►
From YouTube: Ambient Mesh WG Meeting 2022 10 12
Description
No description was provided for this meeting.
If this is YOUR meeting, an easy way to fix this is to add a description to your video, wherever mtngs.io found it (probably YouTube).
A
All
right
recordings,
I
believe
welcome
to
the
second
ambient
weekly
meeting.
A
So
last
week
we
kind
of
discussed
about
giving
some
of
our
newer
folks
introduction
about
ambient
and
save
some
room
for
them
to
ask
questions,
so
we
just
send
out
the
blogs
and
the
videos,
so
I
want
to
take
a
pause
first
of
all,
just
to
ask
if
anybody
have
any
questions
about
Ambien,
they
want
to
ask,
or
would
it
be
helpful
that
we
do
a
brief
intro
or
you
feel
like
you
get
started,
so
we
can
just
get
onto
the
agenda
items
for
today.
C
A
Yeah
I
think
last
week
it
was
the
meech
and
I
think.
Also
there
was
a
lady
leaving.
You
were
interested
in
that
I.
Don't
know
if
you
guys
have
any
specific
questions
or
specific
area.
If
we
go
through
an
intro
on
his
focus
on.
D
D
C
And
I
know:
John
had
started
the
work
item
list,
I
wonder
if
maybe
that
would
be
just
the
best
place
to
start
yeah.
A
It's
a
second
item
on
the
agenda.
Yeah
I
think
that
would
be
good
because
I
think
what
makes
you
asking
for
I
don't
believe
we
have
material
out
there
as
far
as
like
the
blog
and
video,
because
they
don't
get
you
the
code
level
so
that
that's
a
fail.
Ask
okay,
so
yeah!
If
there's
no
particular
feedback,
I
think
we
should
just
go
through
the
agenda.
There's
a
couple
of
topics:
there's
a
Daniel!
Why
don't
you
start?
First.
E
All
right
first,
one
has
to
do
with
the
image
building.
Currently,
the
istio
cuttle,
at
least
that
I
can
find
is
from
the
initial
commit
when
it
was
first
murdered
or
brought
over
to
the
Upstream
repo
mix
testing.
Recent
changes
a
little
more
difficult
because
you
gotta
still
build
everything
manually.
So
I
was
wondering
what
we
could
do
to
get
that
automated
foreign.
E
A
So
I
saw
Justin
you're
presenting
this
task.
Did
you
just
the
task
of
John
created
I
guess
this
is
probably
would
what
Daniel
is
asking.
It
would
be
good
to
track
it
as
a
task
here
right.
So
people
can
try
ambient
on
the
latest
easily
yeah.
C
E
Right
so
if
you
use
the
istio
cuddle,
for
example,
that
was
published
in
the
getting
started
blog,
it
points
at
a
specific
commit,
and
so
all
the
changes
and
Bug
fixes
that
we've
done
since
won't
get
installed
by
it
unless
you
manually
Target
a
different
tag.
But
since
we're
not
it's
not
automated.
In
that
way,
I'm
having
to
do
everything
manually.
D
G
Cuddle,
it's
kind
of
a
misnomer.
It's
really
just
the
istio
release.
Like
all
the
images
you
still
cuddle.
A
All
right
yeah,
so
the
and
the
way
I
look
at
this
right
if
I'm
using
istio
116
that
build
I,
don't
have
to
build
it
right.
There
is
a
way
you
know
of
a
weekly
page,
that
we
teach
people
how
to
pull
out
the
images
so
I
think
the
question
is
it:
should
we
do
something
similar
for
ambient?
So
there's
at
least
one
user
asking
you
know
a
channel
for
sure.
D
If
that's
a
long-term
thing,
then
we
should
definitely
have
Dev
builds
set
up
for
that
branch,
but
we
don't
have
I,
think
anyone
in
the
test
and
release
working
group
anymore,
and
that
would
be
where
all
that
expertise
is
60
days.
Just
to
guess.
A
D
F
D
G
G
Branches
in
general
didn't
have
Branch
protection,
and
so
we
can't
have
something
that
publishes
a
release
with
on
a
branch
without
Branch
protection,
because
then
someone
could
go
just
merge
some
malicious
thing
without
getting
any
approvals
right,
but
recently
or
actually
I
think
the
Pierce.
So
maybe
we
can't
do
it
quite
yet,
but
it's
close
now.
F
Yeah,
if
we
get
one
of
the
other
approvers
I,
think
and
that.
A
G
A
F
I
know
I
have
thought
of
the
same
issue
before
you
know
you're
playing
with
ambient.
How
do
I
get
you
know
something
later
than
what's
in
the
blog
and
then
the
follow-up
to
that
becomes.
How
do
we
update
the
blog
to
maybe
point
to
something
later
because,
obviously
it's
pointing
to
the
original
one
or
do
we
want
to
do
that?
I,
don't
know
what
the
best
way
forward.
There
is
either.
D
So
usually,
blogs
are
static.
They
they
tend
to
be
designed
to
be
point
in
time.
Advice
related
to
a
particular
istio
resource
or
a
particular
release
in
this
case,
because
we
expect
iteration
to
happen
so
quickly.
Would
it
make
sense
to
spin
off
a
part
of
istio.io
for
ambient.
G
F
Well
and
I'm,
assuming
you
know,
as
John
mentioned,
if
it's
a
developer,
they
would
be
able
to.
You
know,
build
everything
locally
anyway,
so
this
is.
This
is
just
the
ability
to
you
know
have
somebody
who
doesn't
want
to
pull
a
commit
and
build
it
to
just
pull
from
something
that
we
built
as
part
of
the
post
submit.
H
F
I
Yeah
Lynn
I
started
looking
to
this
and
I
think
I,
really
like
the
design
and
the
direction
that
how
this
you
know
is
going
in
terms
of
build
I
think
what
we
have
probably
is:
okay,
because
I
find
why
trying
to
do
this,
it
wasn't
too
difficult
or
anything.
I
The
only
thing
I
find
is
that
I
think
of
the
document,
reference
Lane,
your
personal
repo,
and
it
wasn't
any
kind
of
particular
words
about
why
you
cannot
use
the
I.
Still,
you
know,
feel
deployment
files
and
I
look
into
that.
I
find
it's
about
the
note
Affinity,
that's
the
only
difference
it
could
be.
You
know
useful
if
we
add
a
few
words
there
why
the
deployment
file
needs
to
be
from
this.
I
You
know
personal
repository
so
that
that
part
I
think
it's
fine,
but
I
also
noticed
we
have.
We
have
this
on
the
experimental
branch.
I
My
question
is
that
you
know,
because
there
are
two
we
have
the
master.
We
have
this.
You
know
experimental.
A
So
the
sample
I
mean
yeah,
certainly
when
we
write
a
blog
and
we
were
a
little
bit
rushed
to
get
out
in
the
days.
We're
happy
to
maybe
add
that
as
an
additional
sample
in
istio.
If
that's
valuable,.
I
A
So
the
branch,
if
you
look
at
the
meeting
minutes
from
last
week,
yeah
we
kind
of
had
a
lot
of
discussion
on
that.
The
main
concern
about
not
emerging
the
branch
to
master
is
about
the
stability
and
scalability
of
ambient.
Today,
and
also
there
are
re-innovation
in
zetano
today.
So
there's
a
lot
of
changes
coming
through
yeah.
That
which
is
why
it's
in
a
branch
at
the
moment.
I
A
Yeah,
so
I
think
that
this
is
probably
a
good
way
to
get
to
the
next
task
of
which
John
open
right.
That
has
the
list
of
the
tasks
to
go
towards
merging
back
to
either
merging.
Back
to
myself,
we
decided
to
do
a
tool
version
of
istio
right
so
John.
Maybe
we
should
start
to
go
through
that
task.
You
open
now,
because
I
think
we're
done
with
Daniel's
topic.
If,
if
everyone
agrees.
G
Not
really
I
don't
really
want
to
go
over.
Every
task
in
here
I
wrote
down
a
bunch
of
tabs
that
I
thought
needed
to
be
done.
It's
not
a
conclusive
list.
G
A
lot
of
them
are
like
figure
out
what
we're
going
to
do
for
acts
and
then
once
we
figure
it
out,
there'll
be
a
big
list
of
of
specific
items
but
yeah.
Here's.
G
What
I
have
down
so
far
feel
free
to
add
more
and
I'll,
probably
add
more
as
I
think
of
things,
but
if
you're
wanting
to
contribute
I
would
go
here,
find
one
that
sounds
interesting
and
that
you
think
you
know
how
to
do
and
put
your
name
next
to
it
or
convert
it
to
an
issue
and
then
assign
yourself
the
issue.
A
Yeah
so
I
do
have
a
question.
Spuno
dementia
to
its
own
chat,
is
that
necessary
for
data,
because
the
the
reason
I'm
asking
is
I
find
that
it's
easy
to
kind
of
provide
explanation
to
our
user
as
far
as
the
channel,
it's
kind
of
like
a
cni,
it's
part
of
your
infrastructure.
G
D
It
sounds
like
the
larger
question
here
between
I've
seen
some
stuff
about
splitting
Z
Channel,
as
well
as
things
about
how
Z
tunnel
interacts
with
revisions.
The
larger
question
is:
what
does
an
upgrade
look
like
until
we
have
a
really
good
idea
of
what
that
is.
I,
don't
know
that
we
can
make
statements
about
how
it
ought
to
relate
to
revisions
or
what
chart
it
ought
to
be
in.
A
So
I
guess
that
goes
back
to
a
fundamental
question.
How
is
the
Z
tunnel
today
installed?
I
thought
it
seems,
thought
part
of
the
history
of
CMI,
but
I
may
be
wrong.
I
A
Agree
with
this
point
that
we
need
to
think
so
upgrade
too,
but
now
that
I
see
it's
part
of
a
co-d,
it
does
make
sense
too,
to
not
be
part
of
the
seod.
I
G
That's
not
that's
not
how
the
Z
tunnel
works,
though
it
works
more.
Like
the
cni
one
requester.
We
may
have
a
short
period
of
time
where
you
can
have
two
for
a
transition,
but
it's
not
like
sidecars,
where
you
can
have.
You
know
10
different
versions
indefinitely
in
there
each
one
is,
you
know
certain
pods
use
certain
versions,
it's
kind
of
more
big
to
the
infrastructure
layer.
G
I
G
It
would
be
weird
to
have
10
Damon
sets
on
every
node
for
every
version
right,
okay,
I
suppose
we
could
do
it,
but
I
don't
think
we
should.
We
should
really
the
concern
then.
G
A
G
G
We
have
some
design
dock,
that
is
in
progress
that
we
should
probably
share
it
on
how
this
will
work
so
I
think
there's
there's
a
lot
of
discussion.
That's
already
already
happened
here
between
some
folks
yeah.
A
D
A
D
D
G
B
Been
talking
about
it
for,
like
the
internal
Google
gke
case,
I
I,
don't
think
he's
done
much
on
the
external
side.
B
Well,
the
short
answer
is
internally
in
GE,
we're
just
going
to
do
a
Damon
set
upgrade
and
the
downtime
is
quite
minimal
if
you
get
the
settings
right
so
that
should
work
in
OSS
too,
but
he
hasn't
like
written.
That
down.
Does
that.
A
B
And
by
the
way,
I'm
fine,
with
volunteering
Martin
to
own
this
I
I
didn't
tell
him
to
do
it.
He
just
hasn't
gotten
to
it.
Yet.
A
B
A
A
A
B
I
think
we'll
be
ready
to
talk
about
it
informally,
probably
next
week,
we're
like
at
least
internally
at
Google,
we're
kind
of
converging
on
an
opinion
and
once
we've
converged
on
an
opinion,
we'd
like
to
get
the
opinion
of
the
rest
of
the
community,
but
there's
a
couple
of
paths
for
it.
That
could
make
sense.
So
we'll
talk
about
all
of
them.
A
B
Gotcha
and
basically
both
could
could
be
made
the
the
summary
of
where
we
are
now
really.
We
have
a
path
to
get
Envoy
actually
like
dramatically
better,
but
even
if
we
do
that,
it's
not
going
to
be
as
good
as
a
rest
implementation.
So
there's
just
like
a
decision
that
has
to
be
made
about
whether
we
want
to
like
have
the
community
take
on
writing
a
custom
Z
tunnel.
B
We
don't
really
know
you
know,
that's
certainly
up
for
a
debate,
but
maybe
I'll
say
we.
We
think
rust
will
be
significantly
better.
C
We
also
are
able
to
control
it
more
like
so
with
Envoy.
We,
you
know,
because
we're
kind
of
abusing
envoying
a
lot
of
ways
to
implement
a
z
tunnel
for
the
scale
it
wasn't
really
designed
for
this,
and
so
yeah
we'll
be
able
to
you
know
really
just
put
together.
The
features
that
we
need-
and
you
know
for
for
this
for
Teacher
particular
feature
set
that
it's
going
to
do.
A
Yeah
I
was
really
look
forward
to
that
next
week.
All
right,
yeah,
yeah
tall
I,
think
you
have
another
hand.
I
Oh
sorry,
take
them
into
okay.
I
A
Right
is
there
anything
else
we
should
discuss
Jones.
The
initial
task,
certainly
there's
a
lot
of
item
that
needs
help
Andre
go
ahead.
J
Yes,
I
had
a
question
like
last
time.
You
mentioned
that
Solo's
going
to
do
some
stuff
with
cni,
but
the
section
on
the
cni
for
ambient
test
for
beta
is
a
little
bit
small.
So
is
there
some
meat
spanned
on
what
exactly
Solo's
planning
to
do
with
CMI,
and
is
that
part
of
ambient
going
to
be
there
or
is
that
something
separate.
A
Yeah
so
yeah,
so
that
is
one
of
the
items
like
the
long-term
plan
for
redirectioning
Ambience.
So
the
current
way
we're
doing
redirect
from
the
application
container
to
zetano
it's
using
IP
tables
and
routes,
so
we're
actively
looking
yoga
on
our
team
is
looking
at
rewrite
that
using
bbpf
using
socket
map,
potentially
so
yeah
that's.
So
that
is
the
item,
which
is
why
you
are
is
a
sign
for
that.
A
Does
that
answer
your
question?
Andre.
A
Yes,
okay,
cool!
Actually
that
that's
triggered
the
question.
I
have
John.
You
said
that
CI
has
been
found
to
use
the
high
CPU
and
memory
for
small
clusters.
Can
you
provide
some
details
on
that.
B
I
G
Every
time
I'm
running
a
cluster
I,
the
cni
uses
like
a
crazy
amount
of
resources.
Like
it's
100
times
more
than
the
z-tunnel
I
mean
it's
doing,
basically
nothing.
So
we
should
find
out
why
and
then
we
should
fix
it,
but
I
don't
know
why?
Yet.
G
Like
I'm
talking
like
empty
cluster
800
or
you
know,
100
m
CPU
usage
all
the
time.
So
that's
probably
some
Loop
or
something
that
as.
B
A
E
Yeah,
so
the
last
one
is
VMS
I
know
we
haven't
really
put
any
work
into
into
that
part,
but
it
was
something
that
I've
been
looking
at.
So
it's
essentially
the
config.
E
When
you
use
the
ambient
profile
breaks
for
even
Communications
with
site
card
workloads,
we
were
doing
a
little
troubleshooting
today
and
I
thought
I
wrote
this
down,
but
essentially
it's
the
the
traffic
gets
the
East
West
Gateway
for
the
side
card
workloads,
but
then
the
East
West
Gateway
no
longer
seems
to
recognize
how
to
handle
the
connection
and
just
terminates
it.
So
we
we
expected
the
ambient
ones,
not
work,
because
it's
not
configuring
any
endpoints,
but
the
sidecar
workload
should
still
work.
E
Essentially,
yeah
I
don't
have
a
lot
of
experience
when
it
comes
to
Envoy,
so
I'm
not
really
sure
what
to
look
for
here,
but
we
sat
in
a
call
with
Greg,
Lane
and
I
and
another
one
of
our
co-workers
troubleshooting
everything
and
basically
we
could
connect
to
the
sidecar
from
the
East
West
Gateway
using
curl,
and
we
can
see
in
the
logs
that
traffic
goes
from
the
VM
to
the
east
west
Gateway.
But
for
some
reason
in
the
East-West
Gateway,
it's
not
getting
routed
correctly.
A
So
John,
actually,
if
you
can
give
us
some
hints,
that
would
be
helpful,
I
think
maybe
we
can
write
the
code
to
fix
this,
so
the
the
I
guess
the
current
challenge
in
at
least
from
my
perspective,
is
you
know
the
so.
First
of
all,
when
you're
wrong,
let's
say
I
have
sleep
on
my
VM
right
and
I'm,
calling
hello
word
in
my
kubernetes
pod
right
so
on
that
sleeve.
A
So
when
I
run
in
ambient
profile
right,
the
envoy
that's
been
bootstrapped
on
my
VM
I
think
it's
config
as
zetano,
because
today
we
saw
the
the
internal
Upstream
cluster,
which
is
a
very
complicated
debug.
That's
where
we
kind
of
stop
right.
There.
A
Yeah,
that's
the
other
thing
I'm
interested
to
see.
You
know
how
you
know.
How
could
that
happen?
Yeah,
because
in
the
regular
sidecar
case,
oh
just
regular
Envoy
on
the
VM
we
verified
Daniel
correctly.
Remember
we
verified.
There
was
no
reference
to
the
internal
Upstream
cluster
yeah
in
in
your
own
way,
config.
E
Yeah
and
then
the
config
downstairs
saying
some
differences
in
the
tunnel.
I
can't
remember
the
word
now,
basically
we're
seeing
H2
as
one
of
the
tunnel
configs
that
we
weren't
seeing
in
the
non-ambient
side.
So
it's
almost
like
it's
getting
configured
to
connect
to
Z
tunnel
just
because
the
ambient
profile
is
enabled.
G
We're
talking
about
the
the
transport
sockets
in.
G
G
Yeah
I
mean
I,
don't
know,
I'd
have
to
have
to
look
at
it
more
to
debug,
but
you
should
be
able
to
trace
through
the
config
and
see
where
it's
going
wrong
compared
to
you
know
the
non-ambient
profile.
A
Yeah,
so
so
John
I
think
if
you
can
give
us
some
a
hints.
That
would
be
great,
because
Daniel
has
the
environment
with
the
working
well
and
not
working
well
and
we're
kind
of
stuck
on.
You
know
when
the
traffic
gets
to
the
Envoy
internal
medicinal
Upstream
cluster.
You
know
well
with
it.
Oh
you
know.
How
could
this
happen?
First
of
all,
second
of
all,
you
know
what's
to
look
next.
G
G
You
know
finally
going
through
the
East
West
Gateway
and
the
VM
sidecar
and
the
app
sidecar.
That's
quite
a
lot
of
listeners.
So
there's
a
lot
of
places
to
book,
but
you
know
all
the
info's
there
and
they
can
pick
dumps
and
the
logs.
So
foreign.
A
B
I
think
Stephen
would
be
a
good
choice
because
he
he
knows
this
code
pretty
well.
A
Okay,
is
he
around,
or
this
is
the
on
vacation?
Okay,.
G
E
A
G
Yeah
I
yeah,
that's
fine!
The
reason
I
wasn't
a
good
thing
to
do
that
was
I
was
specifically
hoping
to
get
people
other
than
me
and
Steven.
You
know
familiar
with
debugging,
but
that's
that's
fine
as
well.
Yeah.
A
A
Okay,
very
cool
I
think
that's
very
helpful.
So
Daniel,
let's
sing
with
Stephen
offline
for
this
and
I,
saw
there's
some
chats
going
on
with
cost
and
question.
Do
you
want
to
mention
anything.
A
Okay,
I
think
you
comment
about
simplified
sdod
that
would
apply
to
both
rust
and
and
if
we
can
simplify
Envoy
as
well
right.
The
biggest
thing
is
the
weather
is
your
D
supports.
Psycha
today
was
not
psycho
today,.
A
H
I
I'm
personally
interested
in
you
know
having
proxy
less
support
and
other
you
know
kind
of
heading
as
many
implementation
as
possible,
so
I
don't
really
care
all
of
them.
All
of
the
above
is
perfect
for
me,
but
I
think
rust
is
the
best
decision,
for
you
know
fast
development
and
best
quality,
probably.
A
And
we
don't
have
quite
here
so
yeah,
oh
by
the
way
we
have
Steven
now
so
Stephen,
oh
I,
think
was
you
offline
on
this
yeah?
Thank
you.
Does
anyone
have
any
other
topics?
I,
don't
see
any
other
topics
or
anyone
have
any
questions
before
we
close
out.