►
From YouTube: Multi-Network Community Sync Meeting for 20230405
Description
Multi-Network Community Sync Meeting for 20230405
A
All
right
welcome
everyone
in
to
the
multi-natural
community,
sync,
today's
April
5th,
and
we
are
going
today
continue
with
our
cab
discussion.
Last
week
we
stopped
at
the
Talk
on
the
default.
Network
I
see
some
new
faces
here
for
all
that
wants
to
follow
the
dock.
On
the
top
of
this
meeting,
there's
a
link
to
the
pr
that
I'm
to
talk,
and
basically
all
the
discussion
we
do
in
terms
of
apis
is
in
this
link
here
on
the
top
all
right
and
last
time
any
other
topics
to
bring
up.
A
A
We
started
someone
Cube
cattle
deleted
this
object,
but
there
is
a
pod
using
it.
So
since
we
want
to
make
sure
that
the
Pod
cannot
be
removed
when
a
when
the
Pod
Network
cannot
be
removed
when
pod
is
using,
it
referencing
it.
So
in
that
case
we
will
have
a
ready
condition
going
false
and
the
in
a
reason,
I'm
deleted
in
progress,
delete
it
in
progress
that
will
allow
that
would
prevent
Newports
to
attach
to
this
pod
Network,
because
it's
someone's
trying
to
delete
it.
A
A
I
think
please
read
through
the
through
the
description.
I
hope
this
is
kind
of
understandable
and
one
more
thing
I
started
started
but
didn't
finish.
I
started
some
changes
on
creating
this
whole
crd
in
go
code
in
the
CR
I
got
recommendation
to
just
create
a
fork
of
a
KK
repository
and
do
it
there
so
I'm
following
that.
So
and
that
comes
with
like
a
bunch
of
a
changes
because
of
how
the
kubernetes
repository
works.
A
So
I'm,
probably
maybe
hope,
will
try
to
publish
this
by
the
end
of
the
week
so
that
you
can
look
at
it.
There.
I
will
split
it
into
separate
comments,
so
you
can
see.
What's
what
is
the
core
API
versus
the
other
additional
stuff
that
just
need
to
be
done
for
kubernetes
repository,
but
beside
that
you
will
be
able
to
if
you
want
to
import
this
or
just
copy
paste
the
whole
go
code
to
try
out
some
of
the
implementations
or
on
your
own.
A
A
This
cluster
pod
network
is
basically
the
representation
of
today's
Network.
So
whatever
cluster
you,
whenever
you
create
a
cluster-
and
you
have
already
cni
in
it
that
cluster
default
Network
would
represent-
and
there
are
some
specific
some
characteristics
for
it
like
it
cannot
be
removed-
must
be
available
on
all
the
nodes.
So,
basically,
all
the
kind
of
characteristics
that
today's
Network
that
we
have
with
cni
have
yes,
tomorrow.
B
Yep
I
have
the
question
about
the
default
Port
Network
about
the
the
backward
compatibility
of
the
current,
the
kubernetes
So.
Currently,
the
kubernetes
doesn't
have
the
ID
spot
Network,
and
then
we
are
guaranteed
that
the
we
should
work
work
with
current
quantities.
B
So
so
this
means
the
without
the
app
the
so
sometimes
the
cluster
may
not
have
any
port
Network
object,
only
the
the
they
just
putting
the
default
the
cni
at
that
time,
the
so
that
this
cap
is
allowed.
This
situation
I
mean
that.
A
Maybe
maybe
look
at
it
from
this
way
when
we
were
to
when
we
like
the
long
term
goal
right
if
we
were
to
finally
get
an
PGA
I
would
consider
pod
Network
BD
reference
for
any
network
right
with
some
defaults,
which
will
not
change
the
default
behavior
that
we
know
of
today,
for
example,
like
I,
don't
have
to
specify
the
network
in
a
pod.
That's
basically
that
will
mean
that
it
will
default
to
using
the
default
pod
network.
But
what
I
want
to
say
is
with
this
change.
B
A
B
B
So
I'm
I'm
a
little
bit
wondering
that
sometimes
the
India,
the
some
deployments
in
the
case
I'm
a
little
bit
thinking
that
there
is
no
default
Port
Network
object
because
they
they
just
putting
and
then
so
and
then
they
are
they
adding
the
another
Port
Network.
At
that
time,
the
the
cluster
may
have
the
Airport
network
without
the
default.
C
B
So
so
yeah,
what
do
I
say?
Yeah
I'm
still
wondering
the
how
to
express
what
the
the
we
should
add
the
add
some
stuff.
Some
sentence
about
to
mentions
that
the
custom
may
not
have
default
object
at
that
time
of
the
year.
B
C
A
So,
let's,
let's
we
I
had
this
here.
Let's
discuss
that
in
a
second
okay,
okay.
B
A
E
So
yeah,
so
classes
can
live
for
a
really
long
time.
That
would
know
so.
My
question
is
really:
do
you
mean
with
what
you're
writing
here
that
the
pro
the
default
Network
cannot
be
replaced.
A
At
this
for
this
cap,
the
whole
effort
I
think
in
our
requirements
we
said
we
will
not
allow
like
we
will
not
allow
replacing
it
for
the
cluster
level.
E
I'm
good
with
it,
but
I
think
just
make
it
clear
because
right
now
it
says:
well,
you
can't
remove
it,
but
I
can
replace
it
with
another
one
and
have
sort
of
an
old
version.
So
if
that
is
what's
what's
we
want
to
say,
then
we
should
say
that
it
cannot
be
replaced.
A
Our
requirements,
what
we
do
want
to
so
maybe
let
me
know
I
replaced.
A
Depending
on
how
you
think
about
replacing,
because
in
one
of
our
last
phases
to
replacement,
is
to
allow
changing
the
default
on
the
namespace
level,
so
we
would
want
to
do
that,
but
so
we
cannot
replace
it
like
globally,
but
we
can
specify
it
locally
for
namespace.
That's.
E
What
we
use
the
code,
we
can
change
right.
That
was,
but
on
the
global
or
leave
it
open.
So
is
it
is
it
so?
What
I'm
trying
to
say
is
that
is
it
that
a
system
must
have
a
default
Network
always,
and
you
cannot
have
a
system
without
default
Network
we
mean,
or
is
it
that
the
default
Network,
the
sort
of
the
instance
of
some
network
type
that
we
created
is
from
when
we
started
the
cluster
can
never
be
changed.
A
E
A
F
A
To
ensure
we
always
so
that's
what
I
said
in
initially
for
tomo
right,
we
must
make
sure
right
to
represent
the
default.
Network
same
way,
we're
going
to
represent
the
additional,
so
they
are
in
their
par
in
priority.
They
cannot
be
like
oh
yeah
default.
One
is
assumed
and
it's
somewhere
there,
but
we
don't
have
to
create
it
and
worry
about
it.
Now
we
need
to
kind
of
identify
it
and
point
it
out
and
we
have
to
be
there.
A
That's
that's
why,
so
that
we
can
for
two
reasons
if
we
say
that
if
we
just
neglect
it,
it
always
be
like
this
and
if
we
don't
Define
it,
that
would
always
like.
A
A
Yeah
very
Point,
let's
let's
say
I
will
shut
down
the
cubelets
everywhere
and
then
just
change,
my
cni
configurations
and,
and
there
you
go
right
or
even
I
can
do
it
by
by
per
note,
that's
something
that
operation
that
you
can
do
and
what
would
you
do
at
that
point
with
the
default
Network?
That's
what
you
refer
to.
E
A
E
At
the
same
time,
if
I
over
specify
that
maybe
we
make
things
that
people
are
doing
today,
not
legal,
so
to
speak
so.
A
So
so
so
do
we
need
some
sort
of
admin
mode
to
edit
right.
A
E
A
That's
that's
something
that
that's
that's
doesn't
exist
today
right,
so
that's
something!
That's,
let's
put
it
on
the
side.
That's
separate,
I,
think
outside
of
scope
of
this
topic,
but
I
should
be
able
to
address
yeah,
and
let
me
just
think
about
that.
But
unless
we
want
to
someone
has
some
ideas,
support.
A
Okay
yeah,
so
so
yeah,
okay,
gotcha,
that's
that's
a
very
point.
What
else
so,
as
I
mentioned
already.
A
Then
it
can
be
changed
edited,
but
but
but
basically
so
Paul
Network
cannot
can
be.
A
Something
like
that
yeah
that
will
be
yeah.
We
need
to
figure
out.
What's
what
will
that
be?
Because
we
we
say
it's,
the
object
is
immutable.
So
that's
that's
a
big
question
right
because
we're
breaking
our
own
rule
over
here
lots
network
is
immutable
right
so
and.
E
Then
we
would
have
to
remove
it
all.
This
can
be
immutable.
If,
if
you
change
it,
then
each
object
is
immutable
right.
What
do
you
mean?
I
mean
if
you
say
that
you
replace
the
the
you
have
a
pod
Network
and
you
say
that's
immutable,
and
now
you
put
create
a
new
pod
Network
and
you
you
change
sort
of
which
one
is
the
default.
Then
the
objects
are
still
immutable,
but
the
reference.
Okay,
the
the
default,
the
the
reference
call
default-
can
be
changed.
A
That
but
the
name
has
to
be
so
immutable
means
you
cannot
change
the
name,
but
the
name
has
to
be
right.
Oh.
A
Right
so
we
either
so
so,
let's
talk
in
a
second
tomorrow,
you
had
a
hand,
you
want
to
say
something.
B
Yeah,
so
so
the
based
on
the
the
price,
the
comment
I'm
a
little
bit
the
come
up
with
about
the
So
currently
kubernetes
can
put
the
default
Network
for
node
I
mean
that
the
even
we
have
the
one
cluster
in
the
consists
from
three
three
nodes
other
than
each
worker
node.
Each
node
may
have
a
different
cni
configurations,
but
it's
okay
I
mean
that
so
the
if
they
each
cni
is
possible
to
connect
to
each
nose
and
then
it
works.
Fine.
B
A
A
Are
you
anyone
else
so
I
hear
you
I
I
agree
with
you
that's
possible,
but
the
question
is:
is
that
something
that
we
need
to
address?
Because
if
we
are
that's
a
big
deal,
anyone
can
chip
in
or
anyone
from
like
their
experience?
Is
that
something
that
well.
G
I'm
still
catching
up
from
being
out
for
a
few
weeks,
the
default
pod
Network
essentially
you're,
describing
essentially
ethics
versus
ethy,
as
opposed
because
it
there's
no
such
thing
as
the
default
interface.
A
G
G
G
A
G
The
end
of
the
day,
leaving
vrf
society
and
it
really
doesn't
change
anything
I'm
about
to
say,
and
you
some
of
you
are
sick
of
me
saying
it
packet
does
about
look
up
and
based
on
the
row.
It
will
pick
an
egress
of
X
versus
Y
versus
Z
and
that
that
Lookout
can
be
based
on
the
prefix.
It
could
be
ecmp
doesn't
matter
so
I
can
you.
We
can
go
through
all
this
effort
to
pick
a
default
interface,
which
is
what
I'm
reading
default
pod
Network
to
be.
E
H
E
E
A
It
has
nothing
to
do
how
it's
how
you
implemented
it,
and
in
your
case,
yes,
it
can
be
just
a
it's
it's
a
black
hole
down
in
on
the
Note,
where
black
hole,
it's
a
just
routing,
based
on
routing
where
down
on
the
Node
it
selects,
whichever
interface
it
wants.
This
is
not
the
this
is
not
about
selecting
which
interface
this
is
this,
is
you
can
think
of
it?
This.
G
A
So
the
question
is
yes,
so
the
question
is:
can
someone
else?
Collaborate
on
Thomas
example?
Is
that
something
that
we've
seen
production,
I
I
know
it's
possible
as
a
engineering
experiment
just
locally
on
three
nodes,
but
is
that
something
that?
Because
if
it's
a
just
one
percent
of
a
case
I
would
say
we
neglect
it?
But
if
you
say
something
that
anyone
has
a
real
live
situation
and
what
I
may
refer
to
here
is
each
note
has
different
cni
in
the
same
cluster.
B
So
so
much
how
about
addressing
in
the
signal
Network
side
as
or
maybe
they
will.
B
Right
right
writing
down
in
the
some
concerned
area
and
then
the
in
the
to
request
review
we
can
discussing
inside.
B
So
at
least
I'm
I
have
the
yeah
a
little
bit
concerned
about
today.
I
I
don't
have
the
any
conclusion
about
the
yeah.
No
one
used
to
use
the
word
someone's
used,
yeah.
G
A
But
Michael,
that's
that's
fine,
but
what
we're
trying
to
say
here
is
we
are
kind
of
enforcing
on
to
you,
so
your
different
CNN
implementation
would
mean
youth
have
different
pod
networks
because
you
want
to
differentiate
between
each
implementation,
and
here
we
are
saying
and
we're
kind
of
forcing
that
no
for
the
default
Network.
You
have
to
have
one
across
the
board
end
of
the
story.
Right,
that's
what
it
means,
but
that's
and
and
the
question
and
today
you
could
you
have
the
freedom
to
just
do
that.
A
I
can
every
note
will
have
different
cni
for
the
default
stuff
and
and
it
will
work
right
because,
depending
on
how
you
create
your
cluster,
so
do
we
want
to
really
support?
That?
Is
that,
like
a
one
percent
use
case,
and
then
we
can
neglect
it
and
with
this?
So
so
that's
kind
of
the
problem,
but
let's
Fairborn
tone.
That's
thanks
for
suggesting
this
I
will
bring
this
up
on
our
next
Sig
Network
and
ask
the
community.
H
Yeah
so
as
we
meet
I
have
a
bigger,
a
question:
yeah
sorry
I'm
a
bit
new,
so
I'm
a
bit
new.
So
maybe
it's
a
stupid
question,
but
oh
go
ahead.
The
cni
implementation
that
sits
on
the
meat-
because
here
we
are
talking
about
we
Define
a
default
Network
object
right,
so
I
look
at
cni
implementation.
It's
the
way
that
actually
consumes
that
object
and
implements
it
and
and
does
something
with
it
cool.
So
we
could
build,
have
a
different
implementation
on
every
node.
Underneath
this
object.
A
H
H
A
Problem
would
be
the
one
problem.
What
would
it
will
be
if
I
wonder
if
your
implementation
somehow
is
centralized?
It's
it's
a
it's.
It
would
work
if
you're.
If
each
cni
is
a
silo
per
node,
then
yes,
it's
each
one
is
independent.
So
I
agree,
that's
possible,
and
maybe
that's
the
answer
to
this
because
you're
right
then
maybe
this
is
just
an
object
and
on
each
note
the
implementation
can
consider
themselves
as
the
default
and
and
handle
this.
E
A
A
D
A
E
E
You
could
all
I
mean
just
as
well
so
say
that
today
is
under
specified.
Kubernetes
I
mean
the
requirement
is
that
you
have
layer,
3
connectivity
between
the
parts
and
then
the
cni
is
there
to
affect
that.
If
you
use
one
or
many
here,
we're
trying
to
build
up
a
yeah,
so
so
for
for
the
for
the
default
network.
If
you
don't
say
anything,
you
should
be
able
to
fall
back
to
what
to
do
today.
I
think,
that's
what
we
need
to
get
to
what
you
guys
described.
A
You
kind
of
yeah,
if
if
so,
this
is
what
I'm
trying
to
say
on
on
two
things:
first,
what
will
be
the
initial
values
and
then
it
matters
for
first,
maybe
let
me
change
the
order
who
creates
this
right.
So
what
I'm
thinking
is.
This
is
something
that
Tomo
brought
up
very
early.
What
what?
If
my
cni
is
elaborate
and
my
default
network
has
to
have
a
reference
to
some
parameters
right
because
I
control,
what
whatever
the
default
network
is
going
to
use
it's
controlled
here.
Mycni
is
fully
pod
network
based.
A
So
now
it
looks
at
this
object
all
the
time,
even
for
the
default
Network
right.
So
what
do
I
do
here
and
how
do
I
configure
it?
And
if
we
were
to
just
follow
the
mechanics
of
the
default
namespace,
which
basically
is
created
by
I,
think
a
KCM
that
that
won't
fly
right,
because
if
we
say
that
the
object
is
immutable,
KCM
creates
me.
A
Let's
say
just
this
version
of
the
object
right:
even
without
the
provider
just
says,
it
checks
whether
you
provide
V4
or
V6,
and
if
you
did,
then
it
will
just
do
and
use
the
internal
here
and
then
it
will
just
provide
a
create
dashboard
network
with
the
default
name,
and
that
might
be
fine.
But
what?
If
not?
What?
A
If
I
need
to
have
some
specify
my
provider,
because
my
cni
requires
it
or
I
I
need
to
provide
parameters
reference
as
well,
so
what
I'm,
thinking
and
proposing
is
have
a
flag
in
whoever
would
create
this
object
in
okay,
let's
say
KCM,
let's,
let's
assume
KCM
for
for
those
folks
that
are
not
maybe
another
case
in
kubernetes
control
manager.
This
is
one
of
the
main
beside
the
API
container.
A
There
is
a
kubernetes
control
manager
container
that
has
most
of
the
controllers
for
core
controllers
for
kubernetes,
so
KCM
would
have
a
flack
and
basically
in
it
it
will
say:
Do
not
create
default.
Pod
Network,
something
like
that
along
those
lines.
What
it
will
mean
that
it
is,
it
will
be
expected
by
your
CNN
implementation
to
create
that
default
object
and
basically,
before
before
you
create
it,
you
cannot
create
a
pod
Network
pod.
A
You
can
only
create
host
Network
pods,
but
not
pod,
Network
Bots,
and
there
probably
will
be
some
other
conditions
or
how
to
force
someone
to
create
this
object.
Probably
this
can
be
translated
into
a
node
Readiness
where
today,
if
you
don't
create
a
cni
configuration,
kublet
will
say:
never
not
not
that
ready
and
because
there
is
since
cni
missing.
So
until
that
probably
can
be
a
blocker.
That's
how
we
can
kind
of
make
a
dependency
on
this,
so
basically,
until
your
cni
creates
a
default.
A
B
Ahead,
oh
one,
so
yeah
yeah,
let
me
clarify
that
so
in
in
the
so
we
we
have
the
two
modes
right.
So
the
if
the
controller
KCM
creates
the
the
this
stuff.
B
When
is
created,
I
mean
that
the
right
right
there,
let's
imagine,
they're,
killed
up
creating
a
cluster
in
the
key
battle
and
then
there
we
we
already
creating
the
old
worker,
node
and
master
node,
and
then,
after
that,
the
default
object
is
already
created
or
once
the
network
cni
is
configured
I
mean
that
so
current
idea
we
have
the
what
on
the
when
cluster
is
created.
At
that
time,
the
node
is
not
ready
and
then
once
the
cni
configuration
is.
B
A
No,
so
so
when
KCM
creates
it,
it
will
just
create
this,
because
the
only
knowledge
KCM
has
is
about
the
eye
pump.
So
you
see
how
in
KCM
as
well,
you
can
provide
something
called
pod
cider,
not
both
Sider.
It
can
be
I.
Don't
I
forgot
the
argument
name,
but
basically,
today,
KCM
accepts
a
field
called
pod
cider.
If
you
want
to
use
the
kubernetes,
the
KCM
I
pump
controller
as
not
internal
so
and
you
can
do
it
for
V4
and
V6.
A
That's
how
you'd
configure,
basically,
today
your
your
dual
stack,
so
you
will
translate
those
arguments
into
what
whether
I,
Define,
iPhone,
4
or
icon,
6.
and
then
I
will
just
create
a
blind
default
object.
A
A
Something
like
that
right,
so
you
will
have
this,
you
would
you
would
pass.
Let's
say
this
this
this
type
of
argument
and
then
it's
on
you
fully
on
you
to
create
the
the
Pod
Network,
the
default
pod
Network,
you
see.
So
in
this
case
there
is
no
engagement
of
of
of
the.
There
is
no
engagement
of
the
cni
at
all
of
the
implementation.
Only
in
the
other
one
there
is.
That
will
be
the
difference.
B
So
another
question
is
So
currently,
if
we
we're
using
the
the
front
or
open
kubernetes
or
Asylum
at
that
time,
the
IP
address
is
also
provided
by
the
dcni.
So
this
means
the
iPhone
4.
It
should
be
external.
A
B
You
say
that
will
put
the
burden,
maybe
KCM
the
I
mean
that
the
KCM
create
mode
at
that
time.
They
are
somehow.
We
should
provide
the
external,
because
the
the
even
with
this
enhancement
users
want
to
use
usual
previous
kubernetes
implementation
I
mean
that
the
setting
up
deploying
a
cluster
and
then
the
installing
the
psyllium
or
open
kubernetes
at
that
time,
the
they
do
not
and
then
the
if
they
do
not
care
about
the
amount
Network
at
that
time.
They
do
not
want
to
add
additional
process,
I
mean,
but.
A
B
Unless
we
are
changing,
but
so
the
default
Network
requires
the
address
the
B
for
ypb6
address,
one.
A
B
So
so
the
so
at
that
time
the.
B
Yeah,
even
with
the
Casey
and
the
KCM
Auto
creation
mode
other
than
the
somehow,
the
this
mode
should
specify
the
way
to
iPhone
4
and
iPhone
6
and
then
the
which
is
the
mode
is
used
for
iPhone
4
or
iPhone
6
iPhone,
the
iPhone
4
equal.
A
What
you're
saying,
maybe
that's
that
that's
that
is
the
maybe
that
is
the
maybe
so
you're
saying,
because
those
are
the
two
Fields
right
so.
B
So
I
I
mean
that
this
this,
the
coupe,
the
iPhone
for
kubernetes
and
iPhone
6
kubernetes,
should
be
configured
in
case
of
the
KCM,
create
mode.
A
That's
that's
a
that's
a
very
easy,
simple
solution.
There
you
go
yeah
I
I
like
that.
Okay,
thanks
just
add
argument
to
Casey
in
to
expose
those
right
yep
any
comments
on
that
I
wonder
how
that's
going
to
be
taken
by
the
KCM
folks,
but
any
any
other
comments
on
this.
A
Okay,
initial
values,
I
think
we
already
covered
that
right.
What
will
be
so?
Basically,
it
would
mean
that
what
we
kind
of
draw
here
right,
no
provider,
no
parameters
reference.
The
only
thing
you
can,
though,
control
is
those
two
Fields.
The
problem
would
be
if
we're
going
to
expand
this
object
right
for
any
reason
in
the
future.
A
What
we
do
then
right,
so
that's
a
bit
tricky
and
the
other
part
we
don't
have
Antonio
here
is
the
the
service
and
clusters
either
maybe
service
less,
but
there
is
a
object
called
cluster
cider,
that's
going
to
be
better
soon
and
that
cluster
side
they're
supposed
to
replace
the
cider
definition
that
we
have
today
in
KCM-
and
maybe
this
is
good,
because
if
we
expose
those
flags
here
that
indirectly
kind
of
defines
what
sort
of
cluster
I
want
to
have
dual
stack,
single
stack
or
or
one
of
the
two
right.
A
A
A
A
E
We
can't
enforce
it
right,
I
think
it
can
be.
Okay,
I
mean
I,
think
that
sort
of
that
you
don't
have
to
fulfill
everything.
We
just
need
to
know
what.
A
D
E
On
the
other
hand,
it's
it's
always
it's
an
easy
trick
to
make
sure
that
people
that
don't
really
that
are
fine
with
everything
that
that's
there
and
sort
of
say.
Oh
yeah,
you
can
do
everything
you
do
now
if
you
don't
using
multi-network,
but
if
you
do,
then
these
are
the
rules
that
you
will
have
to
use.
Yeah.
A
Sounds
yeah,
let's,
let's
think
about
this
I
think:
let's
do
it!
Maybe
we
can
get
back
to
this
next
week.
If
anything,
let's
sit
on
this
and.
E
A
Can
no
worries
yeah.
Everyone
is
not
a
new
pair.
It's
on
the
mass,
so
no
worries
any
other
comments
on
the
default.
I
would
like
I
think
any
other
things
we
missed
around
default
net,
pod
Network
so
always
created
with
default
namespace
when
not
referenced
it.
It
will
be
the
one
that's
going
to
be
used,
it
will
be
named.
Default
must
be
available
on
all
the
nodes.
This.
I
E
E
A
No,
no,
that's
that's
very
good
because
you're
correct
it
it
it
it.
It
might
be
misunderstood.
Yes,
that's
good!
That's
good!
Let's
be
explicit!
Any
other
comments,
anything
else
we
missed.
A
We
migration
to
a
different
cni
so
for
the
default
Network.
H
A
H
A
Yes,
I
can
go
to
Every
to
every
node
and
modify
my
cni
config
today,
because
this
is
just
a
config
right.
There
is
not
not
much
to
it.
What
what
do
I
do
with
the
default.
So
from
the
API
point
of
view,
we
want
to
block
the
whole
thing
we
so
because,
if
you,
if
you
allow
it
from
the
API
point
of
view,
that
means
the
underhood
under
the
need
underneath
of
the
whole
implementation.
A
Something
can
change
and
we
say
we
don't
want
to
support
it,
so
nothing
from
the
API
point
of
view.
This
is
in
our
requirements
that
we
said
that
the
default
network
will
be
unchangeable
the
same.
It
is
today
because
if
you
were
to
change
it,
that
kind
of
assumes
something
will
happen
under
the
under
the
hood
under
the
need.
So
the
wheel
we
so
this
at
this,
not
at
this
stage.
We
just
do
this.
This
whole
effort.
We
in
our
requirements
said
that
we
don't
want
to
allow
that
we.
H
Don't
want
to
change
the
default
and
I
agree
with
that.
So
my
only
my
comment
is
the
following:
is
that
you
cannot
change
the
parameter,
but
if
you
want
to
change
the
cni,
that
means
that
all
the
cnis
implementation
should
adhere
to
the
parameters
that
we
specify
in
the
default
networking
configuration
so
so.
A
No,
that's
fine,
no
worries,
but
the
nice
Vim
is
that
is
that
correctly
pronounced
yeah.
A
Know
so
so
my
we
need
to
Define
it
so
that
the
case
that
you
want
to
maybe
cover
is
is
is
is
expected
today,
it's
not
defined,
we
don't
know
what
what
can
be
done
or
cannot.
So
this
is
what
we
want
to
say
what
what
Perry's
suggesting
as
well:
let's
have
it
for
properly
Define
or
how
it
should
behave
right
or
we
can
open
it.
What
you
sign!
A
H
It
I
agree
with
that
yeah.
My
second
point
on
migration
is:
how
do
you
come
from,
let's
say
a
release
of
kubernetes
that
doesn't
have
it
to
one
that
supports
it
all
right.
So
there's
a
second
point
on
migration:
how
do
you
get
the
default
network
instantiated
from
an
ex
from
a
cluster,
so.
A
So
this
is
where,
where
I
would
imagine,
that's
a
fair
point
so
I,
so
this
is
where
we
discussed
with
with
tomorrow
right.
Let
me
just
answer
and
then
I
think.
E
I
think
I
think
Vim
opened
up
a
little
bit
of
can
of
worms.
It's
one
thing
in
the
system
where
you
only
have
the
default
Network
there
you
can
describe
how
this
would
work.
I
think
the
the
real
problem
comes
when
you
have
a
system
that
uses
multi-networking,
be
that
using
multis
or
some
other
other
sort
of
cni
function.
That
has
many
networks,
but
those
I
don't
think
we
should
even
try
for
the
default
Network.
We
should
specify
how
that
migrates
over
to
to
this
or
for
any
other
multi-networks
I.
D
Mean
if
somebody's
using
maltes
and
maltes
is
working
for
them,
we
could
just
migrate.
The
default
Network
to
this
model
and
multis
handles
everything
out
of
band
and
they've
they've
got
what
they
had
before
and
yeah.
If
they
want.
If
they
want
the
new
things,
then
finally,
they
create
a
new
cluster.
H
A
I
think
I
think
this
is
only
in
terms
of
the
default
Network
right.
How
we
would
do
that
and-
and
that's
a
fair
question,
because
now,
if
you
want
to
because
we
could
say
this,
if
you
want
to
upgrade
to
130-
let's
say
one:
three:
zero
kubernetes,
your
implementation
I
should
support
what
network
and
this
that's
a
big
restriction.
That's
a
big
big!
Probably
no!
No,
because
you
know
you
put
some
restriction
and
now
every
Etna
has
to
catch
up
to
report
properly.
What
is
support
network
can
handle
it.
A
So
I,
don't
think
that's
a
proper
thing.
That's
tricky
right,
because
then,
how
do
I
like
now,
then
I,
your
your
cluster,
has
to
specify
a
user
has
to
specify
in
KCM
those
two
flags
right.
What's
what
is
the
iPhone
for
external
internal
right?
So
so
that's
that's
a
fair
point
and
that's
that's
of
its
own
is
kind
of
one
pair,
because
how
would
you
do
that
right?
I
would
say
it
is
it
is.
It
is
a
tricky.
E
E
A
B
A
E
A
I
Yeah
so
yeah
I
mean
I,
have
a
question
on
how
cni
creates
a
default
network
default
partner
work
is
per
cluster
and
CNA
like
there
can
be
like
a
different
version
of
CNA
running
on
different
nodes.
So
is
it
like
any
one
of
the
CNA
running
on
each
any
one
of
the
nodes?
Will
you
first
create
a
default
part
Network
and
all
the
CNS
will
be
using
the
same
default
button?
What
network
there.
A
So
it
will
be
up
to
the
implementation.
I
would
assume
either
the
first
cni
that
starts.
It
can
be
direct
like
the
first
cni.
This
is
just
a
binary
right,
I
think
I,
don't
know
Mac
villain
right.
If
it's
just
the
first
binary
starting
up
a
it,
would
need
now
need
access
to
kubernetes
configuration
which
it
doesn't
have
today.
So
that's
where
the
support
would
have
to
come
in
and
then
it
will
just
check.
A
Okay,
is
there
a
default
pod
network,
node
and
I
will
create
it
right
and
if
I
am
the
default
one
right,
but
that
has
to
be
a
properly
supported
and
I'm,
throwing
some
some
ideas
that
that
might
be
completely
incorrect.
But
you
see
that
that's
all
based
on
the
implementation
right,
let's
think
like
the
other
implementation,
might
have
a
centralized
controller
which
can
do
that
for
you
right.
A
I
So
if,
if
you
have
I
think
this
is
anyway,
tomorrow
brought
this
up,
but
if
you
have
a
different
cni
on
different
nodes,
so
they
should
have
some
common
way
of
understanding
the
default
part
Network
yeah.
A
Right
right,
so
so
this
is
something
that
we
need
to
identify
how
big
of
a
use
case.
This
is
right,
so
this
is
something
that
we
mentioned.
So,
let's
reach
out
to
let's
talk
with
the
Sig
Network
folks
and
see.
Is
that
a
valid
use
case
because,
like
in
in
my
in
my
experience,
I,
never
met
that
like
I
have
different
cni
per
node?
That's
never
happened,
always.
A
A
So
that's
something
that,
let's,
let's
ping,
the
Sig
Network
folks
to
for
that
and
that's
that's
it
and
let
them
kind
of
Define
that
yeah,
because
that's
basically
that
that's
their
call,
because
this
is
how
this
is
what
we
want
to
support
in
kubernetes
in
general.
So
let's
ask
them
what
they
want
to
do
right
and.
I
A
Something
that
we
should
worry
about
anything.
A
Okay,
I
think
going
back
to
this
default
transition.
Yeah,
that's
something
that
we
need
to
maybe
stating
some
of
those
statements.
Probably
we
need
to
have
a.
Let
me
just
add
that
integration
story
and
we
need
this
because,
because
without
this
there
is,
we
cannot
have
that
right.
We
need
to
Define
what
will
be
the
migration
story.
A
Sorry,
jumping
back
and
forth.
We
need
to
have
a
migration
story
on
how
to
transition
the
today's
whatever
I
have
today,
and
it
will
create
the
default
Network.
A
Pointers
I
think
we
have
at
least
two
that
we
are
not
resolved.
So
this
is
something
that,
let's
all
think
about
this
I
will
I,
will
try
to
think
about
this
one
as
well
through
the
weekend
for
the
week
to
kind
of
come
up
with
something,
because
those
are
really
good
topics
that
we
need
to
kind
of
identify.
A
No
okay,
I
think
I
will
cut
it
here
for
today,
30
once
item
no.
A
C
A
Should
have
you
yeah
I
was
looking
sorry
I
haven't
looked
at
it.
We
were
looking
at
this.
That's
why
I
was
kind
of
asking
at
the
beginning
to
bring
up
any
any
topics
you
want
to
bring
up.
Let's
bring
it
up
first,
because
yeah
this,
this
first
topic
always
takes
the
whole.
The
whole
thing:
that's
how
this
those
meeting
goes
by
the
way.
Philip
I
think
your
first
time
here
right.
A
Yeah,
let's,
let's,
let's,
let's,
if
you,
if
you're
okay,
to
come
up
to
come
next
week
and
we
can
start
from
this
next
week.
Okay,.
A
A
No
yeah,
we
are
at
the
time
where
we
transitioned
to
that.
We
are
working
with
the
yes.
We
are
facing
that
I
will
add
you
too,
because
I
have
an
invite,
my
own,
that
is,
kicked
off
and
I
I,
sending
those
those
out
to
everyone
and
I
have
to
say
that
one
to
be
weekly,
the
other
one
is
not
updated
in
GitHub
yet
and
I
think
Michael
Michael's
gonna
teach
today,
Michael
Michael
Zappa
is
working
on
it
to
kind
of
update
that,
and
that's
that's
a
longer
process.
A
That's
why
it's
not
there
yet
sorry
Phil!
But
yes,
we
have
a
meeting
next
week.
I
will
add
you
I
will,
as
you
reach
out
to
you
in
in
the
slack,
if
anything,
to
add
you
to
the
invite
I
have,
but
it's
still
the
same
link.
So
at
the
same
time,
everything's
the
same.
The
link
is
the
same,
so
we
meet
at
the
same
time
every
week
and
the
zoom
link
it's
it's
that
doesn't
change,
so
we
will
hear
yeah
the
invite
will
get
updated
soon.
A
Sure
definitely
I
will
add
you
folks
to
if
I
have
your
if
you
could
slack
a
slack.
D
A
Select
me
with
your
emails
in
case
I:
don't
have
them
that
I
can
add
you
to
I,
have
a
Google's,
my
my
my
my
corporate
invite
for
this
as
well,
which
I
was
thinking
to
to
to
obsolete
to
to
this
car,
but
right
now,
it's
kind
of
useful
to
to
have
it
for
for,
for
this
particular
case,
but
beside
that,
keep
in
mind
that
it's
just
having
calendar
that's
easy,
but
but
we
now
we
meet
weekly
and
with
the
same
league,
all
right
all
right,
Philip.
A
We
will
move
to
your
topic
first
thing
next
week
and
then
continue
on.
Hopefully
we
can.
We
can
touch
on
those
two
and
then
the
next
phase
will
be
just
how
we
change
pod
spec
for
the
attachment.
So
I
have
some
examples
here
read
through
those,
and
we
can.
We
can
talk
about
this.
There
is
a
lot,
so
there
is
another
thing
that
and
the
status
of
reporting.
So
let's
let's
focus
that
on
next,
but
hopefully
we
can
resolve
those
as
well
all
right
thanks.
Everyone
meet
you
next
week,
yeah.