►
From YouTube: CNCF Network Service Mesh Meeting - 2019-08-20
Description
Join us for Kubernetes Forums Seoul, Sydney, Bengaluru and Delhi - learn more at kubecon.io
Don't miss KubeCon + CloudNativeCon 2020 events in Amsterdam March 30 - April 2, Shanghai July 28-30 and Boston November 17-20! Learn more at kubecon.io. The conference features presentations from developers and end users of Kubernetes, Prometheus, Envoy, and all of the other CNCF-hosted projects
A
Well,
let's
go
ahead
and
get
started
so
welcome
to
the
next
network
service,
the
mesh
working
group
meeting,
and
so
we
have
this
meeting
every
Tuesday
at
8
a.m.
and
right
now
we
have
two
other
calls
which
are
currently
on
a
hiatus
until
and
so
we
have
a
couple
people
back
who
have
been
taking
a
short
break.
We
have
the
CNC
of
telecom
group,
which
occurs
every
first
and
third
Monday,
the
first
Monday.
It
occurs
at
8:00
a.m.
the
second
call
on
the
third
on
the
third
Monday
of
every
month.
A
On
September
23rd
to
25th,
we
have
four
except
the
talks:
a
telecom
user
group
meeting
and
ACM
F
testbed
tutorial.
We
have
open
source
summit
in
Lyon
France,
which
is
with
a
talk
accepted
by
Ivana
and
radoslav.
We
have
Cube
Khan
and
cloud
native
Col
North
America.
We
are
still
waiting
to
hear
the
results
of
the
call
for
proposals,
but
we
do
have
an
NSM
con
and
we
have
a
call
for
paper
or
call
for
proposals
and
that
are
that
are
currently
open
and
we
also
which
will
close
on
September
13.
A
B
A
C
So
that's
great
and
I'm,
trying
to
like
everyone,
so
hopefully
that
will
reach
I
will
expand
the
audience.
I
think
I'm
caught
up
as
of
like
five
minutes
ago,
so
I'm
sure
there's
a
few
more
to
like
yeah.
That's
awesome,
I
believe
that
counts
is
going
viral,
so
congrats,
we've
gotten
14
more
followers
in
the
past
week,
somehow
there's
still
more
folks
to
follow
so
I
followed
over
50,
more
folks
and
tweeted
and
retweeted
20
times
in
the
past
week,
thanks
to
members
of
our
community
for
capturing
that
we
have
reached
our
1000
commit
mark.
C
That
was
great
to
see
and
slack
and
I
put
my
reposted
bat
with
permission
and
tagged
the
community
member
who
called
that
out
that
one
has
been
a
fun
post.
We've
gotten
a
lot
of
retweets.
Of
that
post
as
well
and
tried
to
share
since
this
week-
and
today
we
had
two
working
group
calls
one
early
morning:
Central
European,
Time
and
one
morning
a
specific
time
and
also
shared
from
last
week's
network
service.
Mesh
working
group
called
the
speck
board
shared
some
information
about
that.
C
This
morning
we
have
a
pin
tweet
for
the
network
service
mask
on
at
cube,
con
co-located
events
and
yes,
that
is
a
separate
event
from
the
service
mask
on
event.
The
network
service
mesh
Cana
is
hosted
by
the
network
service
mesh
community
and,
as
Frederic
mentioned,
the
CFP
is
for
network
service.
Mesh
con
are
still
open
until
Friday
September
13th.
C
So
if
you
have
an
idea,
please
go
to
that
forum
and
fill
out
your
request
this
week,
I'll
continue
announcing
many
events
coming
up
for
a
networking
summit
in
Europe,
open
source
summit
in
Europe,
cube
con
network
service,
con
network
service
mesh
con
and
I'll
continue
liking
and
retweeting.
All
of
these
mess
of
meshes
blog
shares
and
there's
anything
else,
you'd
like
to
see
or
share
on
our
Twitter
page.
Please
let
me
know.
B
D
I
have
one
question
for
you,
so
today
we
hit
our
peak
every
other
week
in
the
Asian
the
time
zone,
and
there
was
this
question
from
attendance
there
should
we
find
a
way
to
somehow
promote
disco
better,
because,
let's
not
attendance
from
from
China
in
the
region
from
and
I
know
that
till
the
2g
had
also
struggled
to
find
the
proper
time
slowed,
something
that
actually
works
for
them.
So
on
any
TV.
Any
any
told
from
this
any
experienced
education.
C
Well,
one
thing
I
learned
from
the
telco
music
group
is
that
sharing
on
the
on
slack
didn't
work
as
well
for
folks
in
different
countries
and
and
maybe
sharing
on
Twitter
may
not
work
as
well.
One
thing
that
did
get
more
interaction
was
posting
a
github
issue
and
posting
a
github
issue
for
the
the
the
the
call
that
would
be
earlier
in
the
day
for
folks
in
a
pack
time
zones.
So
maybe
we
can
create
a
github
issue.
D
C
C
D
A
A
Let's
see
we
have
an
SM.
Now
has
github
actions
available,
so
start
thinking
about
the
type
of
actions
that
you
would
like
to
see,
and
we
should
start.
We
should
probably
start
making
a
couple
proposals
on
this.
I
think
github
actors
it'll
be
one
of
those
things
that
will
really
help
supercharge
our
some
of
our
workflows
and.
D
A
B
So
I
mean
this
is
sort
of
standard
prospectus
for
events
the
why
sponsor
NSM
con
2
is
down
on
the
third
page,
which
is
various
sponsorship
opportunities
and
the
benefits
thereof.
So
if
you
are
part
of
an
organization
that
would
you
could
scroll
down
to
the
third
page,
use
the
Sherlock,
though
okay?
So
if
you
were
part
of
little
further
down?
B
So
if
you
are
part
of
an
organization
that
sponsor
aspects
of
NSM
con,
there
are
a
number
of
sponsorship
act,
opportunities
at
various
price
points,
you
limb,
sponsorship,
session,
recording
of
in
sponsorship,
etc.
So
I
would
encourage
you
to
have
your
folks
reach
out
to
sponsors
and
network
service
mission
at
IO
and
sponsors.
A
Yeah
I
believe
this
is
the
first
time
we've
ever
asked
community
for
money
so
but
yeah.
So
we
have
responses.
We
have
sponsorships
available,
feel
free
to
read
through
this
and
pass
it
to
your
favorite
to
your
favorite
leadership
or
our
PR
s,
and
with
that,
if
you
have
any
questions,
definitely
definitely
get
a
hold
of
us.
You
can
get
a
hold
of
us
through
the
usual
channels
or
at
sponsor
that
were
super
stretched
on
Io
anyways.
B
Noticed
yesterday
that
master
seemed
a
little
bit
wobbly,
so
lots
of
jobs
lost
to
see
eye
drops
on
master,
we're,
failing
and
sort
of.
It
turns
out
that
Circle
C
was
having
difficulties,
bombing,
docker
instances
to
run
our
docker
belts,
and
since
you
know,
people
tend
to
look
at
this
and
go.
Why
is
masters
unstable
I
wanted
to
make
sure
that
we're
aware
of
especially
since
that's
probably
just
go
to
apply
to
PRS
as
well?
That
at
least
yesterday
circle
see
I
was
having
some
fairly
serious
difficulties.
B
E
B
I'm
not
sure
cuz,
I've,
mostly
been
chasing
master
and
master
I,
think
I
think
we
did
have
one
failure
today
on
master
that
was
attributable
to
that
kind
of
a
problem
where
you
know,
generally
speaking,
income
China.
So
your
container
build
failed
and
you
look
at
it.
It's
like
we
weren't
able
to
get
a
docker
engine,
so
it's
amazing
how
unproductive
it
is
to
try
and
make
do
a
docker
build
without
a
docker
engine.
B
So
this
is
just
sort
of
trying
to
go
through
and
grab
some
things
that
are
in
process.
I'd
actually
encourage
folks
to
also
add
some
of
these
things
themselves
to
this
list,
as
we
go
forward
in
the
weeks
because
I
may
or
may
not
catch
everything
that
goes
by,
but
just
to
give
folks
a
notion
of
what's
up
what's
currently
in
motion,
and
you
know
so
they
can
sort
of
track
some
of
what's
happening.
I
know
that
as
communities
get
much
more
active,
it
gets
way
harder
to
keep
track
of.
B
B
F
B
B
The
documentation
on
service
accounts
is
a
tiny
bit
sparse
and
it
looks
like
you're
supposed
to
use
a
token
for
the
service
account
when
deploying
versus
specifying
the
pods
back.
But
it's
not
super
clear
on
the
docks,
so
we're
still
trying
to
chase
that
down
as
a
question
of
what's
the
proper
behavior,
because
fire
tends
to
piggyback
its
notion
of
identity
based
upon
committee
service
accounts
and
some
other
data.
B
Pool
and
then
upcoming
I
think
we've
got
some
work
to
do
is
to
adapt
to
the
inter
domain
work.
That's
coming
in
and
then
probably
look
at
using
open
policy
agent
for
authorizations
aspire
and
spiffy
give
you
authenticate
abilities.
The
idea
would
be
to
use
OPA
so
that
the
network
surface
mesh
can
and.
C
B
Know
we
can
handle
delegation
of
authorization,
decisions
to
the
service
mesh
and
then
the
service
mesh
allocates
those
two
OPA,
because
lots
of
the
world
is
using
OPA
for
authorization
policy
might
as
well
get
with
a
predominant
bandwagon
cool
interdomain
artem.
Do
you
want
to
say
a
few
words
there?
B
B
B
Currently
do
not
that's
fine,
so
the
inter-domain
I
think
make
1298
is
the
last
PR
for
the
initial
letter
domain
work
I
believe
it's
close
to
going
in.
We
have
a
known
issue
with
be
a
nice
selection,
there
being
a
slight
bug
there
across
domains,
that's
being
looked
at.
So
if
you
have
interesting
ideas
there,
that
would
definitely
be
worth
looking
at.
B
We've
got
some
work:
that's
going
on
with
increased
applicability
for
network
service
manager,
so,
for
example,
making
it
easier
to
make
the
exclude
prefix
detection
easier
to
detect.
So
some
of
that's
been
happening.
Sr
v6
support,
so
I
think
Arnhem
is
also
working
on
adding
s
r
v6
support.
Let's
see
if
we
have
our
resident
s
our
v6
fans
on
the
call.
No
so,
but
that's
currently
slightly
blocked
because
of
a
need
to
get
a
little
bit
of
feature.
Extension
in
legato
for
the
VPP
agent,
but
I
expect
that
will
probably
come
together
fairly
quickly.
B
B
G
B
C
B
We've
also
got
an
issue
out
for
training
or
chaining
refactoring
of
net
resource
manager
to
look
at
breaking
network
service
manager,
also
open
to
change
them
similar
right,
I,
hope,
being
network
service
manager
is
kind
of
big
right
now
and
a
little
bit
complex
and
the
hope
is
to
break
it
into
a
chain
of
much
simpler
pieces
so
that
it's
possible
to
say
hey.
You
want
to
go.
Do
this.
Oh
yeah
do
like
a
small
fragment
of
the
service.
B
Here
it
uses
the
same
API
as
the
G
or
PC
call
and
insert
that
into
the
chain.
You
should
be
good.
It
also
should
make
it
much
easier
for
other
people
to
write
network
service
managers
and
other
environments
that
differ
because
they
can
take
off
the
Cheryl's
sort
of
the
sock
pieces
of
a
network
service
manager
and
compose
them
together
in
chains
for
themselves.
B
Okay
and
then
there's
a
certain
amount
of
linearizing
local
to
remote
calls
the
metric
service
manager-
and
this
is
again
trying
to
simplify
network
service
manager
and
as
a
side
effect.
It
should
make
it
trivially
easy
to
do.
I,
don't
know
how
many
folks
remember
talking
about
the
create
proxy
network
service
managers
that
could
dynamically
create.
B
D
B
H
I
was
Rick,
I
had
pretty
good
progress
with
mapping
metrics
with
the
cross
neck
data
I'm
now
trying
to
resolve
another
problem,
I
hit
that
the
statistics
are
on,
the
cross,
connect
monitor
actually
empty,
and
there
are
no
statistics
received
so
I'm
currently
in
testing
investigating
trying
to
resolve
this.
Otherwise,
if
it's
fixed
I
think
it's
a
painting,
a
pretty
good
stage.
B
H
B
Then
I
just
want
to
sort
of
briefly
call
people's
attention
to
the
spec
board.
We've
got
a
lot
of
interesting
things
out
there
aspects
where
people
have
sort
of
written
stuff
for
conversation
with
the
community
about
you
know
stuff
that
we
think
we
probably
may
want
to
do
going
forward.
Specs
are,
of
course,
not
required
to
submit
PRS.
B
That
would
be
too
much
formalism,
but
they're,
often
a
really
good
way
of
collaborating
with
other
people
in
the
community
to
sort
of
figure
out
what
it
is
that
you're
suggesting,
and
so
we've
got
a
bunch
of
them
around
give
the
or
device
plugins
I
think
some
of
the
work
Yvonne
is
doing
is
related
to
her
said
my
interaction.
Spec
we've
got
some
things
where
people
are
looking
at
non
round-robin
selection
rules
to
bring
a
little
network
awareness
to
selecting
candidates
for
network
service
in
points.
B
C
A
Boo,
so
yeah,
and
just
a
note,
so
I
completely
agree
with
that
as
well.
So
don't
don't
think
you
need
a
aspect
you
to
contribute.
If
you
have
something-
and
you
understand
how
how
something
should
be
should
be
changed,
you
know
but
or
you've
had
discussions
with
us,
so
we're
so,
unlike
feel
free
to
just
send
a
simple
request.
So
a
spec
is
just
a
tool
that
we
have
in
our
toolbox
in
order
to
in
order
to
help
foster
communication.
But
it's
it's
not
it's
not
something
that
it's
absolutely
necessary
for
every
commit.
So.
B
D
F
D
D
Page
here
so
I
don't
know
if
what
was
discussed
last
week,
because
I
was
attending
but
essentially
zero
point.
One
was
was
released
like
ten
ten
days
ago,
maybe-
and
but
this
my
my
initial
feeling
while
I
was
reading,
he
preparing
imagery
is
testing
hokum
etcetera
was
that
it's
actually
a
very
solid
for
what
it's
meant
to
do.
I
call
the
examples,
then
everything
was
flown.
Is
he
working
and
I'm
pretty
pretty
much
proud
of
it?
C
D
We're
just
talking
today
here
that,
when
I
joined
in
October
and
SMU
sailing
on
a
single
host-
and
today
we
are
striving
to
make
it
work
on
multi
clusters
right.
So
it's
a
long
way
the
last
tournament.
So
definitely
the
next
steps
will
be
40.2.
I,
particularly
don't
have
any
special
list,
something
that
I
would
want
to
see.
They're,
adding
on
everything
that
we
have
in
the
queue
already
is
good.
Not
not
we
mean
we
have
discussed
some
some
things
internally
and
trying
to
figure
out
what
actually
will
be.
D
Next
from
what
I
can
tell
you,
it's
something
that
we
are
going
to
go
for
like
zero
countries,
we
are
going
to
go
for
for
Capcom
North
America
to
have
it
released,
I
hope
to
see
gosh
water
I
mean
once
we
we
made
our
the
first
to
have
this.
You
know
the
first
one
so
yeah.
Maybe
we
need
to
try
to
figure
out
the
name
just.
B
D
Know
we
to
be
absolutely
absolutely
Frank
here
when
I,
when
I
approached
it
like
very
first
service,
I
was
like
yeah.
Is
it
going
to
be
when
alpha
is
like
pre
alpha
or
alpha
alpha?
Something
like
that,
but
but
my
final
final
feeling
after
I
released
it
after
I
did
all
the
tests
I
think
like
playing
with
it
for
almost
a
whole
day.
It's
actually
pretty
solid
for
what
it's
meant
to
do
and
I'm
really
proud
of
it.
So
we
should
exponentially
if
higher
standards
for
the
next.
B
B
D
Yes,
that's
what
I
was
going
to
say
it
actually
pays
off,
I
mean
even
even
given
the
fact
that
actually
0.1
was
released
with
our
daughter,
testing
system
etc.
It's
is,
is
has
been
tested
on
all
the
other
costs
that
we
we
do
support,
and
it
was
interesting
to
find
I
mean
the
way
that
we
found
that
there
is
a
new
CV
coming
was
because
our
tests
started
failing
from
it
obvious
right,
I
mean.
B
That
was
that
was
super
fun.
It's
like.
Why
are
the
cluster
suddenly
failing,
and
the
answer
was
because
something
changed
how
you
engage
with
certain
aspects
of
the
security
and
everyone.
Everyone
did
the
right
thing
by
just
literally
dumping
the
fixes
on
to
their
clouds.
At
the
same
time
as
the
announcement
out
until
we
pick
your
head
up
and
look
around
I
think.
D
D
So
speaking
of
testing
I
think
that
that
our
test
works
actually
project
is
pretty
cool
in
need
and
I
would
vote
for
it
to
get
its
own
life
in
a
separate
report.
If
someone
asked
me
I
don't
know
if
this
is
the
time
to
talk
about
it,
but
essentially
I'm
sure
that
there
are
a
lot
of
other
project
that
not
can
take.
E
Yes,
I
think
that
topic
deserves
some
discussion,
because
the
troubles
that
we
have
seen
yesterday
with
CI
they've
caused
by
go
mode
downloading
dependencies
and
some
servers
went
available.
That's
why
we
had
some
crash
and
also
the
time
of
the
downloading
all
dependencies.
It's
about
two
minutes
for
every
image
that
we
built
so
rendering
might
help.
B
B
Layer,
caching
would,
and
for
most
things,
because
you're
going
on
time,
you
would
pick
it
up
from
the
layer
cache
that
it
could
run
for
previous
examples.
So
even
though
you're
running
in
parallel,
no,
you
don't
win
every
time,
but
but
since
you
don't
change
mods
very
often
you
find
yourself
in
a
situation
where
you
usually
win
now,
if
we're
not
usually
winning
and
finding
this
in
the
layer,
cash
that's
distressing
and
unfortunate.
B
B
Way
it
works
is
that
when
you
a
build,
comes
up,
it
assists
means
with
the
cash,
and
so
they
come
up,
and
so
cash
is
good
is
preceded
with
them,
but
they're
not
shared
cashes
right.
Every
every
bill
gets
assisted
with
cash,
though
you
go
through.
You
run
you,
let's
see
you
make
a
goemon
change.
You
run
through
each
of
these
builds
gets
associated
to
cash,
and
then
later
it
will
go
through
an
associate
the
cash
cash
with
the
build
again,
or
at
least
a
you
copy,
the
cash
again.
B
E
E
B
Should
be
the
case
that
usually,
if
you've
not
changed
goemon's
and
it's
not
changing,
often
you
get
a
fight.
Akash
library
has
the
go
launch
you're
looking
for,
and
so
you
know
we
download
them
now
again.
I,
don't
know
that
circles,
you
guys
Kashi
working
properly.
They
have
been
having
a
challenging
little
while,
but
I
mean
you
there's
a
trade-off
here
in
terms
of
CI
time.
E
B
No
not
slightly
wrong
information
like
these
depressing
as
the
way
things
are
supposed
to
work
so
I
guess
we
need
to
try
and
figure
out
whether
or
not
we're
actually
not
getting
the
correct
behavior.
If
there's
a
reason
why
you
know
so
like
the
first
thing
I
would
check
is
to
see
if
we
turn
on
docker
later
caching
for
all
our
stuff,
we're
certainly
paying
them
for
not
really
our
caching.
So
a
lot
to
see
okay
well,
let's
definitely
think
of
it
about
that
because
and
stable
stable
builds.