►
From YouTube: TGI Kubernetes 016: Heptio Contour ingress controller
Description
Come hang out with Joe Beda as he does a bit of hands on exploration of Kubernetes and related topics. Some of this will be Joe talking about the things he knows well. Some of this will be Joe exploring something new with the audience. Ask questions, comment and help decide where things go.
This week we will be exploring a new project from Heptio: Contour! This is a new Kubernetes Ingress controller based on Envoy written by Dave Cheney.
Links:
* Contour: https://github.com/heptio/contour
* Envoy: https://www.envoyproxy.io/
* Envoy v2 API: https://github.com/envoyproxy/data-plane-api
* Kris and Justin's Cloud Native Infrastructure book: https://www.amazon.com/Cloud-Native-Infrastructure-Applications-Environment/dp/1491984309
* Weaveworks Flux: https://github.com/weaveworks/flux. Haven't played with it yet but it looks interesting.
* Zalando ALB controller: https://github.com/zalando-incubator/kube-ingress-aws-controller
* Glowforge: https://glowforge.com/
* My paperjs program for the slinky: https://j.hept.io/k8s-slinky
A
Hello
and
welcome
to
CGI
kubernetes.
My
name
is
Joe
Beda,
and
this
is
a
bit
of
an
odd
addition.
Here
we
go,
can
y'all,
hear
me
fine,
I'm
working
from
home
today,
because
it's
actually
a
holiday
here
in
the
US,
it's
Veterans
Day
and
I
tweeted
out
that
I
was
gonna,
do
TGI,
K
and
then
I
realized.
It
was
Veterans,
Day
and
I
figure.
I'll!
A
Do
it
anyway,
so
a
little
bit
of
a
different
setup
here
in
my
home
office,
so
I
just
want
to
make
sure
it
sounds
good,
alright,
cool
nice
to
see
you,
sir,
and
Richard
and
I
you
know,
and
because
of
daylight
savings
time.
A
lot
of
folks
who
aren't
on
daylight
savings
time
got
a
little
confused
on
the
time
here.
A
So
just
for
the
record
I'm,
not
a
fan
of
tail
light
savings
time.
We
actually
did
this
thing
where,
where
we
did
this,
we
have
like
one
big
ammo
file
for
all
the
cigs
and
kubernetes,
and
we
had
the
meeting
times
in
there
and
those
were
expressed
in
UTC,
because
that
seemed
like
really
nice
and
fair
to
actually
sort
of
like
do
it
in
a
way.
A
That's
not
super
time
zone
specific,
but
the
reality
is
that
all
the
sig
meetings
crack
daylight
savings
time,
and
so
now
all
those
things
are
wrong,
and
so
last
week
I
had
to
go
and
modify
that.
So
you
could
specify
a
time
zone
nice
to
see
you
wallet
on
hell,
I'm
sure,
I'm
pronouncing
your
name
or
angel
on
how
Jose
good
to
see
y'all
so
yeah.
So
today
our
plan
is
is
to
go
through
contour,
which
is
this
project
that
hefty
oh
yeah
developers
don't
have
holidays.
Yes,
you
should
have
holidays.
A
Everybody
should
take
some
time
off
when
they
need
it.
That's
a
that's
part
of
being
making
this
industry
work
for
you
for
drink,
oh
good,
to
see
again
so
yeah.
So
today
we're
gonna
go
through
a
project.
This
is
a
hefty
a
project
called
contour,
which
is
an
ingress
controller
based
on
the
Envoy
proxy,
which
is
a
a
CNC
F
project.
A
It's
a
it's
a
load,
balancing
proxy
HTTP
proxy
that
has
been
proven
out
at
lift
the
sort
of
car
sharing
company
and
they've
been
using
it
for
a
while,
and
it's
a
really
interesting
design
that
lends
itself
really
well
towards
kubernetes,
which
is
one
of
the
reasons
why
we
built
an
ingress
controller.
Now
this
particular
okay,
yes,
oh
hi,
Chris,
how's
it
going
so
so
Chris
Nova
is,
is
just
gonna
go
and
climb
for
three
days,
and
so
right
after
this
I
think
she's
gonna
go
climb.
A
A
If
he's
gonna
join
us
today
or
not
josè,
once
you
go
to
New
Zealand,
you
should
go
to
New
Zealand,
Chris,
yeah
and
then
and
then
we're
gonna
talk
about
the
different
ways
to
deploy
something
like
like
contour
and
along
the
way.
We're
gonna
cover
some
of
the
ins
and
outs
and
sort
of
like
where
contour
is,
is
a
project
and
some
of
the
stuff
that's
going
to
be
coming.
A
Burn
nice
to
see
ya,
know
new
aw
nice
to
see
you
also
so,
and
then
I
have
a
new
toy
here
in
my
home
office
I
have
this
thing
called
a
glow
Forge,
which
is
a
cloud
controlled
laser
cutter
for
doing
all
sorts
of
fun
stuff,
and
so
it's
made
by
this
company
here
in
Seattle,
a
friend
of
mine,
runs
it
and
they've
been
working
on
it.
For
two
years
and
yeah,
and
so
I
want
to
give
a
demo
of
that,
so
we
got
a
lot
to
cover
here.
A
So
one
of
the
things
that
I
like
to
do
when
I
start
out
is
is
go
over
some
of
like
what's
going
on
in
you
know
around
the
kubernetes
world,
and
so
I
don't
have
much
today.
But
the
one
thing
I
do
want
to
talk
about
in
folks.
We're
talking
about
it
in
the
chat
beforehand
is,
is
Justin
garrison
and
Chris.
Nova
two
folks
were
very
active
in
the
urban
India
community
wrote
this
book.
That's
now
out
called
cloud
native
infrastructure
and
I
hadn't
ordered
it
and
I
just
ordered
it
just
now.
A
So
you
can
see
you
purchased
this
item
on
today.
So
there
we
go
so
I
I
just
bought
the
book.
I
haven't,
read
it
yet,
but
I'm
really
excited
about
it.
And,
oh,
you
were
just
talking.
Okay,
so
well,
it
said
he
was
talking
about
it
at
a
local
doctor
meetup,
which
is
great
I
know
you
know
Chris
and
I.
You
know
sat
down
and
she
interviewed
me
a
little
bit
forward.
So
I'm
sure
I
show
up
someplace
there,
hey
Elric,
how's
it
going
so
I
haven't,
read
it
yet,
but
I'm
sure
it's
awesome.
A
A
Is
there
anything
else,
that's
going
on
hey
Sam.
Is
there
anything
else
that's
going
on
around
the
kubernetes
world
before
we
move
on
to
sort
of
the
the
main
event
here
around
contour
that
that
folks
want
to
ask
questions
about
or
talk
about
a
little
bit
I
usually
want
to.
Like
you
know,
if
there's
something
I
forgot
something
interesting
going
on,
it's
always
fun
to
be
able
to
bring
that
stuff
up.
I,
think
things
are
a
little
bit
quiet.
You
know,
there's
one
other
project
that
I
think
is
interesting.
A
A
So
Chris
are
you
seeing
so
so
different
people
have
different
delays
on
this
thing
and
I
set
youtube
to
try
and
do
a
short
delay,
but,
depending
on
how
you're,
connected
or
sort
of
like
you
know,
what
kind
of
buffering
you
have
going
on,
you
may
see
different
amounts
of
delay,
so
I
apologize.
If
there's
a
lot
of
delay
there
that
I
it's
one
of
the
downsides
of
doing
the
the
broadcasting
like
this
and
I.
Think
all
the
platforms
pretty
much
including
like
twitch,
have
have
a
similar
type
of
delight.
A
So
that's
unfortunate,
so
I
will
just
talk
on
monologue,
but
you
guys
can
feel
free
to
comment.
It'll
just
be
a
little
delayed,
but
if
flutter
looks
really
interesting
because
it's
a
way
of
checking
your
configurations
in
to
get
and
then
having
having
something,
that's
monitoring
it
and
and
automatically
taking,
what's
checked
in
and
then
pushing
it
and
there's
some
other
cool
features
that
are
going
on
there
too.
So
this
is
something
that's
really
interesting
and
and
one
of
the
things
that
we've
been
looking
at
a
lot
at
hep
Tio's.
A
How
do
you
define
your
application?
Configuration
we've
been
working
on
this
project
called
case
on
it.
We
have
some
really
interesting
things
in
store
there
and
I'm
really
interested
to
see,
if
maybe
that
will
work
well
with
with
something
like
flux
or
a
similar
system
for
essentially
taking
an
application
definition
and
then
and
then
go
ahead
and
deploying
that
my
dogs
here
he's
barking.
My
my
son
is
having
a
friend
come
over,
so
that's
gonna
be
exciting:
okay,
Diego
from
Vancouver
nice
to
see
ya
you're,
at
least
in
the
same
time
zone.
A
A
Dave
Chaney,
who
is,
is
you
know
pretty
well
known
in
the
go
community
started
this
when
he
started
working
for
us
at
hefty
Oh
and
has
made
some
amazing
progress
with
it.
It's
still
a
very
early
project,
the
biggest
feature
and
I'll
just
get
this
out
there
right
now
that
it's
missing
is
it
doesn't
do
TLS
termination
and
so
most
of
the
time
when
people
are
running
ingress,
you're
gonna
run
it
through
with
something
like
certificate
manager
or
cube.
A
Leggo
that'll
work
with,
let's
encrypt
or
or
what-have-you
to
be
able
to
provision
certificates
automatically
that
that's
not
supported
yet
in
contour.
It's
something
that
we're
working
on
and
the
the
TLS
support
in
envoy
itself
is
evolving
to
actually
support
this
use
case
a
lot
better,
so
so
envoy
right
now
doesn't
support
s
and
I,
which
is
the
technology
that
protocol
needed
so
that
you
can
support
multiple
sites,
sort
of
virtual
sites
on
a
single
IP
address.
So
that's
not
supported
yet.
A
A
So
actually,
let
me
back
up
a
little
bit.
I
did
do
an
episode
of
TGI
K
on
ingress,
but
it
would
it
be
worthwhile
for
people
I'm
gonna,
give
a
little
bit
of
a
recap
on
what
ingress
is
I.
Don't
have
my
document
camera
so
I'm
not
going
to
draw
a
fancy
diagrams,
like
I,
did
in
the
previous
episode,
so
go
check
that
out.
A
Under
these
circumstances,
I
want
you
to
load
balanced
it
across
these
things
on
the
backend
and,
as
you
add,
more
and
more
sites
to
an
engine,
X
config.
What
you
find
is
that
config
file
or
that
config
file
tree
gets
bigger
and
bigger.
Ingress
is
a
way
to
describe
that
configuration
in
a
relatively
generic
way
across
a
bunch
of
load
balancers
and
then
collect
all
that
stuff
together,
so
that
you
can
have
a
bunch
of
different
pieces
that
go
into
that
configuration
all
consolidated
for
you.
A
A
So
first
of
all,
here's
the
the
Envoy
website,
one
of
the
interesting
things-
and
this
is
not
a
reason
to
use
a
technology,
but
it's
kind
of
nice
to
know
that
that
there's,
a
certain
community
behind
it
envoy,
is
now
a
project
that
is
part
of
the
cloud
native
computing
foundation.
And
so
what
this
means
is
that
there
is
long
term
governance
of
the
project.
It's
something
that
that
you
know
the
Linux
Foundation
essentially
is
going
to
be
a
steward
of
this
project
for
for
quite
a
while.
A
A
A
The
other
interesting
thing
about
envoy
is
that
if
I
were
to
actually
think
about-
and
you
said,
an
API
driven
load
balancer,
what
I
would
expect
is
that
you
could
actually
call
the
into
the
load
balance
or
be
an
API
and
poke
configuration
into
it.
So
and
hey
I
want
you
to
reconfigure
this
and
that
it
turns
out
that
Envoy
works.
A
The
other
way
you
configure
envoy
with
an
API,
and
then
it
calls
out
to
that
control,
server
to
actually
say
hey
what
configuration
should
I
be
using,
and
the
interesting
thing
here
is
that
the
relationship
between
envoy
and
the
controller-
that's
actually
configuring.
It
is
very,
very
similar
in
concept
to
the
relationship
between
the
cubelet
in
kubernetes
and
the
api
server,
so
the
cubelet
calls
them
to
the
api
server
saying
what
should
I
be
running?
A
Similarly,
Envoy
calls
into
the
on
void,
controller
and
there's
there's
actually
a
set
of
api's
that
can
be
split
across
servers,
but
calls
in
to
the
to
the
the
the
control
plane
server
for
envoy
and
says:
hey
how
should
I
be
configured
and
so
that
model
again
is
a
really
good
fit
for
the
kubernetes
world,
so
Sam
st.
does
it
use
a
gr,
PC
gateway
for
rest
for
the
v2
stuff,
I'm,
not
sure
I'm,
not
sure
if
it's
natively
supports
rest
or
whether
it's
using
the
the
GR
PC
gateway?
That's
a
good
question.
A
I
do
know
the
work
that
Dave's
working
on
he's
made
some
good
progress
with
it
using
these
v2
API
he's
just
using
the
native
gr
PC
stuff.
So
that's
that's
where
he's
at,
and
so
what
you'll
see
is
that
and
if
we
go
down
here,
here's
the
list
of
api's
that
are
being
driven
in
in
v2
and
there's
some
really
interesting
stuff.
Here.
There's
there's
you
know
what
are
your
set
of
backends?
What
are
the
endpoints
for
those
things?
Are
they
healthy
or
not?
A
It's
a
way
to
essentially
sort
of
if
your
health
checking
you
can
take
those
health
checking
results
and
broadcast
that
rate,
limiting
metrics
all
sorts
of
really
interesting
stuff
going
on
here.
I'm
not
gonna,
go
into
all
the
details
of
this,
but
I
think
you
know.
For
me
this
is
the
thing
that
really
makes
envoy
special.
It's
also
written
in
C++.
It's
fast.
It's
really
fast
anytime,
you're
dealing
with
packets
and
connections.
You
definitely
want
to
make
sure
that
things
are
fast
all
right.
A
He
said
he
does
have
the
G
RPC
support
working
in
a
branch
he
just
didn't
want
to
merge
it
before
I
went
ahead
and
did
t
GI
case,
so
he
didn't
want
to
push
new
stuff
right
before
we
went
ahead
and
did
this
let's
see
and
the
authors
are
amazing,
so
the
authors
have
contour
well
Dave's
amazing,
also
and
then
the
the
Envoy
okay
more
stuff
about
I'm,
going
to
sing
more
on
bug
phrases.
Matt
Klein,
who
works
at
LIF,
is
really
great.
A
The
team
at
lifts
are
doing
that
Google's
putting-
and
this
is
for
for
both
sto
in
places
where
Google's
using
envoy
internally
are
putting
a
ton
of
resources
into
envoy
also,
and
so
it's
getting
a
lot
of
performance
and
extensibility
and
hardening
from
a
security
point
of
view
that
that
is
is
really
gonna
benefit
everybody
and
so
a
lot
of
exciting
stuff
going
on
with
envoy
there
all
right.
So
let's
go
ahead
and
get
started
here.
So
this
my
instruction
for
getting
started
now,
I
have
I
have
a
plain
Jane
cluster.
A
That
is
one
of
our
AWS
QuickStart
clusters
that
I'm
going
to
applying
this
to
so.
The
first
thing
we
tell
people
to
do
here
is
a
cute
cute
control
applied
a
chef,
and
then
we
point
through
this
redirector
here
to
actually
get
a
yamo
file
that
you're
going
to
download
now
I'm,
not
proud
of
this.
We
really
wanted
to
make
this
be
as
easy
as
possible.
A
The
redirector
that
we're
using
it
hep
co
doesn't
support
TLS
out
the
gate,
so
this
should
have-
and
this
is
something
I'm
probably
gonna
fix
next
week
there
should
be
an
S
here.
We
really
want
to
do
that,
and
even
so
my
advice
to
people
is
before
running
something
like
this.
You
want
to
dig
in
and
really
understand.
What's
going
on,
most
people
aren't
patient
enough
to
do
that.
A
We
do
give
you
a
little
bit
of
a
rundown
of
what's
gonna
happen
here,
but
let's
go
ahead
and
download
this,
because
that's
a
best
practice
before
we're
gonna
go
ahead
and
actually
run
it.
So
did
you
consider
building
contour
on
top
of
traffic
or
was
a
pretty
easy
decision
to
target
on,
but
given
the
community,
you
know,
honestly,
we
didn't
look
at
traffic
very
closely
in
terms
of
this
stuff.
A
I
think
the
some
of
our
plans
and
where
we
really
want
to
sort
of
take
contour
over
time
is
to
be
able
to
go
and
handle
very,
very
large
sort
of
company-wide,
horizontally,
scalable,
ingress,
installments,
right
and
so
and
and
in
talking
to
Matt
and
the
folk
that
lift.
This
is
one
of
those
places
where
envoys
already
being
used
in
that
capacity
in
in
some
companies
and
so
and
lifted.
You
know
specifically,
and
so
I
hadn't
heard
the
same
thing
about
traffic.
A
So
just
you
know,
looking
at
the
api's
looking
at
the
community
envoy
really
seemed
like
a
slam.
Dunk
for
us
out
of
the
gate
and
and
and
traffic
already
has
an
ingress
controller
that
that
works,
but
but
in
terms
of
sort
of
where
we
want
to
take
contour
over
time,
I
think
I
think
on
boy
ends
up
being
a
being
a
great
fit
for
us.
Okay,
so
let's
go
through
and
I'm
gonna
do
make
der
contour
and
curl
dash
L
we're
gonna.
Do
that!
Let's
see
it's
an
O
that
I'd
do
that
there
we
go.
A
And
then
we'll
look
at
this
and
we'll
see
what's
going
on
here,
alright,
so
there
we
go
now.
This
I
wrote
this
stuff
here
so
I
know
exactly
what's
going
on
here,
and
one
of
the
things
that
we're
gonna
be
doing
probably
over
the
next
month,
is
the
configuration
for
generating
this
is
going
to
be
driven
by
case
on
it.
We're
gonna
actually
start
bringing
bringing
stuff
together
here.
A
What
this
is
is
it's
a
concatenation
of
a
bunch
of
files
that
are
in
the
deployment
directory
here,
so
we're
looking
at
at
this
one
here.
So
it's
all.
These
files
can
cat
it
together,
but
what
we're
gonna
do
is
we're
gonna
create
a
hefty
Oh
contour
namespace,
we're
gonna,
create
a
service
account
for
this
thing
and
and
then
for
the
service
account.
We're
gonna
give
it
permissions
for
to
be
able
to
to
read
config
Maps
and
points
nodes,
pods
in
secrets
right
and
the
secrets
are
so
yeah
Alex.
A
My
keyboard
is
the
is
the
the
new
MacBook
Pros
with
the
shallow
Keys,
but
I
really
like
to
pound
on
the
keyboard.
So
these
things
end
up
being
really
loud.
So
secrets
ends
up
being
so
that
you
can,
you
can
read
TLS
secrets,
it's
probably
not
necessary
for
contour,
where
it
is
now
out,
but
as
we
start
supporting
TLS
in
this
context,
it's
going
to
be
necessary.
A
The
fact
that
most
ingress
controllers,
end
up
being
able
to
read
all
secrets
across
the
entire
cluster
is
not
something
that
people
are
happy
with
and
I
touched
on
this
on
the
previous,
the
previous
episode
on
ingress,
but
it
is
sort
of
where
things
are
right
now
and
so
we're
looking
at
services
and
then
ingress
is.
We
also
need
to
be
able
to
do
right
now.
A
A
When
that
happens,
this
thing
is
going
to
have
to
have
the
ability
to
also
update
the
the
ingress
is
to
actually
good
status
in
there
and
then,
in
this
particular
case,
we're
running
ingress,
we're
running
contour
and
there's
there's
two
containers
in
an
init
container
going
on
here,
so
the
two
containers
are
Envoy
itself
and
and
then
contour,
and
so
what
we
do
is
we
set
up
contour
to
actually
run
as
a
sidecar
right
next
to
on
voice.
So
for
every
envoy
you
have
a
version
of
content.
A
A
You
know
relatively
small
daemon,
so
so
the
overhead
of
running
one
on
a
sort
of
buddy-buddy
situation
is
really
not
that
great,
and
then
we
have
this
init
container
here,
and
this
does
nothing
more
than
writes
out
sort
of
a
bootstrap
JSON
file
for
configuring
Envoy
zoom
in
on
that
yam,
okay,
is
it
too
small?
Here
we
go
we'll
do
that
and
we'll
do
that.
Is
that
easier
to
read
for
y'all.
A
So
so
the
way
when
this,
when
this
pod
starts
up
the
init
container,
runs,
it
writes
a
simple
configuration
file
into
an
empty
durval
um--
and
then
both
envoy
and
contour
start
Conroy's
contours,
gonna
use
the
service
account
to
actually
get
all
of
its
stuff
from
the
kubernetes
world,
and
then
it's
gonna
translate
and
serve
that
to
envoy
/
localhost
locally.
So
so
that's
what
we
have
going
on
there
so
envoy
is
HTTP
to
aware
I,
don't
think
I've
tested
it
yet
with
contour.
One
of
the
things
that's
necessary
really
for
HTTP.
A
Two
to
really
work
is,
is:
is
TLS
those
things
end
up
being
being
pretty
tied,
and
so,
as
we
get
TLS
supported
with
contour
I
think
it
you
know,
it'll,
you
know
I
think
we're
gonna
get.
You
know,
h2
support
for
free
the
interesting
thing
about
employ.
Is
it
supports
h2
both
for
incoming
connect?
You
know
for
upstream
and
downstream
so
downstream.
A
Also
so
so
small
small
digression
digression
here,
sto
does
also
support
ingress
using
envoy,
and
so
there's
a
lot
of
overlap
there,
conceptually,
but
most
of
the
time
when
people
are
using
SEO
comes
along
with
a
whole
host
of
other
concepts,
and
one
of
our
goals
with
contour
was
to
really
have
something
that
was
simple
and
purpose-built
for
ingress.
Essentially
what
people
called
north-south
traffic,
whereas
as
sto
is
taking
on
a
much
much
more
complex
problem
in
general.
A
A
A
A
An
alternate
way
of
doing
this
is
that
you
can
have
multiple
cube
configs.
Then
you
can
set
environment
variables
to
actually
set
which
one
is
live.
This
will
help
you
actually
manage
the
environment
variable
for
being
able
to
set
which
cube
config
like
you
want
to
actually
use.
So
now
you
need
a
cue
control.
Get
nodes.
That'll
show
me
what's
going
on
here,
it's
a
180
to
cluster
and
then,
let's
do
cube
control
applied
this
thing
here,
so
we're
gonna
create
a
whole
bunch
of
stuff.
That's
all
that
we
expected
to
happen
is
happening
there.
A
Let's
see
yeah
and
then
let
me
go
through
what's
going
on
there,
so
we're
running
the
deployment
with
two
of
these
things,
but
you
generally
don't
want
these
things
to
land
on
the
same
node,
and
so
we
have
here,
a
affinity
stands
in
the
amel
here,
which
is
essentially-
and
this
is
this-
is
way
harder
than
it
should
be.
But
essentially
what
I'm,
trying
to
say
is
is
try
not
to
run
to
multiple
contours.
A
On
the
same
note
right,
it's
saying,
based
on
hostname,
make
sure
that
we
have
anti
affinity
so
have
these
things
sort
of
like
you
know,
try
not
to
actually
do
that.
You
know
for
any
other
pod
that
has
the
app
equals
contour
label
on
it.
That's
essentially
what
we're
saying
here
would
be
interesting
if
you
have
time
for
you
to
CPU
limit
that
deployment,
2.4
million
want
to
smash
other
running
production
service.
A
That's
a
really
good
point,
I
think,
and
it's
something
that
I
haven't
been
very
diligent
about
up
until
now,
if
you're
running
kubernetes
in
a
production
environment,
you
really
want
to
make
sure
that
you
set
limits
on
everything.
In
fact,
I
think
it
might
be
a
good
tgia
tik
episode
to
really
go
through
how
limits
work
and
what
we
can
do
to
make
sure
that
people
are
using
them
appropriately.
A
Looking
at
both
limits
in
quota
and
in
what
it
looks
like
when
say
you
run
out
of
RAM,
we
can
sort
of
play
with
things
and
and
and
the
killer
and
stuff
like
that.
So
it's
a
really
good
point
Richard
and
it's
something
that
I
think
it's
probably
worth
its
own
tea
gik
episode
to
really
go
into
that
I'm,
not
gonna,
try
and
do
that
live
because
I
haven't
tested
it
out
yet
but
yeah.
So
this
is
up
and
running
and
we
should
have
these
two
running
on
two
separate
nodes.
A
The
other
method
that
we
have
and
we
have
yellow
provided,
is
that
you
can
run
this
as
a
daemon
set
and
what
that
essentially
says
is
I
want
to
run
contour
on
every
node
or
at
least
every
node
that
matches
a
particular
label.
So
you
could
have
say
a
pool
of
nodes
and
you
could
say
I
want
to
run
contour
on
every
node
in
that
pool.
It's
it's
for
large
production
clusters.
A
It's
a
lot
of
times
what
people
do
is
they'll
have
us
medicated
set
of
nodes
just
for
running
ingress
and
incoming
load,
balancing
that's
really
different
from
everything
else.
Richard
says
yeah.
So
having
dedicated
in
four
nodes
for
different
purposes,
is
it
can
be
good
practice,
so
try
running
spring
booth
class
pass
scanning
sort
of
hundred
seconds
of
time,
yeah
so
yeah.
This
is
one
of
those
things
that
you'll,
see
and
I
saw
this
running.
A
Java
apps
on
Borg
is
that
if
you
don't
start
putting
limits
on
things,
you
know,
Java
can
take
a
long
time
to
actually
get
itself
up
and
running,
which
also
is
a
good
argument
for
readiness
probes
right
because
sometimes,
even
though
the
process
is
running
doesn't
mean
that
it's
ready
to
actually
receive
traffic.
Okay,
so
let's
go
see
cute
control,
namespace,
hefty
Oh
contour
get
services,
Oh
why'd.
We
go
ahead
and
let's
open
this
puppy
up
and
it's
not
ready
yet.
Oh,
let's
go
back
and
hit
refresh
until
it
shows
up.
A
A
A
A
A
This
is
gonna,
create
a
deployment
with
three
replicas
running
quirky.
We're
gonna
set
up
a
service
for
this,
which
is
targeting
port
8080
from
port
80,
and
then
we're
gonna
set
up
an
ingress
saying
anything
that
comes
in,
send
it
to
set
it
to
this
thing,
and
so,
let's
go
ahead.
Keep
control
f,
apply,
yeah
on
tour,
for
example;
okay,
so
we're
creating
the
deployment
the
service
and
the
ingress.
A
This
is
all
happening
in
the
default
namespace,
though
that
should
come
up
pretty
quick,
hopefully,
and
so
now,
if
we
go
back
here,
yay,
okay,
we're
up
and
running
that's
great.
One
of
the
things
that
we
see
here
is
we
see.
X-Forwarded-For
is
this.
Is
the
original
IP
address
I
just
gave
away
my
home
IP
address
I?
A
That
might
have
been
a
bad
idea
who
knows
and
then-
and
then
this
is
the
this
is
the
the
address
that
envoy
saw
the
thing
coming
from.
So
this
is
is
essentially
I
believe
this
might
be
the
the
the
AWS
ELB
address
that
you're
actually
seeing
here,
yeah
Chris,
yeah
I'm,
not
gonna,
talk
about
that
Chris
I'm,
not
gonna,
I'm,
not
gonna!
Call
you
out
there
so
okay,
so
we
got
envoy
up
and
running
here
with
you
know,
configured
through
contour
everything's
working
great.
A
So
that's
it
honestly.
Now,
as
we
get
more
features
into
contour,
including
TLS,
I,
think
it's
going
to
be
worth
another
demo.
What
I
want
to
do
now
is
I
want
to
do
an
advanced
configuration,
an
advanced
deployment
of
on
using
AWS
NLB,
and
this
is
something
that
I
think
I
talked
about
a
little
bit
in
terms
of
at
the
at
the
previous
ingress
episode.
I've
had
a
chance
to
play
around
with
that
and
I
wrote
up
some
instructions
on
how
to
do
that.
A
Let's
see
is
enabling
HSTs
and
ingress
available.
I
think
you
know
a
lot
of
times.
What
happens
is
different.
Ingress
controllers
will
allow
you
to
use
annotations
to
describe
sort
of
whether
you
want
to
enable
certain
features
like
like
HSTs.
That's
something
that
we're
going
to
be
looking
at
with
contours,
in
terms
of
which
annotations
do
we
want
to
support?
How
do
we
want
to
actually
enable
certain
options
like
that?
A
So
you
know,
as
we
start
doing
this,
that's
that
we're
gonna
be
looking
for
people
with
feedback
in
terms
of
what
what
features
we
go
ahead
and
add
there,
and
then
also
you
know
if
you
want
to
get
involved,
it's
a
it's
a
really
nice
codebase,
because
you
know
Dave,
wouldn't
do
anything
else
and
and
so
feel
free
to
submit
some
PRS
there.
We're
definitely
would
love
to
see
that
okay,
so
what
I'm
gonna
do
now
is
I'm
gonna,
actually
tear
down
my
envoy,
my
contour,
I'm
do-kyu
control.
A
A
A
A
A
Think
the
one
like
and
you
know,
look
I'd
search
for
this
before
solando.
Is
this
German
company?
It's
a
German
retailer!
Well,
thank
you
for,
for
you
know,
live
port
scanning.
My
home
IP
address
Chris
I,
appreciate
that
so
so
zalando
is
this
the
this
German
retailer
and
they
wrote
an
Al
B
controller
that
they
had
talked
about.
I.
A
Think
a
Q
Khan
Europe
a
little
while
ago,
I
think
there
is
a
PR
Cooper
needs
to
add
NLB,
but
I,
don't
know
if
there's
a
PR
in
progress
to
add
alb
Chris
might
know
off
the
top
of
her
head,
I'm,
not
sure
quick
sort
of
like
so
early
on.
There
was
ELB
elastic
load
balancers,
it's
now
called
ELB
classic
and
it
could
run
in
either
TCP
mode
or
HTTP
mode.
Both
of
those
things
were
not
super
configurable.
You
couldn't
actually
sort
of
do
different
routes
in
the
HTTP
mode.
A
When
you
did
the
TCP
mode,
it
would
still
terminate
connections
so
that
you
wouldn't
actually
see
the
the
originating
IP
address
coming
in
over
time.
Amazon
is.
It
has
actually
introduced
two
new
types
of
ELB
ELB,
which
is
really
sort
of
a
l7
load
balancer,
something
that
is
appear
to
say
something
like
envoy
and
then
they've
introduced
MLB
network
load
balancer,
which
is
a
tcp
type
of
thing
that
doesn't
actually
terminate
the
connection.
So
it
does
some
fancy
natty
tricks
versus
versus
doing
the
terminating
the
connection.
A
Okay,
so
there
is
one
then
four,
four
four
al
B's
I
haven't
actually
gone
ahead
and
play
with
that.
Yes
could
I
implement
a
Bitcoin
miner
in
the
next
version
you
know
I
probably
should
or
likes
to
do
something
that
takes
a
lot
of
CPU
so
that
we
can
maybe
I'll
do
that
in
preparation
for
the
for
the
for
the
resource
limits,
episode
of
TGI,
K,
okay,
so
so
there's
a
longer
thing
is
actually
I
know
that
that's
something
these
these
guys
are
using
this
stuff
in
anger.
A
So
that's
super
interesting,
but
apparently
there's
there's
other
PRS
out
to
actually
support
a
lbs.
I
haven't
been
tracking
that
stuff
super
carefully
all
right,
so,
let's
just
all
contour
with
NL
bees.
Now,
like
I,
said
there
is
a
PR
out
to
actually
support
NL
bees
natively,
so
you
can
do
type
equals
load,
balancer
and
then
probably
an
annotation
saying
use
an
NL
be
versus
an
e
lb
classic.
That's
probably
where
we're
gonna
actually
show
up.
Oh
look:
there's
a
there's
a
bug
there.
A
A
A
A
So
if
I
go
repo
deployment,
we
have
this
thing
called
DS
host
net.
Okay,
so
we
have
the
namespace
our
back.
This
is
the
stuff
that
we
all
had
before.
These
are
actually
soft
once
within
this
directory,
and
then
we're
going
to
do
a
daemon
set
and
we're
gonna.
Not
only
are
we
gonna
do
a
daemon
set
we're
going
to
a
daemon
sets
that
uses
host
networking,
and
so
what
we're
essentially
saying
is
that
this
is
a
special
container
that
I
don't
want
to
contain.
A
The
networking
I
wanted
to
use
the
host
networking
directly,
usually
like,
if
you're
gonna,
lock
down
your
cluster
at
all.
This
is
something
that
you
need
a
little
bit
more
permissions
to
be
able
to
do
so,
and
typically
this
will
be
most
appropriate
when
you're
running
it
on
a
dedicated
set
of
nodes.
A
It's
a
daemon
set
so
it'll
run
in
every
node
and
it'll,
listen
directly
on
the
it'll,
listen
directly
on
port
80
80s,
where
envoys
can
actually
be
be
running,
so
we're
gonna
go
ahead
and
set
that
up.
So
we're
gonna,
deploy
that
if
you,
if
you
point
cube
control,
applied
a
chef
at
a
directory,
it
essentially
installs
all
the
animals
and
memo
files
in
that
directory.
A
So
that's
up
and
rain.
That's
running
right
now,
but
we're
not
gonna
be
able
to
reach
that
because
it
turns
out
that
that
it's
it's
like
I,
said
running
directly
on
the
node.
There
is
no
service
there.
There
is
no
cluster
IPS,
there's
nothing
right.
It's
just
just
running
on
the
node
I
could
SSH
into
the
node
and
curl.
Oh,
maybe
we'll
do
that.
Well,
the
NLP
is
actually
spinning
up
now
we
have
to
configure
the
NLB
by
hand.
A
So
let's
go
ahead
and
do
that
so
the
first
thing
we're
gonna
do
and
again
this
won't
be
necessary,
as
we
I
was
already
there.
This
won't
be
necessary
as
NLB
support
gets
merged
into
kubernetes.
Hopefully,
that'll
come
in
the
next
version,
but
now
you
can
actually
see
oh
look.
I
can
do
the
classic
load
balancer!
That's
what
we
had
before.
There's
a
lb
and
n
lb.
Let's
go
ahead.
We're
gonna
create
an
NLB.
Let's
give
it
a
name.
Let
me
look
at
my
where's,
my
instructions
here.
A
A
Let's
see
this
V
PC
and
the
contour
V
PC,
okay,
so
us
West
to
a
you,
have
to
essentially
allocate
an
IP
address
from
a
and
I'll
be
per
per
availability
zone.
That
you're
running
in
so
essentially
say
I'm
gonna,
listen,
port
80
in
u.s.
West
to
a
you,
go
ahead.
We're
gonna,
configure
routing
here
and
again:
I
have
like
screenshots
everything
here
that
we're
going
to
have
for
routing
what
we're
gonna
do
did
I
click
on
this
is
Amazon
thinking.
A
What's
going
on
guys,
oh
at
least
one
subnet
must
be
specified.
I
thought
I,
oh
I
want
to
I
want
to.
Actually,
let's
see
what
did
I
say
I
wanted
to
do.
You
want
to
select
the
public
subnet
okay,
so
logically,
and
then
you
can
assign
a
elastic
IP.
If
you
have
one
to
this,
which
is
kind
of
interesting,
you
can
now
have
elastic
ip's
play
double
duty
between
your
load,
balancer
and
and
your
your
nodes,
which
is
kind
of
cool.
Actually,
you
can
do
that
before
you
could
do
that
before.
Okay.
A
So
when
you
create
a
new
target
group,
we're
gonna
call
this
contour.
Also
TCP
is
gonna,
be
talking
to
port
8080.
So
this
is
essentially
the
upstream
servers
for
NLB
and
we're
gonna
point
that
at
target
type
is
instances
protocol
TCP,
you
could
do
advanced
health
checking
we're
just
gonna.
Do
the
same
old,
same
old
air.
Let
me
make
sure
that
that's
essentially
what
we
want.
A
Yes,
so
now
right
now,
you
could
register
the
targets
explicitly
where
I
could
go
through
and
click
these
things
or
we
could
go
through
and
I'll
show
you
how
to
do
this.
We
could
actually
we're
using
a
auto
scaling
group.
We
could
actually
configure
the
auto
scaling
group
to
go
ahead
and
populate
the
the
target
group
for
us,
so
we're
not
going
to
set
any
targets
right
now.
We're
gonna
go
ahead
and
review
this
and
we're
good
to
go
and
I'm
going
to
hit
create
there,
but
we're
not
done
so.
A
Let's
go
to
the
target
group
here
and
I'm
going
to
go
to
targets
here,
edits
and
then
wait.
No,
no!
No!
Actually
I
got
confused.
What
you
do
is
you
go
to
the
you
go
to
the
auto
scaling
group
for
this
and
and
you
edit,
the
auto
scaling
group
and
I
said:
hey
I
want
you
to
actually
go
off
and
no
not
that
one
I
want
you
to
go
off
in
in
populate
contour.
The
target
group.
So
now
what
happens
is
is
Amazon
will
take
care
of
registering
all
the
VMS.
A
Packets
are
coming
directly
into
your
instances.
You
have
to
set
up
your
security
groups
for
those
instances
to
be
able
to
accept
traffic
essentially
from
the
internet.
This
is
really
kind
of
scary
now,
because
I'm
running-
and
this
is
the
way
that
our
QuickStart
does
it,
because
I'm
running
our
nodes
on
a
private
subnet.
This
isn't
that
dangerous
traffic
will
only
come
in
through
this
port
through
the
load
balancer.
A
But
if
you
are
running
on
a
public
subnet
and
your
nodes
had
public
IP
addresses
this
would
essentially
open
these
things
up
for
direct
access
outside
of
your
load
balancer,
which
means
that
there's
a
certain
amount
of
sort
of
steering
attacks
for
traffic,
which
is
a
security
concern
and
it's
easier
to
DDoS,
because
instead
of
DDoS
in
the
whole,
you
know
do
s
in
the
whole
pool.
You
can
actually
pick
out
and
kill
one
note
at
a
time
type
of
thing,
so
it's
a
little
bit
more
dangerous
from
from
a
security
point
of
view.
A
If
you
don't
configure
this
stuff
correctly
so
now
we're
going
to
go
into
the
security
group
here
and
the
right
I
believe
we
set
up
this
security
group
here
and
right
now.
This
is
what
the
security
good
for
all
machines
in
the
cluster
I
believe
yeah.
This
is
the
one
that
we
want
yeah,
because
it's
not
the
API
server.
Okay,
so
we're
gonna
go
ahead
and
edit
that
we're
gonna
add
a
rule
and
we're
gonna
be
saying:
custom,
TCP,
8080
coming
from
everywhere
and,
let's
say,
contour
there,
alright!
A
A
A
It
has
targets,
but
these
are
an
initial.
The
health
checks
haven't
actually
completed.
Okay,
so
it
takes
a
little
while
for
the
health
checks
to
actually
to
actually
go.
Chris
is
live.
Demos
make
me
nervous,
Oh
Chris,
oh
my
poor,
sweet
summer
child.
You
have
not
seen
previous
TGI
case
where
things
have
totally
gone
off
the
rails.
So
you
know
there
is
that
we're
always
only
that
it's
working
okay,
we're
always
flying
without
a
net
here.
A
Not
only
that,
but
NLB
has
some
really
nice
economics,
especially
when
you're
doing
development,
there's
zero
cost
for
running
and
NLB
load
balancer.
If
there's
no
traffic
going
through
it,
which
is
actually
really
really
cool
on
ELB,
they
charge
you
for
the
hour
whether
or
not
there's
traffic
going
through
it
and
I'll,
be
it's
only
the
data
that's
actually
coming
through
in
connections
that
are
coming
through
that
charge.
So
so,
especially
for
demos,
the
the
you
know
or
low
traffic
services,
the
the
cost
structure
of
NL
B's
are
really
really
attractive,
see
ya!
A
That's
pretty
cool
that
actually
works.
It's
gonna
be
even
easier
when
you
can
set
up
and
alby's
the
kubernetes
natively,
because
then
we'll
just
be
able
to
create
a
service
like
we
did
before
and
and
point
it
at
these
things.
So
that's
really
cool
alb
is
really
cool
too,
but
alb
is
is
a
whole
nother
level,
where
you're
essentially
having
all
the
stuff
that
we
might
be
doing
with
Envoy
you're,
actually
going
ahead
and
doing
with
with
Amazon
directly.
A
So
is
it
possible
to
keep
a
long,
colon
collection,
I,
don't
know
about
alb,
but
with
NLP
the
network
load
balancer
that
we
did
here,
I
believe
as
long
as
there's
a
certain
amount
of
traffic
going
on
it
actually
won't
cut
the
connection.
So
as
long
as
the
the
connection
is
not
totally
idle,
you'll
actually
be
good,
which
is
I
think
in
some
ways
more
attractive
than
ELB
classic.
Where
you
know
you,
you
know,
there's
there's
a
limit
to
how
long
you
can
actually
hold
on
to
a
connection
so
yeah
alright.
A
So
that
is
a
quick
run-through
for
tour,
some
of
the
ways
that
you
deploy
it.
What
makes
envoy
so
special
and
why
we
picked
it
and
then
a
little
bit
of
a
demo
in
advance
configuration
with
NLB
yeah,
NL,
balb
e
lb,
classic
it's
very
confusing
Mateus
I!
Don't
blame
you
for
getting
confused
on
that
stuff.
Now,
I
want
to
show
you
my
cool
toy,
my
laser,
my
laser
cutter.
A
So
what
I
did
here
is
I'm
gonna
make
a
kubernetes
linky
out
of
wood,
which
is
totally
crazy
and
I'm
gonna
show
you
I'm
gonna
grab
the
camera
here
and
I'm
going
to
move
it
around,
and
hopefully
I
won't
unplug
the
camera
as
I.
Do
this
and
let
me
go
ahead
and
I'm
gonna
switch
to
the
main
camera.
While
I
do
this,
you
can
see
my
messy
office
here
and
I'm
gonna
set
up
this
camera
waiting
at
this
thing.
A
Okay,
so,
first
time
over
you,
the
things
pretty
big
that
the
piece
of
wood
that
I
have
in
there
is
is
I'm.
Gonna
talk,
Imperial,
I'm,
sorry
guys,
and
you
may
not
be
able
to
still
hear
me
yeah.
So
the
piece
of
wood
that
I
have
in
there
is
eighth
inch
plywood.
It's
it's
birch
plywood!
It
is
about
20
inches
by
12
inches,
which
I
believe
is
about
500
by
300,
millimeters
and
so
I'm
gonna
go
ahead.
A
I'm
gonna
set
up
my
camera,
pointing
at
this
thing
and
make
sure
that
it
actually
looks
good
so
that,
as
the
thing
actually
goes,
you
can
see
it
work
and
then
one
of
the
other
things
you
might
have
seen
there's
a
silver
tube
coming
out
there.
That's
a
dryer
exhaust
tube.
This
thing
burns
things,
and
so
you
want
to
make
sure
that
the
exhaust
doesn't
actually
come
back
into
your
room.
The
glowforge
is
working
on
a
filter
that
fits
under
it.
A
That'll
essentially
go
ahead
and
filter
things
that
you
can
run
it
without
that
exhaust
port,
but
but
that's
not
shipping.
Yet
the
regular
glowforge
is
alright.
So
now
I'm
gonna
switch
back
to
my
screen
and
show
you
what's
going
on
here.
So
this
is
this
cool
javascript
library
called
paper
Jas,
and
it
makes
it
really
easy
to
do
sort
of
like
vector,
graphics,
type
of
stuff
in
the
script
and
so
Oh
three-prong.
Oh,
you
know
Chris,
no,
not
the
tree
puns,
please!
No,
we
could
do
birch
ventures
right.
A
A
You
can
then
go
through
and
export
this
as
SVG
and
then
once
you
exported
it
as
SVG,
you
can
go
ahead
and
upload
the
SVG
directly
into
glowforge,
and
so
I
did
that
a
while
ago,
and
what
happens
is
that
comes
in
here
like
this
now
the
glowforge
itself
has
a
camera.
The
super
wide
angle,
camera
that
can
take
a
picture
of
the
bed
and
you
can
actually,
oh
god,
Chris.
A
No
just
leave
just
leave,
so
you
can
drag
it
into
position
on
on
the
on
the
bed
to
actually
figure
out
where
you
wanted
to
call
this
thing's
pretty
accurate,
but
they're
doing
some
really
funky
math
here,
because
it's
a
wide-angle
camera
which
is
pretty
close
to
the
material.
So
there's
all
sorts
of
like
interesting
mathematics
to
essentially
deal
warp
the
the
image
out
of
this
camera.
So
they
can
actually
give
you
this
positioning
stuff.
A
So
that's
pretty
cool,
but
what
it
means
is
that
I
have
to
go
through
and
I
have
to
actually
put
some
settings
in
here.
So
this
is
about
an
eighth
of
an
inch
thick
I.
Have
these
these
calipers
that
I
use
these
like
cool
digital
calipers,
I,
don't
know
if,
like
you,
can
see
these,
that
I
can
use
to
actually
measure
thicknesses
stuff.
So
these
are
really
useful.
If
you
actually
start
doing
this,
so
I
can
do
so.
A
Eighth
of
an
inch
and
then
through
through
trial-and-error
I,
figured
out
that
if
I
set
the
speed
to
two
to
five
and
then
I
run
it
full
power,
that's
gonna
actually
give
me
a
good
cut,
and
so
now
what
I
do
is
and
you'll
see
it
if
I
press
print
here
so
that
I
think
that
looks
pretty
good.
If
I
press
print
here,
what
it's
going
to
do
is
it
says
scanning
your
material
and
so
what
it's
doing
is
it's
moving
the
head
there.
A
It
shines
a
red
laser
and
then
it
has
another
camera
on
the
head
and
it's
using
that
to
actually
go
ahead
and
really
detect
the
depth
and
fine-tune
it.
So
we
can
focus
the
laser
appropriately.
Let
me
adjust
this
camera
a
little
bit
now.
The
interesting
thing-
and
this
is
good
or
bad
depending
on
your
position-
is
all
of
this
stuff
happens
through
a
web
service,
and
so
the
glowforge
connects,
via
my
home
internet
and
Wi-Fi
up
to
the
glowforge
servers.
A
The
glowforge
servers
then
actually
implement
all
of
the
really
sort
of
funky
law
logic
and
oh
yeah.
Oh,
what
an
error
occurred.
I've
never
seen
it
before.
Let's,
let's
hit
it
again
and
actually
see
what
happens,
yeah
I
needed
Cooper,
Forge,
so
anyways,
so
so
the
and
then
it's
the
glowforge
servers
that
actually
do
all
the
sort
of
like
control
and
math
and
stuff
like
that,
and
so
it
really
creates
an
integrated
experience.
And
this
is
this
is
a
heck
of
a
lot
more
plug-and-play.
A
So
there's
a
little
bit
of
a
pitch
there
around
that
and
and
early
on,
I
help
them
get
started
and
I
pointed
them
towards
kubernetes,
so
they're
actually
running
on
kubernetes,
and
so
this
is
kubernetes
running
a
laser
cutter
to
cut
out
a
kubernetes
slinky.
Okay.
So
now
let
me
go
so
now
it's
ready
to
go.
It
says
it's
gonna,
take
three
minutes
and
52
seconds.
The
warning
here
is
essentially
saying:
hey
you're
using
your
own
materials.
A
We
can't
actually
guarantee
that
you'll
get
a
good
cut,
but
because
for
me
this
stuff
is
so
cheap
I,
don't
care
I've
done
trial
and
error.
I
have
this
thing
dialed
in
so
now,
I'm
gonna
switch
back
to
the
other
camera
and
there's
a
big
glowing
button
here.
I'll
show
you
the
glowing
button,
so
there's
a
glowing
button
and
if
I
go
ahead,
I'm
gonna
press
the
button
here
and
this
thing's
gonna
start
up.
A
The
fan
only
comes
on
when
it's
actually
cutting
it's
generating
quite
a
bit
of
smoke,
but
the
fan
actually
has
enough
airflow
to
move
it
out
and
then
out
my
window,
so
it
kind
of
smells
like
a
campfire,
and
so
you
know
I
guess
my
neighbors
are
getting
a
little
bit
of
a
campfire
impact.
Yeah
burn
baby
burn
and
there
we
go,
and
so
we're
at
three
minutes
and
eight
seconds
left
and
so
you'll
actually
see
this
thing
counting
down
as
it
goes
now.
This
can
both
cut
&
a.gray
engrave.
A
So
what
we're
doing
here
is
cutting,
so
it's
actually
doing
a
very,
very
sort
of
like
with
of
a
human
hair
type
of
cut
around
this
thing.
It
also
can
engrave
where
it
uses
lower
power
and
it
only
sort
of
burns
off
a
layer
of
the
wood.
In
that
case,
it
actually
is
kind
of
like
a
dot
matrix
printer
where
it
does
scan
lines
back
and
forth,
and
it
can,
it
can
do
about
two
or
three
hundred
dpi
is
what
I
I
read
when
you're
doing
that
engraving
it
ends
up
being
a
lot
slower.
A
So
I'll
show
you
like
what
the
engraving,
let's
see,
I,
don't
think
I
have
any
good
engraving
projects
ready
to
show
you.
So
that's,
but
it's
also
pretty
cool.
You
can
do
the
engraving.
It
just
takes
a
lot
longer.
So
this
demo
is
a
lot
better
because
it's
doing
the
cutting
the
total
amount
of
sort
of
like
linear
cutting.
That's
going
on
here.
I'm
really,
not
sure.
But
I
did
this
with
a
piece
of
cardstock
instead
of
wood
and
it
pretty
much
like
stretched
all
the
way
across
my
house.
A
So
there's
you
know
we're
gonna
end
up
with
essentially
a
really
long
thin
piece
of
wood
out
of
this
one.
It's
all
done
so
Matthias
says
I'd
rather
use
it
to
cut
pizza.
So
it
turns
out
that
people
have
used
this
to
cut
food.
If
you
go
on
the
glowforge
forum,
so
I
think
somebody
like
extra
banana,
so
I
don't
know
if
anybody's
done
pizza
yet.
A
Well,
you
could
do
the
taste
that
you
could
build
a
pizza
cutting
template,
so
you
could
use
this
to
essentially,
like
you
know,
make
sure
that
you
can
guide
your
pizza
cutter
correctly
at
all
times,
so
that's
totally
possible,
so
the
materials
of
this
will
cut
it
will
do.
Would
I've
done
up
to
a
quarter-inch
I
think
you
may
be
able
to
do
a
little
bit
thicker
than
that,
but
it
gets
more
difficult.
You
can
do
a
plastic.
Acrylic
works
really
really
well.
You
can
do
cloth.
A
You
can
do
leather,
I,
think
I,
don't
know
where,
when
I
had
a
small
hefty
a
logo
that
I
did
with
leather,
you
can
etch
certain
types
of
metal
and
people
done
all
sorts
of
other
things.
The
one
thing
that
you
don't
want
to
cut
is
anything
that
has
chlorine
in
it.
So
PVC
specifically,
is
a
bad
idea
because
it
actually
generates
boring
gas
as
you're
actually
cutting
it,
which
can
both
kill
you
and
and
go
ahead
and
ruin
the
machine
because
it
can
throw
it
yeah.
So
the
cost
I
think
right.
A
Now,
it's
a
pre-order.
If
you
got
in
early
on
it,
it
was,
it
was
probably
I
think
around
2000
was
the
was
that
the
price
when
it
first
started,
you
know,
since
they've
entered
production,
things
aren't
quite
as
as
expensive
as
I
mean
things
about
a
little
bit
more
expensive,
so
yeah.
So
there
we
go.
It's
done.
Okay,
now
I
can
stop
yelling
and
let's
go
ahead
and
look
at
it.
So.
A
So
this
is
one
of
the
problems
that
we
have
here.
I,
don't
know.
If
you
all
can
see
this,
and
this
is
why
the
company
sells
that
sort
of
proof
material.
Is
that
there's
a
piece
there's
a
there's,
a
piece
where
they
probably
filled
in
a
knot,
and
so
there
was
extra
glue
there
and
it
actually
didn't
cut
through
it
as
much
as
it
could've,
right
and
so
I
could
have
gone
with
higher
power
or
I
could
have
gone
over
it
twice
and
that
would
have
actually
solved
the
problem
there.
A
But
this
is
me
being
cheap,
but
if
I
go
ahead
and
I
tear
this
out
anyways,
what
you
can
see
is
that
it's
actually
a
little
bit
like
in
this
one
again,
because
you
didn't
cut
all
the
way
through
there.
That's
unfortunate,
but
I
have
another
sort
of
similar
one
here
that
we
can
check
out
of
like
so
this
is
8th
inch
plywood.
It
turns
out
when
you
cut
it
really
thin
skin
thin.
It
actually
turns
into
a
slinky,
so
unfortunately
Mike.
This
is
the
first
time
I've
had
one
of
these
things
fail.
A
A
So
one
of
the
things
that
you
can
do
in
and
if
you
go
to
I'm,
trying
to
remember
the
name
of
the
site,
but
but
there's
all
sorts
of
sites
out
there
that
will
actually
help
you
that
will
help
you
sort
of
like
do
boxes
that'll
like
finger
joint
together,
so
that
actually
ends
up
being
really
really
cool.
Do
I
have
to
purchase
additional
fire
insurance.
No,
let
me
go
ahead
and
with
my
camera
back
here.
A
Yeah
I
probably
should
like
what
I
do
have,
though,
is?
Is
you
know,
safety
first
I
do
have
the
fire
extinguisher
right
here
in
case
the
whole
thing
lights
on
fire,
so
that's
kind
of
fun,
so
yeah.
So
with
that
I
think
I'm,
gonna,
sign
off
I
hope
everybody
has
a
great
weekend.
If
you
have
any
questions
about
contour
or
or
even
the
globe
Forge.
You
know
feel
free
to
reach
out
to
me
on
Twitter
and
I'm
happy
to
chat
about
any
of
this
stuff.
A
So
yeah,
yeah
and
I
will
put
a
bunch
of
links
into
the
description
for
this
video
to
a
bunch
of
the
stuff
that
we
talked
about.
So
you
can
actually
find
it
later,
but
you
can
also
go
glowforge
comm
and
see
it.
There
there's
a
really
cool
marketing
video
that
that
they
made
when
they
first
launched
the
crowdfunding
campaign
for
the
thing
so
alright.
Well,
thank
you
very
much.
Everybody
and
I
will
see
you
next
week.
I,
don't
know
what
my
topic
is
gonna
be
yet,
but
I
may
be.