►
From YouTube: Kubernetes SIG network meeting 2019-08-22
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
C
B
C
So
I,
do
you
guys
see
my
screen
yep
all
right,
so
I
put
everything
I
could
in
one
one
entry
right
under
the
agenda?
Will
we
stand
today?
Is
the
core
service
endpoints
work,
because
I
think
is
almost
done
with
waiting
on
them
for
one
final
review
for
a
change
and
B
the
way
we
default
things
for
for
upgrades
downgrades
and
all
of
that
the
proxy
proxy
errs
work.
C
There
are
two
to
two
parallel
PRS
one
is
at
I
think
completion
stage,
just
bending
reviews,
I
have
done
my
own
review
on
it
and
it
could
used
the
pair
more
eyes
two
additional
eyes
on
it.
The
IV
tables
one
is
still
in
progress,
so
I'm
here
I'm
talking
about
this
one.
Now
all
right,
that's
that
one!
If,
in
order
to
merge
an
old
support
services,
doing
the
whole
stack
on
kubernetes
beyond
16,
we
need
the
core
one,
the
first
one,
and
at
least
one
of
the
other
two
all
right.
C
Ultimately
speaking,
we
need
the
other
tools,
but
one
should
be
enough
for
us
to
provide
alpha,
which
the
rest
is
really
a
catch-up
work
from
from
phase
one.
Now
there
is
a
PR
to
support
multi
siders
for
cluster
cider.
So
if
you're
not
aware
of
what's
happening,
we
used
we
used
to
support
only
slash
24
for
v6
and
obviously
that's
a
truly
big
cider
for
for
not
cider,
so
that
one
actually
supports
multiple,
multiple
sizes
for
four
nodes.
C
C
There
is
a
brow
job
PR
for
the
HIV
test,
and
we
need
to
merge
that
there
is
the
agile
lb
support
for
people
running
clusters,
I'm
hoping
that
we
do
this
as
as
like.
Okay,
this
is
how
work
our
provider
will
do
it
alright,
as
I
started,
so
everything
is
behind
the
feature
flag.
Obviously,
the
last
thing,
which
is
also
needed
for
for
the
complete
support
is
people
will
need
IP
mask
agent
that
supports
dual
stack
or
ipv6
and
we
merged
the
change.
We
just
need
a
release
over
there,
that's
where
we
are
right
now.
C
C
E
So
hey
this
is
Tim.
Can
you
hear
me
I?
Can
yes,
great
I
can
hear
you
too
I
I'm
spending,
all
my
free
time,
Matt
between
now
and
code
freeze,
doing
PR
reviews
I've
got
at
least
four
that
are
all
touching
cube.
Proxy
I
am
terrified
of
this.
That's
like
air
traffic
control
and
I,
see
four
747s
and
they're
all
going
to
hit
the
same
runway
at
the
same
time.
E
Everybody
who's
dealing
with
anything
cube
proxy
related,
really
needs
to
be
on
the
ball
with
rebuses.
These
are
not
going
to
be
trivial
tree
bases.
Some
of
them
are
going
to
require
you
to
go
back
through
and
patch
your
stuff
into
different
places
and
code
freezes
next
week.
So
I
want
to
encourage
people
to
be
paying
very
close
attention
if
you
want
your
stuff
to
be
landing
in
time
and
I
hate
that
word
down
to
the
last
week
again,
but
at
least
we're
a
week
in
advance
and
not
two
days
in
advance.
E
Okay,
can
you
assign
it
to
me
so
that
I
can
make
sure
I
find
it?
My
usual
loop
that
they
be
giving
me
today
is
open
all
the
PRS
and
issues
that
are
assigned
to
me
and
see
how
fast
I
can
go
through
them.
If
it's
not
assigned
to
me,
it's,
there
are
things
that
are
getting
lost.
I
know
this,
but
it's
the
best
I
can
do
right
now.
Okay,.
A
Cool
thanks
guys,
okay,
any
more
on
dual-stack.
E
C
E
E
F
E
F
E
A
Okay,
any
other
topics.
People
would
like
to
talk
about
in
the
larger
group
or.
B
E
Awesome
question:
thank
you
for
bringing
that
up
yeah.
So
the
context
is
we
have
we're,
adding
a
field
to
a
stable
API
which
is
acceptable.
We
have
a
general
pattern
when
you
do
this
of
checking
whether
the
feature
gate
for
that
field
is
enabled
and
if
not
blanking
the
field
out
on
on,
creates
and
updates.
E
E
We're
sending
the
default
in
a
different
piece
of
the
rest
processing
stack,
which
means
that
it
would
not
be
as
complete
of
a
solution
as
if
it
was
doing
static
defaulting
so
Kyle
and
I
had
a
very
fun
conversation
with
Jordan
Liggett,
who
helped
I
think
helped
us
figure
out
how
we
can
cover
all
the
bases
on
this
cow.
Is
that
right,
he's
nodding,
I,
guess,
I
think
so
at
least
okay?
E
So
so,
if
that's
right,
then
probably
you
have
a
new
one
in
my
mailbox
and
I
will
go
look
at
it
this
afternoon
and
if
that
has
all
the
bases
covered,
then
then
that's
done.
I.
Think
this
case
is
a
relatively
special
case
in
that
we
have
this
weird
defaulting:
I,
don't
I
thought.
Maybe
there
was
a
general
problem
for
API
evolution
changes,
but
I.
Don't
think
that
this
generalizes
a
special
so
there's
that
so
I'll
follow
up
on
the
PR
on
that
today.
C
It
will
become
a
general
general,
generalized
problem
that
they're,
like
dynamic
defaulting
and
so
on
as
the
API
evolved,
because
some
of
the
things
depend
on
cluster
configuration,
ambient
environment
and
so
on,
and
all
of
this
data
is
not
driven
into
either
to
validation
the
default
thing
of
the
API
process.
Everything
is
left
to
controllers
and
or
stuff
like
I,
be
a
locator.
It's
not
a
today's
problem,
but
it's
it's
gonna
be
becoming
prevalent,
as
things
evolve,
I
think.
Well,
if.
E
It
becomes
more
prevalent
than
at
least
we've
got
an
existence,
proof
where
we
think
we
reasoned
through
it
once
and
in
all
likelihood,
we'll
probably
find
something
that
we
missed,
but
we'll
find
that
on
the
second
or
third
person
who
tries
to
do
this,
and
it's
all
alpha
right
now,
so
I
I
think
we're
gonna
be
ok
on
this
one
and
I
actually
I
do
think.
We've
got
all
the
bases
covered
so.