►
From YouTube: Network Service Mesh WG Meeting - 2018-07-13
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
A
D
D
D
A
B
A
E
I
mean
we
could,
if
some,
if
folks,
wants
to
play
with
it
with
what
we
have
already
I
mean
it
would
be
great
to
have
an
extra
feedback,
or
you
know,
reports
on
some
issues.
If
they
don't
want
to
build
images
by
themselves,
then
having
them
or
the
docker
would
be
great,
it's
not
mandatory,
but
it
would
be
nice
yeah.
A
I
I
think
it'd
be
a
good
idea
to
stew
to
start
building
out
some
of
that,
even
though
the
functionality
is
not
really
there
to
to
run
we're
close
yep,
so
I
think
it'd
be
good
to
start
getting
this,
and
so
that
way
that
it's
been
testing
and
baking
for
a
while.
We
can,
as
especially
especially
with
the
attention
coming
of
better
continuous
integration,
testing.
A
A
There's
there's
a
couple
bugs
that
I
found
in
the
make
file
that
I'm
going
to
fix
today,
but
basically
you
should
be
able
to
type
make
and
and
have
it
work,
I,
don't
know
if
this
works
in
the
OS,
X
or
Windows
environments,
it
definitely
works
in
the
Linux
environment.
So
if
anyone
wants
to
test
on
my
their
Windows
or
our
OS
X
I
make
contributions.
That
would
be
very
welcome.
A
I
had
a
I
had
a
talk
with
actually
I
was
taking
a
look
at
at
potential
option
and
I.
Tell
me
what
you
guys.
Think
of
this,
so
I
was
I
was
thinking
of
perhaps
some
form
of
a
of
a
tunnel
spider
as
a
as
a
mascot.
You
know
it's
constantly
weaving
the
web
and
constantly
building
tunnels
and
and
connecting
things
together,
and
we
can
find
a
way
to
make
it
really
cute.
D
B
A
A
E
E
A
A
D
D
D
Anyone
so
the
one
issue
I
have
I
wanted
to
bring
up.
That's
that
is
hard
to
deal
with
is
there's
a
lot
of
run,
a
lot
of
dot
MD
files.
That
sort
of
tell
you
bits
of
how
to
how
to
do
things
and
I
decided
to
write
one
from
scratch.
So
I
don't
know
whether
I
should
so
I'm
going
to
do
a
but
a
pull
request
with
just
this
file,
and
then
we
hopefully
that'll
start
the
discussion
of
whether
we
want
to
consolidate
this
the
startup
and
build
documentation
or
not
start
up,
build
and
run.
D
So
it's
just.
You
can
wait
to
the
pull
request
to
start
that
discussion.
It
might
be
better
because
it
looks
like
we
have
a
lot
of
things
on
the
agenda
today.
Yeah.
D
The
general
idea
of
having
three
or
four
dot
MD
files
that
we
have
to
if
we're
gonna
have
multiple
ones,
then
we
just
have
to
organize
them
in
a
hierarchy
like
will
agree
with
each
other.
It
makes
sense
and
look
like
oh
you
know
so,
but
it's
just
a
discussion.
We
have
to
go
given
this
I.
Will
I
see
that
note
here?
But
but
let
me
don't
know
why
I
didn't
look
at
this
yet
because
this
must
have
just
pulled
up
in
a
recent
poll
request
so
or
in
a
recent
poll.
D
A
A
C
It
was
a
say
yeah,
so
after
talking
with
Edie
and
I
had
started
some
initial
work
on
this
and
I
think
I
was
going
at
different
approach.
So
I
was
looking
at
kind
of
a
general
overview
versus
focus
on
working
with
the
NSM
project,
so
I'm
redoing
the
write-up
to
be
focused
on
what
do
we
need
in
this
project
from
NSM?
So
where
is
that
integration
point
and
talking
with
them?
C
It
may
be
better
to
open
like
a
issue
or
a
pull
request,
or
something
similar
to
and
the
kubernetes
side,
where
you
have
the
kubernetes
enhancement
proposals,
the
cuts
or
something
like
that.
So
have
a
essentially
a
write
up,
which
could
be
an
issue
or
whatever
and
say,
here's
the
project,
the
CNF
project,
and
how
we're
deploying
these
containers
in
kubernetes
and
where
they
fell
to.
C
We,
where
we
can't
use
the
kubernetes
ways
or
build
them
containerized
without
stepping
outside
of
what
kubernetes
offers,
which
is
the
networking
side
is,
is
what
we're
needing
so
focused
more
on
that
and
highlighting
so
give
an
overview
which
I
had
already
been
working
on.
But
I
want
to
highlight
those
parts
where
we
were
actually
blocked
and
there
are
specific
areas
where
we're
just
fully
blocked
so
I'm
pivoting.
B
Because
I,
you
know,
obviously
being
able
to
meet
those
needs,
is
certainly
among
our
objectives
and
and
I
think
being
able
to
get
something
you
can
use
sooner
rather
than
later.
At
least
for
me
personally
is
something
of
a
priority,
and
so
you
know
getting
an
issue
where
we
could
sort
of
dump
them
out
of
the
Navy
from
there.
We
can
break
through
that,
but
it's
the
smaller
more
actionable
issues
thanks,
B
he's
an
excellent
idea.
C
B
F
Now,
all
those
groups
are
doing
it,
it's
creating
some
SSL
Certificates
TLS
certificates
and
creating
those
as
secrets
in
the
queue
current
is
in
the
Kuban.
It
is
using
the
kubernetes
api
s--
and
all
this
can
be
automated
using
an
init
continuous.
So
when
somebody
is
trying
to
deploy,
they
can
just
deploy
a
UML
file
and
an
init
container
comes
up
and
does
all
this
work
for
us.
So
that
would
be
my
next
step
once
we
once
I
get
some
feedback
about
this
approach
and
what
we
are
doing
it.
F
A
And
definitely
everyone
has
worked
on.
One
thing
source
is
like.
Thank
you
very
much.
You
know,
even
if
it's
not
checked
in
at
the
moment
or
anything
like
that,
like
we're
very
grateful
for
the
for
the
work
that's
been
put
in
just
backtracking.
A
little
bit
recall
one
of
the
rien
ounce
nents.
Remember
that
there
is
a
cloud
native
network
function,
seminar
at
the
open
source
submit
that's
gonna,
be
on
Tuesday
August
28th
in
the
afternoon.
A
A
Review
of
development
activity,
so
we
had
a
few
starting
with
issues
closed.
We,
the
ones
that
were
the
most
prominent
to
me,
where
we
had
the
shell
check
was,
has
been
added
to
our
bills.
So
you
can,
whenever
you
add
any
shell
scripts
it'll
automatically
be
a
part
of
the
build
and
and
help
find
issues
with
portability
or
or
otherwise,
and
the
integration
tests
are
now
Rena
bowl
from
console
from
both
console
and
Travis.
So
you
should
be
able
to
to
run
the
integration
test
and
see
them
work
from
the
pull
request.
A
We
now
have
a
pluggable
architecture
and
it
has
done
a
tremendous
amount
of
work
towards
getting
us
to
this.
To
this
point,
so
basically,
we
we
want
to
have
the
project
set
up
so
that
we
can
add
and
remove
plugins
based
on
the
beads
of
of
whatever
it
is
that
this
being
that
is
being
built.
So
this
is
an
important
step
towards
it.
B
B
We
can
make
it
trivial
for
people
to
plug
in
replacement
components
for
various
things,
but
the
biggest
example
of
that
would
be
game
play
where,
basically,
if
you
have
a
different
data
plan,
you'd
like
to
use
it
with
MSM,
you
really
should
just
be
as
simple
as
write
your
data
playing
plug-in
and
then
you
know,
take
the
MSM
plug-in
and
plug
that
in
to
replace
a
minute.
Has
the
data
playing
that
you
want
to
use
and
then
away
you
go.
A
Yeah
there's
other
other
benefits
as
well,
like
even
the
configuration
management
is
done
as
a
plugin.
So
if
you
want
to
swap
out
the
configuration
management
for
for
one
that
integrates
more
closely
with
your
with
your
operational
environment,
this
this
makes
it
possible
so
so
doesn't
definitely
and
also
greatly
increases
testing,
because
you
can
mock
out
pretty
much
any
any
major
component.
A
E
Yeah
sure
so
to
be
able
to
run
m2m
testing.
Basically,
we
need
three
components:
an
SM
client,
an
SM
as
a
server
and
the
NSC
which
offer
its
channel
or
whatever
service
it
or
offers
to
the
NSM,
and
so
we
had
an
assembler
end
and
an
SM
and
I
just
pushed
lost
beats
for
NSC
and
what
it
does
is
so
the
way
it
works
now
initially,
when
NSC
comes
up
same
as
an
SM
client,
it
calls
for
an
SM
and
basically
gets
the
socket
for
further
communication.
And
then
it
starts.
E
Nse
starts
the
JRPG
server,
which
is
waiting
for
connection
requests
and
when
there
is
a
client
which
is
an
assemble
and
it
needs
to
connect
specific
network
service
to
the
specific
channel,
then
NSM
client
calls
to
the
NS
m
and
then
NS
m
on
behalf
of
the
client,
calls
to
the
NSC
to
get
the
information
required
for
programming
of
the
data
plane.
So
this
chain
is
now
working
and
basically
the
next
step
would
be
to
add
the
data
plane
programming
for
or
for
these
two
NS
m
client
and
an
NSC.
A
B
Probably
worth
noting
there
so
if
I
first
look
backs
the
plugin
and
the
log
plugin
is
also
backed
by
loggers
and
we're
currently
doing
what
I
sort
of
call
and
what
I'm
clots
are
calling
logging
by
label,
which
is
when
you
set
up
a
logger,
a
longer
plugin
used
to
say.
Okay,
these
are
the
sense
of
labels.
I
want
all
of
my
log
messages
to
carry
and
they
they
get
at
it
because
loggers
logs
in
Jason
and
one
of
the
nice
things
about
that
is
when
you
then
go
to
configure
a
lot
of
levels.
A
Yeah,
that's
why
I
had
loggers
onto
the
list
of
stuff
that
was
added
in
as
well.
So
we've
already
spoke
about
the
make
file.
We
spoke
about
the
documents
move,
we've
had
work
more
code
cleanup,
and
this
was
also
peers
merged
in
the
last
two
weeks,
as
opposed
to
last
week's
we
do
not
meet
last
week
so.
A
B
So
not
the
last
scene
are
you
meeting
by
the
one
before
we
want
to
sing
everything?
Is
it
look
hey?
You
know
it
looks
like
there's
frost
here,
one
only
be
to
become
a
student
working
sub-project.
The
other
would
be
to
become
a
career,
Diggy's
working
group
and
the
opinion
that
came
back
from
sig
networking
was.
B
We
would
really
like
to
sponsor
you
as
a
working
group
rather
than
was
a
sub
project,
which
I
think
is
a
perfectly
good
good
answer,
and
so
I
opened
the
pr
for
that
we're
going
through
this
sort
of
the
normal
dance
there.
Where
you
know
the
folks
who
think
networking
who
are
sponsoring
us
need
to
speak
up,
there's
a
little
bit
of
confusion
right
now.
Somebody
closed
the
PR
because
they
thought
it
should
belong
and
sig
networking
proper.
B
B
D
B
Yeah,
so
being
a
being
a
suit
network,
some
project
would
basically
mean
that
we
would
do
the
network
service
mash.
You
know
into
the
kubernetes
repo
and
as
a
sub
project
of
cig
networking
and
in
Signet
working
has
said,
they'd
like
two
things
about
network
service
measures
very
very
much.
The
first
thing
they
like
is
that
we're
actually
technically
orthogonal
to
the
standard
kubernetes
networking.
B
They
think
that
is
awesome,
but
the
second
thing
they
said
is:
they
also
recognize
the
possibility
that
we
could
be
use
to
solve
problems
like
multi-tenancy
for
kubernetes
networking,
and
so
at
this
time
they
don't
really
want
us
to
strictly
live
under
sea
network
in
the
kubernetes
repo.
They
like
the
fact
that
we're
working
independently
orthogonal
e
it
preserves
this
as
a
options
best
for
them,
and
so
for
that
sort
of
thing
a
working
group
is
a
much
better
structure.
I
think
this
is
actually
probably
good
for
us.
B
I
think
we
can
move
much
more
quickly
in
an
independent
repo,
not
as
a
sub
project
right
now,
then
we
would
be
able
to
move
as
a
project
of
state
network
and
more
a
young
growth
state.
Such
that
this
is
really
the
best
fit
for
the
project.
The
point
I've
made
repeatedly
to
save
networking
and
I
think
we
sort
of
generally
agree
about
that
here.
Is
we're
really
perfectly
happy
to
go
with
whatever
disposition?
B
D
B
It
turns
out
network
service
dimension
if
you
were
to
choose
to
apply
to
the
multi-tenancy
problem,
is
an
insanely
clean
solution
to
that
problem?
It
avoids
a
whole
lot
of
problems
that
you
might
otherwise
experience
in
trying
to
slowly
cut
and
see,
and
that
that's
one
of
the
reasons
that
was
expressed.
Why
they
they
like
us
as
a
possibility
for
that.
H
H
B
And
thank
you.
We
should
probably
go
talk
to
them
about
it.
Let
me
go
try
and
get
in
their
agenda
and
get
some
slides
together
and
talk
about
the
other
thing.
I
do
want
to
be
mindful
of
is,
and
this
is
something
right,
probably
what
Tim's
guidance
so
I
may
not
do
it
terribly
immediately.
Is
that
I
want
to
make
sure
that
we
presented
in
a
way
that
is
comfortable
for
the
Signet
working
group?
B
A
Yeah
I
think
short
term.
The
plan
is
pretty
straightforward
and
it
plans.
The
short
term
plan
is
pretty
much
the
same,
regardless
as
to
where
we
end
up
as
a
project
in
from
a
from
an
organizational
perspective
which
is
just
get
the
product
out
there,
get
it
in
people's
hands,
find
use
cases
make
sure
that
we
map
to
them.
A
You
know
so
it's
really.
This
is
really
about
the
long
term.
You
know
where,
where
do
we
set?
How
do
we
interact
with
other
with
other
groups?
What
what?
What
credibility
do
we
have?
Besides,
we
have
something
working
you
know
like
how
do
we
interoperate
with
other
with
other
groups,
so
I,
think
I
I
think
this
is
not
as
important
for
the
short
term
but
incredibly
important
for
for
long
term.
A
So
I've
had
an
action
item
for
the
a
multi-tenancy
group
and
it
doesn't
doesn't
need
to
be
this
particular
week.
We
can
work
even
just
working
out
when
the
right
time
to
talk
with
them
and
keeping
that
on
the
on
the
back
burner
would
be
would
be
enough
for
that.
So
so
I've
added
that
on
is
there
any
other
questions
on
this
particular
topic
or
shall
we
move
on.
A
A
A
So,
and
in
the
long
run
we
with
we
should
also
it
appears,
it
will
also
get
access
to
some
interesting
hardware
as
well,
so
we
should
gain
access
to
DP,
DK,
enabled
Nick
guards
and
there's
also
the
possibility
of
gaining
access
to
two
hardware
that
is
capable
of
speaking
more
esoteric
protocols.
So
not
a
guarantee
at
this
particular
point
for
those
for
those
items
and
at
the
very
minimum
right
now
we're
starting
small.
So
that's
that's
where
we're
at
with
that,
and
one
of
the
things
that
I'm
also
going
to
do
is
I.
A
Have
a
kubernetes
install
installer
for
ansible.
It
works
on
both
send
to
us
and
into
systems,
and
I
wrote
this
for
the
Linux
Foundation
for
their
IT
group,
because
we
needed
it
in
open
daylight
and
turns
out
that
CN
CF
CI
Group.
It
appears
that
they
don't
have
something
that's
equivalent
to
this.
So
I'm,
going
to
pitch
this.
To
that
particular
group.
I
know
there's
a
couple
people
from
that
group
who
are
here
so
it's
a
bit
of
a
preview
on
something
that
we're
going
to
do
on
that.
A
But
basically,
what
we
can
do
is
we
could.
Potentially,
we
could
potentially
use
this
to
install
nodes
on
to
that
cluster.
If,
if
there's
no
current
way
to
do
that
from
from
a
CN,
CF
CI
tool
that
already
exists
and
if
one
doesn't
exist
and
we'll
see
a
wolf
work
out
a
way
to
to
integrate
and
potentially
donate
some
of
this
to
to
the
CN
CF
CI
group.
A
C
A
C
This
is
Taylor
with
the
cross
cloud
provisioner
portion
as
the
cross
cloud
CI
project
does
provisioning
of
packet.
It
can
provision
kubernetes
clusters
on
packet.
C
I've
gotta
issue
with
this
would
be
use
case.
Does
it
fit
what
Innes
emanating
and
I?
Definitely
like
the
idea
of
the
ansible
roles
being
available
to
Linux
Foundation,
the
IT
group
as
a
whole
I
think
that's
going
to
be
really
useful
and
then,
as
far
as
what's
most
useful
for
in
assume.
If,
if
this
works
out
that's
great,
and
definitely
we
can
take
a
look
if
desired
at
the
cross
cloud
stuff,
we
talked
about
that
before
and
I
didn't
know
where
that,
where
things
were
as
far
as
the
automated
testing.
A
Yeah
I
I
believe
yeah,
that's
my
my
understanding,
so
cross
cloud
has
something
to
install
on
packet.
The
absol
role
was
designed
for
originally
for
spinning
running
on
a
set
of
of
VMs
or
bare
metal
hardware
that
that
already
exists,
and
so
a
cloud
provider
was
not
present,
and
it's
in
that
scenario,
but
yeah.
If,
if
it
turns
out,
we
can
just
use
the
cloud
CI,
Grubin,
ATS
provisioner
to
spawn
and
kick
these
things
off
or
we
can
make
a
commit
to
to
help
in
that
path
like
I
think
that
would
be.
H
A
C
There's
really
a
large
number
of
tools
out
there.
So
it's
hard
to
say
this
is
the
one
to
use
without
really
looking
at
the
use
case
in
deciding
the
the
cross.
Quad
CI
has
a
lot
of
components
and
if
you
go
there
or
you
look
at
the
dashboard
at
CNCs,
CI
there's
a
lot
of
different
parts.
It
doesn't
mean
that
all
of
those
are
necessary.
The
the
part
that's
probably
most
relevant,
for
this
would
be
just
the
cross
cloud.
C
The
multi-cloud
provisioner
portion
of
drop
a
Lincoln
in
there
and
it
supports,
will
all
the
clouds
that
are
currently
active
on
the
production
which
AWS
has
your
GCE
IBM
cloud
is
actually
the
container
service,
so
it
does
support
like
using
different
container
services
and
then
packets,
an
example
of
bare
metal.
So
it's
allocating
the
hardware
resources
from
packet
using
the
API
and
then
provisioning
kubernetes
on
this
and
then
OpenStack
we
just
added
vSphere.
C
Recently
it's
could
target
other
things,
VMs
and
or
your
run
hardware
as
well,
because
we've
done
like
pixie,
boot
and
stuff,
but
again
there's
a
lot
of
choices.
I
mean
digital
rebar
and
crib.
They
do
bare
metal.
So
I
think
would
be
good
is
to
look
at
how
the
testing
should
happen.
We're
saying
earlier
running
it
daily
and
looking
at
the
builds,
but
looking
at
what
we
want.
As
far
as
output
I.
A
Okay
and
the
last
see
the
last
major
item
on
the
agenda
is
the
what
we're
doing
at
the
documentation.
So
last
time
we
met,
we
ran
into
a
couple
issues
with.
How
do
you
review
things?
How
do
you
tell
what's
changed
within
the
dot
within
the
documentation
and
github,
despite
the
fact
that
it
stores
everything
in
the
git
repo
and
the
wiki
does
not
extend
the
github
repo
tools
to
that
would
to
that
wiki?
A
The
second
second
step
is:
we
want
to
set
up
Hugo
to
generate
a
a
site,
and
once
we
have
Hugo
generating
a
site,
then
we're
looking
at
a
tool
called
Detlef
eye
which
can
which
is
capable
of
triggering
on
a
bond
a
commit,
rebuilding
the
site
and
then
and
setting
of
setting
it
up
with
hosting
and
then
pointing
network
service
mesh
that
I
out
towards
it.
And
so
so.
A
D
E
A
A
A
B
A
D
D
Quickly,
I
I
am
chair
of
the
NF
e
Seguin
sent
us
so
I
always
have
a
back
agenda
of
looking
at
what
and
asking
people
what
what
packages
we
need.
An
NF
e
in
the
CentOS
distro
and
just
keep
that
in
mind,
and
if
anybody
wants
to
sponsor
this
or
some
combination
of
packages,
we're
talking
about
as
met
as
being
in
the
NF
e
cig
and
available
on
the
sense
on
CentOS
distro,
we
have
a.
We
have
a
process
for
that.
So.
D
A
And
even
if
the
network
service
mesh
itself
doesn't
end
up
in
it,
we
should
start
thinking
about
our
dependencies
and
working
out
what
what
should?
What
should
land
in
there
so
like
well,
I'll,
give
a
simple
example
like
if
TPP
is
in,
it
is
a
dependency
having
a
VP
V
package
would
be
useful
too.
That's.
D
Yeah
yeah
yeah.
Absolutely
that's
a
good
point,
though
a
very,
very
good
point:
anything
you
can
think
of
I
want
a
sponsor
or
anybody
wants
to
sponsor
a
package,
and
so
it
is
something
that
not
a
zero
amount
of
work,
just
pushing
the
package
through
the
system
and
so
on
and
so
forth,
then
building
and
so
on.
All
that
all
this
sort
of
mechanics
of
it
is.
It
would
be
quite
helpful
to
have
people
from
this
community
involved
in
the
NF
ecig
for
CentOS.
D
A
Okay,
well,
thank
you,
everyone
for
for
attending,
and
our
next
meeting
is
scheduled
for
next
Friday.
So
start
thinking
of
agenda
items
throughout
the
week
and
feel
free
to
start
adding
to
them
earlier
on
so
I'd
like
to
try
to
get
agenda
items
in
a
bit
a
bit
earlier
if
we
can,
but
we'll
always
have
the
agenda
bashing
at
the
start
as
well.
So
even
if
you
forget-
or
something
comes
up
the
last
minute
like
don't
don't
sweat
it
like,
we
can
still
still
get
it,
but.