►
From YouTube: 20181114 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
Wednesday
November
18th,
all
right,
November,
14th
2018.
This
is
the
standard
committee
and
office
hours.
If
you
could,
please
add
your
details
or
questions
or
comments
into
the
meeting
notes.
That
would
be
helpful.
I'm
going
tough
down
from
the
agenda
first
up
is
little
mirror
with
strict
DML,
marshalling
and
unmarshal
or
strictly
amun
marshalling
should
be
any.
What
should
we
do
for
113?
You
wanna
go
through
the
details.
A
B
So
what
I
did
instead
is
I
added
the
changes
locally
to
our
verification
of
init
config,
drunken
config
couplets
cookie
box
in
coastal
configuration,
and
but
if
the
feedback
I
want
to
get
here
is
like
do
we
want
to
enable
the
warnings
and
possibly
how
to
proceed
with
them
like?
Should
they
be
block
warnings?
And
that's.
A
The
question
I
commented
on
the
issue
already
and
I
think
it's
good
for
the
people
of
comment
too
I
think
for
right
now,
because
we
don't
want
to
break
anybody
who
might
already
have
misconfigurations,
which
is
probably
a
fair
number
of
to
just
output,
a
warning
as
part
of
whatever
process,
whether
it
be
upgrade
or
in
it
I
kind
of
think.
Certain
workflows
should
fail
like
in
it
having
a
strongly
typed
failure
for
an
it,
but
I
think
for
the
time
being,
just
one
on
everything.
A
In
that
way,
as
when
we
get
officially
get
the
config
to
GA,
we
can.
We
can
switch
that
behavior
from
a
warning
to
error
or
possibly
have
a
flag
that
allows
them
to
be.
You
know,
have
a
flag
that
allows
it
to
be
strict,
I,
think
warning
for
now
it
seems
fine
and
then,
as
we
get
to
GA
with
config,
have
a
comment
and
an
issue
filed
and
I
think.
What's
probably
a
good
path
for
us.
B
B
Yeah,
this
makes
sense,
like
the
PR
doesn't
doing.
My
problem
here
is
at
the
pier
I'm
doing
is
kind
of
it's
not
pretty,
because
we,
you
don't
have
a
good
way
of
doing
this,
so
I
can
I'm,
probably
gonna
updated
today
or
tomorrow
and
I'm
gonna.
Look
for
more
feedback,
my
initial
ideas
to
enable
the
warnings
for
everything
not
throw
errors
at
all.
This
cycle
like
this
is
something
that
we
have
to
agree
upon.
I
think.
D
B
A
Think
I,
like
the
current
warn
approach
and
then
we
can
revisit.
You
know
to
tighten
up
because
if
we
do
add
sub
commands
or
flags
as
because
we
are
promoting
the
tool,
the
GA,
we
would
have
to
go
through
policy
to
do
it
and
it's
just
default.
Behavior.
It's
not
going
to
disrupt
anybody
right
now,
which
I
think
is
the
important
part,
and
then
we
can
always
we
can.
We
can
find
a
path
for
words
in
114.
B
A
I'm
all
for
forewarning
and
getting
it
done
and
ASAP
I
know.
My
current
backlog
is
actually
surprisingly
low.
I've
gone
through
everything
and
there's
there's
only
like
some
minor
PRS
that
I
see
they're
currently
pending.
So
nothing
it's
really
popping
to
the
top
of
the
radar
so
I.
If
you
do
that,
well,
I,
just
ping
me
and
read
so
I'll,
probably
get
it
reviewed
and
done
as
soon
as
possible.
A
B
I'm
gonna
need
some
help,
probably
which
probably
not
gonna,
get
it
to
right.
It's
because
the
warning
is
a
like
I
had
at
the
point
in
the
agenda
about
this
like.
If
we
have
an
error
is
easier
to
unit
test.
If
it's
a
warning,
I
have
to
do
some
like
changes,
so
yeah
I'm
gonna
pick
people
to
review.
It's
like
the
team.
What
is
the
like?
The
deadline
for
this
particular
type
of
PR
I.
A
It
doesn't
really
classify
as
bug
fix
it's
more
like
a
more
clean
up
than
anything
else.
So
Friday,
it's
got
to
be
in
by
Friday,
okay
Oh.
After
this
Friday,
only
critical
bug
fixes
will
be
able
to
go
in
because
I
believe
I
think
we'll
have
one
week
to
actually
get
bug
fixes
in
and
then
after
that,
only
critical
bug
fixes.
So
the.
A
D
D
A
A
We
should
probably
punt
on
both
the
CSR
PR,
as
well
as
the
actual
sub-command
flag,
override
PR,
if
we're
being
strict
now
for
a
little
looser
that
we
could
probably
probably
get
both
of
them
in
I,
don't
see
a
problem
getting
them
in,
but
it
depends
upon
as
a
sig.
If
we're
willing
to
do
this
or
not
it's
it's
a
very
minor
feature,
but
it's
a
it's
a
it's
a
requested
one
from
the
wild.
So.
E
D
F
Your
my
feeling
is
that
the
CSR
change
by
itself
is
pretty
low
risk
and
if
it
was
just
that
I
would
say
we
should
move
forward
because
it's
a
you
know
pretty
isolated
code
path
but
refactoring,
flag
management.
That's
a
pretty
invasive
change,
there's
going
to
be
edge
cases
and
those
edge
cases
are
sort
of
by
the
nature
of
flag
management,
a
large
feature
area
to
test
so
I.
Think
if
it's
all
or
nothing,
then
we
should
go
nothing.
F
F
A
F
B
Couple
of
things
I
wanted
to
say
about
the
CSR
PR,
it's
probably
better
to
remove
the
utilities
we
want
to
add
to
client
go.
Did
this
wasn't
approved
to
my
understanding
by
the
people
who
reviewed
it
and
also?
The
second
thing
is
that
the
PR
adds
something
inside
a
function
that
no
longer
exists.
We
removed
a
function
which
called
to
run.
E
B
F
We
can't
I
put
them
there,
because
that's
where
all
of
the
existing
code
was
and
usually
when
I'm
making
a
change
like
this
I
try
to
make
it
as
unobtrusive
style-wise
as
possible.
But
I
can
just
yank
that
stuff
out
and
move
it
into
our
own
library.
No
problem
yeah.
B
A
Sense
to
me,
I
will
get
rid
of
that.
So
I
think
the
path
forward
is
I.
Think
four
beats
just
on.
First
pass:
swagger
breaches
PRC,
it's
pretty
straightforward.
We
can
review
that
like
immediately
after
this
meeting
and
then
then
Liz
will
have
to
rebase
changes
on
that.
We
can
review
the
CSR
PR
after
the
rebase
because
there's
a
couple
other
shifts
to
and
if
it's
minimal
and
non
evasive
we
accept.
If,
if
we
feel
like
there's
still
issues,
then
we
just
want:
does
that
seem
like
a
reasonable
path
forwards.
D
Reading
through
the
new
policy
checking
what
it,
what
was
the
current
situation
in
open
mean
before
and
I'm,
proposing
a
solution
about
when
and
where
we
have
given
a
warning
for
a
user
that
use
addaperle
are
a
remove
the
feature
gate
if
we
can
find
agreement.
This
is
something
that
probably
we
need
to
have
into
the
release.
I
hope
a
small
change
but
I
have
to
before
I
before
I
want
to
figure
out
what
exactly
have
to
do.
So,
if
you
can
ever
look
at
the
issue,
I'll
link
it
in
the
I'm.
A
Looking
at
the
issue,
I
already
issue
in
detail,
I
think
what's
weird,
though,
is
that
feature
gates
by
definition,
are
not
g8
things
there?
Well,
that's
not
asteroid.
True,
there
could
be
g8
feature
gated
items,
but
those
are
separate
features
as
they
exist
wholly
inside
of
communities
and
the
they
are
also
opt-in,
behaviors
right.
A
So
if
they're,
opt-in,
behaviors
and
not
the
defaults
for
standard
command
line,
flags
I
believe
there
were
still
even
like
this
weird
policy
question
we've
had
with
regards
to
feature
gates
with
upstream
and
cig
architecture.
So,
like
we
had
this
like
nerd
fight
regarding
what
does
it
mean
to
be
a
g8
feature?
Gated
thing
it
should
be
defaulted,
is
the
answer
and
it
shouldn't
have
standard
command
line
arguments
for
the
things
that
means
otherwise,
by
definition,
it's
not
technically
a
g8
well-distributed
thing.
D
A
A
It's
I
think
we
can
probably
make
a
pretty
strong
argument
there
if
we
need
to,
and
if
somebody
other
people,
if
somebody
else
goes
to
have
a
question
about
it,
there's
it's
weird
policy
there
to
say
that
we
have
long-standing
life
cycle
support
for
things
that
are
feature
data,
because
the
the
life
cycle
is
question.
Mark
they're,
not
they're,
not
well.
You
have
to
opt
into
that.
Behavior
I
still
achieve.
B
Something
important
here
is
that
the
the
future
gate
was
already
true
for
a
couple
of
releases.
The
coordinates
feature,
gate,
I,
think
that's
the
biggest
problem
human
colonists
feature
here.
It
was
true
for
for
a
cup
of
releases,
so
we
are
only
including
users
that
have
this
feature
gate
default
for
some.
D
In
my
opinion,
the
function
that
the
feature
is
graduated
because
now
it
you
have
the
toggle
in
your
config,
so
the
underlying
feature
is
graduating.
For
us
it
is
beta
and
it
is
bit
above
the
toggle
dead
and
recorded
and
Cardenas
as
a
compound.
The
problem
is
that
they
feature
gate
was
crucified
EDG.
So
it
is
a
really
away
problem.
We
can
make
argument
for
all
day,
for
any
position
is
just
to
agree
on
an
on
a
common
position
and
and
then
event
will
implement
a
PR
for
making
this
happen.
A
A
We
should
have
marked
it
as
deprecated
in
the
beginning
once
we
promoted
it
to
be
the
default,
but
we've
been
pretty
consistent
on.
You
know
denoting
that
this
is
the
default
and
it's
in
all
of
our
documentation
everywhere,
I
think,
given
them
the
other
changes
to
the
config
and
weighing
the
priority
of
what
it
means
to
add
that
behavior
back
I
would
rather
just
remove
it,
and
if
we
need
to
suffer
the
slings
and
arrows
they
can
come
at
me.
I.
B
Think
that
we
have
the
excuse
that
it
has
been
GA
for
a
couple
of
releases
already
but
I'm,
okay
with
removing
that,
if
you,
if
you
guys
want,
we
can
also
keep
it
deprecated
for
a
couple
of
more
releases,
but
it's
not
going
to
serve
any
functionality
behind
it,
probably
best
remove
it.
In
my
opinion,
perfect.
A
Removing
it
is,
is
prudent
and
it's
also
an
add-on
right.
So
like
the
beauty
of
the
add-on,
is
that,
if
you
want
to
you,
can
always
just
you
know,
change
the
manifest,
because
it's
a
separate
step,
you
know
you
can
patch
apply
an
updated
manifest.
So
for
that
particular
feature,
that's
even
better
right,
because
it's
an
add-on,
you
can
do
it
after
post
and
up
and
there's
there
is
still
an
escape
hatch
for
everyone
else
that
wants
to
set
up
their
clusters.
Then.
B
D
A
We
can
also
we
can
point
them
on
the
happy
path.
There's
two
happy
paths
for
them
to
update
and
especially
as
its
GA
this.
This
will
be
a
transition
issue.
It's
a
very
transitional
issue
because
we
actually
helped
to
do
the
upgrade
path
smoothly
to
two
core
dienes.
It's
only
the
small
subset
of
folks
who
really
wanted
cube
dns
for
some
reason
for
multiple
cycles.
B
F
B
Yesterday,
I
joined
the
first
ever
I
think
kubernetes
long
term
support
meeting
and
one
of
the
main
topics
like
between
team
peppa
and
jordan
legit
was
that
we
need
a
mechanic
to
run
occasional
surveys,
and
this
is
something
that
we
can
do
unless
next
say
what,
by
the
way,
I
think
the
coordinates.
People
already
have
a
service
or
I
can
ask
them
if
they
have
some
numbers
of
this,
but
yeah
mechanic
is
definitely
needed
for
this.
So.
B
A
E
A
So
anybody
who
is
supported
by
app
do
we
collect
that
data
and
we
know
what
Flags
existed,
because
it's
part
of
your
config,
that's
uploaded,
so
we
know
it
so
certain
config
Maps
are
automatically,
can
automatically
be
it
sort
of
always
very
configurable
and
that
you
can
gather
whatever
data
you
want
to.
It's
actually
can
probably
be
a
little
bit
too
powerful
for
some
people
is
liking.
But
if
you
offer
support
services
on
stuff,
you
can
collect
whatever
data
and
find
it
into
the
contract.
A
So
it's
not
part
of
scanner,
because
we
don't
want
people
to
be
creeped
out
by
that.
So,
like
the
free
services,
such
as
scanner,
could
potentially
just
have
a
selective
catch
on
things
like
the
cube
ATM
config.
So
if
we
did,
you
know
we
could.
We
could
have
a
volunteer
clause,
obviously,
of
course,
on
the
front
end
of
scanner.
That
would
basically
say
like
we're.
Gonna
collect
whatever
knobs
you
use
to
configure
your
cluster
just
so
that
we
have
an
overall
view
of
how
people
are
using
and
setting
up
their
environments.
A
A
A
D
One
question
that
I
am
I
didn't
have
time
to
dig
into
button.
He
seems
to
me
that
there
are.
There
is
a
growing
number
of
fish
and
requests
in
the
boffin
in
issue
in
the
Cuban
media
reported,
and
the
requests
in
the
insula
is
like
about
setting
up
each
a
cluster
following
the
Lister
destruction.
On
the
on
the
side.
A
Think
the
instructions
are
hard
to
follow
for
most
folks
as
many
there's
many
steps
and
anything
that
has
more
than
like
one
of
one
of
the
greatest
things
of
Kubb
ADM
is
that
we've
reduced
the
number
of
steps
to
enjoying
for
most
folks,
so
the
two
but
I
think
because
we
have
multiple
steps
that
are
involved.
It's
easy
for
people
to
skip
a
step
or
to
do
something
custom
along
the
way.
I
do
see,
I
think
more
people
are
using
it
because
we
have
the
instructions
out
there
in
front
and
center.
D
You,
like
the
background
that
the
baton
question
is
that
if
people
are
having
problem
because
the
procedure
is
complex
or
if
there
are
underlying
an
issue
or
bad
that
are
surfacing,
but
I
will
have
a
person
on
the
issue
and
try
to
understand,
because
all
these
issue
are
very
very
low
distraction
and
very
very
long
story
of
Friday's
duty.
Do
that
really
difficult
to
follow?
Well,.
F
F
F
Sorry
see
and
I
not
see
all
right,
Oh
too
many
acronyms
Jimmy
yeah.
B
G
From
my
noticing,
the
issues
have
been
tracking.
That
page
tend
to
pay
attention
to
that
page.
The
complex
setup
is
when
we
get
people
to
install
stacked
masters
and
I.
Think
because
the
the
external
@cd
story
is
a
lot
easier
and
was
built
first
for
ku
Vidya.
If
we
had
a
similar
mechanism
for
stacked
masters-
and
we
reduce
the
number
of
steps
required
on
that
page
I
think
we
would
agree
would
see
the
issues
drop,
but
I
guess
it's
all
speculation
at
this
point.
B
A
B
A
Is
it
I
think
just
having
the
promotion
of
GA
in
having
at
least
well-defined
subsections
of
sub
commands
that
are
alpha
and
the
config
being
beta
is
actually
way
better
than
what
we've
been
in
the
past?
So
it's
very
clear
to
consumers
like
any
command-line
options
that
we
have,
that
aren't
explicitly
marked
as
alpha
or
beta
are
very
clear
that
we
support
these
things,
so
upgrade
is
a
well-defined
command
that
explicitly
states
that
is
supported
and
for
saying,
Kubb
ADM
is
now
GA.
A
A
A
A
But
I
just
I
do
think
like
in
the
next
cycle,
as
we
start
to
ramp
down
this
cycle,
maybe
at
around
the
Kukuanas
timeframe,
you're
right
before
coutdown
I
think,
like
maybe
the
week
before
group
con,
we
can
start
talking
about
114
cycle
and
I,
do
think
finishing
off
the
sub-command
shuffle
I
may
be
V
1
beta
2,
depending
upon
feedback.
We
get
and
experimental
control
plane
are
probably
the
most
common
past
for
things
that
that
I
that
I
see.
B
A
A
F
F
B
F
B
A
A
A
B
B
A
A
B
B
A
We
can
probably
take
that
particular
one
that
particular
question
and
you
do-
and
this
can
can
take
a
look
at
that
I
want
to
I
want
to
make
sure
we
go
through
the
rest
of
the
backlog
here.
So
at
least
triage
we
have
everybody
assigned.
That's
working
on
calico
pod,
subnet,
cider
docks
for
comedian
is
felling.
Updating
do
that.
A
B
A
Unfortunately,
yes,
there's
there's
extra
documentation
that
people
well
there's
some
documentation
that
we
need
to
update
it.
There's
not
much.
There's
been
a
ton.
I
did
I
did
I.
Think
I
put
a
block
on
a
couple
of
PRS
where
people
wanted
to
add
the
here's.
How
you
do
things
in
the
backwards
way?
I
said
that
to
make
any
sense
so.
B
A
B
A
D
F
Yeah
I
started
that
yesterday,
after
you
I.
A
A
E
A
A
D
B
B
A
A
It's
there
were
so
many
different
aspects
of
reset
that
didn't
work
right.
It's
not
idempotent
on
a
number
of
fronts,
later,
there's
iptables
rules
and
we
have
like
a
basically
instructions
now,
the
other
serial
inertia
is
definitely
so
I'm.
Okay,
with
closing
this
one
because
we
have,
but
we
have
a
separate
issue
for
the
CRI
and
we
have
instructions
on
how
to
clean
up
the
proxy
goo.
That's
left
over.
A
A
This
particular
one
is
definitely
have
to
go
to
114.
We
need
test
coverage
for
a
che
configurations
and
a
bunch
of
other
configurations.
Liz
started
a
test
suite
we
need
to
once.
We
actually
have
an
H,
a
deployment
tool
in
place,
whether
that
be
coop
spray
or
the
AWS
provider
for
cluster
API,
either
A
or
B.
We
need
to
enable
that
test
suite
as
part
of
stand
up
and
that
standard
periodic
jobs
that
we
that
we
track.