►
From YouTube: CNCF SIG App Delivery Meeting 2019-11-06
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
CNCF SIG App Delivery Meeting 2019-11-06
A
A
A
A
C
D
A
A
B
B
B
Right
so
there
are
some
items
that
didn't
hit
the
agenda
and
so
I'm
just
going
to
rattle
them
off
and
then
we'll
tackle
them
from
top
to
bottom
and
get
them
in
the
document.
So
the
first
item
is
around
the
definition
of
operator.
I.
Think
the
history
here
is,
we
know,
depending
on
who
you
talk
to
an
operator
is
different,
and
we
want
to
make
sure
that
there's
a
canonical
good.
B
That
there
is
a
canonical
definition
of
what
an
operator
is
and
have
it
written
down,
and
the
next
item
that
al
was
proposed
is
is
on
the
operator
framework
versus
the
hub.
There
I
believe
they've
been
submitted
and
we
need
to
figure
out
if
they
should
be
the
same,
submission
or
separate
submissions.
A
B
C
A
A
I'm
back
yes,
great
I
know,
what's
going
on
with
you
so
yeah.
C
A
Know-
and
it
could
be
that
it
is
a
mmmm
two
main
items
that
we
need
to
discuss,
because
it
our
there's
different
implications
right,
have
we
done
before
obviously
was
helm
and
the
truck
museum,
but
I
think
it's
better
to
understand
what
the
future
structure
and
the
management
of
the
project
should
be
and
fully
read.
It
should
be
two
separate
submissions.
G
H
Yeah
I
think
that's
correct
and
also
one
thing
that
we
I
think
there
are
some
discussion
or
on
it
already.
Actually,
people
are
using
operator
for
work
mostly
for
developing
operators,
so
the
prediction
of
a
pretty
hard
and
opressor
lifecycle
management
knows
things
like
that
are
not
quite
in
the
scope
of
many
people
are
using
operators
and
I.
Think
that
is
the
discussions
and
great
most
skill
about.
G
Well,
some
of
this,
if
I
understand
it
right
is
there's
another
submission
for
Kudo,
which
is
another
operator
based
project,
and
it
takes
a
different
route
than
the
operator
hub.
It
is
also
I,
don't
know
the
ins
and
outs
of
it.
I
know
there
are
other
operators
out
there,
such
as
the
one
that
the
Microsoft
folks
wrote
for
it's
called
rudder
and
that's
another
example
of
an
operator.
It's
not
even
written
and
go,
and
it
doesn't
use
things
like
chargeback.
G
Other
things
like
that
that
are
included
in
the
operator,
lifecycle,
manager,
right
and
operator
lifecycle
manager
operator
have
an
operator
framework
are
all
bundled
together.
So
what
does
all
of
this
mean
when
there's
these
other
things
outside
of
the
one
submission
like?
How
does
it
all
play
together?
And
what
does
that
mean?.
A
Yeah
and
I
think
that's
what
everybody's
a
bit
unclear
about
right
now,
because
it's
like
not
just
the
project,
it's
a
couple
of
bits
and
pieces
and
the
question
should
really
go
back
to
red
hat
whether
they
want
to
potentially
separate
the
three,
because
it's
still
three
different
discussions.
Kind
of
hi.
I
So
can
you
ask
the
question
again
about
aqua
framework
versus
operator
hub.
A
A
Repeat
we
had
yesterday
a
presentation
also
to
that
you
will
see,
and
a
couple
of
questions
came
up
with
guarding
the
operators
framework
submission,
first
of
all
operator
and
operator,
lifecycle
manager
that
they
are
bundled
together
and
obviously
serving
two
different
purposes.
Plus
also
operate
their
hub
and
how
this
should
like
all
fit
together,
because
it's
different
to
have
the
project
they
over
the
framework
project
inside
CNCs
versus
two
more
less
them
discuss
how
we
can
manage
and
run
the
entire
operator
hub
and
plus.
I
I
Efforts
are
getting
to
be
pretty
automated
and
what
we're
looking
for
in
donating
that
piece-
and
this
is
Diane's
point
of
view-
is
to
in
the
review
process
more
people
involved
in
in
that.
So
by
donating
it
it
becomes
not
such
a
Red
Hat
centric
process
or
resource
sub
process,
but
we
would
still
put
resources
on
it
to
obviously
to
help
with
the
CIO
CD
process
of
getting
things
into
operator.
I
I
know,
and
it
would
be
nice,
in
my
opinion,
to
have
a
non
vendor
space
or
non
branded
a
CNC
F
branded
space
for
all
of
these
operators
and
services
that
had
players.
We
have
had
interest
from
AWS
and
Google
in
resourcing
some
of
the
the
review
and
upload
process,
but
it's
pretty
highly
automated
at
this
point
so
operator
hub
itself
is,
is
almost
self-sustaining,
but
you
can
never
say
that
a
hundred
percent
and
the
other
projects
it's
it's
kind
of
being
presented.
I
think
as
a
bundle,
because
they
all
sort
of
feed
each
other.
G
I
might
be
able
that
this
is
Matt.
Farina
I
might
be
able
to
fill
it
in
this
started
in
the
TOC
meeting
and
the
context
is
there
are
operators
that
do
things
very
differently
than
the
operator
framework,
one
of
the
other
ones
that's
been
submitted
to
the
CNC
after
it
is
no
there's,
also
other
ones
that
have
been
written,
and
so
how
do
they
play
out?
Because
the
operator
hub
plays
very
much
into
the
way
operator,
lifecycle
manager
works
and
the
operator
framework
and
these
other
operators?
They
do
things
very
differently
from
that.
G
So
what
does
it
mean
for
them?
And
some
of
these
are
written
like
Kudo.
Some
are
written
in
entirely
other
languages,
they're
not
even
go
based
and
they're
just
different,
and
so
what
does
that
mean?
Because
they
don't
fit
the
current
operator
hub
model
and
so
do
they?
Should
they
fit
into
that
flow?
You
know
a
lot
of
operator
lifecycle
manager
is
very
Service
Catalog
like
and
there's
a
lot
of
operators
out
there
that
don't
fit
that
mold.
So
how
does
all
of
this
work
together?
And
what
does
it
mean
should
operator
hub?
G
I
I
actually
think
those
are
awesome
questions
because
we
I
mean
I've
thought
about
them
as
sort
of
the
community
person
behind
it,
and
we've
done
some
outreach
with
the
helm
folks
and
with
coop
builder,
and
had
conversations
about
that.
The
goal
of
the
lifecycle
management
is
really
to
get
more
mature
operators
out
there
in
the
wild,
and
so
we
can
support
day
to
ops
kind
of
things,
but
by
moving
it
into
the
CNCs.
The
governance
of
what
goes
into
operator
hub
dot
IO
becomes
an
open
and
transparent
thing.
I
So
that
would
be
then
I
think
by
adopting
it
into
operator
hub
that,
oh,
that
conversation
becomes
something
that
people
will
have
more
influence
over
and
be
able
to.
Maybe
extend
operator
have
that
IO
to
do
all
the
other
types
as
well.
So
I,
don't
think
the
donation
of
operator
hub
do,
as
is
precludes
it
from
extending
to
adopt
other
forms
of
operators,
but
I
think
in
my
humble
opinion,
again
those
ones
that
don't
they
would
have
to
be
some
metadata
tagging
things
show
showcasing.
I
You
know
maybe
what
level
of
maturity,
because
that's
fairly
been
with
operator
help
one
of
the
most
significant
things
is
being
able
to
show
the
level
of
automation
and
in
place.
So
I
think
that
that
is
where
some
of
the
disparity
comes
in,
but
I
don't
think
we
once
it's
in
ciencia
I,
don't
think
anything
for
cludes.
G
Let
me
ask
this,
though,
should
it
be
a
separate
project
from
the
operator
framework,
so
that
way,
let's
just
imagine
a
world
where
both
kudo
and
the
operator
framework
have
both
been
submitted
and
become
successfully
part
of
the
scenes
yeah
if
they're,
both
going
through
that
process.
Now,
if
I
understand
the
backlog
of
issues
right
and
they're,
both
operator
based
because
there's
two
projects
should
the
operator
hub
not
be
coupled
to
any
of
those
projects,
so
that
way,
multiple
projects
can
feed
into
it
again.
I
I
I
think
that
once
the
donation
and
the
adoption
of
it
is
done,
it
is
up
to
the
CNCs
how
how
thank
you,
yeah
so
I
think
we're
just
trying
to
get
it
out
there
into
an
openly
governed
fashion
and
and
have
to
make
sure
I
know.
From
my
perspective,
I
would
like
to
continue
to
have
it
support
the
operator,
lifecycle,
management
and
and
yeah.
G
The
reason
I'm
sorry
the
reason
that
I
bring
this
up
is
when
you
submit
it,
it's
not
that
the
CNC
F
controls
it
the
CNC
F.
Let's
each
individual
project
have
their
own
governance
and
they
are
seen
as
each
individual
projects,
okay,
and
so
it
isn't
donate
it
to
the
CNC
f,
as
in
its
the
CNC
F
will
go
control
it.
The
CNC
F
lets
the
project
has
its
governance.
G
So
if
it's
part
of
operator
framework,
then
that
operator
framework
and
that
whole
thing
will
have
one
set
of
governance,
and
that
would
fall
under
that
and
it
would
be
separate
from
the
governance
of
Prometheus
or
Envoy
or
any
of
the
other
ciencia
project,
and
if
CUDA
landed,
it
would
have
its
own
governance
structure.
That
is
separate.
In
fact,
if
you
look
at
the
CNC
our
project,
each
one
of
them
has
slightly
different
governance
that
works
for
them.
I
That
group
becomes,
and
whatever
its
name
but
I
the
opportunity
here
is
collaborate
openly
and
to
get
more
and
and
the
possibility
of
it
being
adopted
as
an
incubated
project
allows
it
to.
You
know,
move
forward
and
have
more
people
under
the
umbrella
of
the
operator
framework.
We
doing
that
collaboration
with
other
projects.
So
it's
a
semantic
thing.
I
got
it.
If
you
need
to
separate
it
out
and
in
order
to
get
operator
framework
done,
then
that's
that's
the
TOC
is
call
but
I.
I
You
know,
I
personally,
think
that
the
that
we
need
for
operator
framework
and
for
the
operator
hub,
IO
together,
is
for
them
to
grow,
underneath
one
project
and
collaborate
with
all
the
other
ones.
So
that's
just
you
know
the
the
way
that
I
would
like
to
see
this
move
forward,
but
yeah,
that's
really.
The
TOC
is
call
whether
they
want
all
the
pieces
or
they
separate
it
out.
I'm.
A
I
There's
a
there's
their
scaffolding
for
ansible
and
helm
and
Python
and
and
other
languages
already
in
place
to
to
turn
pretty
much
any
of
those
into
operator
hub
service
offering
or
offerings
so
it
you
know,
it
just
depends
on
how
we
extend
yeah
thanks
putting
money.
You
can
add
my
name
in
more
too.
So
it's
it's
pretty.
If
you
go
to
operator
hub
dot
IO,
you
can
find
the
way
to
add
anything
in
that's.
A
Yeah
I
think
he
did
the
thing,
maybe
that
you
see
also
this
time
of
it
well.
Worried
might
be
the
wrong
word
think
what
needs
further
discussion
is
to
be
create
a
code
standard
where
an
operator
looks
like
right
now
we
saw
involving
other
people
who
have
operated
the
frameworks
like
the
Twitter
were
mentioned
for
their
requirements
and
because,
as
soon
as
we
as
appointed
as
a
CNCs
project,
that's
pretty
much
what
we
do.
This
is
where
the
operator
has
to
look
like,
and
this
custom
with
the
other
folks
as
well.
Yeah.
I
Yes
or
no
I
mean
I,
think
there's
one
of
the
things
that
the
CNC
F
has
I
think
been
clear
about.
Is
that
I
mean
if
optics
might
be
in
mark
how
people
market
is
but
they're,
not
setting
gold
standards
for
projects
or
I
mean
if
you
look
at
container
registries,
for
example,
there
are
multiple
container
registries
out
there.
You
know,
that's
not
really
what
we're
doing
here.
Well,
so
I'm,
you
know
I'm
just
putting
that
out.
I
There
I
think
what
what
I
see
this
more
is
an
opportunity
to
collaborate
with
other
people
who
are
trying
to
solve
the
same
problems,
and
you
know
come
up
with
some.
What
I
will
what
I
am
personally
this
is:
is
this
that
I,
like
about
the
operator
framework
and
the
lifecycle
management
is
that
it
gets
us
beyond
basic
installs
and
of
operators
and
things
that
can
do
and
what
I?
What
I
think
is.
I
Is
it
creates
a
pattern
for
things
to
get
to
have
more
maturity
and
be
more
helpful
in
the
life
cycle
of
operators
and
that
service
on
any
kubernetes
platform.
So
that's
that
I
think
that's
more
of
the
goal
of
operator
hub.
That
IO
is
to
bring
whom
to
show
what
level
of
support
you're
going
to
get
for
the
different
things
and
so
yeah
it's
more
of
a
pattern.
G
Don't
you
let
me,
let
me
ask
this
in
so
I've
got
two
questions
or
a
comment
in
a
question.
One
of
the
things
that
this
that
I've
heard
in
the
TOC
meetings
repeatedly
is
the
CNC
F
isn't
going
to
bring
that
collaboration
the
project
they
that
should
be
out
there,
collaborating
whether
it's
in
the
CNC,
F
or
not.
G
That
is
just
one
of
the
pieces
of
feedback
they've
given
so
to
say
to
have
in
the
CNC
F
will
bring
the
collaboration
normally
one
of
the
things
they
push
back
on
that
they
say
you
should
have
that
collaboration
anyway.
I've
heard
that
and
a
number
of
calls
my
question
would
be
so
we
talk
about
something
like
operator,
lifecycle,
manager,
right,
let's
go
and
say,
and
this
is
something
that
I've
done
where
I'm
at
well.
Let's
say:
I
have
an
operator
packaged
up
as
a
helmet
chart.
G
There's
no
operator
lifecycle,
manager,
involved
and
I'm
doing
this
to
create
an
entirely
separate
Avenue
as
an
example.
It's
only
for
an
example
with
the
operator
hub
now.
How
would
something
like
that
be
positioned
and
displayed
and
fit
in
to
the
operator
hub?
Because
it's
a
different
flow
from
the
other
tooling,
that's
that's
proposed
here.
You.
I
C
I
Don't
think
it
does,
it
injects
some
more
llamó
around
it
to
make
it
work
in
the
flow
flow,
but
yeah
I'm
gonna
reserve
judgment
because
I
just
should
I
shouldn't
speak
where
I
don't
know
exactly
the
anti-venom
apologies,
but
I,
don't
think
so.
It's
pretty
simple
scaffolding
that
gets
added
and
wrapped
around
it.
So
there
is
a
path
for
anything
and
or
there
is
a
pattern
for
anything
new
things
like
whether
it's
Kudo,
which
I
don't
know
much
about
or
could
you
know
COO
builder,
there
are
collaborations
going
on
already.
I
What
we
just
really
want
to
get
I
think
is
the
lift
from
more
eyeballs
on
the
project
that
you
get
by
becoming
an
incubated
project,
and
you
know
then
probably
myself
and
other
people
who
come
to
the
project
will
be
doing.
You
know
it's
still
driving
that
collaboration,
Daniel
you're
myself
have
been
doing
most
of
it
as
well
as
Matt,
Dorn
and
a
few
other
folks
and
Chris
short,
but
there
you
know,
there's
you
know
we're
gonna,
be
honest,
there's
primarily
Red
Hatters
working
on
it
on
the
outreach
right.
I
B
B
B
I
Thing
I
would
ask
is,
and
it's
normally
I
just
jumped
on
the
call,
so
it's
normally
Aaron
Boyd
and
Daniel
Messier,
who
have
been
answering
most
of
the
questions
for
the
TOC.
So
if
you
put
it
in
an
email
thread,
let's
get
them
to
answer
the
official
ones,
because
I'm,
basically
just
the
manager
behind
operator
hub,
do
so
I'd
like
them
to
be
able
to
weigh
in
it
and
I.
G
Can
can
I
make
two
suggestions
of
things
to
work
out
in
this
one?
Is
the
TOC
asked
about
the
definition
of
an
operator
not,
and
so
that
was
a
specific
ask
of
the
TOC
to
come
back
with
something
in
writing,
and
so
we
should
probably
chase
that
down.
Hopefully,
it's
pretty
easy
since
chorus
to
find
it
in
a
manner
that
everybody's
been
running
with
a
few
years
ago,
and
then
the
second
will
action
I
would
suggest
is
there
is
Kudo,
is
an
operator
and
they
have
an
open
submission
to
the
TOC.
G
So
I
would
suggest
sig
app
deliver
actually
reach
out
to
the
project
on
it,
because
a
lot
of
people
who
submit
projects
still
don't
know
that
it
needs
to
go
through
a
cig.
First
there's
a
bunch
of
communication
and
stuff
there's
even
been
some
due
diligence,
that's
gone
before
the
TOC
and
they
said
hey.
What
did
the
cig
review
this
first
and
help
with
it?
They're
like
we
didn't
know
that
that
had
to
happen.
G
So
it's
probably
a
good
idea
to
have
the
Stig
actually
reach
out
to
kudo,
to
loop
them
in
and
maybe
bring
them
in
and
start
doing.
The
due
diligence
that
the
TOC
Xbox
I
would
suggest
those
is
to
next
steps,
because
that'll
kind
of
help
see
how
everything
fits
together,
because
that's
you
know
see
I
think
wanted.
G
I
And
I
don't
know
how
many
of
you
are
from
the
sig
are
coming
to
coop
con
in
San
Diego
in
a
couple
of
weeks.
If
you
want
to
have
a
face-to-face
about
this,
I
have
a
room
at
the
at
the
Marriott.
That
I
would
happily
schedule
a
time
to
deep
dive
in
further
and
bring
Erin
and
not
Daniel,
but
Rob,
some
ski
and
the
other
engineers
to
on
this,
and
we
can
talk
about
it
more
in
further
depth.
I'm,
not
sure
whether
you'd
like
to
do
that
or
not.
If
we
have
time.
A
Yeah,
let's
see
we
can
get
together,
I
might
I'm
trying
to
be
available
and
see
was
I
was
like
other
people
as
well.
I
wasn't?
We
won't
have
time
during
the
discussion
per
se,
but
we
can
try
to
find
the
time.
Yeah
I
think
there's
just
a
couple
more
things
that
I
think
still
need
some
some
homework
here.
A
What
is
what
are
the
requirements
to
run
it,
and
eventually
somebody
has
to
provide
the
infrastructure
for
operator
puppet
if
it's,
if
it
is
submitted,
and
even
if
somebody's
sponsoring
this,
we
should
have
a
better
understanding,
what's
really
required
to
run
it.
The
second
point
is
really
on
the
governance
model,
where
I'd
also
like
to
better
understand
how
you
envision
the
future
governance
model
to
work.
How
like
operator
can
make
it
into
the
operator
hub.
A
B
A
Happy
to
do
so
and
then
I
think
I
miss
Mona
splitting,
where
it's
two
separate
submissions
we
can
discuss
later
by
the
way
I
just
read
the
website
again
of
operator
hop
about
publishing
and
submitting
it
I
think
this
is
also
kind
of
like
a
bit
ambiguous
because
there's
40
operator
your
operators
should
be
able
to
be
managed
by
the
operator
lifecycle
management,
but
it
also
a
lifecycle
manager
and
what
I
think
comes
to
automated
testing
I
think
it
has
the
operator
lifecycle
manager
as
a
requirement
in
there.
Yeah.
I
A
I
Worries
I'm
sure
I
misspoke
at
least
5
times,
so
that's
we'll
clean
it
well
I'll
go
through
the
notes
and
make
sure
that
the
folks
who
are
on
the
engineering
side
answer
your
questions.
B
A
A
So
this
was
like
the
first
delivery
that
at
the
first
deliverable
we
were
planning
to
work
on
and
I
think
Harry's
done
a
great
job.
Moving
this
point
forward,
I
think
the
key
piece
in
there
is
really
that
diagram,
which
is
a
bit
further
down
on
the
yeah
and
then
on
the
reference
model,
where
we
have
application
definition,
packaging
and
so
forth.
B
A
So
yeah
I
think
we
should
35
what
really
has
to
be
defining
the
individual
layers
and
where
the
various
is
like
and
a
certain
logical
flow
of
components
and
also
which
projects
are
the
people
are
using
for
application
delivery,
see
themselves
at
which
layers,
which
other
helps
us
with
the
landscape
work
so
where
we
would
put
them
so
because
they
validate
two
things.
If
we
can
define
a
landscape
based
on
the
model,
it
will
show
that
the
model
actually
works
for
categorizing
projects.
A
G
Yeah
and
I
would
actually
expect
many
projects
to
cross.
Many
of
the
different
layers
are
the
things
you're
calling
topics
in
this
because
they're
gonna
be
targeted
around
solving
and
user
problems,
and
when
you
try
to
do
that,
you're
gonna
end
up
touching
many
of
these.
In
order
to
create
a
good
user
experience
or
to
provide
usability
in
the
solution.
That's
targeted
have
as
saket
delivery
captured,
who
different
roles
and
actors
are
and
their
needs,
and
things
like
that
at
all.
Yet
I.
A
Think,
honestly,
we
haven't
come
any
further
than
what
you
see
on
the
screen
right
now
really
on
this
one.
Specifically,
some
parts
were
in
the
Charter,
but
most
of
the
reasons
efforts
were
really
about
like
project
presentations
and
not
so
much
on
the
model.
That's
why
we
I'm
trying
to
push
the
focus
back
now,
a
bit
more
in
the
application,
with
every
model
here,
yeah.
G
I
mean
if
you're
trying
to
create
a
usable
solution
for
certain
end
users
like
an
application
developer,
an
application
operator,
you're
gonna
end
up
having
tools
that
cover
a
lot
of
the
different
layers
here,
just
in
order
to
create
something
that
has
good
usability.
You
know,
if
you're
doing
the
UNIX
model,
where
you've
got
lots
of
things
separated
into
small
slices,
possibly,
but
quite
often
you
have
to
chain
those
together
and
act,
a
lot
of
the
app
developers
and
app
operators
out
there.
A
B
Actually,
this
is
another
thought.
Is
that
we're
going
to
be
meeting
in
person
in
two
weeks
like
two
weeks
me
today,
I
think
so
why
don't
we
prepare
something?
So
we
I'm?
We
will
be
doing
this
so
that
we
can
actually
show
this.
So
we
can
get
some
consensus
with
a
much
larger
group
of
people
and
then
what.
B
A
Yeah
thanks
for
pointing
up
and
actually
I
said
that
session
dedicated
to
the
application
delivery
model
at
cube
con,
so
everybody
that'll
come
join
that
session
and
I
think
it's
actually
called
a
workshop
and
I
think
it
should
be
a
workshop
where
everybody
say:
okay,
what
do
we
want
to
have
like
in
the
application
definition
piece?
What
do
we
want
that?
A
E
A
A
B
I
B
I
I
do
have
someone
graciously
gave
me
a
room
at
that-
that's
like
Marriott
for
the
week.
Well,
if
we
wanted
to
have
a
separate
meeting
to
talk
about
any
of
the
operator
hub
stuff
or
something
else,
adapt
related
I'd,
happily
schedule
some
time.
It
only
fits
15
people
and
I'll
add
some
notes
to
the
meeting
there
with
follow-up.
Some
of
the
questions
that
were
asked
I.
E
I
B
All
right
all
right!
Well,
in
that
case
Diane
we
will
be
following
up
with
you
and
I
think
there
are
a
couple
of
people
that
we
would
like
to
have
a
conversation
around
that,
but
we'll
have
to
keep
in
mind
the
amount
of
space
that
we
have,
but
we'll
work
out
the
logistics
for
that
all
right.
Well,
thank
you,
Diane
for
popping
in
here
from
whatever
else
you
were
doing.
That
is
appreciated
and
thank
you.