►
From YouTube: Kubernetes SIG Network meeting 20200430
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
A
A
A
Which
one
is
that
there
we
go
share
that
how's,
that
can
everybody
see
yep,
that's
good,
excellent,
all
right,
so
I
went
through
and
loaded
up.
All
of
the
unresolved
I
actually
went
through
a
few
before
and
either
addressed
them
or
or
closed
them.
It
looks
like
a
few
other
people
were
doing
the
same.
So,
let's
see
this,
is
it
the
entirety
this
this
set
of
tabs
is
the
entire
unresolved
triage
queue.
A
A
Procedurally,
we
don't
remove
the
triage
unwritten
unresolved
label
until
we're
sure
it's
a
bug
or
feature
that
we
want
to
keep
and
if
it's
not
something
we
want
to
keep,
then
we're
gonna,
try
to
aggressively
close
issues
that
we
can't
reproduce
on
the
assumption
that
they
can
always
be
reopened.
If
somebody
else
comes
along
and
can
reproduce
them
that
helps
to
keep
the
list
a
little
bit.
A
Or
you
punt
to
another
saying
or
something
and
every
sing
as
different
modes
I
actually
have
not
gotten
an
update
on
the
automatic
triage
state
machine
that
they're
working
on
for
contra
vex,
but
the
last
cap
I
saw
looked
like
it
was
gonna
help
us
out
a
little
bit
by
making
the
triage
label
more
automated,
so
jumping
in
cluster
recovering
failure,
905
nine
four
not
assigned
to
anybody.
This
user
did
something
like
turned
off
their
cluster
and
is
complaining
that
when
it
came
back
up,
there
was
something
wonky
with
networking.
A
A
D
F
A
A
G
E
A
A
A
G
F
C
F
G
E
F
The
dude
from
the
same
Lord
you
can,
you
can
expose
their
this
source.
You
know
you
can
create
a
a
pocket
with
the
solved
a
one,
two,
seven:
zero,
zero
one.
You
know.
But
what
does
that?
Let
you
do
well
God,
let
you
you
are
not
going
to
receive
the
packet
back,
because
the
server
is
going
to
to
return
to
their
local
host.
Well,
III.
I'm,
sorry
well,
I
think
that's!
This
is
the
attack,
but
I
come.
A
C
C
D
F
C
A
I'm
surprised
that
the
target
kernel
will
actually
successfully
respond.
I
guess
that's
what
that
enables
okay,
Wow,
okay
I
have
some
vague
recollection
of
something
some
other
corner
case
in
iptables
that
needed
this,
but
I
honestly
cannot
remember
what
I
like
the
idea
of
just
opening
a
bug
and
seeing
what
breaks
volunteers
are
opening
a
PR
rather.
C
A
A
A
A
F
B
F
Predestination
and
port
destination-
and
my
guess
is
the
contract-
is
not
allowing
one
of
the
connectors
but
its
ability.
You
need
to
drill
through
and
digest
it's
a
lot
of
information,
but
this
world
is
cry
and
it
will
commented
also
Lars
commented
in
in
another
one
that
is
very
similar
and
it's
it's
not
the
same,
but
I
think
that
the
discussion
is
is
very
informative.
F
I
C
J
D
A
So
I
know
that
there
are
folks
who
have
done
things
like
expose
their
docker
registry
on
a
node
port
and
then
access
it.
That
way
that
you
know
they'll
do
their
image
pulls
from
local
host
:,
three
thousand,
whatever
thirty
thousand
whatever,
and
that's
their
established
API
to
their
private
registry.
That
avoids
the
TLS
the
need
for
TLS
to
the
registry,
which
may
or
may
not
be
a
good
idea,
but
is
being
done
anyway.
D
A
A
All
right,
I
will
I'll
read
this
one,
because
I
think
it's
interesting
and
yeah.
It
aligns
with
the
other
problems
cool
all
right.
Let's
call
that
a
wrap
for
triage
there's
more.
If
people
wanted
to
look
at
them,
but
not
a
ton
more
and
we're
getting
into
the
ones
we've
seen
already
all
right
agenda
today,
issue
triage
time
box,
ten
minute
discussion
on
dual
stack
services:
I'm
assuming
that's
you
Cal
I,.
H
H
The
first
one
will
be
four
or
six
just
like
politics
all
right.
There
are
a
couple
of
questions
that
we've
raised
in
the
first
two
way.
He
just
and
specific.
Those
are
the
two
pages
we
need
to
help
either
you
comment
on
or
ask
your
users
for
that
specifically
round
two
questions.
What
will
happen
after
you
convert
the
cluster
to
dual
stack?
Should
you
could
expect
your
existing
services
to
automatically
be
to
list
act
or
not?
H
To
up
see,
I
want
this
cluster
to
be
a
dual
stack
right,
even
without
the
hope
seaboard
should
the
existing
service-
let's
say
nginx
website
should
start
working
with
an
internet
controller
on
top
of
our
website
should
I
start
to
automatically
be
converted
to
a
dual
stack
as
an
existing
service
will
carry
another
IP
or
not.
The
second
part
of
this
question
of
the
next
question
we're
trying
to
answer
and
is,
let's
assume
for
the
sake
of
discussion,
my
user
and
I
want
to
create
a
new
service.
H
I
have
an
option
of
providing
a
family
or
not
right
if
I
provide
it
if
I
did
not
provide
a
family
specifically
if
I
said
IP
family
is
null
right
field
should
the
services
that
will
be
created
as
a
result
of
my
action.
Should
it
be
a
single
family
according
to
the
cluster
default,
or
should
it
be
a
dual
stack
for
the
Torah
stack
service
to
solve
primarily
the
two
questions
that
were
asked?
What
will
happen
to
listing
services
after
though
after
agreed
and
what
will
happen
to
new
services
with
an
L
family.
A
If
I
can
modify
this
statement
just
a
tiny
little
bit,
let's
even
take
out
the
presumption
that
there's
a
normal
family
say
you
don't
say
anything
about
IP
family,
what
happens
and
clearly
there's
been
so
thank
you
forced
all
to
the
people
who
have
already
commented,
there's
a
fair
number
of
differing
opinions
already,
which
is
interesting
and
every
time
somebody
posts
their
own
story
with
it.
I
get
more
insight
into
what
people
are
thinking.
I
think
we
need
more
need
more
perspective.
D
A
That's
a
fair,
that's
true
and
that's
always
happens
with
these
docks.
I
guess
I
didn't
really
intend
for
conversation
to
happen,
but
I
should
have
known
that
they
would
so
one
thing
people
can
do
is
just
add
a
suggestion
like
at
the
at
the
bottom
of
the
second
page,
before
the
like.
Don't
read
below
this
line
line.
Just
add
your
own
responses
there
as
a
suggestion
that
way,
it'll
be
attributed
into
your
name,
but
everybody
can
still
see
it.
F
A
So,
there's
a
lot
of
corner
cases
that
only
pop
up
when
you
upgrade
when
you
enable
the
gate
and
convert
to
dual-stack
at
the
same
time,
let's
ignore
those
corner
cases.
Let's
assume
for
that
thing.
For
the
conversation
here,
you
have
a
cluster
with
the
the
feature
is
GA
right
it.
The
gate
is
enabled
you
have
it
in
single
stack
mode.
Now
you
want
to
convert
to
dual
stack.
Maybe
we
should
stop
saying
upgrade
and
say:
convert.
H
So
through
two
of
two
opposing
I
think
probably
three
posting
news
right
now
is
around
around
ddddd.
Come
from
the
fact
that
okay
I
haven't
a
new
service,
I
didn't
know,
family
can
eat
a
service.
If
the
target
cluster
is
dual-stack,
then
it
should
be
a
dual
stack.
So
that's
one
family
of
opinions
right,
the
other
family.
H
A
That's
really
it
like.
It
became
clear
after
we
had
gotten
this
far
with
it
that
for
at
least
some
people,
we
had
missed
what
their
expectations
were.
I'm,
not
saying
that
you
know,
Clayton's
expectations
were
necessarily
correct,
but
that
we
didn't
get
what
he
expected
and
so
I
want
to
understand.
If
he's
in
the
norm,
or
if
he's
the
oddball
and
I'm,
absolutely
confident
that
whatever
we
decide
is
the
right
thing
to
do,
we
can
figure
out
how
to
do
safe,
conversions
and
upgrades,
but
we
need
to
get
agreement
on
what
is
the
right
thing.
H
So
I
never
came
into
this
meeting.
Thinking
that
we'll
gain
consensus,
all
right,
I
just
wanted
to
make
sure
that
everybody
goes
through
the
talk,
but
we
understand
what
we're
trying
to
achieve
here
and
we
want
to
take
this
thing
right.
The
thing
that
also
I
will
just
I
said
it
and
I
didn't
see
a
lot
of
tears
or
agony
or
shouts,
which
was
multiple
cluster
IDs,
which
is
impressive.
All
right.
C
H
H
A
You
truthfully,
hopefully
we're
talking
about
10
to
20
minutes
of
your
time.
It's
you
shouldn't
need
to
write
a
book
on
this.
All
right.
Next
topic
suggested
five
minutes,
but
we'll
see
standardized
issue
dealing
policy
is
90
days
too
much
for
waiting
for
a
users
response
Oh,
for
this
is
for
waiting
for
user
timeouts
on
issues.
Are
we
supposed
to
direct
support,
requested
slack
and
Stack
Overflow
I?
Don't
know
who
wrote
this,
but.
K
Jj
ruled
in
it
in
this
light
channel.
This
was
something
else
so
that
probably
arisen
talking
with
Antonio.
The
thing
is
that
we
have
a
lot
of
issues
waiting
for
some
some
some
user
response.
Like
this
happened.
Ok,
can
you
provide
me
more
information?
No,
and
it
keeps
waiting
like
in
30
days,
40
days
and
so
on,
and
then
we
wait
to
the
bots
to
rolls
and
then
to
close
the
day
issue.
K
So
the
question
is:
if
we
want
also
to
to
reduce
the
number
of
the
issues
to
do
something
that
directors
to
the
important
one.
If
this
is
this,
isn't
too
much
like
waiting
for
force,
the
user
to
provide
information
and
also
I've,
been
I've,
been
doing
some
some
issue
triage
and
seeing
a
lot
of
questions
that
are
already
answered
in
in
Stack,
Overflow
or
in
discuss.
I
know
that
you
are
dealing
with
we
stiffened
to
to
improve
the
the
issue
triage,
so
I'm,
just
just
asking.
A
We
should
probably
document
them
under
community
somewhere,
but
so
the
way
I
see
the
the
fate
ABAT
the
90
days
thing
is
really
the
safety
net
underneath
everything
like
if
nobody's
touched
it
in
90
days,
it's
just
gonna
go
away
right.
It's
clearly
not
important
to
enough
people
to
keep
around
I
have
generally
kept
to
a
30
days
policy,
two
to
two
cycles
of
triage.
A
I
H
A
I
A
A
I
A
A
J
A
A
D
A
F
A
I
If
there's
anything
else,
there
is,
you
know
Dan's
here,
Casey's,
not
here,
I,
don't
know
if
there's
any
other
issues
on
that
our
Valley
Network
policy
validation
kept.
If
there's
anything
else,
we
could
discuss
them
now.
I,
don't
think,
there's
anything
left
then
so.
But
it's
been
kind
of
lingering
for
a
while
and.
E
B
I
H
A
B
I
K
I've
seen
to
the
mailing
list
and
I've
set
up
also
a
shared
calendar
after
the
meeting,
because
my
meeting
invite
was
wasn't
working
I'm
going
to
send
again
the
meeting
the
meeting,
invite
the
the
calendar
invite
and
you
can
take
a
look
Mike,
but
that's
happening
every
we
are
still
deciding
if
this
is
bi-weekly
or
not,
but
that's
in
every
Monday
1
p.m.
Pacific
time.
Thank.