
solution.
I
think
we
should
just
error
out
my
my
comet.
There
was
that
you
know
for
the
hope
here.
My
comet
is
that
we
are
getting
into
the
picture
of
managing
flag
mixtures
again,
which
is
pretty
bad,
but
yeah
I
think
we
should
instead
of
a
warning
question,
probably
just
error
out.
A
A
A
How
much
time
we
have
ten
minutes
all
right.
We
had
a
discussion
with
him
about
bumping
the
doctor
version
in
Cuba
diem
that
this
may
is
also
populating
the
validators,
but
we
still
don't
have
a
test
team
for
a
pump
which
I
guess
we
should
either
coordinate
with
them,
but
also
till
found
that
some
of
the
distances
still
want
version
18
instead
of
19.
So
I
guess
I
read
this
agenda,
but
maybe
we
should
talk
about
it
when
the
districts
are
more
up-to-date,
I.
A
A
A
Alright,
this
is
guest
it
merged,
but
yeah
it
worse.
What
we
have
to
change
it
again
alright.
So
this
is
a
going
back
to
this
topic.
It's
basically
somebody
from
the
mini
cube
maintainer,
which
is
using
eBay
DM
already
by
the
way
they
found
out
that
Cube,
a
DMR
set
is
removing
this
entire
directory,
which
is
true,
but
also
another
package
which
is
installed
by
the
way
which,
with
a
package
manager
that
sportmen
installs
networking
configuration
inside
this
EDC
CNI
father.
A
So
by
calling
cube,
ATM
receipt,
you'll,
always
nuke
files
maintained
by
the
external
package,
so
the
users
basically
proposed
sorry,
the
mini
cube
folks
propose
that
we
should
take
a
package
manager,
walking
and
see
hey
this.
This
file
is
maintained
by
package
manager.
This
is
possible,
but
the
problem
is
that
I
really
don't
want
to
get
cube
medium
into
the
package
manager
picture
at
all.