►
From YouTube: Antrea Community Meeting 09/27/2021
Description
Antrea Community Meeting, September 27th 2021
A
A
Unless
you
live
in
your
west
coast
of
united
states,
where
it
will
still
be
monday,
tuesday,
monday
september,
the
27th,
and
for
today
we
will
have
a
wedding
presenting
us,
an
obvious,
flexible
pipeline
design
for
open,
with
which
I
do
not
have
any
other
topic
on
the
agenda.
But
then
there
will
be,
of
course,
open
discussions
and
we'll
probably
just
wait
a
few
more
seconds.
As
I
see
team
members
are
still
joining
and
then
we
will
get
started.
A
Okay,
it
seems
that
we
can
get
started
now.
So,
as
I
said,
the
agenda
in
the
agenda
for
today
we
have
the
presentation
about
the
obvious,
flexible
pipeline
design.
So
winning
please
go
ahead.
The
floor
is
yours.
B
First
now
I
share
my
screen
now
so
flexible
pipeline.
First,
I
want
to
introduce
a
generic
pipeline
template.
I
want
to
introduce
10
states
for
the
pipeline,
including
this
one.
You
could
see.
B
We
should
allow
packets
to
escape
some
stages
in
a
pipeline,
even
if
there
are
tables
in
the
stages.
Firstly,
the
photo's
classifier
stage.
The
classifier
stage
is
used
to
classify
traffic.
For
example,
we
can
use
the
classifier
switch
to
mark
the
packet
source
like
what
unsure
is
currently
doing
and
also
we
we
could
use
the
classifier
switch
to
classify
the
package
for
different
pipelines.
B
So,
additionally,
we
will
even
use
a
different
pipeline
for
ips,
unicast
or
multicast
in
future.
I
guess,
then,
the
second
stage
is
validation
stage.
We
could
use
the
validation
stage
to
have
some
pre-check
free
checks
for
the
package
before
they
forwarding
or
routing
the
package,
or
do
not
polishing
policy
something
else
so,
for
example,
they
currently
use
the
spoof
guard
table
to
check
the
ip
import
number
and
the
micro
drives
for
the
packets
to
avoid
rp
chip
in
the
in
the
package.
B
So
we
could
also
use
the
sky
table
in
the
validation
stage
after
using
the
variable
pipeline.
Besides,
we
also
use
the
connection
track
for
the
connections,
so
for
a
reply
package
all
for
the
existing
established
connections
we
might
to
might
need
to
check
the
natural
space.
Then
we
could
use
the
validation,
validation
stage
to
check
connection
state.
B
Then
the
third
stage
is
pre-loading
stage
where
well
packaged
until
the
prerogative
stage.
It
has
passed
the
private
two
stages
and
they
use
this
stage
to
perform
the
night
on
the
package,
if
necessary
or
any
other
operations
on
the
package
before
performing
routing
since
oi
supports
performing
d9
connection
track
zone
and
the
dean
knight
on
the
reply
package
should
be
performed
in
a
table
in
the
ballistic
validation
space.
Just
as
I
mentioned,
and
then
the
next
stage
is
the
egress
security
stage.
B
The
this
stage
is
used
to
perform
the
network
policy
equalize
rules
and
then
is
for
the
routing
stage.
We
could
use
this
page
to
route
the
package
according
to
the
legislative
destination
address,
just
like
the
existing
code
or
reset
the
destination
mic
address
according
to
the
destination
address
and
then
decrement
the
tpl.
If
the
packet
is
forwarding
only
inside
the
os
and
the.
B
In
this
stage,
this
this
operation
might
be
taken
on
service
traffic.
If
we
find
the
soft
address
and
the
destination
address
is
the
same
that
we
might
take
last
night
to
avoid
the
traffic
not
for
not.
No
is
not
forwarding
correctly
the
same
as
the
d
knight,
the
the
same
as
the
pressing
stage,
the
same
as
the
denied
traffic
they
also
perform
night
on
the
reply
package
in
the
validation
stage,
then
we
come
to
the
switching
stage
it
is
used
to.
B
B
If
a
connection
has
passed
the
sea
grass
and
ingress
now
policy
check,
we
need
to
commit
the
package
to
commit
the
connection
to
a
generic
contract
soon,
and
this
operation
is
useful
to
is
have
to
apply
now.
Policy
rules
on
the
sequencing
package
and
the
last
stage
is
absolutely
just
like
what
they
have
done,
yeah
and
sure
for
the
layer,
2
output
table,
and
in
this
stage
we
will
output
the
package
to
os
parts
so
that
the
packet
will
leave
the
os
pipeline.
B
This
is
for
the
stages
and
for
pipeline
time
template.
It
is
an
also
a
new
concept
introduced
in
the
velocity
pipeline
pipeline.
Template
is
used
to
describe
the
expected
os
pipeline
for
feature.
B
For
example,
a
connectivity
feature
or
service
or
network
policy
feature
can
declare
one
or
many
pipeline
templates
according
to
the
use
case,
for
example,
if
the
connectivity
feature
has
different
requirements,
important
networking
and
vm
networking
it,
it
can
be
defined
two
different
pipeline
templates,
for
example,
for
the
part
connectivity
we
might,
you
might
support
a
layer
to
early
stream
loading
and
a
little
forward
forwarding,
but
vm
connectivity.
You
might
only
forward
the
package
from
uplink
to
the
bridge
directly
and
the
template
might
be
simpler
might
be
easier
than
the
pos
connectivity.
B
This
is
the
struct
definition
for
the
pipeline
template
we
might
to.
We
might
need
to
declare
the
state
tables
the
the
expected
contract
zones
and
the
facial
ids.
We
can
template
official,
declares
the
expected
tables
in
the
pipeline
template
and
organizes
the
tables
according
to
the
aesthetic
stages.
B
It
is
not
necessary
to
declare
all
static
stages
in
a
pipeline
template,
but
it
must
have
declared
tables
in
which
the
facial
needs
to
install
a
new
block
entry.
I
mean
that
if
a
feature
needs
to
install
flow
entry
in
a
table,
then
it
must
declare
the
table
in
each
type
in
its
template.
B
B
For
table
request,
which
is
used
in
the
speech
space
tables,
the
structure
is
like
this:
we
have
a
name
for
the
table
picture
table
and
we
have
the
os
table
in
the
structure
and
we
have
the
page
id
and
the
features
since
we
could
declare
a
table
in
different
pipeline
templates
by
the
features.
So
here
we
use
a
size
for
the
features
that
means
the
table
might
have
have
many
features
and
for
free
table
requests.
B
We
should
include
the
table
the
feature
table
and
the
priority.
The
priorities
used
to
organize
the
table
sequence.
B
The
package
might
enter
the
table
priorities
than
the
other
tables,
with
the
lower
priority
and
for
the
council
zoom
request.
We
should
declare
expected
control
zone
ids
in
the
template
for
pipeline.
B
As
I
mentioned,
the
pipeline
template
is
defined
by
feature,
but
for
os
we
have
a
pipeline
generated
by
the
templates
introduced
by
the
features
a
pipeline
is
built
according
to
the
trojan
chosen
pipeline
templates.
In
the
agent
initialization
phase
and
the
tables
in
different
templates
are
organized
according
to
the
stage
table,
id
is
generated
dynamically
and.
B
And
the
ideas
of
the
tables
are
secret
pipeline
that
we
have
the
fixed
number
of
table
ids,
but
for
flexible
type
of
flexible
pipeline
the
ids
are
dynamic
and
the
I
the
the
number
is
frequency
my
pipelines
might
exist.
At
the
same
time
I
mean
the
interagency,
we
might
have
a
multiple
pipelines,
and
so
we
could
differentiate
the
pipelines
according
to
the
package
protocol
like
app
app
package
and
ipv4
and
ipv6,
so
that
for
the
different
protocol
for
the
package
with
different
protocol,
it
enters
a
region.
It
enters
a
unique
pipeline.
B
I
mean
that,
except
for
the
classifier
table,
the
table
0
a
package
will
not
enter
the
pipeline
for
different
particles
anymore,
then,
for
the
packet
of
walking
paths,
a
packet
is
walking
from
the
first
table
to
the
left
table
in
a
pipeline
if
it
is
not
dropped
by
any
table
in
the
pipeline
by
sport.
B
We
support
the
three
types
of
the
table
summit
actions
the
year
of
law
action.
The
first
one
is
go
to
the
next
table.
This
is
the
action
if
a
packet
is
backed
to
leave
the
current
cable
by
not
output,
to
a
pulse
or
sent
to
controller,
and
if
the
current
table
is
the
last
variance
stage,
the
next
cable
should
be
the
first
one
in
the
next
valid
stage.
B
A
valid
stage
means
that
tables
are
included
in
the
stage
the
second
table
submit
action
is
to
resubmit
packages
to
one
or
more
tables.
Here
we
usually
resume
the
packet,
but
not
go
to
tables,
go
to
table
actions
in
the
open
flow
entries,
and
this
action
is
useful
for
the
service
traffic.
If
we
want
to
compose
a
social
affinity
or
some
dynamic,
true
choosing
the
end
points
use
cases.
B
The
target
tables
in
the
resulting
actions
must
be
in
the
same
stage
as
as
the
current
table
and
the
tables
are
declared
in
by
the
facial
of
this
of
the
current
flow.
It
is
the
restriction
limitation
for
the
action
and
the
the
third
resume.
The
table
action
is
go
to
a
stage.
The
package
packet
actually
goes
to
the
first
table
in
the
target
stage,
and
the
target
stage
must
be
the
be
run
later
than
current
stage.
B
This
action
is
useful
for
some
special
traffic,
like
the
panel
traffic
after
they
performance
performing
the
perform
the
less
resulting
on
the
time
on
the
package
and
find
the
packet
that
needs
to
output
to
the
panel.
Then
they
could
ignore
the
ingress
policy
rules
that
we
could
use
this
this
section
to
to
escape
the
increased
policy
rules
following
and
then
is
an
example
for
the
pipeline
generation.
B
We
have,
or
I
have
defined
some
templates
for
different
officials.
Firstly,
for
the
power
connectivity
for
power
connectivity,
we
might
want
to
classify
the
product
source,
so
we
have
the
classifier
page
transfer
table
and
we
want
to
perform
the
subject
on
the
packet
and
then-
and
we
will
also
commit
the
package
to
a
contract
zone
in
the
contract
stage.
B
So
we
have
declared
a
routing
stage
on
the
in
the
feature
and
including
the
tables
they
might
use,
then,
is
for
the
switching
switching
stage.
We
need
to
use
the
layer,
2
forwarding
calculation
table
to
to
calculate
the
target
or
pause
number
according
to
the
destination
microdrives,
and
the
last
stage
is
for
the
output
table.
We
might
output
the
package
to
the
target
or
export
number.
This
is
for
pulse
connectivity,
and
then
we
come
to
a
second
facial
nerve
policy.
It
has
all.
B
It
has
also
introduced
its
own
target
stage
and
the
tables
in
the
template.
Then
we
have
a
sort
of
service.
B
Facial
solace
is
introduced,
is
its
own
tables
and
the
first
temple
phosphate
is
egoized
and
so
on,
and
we
have
a
vm
connectivity
then,
after
all
these,
the
last
templates
brought
up
for
the
public
traffic
and
after
they
introduced
declared
public
templates
and
in
ensure
agent
initialization
space.
They
must
generate
those
flexible
pipelines
for,
according
to
the
templates,
then
they
can.
We
have
this
table
generation
and
the
table
id
is
frequency
all
the
states.
B
All
the
tables
in
the
same
page
is
one
by
one
organized
then
I
I
introduced
in
the
example
for
the
pipeline
template.
We
have
connectivity
for
pod
and
for
vm,
and
I
think
this
as
used
about
two
different
use
cases
for
intro
agents.
Then
we
might
have
another
pipeline
for
vm
case.
The
intro
agent
is
deploying
on
the
vm
case
and
the
table
ids
are
different
from
what
this
must
be
in
the
pod
traffic.
B
I
could
see
that
for
the
egress,
for
example,
the
egress
rule
table
in
ram
connectivity
pipeline
if
type
table
id
is
in
file,
but
for
the
pod
traffic
the
egress
rule
table
id
is
guarding.
We
have
four
to
have
troubleshooting
the
os
pipeline,
overflow
pack
open
flow
entries.
We
have
introduced
another
pr
to
modify
the
table
names.
Then
after
we
dump
the
open
blue
entrance,
we
could
see
the
open
floor
table
name
by
another
table
id.
B
Then
every
official
must
focus
on
its
its
own
tables
and
the
flows
in
the
tables.
It
could
even
know
the
table.
Id
number
and
animal.
B
The
last
topic-
the
last
thing
I
want
to
say,
is
about
a
question
in
the
flexible
pipeline
design.
Is
that,
as
I
introduced,
we
might
introduce
different
contract
zones
in
the
in
different
stages.
For
example,
we
use
a
contract
zone
for
the
night
in
the
per
routing
stage
and
we
use
another
control
zone
in
the
post-resulting
stage
for
internet.
They
either
has
a
third
contract
zone
for
the
generic
connections
in
the
contract
stage.
Then
we
might
have
straight
contracts.
B
The
worry
about
this
design
is
the
performance,
so
we
have
a
question.
Is
that
shall
we
use
a
single
city
contract
zone
for
the
generic
connections,
I
mean?
Shall
they
share
the
control
zones?
They
follow
the
how
to
power.
The
connections
with
the
dnx
contract
zone
always
just
use
a
single
contract
zoom,
and
this
is
the
the
candidate
issue
about
the
performance.
I
have
thought
about.
The
threshold
design.
B
A
I
know
we
can
eat
you,
we
can
hear
you.
Oh
okay,
yeah
speaking
of
I
was
just
waiting
if,
if
there
was
anyone
that
was
willing
to
come
up
with
any
question,
I
I
I
mean
I
personally,
I
do
not
have
any
question.
I
have
to
say
that
your
presentation
was
very
clear
me.
Everything
makes
sense
to
me
here.
I
I
need
to
review
the
documents
to
get
a
better
idea
of
how
these
will
work.
A
Let's
say
that
the
only
question
that
this
might
be
very
trivial,
but
I
mean
it's
not
really
a
question
is
a
this
design.
In
your
opinion,
what
kind
of
benefits
will
bring
to
andrea.
B
B
I
mean
that
every
feature
might
have
different
control
results
and
I
don't
need
to
think
about
how
to
differentiate
the
package
in
in
a
zone
which
is
shared
by
different
features,
but
that
the
the
bad
is
performance,
because
we
might
answer
three
zones,
and
you
might
know
that
the
performance
is
worse.
If
they
enter
or
enter
a
new
contract
school,
then
they're
not
entering
the
contract
zone.
C
A
All
right,
it
appears
that
it's
all
about
this
topic,
so
thanks
a
lot
queening
for
this
presentation.
That
was
a
really
nicely
presented
and
very
informative.
A
As
for
your
document,
I
believe
is
that
correct.
Do
we
do
have
an
issue
on
github
attracted
and
the
document
link
is
available?
There?
Is
that
correct.
A
And
good,
so
I
believe
that
that's
all
for
this,
this
topic
is
there,
since
we
don't
have
any
other
topic
on
the
agenda,
we
are
now
in
the
open
discussion
stage.
Is
there
anything
that
you
like
to
bring
up
for
discussion
today.
A
All
right,
it
seems
that
we
don't
have
any
topic.
I
just
had.
The
one
quick
question
for
for
everyone
here
is:
how
are
we
doing
with
regards
to
code
review
now?
Is
the
our
is
the
mean,
let's
say
the
wait
time
for
prs
to
be
reviewed
decreasing,
or
do
we
still
have
like
longer
than
longer
than
expected
times
to
get
reviews
on
prs?
A
I
mean
for
the
prs
that
you're
submitting,
do
you
feel
like
review
times,
are
within
what
we'll
be
expecting
or
do
you
feel
like
you
know,
we
should
probably
try
and
reduce.
I
mean,
let's
say
mean
time
to
review.
C
I
think
we're
doing
our
best,
solitary
and
we're
also
rolling
out
a
new
tool
to
be
able
to.
I
think,
did
I
cover
this
a
couple
weeks
ago.
I
don't
remember,
but
basically
to
be
able
to
assign
pull
requests
to
the
right
set
of
people
and
enforce
that
we
get
enough
reviews
for
hpr.
So.
A
Thanks
anton
that's,
that's
what
I
wanted
to
talk
about,
because
I
don't
think
that
we've
spoken
about
it
into
the
in
in
in
the
meeting
here.
But
so
you,
you
rolled
out
the
tool
and
you're,
seeing
that
it's
bringing
some
benefits
right.
C
Well,
we
literally
just
started
out
this
weekend.
It's
not
like
automated!
Yet
right,
it's
yeah
thanks!
Oh
actually,
let
me
put
the
issue
in
the
chat.
I
thought
that's
what
it
was,
but
yeah
I
put
the
link
in
the
chat,
so
we
just
rolled
it
out
this
week
and
right
now,
it's
not
automatic.
It
only
applies
to
a
pr's
that
have
a
specific
label.
The
label
is
review
manager
test.
Of
course,
in
the
long
term
it
will
apply
to
all
prs.
C
So
if,
if
I'm
leveling
labeling
your
pr
with
this
it,
it
just
means
that
I
want
to
do
a
test
run
of
the
two
on
that
specific
pr
and
perfect
yeah-
and
I
guess
I
don't
really
need
to
get
into
the
details
of
those
people.
Interested
people
can
can
take
a
look
at
the
issue
and
comment
on
the
issue
if,
if
they
want
to,
but
basically
the
idea
is
that
we're
kind
of
gonna
label
prs
by
area
kind
of
like
how
it's
done
in
a
way
in
upstream
communities
and
based
on
which
area
labels.
C
We
apply
to
a
pr
the
owners
for
those
specific
areas
or
are
going
to
get
assigned
as
reviewers
and
and
then,
when
you
get
enough
approvals
from
those
people,
the
pr
can
be
merged.
Basically,.
A
Okay,
that's
great
that's
great,
and
you
know
connected
to
these.
If
you
remember,
like
a
few
sometimes
ago,
we
had
michael
running
some
michael
wong
running
some
project
health
metrics
on
andrea
is
this.
I
don't
think
this
is
an
activity
that
we
are
doing
periodically
right
perhaps
could
be
a
good
idea
to
run
these
metrics
like
every
quarter
to
see.
C
That
maybe
people
don't
have
the
link,
but
now
that
we're
cncf
project
we
actually
have
access
to
a
real
time
graph
and
a
dashboard
where
we
can
see
the
average
time
it
takes
for
a
pr
to
be
reviewed
and
for
new
issues
to
be
commented
on,
and
things
like
this.
A
Yeah
good,
that's
that's
very
useful.
That's
very
useful
to
have
thanks
thanks
antonio.
Yes,
I
should
have
access
to
it,
but
probably
I
was
ignoring
it
all
right,
thanks
a
lot
for
thanks
a
lot
for
the
clarification,
okay,
so
good!
That's
all
I
wanted
to
discuss
today
anything
else
that
we'd
like
to
bring
up.
C
Yeah
thanks
genji
yeah.
I
think
everyone
should
try
to
get
into
the
habit
to
tag
their
to
label
their
pr.
Sorry
with
the
appropriate
labels
area
labels
and
if,
if
someone
thinks
that
a
label
is
missing,
which
is
definitely
possible,
they
can
like
open
a
pr
to
add
this
label
or
open
an
issue,
and
we
can
add
this
label
and
yeah.
I
mean
the
the
better
we
label
prs.
We
are
with
the
correct
labels,
correct
area
labels,
the
better
the
process
is
going
to
work.
D
C
Yeah
and
everyone
can
add
the
labels
to
other
people's
pr,
so
I
mean
you
can
comment.
You
can
also
add
labels
directly,
of
course,
if
I
think
every
author
should
take
the
habit
to
label
their
own
pr
when
they
open
it,
but
as
a
reviewer
or
someone
who
just
stumbles
upon
the
pr,
you
can
also
add
those
labels.
I
expect
external
contributors
may
not
be
like
totally
familiar
with
the
project
and
so
yeah.
C
And
I
found
the
cncf
dashboard
for
entry,
so
let
me
put
the
link
in
the
in
the
chat
here
and
people
can
you
can
take
a
look
at
this
and
should
be
able
to
browse
some
metrics.
A
Thanks
for
sharing
this
sorry,
I
was
talking
to
myself
for
a
minute
and
and
well.
A
Also,
oh,
where
do
it?
Oh,
this
is
for
all
the
cncf
projects.
C
Yeah,
but
if
you
you
can
scroll
down
and
you'll,
see
a
lot
of
metrics
like
there
is
our
time
to
engagement,
yeah.
A
Yeah
no,
no,
I
was
just
you
know.
Obviously
I
was
struggling
to
see
the
entry
icon,
but
that's
just
my
brain.
When
there
are
so
many
icons,
I
they
look
all
the
same.
To
me,
I
found
a
f.
I
found
andrea
and
yes,
I
can
see
that
there
are
all
the
matrix
yeah
good.
A
A
Okay,
so
great
thanks
for
sharing
this
anton
and
well,
let's,
let's
see
how
the
project
matrix
improve
with
the
tool
that
you
have
developed,
and
you
know
with
identifications
of
reviewers
for
areas
and
said
that
is
there
anything
else
that
we'd
like
to
discuss
for
today.
D
C
I
don't
know
if
you
can
group
by
by
by
like
sandbox
projects,
for
example
versus
graduated
or
incubating,
but
I'm
pretty
sure
you
can
see
aggregated
metrics
for
all
the
projects.
D
A
Okay,
so
that's
that's
great
and
is
there
anything
else
that
we
would
like
to
discuss?
I
mean
that's,
I
think
that's
all
about
these
metrics
and
whatever
that's
a
nice
thing
to
see
to
have.
I
do
not
think
that
we
can
compare
directly
project
versus
project
right.
We
can
probably
compare
ourselves
versus
global
metrics.
A
Now
that
that
will
make
sense,
you
don't
want
this
to
become
like
you,
don't
want
to
gamify
this
too
much.
Otherwise,
people
will
start
gaming
so
which
is
not
a
good
idea.
A
Okay,
so
growth
great
any
any
other
topic
for
today,
because
otherwise
we
might
close
the
call.
C
I
guess
just
one
more
thing
I
wanted
to
add
is
that
the
mapping
from
area
label
to
reviewers
and
approvers
is
is
part
of
the
entry
repository
and
I
put
the
link
in
the
chat,
so
anyone
can
add
themselves
as
reviewers
and
then
they
would
get
notification
a
notification
and
they
will
get
a
sign
if
a
pr
is
open
in
that
specific
area.
So
if
you
want
to
add
yourself
to
an
area,
everyone
should
be
welcome
to
do
so.
We
have
an
initial
list,
but
anyone
can
become
a
reviewer.
A
A
Now
our
next
community
meeting
should
fall
if
I'm
not
mistaken
during
cubicon
itself,
so
you
I
would
like
to
first
ask
if
there
is
a
if
you
would
like
to
move
the
meeting.
If
you
have,
if
you
need
to
attend
the
cube
concessions,
you
know
if
you
are
like
kind
of
busy
with
preparing
stuff
for
cubicle
or
whether
it's
okay
to
keep
the
meeting.
At
the
same
time,.
A
This
is
correct.
Tuesday
next
meeting
will
be
monday,
11th
u.s
time,
and
I
think
it
should
be
the
opening
day
of
cubicon,
which
typically
you
know
is,
if
I
remember
correctly,
agenda
is
just
keynotes,
but
in
any
case,
if
we
want
to
move
the
meeting
to
move
the
meeting
to
make
things
easier
for
everyone,
we
can
do
that.
Usually
that's
being
9
pm
u.s
time
should
not
be
a
problem,
but
you
know
just
just
that
we
can
discuss
these
awesome
slack.
A
All
right,
it
seems
that
no
feedback
means
that
we
are
okay
with
keeping
the
meeting,
so
we
keep
the
meeting
for
as
scheduled
and
and
well
that's.
I
believe,
that's
all
for
today,
unless
everyone
has
a
any
other
topic
to
bring
up.
A
Assuming
that
I'm
not
talking
to
myself,
of
course
well,
it
appears
that
that
is
all
for
today.
So
I
would
like
to
as
usual
to
thank
everyone
for
joining
most,
in
particular,
many
thanks
to
winning
for
today's
presentation
and
all
the
links
for
the
things
that
we
have
discussed
today
are
shared
on
the
chat.
So
thanks
again,
thanks
again
to
everyone
for
joining,
I
wish
I
wish
all
of
you
a
good
evening
good
morning
or
good
afternoon,
ever
that's
all
for
today's
meeting
and
talk
to
you
in
tweaks
time.