►
From YouTube: Network Service Mesh Meeting - 2019-06-04
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
B
B
So,
starting
from
the
starting
of
top
so
welcome
to
the
next
network
service
mesh,
meaning
we
have
three
meetings.
We
have
this
one
which
occurs
every
Tuesday
at
8:00
a.m.
Pacific
time
and
SM
document
call
which
first
weekly
on
Wednesday
at
8:00
a.m.
Pacific,
and
we
have
the
MSM
use
case,
which
appears
every
second
fourth
Monday
at
8:00
a.m.
Pacific.
We
do
have
a
NSM
use
case
meeting
coming
up
this
next
Monday.
B
We
are
also
participating
in
the
CNC
up
telecom
user
group,
which
occurs
every
first
and
third
when
they
had
8:00
a.m.
Pacific,
I
noticed
a
trend
here
and
so
that
one
just
occurred
so
it'll
be
Matt
next
week,
but
the
week
after.
For
that
one,
there
is
also
a
CNC
F
networking
working
group,
which
occurs
every
two
weeks
on
Tuesday
at
9:00
Pacific
time
we
so
we
can
remove,
cube
con
you,
we
don't
we've
already
gone
over,
that
it.
B
We
have
cube
con
China
coming
up
and
we
have
an
intro
and
the
maintainer
talk,
and
there
is
also
the
second
telecom
user
group
work
working
group
kickoff
that
is
occurring
as
well.
We
have
the
BBK
user
space,
which
occurs
a
week
before
ons
I,
believe
it's
like
the
18th
and
19th
of
September
in
Bordeaux.
So
I've
submitted
a
paper
into
that
talking
about
similar
to
Fido.
B
B
C
C
B
So
we
have
cube
con
coming
up
in
November.
They
call
for
paper
ends
July
12th
and
that
is
I
was
going
to
that's
gonna,
be
on
from
November
18
through
20.
First,
there
is
a
new
to
be
announced
to
educating
world
coming
up
as
well,
and
so
that
one
maybe
I
need
to
talk
with
the
organizer
to
see
exactly
where
he
wants
to
take
it.
But
if
there
is
a
networking
component
or
networking
path
which
sounds
like
there
is
going
to
be,
then
we
may
want
to
see.
B
B
E
E
You
and
I
tried
to
search
for
anyone
mentioning
us
by
our
twitter
handle,
as
well
as
NS
m
network
service
mesh,
all
one
word
or
three
words
and
tried
to
retweet
as
much
as
possible,
so
we
gained
45
more
followers
over
the
course
of
the
last
three
weeks
I
slowed
down
with
following
folks
and
followed
over
just
over
50
new
folks,
and
we've
got
a
total
of
133
tweets
and
fun
fun.
News
I
actually
went
to
a
park
in
Barcelona
where
there
was
a
labyrinth
and
statues
of
Ariadne.
E
It
was
a
beautiful
space
in
Barcelona
and
I.
Think
I've
posted
the
recap,
video
of
the
intro
and
deep
dive,
and
if
there
are
any
other,
oh
actually,
I
can
also
post
a
recap
of
the
FCI
o
presentation,
because
those
videos
have
been
released
as
well
as
the
Linux
Foundation
networking
video,
so
I'll
get
those
out.
This
week
too,.
A
B
F
F
We
have
the
branch
now
I
still
have
to
figure
out
how
we
are
going
to
publish
the
images,
because
you
know
the
credentials
are
somehow
in
the
CIA
automatic
way
of
publishing
the
these
images.
But
that's
something
to
to
be
to
be
feared
out
and
to
happen.
I
think
that
we're
in
a
good
situation,
mainly
in
a
good
position
now.
A
E
F
B
Don't
forget
that
a
lot
of
this,
our
infrastructure,
not
our
infrastructure,
explicitly,
but
a
lot
of
the
go
infrastructure
implicitly
runs
on
and
docker
runs
on
Google
Cloud,
actually,
dr.
one
doesn't
but-
and
so
so
there
may
have
been
some
issues
that
we
that
we
may
have
had
intermittently,
because
of
that,
something
that
we
should
consider,
though,
is
there's
a.
B
We
can
spin
up
anathema
server
in
some
of
these
locations
that
can
help
store
some
of
these
some
of
this
information
some
of
its
context,
so
it
may
make
sense
to
to
bring
up
a
proxy
and
store
some
of
this
and
for
some
of
the
packages
to
help
with
the
go
mod.
If
that
becomes
an
issue
over
time,
although
a
won't
help
with
network
connectivity
out
for
the
CI,
so
that's
the
only
probably
run
into
unless
we
can
work
out
what
clout
they're
running
it
in
what
region,
but
we
probably
won't
get
that
visibility.
A
B
We
did
claim
to
have
backups
on
line
two
to
deal
with
this
problem,
but
yes,
there
they're
back
their
backup
system
is
relying
upon
their
rather.
Their
messaging
service
is
really
reliant
on
their
production
system,
which
I
guess
works
for
most
things.
As
long
as
your
networking
isn't
that
thing
that
fails,
yeah.
A
B
So
this
so
this
is
basically
a
more
verbose
version
of
Ted's
Excellent
code
of
conduct,
and
so
basically
it's
just
to
the
main
things
is
banning
the
use
of
behavior
around
sexualized
language
or
imagery
personal
attacks,
trolling
insulting
harassment
in
the
public
and
private
docks
and
people
who
don't
wish
to
be
doxed
without
explicit
permission
and
a
catch-all,
unethical
or
unprofessional
conduct.
The
one
thing
to
call
out
on
the
bottom
is
the
the
section
on
maintaining
and
mediator.
B
So
if
there
is
a
problem
between
a
community
member
and
the
and
the
maintainer
Zoar
a
problem
between
maintainer
and
maintainer,
they
do
offer
the
services
of
a
mediator
which
is
fishy.
Chowdhury
I
have
not
met
me.
She
shatter
II,
so
I,
don't
know.
I
can't
say
anything
for
against
in
that
scenario,
but
I
do
have
some
confidence
in
in
the
people
that
the
CN
CF
has
brought
on
for
this
kind
of
stuff.
So.
A
B
B
A
B
A
A
Cool
so
I
think
it's
probably
good
for
people
to
go.
Take
a
look
at
redo.
This
I've
taken
some
time
to
look
through
it
in
detail
and
was
actually
super
pleased,
but
the
the
looking
through
notice
Hale
did
lead
me
to
realize
the
degree
to
which
it's
probably
productive,
to
give
some
folks
time
to
look
at
it
in
detail.
A
B
B
F
F
A
F
F
I
think
that
I
don't
know-
maybe
maybe
maybe
I
need
to
do
a
pass
here
again
and
check
check
what's
on,
but
from
my
personal
tastes
and
my
experience
with
the
examples
I
mean
anything
behaves
pretty
pretty
much
stable
for
for
the
type
of
project
that
we
I
mean,
like
the
status
of
the
project
that
we
are
I
mean
I,
don't
see
any
obvious
huge.
You
know,
crashes
and
misbehavior.
F
I
think
that
this
is,
of
course,
due
to
the
to
the
extensive
integration
testing
that
we
are
doing
and
the
work
that
we
did
in
the
last
couple
of
months
for
enabling
other
clouds
which
are
introducing
other.
You
know
problems
in
terms
of
delays
being
slow
being
fast
whatever,
but
then
Indian
we
obviously
have
tested
a
lot.
So,
at
least
from
my
point
of
view,
whatever
we
have
in
the
release
branch
is
something
that
we
should
be
proud
of.
So,
thank
you
all
for
putting
your
efforts
with
patches
and
with
whatever
you
are
helping.
C
A
G
A
F
A
F
Yeah
I
mean
adding
more
tests
and
yeah,
so
I
think
that
we
should
move.
Do
we
have
around
key
on-prem
on
the
call
I
mean
that's
all
for
me.
From
for
the
Andromeda
release,
I
mean
I
expect
that
before
keep
Kanchana,
we
should
get
attack
and
published
images,
something
that
can
be
shown
around.
That's
that's
my
my
expectation
and
cool.
B
F
So
the
road
map,
yeah
I,
think
that
we
said
this
last
time.
I
would
really
like
to
keep
this
this
item
here
for
us
to
remind
us
where
we
are,
what
we
do,
maybe
add
things:
remove
things
kind
of
checking
quickly.
The
status
just
just
a
kind
of
open
discussion
here,
so
release
is
clear,
more
or
less
the
examples.
We
said
it's
kind
of
a
little
bit
tight
with
the
next
topic.
Cnc
have
testbed
in
everyone's
or
me
Andy
Taylor.
F
We
were
just
having
a
good
nice
conversation
preparatory
conversation
for
the
coastal
two
together
this
week.
So
we
have
scheduled
a
number
of
calls
with
people
that
are
interested
in
to
enabling
NSM
for
the
CN
CN,
f,
CN,
CF,
CN,
f,
testbed.
We
have
summer
of
plan
there.
I
would
really
like
to
see
this
in
a
moment
based
on
the
examples
and
I.
Don't
know
we
have
discussion
tomorrow,
we'll
figure
it
out.
F
A
E
F
That's
the
point
yeah
yeah,
but
that's
a
conversation
that
we
probably
should
should
take
offline,
but
I
mean
it's
I.
I
can
have
it
now
because
we
have
today
merge
the
PR
where
we
are
renaming.
There's
no
examples.
Folder
in
the
main
report
anymore
right
I
mean
we
moved
everything
to
testing
the
sidecar
container
belongs
the
sidecar
folder,
so
everything
is
moved
to
other
places.
So
examples
are
existing
by
the
virtue
of
just
being
part
of
the
history
and
also
being
part
of
the
RCI.
F
Maybe
we
need
to
figure
out
some
plan
for
completely
removing
the
things
there,
but
okay.
This
is
something
that
that
we
can
discuss
like
also
so
for
mesially.
I
know
that
Lumina
was
supposed
to
have
something
there.
I
don't
know.
What's
the
start,
was
there
unless
Prem
from
peers,
or
at
least
give
us
some
name
that
can
join
yeah
yeah.
D
F
F
Technical
telco
user
group
meeting
and
actually
today,
someone
I,
don't
remember
name
the
name,
but
there
was
I
mean
there
was
an
initiation
of
a
white
paper
like
preparation
document.
So
SMI
was
mentioned
there
I,
don't
know
how
these
merges,
but
I
think
that
we
should
just
keep
a
nice.
But
that's
enough
from
now
crippled
forwarding
plane
runs
love.
Who
is
here?
He
started
doing
something
as
I
said
in
the
chat
today.
I
hope
that
we
learned
some
things
along
the
way
of
you
know
enabling
PPP
with
ipv6
and
not
bugging
things
there.
F
So
I
think
that
we
learned
some
things.
Maybe
we
have
still
a
lot
to
learn.
I,
don't
know
you
see,
but
we
are
definitely
on
the
path
to
to
move
this
forward.
I
hope
to
get
some
patches
for
having
like
come
today.
Pvp
is
very
deeply
embedded
in
our
test
integration
testing,
so
we're
planning
to
have
some
abstraction
there
so
that
you
can
switch
back
and
forth
between
the
different
implementations,
which
I
will
hope
will
make.
Some
people
feel
a
little
bit
better,
because
I
have
heard
some
opinions
about
this.
Everybody.
A
F
A
So
see
if
it
makes
sense
to
you,
you
know
and
see
you
see
kind
of
what
makes
sense
then
part
of
it
was
came
out
of
comments
from
conversation
at
UConn.
Where
folks
were
like
look,
you
know
the
spire
I
got
the
spiffy
inspired
guy.
Is
it
just
about
solved
Federation,
which
is
gonna,
be
super
important,
so
I'm
happy
to
hear
or
someplace
else
talk
through
a
little
bit
of
that
my
thoughts
there
and
I'm
very
much
open
to
other
thoughts
and
ideas,
and
then
please
God
review
because
see.
E
A
D
F
A
I've
talked
to
some
of
my
internal
security
folks
about
the
general
gist
of
it
might
even
even
the
security
fellow
here
at
Cisco.
You
know
yeah,
and
he
was
positive
ish
about
the
issue.
Now,
admittedly,
he
didn't
went
on
to
give
a
talk
about
how
you
continue
to
secure
things
in
the
post.
Quantum
computation
of
the
world.
E
A
A
B
B
So
now
you
have
two
or
more
DNS
servers
that
you
have
to
resolve
against,
and
so
one
of
them
being
your
primary
kubernetes
resolver,
plus
any
additional
secondary,
yeah
DNS
resolvers
that
you
that
you
attach
on,
and
so
this
one
of
the
solutions
that
we're
looking
at
is.
If
so,
if
you
look
at
the
second
image,
so
is
adding
in
a
coordinate,
sidecar.
B
A
B
A
A
Normally,
if
you
get
a
positive
response,
you
would
actually
like
to
pass
through
the
first
positive
response
you
get.
You
do
have
to
be
a
little
bit
more
cautious
with
negative
DNS
responses,
because
you
don't
want
to
pass
through
the
first
negative
DNS
response
you
get
because
it
may
just
be
that
somebody
else
has
a
response,
that's
positive
and
they
haven't
gotten
it
to
you
yet
so.
B
When
you
say
first
response
as
well,
is
that
literally
like
the
first
response,
regardless
as
to
which
one
it
is,
or
you
mean
first
as
in
like
the
kubernetes,
would
get
priority
and
that's
what
sort
of
be
the
first
response
and
then
VPN
one
regardless
of
the
time
ordering
would
be
like
the
second
response.
I.
A
Was
thinking
in
terms
of
time
ordering,
because
you
get
very
long,
latency
Ziff,
you
wait
if
you
run
through
the
materially,
it's
much
more
efficient
to
run
through
them
in
parallel
and
the
kubernetes
dns
is
going
to
have
a
huge
latency
advantage
in
that.
Not
only
is
it
running
in
the
cluster,
but,
generally
speaking,
the
kubernetes
dns
is
now
moving
to
per
node
caches
for
dns
and
as
a
result,
you
would
expect
that
if
kubernetes
DNS
has
something
to
say
about
it,
it
will
look
something
to
say
about
it
really
fast.
B
Okay,
so
something
we
should
probably
consider
in
this
scenario
as
well
is
making
sure
that
we,
of
course,
for
certain
domains.
If
you,
if
we
can
say
like
we
want
everything
from
example.com
like
all
requests,
for
example,
uncom
to
go
only
through
VPN,
one
or
VPN
two
so
I
think
we
should
I
think
it
would
make
sense
to
add
to
add
something
in
like
that.
Just
so
that
you
can
say
things
in
my
private
corporate
intranet,
you're
not
pulling
the
public
addresses
you're
only
playing
the
private
addresses
for
those
probably.
A
No,
it's
it's
having
beer
with
the
really
deep,
with
the
really
deep
DNS
guys
they
will
just
endlessly
about
it,
but
yeah
everybody
does
it
so,
but
I
mean
might
make
what
I
would
encourage
cuz
we're
getting
down
to
time
here
shortly
is
folks
to
please
go.
Take
a
look
at
this
and
comment.
This
is
gonna,
be
super
important
for
us
to
support
going
forward.
Yeah.
B
A
A
We've
got
a
good
story
where,
if
anyone
that
Rick
service
manager
and
network
service
manager
data
plane
dies
everything
rican
verges
and
if
you
lose
the
network
service
in
point,
we
auto
he'll
Andre
had
some
really
smart
ideas
where
we
could
literally
lose
all
of
the
network
service
managers
and
all
the
network
service
administered
data
planes
at
the
same
time
and
still
reconverge,
which
I
think
we're
tentatively
calling
resiliency
v2
folks
have
other
ideas
for
naming.
That's
awesome,
dynamic
rewiring.
A
So
we
have
Auto
healing
already
in
the
system,
meaning
that
if
you
lose
a
network
service
endpoint,
we
can
automatically
reconnect
a
client
to
another
network
service
endpoint
to
continue
good
continuity
of
service,
and
we
know
we
can
use
the
same
mechanism
so
that
if
you
change
the
network's
the
policy
for
the
network
service,
we
are
capable
of
dynamically
rewiring.
A
That
mechanically
and
I
think
there
are
even
ways
for
us
to
do
it
safely,
because,
obviously
you
don't
want
to
do
it
to
everything
all
the
time
that
could
be
destructive
and
that
might
that
should
allow
us
to
do
a
really
cool
example.
Where
imagine
that
you
have
a
pot
of
it's
consuming
a
network
service
and
something
is
weird-
and
you
decide
you
know
after
the
pot
is
burning
for
a
while
gee.
Wouldn't
it
be
nice
if
you
had
a
packet
capture
box
between
the
pond
and
everything
else
that
was
going
on.
A
B
A
A
So
we
at
least
know
what
the
bogies
are
right.
What
are
the
things
we
could
go
work
on
that
people
might
want
to
go
work
on
in
the
community,
because
I
know,
for
example,
like
there's
a
bunch
of
stuff
that
you
had
on
the
spec
board.
Matthew
around
metrics
and
I
know
we
got
some
of
that
done,
but
I
don't
know.
If
we've
got
all
of
that
done
and
like
Huebert
integration,
I
know
you
were
interested
in
that
Jeffrey
I'd
want
to
make
sure
that
we
got
that
represented
correctly.
A
We've
got
some
new,
interesting
things
coming
in
around
network
or
a
load
balancing
where
people
have
got
suggestions
for
that,
and
so
I
just
wanted
to
make
sure
to
walk
through
the
speck
board,
make
sure
we
were
capturing
all
the
specs
and
getting
them
filed
correctly,
so
they
could
be
worked
on
it.
People
were
just
in
working
on
them.
A
A
F
A
A
F
F
B
Will
cover
it
yeah
pretty
I
did
have
a
brief
talk
with
Geoffrey
about
the
split
data,
plane
and
I.
Think
the
concept
was
that
you
may
have
to
do
planes
that
have
different
capabilities.
That
would
be
complementary
on
the
same
node,
and
so
the
question
would
be
how
to
make
sure
like
if
one
of
them
is
supported
by
F
and
the
other
one
supported.
That's
aa
really
then
like.
How
would
you
like?
How
would
you
forward
them
on
to
the
right
things?