►
Description
Kubernetes SIG Network KPNG Subproject Weekly APAC Meeting for 20220105
A
Hey
hi
everyone,
this
is
5th
jan
2022
and
we
are
at
the
kipping
apec
meeting
before
we
get
started.
This
meeting
is
under
the
cncf
code
of
conduct,
so
it
just
means
be
awesome
to
everyone
else.
Also
this
will
be
recorded
and
put
on
youtube
or
something
that's
also
just
be.
Mindful
of
that,
I
don't
see
anything.
Much
added
to
the
agenda
is
such
a
couple
of
things
that
I
had
in
mind,
one
of
them
being
the
users.
A
These
changes
needed
to
you,
know
the
store
command
stuff
to
register
user
space
sort
of,
and
then
I
try
to
run
the
cupping
local
upscript
that
we
usually
run
the
ci.
B
A
I'm
not
sure
maybe
neha
and
friedrich
can
tell
me
if
I'm
wrong
about
that,
so
also
for
anyone
joining
if
they
just
want
to
try
out
cupping.
Maybe
this
is
the
script
that
you
can
look
at,
which
will
just
get
a
kind
cluster
up
without
q
proxy,
but
have
coping
on
it.
A
So
I
was
trying
to
run
this
script
in
order
to
see
if
user
space
works,
and
then
there
were
a
couple
of
bugs
that
I
noticed
while
building
the
image
and
then
turned
out
that
I
had
to
also
add
some
changes,
some
go
marching
just
to
the
cmd
directory
and
and
now
I
think
that
this
is
almost
ready.
So
maybe
the
next
step
would
be
to
run
cupping
local
up
and
I
think,
by
default,
this
gets
cupping
in
the
nft
mode,
using.
B
A
Backend
so
now
the
cool
part
would
be
to
edit
that
config
map
and
replace
nft
with
user
space
and
see
what
all
fails
and
obviously
things
will
fail,
looks
like
it
is
ip
table.
So
maybe
if,
if
this
is
set
to
user
space
lin,
then
we
should
be
able
to
see
cupping
running
in
user
space
backend
and
things
are
going
to
fail.
I
think
we'll
just
we
just
need
some
volunteers.
A
If
anyone
is
looking
to
hack
around
this,
then
maybe
this
can
be
a
good
project,
but
in
all
you
got
to
do
is
check
out
this
branch
and
run
keeping
local
up
with
user
space
lin
back
end
and
see
what
all
fails
and
then
maybe
create
an
issue
or
tell
us
in
this
pr
or
you
know,
or
just
tell
us,
on
the
slide
channel,
whatever
works.
A
C
I
think
yeah
over
the
holidays.
I
have
not
put
any
work
in
this
at
all.
I
think
douglas.
We
we
refactored
everything
made
it
much
better.
Maybe
he
can
talk
about
this.
Maybe
he
had
some
stuff,
but
no
updates
from
my
side.
At
the
moment
I
have
to
get
back
on
track.
Basically.
B
Yeah
sure,
thank
you
freddie.
Basically,
I
just
did
a
few
chains
on
top
of
the
the
work
that
have
been
done
and
I
test
again
epvs
ip
tables
and
it
seems
the
capping
is
working
well.
So.
A
D
No
not
really
because
I
was
also
out
last
week.
Full
so
did
not
work
anything,
and
so
actually
the
last
update
was
like.
I
was
working
with
anusha
on
this
issue
162.,
so
we
I
mean
we
had
one
call
with
last
time,
but
I
mean
just
to
understand
bit,
but
still
we
need
to.
I
mean,
try,
try
and
test
out
the
things
there.
D
So
so
we
will
continue
our
analysis
on
that
issue
and
we'll
yeah
and
also
like
want
to
understand
few
of
the
refactoring
which
douglas
did
so
yeah.
That
is,
that
will
check
on.
D
I
see
that
there
is
a
unit
test,
also
written
for
session
affinity,
and
I
do
you
know
like
if
the
fix
is
in
also
like
the
I
think
amin
has
some
pr
created,
because
there
was
one
failure
related
to
that
also
an
end-to-end
test.
D
D
E
E
A
See
also
if
anyone
has
joined
this
meeting
for
the
first
time,
I
think
ranish,
where
feel
free
to
introduce
yourself.
G
Yeah
I've
been
like
been
around
in
this
meeting
and
just
trying
to
learn
whatever
I
can
like
recently.
I
just
came
up
with
something
if
you
can
help
me
understand
that,
to
be
great
like
I
was
trying
to
learn
about
calico
cni
and
came
about
this
bgp
body
gateway
protocol
of
calico,
and
I
wasn't
really
sure
how
it
works.
A
F
So
that
was
a
little
bigger
review.
I
think
doug
had
a
look
at
it
and
yeah,
so
he
had
given
many
good
comments.
So
yeah
only
thing
is
yeah
still,
it
is.
You
know,
kind
of
work
in
progress
type
of
code
right
in
the
sense
that
still
we
are
adding
lot
of
functional
code
and
we
are
not
at
much
concentrating
on
okay,
okay.
So
where
do
do
we
add?
Okay,
too
many?
F
You
know
enough
comments
about
the
functions
or
methods
or
you
know
a
lot
of
debug,
logs
or
even
ut
code,
so
those
kind
of
things
will
follow
in
in
subsequent
patches
but
yeah.
So
first,
I
think
me
and
vivek
were
kind
of
trying
to
add
all
functional
code
so
that
you
know
e2e
test
cases
pass
so
that
patch
was
about
that,
and
so
with
that
patch
yeah,
I
could
see
many
test
cases
are
passing
at
least
in
github
actions
right,
so
some
300
plus
test
cases
pass.
F
So
still,
I
have
to
add
many
minor
features.
So
right
now,
I'm
working
on
external
ip
support
for
all
the
services,
so
I'll
be
raising
that
patch
in
coming
days,
one
or
two
days
or
something
along
with
minor,
missing
functionality
like
okay
session
affinity
or
all
that
minor
things.
Basically,
I
am
doing
that
incremental
implementation,
so
step
by
step,
so
that
I
will
raise
it
in
couple
of
days
and
yeah
I
mean
yeah.
We
me
and
vivek
have
discussed
quite
in
length
in
our
individual.
F
You
know
our
discussion
that
yeah
ut
is
also
an
important
aspect
right
ut
code.
So
it's
it's
an
important
code
to
add
and
I
think
the
I
think
we
know
in
go
right.
We
don't,
I
think,
maybe
if
you,
if
any
of
you
know
well
about
golang
right,
so
mocking,
has
been
a
challenge
right,
so
so
even
coup
proxy.
I
could
see
that
the
way
they
have
circumvented
certain
challenges
is
using
interface
kind
of
an
implementation,
and
then
you
know
pass
certain
dummy
methods
and
that's
how
they
have
done
so
yeah.
F
F
F
Yeah,
so
that's
the
thing
from
my
side
yeah
anyway,
nothing
much
other
than
this
time
being
yeah.
F
F
Like
you
know,
how
do
you
mock
a
method,
say
method,
a
calls
internally
method,
b
or
say
method
c
internally,
so
you
may
need
to
mock
them
as
per
your
needs
or
test
cases.
So
that
is
a
little
bit
challenging
in
golang,
so
me
and
vivek
kind
of
explored
a
bit
last
month.
How
do
we
do
it
and
all
we
couldn't
get
any
satisfactory
answers
in
the
in
the
net?
Basically,
different
people
have
followed
in
all
different
approaches,
right
so
hey
by
by
friedrich,
so
yeah.
F
So
that's
that's
what
I
meant
that
you
know
that
using
interfaces
is
what
is
done
in
current
cube
proxy
also,
so
it's
a
little
bit
overhead
code,
like
you
know,
even
for
a
very
simple
say,
some
kernel,
api
kind
of
a
thing
right.
If
you
just
see
the
cube
proxy,
you
want
to
get
the
list
of
ip
addresses
of
a
worker
node,
it's
just
it's
a
simple
kernel
api.
But
but
you
know,
if
you
have
to
mock
it,
you
have
to
enclose
it
within
an
interface
implementation.
F
A
Yeah-
and
maybe
we
can
use
this
meeting
like
how
you
said
to
get
some
help
around
that
as
well
like
you
know,
folks
are
up
for
taking
things
around
that.
So
one
one
of
the
things
that,
like
I
was
asking
help
was
around
the
user
space.
You
know
testing
user
space,
current
port
like
before
merging
it.
You
know
just
trying
to
get
a
hack,
local
up,
cupping,
local
up
and
then
changing
the
mode
back
into
user
space
and
then
seeing
what
works.
What
does
it
so?
A
Similarly,
you've
been
asking
helper
on
the
united.
A
Help
you
know
in
the
apec
region
and
then
basically
brainstorm
more
in
this
meeting
or
things
like
that
like
right
now
I
see
one
around
user
space
and
you
know
the
ipvs
stuff
that
you've
been
talking
about
so
around.
The
ci,
like
you
know,
which
are
neha
free
range,
douglas
and
nature
will
be
having,
like,
obviously
been
needing
some
help
another.
So
maybe
we
can
have
some
good
first
issues
or
things
like
that,
and
then
ask
folks
to
help
us.
F
Yeah
yeah
yeah.
This
will
be
a
good
forum
to
brainstorm
about
best
practices.
What
we
follow
or
what
we
know
yeah
true
intro,
yeah
doug,
you
you
have
any
idea
doug
I
mean
about
any
other
approach,
other
than
interfaces
using
interfaces
for
mocking
methods.
B
So,
no
actually,
I
need
to
dig
into
it,
but
other
than
that,
the
interface
I
don't
know.
F
Yeah,
just
maybe
I
think,
yeah
yeah.
This
is
anyway
a
good
forum.
F
A
Yeah
sure
yeah
that
that
goes
like
without
saying
we
can
obviously
check
with
everyone
else
who
are
currently
working
on
coming,
but
we
can
also
use
this
as
a
forum
to
get
you
know.
People
are
trying
to
get
into
cupping
and
you
know
try
to
get
them
once
we
have
kind
of
thought
about
all
of
these
complex
problems
and
when
it
comes
down
to
implementing
things
like,
we
can
surely
take
help
from
people
who
are
just
looking
to
you
know.
B
F
No,
no,
nothing!
I
I
was
just
trying
to
say:
maybe
this
this
is
in
a
way
one
good
way
to
get
into
code.
Like
you
know,
I
think
earlier,
I
think
neha
was
interested
also
to
jump
into
this
code
arena
right
like
she
wanted
to
understand,
or
even
other
folks
who
want
to
understand
the
backend
code.
One
best
approach
is
also
start
writing.
Ut
and
you
will.
You
will
really
learn
the
methods.
Okay,
what
this
method
a
does
method
b
does
and
all
that
yeah.
So
we
can
work
on
that
now.
If.
D
Okay,
yeah,
we
can
surely
work
on
this
yeah
yeah,
so
oh
we
vivek.
We
want
to
discuss
on
that
ready
states
right
that.
D
So
I
think
that
issue
was
kept
on
hold
because
there
were
other
issues
like
protobuf
issues,
and
I
mean
there
was
a
few
issues
like
mostly
we
were
fixing
on
protobuf
and
other
failures,
that's
it
and
and
our
ready
state
also.
I
think
we
discussed
last
time
like
locally
it
was
working
sometimes
and
on
ci
it
was
failing.
So
there
were
few
discussions
right
like
this,
so
I.
E
D
E
D
Yeah,
but
that
changes
I'm
not
doing
that
ready
state
change,
because
for
that
I
need
to
come
and
understand.
First,
the
code
base
at
least.
E
D
Yeah
he
told
the
last
friday
correct,
but
at
least
there
were
two
issues
on
ready
states.
One
was
related
to
the
struct
changes
and
other
was
related
to
the
controller
side.
So
there
are
two
different
issues
already
yeah
both
are
related,
but
there
were
two
issues,
so
one
of
them
he
assigned
me,
and
mostly
I
will
check
that
once
this
controller
side
changes
are
in
because.
D
Okay,
okay,
yeah
yeah,
so
maybe
I'll
need
some
help
on
that
side.
How
to
do
that?
Maybe
I'll
help
you.
D
F
B
G
F
Yeah,
if
you
see,
I
think
six
thousand
plus
seem
to
be
skipped,
so
I
wanted
to
know
what
is
the
plan?
Do
you
have
any
idea
or
like
or
we'll
check
with
jay?
What
is
the
plan.
D
F
Yeah,
because
one
of
the
thing
which
is
kept,
I
see,
is
a
load
balancer.
So
I
know
which
is
a
little
complex
area
and
it's
a
very
cloud
provider
specific
area
because
load
balancers
always
come
with
cloud
provider
type
of
deployment,
so
that
is
also
networking
but
yeah.
Anyway.
We
need
that
real
setup
to
run
those
test.
Cases.