►
From YouTube: Antrea Community Meeting 10/25/2021
Description
Antrea Community Meeting, October 25th 2021
A
A
This
is
likely
going
to
be
a
short
meeting.
So
if
anyone
has
topic
proposals,
anything
that
you
would
like
to
discuss,
please
come
forward
because
today
is
just
open
discussion.
B
A
Yeah,
that
was
that
was
the
same
question
that
I
had
yeah.
Let's
wait
for
to
see
if
chan
or
others
can
join
and
then
we'll
probably
discuss
a
little
bit
of
the
status
of
the
release.
A
But
while
we,
our
child,
has
joined
80s
good
afternoon,
chan,
hi
salvador,
sorry,
no,
no
worries.
So
today
we
have
open
discussions
because
there
is
no
pre
scheduled
topic
for
the
agenda
and
we
were
wondering
if
we
could
use
this
meeting
for
a
general
update
on
the
status
of
the
1.4
release
and
we
selected
you
as
the
lucky
one
for
providing
this
update.
C
Okay
sure
so
so
far,
one
major
feature
which
is
proxy,
oh,
has
been
merged
of
all
platform,
has
been
supported
on
both
linux
and
windows
and
another
major
feature.
We
know
that
it's
android
ipad
and
I
started.
C
I
started
reviewing
the
related
patches
late,
but
I
think
changing
and
has
blue
them
for
a
while,
and
I
think
the
most
mostly
code
look
looks
good
to
me
and
I
have
only
some
suggestion
about
simplifying
the
code
and
improving
the
test
and
some
minor
style
improvement.
C
A
No,
no,
no,
that's
that's
a
good
update
and
talking
about
andrea
ipam.
I
know
that
the
we've
done
verification
we've
done
also
integration
testing,
and
we
know
that
the
feature
end-to-end
testing
and
I
believe
that
zhang
has
also
verified
these
in
other
integrations.
A
So
let's
say
that
the
feature
itself
is
probably
ready
for
being
alpha
in
1.4.
My
only
comment
regarding
the
overall
studios
of
the
feature
is
that
we
have
introduced
the
cr
that
needs
validations.
You
know
like
just
standard
user
level
validations,
but
so
far
we
don't
have
yet
any
validation.
A
It's
probably
far
from
ideal,
but
maybe
since
I
don't
think
that
we
want
to
defer
further
andrea
1.4,
maybe
one
idea
could
be
to
document
that
you
know.
Users
must
be
careful
to
set
properties
to
to
carefully
set
properties
because,
for
instance,
you
know
we
are
not
even
validating
the
start.
Ap
of
a
range
must
come
before
the
and
the
final
ip
of
the
range.
A
A
About
the
anterior
ipad,
but
okay,
I
am
talking
about
entry
ipam
as
well.
No
diaphragm,
it's
relatively
easy!
Okay,
because
you
know
node.ipam,
we
simply
took
the
same
feature
which
is
in
the
cubecontroller
manager,
and
we
imported
it
into
one
three
or
four
scenarios
where
node.ipam
is
not
running
the
controller
manager
yeah.
I
was
talking
about
entry
ibm,
you
know
we
added
the
we
added
the
cr
for
entry
ipam
and
that
cr
might
need
value
needs
validations,
which
I
believe
we
cannot
date
in
the
next
release.
D
For
example,
do
we
want
to
validate
a
load?
Deletion
can
be
done
if
the
I
located
from
a
pool
in
the
know.
Let's
know
you
can
you
can
stand
approval?
You
cannot
string
the
proof,
something
like
that.
D
Yes,
yeah,
that's
right!
Yes,
actually,
I
think
anna
give
as
the
meat
saying
it's
probably
a
little
risk
to
risky
to
finished
the
validation
part
in
one
week.
D
So
there
was
a
discussion
to
say:
are
we
okay
to
defer
the
violating
part
to
the
last
race?
Yeah?
Probably
probably
we
can.
We
can
also
talk
about
this
special
crashing.
A
little
yeah.
A
A
Know
I
mean
what
is
what
is
clear
to
me:
is
that
it's
a
risk
for
this
release
to
add
validation
and
not
because
it's
it's
difficult,
but
mostly
for
a
timeline
constraint
thing.
We
would
like
to
release
andrea
1.4
at
the
end
of
the
week,
and
I
don't
see
this
happening
in
three
days,
so
this
is.
This
is
the
only
concern
that
I
have
on
my
side
and
therefore,
since
since
I
also
believe
that
maybe
we
don't
want
to
defer
the
release
further.
A
What
I
was
thinking
is,
we
might
want
to
document
that
at
the
moment
that
the
sierra
doesn't
have
yet
validation.
The
only
thing
that
I
do
not
know-
and
maybe
we
should
probably
seek
sick
feedback
from
the
developers
from
and
run
about.
This
is
that,
if
we
do
some
mistake,
like
you
know,
the
one
that
john
june
was
mentioning,
will
there
be
consequences
on
on
the
running
system
will
be?
A
So
this
is
the
only
the
only
feedback
that
I
think
we
might
need
from
from
online
run
to
decide
whether
it's
too
much
of
a
risk
shipping
the
feature
without
any
cr
validation.
C
A
Okay,
yeah,
maybe
you
know
that's
that's
a
good
idea.
I
think.
Perhaps,
if
you
can,
if
you
can
share
these
pointers
with
anna,
maybe
you
know
maybe
perhaps
on
the
slack
channel,
then
we
can
figure
out
with
there.
If
this
is
something
that
we
could,
we
can
quickly
add
to
the
current
release
and
maybe
in
maybe
you
know
in
either
it
over
it
in
the
next
release
and
improve
it.
A
Okay,
yeah,
then
you
know
the
other
thing
that
was
still
pending,
as
far
as
I
know
is
merging
back
the
feature
branch
into
the
master
branch,
but
since
run
has
been
periodically
synchronizing
the
feature
branch.
This
should
not
be
a
problem.
D
C
D
For
the
implementation,
I
said
so
back
to
the
question:
could
we
releases
feature
without
validation?
D
What
you
guys
think
I
do
have
a
concern.
It
can
cause
some
serious
criticism
issues
in
my
mind.
For
example,
if
you
delete
the
pool,
you
need
a
poor.
D
Yeah,
I
I
think
I
guess
maybe
if
you
just
need
to
prove
me,
maybe
it's
not
too
bad,
but
if
I
I
just
assume
it
can
be
something
sweet.
We
did
it
before.
You
added
the
bike.
We
did
in
some
ip
and
you
had
an
ip
bike.
D
A
No
you're
right,
so
I
I
for
instance,
I
don't
have
a
very
cool
I
mean
at
the
moment.
I
don't
have
a
very
clear
idea.
What
could
happen?
For
instance,
you
know
we
are
storing
the
we're,
also
storing
the
current
allocation
status
in
in
in
the
in
the
cr.
So
I
don't
know
I
see
if
one
features
there
and
modifies
the
allocation,
the
allocation
states,
then
what
happens?
You
know
if
there
will
be
overlapping
a
piece?
A
I
think
that
unexpected
changes
into
pool
configurations
can
also
lead
to
might
also
lead
to.
You
know
to
re-allocation
allocating
the
same
ip
address
twice.
I
don't
have
a
reproduction
case
at
the
moment,
but
I
think
that
this
is
something
that
can
happen.
Therefore,
you
know
it's
it's.
It
is
a
little
bit
tricky.
That's
why
I
was
I
mean
I
I
I'm
thinking
if
we
need
to
release
it
for
1.4.
A
In
any
case,
even
if
we
had
some
basic
validation,
we
will
not
be
able
to
have
the
full
range
of
what
I
mean,
the
full,
the
full.
What,
if
experience
to
make
make
a
call
on
what
can
be
a
safe
change
or
what
can
be
a
risky
change?
A
D
A
All
right,
let's
say
that
maybe
maybe
now
we
have
the
elements
that
we
have
more
elements
to
make
a
an
assessment
there.
We
have
a
pr
chance
sheet,
the
pr
for
adding
validation
to
the
external
people
we
can
build.
A
We
can
build
on
that
and
then
for
the
I
will
say:
let's,
let's
also
ask
anna
she
since
she's
the
sub
magic,
my
subject
matter
expert
for
the
potential,
the
potential
side,
effects
of
missing
validation,
and
then
maybe
you
know
by
tomorrow,
we
can
make
a
call
on
whether
we
want
to
include
this
feature
on
393
or
1.4.
D
Yeah,
I
have
another
another
question
for
ipam.
We
don't
have
otherwise
on
here,
but
yeah.
Maybe
I
can.
I
can
bring
bring
the
hub
here
and
at
least
I
can
get
some
input
from
you
guys.
D
I
was
wondering,
should
we
unify
them
to
be
under
one
feature,
gate
quote
until
I
found
out.
I
know
there
are
different
things
yeah,
so
it's
just
like
the
name
is
a
little
confusing.
I
think
that's
just.
A
That
that
is
right,
I
I
agree
with
you.
I
mean
it's
confusing
because
they
are
both
ibm
but,
as
you
say
there
are,
there
are
two
different
things.
A
The
my
only
concern
is
that
I
for
sure
the
users
that
we
need
node
ibm
will
probably
most
cases
they
will
not
need
entry
ibm
as
well
so,
and
likewise,
I
also
believe
that
the
users
that
will
need
entry
album
in
most
cases
they
will
not
need
no
diabolum,
no
ibm,
must
be
enabled
only
in
some
particular
kubernetes
distribution,
where
the
cube
controller
manager
is
not
running
this
capability.
A
I
personally,
I
don't
see
a
benefit
in
having
them
under
the
same
feature
flag,
since
they
are
a
little
bit
confusing
here
right,
a
bit
a
bit
confusing.
I
don't
know
if
we
want
to
do
something
else
for
make
make
clear
that
they
are
two
different.
They
are
actually
two
very
different
things.
D
For
the
feature
gate,
probably
I
can
agree
with
your
point
and
for
the
config
parameter.
What
do
you
guys
think,
for
example,
let
me
see.
D
B
B
A
Yeah
there
could
be
a
case.
It's
it's
a
case
like
where
a
user
has
a
distribution
where
you
know
there
is
no
no
diaphragm
in
the
distribution
and
at
the
same
time
they
do
not
need
they
need
to
do
both
per
node
subnets,
and
they
have.
You
know
the
anterior
subnets
as
well,
that
that
could
be
a
use
case
where
one
could
need
both.
B
The
thing
is,
since
you're
gonna
take
totally
different
parameters
of
config
parameters.
I
feel
it
kind
of
makes
sense
to
keep
them
as
two
separate
config
groups.
A
D
I
still
feel
it's
pretty
confusing
to
have
two
ipad
things.
I
mean.
A
Yes,
the
the
thing
is
that
know
the
iphone.
I
have
to
say
that
another
ipam
it
makes
sense,
calling
it
another
iphone,
because
that's
yeah.
A
That's
also
used
for
kubernetes
for
entry
ipam.
You
know
we
came,
but
we
came
with
this
name
through
several
iterations.
Now,
if
we
have
to
change
it,
I
will
say:
let's
change
it
before
we
release
it.
If
we
have
to
change
it
to
some
other
to
some
other
name,
but
the
thing
is
whatever
we
call
it,
it
will
call
it
something
ipam
and
therefore,
in
a
way
or
another,
it
will
end
up
being
somewhat
confusing,
with
node
diaphragm.
D
D
C
A
No,
I
I
agree
that
flexible
ipam
is
to
abstract,
and
I
agree
with
the
choice
of
calling
andrea
ipam,
because,
to
be
honest
with
you,
we
don't
know
what
will
be
the
features
that
we
will
add
to
this
capability
in
the
next
next
upcoming
release
and
calling
it
something
like
flexible
album.
It
also
has
implications
on
the
functionalities
that
will
offer
calling
it
on
three
ibm
is
a
little
bit,
let's
see
more
generic
in
terms
of
features
but
specific,
meaning
that
it's
specific
to
entria.
D
A
D
And
I
think
especially
later
it's
possible.
We
want
to
reuse
this
ipad
functionality
for
all
use
case,
for
example
the
secondary
interface
case.
We
also
need
ipl
location,
so.
B
C
D
D
A
Yeah
so
so
you
meant
separating
the
data
path
aspect
from
the
ibm
aspect.
Right,
that's.
A
Yeah,
no,
I
think
I
think
it
might
be
a
good
point
yeah.
It
might
be
a
good
point.
Let's,
let's
see
what
the
others
think
of
it.
A
Well,
it
seems
that
there
are
no
strong
opinions
on
the
mother.
Therefore,
I
will
suggest
maybe
we
can
continue
discussing
on
this
on
this
matter
on
the
onslaught
and
on
the
other
channels,
where
we
are
disc.
Having
all
the
conversations
about
this
feature
you
know
github
and
so
on,
and
so
that
we
can
reach
a
conclusion,
hopefully
before
the
release.
A
Okay,
so
let's
say
that
we
have
to
provide
follow-up
after
the
meeting,
both
up
about
the
validations
that
must
be
in
before
we
ship
this
feature
and
regarding
whether
we
want
to
have
a
single
feature:
flag
or
a
distinct
feature
flag
for
the
ipad
driver
and
the
data
path.
Enablement,
all
right
and
any
other
question
about
ip
star.
Ipam
features
that
have
been
introduced.
A
All
right
cool
any
other
question
in
general
about
the
1.4
release.
A
A
A
Okay,
so
I
just
wanted
to
check
with
you
is
you
know
that
has
been
quite
some
time
ago?
There's
been
cubicon
and
I
I.
A
Yes,
I
have
to
say
that
I
did
not
have
any
time
to
attend
it
at
all
due
to
other
issues
on
our
side.
But
I
wonder
if
anyone
had
some
experience
to
share
for
cubicon
from
cubicon
related
or
not
related
to
andrea.
B
Well,
I
can
say
that
our
office
hours
went
pretty
well,
I
don't
think
we
met
since
that
happened
right.
Is
he
the
first
meeting
since
kimcon
or
is
that
the
second
meeting
since
we.
A
Had
the
other
one
during
cubicon
and
during
cubicon,
the
updated
wedding
is
that
was
indeed
that
the
weirdest
meeting,
I
think.
B
Yeah
so
the
office
hours
112,
I
think
because
they
sent
us
the
attendance
numbers
and
he
was
somewhere
around
65
to
all
visitors
and
I
think
at
any
given
time
where
about
25
people,
and
so
I
think
this
is
like
the
kubecon
entry
event
we've
done,
which
has
had
the
best
attendance
and
there
was
quite
a
lot
of
engagement
from
people.
We
had
some
good
questions
in
the
chat,
so
I
think
it
was
like
a
very
good
experience
that
we
will
definitely
repeat
at
future
kubecons
if
possible.
B
So
we
gave
like
a
project
a
status
update,
essentially
what
happened
during
the
last
six
months
to
one
year
and
then
we
took
questions
as
as
we
touch
different
topics
and
and
new
features.
A
I
think
that's
great
and
my
feeling
is
that
now
being
a
cncf,
sandbox
project
has
helped
a
lot
in
this
direction,
and
I
mean
in
the
direction
of
being
having
community
being
recognized
by
the
community.
A
And-
and
I
believe
it's
all
for
today,
so
I
would
like
to
thank
everyone
for
joining
and
and
well
we'll
talk
again
in
the
two
weeks
time
only
a
reminder,
if
I'm
not
mistaken
for
the
next
meeting,
the
united
states
will
have
a
switched
off
daylight
saving
time,
which
means
that
adjust
for
the
next
meeting
the
us
should
be
in
the
u.s.
It
should
be
at
10
p.m,
instead
of
9
p.m.
A
It
should
go,
go
back
to
6
a.m
for
europe,
but
it
should
change
for
china
where,
instead
of
12
pm,
it
should
be
1
pm.
So
just
keep
an
eye
on
your
calendar,
which
has
been
updated
by
the
way
and,
let's
not
get
confused
with
time
zone
changes.
A
Okay,
so
thanks
everyone
and
I'm
going
to
stop
the
recording
now,
and
I
wish
everyone
a
good
morning
a
good
afternoon
or
good
evening
or
good
night.