►
From YouTube: April 17, 2019, Istio Environments WG
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
Just
one
moment
this
is,
we
have
a
clear
use
case
to
support
user
customizable
of
multiple
base
images,
because
it
is
critical
for
many
customers
who
standardize
on
a
particular
image.
So
if
you
have
an
organization
list
is
standardizing
and
16
want
to
first
want
to
be
able
to
provide
proxies
that
use
exactly
the
image
that
is
including
ability
for
people.
C
A
Build
their
images
with
whatever
is
certified
as
an
environment.
That's
one
reason
that
we
must
support
arbitrary
base.
Images
and
second
reason
is
that
for
performance
point
of
view,
startup
time
and
a
lot
of
other
things
improve
if
the
application
is
using
the
same
base
images
as
a
proxy
because
loading
an
application
that
is
based
on
one
two
and.
A
Is
based
in
guntur
distress
is,
you
know,
slow
and
requires
loading
downloading
to
images,
and
it's
not
ideal
right.
So
we
want
to
support
the
series
of
proxy
images
that
you
can
customize
and
you
can
say,
I
want
to
use
proxy,
be
to
slash
they
be
on
or
they
be
on
proxy
be
to-
and
you
know
basically,
most
common
distributions
should
be
supported
and
distributed
out
of
the
release,
including
destroyers,
of
course,
and
we
should
provide
some
way
for
people
to
build
their
own
customized
image
physically.
A
B
I
think
what
we
discussed
last
week
was
totally
consistent
with
that
guy
pretty
much
what
we
talked
about.
We
didn't
talk
about
much
about
the
other
distros
or
the
base
customization,
but
we
did
talk
about
making
sure
dis.
Realist
was
isolated
from
the
existing
base.
Images,
so
it'd
be
like
a
different
type
of
base
image
which,
with
rotisseries.
A
People
know
that
it's
already
possible,
so
anyone
can
build
an
immediate
proxy
with
arbitrary
base
image.
One
story:
by
simply
doing
a
local,
you
know
make
creating
whatever
image
they
want
as
a
base.
Labeling
us
labeling
it
locally
with
east
your
nightly
base
or
whatever
it's
a
label
and
then
doing
make
make
proxy,
because
for
the
way
it
is.
A
B
A
A
Up
to
handle
multiple
base
images,
multiple
proxy
images,
it's
around
the
patient's
that
is
present
already,
you
can
use
it
and
you
can
specify
which
proxy
image
to
use
for
each
particular.
So
suppose
you
are
familiar
with
the
stability
cluster,
it
has
a
bunch
of
4000
and
other
applications
and
some
for
your
images,
each
each
photo
step.
One
has
a
photo
client
and
there
are
some
one
rotation
to
control
each
photo
to
use
the
different
base.
Image
is
different
whatever
to
not
use.
A
B
A
It
will
be
much
better
to
do
it
this
way
with
with
the
stability,
because,
first
of
all,
the
stability
is
also
ensuring
that
there
are
no
leaks
crashes,
whatever
instability
and
you
can
is
far
more
efficient
because
you,
you
just
write
1
ml
file
for
each
proxy
professionals
that
you
want
to
test,
you
put
it
there
and
it's
automatically
tested,
and
you
don't
have
to
do
pretty
much
missing
us
so
here
in
combination
as
long
as
you
have
something
that
has
a
client
in
a
server
and
48.
So
this
is
something
easy
to
to
Todd.
A
E
B
A
You
can
compare
because
then,
by
running
the
side
by
side
and
having
six
or
none
of
them.
If,
for
example,
distress
happened
to
for
whatever
reason
to
be
20%
slower
than
we
do,
you
will
see
it
on
the
graphs,
so
you
can
see
side-by-side
how
they
compare,
because
there
may
be
a
regression
in
whatever
base.
A
B
A
Use
to
verify
that
you
don't
have
memory
leaks,
we
don't
have
performance
aggressions,
we
don't
have
stuff
crashing
after
two
or
three
hours,
so
it's
a
test
of
transport,
24
hours
or
even
longer
for
weekly
and
its
intended
to
cover
as
many
scenarios
as
possible
and
the
tests
are
based
on
metric
space.
There
is
some
some
scripts
that
is
looking
at
inanimate
things
today.
If
I
said,
oh
five,
what
please
there
know
whatever
okay.
B
A
B
A
A
A
Now
that
would
allow
you
to
to
to
run
the
Installer
in
the
test
locally,
because
it's
starting
kind,
the
reticent
talker,
so
you
can
test
locally
and
do
whatever
you
need
probably
soon
would
have,
will
also
runs
a
version
screen
and
part
of
the
stability
cluster
running
on
on
kind
as
well
for
people
who
want
to
develop
and
to
make
quick
tests
and
yeah.
We
need
more
testing,
more
people
to
look
at
it
and
to
try
different
combinations
because
being
decomposed
and
modular
there
many
permutations
and
combinations
that
need
to
be
to
be
tried.
G
A
However,
we
want
to
support
multiple
CIC.
This
distance
is
particularly
for
installer,
because
one
of
the
goal
is
to
allow
integration
between
you
know
different
enterprises
and
organization
and
vendors
using
different
city
systems
and
the
installer.
So
if
some
organization
is
a
concourse
internally,
the
cleaners,
the
staff
that
recipe
did
decisions
using
whatever
they
should
be
able
to
integrate,
but
there'll,
be
one
official
history
or
CI
CD
system,
which
is
TBD
bill,
Couture
or
crown
and
multiple
other
systems.
That
could
be
supported
and
people
are
welcome
to
contribute
other
integrations
for
the
city
systems.
A
A
Series,
II
60,
set
of
tests
sed
system
is
ready
exactly
the
same
things
that
happens
on
into
my
test,
so
the
CCD
the
main
points
that
was
made
him
in.
So
let
me
let
me
rephrase
there
are
two
ways
to
taste
the
installer
one
is
what
the
CSU
system
is
using
and
that
we
make
test,
which
was
start
kind
or
mean
cube,
basically
mini
cube
root.
A
That's
one
method,
the
other
is,
if
you
specify
some
parameters
in
to
run
against
than
your
cluster
and
exactly
our
cluster
can
be
a
performance
cluster,
which
is
you
know,
a
large
cluster
or
it
can
be
a
small,
but
in
this
class
the
routine
anything
in
the
past
Aretha
anyway.
So
the
idea
is
that
make
tests
would
be
the
single
thing
that
people
need
to
worry
about
and
based
on
parameters.
It
will
be
pointed
to
different
environments,
and
the
default
would
be
confidence,
hey.
A
C
A
A
B
I
didn't
want
to
talk
about
the
deprecation
of
the
Installer
I.
Think
that
makes
a
lot
of
sense.
An
existing
Installer,
assuming
we
can
get
the
new
installer
working.
It
sounds
like
it'll,
be
ready
to
go.
I
guess
the
question
I
have
is:
how
are
we
gonna
communicate
that
we
can
communicate
that
to
this
group?
But
there's
like
ten
people
on
the
call
so.
A
A
A
A
B
I
I
H
K
H
D
A
A
B
A
F
Yeah
well
so
operator
we're
intending
how
the
gates
to
be
based
on
the
Installer
charts.
They're
nice,
not
invisible,
yeah,
pretend
or
are
so
so
you
know
by
by
sort
of
transitive
relationship.
It
should
be.
It
should
be
other
parts
as
well
in
the
sense
that
you
can
turn
on
and
off
components
as
as
you
choose,
but.
A
One
thing
that
we've
been
discussing
and
trying
to
do
is
to
have
a
sort
of
porous
man
operator,
meaning
that
the
new
installation
install
script.
In
fact,
we
use
for
test,
for
example,
in
certain
historical
satori
and
part
of
the
tests.
We
basically
the
docker
image
that
include
salsa
or
the
chart
is
in
store,
so
we
are
trying
to
have
a
small
port.
Basically,
that
is
running
a
shell
loop,
which
is
reading.
What
is
the
Tamil
check
if
it
changed
apply
in
store
that.
A
That's
in
fact
based
on
organizational
units
or
whatever
we
need
for
her.
So
so
it
is
basically
a
very
simple
version,
which
is
intended
also
to
to
upgrade
to
make
sure
that
we
can
upgrade
and
and
have
a
simpler
way
to
integrate
the
performance
faster
with
with
the
rest
and
to
simplify.
The
testing
gives
us
optimizations
for
kind.
If
we
use
this
approach,
but
it
will
also
be
some
things
that
would
be
kind
of
an
early
access
operators
that
is
kind
of
tweaking
and
whenever
you
change
one
is
that
when
something
happens,
okay.
B
L
Yes,
yeah
I,
just
saying
that
we've
been
working
with
Martin
and
the
the
folks
at
ones
iCloud
talking
about
this
over
the
last
few
weeks
and
well
then
read
out.
We
already
have
an
operator
up
and
running
that
does
uses
their
hound
charts
to
as
a
template
mechanism,
and
we
put
some
resource
that
we
have
essentially
maps
on
to
what
you
would
traditionally
put
in
values,
yeah
Mille
and
then
run
through
the
templates
and
then
deploys
the
controller
in
that
way.
L
B
The
Catalina's
has
been
on
pto
a
couple
of
weeks.
He's
missed
out,
I
have
unfortunately,
yes,
so
Kevin
we're
doing
new
charts.
We've
been
working
that
for
what
cosmos
who's
been
working
out
for
a
couple
of
months
actually
and
there's
an
SEO
installer
directory,
which
contains
new
charts,
I,
don't
know
what
the
impact
will
be
that
significant,
but
I,
don't
know
if
it
won't
be
I,
don't
know,
I
haven't
evaluated.
Give
you
mean
new.
A
L
A
Any
large
parties
is
intended
to
allow
operator
to
kind
of
do
a
smart
migration
when
you
upgrade
from
one
one,
two
one
two
or
other
upgrades
or
you
change
any
component.
You
should
be
able
to
do
it
smoothly.
What
you
have
you
know
the
new
version
running
your
shift.
Traffic
operator
can
do
smart
things.
So
if
it
you
look
for
five,
what
reasons
are
all
bad
I
mean
what
key
needed
is
doing
effectively.
Yeah.
A
F
You
know
a
restructuring
and
a
cleanup
of
values
or
llamó
and
during
the
review
was
quite
a
bit
of
back
and
forth
about.
Does
it
make
sense
to
continue
to
support
this
all
these
parameters
as
an
installation
API,
because
it's
quite
likely
that
that
many
of
those
settings
will
move
into
CR
these
in
the
future?
F
So
all
the
config
maps
that
are
currently
sort
of
installed
as
part
of
home
charts
are
likely
to
be
refactored
as
CR
these
and
more
concerning
directly
or
customized,
because
yeah
so
so,
customized
I,
guess
that's
that's
something
else,
but
the
point
being
that
we
don't
necessarily
want
to
introduce
a
new
large
API.
Your
specification.
F
Deprecate
it,
you
know
at
one
quarter
later
so
so
I
out
of
after
all,
that
I
guess
where
we're,
starting
with
this,
to
mainly
use
values.
Looking
angle
will
mostly
just
continue
to
use
values
of
gamma.
There
will
be
minimal
API
that
we
do
need
for
the
operator
and
some
things
that
are
not
likely
to
end
up
a
CR
knees.
B
A
Look
at
what
we
have
in
part
is
like
UML
yeah.
You
can
speak
it
in
several
buckets.
One
pocket
is
stuff
that
customized
us
like
overriding.
What
kind
of
things
in
you
know,
resource
usage
note
under
stuff
that
it's
already
defined
by
their
name
is
API,
because
we
are
just
kind
of
putting
of
our
users
that
is
overwriting
a
protein
and
all
those
probably
move
to
customize
I'm,
not
sure.
Well,
right.
Sorry,.
A
A
And
all
those
things
are
either
already
covered
by
straight
guys,
and
why
is
the
camera
is
just
an
imported
way
to
replace
or
to
create
a
CR,
or
they
are
things
that
belong
to
network
like
default
value
for
retrying,
whatever
you're
using?
So
all
the
settings
should
really
be
part
of
the
networking,
API
or
part
of
the
mixer,
API
or
part
of
whatever.
Okay.
A
Some
of
them
are
already
part
of
the
Louvre's
api's
and
left
there,
because
we
history
or
whatever,
and
some
of
them
should
be
because
we
reality
is
not
an
installer
issue.
It's
an
operation.
If
you
enter
an
initial,
you
may
change
it
at
any
time
and
you
may
change
it
and
you
know
it's
not
really
something
you
only
worry
about.
That
is
for
that.
That's.
A
Any
general
installation
should
have
nothing
specific
to
installation
in
mind,
because
that's
what
I'm
saying
we
are
arguing,
because
in
practice
I
cannot
I
still
don't
know
anything
that
would
you
want.
You
want
to
set
once
again
stole
and
never
change
again,
so
practically
everything
and
then
upgrade
is
I.
A
F
Sure
that
they
help,
but
for
things
like
like,
for
example,
resources
right
could
be
Navy's
resources,
so
one
way
you
could
do
that
is
through
customized.
Is
that,
but
yes,
you
know
customized
may
not
be
the
most
convenient
way
to
expose
that
to
to
users,
particularly
if
you
have
defaults
like
there
are
system-wide.
So
then,
then
you
have
to
customize
every
single
right,
so
I
mean
let's
maybe
not
revoke
them
anyway,
but
I.
Think.
A
D
B
F
F
F
A
One
requirement
that
is
coming
and
and
and
again
a
lot
of
customers
I
mean
is
that
they
want
to
be
able
to
operate
from
1
0
to
1
1
and
1
1
1
twin
going
forward,
so
we
cannot
break
them.
So
whatever
operator
is
doing
it
before
it
is
released
and
and
officially
better
quality,
it
must
have
a
way
to
take
a
value.
So
can
you
convert
it
whatever
and
work?
Ok,.
B
F
A
A
B
You
there
cause
I
think
we
need
to
slow
down
the
change.
Martin
I
get
the
impression
from
you
that
the
TLC
had
wanted
to
basically
be
rid
of
values
at
yeah
mall
permanently
and
want
to,
and
then
translate
that
into
some
other
form
of
API.
That's
built
into
is
to
you
already
that
misunderstand
that,
because
I
think
that's
a
little
aggressive,
well
I.
F
A
Your
customers,
I,
guess
the
two
pockets
and
some
pocket
I
don't
have
any
and
I
can
get
which
would
be
operator
specifics.
We
fight
if
we
identify
things
that
are
specific
to
operator
and
will
never
be
something
that
belongs
to
one
of
the
existing
working
groups,
or
it
is
that
for
okay,
but
so
far
some
packet
is
sent
upon.
You,
I
haven't
seen
anything
that
is
really
really
I
can
say
hey.
This
cannot
be
it's
not
even
any
specific
or
it's
doing
place
this
song.
A
B
K
A
Not
entirely
I
mean
people
will
keep
upgrading
with
whatever
they
are
using
today,
because
we
are
not
going
to
require
the
operator.
In
any
case,
we
expect
people
to
be
able
to
use
whatever
CI
CD
systems
they
are
using
and
whatever
process
they
have
in
place,
including
have
given,
because
there
are
organizations
that
ready
to
deploy
bit
until
I.
So
we
must
put
up
great
parts
for
everyone
and
we
cannot
force
people
to.
It
could
be.
A
A
President
people,
who,
for
example,
hope
if
I
had
a
present
in
a
previous
environment
and
they
explained
how
they
you
know,
check-in
or
the
Yammer
files
into
their
CD
system,
used
I,
don't
know
what
systems
they
are
using
to
deploy
to
the
stage
in
production
and
whatever.
So
they
have
an
infrastructure
built
to
deploy
yeast
your
changes
and
review
them,
and
we
cannot
tell
them
that
hey
stop
using
your
processes
for
this
job
right.
C
C
F
A
F
D
A
You
know
it's
not
a
properly
I
mean
he's
doing
general
is
recommending
people
that,
when
they,
when
they
deploy
your
applications,
they
create
v2,
they
do
use
routing
rule.
That
means
that's.
The
whole
point
of
is
to
allow
people
to
smoothly
and
safely
operate
their
own
version
of
fabrication,
and
any
team
is
being
kind
of
the
same
thing.
It's
crazy
that
our
own
installer
is
is
the
same
thing.
I.
B
A
A
E
C
A
E
C
E
A
Let
me
we
are
not
saying
that
we
are
going
to
have
a
duplicate
environment,
completely
we're
seeing
that
each
component
of
SEO.
Let's
say
you
want
to
upgrade
easier
to
next
version.
You
away
each
component
independently
and
different,
with
a
rolling
upgrade
that
is
happening
today.
So
today,
when
you
upgrade
in
place
effectively
means
that
you
start
you
have
the
old
old
running
version
1-0,
you
start
a
new
port
running
version,
1,
1
and
slowly
you,
you
kind
of
kills.
All
the
old
wooden
is
ready.
So
you
don't
have
traffic
loss
so.
C
A
No,
but
again
we
are
not.
We
are
trying
to
have
this
process
to
be
controllable
by
user,
so,
instead
of
doing
a
rolling
upgrade
where
well
kubernetes
take
care
of
a
plume
of
moving
stuff,
any
sub
ratings,
a
pilot,
you
you
you
you
you
use
operator
or
whatever,
and
you
have
the
pennies.
As
if
I
wrote
you,
you
create
a
new
new
instance
of
pilot,
a
new
version,
and
then
you
have
control
over
when
applications
shift
result
is
any
business
of
pilot.
A
E
A
E
A
And
the
in
place
parties
in
order
to
prevent
to
to
allow
this
kind
of
configuration,
because
pilot
depends
on
kind
of
CRS
and,
and
you
know,
study
configuration
because
that's
that's!
That's
one
problem.
If
you
also
update
values
document
a
configuration,
should
you
say
given
a
call
out
effectively,
so
if
you
want
to
change
MPLS
or
if
you
want
to
change
any
setting,
it's
more
or
less
equivalent
to
the
rolling
upgrade
or
final,
because
you
are
changing
the
behavior
and
and
it
may
break
or
not
applications.
A
B
Guess
back
on
the
operator,
it
sounds
like
we
need
to
do
an
API
review
in
this
workgroup,
not
in
the
TSP
of
what's
happening
there
right,
because
I
haven't
haven't
seen
that
the
API
review
at
all
I,
don't
even
know
what
that
looks
like
I.
Don't
think
people
in
this
worker
do
know
what
they
dad
looks
like
as
well,
so.
F
A
F
G
B
F
Because
it's
gonna
come
out
of
a
separate
repo
I
think
we
can.
We
can
take
a
look
at
how
mature
it
is
and
how
stable
it
is
and
label
it
appropriately
as
we
as
we
approach
one
two,
but
you
know
I
I,
think
it's
realistic
to
have
some
high
I
mean
I
I,
don't
know
how
he
wants
to
actually
communicate
the
the
stability
and
sort
of
release
quality
level
of
the
operator
at
one
two,
but
we
will
have
something
that
will
work
with
1,
but
you.
F
B
A
Want
do
it,
we
should
start
the
beginning
son
of
it's
like
you're
kind
of
customization.
Oh,
don't
make
any
sense
in
production
that
ceremony
for
testing
and
for
for
the
CD
system
and
whatever,
and
since
the
union
store
had
already
Jesus
customized
and
can
can
apply
customized
template.
We
have
a
bit
more
flexibility.
People
hey,
if
you
want,
is
advance
taking
it's
customized
yeah.
B
B
B
A
A
A
B
B
So
you
mind
if
I
take
the
last
five
minutes
or
okay,
so
Kevin
to
answer
your
question,
the
project
was
called
sundar,
geisha
I
haven't
started
the
project
externally
and
I'll.
Tell
you
why,
during
the
original
I
care,
gotta
kick
excuse
me
during
the
original
kickoff
meeting,
I
was
challenged
to
learn
MCP
first,
so
I've
been
trying
to
take
a
look
at
how
MCP
works
to
see.
B
If
we
can
use
MCP
I
think
that's
actually
a
really
good
idea
going
off
and
writing
our
own
stuff
without
understanding
how
the
current
mCP
implementation
operates
is
very
important.
So
if
we
can
learn
that
I
think
we
can
solve
replicated
gateways
service
centuries,
which
is
pretty
straightforward,
I'd
like
to
keep
that
work,
there
are
just
two
parts:
there's
the
figuring
out
how
it
works
and
then
actually
implementing
that
those
are
two
separate
things.
I
haven't
started
that
mainly
because
I've
been
slammed.
B
L
B
I
know
there's
a
lot
of
interest,
and
you
know
the
the
first
part
is
really
about
training,
so
we
can
learn
how
this
stuff
works
and
the
second
part
is
really
about
doing
it
somewhere
else,
like
mostly
in
pilot
emptying
the
implementation,
mostly
by
weight
around
mCP
D.
So
the
first
part
the
training
is
challenging,
but
I
think
that's
worthwhile
for
our
community
to
go
through
some
training
there.
So
yeah.
B
Yeah
I'm
a
bit
overloading
myself
but
I
hope
to
get
to
it
kind
of
in
May
and
I
I
guess
I
was
premature.
I
thought
we'd
have
more
a
little
sorry
less
work
and
whatnot
through
them.
We
we
actually,
if
you
look
at
the
github
link,
maybe
I'd
like
you
myself
to
free.
If
you
look
at
the
github
link,
there's
a
bolo
to
work
there
from
one
too,
and
it's
like
six
weeks
at
my
Wix
has
something
like
that.
So
I
mean
we
and
I
don't
know
how
much
of
that
were
actually
gonna
complete,
don't.
A
L
I
was
under
the
impression
this
was
kind
of
before
I
went
away
on
kind
of
the
the
break
that
I've
just
had
that
the
one
to
work
was
primarily
going
to
be
around
stabilizing
the
test
framework
and
the
infrastructure
there
and
trying
to
automate
the
pull
request.
Process
and
removers
met
as
much
of
the
manual
steps
as
possible.
Is
that
not
still
of
it?.