►
From YouTube: CNCF SIG App Delivery 2020-02-19
Description
CNCF SIG App Delivery 2020-02-19
A
B
B
B
Hello
and
welcome
to
the
February
19th
seems
the
FCX
I
got
deliberate
meeting,
I'm
Brian
Myles
and
today
we're
going
to
talk.
We
actually
have
a
fairly
short
gym
today,
but
today
we're
going
to
talk
about
judo
and
is
it
it's
kudo
and
pam
graduation
requirements
or
questions
from
mat
arena
so
to
get
us
started.
B
D
B
D
D
And
if
you
look
at
what
kubernetes
is
it's
a
you
know,
it's
declarative
data,
that's
operated
on
by
a
series
of
controllers,
and
most
of
us
already
know
that
this
so
I'm
gonna
go
quick.
But
it's
what
I'm
doing
here
is
building
up
a
little
bit
of
the
kudo
view
of
this.
This
story,
so
we
could
enjoy
like.
What's
a
controller
right
and
a
controller
is
really
a
really
a
reconciliation.
Loop
of
I
have
an
actual
state
in
a
desired
state.
Please
advance
me
towards
that
state
right.
D
You
know
I
like
this
particular
definition,
but
I
wanted
to
pick
this
one
to
enhance
upon
it
a
little
for
the
perspective
of
why
we
why
we
ended
up
creating
Kudo,
and
that
is
you
know,
we
know
what
a
controller
is,
and
so
why
not
just
call
them
all
controllers,
and-
and
what's
this
difference
in
an
operator
and
it
comes
down
to
controllers,
reconcile
state
and
they
don't
care
what
the
underlying
workload
is.
If
you
think
about
the
controller
manager
in
kubernetes,
it
cares
that
it
has
a
deployment.
D
It
doesn't
care
about
what
its
deploying
right,
then
that
goes
for
any
resource
that
that
is
maintained
by
that
connect,
controller
manager
or
the
scheduler
or
anything
else,
and
so
controllers
plus
resources
doesn't
really
make
sense
in
general
terms,
and
one
thing
I
want
to
talk
about
and
harp
about,
because
this
is.
This
is
very
central
to
the
way
CUDA
looks
at
the
world.
D
So
so,
for
example,
for
application
backups
right
now
and
restores
we
might
use
something
like
bolero
or
you
know,
they're
they're
scaling,
there's
a
couple
other
things
that
we
need
to
do,
and
so
what
we
can
say
is
then
is
from
from
the
perspective
that
I'm
talking
about
is
the
operators
control
it's
well
optimized
to
the
requirements
of
a
given
domain,
so
we've
set
out
okay,
great
Jared,
you've
convinced
me:
let's
operate
it
all
the
things
right.
We
want
some
Postgres,
let's
write
an
operator,
we
want
any
some
migrations
right
and
operate.
D
You
know:
I'm,
writing
more
and
more
lines
of
operator
code
for
less
and
less
less
domain-specific
code,
and
this
is
kind
of
a
prolific
or
a
proliferation
problem
that
I
want
to
I
want
to
talk
about
more,
but
to
give
a
really
concrete
example.
Here
we
look
at
the
elastic
operator,
elasticsearch
operator
and
elastic
is
a
massive
Java
app.
It
has
65
thousand
lines
of
application
level
code
for
the
operator
right,
that's
not
including
dependencies
on
controller,
runtime
or
client
go
or
a
million
other
dependences
that
you
need
to
interact
with
the
kubernetes
api
server.
D
So
you
get
all
this
setup
right,
you've
written
your
operator
and
you
think,
everything's
great
and
then
kubernetes
1.17
drops
and
removes
deployments
view
on
beta
one
or
something
changes
right
and
now
all
of
your
dependencies
are
out
of
date
and
nothing
works
with
each
other.
And
so
it's
not
just
a
matter
of
you
know.
You
see
this
and
I'm
done,
and
you
do
this
and
I'm
done,
and
you
do
this
and
you're
done.
You
actually
have
to
continue
this
change
management
over
time
and
a
very
large
scale
of
something.
D
So
you
need
go
developers
who
new
communities
very
idioms
very,
very
well,
or
you
need
to
invest
a
lot
of
time
into
learning
those
idioms
in
order
to
write
an
operator
right
and
operators
are
potentially
useful
beyond
the
category
seen
today
and
I
want
to
harp
on
that
a
little
bit
later
than
something
we've
seen
in
Kudo.
That
actually
surprised
us
about
that,
and
what
what
we
really
should
be
able
to
do
is
enable
new
pathways
for
developing
native
applications
on
top
of
kubernetes.
D
That
interact
well
with
that
with
with
the
post
environment
that
it's
in.
So
this
is
where
I
want
to
introduce
Kudo
the
kubernetes
universal
declarative
operator.
It's
it's
right
now,
a
tool
for
application,
sequencing
and
resource
ordering,
but
it's
a
lot
more
than
that.
It's
really
about
shipping,
an
application
with
its
runbook
I
mean
I'm,
going
to
open
chat.
Just
so
I
can
keep
track
of.
If
there's
any
questions,
no,
it's
not
right!
D
We
try
to
use
this
a
lot,
because
what
we're
talking
we're
talking
about
his
writing
operators
using
crts
and
taking
kubernetes,
then
that
could
be
a
sword
to
the
hilt
by
saying
everything
looks
like
a
kubernetes
object
and
use
it
as
much
as
possible.
So
when
we,
when
we
go
to
make
decisions,
we
try
to
ask:
how
do
we
do
this
in
a
kubernetes
native
way
with
that
kubernetes
resource
data
model?
D
That's
done
in
this
plugin
should
be
done,
be
doable
with
raw
cube.
Ctl,
so
if
we
look
at
what
kikuno
is
today,
we
have
a
stable
and
generally
available
version
of
kudo
that
does
some
declarative
sequencing
of
resources
and
can
run.
You
know,
deployments
updates,
upgrades
and
some
custom
plans.
So
if
you
want
to
do
a
backup
or
you
can
you
can
view
that
sort
of
functionality
now
in
that
base,
core
functionality
is
stable.
D
While
we
work
on
1.0
features
that
are
really
centered
around,
how
do
we
make
life
easier
for
both
operator
developers
and
end
users
and
we're
shooting
for
a
1.0
by
puke
on
China,
so
in
progress
or
in
play
or
planning
for
that?
Right
now
we
have
user-defined
commands
for
the
kudos
Eli.
So,
whereas
we
want
everything
to
be
declarative,
we
recognize
that
can
be
hard
for
end-users,
and
so
we,
you
know
what
one
thing
is
to
be
able
to
add
in
sub
commands
into
the
Kudo
CLI.
D
Under
the
hood,
that's
going
to
be
a
CRT,
but
we
at
least
offer
that
imperative
shell
to
make
some
scripting
easier,
we're
working
towards
graph
based
sequencing
sort
of
resources.
If
you
can
think
about
terraform
but
applied
to
kubernetes,
see
RDS,
that's
what
we're
talking
about
here
so
that
we
we
can
get
out
of
manual
dependency
management
and
allow
kudo
to
really
do
a
bunch
of
features
for
you
around
sequencing
and
ordering,
as
well
as
drift,
detection
and
and
expiring.
D
The
right
areas
of
you
up
when
you
go
to
update
we'll
be
doing
some
automatic
sandbox
saying
a
big
one
for
us
to
see
our
D
schema
management
and
declared
your
day
to
operations.
So
right
now,
back
in
the
architecture
slide,
you
saw.
We
had
an
instance
we're
working
on
replacing
that,
where
Kudo
will
manage
the
schema
for
various
CRTs
on
the
user's
behalf.
D
So
instead
of
creating
a
topic
via
the
Kafka
API,
you
would
create
a
broker
CRD
and
have
a
create
update,
delete
plan
that
then
forces
everything
possible
to
be
declarative
and
represented
inside
of
kubernetes.
And
you
could
you
can
take
this
pretty
far.
We
want
to
see
how
far
we
can
drive
that
before
it
becomes
a
problem.
We've
already
identified
some
areas
where
you
need
that
inherit
of
hash
write,
a
restore
of
your
data
is
really
hard
to
make
very,
very
declarative,
but
it
back
up
instead
of
declare
I
want
to
back
up
kudos.
D
We
want
Guto
operators
to
be
able
to
easily
work
with
other
operators,
whether
or
not
they're
written
in
Kudo
or
queue
builder,
directly
or
operator
SDK
or
whatever
else,
so
just
a
quick
drop
into
the
our
community.
So
far
we
have
a
lot
of
members.
In
slack,
we
have
a
lot
of
stars.
We
have
a
lot
of
contributors.
Everything
we
do
is
out
in
the
open,
open
governance
and
we
followed
a
cap
process
since
the
beginning
for
introducing
Oliver
enhancements,
and
we
were
the
first
operator
focused
tool
in
the
crew
brains
podcast.
D
We
found
out
that
someone
back
in
a
few
months
ago
put
us
as
the
the
top
recommended
to
land
kubernetes
operator
Docs,
which
was
a
pretty
cool
thing
to
discover.
We
have
an
upcoming
workshop,
we're
talking
about
all
over
the
place
and
we
have
people
who
are
not
us
even
writing
blog
posts
about
this.
So
we've
started
a
flywheel.
D
In
fact,
we
have
people
interested
and-
and
we
have
people
interested-
not
just
writing
their
databases
but
also
orchestrating
microservices
large
large
micro
service
applications
using
a
Kudo,
and
so
that's
that's
what
I
was
talking
about
earlier.
When
I
said
you
know,
people
are
wanting
to
do
this
for
just
beyond.
You
know
run
my
Postgres
right.
People
are
looking
at
this
for
their
rails,
app
or
for
for
their
other
tooling,
as
well,
and
building
out
a
cohesive
ecosystem.
D
D
So
looking
at
CUDA
on
the
CNC
of
ecosystem,
there's
a
bunch
of
things
that
we
want
to
bring
bring
out
and
and
and
there's
there's
reason,
there's
value
we
see
in
some
of
the
stuff
that
we're
working
on
with
other
people
in
the
CNC
F
and
for
the
community
at
large.
One
thing
we're
working
on
is
the
Cooper
nays
operator
interface
and
what
that
is
and
I'm
going
to
post.
D
All
these
slides,
too
I
wanted
to
save
some
time
for
discussion,
but
really
what
we're
talking
about
is
a
specification
that
enables
compositions
of
operators
and
other
tools
so
that
we
can
start
building
ecosystems
right
and
building
tooling.
That
consumes
that,
so
that
end
users
don't
have
to
learn
a
new
tool
for
every
single
operator
or
a
new
way
to
interact
with
every
single
operator
and
operators
don't
have
to
manually
interact
with
each
other.
If
you
have
a
cough
copper
term,
you
want
to
get
a
connection
string
out
of
a
zookeeper
operator.
D
That's
what
Co
is
attempting
to
solve
here.
If
you,
if
you
want
to
write
a
CLI
tool
or
if
you
want
to
write
a
UI
or
dashboard
that
that
actually
consumes
the
behaviors
of
an
operator,
this
operator
can
do
a
backup
or
restore
that's
what
koi
is
intended
to
solve,
and
it's
really
important
to
note
that
this
is
really
supposed
to
be
not
be
competitive,
but
suppose
we
compatible
with
other
specifications
that
aren't
really
covering
this
particularly
use
case.
D
Another
one
that
we've
we've
talked
to
a
lot
of
people
about,
and
we
have
a
lot
of
users
of
as
as
cuddle,
which
is
our
kubernetes
test
tool
and
it's
a
declarative
framework
for
writing,
conformance
and
ETA
tests
for
kubernetes
resources.
It's
based
on
sir
shins.
We
have
a
lot
of
companies
using
it
to
test
really
anything
on
kubernetes,
but
not
just
operators
of
not
just
kudo
and
for
us
and
we
the
reason
we
developed.
D
This
was
to
enable,
like
conformance
and
certification,
matrices
of
various
distributions
and
various
kubernetes
versions
and
different
operators
that
you
can
quickly
at
a
glance,
see
that
this
operator
works
on
these
versions
of
kubernetes
and
not
these
versions
of
kubernetes
and
also
determine
how
mature,
how
many,
how
many
different
yeah
features
on
a
maturity
model
that
these
operators
supply
right
now.
The
best
way
to
use
that
is
in
the
the
as
a
sub
command
and
cube
Caudill
kudo.
Well,
we
just
moved
it
out
into
its
own
kernel,
repo.
You
can
totally
say
cute
cuddle
cuddle.
D
If
you
really
want
to
that's
what
I
do,
but
it
will
be
we're
working
towards
a
first
release
of
the
of
the
standalone
cuddle
as
it's
cool,
we
have
a.
We
have
a
lot
of
community
stuff,
we're
working
on
the
coin,
cuddle
site,
I,
think
they're,
just
redirects
for
now
to
the
repos
or
they're
they're,
not,
but
in
the
coming
days
those
two
sites
will
be
up.
Kudo
Deb
is
up.
D
We
have
Kudo
on
the
Cooper
like
we
have
github,
and
you
know
there's
a
lot
of
reasons
that
we
think
we
can
bring
a
lot
of
value
into
the
CNCs
sandbox
and
and
and
running
this
experiment
out
in
the
open.
And
so
what
we
really
want
to
do
is
enable
some
really
public
experimentation
with
operators,
but
also
conforms
the
certification
of
workloads
on
top
of
kubernetes.
We
really
want
to
align
that
application
delivery
with
other
CNCs
projects
and
work
with
helm
and
others
that
get
in
to
establish
standards
for
distribution
of
workloads
on
kubernetes.
D
We
we
really
want
to
provide
a
really
lightweight
mechanism,
a
programming
model
for
users
to
develop
these
kubernetes
native
applications
and
operators.
Like
I
said
kudos,
not
the
end-all,
be-all
tool,
but
we
think
there's
there's
a
room
here
for
a
lightweight
tool
that
makes
this
a
lot
more
accessible
to
people
who
aren't
kubernetes
domain
experts,
and
then
we
what
we
really
want
to
bring
this
both
our
tooling
and
our
standards
and
advance.
That's
the
incubating
and
graduated
as
the
project
evolves
or
or
gets
some
of
these
integrated
into
kubernetes
directly,
depending
on
that
tool.
D
So
you
know,
if
you're
looking
about
kudos
all
about
it's
really
about
enabling
teams
to
build
operators
in
a
very
accessible
and
maintainable
way,
and
we
think
the
inclusion
in
CNC
F
sandbox
grows
that
mission
and
with
that
on
Jeff
Dylan.
We
have
a
few
other
members
of
the
team
here
and
thank
you,
everyone
for
your
time.
B
D
B
E
C
B
We've
created
a
a
questionnaire
that
needs
to
be
filled
out,
and
it's
more
on
the
it's
more
on
the
project
maintainer
to
get
this
out
and
then,
depending
on
whether
it's
a
sandbox
or
it
is
an
incubation
project.
Then
we
work
with
Amy
to
get
this
in
front
of
the
TOC
and
they
do
what
they
do.
So
that's
what
that's!
What
we're
trying
to.
A
Yes,
we
recommend
this
or
no
here's
some
other
things
that
we
think
might
be
a
better
fit,
and
that
part
is
that
I
think
we're
working
through
right
now,
it's
like
what
happens
if
it's
not
fit,
but
in
the
case
of
like
hey
Argo
Argo
was
ready
to
be
able
to
go
through
over
TOC,
so
being
able
to
say,
hey,
we've
made
the
recommendation
TOC
now,
please
review
based
on
our
recommendations
and
what
we've
gone
through.
Does
that
help
that
I'm
sure.
D
C
D
C
C
D
E
It
is
in
Matt,
this
is
Diane
Mueller.
Thank
you
for
for
asking
all
these
questions,
because
it
is
very
confusing
and
you
know
I
know
we
went
for
an
incubation
status
with
the
operator
framework
and
now
it
feels
a
little
bit
like
we're
in
limbo,
so
I,
don't
really
know
and
and
I
ought
to
know,
but
I'm
asking
again
what
the
status
is.
Is
it
now
to
go
to
the
TOC
and
we
need
to
request?
B
C
C
I
have
questions
for
home
here
and
I've
had
so
many
questions
about
Kudo,
because
there's
just
so
many
changes,
and
so
much
of
this
is
new
and
reading
the
docs
that
they've
got
is
is
a
little
confusing
and
we're
talking
new
process
and
you're
like
right
over
this
hump
of
you
came
in
during
the
old
process.
You've
got
it,
but
they
had
to
stop
because
the
TOC
couldn't
vote
another
new
teo
season.
What
process
I'm
sure
there's
a
lot
of
confusion
there
yeah
yeah.
E
And
we
we've
been
in
limbo
now
for
almost
four
months,
and
this
is
it's
not
a
comfortable
place
to
be
and
we
keep
trying
to
go
through
whatever
hoop
everybody
puts
in
front
of
us
and
get
all
the
information
out
and
Gerard.
This
is,
you
know,
listen
and
learn
and
there's
nothing
saying
that
this
process
isn't
going
to
change
again.
E
So
you
know
with
the
new
TOC,
so
you
know
my
druthers
is
I
would
really
just
like
to
get
to
a
vote
at
the
TOC
as
soon
as
possible,
preferably
before
coop
Connie
you
so
that
we
could
deal
with
it
there
and
have
some.
You
know,
recruitment
for
new
bodies
and
external
parties
to
work
and
collaborate
like
Jared
and
Matt,
and
everybody
else
at
that
event.
So
you
know
this
will
be
the
second
koukin
opportunity.
We've
missed
to
bring
everybody.
C
E
We
and
we
will
I,
mean
I've,
got
a
meeting
room
set
up
and
everything
else,
but
it's
just
this
is
getting
to
be
redonkulous,
as
the
word
I
use,
the
most
and
and
Chris
I
know
you're
on
the
call
and
Brian
and
Amy
I
know
you're
there,
but
it's
like
if
we
could
just
get
it
on
the
agenda.
I,
don't
care
what
we
have
to
do
to
educate
the
11
new
members
of
the
TOC
we'll
do
it.
We
just
need
to
get
this
done.
A
No
I
very
much
appreciate
that
we
are
having
conversations
to
be
able
to
help
move
all
of
this
forward
and
to
Matt's
point.
You
he's
correct.
You
guys
are
kind
of
in
that
weird
edge
case
between
here's
process.
Here's
bringing
TOC
comes
in,
but
thank
you
for
bringing
it
up.
I
will
take
a
look
at
this
with
Brian
and
with
Chris.
Today,.
C
C
We
just
heard
about
the
operator
framework,
we'd,
really
love
to
know.
What's
going
on
when
things
are
happening,
so
at
least
we
feel
like
we're,
not
forgotten,
and
it
forces
you
to
not
forget
us
and
keep
going.
So
that's
the
thing
I
would
ask,
but
with
now
with
that,
what
I
wanted
to
do
is
I'm.
Looking
at
this
new
process
and
helm
wants
to
go
for
graduation,
we've
had
all
the
eyes
dotted
and
t's
crossed
for
some
time.
C
Everything
we're
doing
now
mostly
goes
above
and
beyond
the
graduation
requirements
and
so
or
just
filling
out
the
right
things
like
you
want
to
know
who
the
users
are
in
a
and
a
markdown
file
in
a
special
format.
Right
you've
got
this
place
where
you
need
to
do
something
like
okay,
so
you've
got
to
take
our
users
and
get
them
vacuumed.
Some
of
them
anyway,
get
them
documented
over
here.
C
Most
of
that
was
supposed
to
be
done
for
incubation,
but
we
went
in
for
incubation
quite
some
time
ago,
so
we
don't
have
a
due
diligence
form
like
that
and
so
I'm
wondering
what
does
the
process
look
like
now,
we'll
go
we'll
work
out
the
due
diligence
we'll
get
with
our
toc
sponsor,
who
is
still
on
the
TOC
and
we'll
start
working
through
some
of
these
things.
But
what
else
do
we
need
to
do?
Do
we
need
to
come
here
and
give
a
presentation,
maybe
in
two
weeks
that
walks
through
the
due
diligence?
C
B
A
C
Since
I
just
learned
that
we
need
to
do
it
yesterday
and
I'm
prepping
for
a
webinar
to
give
next
Tuesday
I
have
not
started
it
yet.
Okay,
the
plan
is
to
start
it
either
this
afternoon
or
more
likely.
Tomorrow
morning
we
know
most
of
the
things
it's
just
a
matter
of
going
and
filling
it
out.
So
probably
sometime
tomorrow
is
when
I
will
start
work
on
it.
Okay,
so
I
can
probably
get
started
on
it
and
then,
if,
after
the
meeting
in
this
timeslot,
you
could
give
me
some
direction.
I'd
appreciate
it.
Yes,.
A
A
B
F
Hi
yeah
I
am
another
supplicant
to
the
process
of
moving
through
this.
This
gauntlet
and
I'm
here
representing
some
of
the
core
team
of
cloud
nebula
PACs
project.
Basically
just
checking
in
and
sounds
like.
We
did
a
pious,
we
did
a
PR
in
January
and
I'm
reading
the
comments
on
the
PR
that
are
saying
we
need
to
do
this
same
due
diligence,
template
I,
have
a
guess,
a
couple
clarifying
questions.
F
My
I
had
used
the
template
as
a
means
of
kind
of
figuring
out
what
information
we
might
need
in
the
PR,
but
and
when
I
read
this
template
it,
it
sort
of
seems
like
it's
for
people
who
are
leading
or
like
us,
the
quote
is
leading
or
contributing
to
due
diligence.
Should
we
be
filling
it
out
ourselves
or
does
someone
else
have
to
fill
it
about
out
about
our
project?
No.
F
F
B
B
B
D
Sure
so,
right
now
we
have
a
repository
model
and
actually
can
ask
Aaron
can
can
talk
more
to
that
in
the
Kudo
channel
as
well,
but,
basically
right
now
you
can
stand
up
your
own
kudo
repository
and
package
and
put
your
operators
into
into
that
repository.
The
reason
we
haven't
come
up
with
a
more
centralized
we
we
do
have
like
our
stable
repository.
That's
your
default
public
one
and
getting
into
that
is
just
a
matter
of
peering
into
our
operators,
repo,
the
the
operator
and
then
cutting
a
version
of
it
and
doing
a
release.
D
The
reason
we
don't
have
a
more
formal
way
of
doing
that.
Right
now
is
given
some
discussions.
We
had
a
cube
con
with,
with
a
few
other
teams,
helm
and
and
and
operator
framework.
We
wanted
to
wait
and
see
how
how
kudo
could
fit
from
a
distribution
or
a
wider
distribution
method.
How
could
fit
into
those
frameworks
right
and
and
and
participate
there
rather
than
going
creating
like
a
fourth
kudo
right
so
repository
model
for
now
happy
to
talk
about
it
in
the
in
the
kudo
channel
and
walk
you
through
that?
D
B
All
right,
well
with
that,
just
one
more
update
last
week
about
the
work
group
for
air-gapped
about
meeting,
invite
saying
now
that
I
know
where
to
send
it
will
be
coming
out
very
shortly
and
it's
gonna
be
next
Tuesday
around
this
time,
so
we'll
meet
then
decide
if
that's
a
great
time
and
and
then
we'll
get
a
cadence
on
the
calendars.
Sound
good.