►
From YouTube: Kubernetes SIG Network 20171102
Description
SIG Network meeting from November 02, 2017.
A
B
I,
don't
have
much
to
say
on
that
other
than
there
is
no
longer
like
a
someone
on
call
effectively
for
curbing,
ladies
to
help
make
sure
the
cement
queues
are
filtering
through
and
there's
not
too
many
tests
backing
up.
There's,
of
course,
a
lot
of
people
that
just
do
that,
naturally,
especially
if
they're
in
the
sig
testing
group,
but
it
has
been
delegated
now
to
each
sake,
to
make
a
bigger
effort
to
take
a
look
at
the
stability
and
and
they're
flakes
as
far
as
their
and
ends
go
and
everything.
B
So
I
posted
a
link
in
there
of
the
dashboard
that
we
have
and
made
a
couple
of
additional
updates.
The
last
two
in
there
for
the
Calico,
the
flannel
they're
not
actually
owned
by
us,
but
I
put
them
in
there
kind
of
as
an
FYI
for
individuals
that
are
interested
and
if,
for
whatever
reason,
cuvee
DM
or
something
broke,
they
see
and
I
plugin
that
they
own
and
there's
a
mailing
list
also
attached
in
there.
That
will
be
populated
with
test
failures
on
a
given
schedule.
I
think
three
is
the
default.
B
So
if
there's
three,
those
in
a
row
for
a
given
test,
suite
it'll
send
a
message
to
that
group
and
so
far
it's
been
a
little
busy.
So
I've
had
this
at
the
digest
mode,
but
that
just
means
that
we
need
to
take
a
look
at
these
flakes.
I
know
that
there's
a
few
that
have
been
fixed
and
there's
a
couple
that
are
kind
of
higher
priority
in
the
Alpha
features
group
that
involve
I,
think
performance
tests,
type
things
like
general.
B
A
Yeah
so
I
think
in
my
my
mind
you
know
they've,
there's
no
longer
this
kind
of
centralized
person
policing
this.
So
we
need
to
make
sure
that
we
are
doing
our
part
to
keep
these
tests
screen
and
question
I
would
raise
is
do
we
think
we
have
the
right
process
today,
or
should
we
be
defining
a
process
to
make
sure
that
that
happens?
C
D
D
B
E
C
F
H
B
D
B
F
D
D
D
D
D
G
F
Interests
me:
okay,
I,
don't
think
we
have
I
think
everyone
should
take
a
look
given
the
complexity
of
this
and
we
can
probably
next
meeting
or
send
out
an
email
as
to
how
should
we
do
this?
Probably
we
should
take
that
summary
put
it
in
a
doc
and
then
sort
of
people
who
feel
like
they
own.
A
piece
of
it
should
put
a
little
blurb
saying
like
oh,
this
I'd
realize
this
and
it's
because
of
X
or
maybe
if
the
entry
doesn't
have
anything
there,
then
then
we
talk
about
it.
A
I
G
Testing
like
we
talked
about
it
more
or
less,
every
cycle
and
I
mean
every
it's
not
just
our
say.
Every
every
sink
has
this,
but
at
some
point
we
reach
a
point
when
everybody's
looking
everybody
else
to
do
the
work,
because
we're
all
busy
doing
other
things
like
do.
We
need
something
more
heavy-handed
to
try
to
get
this
back
in
order.
We
have
a
lot
of
failing
tests
where
last
I
looked.
We
were
among
the
worst
of
this
things
like
do.
D
G
A
It
has
been
several
cycles
that
I
think
we've
had
the
same
conversation.
Alright.
G
We're
coming
up
on
code
freeze
in
two
weeks,
so
by
the
time
we
meet
again,
it
will
be
literally
code
freeze
week
date.
Probably
right
it's
the
third
day,
so
we
don't
have
any
more
opportunities
to
spend
between
now
and
code.
Freeze,
so
I
expect
that
in
the
next
two
weeks
everybody's
gonna
be
pushing
like
hell
to
get
their
PRS
in
and
not
focusing
that
much
on
tests,
and
then
we've
got
a
relatively
short
stability
window
before
we
try
to
cut
a
release
in
December.
G
A
E
E
Yes,
we're
having
the
opportunities.
The
original
proposal,
so
Google
Docs
under
fo
is
a
new
proposal
is
ready.
We
were
broadcasting.
Is
me
released.
G
J
G
G
No,
probably
by
the
time
we
go
beta,
we're
committing
that
you
probably
are
going
to
do
this
so
I
think
it
made
a
time
is
probably
worthwhile
to
move
it
up.
The
downside
of
moving
it
to
GCR
is
that
we
can't
yet
hand
out
the
ability
for
people
to
push
on
their
own,
and
so
you
know
once
you
put
it
there,
then
you
have
to
wait
for
somebody
who's
on
the
short
list
of
people
to
do
the
push
lever
and
so
you're
better
off.
Not
do
that
until
you
have
to
all.
A
Cool
so
next
topic,
I
added,
which
was
just
an
update
on
keep
ten
sessions.
So
we
talked
a
little
bit
last
time.
We
actually
have
two
different
sessions
scheduled,
which
I
don't
think
we
had
talked
about.
Last
time
we
have
the
Signet
work
update,
which
is
an
80-minute
speaking
slot,
and
we
have
the
Signet
work
deep-dive
session,
which
is
more
of
a
kind
of
roundtable
get
together
and
talk
about
some
hard
problems.
A
A
A
G
A
G
It's
not
like
thinking
about
it
will
actually
have
any
shortage
of
things
to
talk
about,
but
I'm
not
sure
how
much
deaf
people
who
aren't
already
on
these
calls
will
really
want.
Out
of
this
right.
We
could
go
into
great
detail
talking
about
the
fund.
It
is
ipv6,
but
I.
Don't
think
anybody
outside
this
group
who's
gonna
care
that
deeply,
except
for
the
headline
right
working
on
ipv6.
D
A
A
G
A
D
G
Yeah
I
think
it's
probably.
We
should
spend
some
of
the
time
going
back
through
all
the
things
we've
done
this
year,
but
I
think
as
an
update.
I,
don't
know
if
retrospective
is
the
best
use
of
so
maybe
you
know
assume
we're
shooting
for
14,
not
18
I
would
spend
no
more
than
10
on
the
retrospective
and
then
the
rest.
G
A
Next
time
so
I
think
that's
yeah,
that's
the
kind
of
trickier
one.
We
do
also
have
this
deep
dive
session,
which
I
know
we
had
talked
about
kind
of
doing
in
an
ingress
forum
as
part
of
that,
and
there
probably
other
topics
as
well.
It
would
be
good
to
go
into
this
with
knowledge
of
some
of
the
things
we
want
to
be
talking
about
and
the
last
time
it
all
kind
of
was
on
the
fly
which
which
worked
out
but
I
think
we'll
get
more
done.
F
A
Under
it,
so
last
time
we
had
this,
a
lot
of
people
showed
up
I
think
this
was
Berlin
and
there
were
like
a
hundred
people
or
something
in
the
room.
So
we
ended
up
splitting
into
you.
Multiple
multiple
sessions
like
subgroups,
that
talked
about
specific
topics
and
we
broke
the
meeting
into
two
halves.
A
G
J
G
C
D
G
G
L
Hi,
this
is
Prateek
sorry
to
interrupt
on
the
previous
topic.
I
just
wanted
to
bring
up
so
I've
been
working
on
this
tool,
it's
very
early
stage
and
I'm,
not
sure.
If
that
will
be
enough
interest
on
this.
It
sort
of
helps
it's
based
on
CR
D
and
what
it
helps
is
like
user
can
create
an
object
and
say:
I
want
to
trace
one
particular
part.
So
what
I
have
been
facing?
The
issue
is
like
in
random.
L
This
path
is
traced
from
here
to
here
and
it'll
chain
like
debugging,
where
the
problem
is
on
the
networking
path
between
two
hosts
on
the
same,
but
two
parts
on
the
same
host
or
maybe
on
different
coasts.
So
if
there
is
any
interest,
I'm
very
early
state-
so
maybe
we
can
talk
about
that
later.
But
that
sounds.
G
A
Yes,
the
next
was
validating
duplicate
target
ports,
so
Chris
raised
this
and
I
agreed.
We
should
talk
about
it
just
to
try
and
reach
an
agreement.
Quicker,
Tim
I
know
you're,
aware
of
this
one
yeah,
so
it
looks
like
in
1.8
we
added
some
extra
validation
on
services
that
have
broken
some
yeah
use,
slash,
abuse
cases,
yep.
G
So
they
have
80
and
443
both
to
the
same
target
port,
even
though
the
sort
of
that
the
incoming
protocols
were
different,
the
back
end
protocols
were
the
same
and
well
that's
sort
of
an
abuse
of
service.
It's
not
egregious,
and
it's
existed
for
far
longer
than
this
idea
has
so
I
guess
my
feeling
is.
We
have
to
roll
that
back
and
rethink
it.
A
A
G
G
F
A
K
Oh
yeah
I
have
some
mic
issues
before
related
to
the
cube
concession,
so
we
are
currently
working
on
your
cue
proxy
implementation
in
user
space
using
VPP
I,
don't
know
whether
this
topic
will
be
interested
in
the
Yukon
session
or
it's
a
will
be
a
separate
sig
able
to
contact
me
I.
Think
it's
Q
proxy
will
be
interested
a
networking
session
as
well
I.
Think
so
so
this
is
it
okay
to
bring
up
the
decision
or.
A
K
K
A
A
N
N
G
Be
part
of
that
conversation
I
think
we
can
try
to
get
a
hangout
or
a
zoom
or
something
running,
but
I'm
wary
of
what
we're
committing,
because
life
is
notoriously
not
good.
Although
again
Kahn
did
respond
to
my
Twitter
statement
of
that
same
effect,
saying
they're
spending
a
lot
of
energy
to
make
sure
the
Wi-Fi
at
this
conference
is
really
good.
We'll.