►
From YouTube: Contour for Kubernetes Ingress, and K8s Operators: DevOps and Docker Live Show (Ep 73)
Description
Starts at at 3:45. I'm joined by Steve Sloka of VMWare. We're talking about a set of #Kubernetes Operators that Steve maintains including Contour (Envoy Ingress) and Gimbal (multi-cluster ingress).
Support this show on Patreon! It's the #1 way to support me interviewing DevOps and container experts, and doing this Live Q&A. https://patreon.com/BretFisher
Show Notes:
https://projectcontour.io/
https://github.com/projectcontour/gimbal
https://github.com/operator-framework
https://operatorhub.io/
https://kubernetes.io/docs/concepts/extend-kubernetes/operator/
https://coreos.com/operators/
Talk on Contour/Gimbal from KubeCon https://youtu.be/-9U3bLPQDKk
You can get the best coupons for my Docker, Kubernetes, and Swarm courses at https://www.bretfisher.com/courses
Listen to my podcast of this show's past episodes at https://www.bretfisher.com/podcast
Signup for my newsletter to get short emails weekly full of useful things while learning and using containers
https://www.bretfisher.com/newsletter
A
A
Hey
Internet,
my
name
is
Brett.
Welcome
to
the
show.
This
is
where
we
do
a
YouTube
live
every
week
and
I
have
guests
and
friends
where
we
talk
about
everything,
containers
and
docker,
swarm,
kubernetes
DevOps,
the
cloud
all
that
stuff,
all
the
buzzwords
I'm
sure
there's
something
I
left
out
there,
CN
CF
like
coop
con,
you
name
it
like
all
that
stuff,
and
so
we
have
a
lot
of
fun
talking
getting
into
the
tech
talking
about
new
projects,
things
that
are
people
are
doing
interesting,
stuff
and
taking
your
questions.
A
So
the
cool
thing
about
you
being
live
if
you're
watching
this
live.
Is
that
you
get
to
ask
questions
of
my
guest
and
we
will
sort
of
dive
into
whatever
you
want
to
talk
about.
It
focus
on
the
topics
we
have
today
so
today
we're
talking
about
ingress
again,
but
this
time
we're
talking
about
different
projects.
So
with
me
on
the
show
today,
I
need
to
have
like
an
announcement
sound
murmur.
We
have
Steve
sloka
from
VMware.
Welcome
to
the
show.
Ok,.
A
We
talked
briefly
before
the
show
about
the
history
of
you
with
VMware
and
how
you
you
know.
Basically
you
have
this.
You
know
this
history
of
your
your
work
with
containers,
basically
and
how
you
got
involved
with
docker,
and
one
of
things
I
always
like
to
talk
to
people
about
is
how
they've,
Ridge
'''l
II
found
containers
how
they
originally
got
into
docker
and
all
that
stuff.
So
what's
your
story
sure.
B
Yeah
so
before
VMware
in
previously
that
helps
you
know.
I
worked
a
very
large
health
organization,
so
big
80,000
person,
enterprise
do
typical.
You
know
a
ticket
in
two
weeks
to
do
anything
so
to
get
a
server.
You
know,
take
it
in
two
weeks
to
get
an
application
on
the
server.
Take
it
in
two
weeks
that
sort
of
thing.
So
we
found
that
this
is
early.
What
2015,
ish
docker
just
became
kind
of
new.
B
When
we
found
out,
we
could
automate
our
applications
or
development
so
deploying
Java,
deploying
our
Engler
apps
at
the
time
docker
was.
It
was
very
easy
to
get
the
same
versions
everywhere.
You
know
the
same
story
that
docker
sells,
but
the
problem
then
became
was
hey.
We
solved
our
dev
problem,
but
how
do
we
solved
this
in
production?
B
How
we
shipped
this
yeah
again,
because
it
was
hey,
hey
IT
person
go
deployed
his
app
for
us
and
then
troubleshooting
that
sort
of
thing
so
that
it's
the
the
feedback
loops
were
very
long
and
it
made
it
hard
for
us
which
changes.
So
we
came
up
with
we
thought.
Well,
how
can
we
ship
these
docker
containers
to
production
at
a
time,
crew.
A
B
Was
just
new
so
we
started
using
it
in
I.
Think
December
of
2014
right
after
God
became
released
as
the
data,
so
we
started
shipping
things
off
of
that
and
so
from
there
it's
just
kind
of
blow
up.
So
we
my
role
kind
of
moved
from
you
know
just
normal
stuff
software
engineer
to
software
engineer
in
kubernetes
and
containers
right.
A
A
B
A
B
A
B
So
it's
so
yeah,
so
I
work
from
home
so
and
I
work
that
yeah
I
work
on
the
team.
I
work
with
here
at
for
contour
is
I'm.
Half
of
them
are
in
Australia,
so
we
have
a
different
time
zone
metric
as
well
right,
cuz,
we're
you
know
a
few
hours
apart,
so
yeah,
it's
interesting.
So
you
learn
to
work
asynchronously.
You
learn
to
work
through
github
issues,
that
sort
of
thing
and
then
every
not
a
game.
You
have
to
chat.
So
you
stay
up
late.
Writing.
A
A
So
for
those
of
us
we
were
just
talking
about
before
the
show
like
working
from
home
has
really
been
an
advantage
for
us
didn't
really
ever
think
of
that
as
being
an
advantage
of
needing
to
hibernate
at
home.
But
now
we
were
talking
about
that.
All
of
our
friends
are
now
learning
how
to
be
remote
and
I'm.
Just
guessing
that
you
are
now
like
the
office
expert
on
the
best
microphones.
The
best
webcams,
like
the
whole
office
set
up
at
home,
like
you're,
the
Guru
yeah.
B
A
B
A
It's
funny
when
you
do
this
for
a
while,
you
take
it
for
granted.
You
just
assume
that
everybody
knows
all
the
right
tech
for
their
home
office
to
make
it
pleasant
for
them,
and
others
right
and
you
know,
cuz
webcam
still
suck
in
laptops
like
laptop
webcams,
are
horrible,
they're,
no
better,
and
they
were
10
years
ago.
So
it's
nice
to
you,
know
it's
nice
to
always
have
someone
in
the
team
that
actually
knows
this
stuff
that
can
help
everybody
improve
it.
A
A
Definitely
on
the
to-do
list
to
get
more
examples
of
ingress,
especially
once
we
start
getting
in
the
course
into
operators
and
more
the
CRD
stuff,
where
you're
customizing
on
top
of
kubernetes.
So
can
you
tell
us
a
little
bit
about
like
maybe
what
the
separation
is
between
in
void,
contour
and
how
those
are
different
than
just
like
the
engine
decks,
one
that
everybody
knows
about?
What's
that
all
about
yeah.
B
Sure
yeah
contour
is
an
ingress
controller
for
kubernetes,
so
ingress
in
in
general.
Today,
there's
a
couple
versions
of
that.
But
if
we
say
ingress,
you
think
about
the
V
1
beta
1,
so
the
kubernetes,
if
you
want,
they,
don't
want
an
API,
so
there's
a
contract
there.
So
if
you
think
of
an
any
generic
ingress
controller,
it
needs
to
implement
that
contract.
B
So
contour
does
that
and
its
job
is
to
watch
the
cluster
recruit
babies
cluster
for
things
like
secrets,
things,
services
and
points
all
the
information
that
it
needs
to
go,
build
that
that
or
fulfill
that
contract
between
your
request
and
the
back-end
pod.
So
in
the
contour
sense
it's
getting
stop
it
does
that
it
sets
up,
watches
on
the
API
and
kubernetes
and
then
pulls
the
information
back
and
then
it
builds
a
configuration
and
then
it
passes
that
configuration
down
to
envoy
envoys
job
in
contours
world
is
to
be
the
data
path
component.
B
So
all
traffic
or
actually
routes
through
envoy
in
a
technical
sense
contour,
is
the
configuration
server
for
envoy
so
implements
envoys
XDS
protocol,
and
so
when
envoy
spins
up,
we
give
it
enough
information
to
go,
find
contour
once
it
finds
contour,
then
we
can
pass
the
information
down
to
envoy
yeah
and
then
you
know
have
it
configured
for
all
the
routes
and
different
things
that
it
needs
to
do.
And
then,
when
a
request
comes
through
on
whose
job
is
throughout
that
where
it
goes.
A
I'm
gonna
get
into
that
and
start
some
examples.
That's
glad
you
prepared
some
of
those.
The
the
the
ingress
is
that
we
all
think
about
today.
Those
are
largely
one
cluster
type
of
solutions
right
so
eventually,
once
people
get
the
hang
of
kubernetes
and
they
start
to
really
love
it
and
start
to
use
it
more,
they
end
up
with
multiple
clusters:
there's
really
no
shop.
That
I've
seen
that's
fully
invested
in
kubernetes.
That
has
just
one
cluster
and
so
there's
then
this
becomes
these
challenges
around.
A
B
So
so
people
was
another
project
that
came
out
of
FTO
and
so
Gimbels
idea
is
to
take
kubernetes
and
make
it
a
software
based
load
balancer.
So
you
think
of
a
load
balancer
you
want
to
have
you
know
redundancy
and
uptime,
so
we
can
leverage
screw
Briones
to
do
that.
So
we
can
add
capacity
by
adding
notes
to
the
to
that
cluster.
We
can
use
the
built-in
HJ
capabilities
of
kubernetes
to
to
make
it
redundant,
but
essentially
what
what
gimble
does
is.
It
goes
and
allows
you
to
send
traffic
to
multiple
different
clusters.
B
So
today
we
support
kubernetes
and
OpenStack
and
I
have
a
PR
I
need
to
finish
to
send
it
to
these
sphere
clusters,
so
you
can
have
any
number
again
of
upstream
clusters
and
then
traffic
will
hit
this
gimble
cluster
and
then
it'll
decide
where
that
should
go
within
the
from
you
know,
within
of
those
clusters.
So
now,
instead
of
thinking
of,
if
you
apply
the
same
idea
of
ingress
to
a
single
cluster,
you
can
apply
this
now
to
multiple
clusters.
B
So
this
solves
the
problems
of
you
know
what,
if
you
know
what
cluster
needs
to
go
down
for
maintenance
or
you
need
to
upgrade
a
cluster
or
something
you
can
help
do
that
through
the
central
software
base
load
balancer
in
a
sense,
so
the
only
one
thing
I
want
to
call
out
that
we're
looking
at
I
think
to
gimbal
today
is
gimble
requires
having
routable
pot
IDs
right.
This
isn't
fit
every
customer
situation.
A
B
A
I
can
see.
I
definitely
can
agree
with
that
trend,
because
you
know
when,
when
it
was
challenging,
as
the
management
of
a
cluster
itself
gets
easier,
it's
easier
to
deploy
update.
You
know
that
stuff,
you
know.
Originally
it
was
challenging
enough
to
just
get
one
set
up.
You
didn't
have
all
these
cloud
providers
making
one
for
you,
and
so,
especially
if
you're
going
to
vanilla
upstream,
you
didn't
want
to
have
to
make
a
lot
of
them,
because
there
wasn't
any
tooling
around
managing
multiple
versions.
A
This
idea
of
having
multiple
clusters,
and
possibly
even
shifting
from
one
cluster
to
another
as
an
upgrade
path,
is
more
sustainable.
So,
okay,
so
we
have.
This
sort
of
view
of
a
proxy
there's
a
management
engine
on
top
of
that
or
one
cluster,
and
then
there's
this
other
management
engine
and
proxy
that
surfaces
across
those
is
it
kind
of
like
the
one
cluster
in
front
of
many
clusters.
Is
that
kind
of
how
this
ends
up
blank
laying
out
where
you
have
all
these
different
layers
of
proxies
all
the
way
down?
Yes,.
B
A
Cool
and
so
since
it's
writing
two
pods,
that's
the
directly!
That's
the
problem.
You
were
saying
with.
Potentially,
if
you
don't
have,
if
you're,
so
it's
not
using
services
on
the
backend
clusters,
it's
using
pods
like
like
a
normally
ingress,
would
do
right.
So
that's
I,
guess
probably
the
IPS
maybe
aren't
routable
between
clusters.
I
could
see
how
that
would
be
a
tricky.
Well.
I
know
you
brought
some
demos
and
you
know
we're
on
YouTube.
So,
let's,
let's
do
some
cool
visualizations
I'm
gonna
turn
it
over
to
you
on
the
screen
share.
B
A
B
A
B
Yeah,
so
quick,
quick,
I
can
stack
round
of
how
contour
works
and
get
me
talked
about.
You
know
the
architecture
of
it.
How
you
know
contours
the
ingress,
control
or
an
envoy
is
the
data
path
component
to
dig
into
a
little
bit
of
IT
speak
of
how
contour
works
is
you
can
think
of
contour
as
a
translation
layer
between
kubernetes
and
envoy
right
so
services
in
kubernetes
secrets
and
kubernetes
endpoints,
all
those
different
things
get
mapped
through
contour
into
envoy,
isms.
I
guess
you
could
say
so
again.
This
is
contours.
B
Job
is
to
help
take
these
different
objects
and
route
them
into
here.
One
of
the
things
we
came
up
with
hep
do
when
we
started
out
contour
now
p.m.
mark
was
this
problem
of
ingress
today,
you
guys
do
different
slide,
so
ingress
be
one
beta.
One
has
some
some
things.
So
if
I
presented
these
ingress
objects
right,
so
here's
a
bunch
of
llamo
I
know
this
could
be
confusing,
but
essentially
what
these
different
ingress
objects
here
demonstrate
are:
there's
two
different
teams
right.
B
We
have
one
team
here
in
team,
a
and
one
heat
team
here
in
team
B
and
what
they
have
is
they
have
the
same
domain
name.
This
is
project
contoured
at
I/o
in
this
example
and
/blog
right,
so
they
both
use
the
same
host
of
the
same
path,
but
they
target
different
services
in
kubernetes,
so
they
go
to
different
places
and
there
are
also
different
namespaces.
B
So
conceptually
to
a
team
makes
sense
because
you
manage
your
own
ingress
resources,
you're
gonna,
you
know
deal
with
this
yourself,
but
when
the
ingress
controller
goes
to
process
this,
it's
gonna
take
a
sum
of
all
these
objects
and
then
smash
them
together.
It
has
to
come
up
with
one
routing
routing
table
out
of
this,
so
the
problem
here
is
that
I've
got
two
different
domains,
but
this
the
same
domain
with
the
same
hat
for
different
services.
So
what
ship?
What
happens
here
and
really
it's
undefined?
B
It's
up
to
the
ingress
controller,
to
define
how
this
should
happen.
Does
the
last
one
in
wind
is
the
first
one
in
wind:
they
both
fail
and
then
you
have
a
downtime
production.
So
this
is
kind
of
what
we
came
up
with
first
off
was
this
thing
called
ingress
route,
and
then
we
switched
it
to
this
idea
of
we
had
a
new
version
of
that
called
HD
proxy,
and
the
goal
here
is
to
support
these
multi
team
clusters
right.
B
A
B
A
B
Great
yeah,
so
you'll
see
you'll
see
a
lot
of
that
in
that
talk
as
well,
but
here's
the
idea
of
this
delegation
idea
right.
So
what
we
want
to
solve
is
the
problem
of
having
multiple
people
be
able
to
have
conflicts
with
your
ingress
resources.
So
in
this
example,
what
I
have
is
I
have
this?
What
we
call
our
route
domain
right,
so
I
have
our
group
ingress
resource
and
it
has
this
domain
of
project
contour.
That
I
am
when
I
want
to
have
is
have
another
team
or
this
child
resource
own
slash
blog.
B
So
this
path,
slash
blog
and
what
it
does.
Is
we
have
this
idea
of
delegation,
so
if
you
think
of
DNS
in
the
world
right,
I
have
a
DNS
server,
so
she
owns
D
sloka
com.
If
you
create
a
server
and
say
hey,
I'm
gonna
serve
records
for
C
sloka
com,
nothing's
gonna
break
with
my
domain.
Unless
someone
who
has
authority
actually
delegates,
you
know
an
a
record
or
something
off
to
you
to
that.
To
that
server,
and
this
kind
of
follows
the
same
pattern.
B
Nothing
happens
until
the
route
you
know
processes
and
include
and
gives
this
other
resource
the
proper
permissions
to
do
what
they
want
to
do,
or
the
proper
inclusions.
So
in
the
same
example
just
showed
where
we
have
the
same
domain
with
the
same
path.
This
person
doesn't
get.
There's
no
conflict
here,
just
contour
we'll
throw
this
out
because
it's
not
there's
no
authority
through
this
inclusion
model.
B
A
You
know
a
lot
of
this
stuff
in
kubernetes,
unlike
anything
any
I,
say
just
about
every
tool
out
there.
It
a
lot
changes
when
you
need
a
team
to
manage
something.
We
all
get
started
with
these
tools,
especially
container
tools
and
it's
one
person,
and
even
some
of
the
simplest
tools
like
one
of
the
people
in
the
common
things
in
my
courses,
is
that
docker
machine
tool
that
was
has
been
around
for
a
long
time
to
help
you
provision
a
simple
little
darker
machine
people
get
used
to
that
tool.
A
Oh
man,
this
is
a
problem
like
I
had
a
neat
URL
that
worked
for
me
and
then
someone
else
came
in
and
through
on
their
own
route,
and
it
suddenly
is
now
taking
precedence
over
mine
and
not
yet
you
know
people
aren't
used
to
necessarily
studying
every
single
existing
ingress
just
so
they
can
make
a
new
rule.
So
it's
a
great
problem
you're
talking
about,
because
I
don't
have
a
feeling.
The
way
that
people
discover
this
problem
is
by
doing
it
accidentally
in
breaking
stuff.
So.
B
Yeah
sure
cool,
so
what
I
have
here
is
this
is
a
generic
ingress,
our
HTTP
proxy
object.
So
this
is
our
custom
CRD
and,
if
you're
familiar
with
ingress
that
we
showed
on
a
slide
ago,
but
it
looks
very
similar
to
how
ingress
looks
today.
You
know
you
have
this
virtual
host,
Pat
or
speck
here
when
you
define
the
the
fully
qualified
domain
here.
I
have
some
TLS
certificates,
so
we're
provisioning
certificates
through
let's
encrypt
and
then
here
I
have
all
my
routes
right
in
this
case.
This
is
the
soup
generic
route
or
I'm.
B
Just
gonna
say:
hey
everything
goes
to
this
one
back-end.
This
is
called
route
app.
So
what
we
can
do
is
we
can
take
a
look
at
get,
try
and
get
all
the
proxies
in
my
cluster
and
show
that
I
have
just
this
one
here,
let's
go
ahead
and
apply
this
again,
we'll
make
sure
that's
clear,
okay,
so
what
I
can
do
is
if
I
hop
over
here
to
my
url
I
can
go
ahead
and
party
for
this
right
and,
if
I
hit
slash
I,
get
the
default
app
right.
B
Nothing
too
too
dramatic
here
if
I
want
I
can
go
here
and
add
a
second
route,
so
here
we're
gonna,
pull
in
a
second
set
of
routes
here
now,
let's
talk
about
these
conditions
for
a
second,
so
here
I
have
these
routes
right.
So
my
condition
here
was
slash.
That
was
everything
if
you're
defining
one
called
secure.
B
So
what
happens
on
conditions
here
is
conditions
make
up
all
of
the
things
that
need
to
be
met
for
that
path
to
route
properly,
so
here
we're
showing
just
prefixes,
but
what
we
can
do
as
well,
so
we
can
pull
them
things
like
headers,
so
I
could
say
you
know
slash
routes
to
this
route
app,
but
then
slash
secure,
you
know
maybe
must
contain
the
user
agent
of
Firefox.
So
maybe
our
secure
site
is
only
gonna,
allow
the
Firefox
browser.
B
So
if
I
go
ahead
and
apply
this
one,
what
I
have
now
is
like
two
two
routes,
so
I've
got
my
slash
and
my
slash
secure.
So
if
I
hop
over
here
to
Chrome,
everyone
still
works.
Fine,
if
I
hit
slash
secure
here
in
my
browser,
you'll
see
I,
just
look
at
the
default
app
site
and
we
get
this
secure
site
and
that's
because
of
those
conditions
right
we're
looking
for
the
some
of
the
slash
secure
as
long
it
along
with
this.
Some
of
it
has
to
contain
Firefox.
B
If
I
pull
up
my
firefox
browser,
which
I
have
here
if
I
hop
over
to
here,
you
see
the
default
site
works
like
it
/
secure,
then
I
do
now
give
a
secure
site
right
and
that's
because
again
we're
meeting
all
of
the
criteria
of
having
the
path
slash
secure
as
well
as
this,
the
user
agent.
So
this
model
kind
of
flows
through
so
I
could
have
as
many
headers
as
I'd
like
and
we're
leave
left
this
open
for
other
types
of
routing.
B
So
that's
just
the
generic
like
hello
world,
so
I
guess
now.
What
I
want
to
do
is
go
ahead
and
look
at
like
let's
go
at
a
team.
Let's
do
some
delegation
things
real
quick.
So
what
we
gonna
do
is
go
create
a
marketing
team.
This
is
the
end.
This
is
my
generic
example.
So
I'll
go
ahead
and
create
the
controlled,
create
EFT,
marketing
and
we'll
do
a
one.
Ok
name
stays
for
our
team
and
they'll
go
ahead
and
deploy
a
sample
app
for
them.
B
Ok,
so
I
created
a
sample
blog
site.
What
I
want
to
do
is
give
them
again
that
/blog
example
that
we
did
so
I'll
click
that
and
I'll
click.
One
more
thing
will
create
the
Parsee
for
them
all
right.
So
now,
if
I
get
a
head,
that's
a
little
bit
if
I
get
my
proxies
again,
what
you'll
see
is
I've
got
two
of
them
like
so
you
had.
B
We
had
this
route
before,
but
now
I
have
it
to
have
this
marketing
one,
and
you
can
see
the
status
is
orphan
right
and
that's
because
no
one
has
a
delegation
path
to
them,
so
we
haven't
included
that
one
yet
into
the
into
the
mix.
So
this
is
the
example
of
someone
creating
some
sort
of
resource
contours.
Seeing
that
is
saying:
hey
you
don't
have
permissions
quite
yet,
so
what
we
can
do
is
we
can
go
and
enable
that
and
we'll
do
that
too.
B
This
idea
to
include
it's
this
one
here,
so
an
included
in
contour
works
like
you
would
in
a
programming
language.
So
if
you
had
say
C++-
and
you
include
a
header
file
when
you
go
to
compile
that
the
compiler
is
going
to
go
inject-
that
header
file
first
into
your
code
and
then
compile
the
rest
of
the
application
and
that's
kind
of
how
these
includes
work.
What
contour
does
is
it
looks
to
these
includes
and
then
adds
them
to
the
top
of
the
of
the
ingress
kind
of
tree,
and
then
it
then
process
it
through.
B
So
here
we
go
so
now.
This
is
again.
This
is
our
route
here
and
we're
gonna
delegate
off
from
to
the
blog
site
in
the
marketing
namespace,
which
we
just
created
and
here's
the
name
blog
site
is
the
name
and
then
spaces
marketing,
I'm
gonna,
give
them
slash
blog.
So
if
I
go
ahead
and
apply
my
route
again
so
now
we
just
edit
that,
and
now
we've
created
a
the
link.
So
if
I
go
ahead
and
get
my
proxies,
you'll
see
now
that
they're
both
valid,
although
I,
have
a
bad
service
here.
B
A
But
what
you're
doing
here
is
you've
added
this
a
couple
of
custom
resources
types
here
right
so
that
way,
instead
of
having
to
do
a
bunch
of
annotations
and
putting
all
this
stuff
and
just
key
values
that
aren't
yamo
formatted
properly,
and
all
that
your
because
you're
extending
this
api
of
kubernetes
you're
getting
this
rich
spec
interface.
That
is
unique
to
the
invoice.
Contour
scenario
right,
isn't
that
house
right.
B
B
So
what
we
can
do
now
is
we
can
kind
of
define
it
more
and
we
can
have
it.
You
know
be
richer
for
what
we
need
to
do
so.
Yeah,
there's
a
lot
more
here
than
we're
actually
showing
like.
If
you
wanted
to
look
at
adding
headers
to
requests
like
request,
headers
or
response
headers
managing
the
the
load
balancing
algorithm
you
apply
to
when
these
request
comes
through.
All
that
is
in
this,
this
CR
T
that
we've
defined.
B
A
A
Of
used
to
configurations
just
spitting
back
that
we
did
not
do
that
right
and
that's
one
of
the
challenges
with
with
annotations
is
that
it
you
basically
have
to
do
it
hope
it's
right
and
then
go
check
logs
to
see
if
there's
anything
complaining
in
the
app
right
so
and
I
think
some
people
that
are
new
to
this
stuff.
It's
not
as
if
we're
extending
the
existing
ingress
features
in
kubernetes
we're
basically
in
this
case
we're
just
replacing
them
with
a
more
feature-rich
version.
That
does
exactly
the
same
thing.
A
Only
more
it's
not
really
disadvantage
to
it.
It's
just
it's
an
extra
step.
You
need
to
do
on
your
your
cluster
by
adding
in
this
custom
resource
definition,
and
we
kind
of
we
did
we
don't
you
know
all
of
the
different
CR
DS.
You
can
look
at
out
there.
They
always
provide
you
all
that
information
so
that
you
don't
you
know
that
command.
You
need
to
apply
to
basically
import
it
and
change
that
stuff
in
your
cluster.
So
it's
not
terribly
complex
to
get
set
up.
A
B
Absolutely
and
you
get
you
get
better
type
safety
as
well,
so
we
can.
We
can
do
some
things
like
the
status
stuff,
that
I'm
doing
I
can
get
better
feedback
I'm
having
a
minute
of
it.
You
see
are
these,
but
you
have
to
buy
into
it.
You
know
you've
got
to
apply
this
year
to
you
and
use
it
in
stuffs.
All.
A
B
A
B
Yeah
for
sure
cool
something
I
got
this
fixed
up
on
that
sure,
okay,
what
was
that
but
cool?
So
now
we
have
is
we
have
two
different
proxies
and
so
the
route
now
delegate
permissions
off
or
the
path?
Slash
blog
once
you
that
here
here
this
flash
blog,
if
delegated
off
to
this
marketing
team
right,
so
I
look
at
the
marketing
team
what
they
actually
go
and
create
its.
This
is
all
they
have
right
now
right.
So
this
is
again
namespace.
B
This
is
called
blog
site
and
all
they're
defining
is
to
hey
I'm
gonna
route
to
dubbed
up
that
blog
and
theory
like
we
don't
have
it
here,
but
because
we
didn't
define
it,
there's
an
inherent
or
an
idea
of.
If
the
slash
is
there
right,
you
don't
define
it.
It
just
uses
slash
so,
but
result
here.
What's
gonna
happen
again,
like
I
said
it's
an
include,
so
it's
like
a
include
on
a
programming
language.
B
So
what's
gonna
happen
here
is
this
is
gonna
at
being
slashed
blog
is
what
gets
passed
here
to
dapat
right
because
of
that
inclusion
model.
So
if
I
go
ahead
back
to
my
terminal
here
in
the
browser,
I
go
to
slash
blog.
What
I'll
get
now
is
I,
get
the
blog
site
right
and
that's
the
again.
That's
that
Pat
and
what
we're
seeing
is
because
we
have
that
inclusion
model.
You
know
this
team
can
then
go
and
manage
that
path.
They
own
that
slash
path.
Now
that
slash
blog,
they
can
do
better
doing
with
it.
B
They
can
make
their
own
delegations
that
can
pass
it
off.
So
it's
kind
of
a
one-time
setup
of
permissions
and
then
you
just
can
do
what
they
want
with
it
right,
yeah.
So
the
only
cool
thing
I
guess
we
could
touch
on
together
when
it
killed
too
much
time
is
there's
ability
to
have
multiple
services
here
in
contour
as
well.
So
with
this
with
ingress
today,
you
can't
have
multiple
pub
streams,
so
you
can
do
things
now.
I
could
have
like
the
service
too.
B
You
know,
I
can
route
between
across
many
services
and
I
could
even
do
waiting
and
it's
like.
If
you
wait,
you
know
80
and
then
you
know
you
know
20
or
something
and
the
weight
based
allocations
across
multiple
up
streams.
That
sort
of
thing-
and
if
you
put
your
gimbal
hat
back
on,
if
you
think
of
these
as
clusters
and
not
different
services,
then
you
can,
you
know,
weight
based
routing
across
multiple
clusters
or
that
sort
of
thing.
Yeah.
B
A
Yeah,
that's
cool
I
definitely
would
like
to
update
people
on
the
progress,
because
I
know
that
over
the
last
year
at
the
coop
cons
there's
been,
you
know,
discussions
and
lot
of
stuff
offline
off.
Obviously,
on
this
v2
that
I've
I
talked
a
little
bit
about
I.
Think
in
the
recent
update
to
my
courses,
but
I
didn't
have
any
hard
evidence
to
show
anybody
other
than
here's
a
video
from
kook
on
a
year
ago
that
you
know.
B
Yeah
for
sure
yeah,
so
it's
definitely
a
thing
it
still.
It's
still
some
ways
off
I
think
from
becoming
ready
to
start
using
the
contract.
She
has
support
for
it,
so
we're
going
to
start
using
contour
as
a
testbed.
So
once
we
start
iterating
through
those
those
different
designs,
you'll
be
able
to
actually
play
with
them
for
real
in
contour.
If
folks
learn,
don't
do
that
yeah.
A
B
Yeah
absolutely
yeah,
so
contour
is
you
know
it
supports
all
the
same:
the
ingress
d1
they
don't
want
API
for
sure
yeah,
so
you
can.
If
you'd
like
this
is
just
another
take
on
it.
So
Connor
has
multiple
versions,
so
it
processes,
B,
1,
beta,
1,
ingress,
the
processes.
Linker
sprouts.
You
know
this
HTTP
proxy.
If
you
have
now
and
then
also
the
new
service
API
switch
is
the
ingress
v2
and
the
v1
will
come
out
here
and
I
think
and
to
119
what
its
gonna
get
released.
B
Yeah,
so
here's
how
you
could
do
yeah,
so
you
could
say
you
could
you
know
spin
up
your
your
version,
your
new
version
and
give
it
zero
percent,
and
once
you
have
it
zero,
then
you
could,
you
know,
swap
these
weights
around
so
maybe
give
it.
These
are
just
arbitrary
numbers.
These
weights
arrives,
I
usually
make
them
add
up
to
100,
just
cuz
in
my
head.
That
makes
sense,
but
you
could
do
you
know
77
and
5
that
make
sense
to
you
yeah.
So
you
could
get
the
weights
back
and
forth.
Okay.
B
Have
precise
not
percentage,
yeah,
8005,
yeah
yeah,
and
you
can
do
that
yeah
and
you
can
swap
it
that
way.
You
can
also
do
some
things
where
you
I
know
some
folks
like
to
do
this
with
look.
If
you
create
a
new
app,
you
put
it
in
a
different
name
space
and
some
folks
will
do
like
a
Bluegreen
deployment
right
where
you
just
switch
traffic
all
at
once,
and
you
can
do
that
through
inclusion
models.
B
Sometimes
so
maybe
I
have
a
marketing,
you
know
new
new
version
or
something
and
then
I
can
instantly
swap
from
one
name
stays
to
another
from
the
root
and
then
manage
how
traffic
goes
from
A
to
B.
So
that's
what
the
inclusion
or
you
can
do
it
just
with
with
the
weights
we
talked
about
here.
It's
all
please
and.
A
B
A
B
So
there's
we
do
have
refined
dashboards
that
we
work
so
I
could
do
a
quick
peek
of
that.
So
what
I
did
here?
I
guess
I
did
too
quick.
I
just
did
I
port
forward
it
to
my
refinery
and
then
what
we
can
see
here
is
now
on
the
screen.
I
have
sort
things
so
I
can
look
at
this
is
this?
Is
the
envoy
dashboard?
So
this
stuff
here
is
showing
you
the
open
connections,
so
we
actually
have
a
new
thing
called
a
shutdown
manager.
B
We
add
it
to
envoy
I
mean
this,
helps
you
roll
out
a
change
to
envoy.
So
if
you
had
to
like
roll
out
a
new
version
of
envoy
or
reallocate
resources
in
your
cluster,
what
you
don't
want
to
have
is
as
traffic's
coming
to
your
cluster,
you
don't
want
to
drop
connections.
So
what
this
does
is
lets
you
shutdown
cleanly
by
by
waiting
for
connections
to
drain
out.
So
this
is
the
way
that
you
visualize,
you
know
what
are
open
nob.
So
there's
nothing
in
that
cluster.
Right
now,
yeah.
B
You
can
see
all
the
different
envoy
connections
in
terms
of
the
200
sponses
300.
That
sort
of
thing
we
have
some
metrics
on
our
proxy
resources
like
how
many
are
existing,
how
many
updates
contour
is
doing
to
the
different
Envoy
listeners
clusters
and
route
services
I
believe
there
was
some
metrics
as
well:
we've
had
yeah
there's
an
API
server
stat,
so
how
many?
And
we
found
some
things
where
we
have
really
busy
clusters
in
kubernetes
and
that
can
cause
some
issues
with
contour,
because
it's
got
to
turn
through
lots
of
different
different
objects.
B
A
Right
in
case
you
don't
have
those
metrics
somewhere
else
already
yeah
I
mean
the
thing
you
mentioned
about
service
training
I
feel
like
we
could
do
a
whole
podcast
on
that.
Like
me,
I
talk
about
it
a
lot
when
I
start
talking
about
day
two
operations,
because
you
know
a
lot
of
there's
a
lot
of
talk
about
out
there
around.
You
know
how
to
properly
update
services
and
other
types
of
resources,
and
we
talked
about
backups.
A
Let
me
talk
about
maintenance
and
all
that
sort
of
stuff
of
rolling
upgrades
all
that
stuff,
but
when
it
starts
to
into
application
actually
getting
an
application
game
replaced
by
another
application,
and
that
do
you
need
to
be
so
nuanced
and
so
detailed
and
exactly
how
your
app
is
shutting
down
and
how
its
monitoring
connections
and
how
so
many
apps
like
Python
and
nodejs,
don't
track
those
by
default.
So
you
can't
just
necessarily
depend
on
your
app
to
not
just
terminate
every
single
upload.
A
We
have
people
that
are
you
know,
clients
that
are
complaining
around
our
maintance
time,
but
I
thought
we
weren't
supposed
to
have
a
maintenance
time
anymore
and
so
I
think
I
feel
like
having
more
visibility
into
that
is
such
a
huge
thing
and
I'm
glad
you
mentioned
that
because
I
for
those
of
you
watching
cannot
be
understated.
Please
test
a
lot
during
your
rolling
updates,
whether
you're
gonna
be
rolling
updates.
A
Blue
greens,
canary
whatever
make
make
sure
you're
doing
all
the
things
that
users
would
be
doing
on
your
platform,
especially
if
they're
doing
things
like
uploading
or
downloading
large
large
things,
because
I
think
that's
what
really
gets
companies
in
the
trouble
when
it's
a
static
website.
No
one
notices,
because
you
just
hit
refresh
and
the
things
back
up,
but
there's
so
much
settled
subtlety
to
those
connections,
shutdowns
that
I
think
a
lot
of
people
glaze
over
that
thinking
that
the
load
balancer
is
going
to
somehow
magically
protect
them
from
it.
Let.
A
B
A
Just
like
you're
just
sitting
there
going
come
on
people
close
your
browser
go
away.
What
this
new
version
finish
like
and
go
home,
yeah
yeah
cuz,
usually,
if
you're
the
operator
that
day
you're
not
going
home
until
the
you
know,
all
the
connections
are
gone,
and
so
maybe
maybe
not
hat.
Don't
have
these
six-hour
wait
states
in
your
HTTP
connections,
where
you're
just
right,
you're
palling
after
six
hours,
like
maybe
don't
do
that?
Maybe
have
them
30
minutes!
A
20
minutes
that
way,
you
can
actually
have
these
short
deployments
because
yeah
these
diplomas
can
get
really
these
updates,
but
in
particular,
can
get
really
long
is
if
your
hat,
if
you
have
these
long
connections
and
WebSockets
and
stuff
like
that,
well
I
kind
of
threw
that
curveball
with
the
with
the
visualization
and
it
didn't
realize
that
you
were
gonna,
have
all
that
stuff
ready
to
go.
So
what
else
you
got?
You
got
anything
else
in
there
that
you
haven't
shown
off
I.
A
B
B
Unless
you
really
really
want
that
things
like
that,
we
talked
about
those
annotations
right,
there's
a
common
one.
That
says,
when
I
get
a
request
to
a
plain,
HTTP
endpoint,
redirect
to
a
secure.
You
know
HTTPS
website,
but
in
contour,
when
you
specify
a
certificate
on
a
domain
name
or
a
V
host.
Well
I'm
actually
do
that,
for
you
cuz,
we
assume
you're
gonna
want
that
right.
So
why
imitate
all
of
your
objects-
and
maybe
this
one
I-
have
that
redirect
yeah.
A
A
A
Like
you
know,
none
of
us
we're
doing
this
stuff
five
years
ago,
like
this
is
all
new
stuff,
and
it's
always
overwhelming
for
people
to
add
these
new
layers
of
tech
to
their
stack
and
keeping
the
yam
well
short
and
keeping
the
defaults
saying
is
I,
think
always
a
good
thing
and
having
what
is
it
doctor
used
to
say
batteries
included,
but
swappable
I
think
is
what
they
used
to
call
it.
I
always
I,
always
like
that
term,
so
very
cool.
So
how
can
people
get
involved?
So
we've
got
the
I've
got
over
here.
A
A
You
talked
a
little
bit
about
the
operators,
though,
and
how
is
how
exactly
for
those
that
aren't
familiar
with
necessarily
how
these
operators
and
CR
DS
and
all
these
things
play
in
the
part
here,
because
we
have
had
a
great
discussion
on
this
show
about
operators,
what's
what's
the
difference
between
adding
the
CR
D
so
that
you
have
this
ingress
route,
HTTP
proxy
resource
and
the
operator?
What's
what's
the
difference
there?
Yes,.
A
B
So
generically
what
contour
is
it's
a
controller
right
so
that
it's
a
controller,
that's
watching
for
a
set
of
resources
and
then
taking
action
on
them
when
I
think
of
an
operator
I?
Think
of
it
having
production
knowledge
about
deploying
new
things
so,
like
a
few
years
ago,
I
helped
to
help
initiate
the
first
elastic
search
operator,
what's
be
tied
to
it
a
little
bit
about
before
it.
So
this
job
was
to
have
CDs
but
then
create
new
resources
based
on
what
you
defined
in
those
in
those
yamo
and
what
it
did
was.
B
It
was
able
to
look
at
the
cluster
and
take
action.
So
again,
it's
opinionated
that
it's
gonna
help
manage.
You
know
a
complex
application,
so
a
database
or
something
like
that
which
has
more
operation
when
I
was
around
it.
But
generically
act
controllers
are
just
but
they're
exactly
the
same
thing
right.
They
still
have
these
watch
loops
and
control
patrol
loops.
That
you're
implementing
is
just
more
of
in
my
head.
How
is
it
applying
that
to
a
cluster
okay,
Sam.
A
Okay,
yeah
cuz
I
mean
when
you
get
all
the
fancy
this
out
of
the
way
and
we
try
to
define
the
operator
framework
and
we
talk
about
operator
hub
and
all
these
things
the
simplest
way
I've
been
able
to
come
up
with
it
and
I'm
probably
wrong
in
some
cases
on
this.
But
a
CR
D
to
me
tends
to
mean
that
you're
using
the
standard,
cube
control
command
line
to
apply
Y,
Amal
and
you're
just
using
a
different
type
of
resource.
A
So
your
Y
Amal
kind,
the
little
thing
at
the
top
says
kind
is
different,
and
then
you
get
this
different
spec,
but
once
it
becomes
an
operator
in
my
experience
it
tends
to
be
an
extra
optional
command-line,
whether
it's
in
addition
to
the
cube,
control,
command
line
or
maybe
exposes
itself
in
a
web
portal,
or
maybe
some
third
party
additional
tooling,
but
it
seems
to
abstract
away
a
lot
of
that
automation
and
intelligence
that
you
don't
use
stuff
no
longer
have
to
know
into
something
else.
Is
that
would
you
agree
with
that?
I
think.
A
B
A
It's
interesting
to
see
some
of
these
new
ideas
about
what
an
operator
might
do
and
sort
of
The
Grudge
work
of
the
day
to
day.
You
know
how
do
I,
how
do
I
do
a
backup
of
this
server
and
make
sure
that
backups
constantly
being
done
and
makes
sure
that
it
doesn't
backup,
but
I
don't
have
to
set
up
like
all
these
extra
systems
just
to
do
a
simple
backup
and
be
notified
if
it
failed
right.
So
that's
really
the
kind
of
exciting
stuff
I'm
seeing
there.
A
A
B
We
just
done
it
so
really.
What
gimbal
does
is
it
actually
didn't
explain
this?
What
it
does
is,
it
goes
and
replicates
all
the
services
and
end
points
in
each
upstream
cluster.
So
it
looks
it
so
it'll
take
every
namespace
to
say
you
have
a
marketing
namespace,
it'll
it'll,
take
all
the
endpoints
in
that
one
and
copy
it
to
the
marketing
space
of
your
gimbal
routing
cluster.
So
now
you
have
a
copy
of
all
those
services
and
endpoints,
so
essentially
we're
using
kubernetes
as
our
own
data
store.
All
right.
B
We
don't
have
to
have
our
own
database
or
something
we're
just
using
xyg
right
to
do
that.
So
once
we
have
all
those
in
the
gimbal
cluster
now
you
can
just
create
your
proxy
resources
and
that
gimbal
routing
cluster
and
then
the
same
endpoints
show
up
there.
So
as
you
change
applications
on
the
backend
clusters,
you
know
gimbal
Wommack.
They
synchronize
all
the
different
services
and
endpoints
to
make
them
show
up
in
that
routing
cluster.
So
you
would
define
all
of
your
ingress
resources
in
that
routing
cluster,
not
in
the
upstream.
Ok,.
B
B
Route
that
way
and
then
you're
off
to
the
races
and
you
have
the
same
different
inclusion
model.
So
you
apply
all
that
now,
you
can
have
lots
of
teams
be
self-sufficient
and
you
can
also
bridge
other
world
so
like
if
you
had,
in
this
case
OpenStack
or
something
that's
not
containerized,
where
you
have
different
endpoints
looking
out
somewhere
else
like
I
said,
but
I
have
a
vSphere
discover
that
I
have
to
go
merge
yet.
B
A
All
right
we've
got
some
questions
in
chat.
Some,
let's
see,
can
you
contrast
to
traffic
plus
I
use
that
in
my
local
development
work
with
swarm
and
it's
very
lightweight,
almost
exactly
the
same
laptop
experiences,
production,
Enterprise,
layer,
7,
yada,
yada,
ok,
so
I,
don't
know
how
much
Steve
knows.
Traffic
and
I
don't
know
contour.
So,
okay,
let
me
just
talk.
I
set
up
top
of
a
head,
so
contour
is
there
any
less
encrypt
functionality
in
there.
B
B
A
B
Well,
they'll
get
written
to
I,
guess
I
can
saying
there's
too
many
as
you
can
get
start,
so
you
can.
Actually
you
can
imitate
with
an
ingress
object,
say,
hey,
go
get
me
a
cert
that
way
yeah
or
you
can
create
a
certain
manager,
C
or
D
object,
type
and
say:
hey
I
want
to
start
with
these
parameters
and
then
certain
tangibles
process
out
the
same
way
which
eliminates
having
to
have
that
ingress
resource.
Okay,.
A
B
A
B
So
you
can
have
you
can
scale
it
out
as
much
as
you'd
like.
So,
if
you
need
more
more
Network
losing
words
now
more
network
capacity,
you
had
more
on
voice
to
the
to
your
to
your
fleet,
sure,
and
then
you
can
have
as
many
on
contours
as
you'd
like
as
well.
So
each
contour
will
respond
like
I
said
on
what
connects
to
contour
it
for
its
configuration.
B
A
B
B
So
by
default
we
deploy
contour
as
a
deployment
I'm
going
to
play
envoy
as
a
daemon
side.
Okay,
so
you
get
one
per
node
envoy
kind
of
with
CPUs
and
and
disk
I/o,
so
it
doesn't
help
to
put
multiple
envoys
in
the
machine.
You
don't
gain
a
lot
from
that,
so
by
using
a
dataset,
we
can
just
have
one
per
four
node.
But
again
it's
up
to
you.
There's
no
requirement
for
on
that.
You
could
have
them
all.
B
A
In
the
model,
so
now
you're
you're
scaling
back
on
the
number
of
contours.
So
it
almost
sounds
to
me
like
your
invoice.
Your
invoice
scaling
is
based
around
your
load
of
of
client
connections
and
maybe
your
your
contour
scaling
is
in
terms
of
scaling
for
throughput
or
whatever.
It's
scaling
more
on
just
the
change
rate
and
whether
or
not
you
want
fault
tolerance
is
that
yeah.
A
A
It
doesn't
actually
need
to
skate
like
it.
In
other
words,
if
you're
suddenly
getting
millions
of
hits,
it's
not
like
contour
needs
to
scale
it's
the
invoice.
That's
really
doing
all
that
work
right,
because
we're
really
just
talking
about
the
controller
acting
on
changes
in
the
API.
So
if
you're
not
doing
a
lot
of
API
changes-
and
maybe
you
can
get
away
with
just
two
two
contours-
I'm
curious
for
that
leader
election-
is
that
doing
raft
something
or
is
that
I'm
different
I?
Think.
B
A
A
All
right,
let
me
see
if
there's
any
other
questions
here
running
the
complete
kubernetes
contour,
a
gimbal
stack
on
a
laptop,
so
developer
experience
is
the
same
as
production.
Is
this
practical
or
just
heavy
biker?
My
opinion
is
that's
heavy.
If
you
have
to
have
communities
to
develop,
you
might
be
doing
it
wrong,
but
Steve.
What
do
you
think
so.
B
Yeah
I
go
back
and
forth
on
some
of
that,
depending
on
you
know
how
much
stuff
you've
got
running.
Actually
in
the
project
contour
IA
website,
there's
a
there's,
a
blog
post.
I
wrote
how
to
use
kind
so
kind
as
kubernetes
and
docker.
So
it's
a
way
to
spin
up
kubernetes
with
just
the
docker
client.
So
it's
a
neat
way
to
get
quick
clusters.
So
I
show
you
how
you
can
do
that
easily
on
your
machine.
B
So
if
you
want
to
deploy
a
contour
or
traffic
or
whatever
ingress
control
you'd
like
I
should
have,
you
should
go
to
swap
in
contour
for
that
yeah.
So
it
guys
depends
on
you.
Whatever
makes
sense
for
you.
You
know
it's
I
used
to
run
a
lot
of
stuff
locally
and
then
at
some
point
it
got
heavy
and
we
deployed
out
to
share
it
infrastructure.
But
then
there's
there's
pain
in
that
because
when
it
breaks
you
can't
just
destroy
it
as
easily
as.
A
A
A
It
does
I
for
one
know,
no
one
that
wants
the
development
kubernetes.
There
are
some
people
that
maybe
think
they
have
two
or
they're
told
to,
but
oh
my
goodness
like
who
needs
another
tool:
that's
gonna,
slow,
their
CPU
and
eat
their
battery
and
and
the
reality
is
as
a
KU
Bernays
was
never
designed
to
do
that,
like
none
of
the
original
scope
of
it
was
designed
to
be
a
development
platform
on
your
local
machine.
I
know,
there's
lots
of
tools
that
do
it.
A
A
I'll
just
say:
hey,
look
what's
wrong
with
docker
compose
like
that,
keeps
us
minimal
or
increasingly
the
the
shops
that
I
work
with
they're
Jun,
less
they're,
maybe
like
a
ruby
shop
or
maybe
some
shops
where
they
just
don't
have
an
easy
way
to
manage
the
state
of
their
at
local,
app
development,
so
they're,
if
they're,
switching
between
different
apps
and
they're,
constant
that
worried
about
like
Ruby
versions
and
version
management
stuff
like
that
is
a
thing
I.
What
I've
been
finding,
though
lately
is
a
lot
of
shops.
A
I'm
working
with,
they
really
are
focused
on
developer
happiness
and
those
developers
are
not
happy
if
they
have
to
develop
in
a
container
all
day.
So,
even
though
there
are
using
containers,
what
they
are
using
them
for
is
only
for
spinning
up
all
the
backend
services
that
I'll
run
on
like
localhost
and
duck
or
desktop,
or
something,
and
that
can
be
on
kubernetes
or
run
compose
or
run
swarm
or
whatever
you
want,
and
then
it
just
lets
them.
A
They
let
them
be
on
their
native
OS,
because
their
apps
are
designed
for
that
and
all
their
advanced
tooling,
and
everything
is
happier
that
way.
I
keep
thinking
that
we're
gonna
get
to
a
way
where
we
have
this
sort
of
sweet
spot.
Where
all
of
these
you
know
all
the
the
dependency
injectors
and
all
that
stuff
will
all
be
container
friendly
for
local
development,
but
there's
still
a
there's
a
whole
there's
a
reason.
A
B
Yeah
I
know
a
few
years
ago,
I
wrote
a
tool
and
what
it
was
to
help
arc.
We
had
at
the
same
time,
word
hit
like
too
many
folks
and
stuff
that
didn't
have
all
the
tooling
to
spin
up
actually
compile
the
code.
So
we
wrote
a
tool
that
would
actually
deploy
feature
branches
to
namespaces
and
kubernetes
I.
Think.
A
B
A
A
A
A
And
I
think
I
think
that's
it
for
the
questions.
I
think
it's
all
just
good
comments
about
the
pros
and
cons
of
all
this
stuff
of
compl,
a
complex
local
development
or
simple
to
development,
and
just
saying
hey,
look
it's
hopefully
CI
is
gonna
work
it
out.
If
we
have
any
cross
architecture
problems,
hopefully
CI
will
help
us
or
maybe
just
running
the
local
tests.
I
think
I'm
also
a
fan
of
seeing
doing
as
much
as
possible
so
that
you're,
all
your
CI
tests
can
also
be
run
locally.
B
A
A
B
A
They
not
too
distracted,
but
I'm
for
those
of
you
still
watching,
I
will
be
putting
out
some
information.
Ducker
actually
published
their
very
first
road
map
here
in
2020.
So
now
they
have
this
open-source
roadmap
for
all
their
tooling
on
github,
and
you
can
you
to
can
request
new
features
of
their
of
their
tooling
and
a
several
of
those
are
around
shrinking
that
shrinking.
The
binary
is
getting
a
lot
of
the
tooling
out
of
there.
A
That
seems
to
be
the
easiest
way
other
than
just
kind
and
running
Cooper
neighs,
without
burning
down
your
your
battery
in
a
couple
of
hours,
so
we're
kind
of
hoping
that
it'll
get
better
for
that.
But
if
you're
on
lepto,
if
you're
on
linux-
and
you
ain't
gonna,
worry
about
right,
it's
all
day-
cool
well,
Steve,
thanks
so
much
for
being
on
the
show.
We
have
some
people
talking
in
chat,
but
nobody
has
asked
any
further
questions,
so
we're
gonna
wrap
it
up
at
our
hour.
A
What
is
what
okay,
so
people
should
get
involved
by
going
to
the
github
repos
right,
I,
think
I
pointed
those
out.
There's
the
project,
contour
repo,
there's
actually
a
whole
organization
for
that
right.
So
we've
got
you
know:
we've
got
the
the
contour
in
here.
We
got
gimble
in
here,
so
people
can
look
at
the
stuff
download
the
interest,
f--
file
issues,
you're,
probably
gonna-
be
involved
with
that
somewhere.
A
A
A
Cool
well
thanks
again
for
being
on
the
show
I
look
forward
to
hearing.
Unfortunately,
I
won't
be
seeing
you
in
coop
con
in
a
couple
of
weeks,
but
I
look
forward
to
the
updates
of
these
projects
and
getting
getting
into
them
myself
and
learning
a
lot
about
them,
and
hopefully
we'll
have
you
back
on
the
show,
with
the
next
major
release
and
all
a
lot
of
new
stuff.
Maybe
when
we
do
ingress
v2
we,
when
that
comes
out,
we
can
see
how
contour
works
with
that.
So
yeah
sounds
good
yeah
thanks
again,
all
right.
Everyone.
A
Next
week
on,
this
show
on
Thursdays
I'm
live
almost
every
week,
so
come
back
here.
Otherwise
also
this
show
is
now
supported
on
patreon.
I
probably
should
have
mentioned
that
earlier.
So,
if
you
want
to
keep
this
show
going
and
all
the
other
free
activities
of
three
things
I
put
out
there
like
the
newsletter
and
the
github
's
and
all
that
stuff,
please
help
me
on
patreon.
All
of
your
patronage
is
a
really
greatly
appreciated.
So
we
will
see
you
next
week
back
here
on
you,
life.