►
From YouTube: Technical Oversight Committee 2021/03/05
Description
Istio's Technical Oversight Committee for March 5th, 2021.
Topics:
- Extensions & Telemetry 2021 and 1.10 Roadmap presentations
- Roundtable discussion of IstioCon takeaways
A
B
Yeah,
all
right,
you
guys
are
going
to
win
this
game
of
checking
I'll,
take
notes,
I'm
presenting
it
just
taking
my
computer,
okay,
no,
it's
okay,
you're,
all
right
yeah,
but
it's
silly
things.
It's
just
make
sure
everyone's
mentioned
for
a
release
that
we're
shipping.
Actually
that's
kind
of
important,
never
mind.
Yeah
louie
would
be
great
if
you
could
take
notes.
A
C
A
B
B
C
A
C
B
G
B
B
C
G
B
B
H
B
H
C
H
B
G
I
J
Work
that's
been
put
in
for
the
xdot
z
stuff
in
mesh
config,
and
we
know
that
in
the
discussions
for
those
apis
use
of
extension
providers
has
come
up,
and
so
I
put
on
our
roadmap
some
you
know
contributing
to
whatever
the
a
better
extension
providers.
Api
is
here,
because
I
think
there's
there's
a
lot
of
interplay
between
the
wasm
extension
api,
the
telemetry
api
and
the
extensions
provider
api,
whatever
form
that
ends
up
taking.
B
B
J
Yeah,
I
think
it
absolutely
needs
a
bunch
of
scrutiny.
We
just
met
this
morning
to
talk
through
some
of
that.
Daniel
is
doing
a
good
job
of
trying
to
shepherd
it
along,
but
it
is
going
to
need
input
not
just
from
the
telemetry
and
extensions
working
group,
but
but
from
others,
and
you
know
separating
out
we're
trying
to
keep
it
fairly
well.
Scoped
nearly
narrowly
focused
so
that
it
doesn't
fall
into
a
general-purpose
envoy,
filter
api
and
some
of
the
same
traps.
But
thank
you
yeah.
J
J
Well,
I
was
going
to
ping
a
couple
of
you
after
this
meeting
with
the
doc.
Maybe
just
start
getting
some
early
early
looks
at
it.
I
think
daniel
is
going
to
take
a
pass
that
based
on
some
feedback
by
tuesday
and
there's
a
wasm
sig
meeting
on
tuesday
and
hopefully,
after
that,
we
can
share
it
more
broadly.
I
mean
the
big
thing
for
me
is
the
alignment.
J
I
think
the
alignment
between
that
and
some
of
the
work
that's
going
to
go
on
with
extensions
providers,
making
sure
that
we
we
are
aligned
there
and
I
think
the
bigger
decisions
are
going
to
be
the
things
that
josh
was
alluding
to
is
just
like.
How
can
we
keep
it
narrowly
focused
on
dynamic
extensions?
How
do
we
avoid
trying
to
make
it
too
general
and
potentially
have
backwards
compatibility
issues
so,
to
the
extent
that
the
toc
members
are
aware
of
that
and
can
provide
feedback
there?
That
would
be
super
useful.
F
B
F
J
Yeah
and
then
there's
a
special
interest
group
inside
of
the
extensions
islamist
working
group,
that's
meeting
to
discuss
some
of
the
extension
stuff
on
tuesday
afternoons.
So
if
you
don't,
if
you
can't
find
that
information
make
sure
to
reach
out
to
me
and
I'll
make
sure
that
you're
aware
of
that?
Okay.
A
Yes,
so
doug
you
know.
Obviously
this
is
a
pretty
well
both
the
definition
extension
and
any
refactoring
or
improvements
in
the
extension
api
are
pretty
impactful,
so
yeah
I
think
most.
Certainly.
I
would
want
to
make
sure
that
I'm
there
and
providing
timely
feedback
to
you
guys,
so
you
can
make
progress
and
that
we
can
get
get
some
of
this
stuff
done
actually.
So
if
you
can
make
sure
that
toc
members
know
when
and
where
those
discussions
are
happening,
so
we
can
show
up.
J
A
E
J
Thanks
sure
yeah
we
have
to
balance
europe
west
coast
and
other
time
zones
too.
So
yeah
we'll
see
them
to
do
a
better
job,
all
right.
Okay,
the
there's
some
work
left
from
transferring
over
from
mixer-based
telemetry
and
that's
where
the
metrics
expiry
comes
in.
This
is
going
to
take
some
design
work
and
some
some
work
and
probably
an
upstream
envoy.
But
that's
why
that
that
line
is
there
and
then
the
doc?
J
J
So
there
was
some
work
that
was
started
on
that
for
one
nine,
it's
been
way
late
a
little
bit.
I
I
think
we
need
to
to
really
find
some
time
this
year
to
prioritize
updating
that
this
includes
things
like
adding
multi-cluster
dashboards,
adding
dashboards
that
show
vm
features
better
using
canonical
service
to
provide
new
information
that
that
sort
of
thing.
J
E
J
J
The
last
three
items
on
the
roadmap
are
things
that
we
think
would
be
useful.
I
don't
know
how
much
progress
will
be
made
resource
metrics
we've
had
a
lot
of
people
ask
for
metrics
around
configuration,
especially
istio
configuration
objects.
We
had
some
of
that
in
some
levels
at
various
times
in
the
past,
so
bringing
that
out
from
what
its
experimental
state
into
a
more
formal
state
and
making
it
useful
and
documented
is
one
of
the
the
goals
for
2021.
J
You
know
taking
taking
advantage
of
the
ability
to
put
exemplars
in
tracing
just
to
make
telemetry
more
useful
and
to
show
links
between
traces
and
metrics
in
logs.
So
that's
a
feature
that
I'm
sort
of
interested
in,
and
I
mean
it's
a
p3
because
I
don't
know
that
how
much
time
I'll
have
to
devote
to
that.
But
this
has
been
sort
of
a
a
long-standing
goal
to
provide
even
more
detailed
and
useful
telemetry
out
of
istio.
A
So
one
of
the
things
that
we
haven't
really
ever
done
an
sdo
is
pick
a
kind
of
default
solution
like
we
picked
prometheus
for
log
indexing
and
searching
as
an
out
of
the
box
experience,
I'm
not
sure,
but
I
get
the
sense
that
kubernetes
might
be
getting
close
to
nominating
something
in
that
space.
I'm
trying
to
remember
what
the
hell
the
name
of
it
was,
and
it
wasn't
elastic.
E
A
J
A
A
A
E
Follow
a
few
well-known
proxies,
but
louis
before
we
dive
into
the
details
of
this,
we
went
in
a
different
direction
for
other
metric
solutions
right,
so
we
actively
moved
prometheus
and
jaeger
out
of
the
default
chart
and
it's
an
opt-in
by
the
users,
because
we
had
production
issues
where
people
thought
this
was
production
ready,
which
was
all
in
memory.
We.
E
J
E
No,
I
really
like
it.
I
think
this
looks
like
a
reboot
of
a
lot
of
stable
features,
so
I
will
just
reinforce
what
josh
said
and
we
should
make
sure
we
carry
the
users
with
us
and
not
break
them,
but
if
we
implement
most-
even
if
not
all
of
this,
I
think
the
users
are
going
to
have
a
much
better
experience
going
forward.
C
J
Docker,
do
you
want
to
just?
No,
I
we
don't
have
a
separate
dock.
You
can
just
if
you
scroll
to
the
right
there
and
cover
look
at
that.
I
mean
the
big
work
for
110
are
really
the
apis,
but
it's
getting
the
wasm
extension
api
and
the
telemetry
api
in
a
state
that
they
can
be
used
and
start
being
played
with,
so
that
we
can
get
feedback
from
a
release
on
the
concepts
there
and
moving
forward
some
work
on
documenting
and
supporting
the
wasm
experience.
J
F
J
So
that
was
the
that
was
the
that's
that
third
item
rate
limit.
Api
kawat
is
involved
in
both
that
and
the
wasm
extension
api
design,
and
so
there's
some
discussion
this
morning
about
whether
or
not
we
can
make
a
single
api
cover,
both
those
use
cases,
and
so
I
think
that
design
needs
to
settle
out
a
bit
before.
I
know
if
they're,
if
we're
going
to
have
separate
apis.
But
yes,
the
goal
is
to
get
that
off
of
envoy
filter
and
have
a
way
to
represent
that.
C
C
E
E
J
So
well,
there's
going
to
be,
I
mean
so
I
think,
for
the
telemetry
api,
the
design.
You
know
we
it's
starting
we're
going
to
pick
a
subdomain
like
tracing
or
something
and
go
there
and
get
that
sort
of
resolved,
and
you
know-
and
I
think,
there's
some
debate
about
whether
it's
tracing
our
metrics
to
go
forward,
and
so
we
can
choose
to
add
domains
or
hold
back
domains
subdomains
within
telemetry.
As
part
of
that
api.
J
E
E
C
A
A
C
J
A
A
F
F
Love
your
help,
at
least
from
you
know,
validation.
You
know,
try
the
different
scenario,
maybe
for
110,
given
you
know
I'll,
be
missing
most
of
the
d
car
date
of
110
because
of
solo
account,
but
I
should
be
able
to
provide
perspective,
maybe
help
you
try.
Different
steps
to
validate.
You
know
make
improvements.
E
H
Yeah,
so
we
don't
have
anything
specifically
for
110,
but
for
the
year
we
really
want
to
start
focusing
on
some
of
the
not
reactive,
whatever
the
opposite
of
reactive
is
proactive.
Thank
you.
So
things
like
threat
modeling.
We
don't
currently
have
an
expert
in
that,
but
that
would
be
really
useful
for
istio.
That
would
that's
something
that
francois
brought
up
before
he
left
dependency
scanning
code
scanning
and
automated
testing
of
cve
fixes.
H
I
C
Yeah,
that's
fine!
I
just
want
to
understand
sort
of
what
the
priorities
were
and
what
whatever,
because
we
have
one
p0
and
three
p
ones
here-
should
be
good
to
know
kind
of
where
you're
gonna
be
tackling
them.
What
sounds
like
reactive,
all,
the
reactive
stuff,
product
security
always
does,
and
then
police
dependency
scanning,
okay,.
E
C
G
Yeah
sure
yeah,
so
for
the
pr,
so
I'm
just
looking
for
feedback,
there's
been
some
discussion
over
the
last
couple
months
about
you
know,
feature
status
and
on
seo
io
or
sorry
on
the
community
markdown
or
whatever
we
have
a
set
of
you
know.
This
is
what
it
means
to
be
stable
and
experimental,
and
you
know
all
of
these
other
feature
stages.
You
know
users
have
been
asking
us,
you
know.
Can
we
get
more
clarification
on
the
istio
io
website
for
this?
G
So
this
is
just
trying
to
reconcile
between
the
community
one
and
what
we
display
to
users
so
that
there's
kind
of
just
one
source
of
truth.
So
I'm
just
looking
for
feedback
on
this
there's.
Obviously,
some
wording
where,
like
they
were
talking
about
monthly,
cadence
and
whatnot
on
the
community
page.
So
I'm
just
looking
for
feedback
on
that
and
then
I
guess
some
guidance
and
we
can
talk
about
that
later
after
I've
tried
to
incorporate
some
more
feedback.
E
G
Okay,
okay!
Well,
then,
that
makes
it
easier
great,
and
then
we
had
talked
about
this
a
couple
of
weeks
ago
and
I
just
wanted
to
follow
up
on
creating
a
google
group.
I
know
louis
said
he
was.
He
was
doing
some
experimenting
with
google
groups
for
release
announcements
or
things
of
that
nature.
So
I
just
want
status
and
if
we
can
get
that
approved
and
kind
of
get
that
published
for
people
to
have
just
you
know
an
email
mailing
list
for
first
year,
so.
A
The
I
I
was
doing
research
on
document
and
domain
ownership
stuff
within
the
kind
of
the
google
ecosystem.
I
think
craig
picked
up
some
of
that.
So
then
you
were
also
looking
at,
but
I
don't
think
it's
the
same
thing
as
the
google
group
stuff
like
we
can
just
create
a
google
group
at
publicgooglegroups.com.
G
I
think
it's
just
like
other
communities
like
kubernetes
and
goling,
for
instance,
already
have
this
and
I
think
it's
it's
I'm
not
sure
how
many
people
actively
read,
discuss
daily
versus
if
they
were
to
actually
get
an
email.
Hey.
You
know,
here's
a
cve
announcement.
You
know
please
go
update
your
istio.
G
A
A
F
C
A
F
Yes,
I
spent
tons
of
time
for
that
so
apologize.
If
I
was
ignoring
your
review
requests,
it
was
very
time
consuming
to
produce
issue
car
along
with
many
of
other
folks.
So
I
think
louis
before
you
join,
I
apologize
to
folks
that
I
got
the
data
wrong.
So
on
the
first
day
I
saw
3700
people
on
the
day
one
keynote
and
then
on
the
night
of
the
china
keynote
I
saw
3
900
people,
so
it
turned
out.
I
learned
on
friday.
F
Those
are
the
registration
number
and
the
crowd
costs.
Somehow,
when
you
log
in
you
saw
the
registration
number,
but
you
thought
it
was
the
live
attendance
number.
So
that
was
not
the
live
attendance
number,
which
explains
why
throughout
the
week
you
actually,
I
think
we
reach
4
000,
either
on
thursday
morning
or
friday
morning.
So
that's
explained
the
increase
of
the
number,
but
the
good
news
is
on
friday
because
I
had
to
give
a
closing
keynote.
So
I
got
all
the
data
we
did
receive.
F
Really
close,
I
I
don't
recall
if
it's
over
or
slightly
under
3
000
live
attendees,
so
this
doesn't
mean
all
the
attendees
show
up
for
every
single
session.
It
means
you
know.
Maybe
this
person
showed
up
on
this
session.
The
other
person
showed
up
on
a
different
session,
but
they
are
counted
as
one
attendee,
a
unique
attendee.
F
On
the
first
day
of
the
keynote
we
had
the
most
people,
which
is
around
a
thousand
live
attendee,
attended
the
first
day
keynote
we
had
about
1500
attendees
watch
the
replay
on
crowdcast
just
before
friday
ends,
but
it's
unclear.
If
those
people
watch
the
replay
actually
are
also
the
same.
People
attended
life
because
we
don't
have
that
stats,
but
we
do
know
those
1500
watch.
The
replay
are
part
of
the
registration
number
because
you
have
to
register
to
be
able
to
see
the
live
or
the
replay.
F
Yeah
still
very
impressive,
I
should
say
even
though
I
I
screw
up
the
number
on
twitter,
but
I
was
super
super
excited
about
the
attendance
number
and
I
think
I
I
actually
went
to
a
lot
of
talk
just
to
see
what
people
message
like
chat
history
for
the
talk.
A
lot
of
them
are
super
positive
people
are
excited.
F
You
know
there
are
really
good
talks
like
ebay
and
airbnb.
They
really
drill
down
to
details
of.
As
to
you
know,
this
is
their
architecture.
This
is
how
they
move
to
istio.
This
is
their
future
plan,
so
really
impressive.
I
think
throughout
the
conference
most
of
the
talks
are
got
a
lot
of
thumbs
up
at
the
end,
I
would
say
the
organizer
committee
did
a
really
good
job,
selecting
these
talks,
so
the
speakers
are
very
well
prepared
and
there
was
actually
a
lot
of
live
demo
on
different
talks
too.
So
I
was
super
impressed.
A
E
F
F
I
think
people
said
a
lot
of
good
things
about
helm3.
I
think
people
clearly
appreciated
that
the
test,
automation,
icon
we
added
on
istio.io.
I
think
that
was
received
really
well.
People
certainly
love
that
feature.
You
know
they
know
that
you
know
when
they
go
through
the
steps.
You
know
it
can
actually
work.
F
So
that
was
a
huge
thing
and
I
also
took
notes
on
t-mobile,
I
think
was
they
were
running
100
meshes
so
that
come
up
in
the
chat
history.
People
were
impressed.
They
were
running
that
much
number
of
meshes.
I
think
these
are
the
key
ones.
Oh
and
then
the
other
thing
was,
interestingly,
both
salesforce
and,
I
think,
was
ebay.
They
both
were
building
abstraction
on
top
of
the
istio,
so
that
was
interesting,
salesforce
build
it
using
helm
and
ebay
was
building
using
their
own
customer
resource.
E
Yeah,
just
a
few
more
things
to
add
len
on
yours,
I
think
one
thing
that
jumped
out
to
me
was
multi-directio
was
quite
hard
and
most
of
the
users
had
issues
with
it.
That's
something
we
should
think
about
this
militancy,
I'm
talking
about
is
making
sure
a
developer
from
one
team
or
an
operator
from
one
team
cannot
create
resources
which
can
conflict
or
affect
the
other
team's
deployment.
E
This
was
mentioned
few
times.
Secondly,
I
echo
what
lynn
said
around
abstractions
many
people
stress
that
the
stability
was
not
up
to
the
mark
as
they
would
have
lied
and
they
had
to
create
a
lot
of
abstraction,
around
testing
and
automation,
to
make
sure
the
new
releases
don't
break.
So
in
that
light,
I
think
our
roadmap
projection
for
making
histo
boarding
went
well.
So
hopefully
we
deliver
on
that
and
then
third
we've
got
a
fair
amount
of
press
pickup
based
on
the
quality
of
talks,
the
production
use
cases.