►
From YouTube: TGI Kubernetes 053: The AWS Cluster API
Description
Come hang out with Kris Nova as she does a bit of hands on hacking of Kubernetes and related topics. Some of this will be Kris talking about the things she knows. Some of this will be Kris exploring something new with the audience. Come join the fun, ask questions, comment, and participate in the live chat!
A
Hey
everyone
and
welcome
to
TGI
K
Fridays
at
1
p.m.
Pacific
live
from
the
the
hefty
O
Studios
here,
hello
how's
everybody
doing
today
feels
good
to
be
back
in
Seattle
I
was
in
Portland
last
night
in
this
morning,
I
flew
in
just
40
GI
case,
especially
today.
So
how
is
everyone
everyone
doing
I'm
in
getting
stuff
pulled
up
right
now,
it's
good
to
see
folks,
let's
see
I'm
gonna
pulp,
the
chat
we
all
know.
A
This
is
my
my
favorite
part
of
the
week
saying
hi
to
everyone
all
right,
so
we
have
dig
Olive,
saying
yeah,
Jason,
Phillip,
Martin,
hello
from
Paris
good
to
see
you
love
Maddy.
It's
always
great
to
see
you
Mateus
good,
to
see
you
again.
Hopefully
you
can
it.
You
can
help
me
out
again
if
I
get
off
in
the
weeds
good
to
be
here,
we
have
helped.
You
I
think
that's
George
on
the
hefty
owe
chat
so
he's
here
as
usual,
helping
out
with
the
hack
indie
and
some
of
the
other
things.
A
He'll
he'll
put
us
his
stuff
in
the
chat
and
explain
it
to
everyone.
Marco
good
to
see
you
I.
She
says
hello
from
across
the
lake
I'm,
assuming
you
mean
Lake
Washington,
which,
in
which
case
hello,
good
to
see
you
but
yeah.
Welcome
we're
gonna
be
talking
about
the
cluster
API
today,
I'm
really
really
excited
about
this
I
I
spend
a
lot
of
time
doing
you
know
public
speaking
about
this
type
of
stuff
and
I've.
A
You
know
this
is
like
sort
of
a
reflection
of
a
lot
of
work
that
myself
and
a
lot
of
people
here.
I
have
do
in
Google
and
in
other
places,
have
a
have
been
working
on
for
quite
some
time,
so
getting
to
actually
live
demo.
It
and
explain
it
to
people.
Finally,
with
like
concrete
examples,
is
going
to
be
really
rad
so
I'm
super
pumped
about
that
it
looks
like
we
had
some
other
folks
to
join
us.
A
We
have
hello
from
Montreal,
hello,
Eric,
Steve
good,
to
see
you
happy
Friday
Duffy
whoops,
that
was,
it
was
Duffy
masquerading
as
hefty.
It
was
not
George.
Okay,
well
Duffy.
If
you
wanted
masquerade
its
hefty,
oh
and
share
the
hack
indeed
feel
free
to
Roy
happy
Friday,
Chris
Jason
good,
to
see
you
Jason
when
Maddy
the
soy.
This
is
really
great
because
it's
like
all
of
the
cluster
API
people
are
in
the
chat
for
TGI
K
today.
So
it's
just
like
the
old
Bing
got
back
together
for
TGI
K.
So
that's
really
exciting.
A
Okay,
so
let
me
grab
a
quick
drink
and
then
I'm
gonna
get
stuff
set
up
in
wall.
I
will
jump
right
in
oh
also
feeling
much
better
since
last
week
last
week
was
like
crazy.
I
was
sick
but
still
managed
to
get
it
and
do
it
still
had
a
lot
of
fun
ended
up
going
I
think
like
two
hours
or
something
but
a
week
of
laying
laying
around
the
house
and
healing
really
did
the
did
the
job
here.
So
this
is
like
one
of
my
first
days
back
and
I'm
I'm
really
pumped
about
it.
A
Okay,
so
without
further
delay,
let's
jump
into
our
hack,
MD
and
I
can
share
this
too.
If
we
don't
have
a
link.
Yet
so,
let's
see
screen
+
face.
We
want
this.
One
here
so,
let's
see
how
do
I
want
to
share
this
hack,
em
D,
so
I
can
come
here
and
I
can
go
down
to
my
secret
window
that
nobody
else
sees
and
I
can
paste
it
in
and
there
we
go.
A
That's
me
as
the
hep
D,
you
account
if
anybody
wants
to
go
and
join
the
hacking
D
via
take
notes
as
we
go
through.
The
episode
feel
free
to
it's
just
good
old
github
markdown
on
the
left
and
notes
for
the
episode
here
on
the
right
feel
free
to
take
notes
that
there's
anything
you
you
want
to
share
or
ask
or
I
think
that
would
be
relevant,
write
down,
feel
free
to
do
just
type
them
up
here
on
the
left,
side
and
they'll
show
up
here
on
the
right
side.
A
Let
me
adjust
my
screen
really
quick.
There
we
go
that
looks
much
better.
Let's
see
what
the
Maddy
says,
it
says
huge
thanks
to
you
have
do
for
making
managing
kubernetes
and
C&I
available
for
free.
You,
rock
you're,
welcome
Linate
happy
to
help
out
oh
and
Chuck's
here
too
good
to
see
you
chuck.
It's
been
like
what
10
minutes,
since
we
saw
each
other
so
good
to
see
you
again
cool,
so
we're
going
to
start
off.
A
I've
got
a
lot
of
links
in
here
for
a
cluster
API
and
I've
got
some
other
this
week
in
kubernetes
thing
that
are
exciting.
We've
got
a
little
go
demo
thing
for
today.
Sometimes
I'll
do
like
a
going.
You
know,
five-minute,
learn
something
and
go
and
we'll
be
doing
one
of
those
today
and
then
we're
going
to
jump
into
the
cluster.
Api
and
I
have
a
feeling.
A
Today's
gonna
be
a
lot
of
me
talking
about
like
sort
of
the
design
here
and
looking
at
a
couple
of
different
repositories
and
then
actually
running
it
and
checking
out
logs
and
and
seeing
what's
going
on,
Sean
Smith
says
happy
Friday,
everyone
here
in
Seattle.
It
is
a
great
Friday
here
in
Seattle
like
it's
really
like
I
can
look
out
the
window
over
there
and
it's
super
sunny
and
flying
in
from
Portland
this
morning,
I
got
to
see
all
of
the
mountains,
like
literally
all
over
the
mountains,
and
it
was
really
exciting.
A
So
yeah
that's
a
good
day
here
in
Seattle,
okay,
so
let's
skip
past
cluster
API.
This
is
the
cluster
API
API,
which
more
on
that
later,
this
is
the
AWS
specific
stuff
we're
going
to
be
looking
at
today
and
let's
go
right
into
this
week
in
kubernetes.
So
I've
only
got
a
couple.
This
week
it's
been
kind
of
a
quieter
week.
Suresh
says:
hey,
everyone
greetings
from
hamburg
and
we
have
somebody
from
turkey
good
to
see
you
as
well.
A
So
this
first
one
here
is
cubic
dual
pseudo,
which
I
just
looked
at
this
about
an
hour
ago.
I
don't
know
very
much
about
it
other
than
the
fact
that
the
name
was
extremely
enticing,
so
I
figured
it'd
be
good
to
just
kind
of
go
through
this.
Together
we
pulled
this
off
of
the
the
our
kubernetes
subreddit
and
just
anytime.
I
see
anything.
That's
called
like
something
pseudo
or
people
are
trying
to
do
something.
Pseudo
I,
like
I,
get
super
interested
I
want
to
see
what's
going
on
there.
A
Also,
why
we're
on
the
topic
of
pseudo
I
just
want
to
be
very
clear
that
I
intentionally
pronounce
it
super
user
doe.
The
instead
of
super
user
do
even
though
I
think
technically,
it's
pronounced
su,
do
but
I'm
a
pseudo
ur.
So
if
you
want
to
drop
a
+1
for
a
pseudo
and
plus
2
for
su,
do
Toofer
do
that.
Rhymes
feel
free
to
vote
in
the
chat,
I'm
just
curious
what
people
think
and
how
they
pronounce
that
word
anyway.
A
Let's
see
what's
going
on
here,
this
project
does
not
really
introduce
a
key
Bechdel
plug-in,
but
a
concept
of
how
to
provide
a
like
system
for
kubernetes
access.
It
says:
kubernetes
cluster
admins
have
great
power.
This
means
that
they
make
a
mistake.
They
could
cause
the
cluster
to
become
unhealthier
insecure.
Just
like
a
Linux
system.
A
Moving
on
a
simple
keep
Bechtel
F
on
a
file
wrongly
namespace
can
end
badly.
So
I
think
this
is
the
solution
here.
To
reduce
the
surface
or
unwanted
expected
actions.
You
can
reduce
the
default
privileges
of
a
cluster
admin
when
a
cluster
admin
needs,
but
a
privilege
he
or
she
can
do,
can
switch
to
the
group
systems
masters
or
any
other
user.
Okay,
so
it
looks
like
all
this
is
doing.
Is
its
setting
up?
Some
are
back
rules
and
roles
to
help
give
users
a
like
a
more
pseudo
like
experience
with
managing
their
clusters.
A
In
this
first
role,
we
have
this
look,
it's
called
an
impossible
person
ater,
so
I
have
a
feeling.
This
is
gonna,
be
like
our
negative
check
here
and
it
looks
like
we
create
a
few
other
roles.
We
have
cluster
admins,
we're
giving
them
the
cluster
role
and
impersonate
ur,
and
then
here
we
just
do
a
queue
Bechdel,
as
this
is
a
handy
little
flag.
A
If
you
don't
know
about
this,
yet
this
allows
you
to
run
queue
Bechtel
as
a
specific
user,
and
you
can
also
run
it
as
a
specific
group
defined
in
our
back
as
well
and
I.
Think
it
just
kind
of
packages
this
whole
are
backing
system
up
and
gives
it
a
handy.
Little
alias
here
called
queue,
Bechdel's
pseudo.
So
this
is
a
fun
little
program.
A
It's
something
like
this
is
probably
a
good
example
to
at
least
learn
how
to
start
creating
things
like
our
back
roles
and
groups
and
how
to
start
using
them
and
probably
learn
a
little
bit
more
about
our
back.
If
that's
something
you're
interested
in
I,
don't
know
if
I
would
enjoy
the
idea
of
having
pseudo
on
a
cluster
but
yeah,
maybe
it
could
be
useful.
Hey
I,
don't
know
so.
Moving
on,
oh
wow,
everybody's
voting,
I
wanna,
see
what
everybody's
voting
for.
Let's
see,
we
have
Justin
who
finally
made
it
from
Vancouver.
A
We
have
a
ton
of
Plus
Ones
Duffy
says
this
is
just
a
raptor
freak.
You
Bechdel
as
I
think
it's
neat
stuff
I,
really
like
this
pattern.
Cuz
from
the
audit
side,
you
get
the
benefits
as
well.
Okay,
so
Duffy
brings
up
a
good
point,
which
is,
if
you're,
actually
doing
an
audit
on
your
you're
kubernetes
resources
having
a
unique
group
defined
for
that
would
give
you
the
ability
to
go
and
see
which
group
made
which
mutations
to
your
cluster.
A
So
it
gives
a
little
bit
of
a
paper
trail
for
people
interacting
with
their
cluster
in
the
case
of
somebody
actually
going
in
and
doing
something,
malicious
or
goofing,
something
up
or
you
know,
providing
the
wrong
namespace
or
whatever
we
can
actually
go
and
to
tie
that
back
to
some
sort
of
our
back
configuration.
So
that's
handy
and
ylim
addy
says:
did
you
get
this
working
with
KTX
Duffy
and
what
is
the
Madi
talking
about?
Oh
Sean
says
interesting
was
in
a
way
of
the
a
staff
from
Quebec
gold
yeah
speaking
of
KTX.
A
Well,
we
can
look
at
that
a
little
bit
later,
we're
gonna
be
using
some
of
it,
but
yeah.
That's
a
good
way
of
managing
cube
configs
one
of
our
engineers
here,
Amy
tweeted,
about
managing
cube,
configs
and
I
can
do
a
little
demo
of
how
I
do
it
here
in
a
second
but
yeah
we're
getting
off
in
the
weeds.
Let's
go
back
to
our
s
this
week
in
kubernetes,
so
this
next
one
was
exciting
because
I've
been
thinking
a
lot
about
this,
which
is
a
bare-metal
install
of
kubernetes
and
yeah.
A
You
can
actually
come
to
this
actually
is
a
pointer
to
the
subreddit
here.
I
want
to
see
what
folks
are
saying:
it'll,
Maddie
nope,
it's
stuffy
this
time.
The
audit
even
will
show
you
who
the
impersonator
is.
That's
good
to
know,
I
think
Duffy's,
talking
about
the
the
sudo
tool
we
just
looked
at
okay,
so
in
this
example,
I
think
all
that
we're
really
doing
is
we're
gluing
together
a
lot
of
kubernetes
set
up
for
what
people
would.
A
A
But
you
can
actually
see
that
we
are
defining
some
various
repositories
and
adding
keys
for
packages
like
kubernetes,
just
regular
kubernetes
and
we're
installing
docker
and
we're
doing
some
Python
new
things
as
well
and
then
ultimately
down
here
at
the
bottom.
We
get
to
this
cue,
bad
mini
little
phrase,
and
you
can
see
that
we're
installing
cubic
double
and
keep
admin
and
a
few
others.
A
A
It
shows
how
you
have
to
know
certain
pieces
of
information
about
other
parts
of
your
system
as
you
bring
up
a
kubernetes
cluster
if
you
go
through
here,
you're
bound
to
find
stuff
like
take
this
token
and
copy
it
over
to
this
other
server,
which
is
where
tools
like
cops
and
cubic
corn
come
in
handy
because
it
sort
of
does
that
magic
for
you.
But
this
is
a
good
resource
if
you
just
want
to
understand
what
it
takes
to
get
kubernetes
up
and
running,
let's
see
what
folks
are
saying
in
the
chat
here.
A
Sean
says:
yeah
I
just
use
some
links
to
manage,
keep
configs
I
was
planning
on
looking
do
KTX
after
seeing
amy's
tweet
and
the
responses
Sean
yeah
we
can
do
later
on.
If
I
forget
remind
me
once
we
start
looking
at
the
kubernetes
clusters
that
we're
bringing
up
I
can
do
a
quick
demo.
Actually,
you
know
what
I'm
just
gonna
do
it
now?
Why
not
so
here's
my
terminal?
This
is
like
specifically
for
you
Sean.
A
A
This
is
just
a
really
handy
little
bash
program
and
I.
Think
to
install
it
like
it's
just
like
a
one-liner
like
you
just
paste,
something
in
and
I
think
I've
even
talked
about
this
before,
but
it's
called
this
cubed
CTX
tool
here
and
there's
an
example
of
how
to
run
it
and
then
I
just
alias
this
to
KTX.
So
if
we
go
back
to
my
terminal,
you
can
actually
see
that
alias
grep
KTX.
You
can
see
that
I
just
have
Q
CTX,
alias
to
KTX
and
I.
A
Think
Chuck,
one
of
the
folks
on
the
call
also
has
this
tool
called
just
KTX,
which
is
confusing.
Let's
see
if
we
can
find
it
I
wonder
if
I
can
take
Chuck,
because
I
alyas
mind
to
the
same
name
as
his
project,
which
is
hekia
labs
here,
ahead
to
your
labs,
KTX
and
I.
Think
this
is
sort
of
does
the
same
thing.
I
just
allows
you
to
manager
cute
configs
in
the
same
sort
of
way,
so
anyway
managing
keep
configs.
A
It's
a
thing,
it's
possible
because
Q
Bechtel
itself
will
actually
munge
a
lot
of
configuration
declaration
bits
into
the
same
cube,
config,
and
then
it
just
has
this
one-liner
that
says
contacts
that
allows.
You
to
switch
between
the
different
configurations
in
there,
so
that's
why
you
can
manage
multiple
clusters
and
you
can
switch
between
them
just
by
redefining
the
context
field
and
your
your
cube
convey
amyl
anyway
back
to
our.
Where
were
we?
A
Nobody
looked
at
this
back
to
our
installing
kubernetes
bare-metal
here.
So,
if
you
want
to
check
this
out,
feel
free
to
come
check
it
out,
it's
an
example
looks
like
it
was
working
for
a
lot
of
folks,
and
it
looks
like
it's
a
good
step
by
step
guide
for
getting
kubernetes
up
and
running
in
a
bare
metal
data
center,
which
I've
been
doing
a
lot
of
thought
about
this,
and
there's
been
some
folks
at
a
publishing
company
I've
been
working
with
who
has
yet
to
be
named.
A
Who
is
very
interested
in
sort
of
a
cloud
native
infrastructure
and
bare
metal
II
type
of
book
that
somebody
might
be
writing
in
the
future
anyway.
More
on
that
later,
let's
see
what
folks
are
saying,
as
she
says,
pro
tip
for
folks
dealing
with
multiple
clusters
show
context
as
part
of
the
bash
prompt
that
will
show
the
cluster
you
were
appointed
to
yes,
so
that
goes
back
into
the
editing.
Your
bash
prompt
a
lot
of
folks
like
to
do
that.
A
I
personally,
like
to
keep
my
bash
profile
as
simple
as
possible,
but
that's
for
my
own
crazy
reasons
and
the
fact
that
I
try
to
keep
things
as
simple
as
I
can
Shawn
says:
yeah
I
generally
keep
a
cluster
per
file,
but
have
a
context
per
namespace
in
my
cube,
configs
seems
good
thanks.
Chris
hearts
and
smiley
faces
all
right
tract
here.
This
last
one
I'm
gonna
skip
to
the
bottom.
Here.
It's
deploying
spark
on
kubernetes.
A
This
is
something
that
has
been
not
only
as
something
that
me
and
my
girlfriend
talked
about
a
lot,
because
she
does
a
lot
of
work
on
spark.
But
a
lot
of
folks
ask
about
spark
on
kubernetes
as
well.
I'm.
Definitely
not
a
spark
expert,
but
I
am
a
kubernetes
expert
and,
at
the
end
of
the
day,
kubernetes
advertises
to
be
able
to
run
any
application.
So
in
theory,
if
you
look
at
spark
as
just
another
application,
despite
its
complexity,
it
should
be
possible
to
actually
get
a
healthy
cluster
up
and
running.
A
In
kubernetes
and
I
know.
We've
looked
at
things
like
spark
controllers
and
operators
and
spark
helm,
charts
and
a
couple
of
others.
So
this
seems
to
be
another
attempt
at
getting
spark
up
and
running
in
kubernetes
and
the
first
thing
that
I
noticed
when
I
came
and
checked
out
this
this
example
here
was:
we
are
actually
building
a
java
container,
we're
using
a
JDK
8.
A
So
already
a
little
bit
excited
slash
nervous
because
we're
running
Java
in
a
container-
and
we
all
know
that
there
are
some
parallels
with
the
JVM
running
inside-
of
a
containerized
system
or
inside
of
a
container
itself,
especially
before
a
job
at
10,
so
curious
to
see
how
that
would
all
play
out.
And
it
looks
like
we
do
some
we're
building
some
spark
Hadoop
containers
as
well,
and
then
we
actually
have
a
deployment
for
the
spark
master.
And
then
we
scroll
down
and
I
think
we
have
a
deployment
for
the
spark
service.
A
And
then
we
have
spark
workers
down
here.
So
this
is
how
we
would
actually
build
a
distributed
spark
system.
It's
exciting,
because
this
actually
goes
from
zero,
all
the
way
to
getting
a
spark
installed
in
kubernetes.
And
then
it
goes
a
step
further
and
it
actually
goes-
and
it
gives
you
this.
A
This
example
down
here
at
a
bot
at
the
bottom
running,
PI
spark
to
actually
get
you
like
a
spark
query
or
whatever
it
would
be
called
up
and
running,
and
you
can
actually
go
and
execute
something
to
get
spark
which
I
think
is
neat
because
a
lot
of
folks
don't
go
that
extra
step.
So
this
is
a
good
sort
of
round
robin
approach
to
just
running
a
simple
PI
spark
command
in
sparkling
kubernetes,
so
very
exciting.
Timmi
timmi
joined
hi,
Tammy,
howdy,
Chris,
I'm,
late
I'm.
Sorry,
it's
ok,
Timmy!
Just
don't
let
it
happen
again.
A
So
a
decent
says
good.
You
mean
from
Russia
good
evening.
Everyone
thanks
folks
for
joining
we're
still
doing
the
coup
Nettie's
this
week,
part
of
the
intro
here
and
we're
almost
done
I
think.
Actually
that
was
the
last.
Like
blog
post
e
thing,
we
have
so
let's
go
ahead
and
close
some
of
this
and
we'll
do
our
quick
little
go
lesson
and
then
we'll
jump
right
into
the
cluster
API.
Okay.
A
Anyway,
why
folks
are
answering
I'm
going
to
start
to
explain
what
we
have
going
on
here?
So
the
first
thing
we
have
I
want
to
bring
attention
to.
Is
this
main
function?
Here
is
the
main
entry
point
of
the
program
and
what
we
have
here
is
we
have
this
other
function
called
T,
gik,
execute
I,
don't
know
why
I
exported
this,
but
whatever
it's
just
a
demo,
so
we're
fine
for
right
now,
but
in
this
T
gik
execute.
A
You
can
see
we're
passing
in
this
thing
and
what
this
is
is
just
in
a
decent,
Mike
Shawn
looks
like
a
lot
of
people
like
function
types,
so
this
is
good
because
I
really
like
to
use
these
and
I
feel
like
it's
really
easy
to
be
intimidated
by
these.
As
a
software
engineer,
because
you
can
solve
this
problem
in
other
ways,
but
this
just
allows
for
a
lot
more
flexibility
as
you're
crafting
your
program
and
mapping
things
together.
A
I
think
the
mapping
is
really
the
important
part
here
and
we're
gonna
look
at
like
an
example
of
this
map
here
in
a
second
after
we
sort
of
lay
out
the
scaffolding
of
what
we're
looking
at
here
in
front
of
us
so
yeah.
All
we
do
is
we
call
this
T
tik
execute
function
and
in
this
T
gik
execute
function.
We
pass
in
something
which
is
renamed
to
the
variable
F
here
when
we
actually
enter
the
function
definition
and
then
we
call
that
function
and
we
pass
in
this
string.
That
says:
welcome
to
TG.
Okay.
A
Now,
if
you
look
at
the
function,
definition
for
T
GI
k
execute.
You
can
see
that
the
first
parameter
here
is
of
type
T,
gik
func,
t
GI,
k,
func
type
is
defined
up
here
on
seven
and
we
are
defining
that
as
any
function
that
just
accepts
a
string
as
its
first
parameter.
So
this
is
cool,
because
you
could
actually
do
a
couple
of
different
types
of
functions
here.
A
You
could
have
like
another
func
type,
which
would
be
like
a
func
that
accepts
an
INT
and
end
in
a
string,
and
then
that
could
return
an
error
and
maybe
an
int
as
well.
So
we
can
have
all
types
of
functions
here
that
we
can
define
them.
However,
we
want-
and
we
can
put
types
inside
of
them
and
we
can
pass
those
around
as
any
other
expression
in
the
go
programming
language.
So
anyway,
for
this
first
one
we
kept
it
very
simple.
We
have
a
function
that
accepts
a
string
and
returns
nothing.
A
So
here
these
two
functions
satisfy
the
criteria
for
number
line.
Number
seven
here,
which
we
have
one
function
called
simple
echo
that
takes
a
string
and
we
have
another
function,
called
exciting
echo
that
also
takes
a
string.
Neither
of
them
return
anything.
So
that
tells
us
that
both
of
these
can
be
implicitly
passed
in
as
a
TGI,
kay
func,
meaning
this
T
gik
execute
function
can
accept
either
exciting,
echo
or
simple
echo,
and
then
we'll
pass
the
string
to
it.
A
A
We
get
one
with
an
exclamation
point
in
one
without
an
exclamation
point,
and
if
you
actually
look
at
the
functions
implementation
details,
you
can
see
that
we
do
just
that
in
the
simple
echo
we
print
it
out,
just
as
is,
and
in
the
second
one
we
print
it
out
with
an
exclamation
point
here
at
the
end.
Okay.
So
why
is
all
this
important?
This
is
important,
because
this
function
doesn't
change.
A
We
never
have
to
change
line
25
and
in
fact
this
function
could
live
in
a
different
package
or
can
do
some
sort
of
crazy
look
up
or
hit
some
sort
of
API
to
generate
the
input
for
the
different
types
of
functions
that
go
in
here.
So
as
a
software
engineer
that
allows
us
to
do
things
like
take
these
different
types
of
functions
that
are
defined
and
we
can
have
conditional
logic
with
it.
So
you
know,
if
not
true,
we
can
call
the
exciting
echo
one.
A
Otherwise
we
want
to
call
the
the
simple
echo
one,
and
this
allows
us
to
sort
of
have
this
really
simple
and
really
clean
logic
here
without
having
to
actually
go
through
and
change
any
of
the
implementation
detail
for
the
function.
So
this
is
sort
of
like
a
higher
level
and
made
a
construct
that
says
map
according
to
whatever
logic
you
want,
and
then
these
functions
themselves
can
live
anywhere
else
in
your
code
or
maybe
in
somebody
else's
code
that
you
don't
have
to
deal
with
a
decent.
A
What
do
you
think
about
the
generics
proposal
and
go
well
I?
Guess
it
doesn't
really
matter
what
I
think,
because
it's
happening,
but
in
general
I've
been
waiting
for
generics
for
quite
some
time.
I'm
not
super
familiar
with
the
proposal
itself,
a
decent.
If
you
want
to
share
a
link
to
I,
think
it
was
Russ
who
wrote
it.
The
proposal
in
the
hack
Andy,
we
can
pull
up
and
look
at
it
a
little
bit
later,
but
in
general
I
think
it's
a
good
idea
and
I'm
at
this
point
I'm
so
ready
for
it.
A
That
I
am
swallowing
my
strong
opinions
about
the
implementation
behind
the
scene.
Scenes
also
just
in
general.
Generics
is
an
extremely
hard
problem
to
solve
programmatically
behind
the
scenes,
so
I
I
do
want
to
take
it
with
a
grain
of
salt.
Whenever
I
look
at
whatever
implementation
that
go
team
comes
up
with
I
know
it's
not
going
to
be
easy
for
them,
and
they're
probably
already
put
a
lot
of
more
thought
into
it.
Then
I'm
gonna
be
able
to
put
into
it
and
afternoon
of
reading
a
proposal
so
yeah.
A
If
you
want
to
share,
have
you
sent
please
I?
Please
do
so
okay,
so
anyway,
this
is
using
function,
types
in
go
and
you
can
see
here.
We
actually
the
one
thing
I
wanted
to
call
out
was
we
don't
actually
call
the
function
here
by
adding
parens
to
the
end.
We
just
typed
the
name
of
the
function
that
we
wish
to
call,
and
then
we
actually
reference
that
here
anonymously
or
dynamically
with
the
parens
down
here
on
line
32
so
anyway
function
types
in
go,
they're,
handy
for
mapping
your
programs
together
now
without
further
ado.
A
Let's
look
at
beauty,
cluster
API,
okay.
So
what
we
want
to
do
is,
let's
start
off
with
this
repo
here,
and
this
is
going
to
be
like
a
little
bit
luxury
for
the
next
five
or
so
minutes.
So
if
you
wanna
go
get
a
cup
of
coffee
or
whatever
I'm,
probably
going
to
have
some
diet
coke,
because
I
go
through
this,
we're
just
going
to
sort
of
talk
about
the
cluster
API
and
what
it
is
so
like.
I
have
a
slide
on
this.
A
That's
like
a
long
long
time
ago,
in
a
galaxy
far
far
away
like
folks
at
Google
and
self
got
together
and
started
this
cluster
epi
project,
and
now
it's
totally
turned
into
a
thing,
but
basically
what
you
need
to
know
about
the
cluster
API.
Is
this
really
complicated
diagram?
I'm
gonna
redraw
that
in
a
very
simple
way
here
to
maybe
give
folks
a
better
understanding
of
what
what
we're
doing
here.
So
here's
a
duct
camera.
A
So
basically,
all
we're
doing
now
is
we're
defining
two
things,
one
of
which
is
a
cluster
and
then
underneath
a
cluster.
We
have
multiple
types
of
machines
right,
and
these
are
API
types
in
kubernetes
and
because
they're
API
types
we've
decided
to
come
together
and
write
an
abstraction
as
a
community,
that's
meaningful
to
all
cloud
providers.
A
In
other
words,
this
specific
machine
might
represent
an
instance
in
Amazon
ec2
cloud,
or
it
could
represent
a
virtual
machine
in
Microsoft
Azure
cloud,
but
ultimately
it's
still
going
to
be
called
a
machine
and
we're
still
going
to
sort
of
package
the
information
up
in
the
same
way
and
nadir
says
yes,
I
diagrams
out-of-date,
which
is
why
I'm
kind
of
drawing
my
own
here
and
then
we
have
this
concept
of
in
number
of
machines.
So
you
know
we
could
have
and
effectively
attach
as
many
machines
as
we
wanted
to
to
this
higher
level
cluster
object.
A
And
then
this
cluster
object
actually
represents
a
kubernetes
cluster
and
we're
going
to
go
and
we're
actually
gonna
look
at
the
fields
in
each
of
these
machine
types
behind
the
scenes
now.
I
think,
what's
important
to
know
is
that
these
definitions
ultimately
can
one
exist
in
the
form
of
EML
and
to
end
up
in
some
database
somewhere.
That
database
could
be
at
CD.
That
database
could
be
another
at
CD
server.
I
mean
in
theory
that
database
could
be
any
old
thing
you
want.
A
It
could
be
in
s3
in
the
form
of
BML
or
Jason
if
you
wanted,
but
that
is
how
we
would
represent
these
cluster
objects
after
we
have
this
in
place.
We
would
then
have
these
other
things
called
controllers
or
operators
that
we've
talked
a
lot
about.
That
would
sort
of
like
exists
over
here,
so
we're
just
gonna
call
this
an
operator
will
say
operator
for
short
and
all
that
that's
gonna
do.
A
Is
it's
going
to
sort
of
do
this
reconciliation,
negotiation
and
update
leap
between
wherever
that
that
information
is
stored
and
this
operator
would
then
actually
go
in
all
of
its
wonderful
greatness
and
sprinkle
all
of
this
magic
out
into
the
cloud
and
actually
make
whatever
you
defined
here
and
here
happen
in
the
cloud.
So
the
whole
thing
what
exists
says:
I
created
a
new
machine,
gamal
file,
I
update
that
it
gets
saved
in
a
database.
The
operator
sees
this
new
yo
file
and
then
it
goes
in
and
takes
some
sort
of
action
in
the
cloud.
A
The
whole
point
of
this
whole
set
up
is
so
that
we
can
not
only
install
kubernetes
clusters,
but
so
that
we
can
actually
go
and
mutate
them
after
the
fact
as
well.
In
other
words,
if
you
wanted
to
create
or
change
or
delete
or
modify
this
machine,
you
no
longer
have
to
SSH
into
it
or
actually
go
do
anything
manual,
but
you
would
rather
have
to
go
and
just
change
some
sort
of
record
in
kubernetes
somewhere,
and
this
operator
should
take
care
of
the
rest.
A
A
The
way
I
did
right
now
we're
actually
creating
our
own
Etsy
server
with
the
cluster
AWS
stuff
that
we're
looking
at
today,
and
the
reason
we
have
this
own
@cb
instance
is
because
we're
doing
what
Duffy
said,
which
is
we're
running
an
aggregated
API
server
in
kubernetes.
All
you
really
need
to
know
about
that
is
that
it's
going
away
and
it's
an
outdated,
primitive
that
we
used
for
a
couple
of
dependency
reasons
and
I.
Think
a
lot
of
the
folks
even
talked
about
that
going
away
as
early
as
next
week.
So
for
all
intensive
purposes.
A
That
is
relatively
new
and
I'm
kind
of
intentionally
and
a
little
bit
mischievously
pushing
on
a
lot
of
the
engineers
here
at
have
do
and
a
lot
of
the
other
engineers
an
open
source
by
demoing
this
today,
just
to
get
it
up
and
in
front
of
people
which
of
course
makes
me
happy
and
I
think
terrifies
them
a
little
bit.
So
thank
you
to
the
the
engineers
who
are
letting
me
run
this
code
today.
It
means
a
lot
to
me.
A
So
anyway,
that's
cluster
epi
in
a
nutshell,
and
let's
go
back
here
and
I
think
my
diagram
was
maybe
I,
don't
know,
probably
just
as
confusing,
but
either
way
on
this
repository,
kubernetes
SIG's,
slash,
cluster
API
is
sort
of
a
really
great
example
of
a
bad
name.
In
my
mind,
it
does
define
an
API
if
you
actually
go
and
you
drill
down
into
like
I,
think
it's
pkg
api's
cluster
v1
alpha
one.
A
You
can
actually
find
like
here's
where
the
actual
API
is
defined,
but
then
there's
this
whole
other
like
chunk
of
stuff
here
in
the
repo
that
really
doesn't
have
anything
to
do
with
the
API
at
all.
It
is
more
of
like
a
framework
or
some
go
code
to
actually
help
you
get
like
your
own
flavor
of
this
up
and
running.
A
Let's
see
what
Chuck
says,
Chuck
says
can't
be
confirmed,
terrified,
sorry,
Chuck
and
Sirach
says
yeah
I
agree
with
Duffy.
Even
AWS
has
introduced
AWS
service
operator
or
alternative
to
open
service
broker
and
Service
Catalog
to
integrate
other
cloud
capabilities.
Yeah
I
think
the
operator
and
the
the
CID
pattern.
I
mean
that's
clearly
winning
here
and
I.
A
Think
we're
all
kind
of
in
favor
of
it
I
think
we
had
to
jump
through
a
few
hoops
as
a
community
to
get
the
software
where
it
needed
to
be
before
we
can
adopt
it
again
and
that
push
already
went
into
this
repository.
It's
now
just
a
matter
of
getting
it
into
the
the
other
one,
the
cluster.
What
is
it
API
provider
AWS
repository,
which
is
the
second
one
we're
gonna
look
at?
A
We
just
have
to
go
ahead
and
make
that
change
without
breaking
what
we
already
have
running
in
here,
which
we're
gonna
be
running
today
now
dear
says,
also
terrified
Chuck
smiling
outside
screaming
inside
yeah
I
mean
I'm
gonna,
be
nice
like
I'm,
not
I'm,
not
gonna
do
anything
mean
or
anything
I
promise
we're
gonna
have
fun
but
anyway.
So
this
new
repository
that
we're
looking
at
is
an
AWS
specific
implementation.
I
think
is
the
correct
word
to
use
there.
A
I'm
actually
gonna
draw
another
diagram
here,
because
there's
some
other
vernacular
I
want
to
kind
of
get
squared
away
before
we
move
back
in.
So
let's
see,
let's
go
back
to
my
doc.
Camera
I
do
love
this
doc.
Camera,
it's
really
nice
to
have
okay.
So
here
we
have
this
broader
level,
repo
called
cluster
API
that
really
we
can
think
of
this
to
use
Tim
Tim
st.
Clair's
vernacular.
A
Chuck
is
smiling
and
Duffy
says:
y'all
got
this
cluster
API
in
great
hands,
yeah
great
hands
here,
look
all
great
I'm,
drawing
already.
Also
like
speaking
of
great
hands.
You
can't
even
tell
that
my
hand
was
broken
right
so
anyway,
cluster
API,
so
underneath
the
cluster
API,
we
can
have
different
breakouts
for
all
of
the
different
clouds.
So
in
this
first
one
the
one
we're
going
to
be
looking
at
today
is
gonna
be
AWS.
But,
as
I
mentioned
earlier,
there's
a
couple
of
other
implementations
out
there.
A
We
know
we
have
one
for
gke
and
then
you
know
we
could
have
like
any
other
amount
of
implementations.
We
wanted
to
hear
as
well,
but
the
one
for
AWS
is
exciting,
because
we
really
haven't
seen
like
this.
This
type
of
system
come
into
AWS,
yet
this
sort
of
declarative,
kubernetes
system
and
furthermore,
it's
exciting,
because
the
whole
cluster
API
is
set
up
so
that
each
cloud
can
also
have
its
own
subset
of
types
of
kubernetes
clusters,
so
underneath
AWS-
and
we
we
actually
have
a
word
for
this.
The
word
here
is
variant.
A
A
Another
example
could
be
maybe
a
private
subnet,
which
is
what
we're
going
to
be
running
today
that
only
allow
instances
to
be
accessed
via
a
bastion
server,
so
you
would
be
able
to
define
in-
and
this
goes
on
and
on
any
number
of
types
of
infrastructure
behind
the
scenes,
and
then
that
would
sort
of
bubble
up
to
this
broader
level.
Aws
implementation
that
then
shares
a
bunch
of
logic
and
a
bunch
of
API
definitions
with
this,
like
master
level
cluster
API
bit
that
we're
talking
about
here.
A
So
this
is
just
a
logical
way
for
engineers
to
start
to
structure
their
code
so
that
we
can
keep
a
lot
of
the
same
logic
that
comes
in
up
here.
The
cluster
API
trickled
in
all
the
way
down
here
and
all
these
different
variant
levels
as
well,
if
you're
an
object-oriented
programmer.
This
is
just
classic
inheritance
pattern.
Just
re
implemented
in
the
form
of
go
and
kubernetes
scripting,
I
guess
I
should
say
software
I
got
in
trouble
for
using
the
scripting
word
once
anyway.
A
That's
how
the
cluster
API
is
structured,
and
if
we
go
back
to
our
repo
here,
we
are
now
looking
at
the
AWS
implementation
of
the
cluster
API
that
we
just
looked
at
a
moment
ago.
Okay,
so
that's
all
everything
fits
together.
Now,
let's
actually
look
at
running
this
thing
and
creating
a
cluster
in
Amazon.
A
So,
to
start
off,
let's
look
at
my
Amazon
cons
and
was
just
validate
that
we
are
in
fact
sort
of
like
starting
at
scratch
here
or
as
close
to
scratch,
as
we
can
get
it
I
like
to
keep
my
us
west
to
as
empty
as
possible.
This
is
sort
of
like
my
magical
region
that
I
always
do
demos
and-
and
the
first
thing
I
do-
is
I
sort
of
check
this
resource
overview
here,
and
this
looks
good
I
know:
I
have
three
security
groups:
I
gotta
keep
for.
A
Like
my
own
personal
reasons,
a
handful
of
snapshots
in
case
I
ever
need
to
get
back
to
some
FEMEN
that
I've
worked
on
for
other
projects
and
then
zero
running
instances.
It's
like
the
big
one
I
want
to
check
for
so
we
know
we
don't
have
any
instances.
You
know
we
don't
have
any
elastic,
IPS
or
load
balancers,
and
we
know
we
expect
three
security
groups.
Yes,
these
all
look
like
the
security
groups.
I
want
and
let's
go
check
our
VP
see
we
should
have
one
which
is
my
default
default.
A
Vp
see
ya,
okay,
so
for
all
intensive
purposes.
This
is
a
fairly
empty
region,
meaning
we
don't
have
any
infrastructure
deployed
here,
or
at
least
any
infrastructure,
that's
relevant
to
the
cluster
API
that
were
going
to
be
talking
about
today.
So
if
you
go
back
to
the
hack
in
D,
and
then
this
cluster
API
and
cluster
API
API
bets
are,
this
is
the
broader
level
cluster
API.
This
is
the
AWS
implementation,
which
is
exciting
Sean
Smith
says
so
many
public
keys.
Yes,
I
have
a
lot
of
public
keys.
A
It's
a
it's
kind
of
a
personal
reason:
I
take
security
very
seriously
and
I
have
small
scripts
that
help
me
switch
public
he's
often,
but
yeah
I
do
have
a
lot
of
public
keys.
Don't
tell
anybody
anyway,
if
you
go
here,
I
lost
my
train
of
thought.
Where
was
I?
A
Oh
here's
the
go
to
draft
designs.
If
you
go
here,
we
can
pull
up
the
AWS
cluster
API
repo,
and
we
want
to
go
right
into
this
Doc's
directory
here.
So
our
friends
at
the
AWS
cluster
API
working
group
have
been
working
diligently
on
this
getting
started
and
this
is
like
I
can
just
like
feel
Chuck
and
everyone
cringing
right
now.
This
is
the
cluster
API
provider,
AWS,
quick
and
dirty
getting
started
guy.
That
I
was
able
to
get
up
and
right
here
today
that
we're
gonna
get
up
and
running
right
now.
A
A
If
you
need
to
know
why
that
is
feel
free
to
ping
me
afterwards,
because
we
clearly
need
to
go
out
and
get
a
beer
and
talk
about
that
face
to
face
a
set
of
AWS
prevent
credential
sufficient
to
bootstrap
the
cluster,
see
bootstrapping
AWS,
Identity
and
Access
Management
with
cloud
formation,
so
I
do
not
have
actually,
let's
see
if
I
can
do
this,
how
do
I
want
to
do
this?
I'm,
not
gonna,
show
you
all
my
my
bash
profile
here.
A
But
if
you
do
look
at
my
bash
profile,
it's
a
beautifully
written
bash
profile
and
I
have
a
lot
of
these
export
commands
defined
in
there.
Let's
say
things
like
AWS
region
and
that
is
set
to
us,
West
you
in
fact,
if
I,
let's
see,
if
I
can
do
this,
and
are
you
gonna
like
make
sure
I,
don't
fat
finger,
this
commander,
I'm
gonna
show
the
world
all
of
my
secrets
grab
eye
region.
A
It
says
an
AWS
I
am
role
to
give
the
cluster
API
control
plane
more
on
that
a
little
bit
later,
we
need
mini
cube,
and
this
is
a
really
interesting
dependency
here
and
I'm
gonna
spend
a
little
bit
of
time
talking
about
why
we
have
this
dependency
and
I
might
even
draw
another
diagram
I'm
sure
yet,
but
we
had
a
proposal
come
out
of
if
anybody
can
find
this
and
drop
it
in
the
Haakon,
V
dowdy
handy
I,
think
was
Jessica
Chen
at
Google
wrote
a
proposal
that
came
through
said
cluster
lifecycle,
the
cluster
API
folks.
A
That
basically
said
in
order
for
us
to
bootstrap
a
kubernetes
cluster.
We
think
it's
a
good
idea
to
create
this
sort
of
ephemeral,
bootstrap
cluster
first,
that
can
run
either
on
your
local
system
or
somewhere
else.
Josh
says
the
secret
access
key,
no,
not
going
to
grep
for
AWS
secret
access,
key
sorry
Josh,
but
we
would
have
this
ephemeral
bootstrap
cluster,
that
we
can
then
use
to
actually
deploy
controllers.
That
would
then
create
a
cluster
somewhere
else.
A
So
I
am
going
to
draw
a
diagram
here
because
I
feel
like
this
is
important
to
make
sure
we
get
the
language
right
here.
So
last
diagram
I
promised
folks.
So
here
is
your
local
workstation
and
then
you
know
I'm
gonna
draw
like
a
little
terminal
there.
That's
your
terminal
sign
for
your
local
workstation
and
on
your
local
workstation.
You
ultimately
want
to
end
up
creating
resources
down
here.
In
today's
example,
it's
going
to
be
AWS.
A
A
After
mini
cube
is
up
and
running,
we're
then
going
to
deploy
resources
to
mini
cube
after
those
resources
are
then
up
and
running
in
mini
cube.
Those
resources
can
then
go
and
actually
create
things
in
Amazon
behind
the
scenes,
and
that's
where
we
get
a
list
of
these.
You
know
Amazon
resources
down
here,
and
this
could
be
everything
from
a
load
balancer.
A
It
gives
us
sort
of
a
launching
a
launching
ground
here
to
deploy
these
small
bits
and
reuse
them
not
only
as
running
and
controllers
here,
but
also
is
running
as
controllers
here
as
well.
So
that's
why
we
have
this
dependency
on
mini
cube
and
it's
interesting
because
you
really
don't
even
have
to
use
mini
cube.
Really.
All
we
need
is
some
sort
of
kubernetes
cluster
that
we
can
deploy
our
controllers
and
just
point
those
controllers
to
whatever
Amazon
account
we
want.
So
in
theory,
this
could
be
a
Raspberry
Pi.
A
A
It's
Turtles,
all
the
way
down.
Justin
says:
I
have
a
good
use
for
raspberry
pies,
now
rat
okay.
So
let's
actually,
let's
actually
get
this
mini
cube
cluster
up
in
writing
as
soon
as
possible.
So,
let's
go
back
to
our
our
docks
here
and
talk
a
little
bit
more
about
them.
So
after
we
have
mini
cube,
we
of
course
need
queue
Bechdel,
so
we
can
do
all
the
things,
and
then
we
have
some
optional
commands
here.
You
know:
go
and
JQ
and
homebrew
and
the
AWS
command-line
tool.
A
Sean
says:
yeah
I
have
a
Raspberry
Pi
kubernetes
cluster
and
Justin
Lutz,
pretty
nice
to
play
around
with
yeah.
So
that
would
be
a
really
great
example.
It's
like
what,
if
you
just
had
your
magic
Raspberry
Pi,
that
you
carry
it
around
with
you
that
had
all
of
the
cluster
API
stuff
on
it
and
you
could
just
like
sit
down
anywhere
hook
that
thing
up
to
Wi-Fi
and
then
poof
like
you,
you
use
that
cluster
to
create
a
cluster.
A
That's
a
really
cool
fun
example
that
I
think
would
be
handy
just
having
like
your
teeny,
tiny
Raspberry
Pi,
that
sort
of
spawns
itself
out
into
the
cloud
into
these
huge
amount
elastic
clusters.
Let's
see
we
have
to
dig
Olive.
One
thing
that
I
noticed
about
one
of
the
implementations
is
that
it
uses
the
CA
key
material
of
the
mini
cube
designed
to
join
the
tokens
for
the
nodes
at
the
new
clusters.
We
should
make
sure
to
decouple
those
things
from
mini
cube.
Yes,
that's
a
really
great
point.
A
Let
me
go
back
to
my
dock
camera
here,
see
we're
only
45
minutes
in
and
I've
literally
just
been
ranting
for
the
past
45
minutes.
We
actually
have
an
ear
and
any
code
yet
I
feel
like
that's.
Okay,
though
this
is,
this
is
a
very
important
subject
for
me,
so
we
can
really
go
off
in
the
weeds
here
anyway.
A
Dig
Olive
brings
up
a
good
point,
which
is
we
effectively
I
wish
I
had
like
a
different
color
pen.
You
see
I
wonder
if
this
is
a
different
color
time
yeah,
it
kind
of
is
so.
We
have
to
kubernetes
clusters.
This
is
the
AWS
cluster.
That's
all
of
the
wonderful
AWS
infrastructure,
and
this
is
just
mini
cube,
running
on
some
sort
of
hypervisor
in
my
case
VirtualBox.
But
the
point
that
dig
all
that
brings
up
is
that
there
are
going
to
be
a
subset
of
TLS
certs
and
certificate
information
for
each
of
these
clusters.
A
A
Okay,
so
Justin
says
Shawn
I
just
got
one
set
up
based
on.
It's
got
handsome
in
his
blog
I'm,
assuming
Justin
talking
about
his
raspberry
pie.
Duffy
coolly
says
the
cluster
CA.
Would
be
defined
in
the
cluster
level
right,
yes,
that
is
well,
it's
no
I,
don't
think
we
actually
define
the
cert
material
explicitly
Duffy.
Somebody
fact
check
me
on
that.
Keep
me
honest,
but
I'm
pretty
sure
that's
the
way
it
works.
Shawn
Smith
says:
okay,
I,
think
Shawn
and
Justin
are
talking
about
raspberry
pies.
A
You
guys
have
fun
talking
about
it
as
great
pies,
okay,
so
anyway,
a
good
point
to
go
out.
Olive.
Let's
go
back
to
my
screen
here.
So
I
think
we
have
all
of
our
requirements
in
place
on
my
local
system
and
we'll
start
off
with
a
mini
cube
command
here
in
a
second.
But
let's
just
go
ahead
and
keep
going
through
these
docks
to
make
sure
we
didn't
skip
anything.
It
says,
okay,
so
you
want
to
grab
the
latest
cluster
cuddle
release.
So
I
guess
the
the
interesting
history
here
for
a
cluster
cuddle
is.
A
We
were
on
an
open
source
call
earlier
this
week
and
I
had
noticed
in
the
notes
Jason,
which
is
D
tie
beer
here,
Jason
tiberias
had
mentioned
that
the
the
cluster
AWS
implementation
was
ready
for
folks
to
use.
So,
of
course,
I
was
like
okay.
That
means
we're
gonna
demo
on
TGI
can
on
Friday,
and
here
we
are
so
anyway,
the
two
binaries
that
we
released
three
days
ago.
So
that
gives
you
sort
of
an
idea
of
how
new
this
thing
is:
are
the
Linux
and
the
one
on
Linux
I'm?
Sorry?
A
Let's
see
if
I
can
find
it.
Cluster
aw
Adam
see
if
I
can
find
a
better
pronunciation
of
this
cluster
awesome,
I,
don't
even
know
anyway,
we
can
do
cluster
aw
s
Adam
help,
and
then
you
can
see
that
it's
just
good
old
Cobra
commands
and
we
have
alpha
which
we
can
run
and
then
I'm,
assuming
underneath
alpha.
We
have
a
bunch
of
alpha
commands,
yeah,
bootstrap
and
ec2
and
stuff,
but
basically
this
whole
thing
is
in
place
and
it
creates
some
cloud
formation
for
us
that
helps
us
bootstrap.
A
The
I
am
rules
for
our
cluster,
so
if
we
actually
run
the
command
that
it
says
so
in
the
documentation
here
which
I've
already
run
it's
this
cluster
AWS
Adam
alpha
bootstrap
create
stack
command,
it
runs
the
sort
of
item,
potent
cloud
stack
creation
and
that
actually
was
really
fast.
I
just
mean
it,
and
you
can
see
it's
outputted,
a
ton
of
instance,
profile
and
managed
policy
and
roll
information
and
then
ultimately,
a
user
down
here
at
the
bottom
that
we're
going
to
use
to
authenticate
with
programmatically
as
we
move
forward,
and
it
did
not.
A
It
did
all
this
through
cloud
formation
and
if
this
was
the
first
time,
I
ran
it,
it
would
have
taken
a
you
know,
30
seconds
45
seconds
or
so,
and
it
would
have
needed
all
of
the
AWS
credentials
defined
to
actually
go
and
create
new
I
am
information.
So
whatever
you
authenticate
with,
we
need
the
permissions
to
then
create
more
permissions,
so
a
decent
says:
cluster
Ross
enum
kind
of
like
it.
A
Okay,
let's
go
back
to
Google
Chrome
here,
okay,
so
we've
ran
that
and
then
we
can
actually
go
and
if
we
go
into
on
the
ec2,
actually
not
easy
to
I
want
CloudFormation
up
here
at
the
top.
You
can
actually
see
we
have
this
cluster
API
provider
and
then
this
all
matches
what
we
just
saw
in
my
terminal-
and
you
can
see
that
it
all
is
create
complete.
A
So
we've
created
all
this
already,
so
that
is
done
in
out-of-the-way
and
if
we
go
back
to
our
documentation
here,
we
can
scroll
past
to
this
SSH
key
pair
snippet
I
feel
free
to
come
in
and
read
it
if
you
need
to
actually
create
a
new
key
pair.
As
mentioned
earlier,
I
am
insanely
over
cautious
about
my
SSH
key
pairs.
So
I
have
already
done
all
of
the
CGT
things
behind
the
scenes
that
folks
at
home,
don't
need
to
worry
about.
A
It
will
just
magically
work,
but
if
you
are
doing
this
for
your
your
a
first
time,
there's
good
instructions
here
for
you
to
get
a
new
key
pair
and
actually
create
that
in
Amazon,
so
scrolling
down
to
setting
up
to
mini
cube
for
folks
at
home,
who
have
never
set
up
many
queue
before
there
is
this
instructions.
Here
you
can
go
and
set
up,
but
the
TLDR,
if
you're
running
on
the
Mac
like
I,
am
here,
you're
gonna
want
to
have
VirtualBox
installed,
or
some
other
hypervisor
figure.
A
I
think
we
can
use
X
hive
is
another
one.
I
just
use
VirtualBox
because
I
already
had
it
for
some
other
projects
and
we
want
to
run
these
mini
cube
commands
after
we
get
the
mini,
cube
binary
into
our
path
so
to
run.
This
I'm
already
ran
this,
but
it's
gonna
run
it
again
for
good
measure.
We're
gonna
go
in
order
to
say
many
keep
config
set
and
that's
it.
A
It
doesn't
give
us
any
output
or
saying
that
it
can
fix
that
or
whatever,
but
we
do
need
to
set
it
to
one
point:
nine
point:
four
for
our
aggregated
API
server
dependencies
that
we
talked
about
a
little
bit
earlier
in
the
episode
m'dear
says
if
anyone's
wondering
I'm
the
powershell
on
linux
user
nadeer.
Do
you
really
run
PowerShell
on
Linux
like
like
I?
Don't
know
if
this
is
a
troll
or
not?
A
If
you
do
that's
really
impressive,
if
not
lol
I
choose
nodes,
how
is
that
even
possible
when
I
worked
at
Microsoft
I
tried
to
run
PowerShell
unlimited
Lennox
for
a
day
anyway,
we
can
run
this
mini
cube,
config
set
bootstrap
or
keep
admin
command,
and
all
this
does
is.
It
tells
us
that
we
do
want
to
bootstrap
our
mini
Q
cluster
with
cube
admin,
instead
of
what
other
other
mini
cube.
Bootstrapping
methods
are
available.
Okay,
so
we've
set
those
and
those
exist
they
persist.
A
So
now,
whenever
I
do
a
mini
cube
setup,
it
will
be
with
those
configuration
settings.
So
what
we're
gonna
do
here
is
we're
actually
going
to
bypass
and
inject
some
of
our
own
custom
documentation
here
to
give
folks
at
home
a
better
idea
of
what's
actually
going
on.
So
all
we're
really
gonna
do
is
we're
gonna
manage
the
mini
cube
stuff
ourself,
instead
of
letting
you
cluster
couple
do
it
so
that
we
can
actually
go
in
tail
logs
and
see?
What's
going
on
on
this
intermediate
or
bootstrap
cluster
bob
says
for
what
it's
worth?
A
X
is
deprecated
Oh,
interesting,
Bob
I
do
not
know
that
kind
of
a
bummer,
though
I
know
some
good
friends
of
mine
and
worked
on
the
x-height
project
for
a
while.
So
anyway,
what
we're
gonna
do
is
we're
gonna,
actually
start
a
mini
cube
cluster
now,
and
so
many
cube
is
pretty
handy
because
you
can
come
in
and
you
can
actually
like
run
the
mini
cube
command
and
it
will
like
do
all
of
the
creation
e
kubernetes
stuff
for
you.
A
So
if
you
never
ran
it
before,
you
can
actually
come
in
and
type
mini
cube
and
you
can
do
age
and
it's
just
another
Cobra
command,
like
all
the
other
ones.
We
see,
and
you
can
see
we
have
all
of
these
sub
commands
here.
The
one
that
we're
going
to
use
is
start-
and
it
just
says,
starts
local
kubernetes
cluster
and
because
of
those
two
config
lines
we
defined
earlier,
we
know
we're
gonna
get
kubernetes
cluster
of
type.
A
A
These
ones
here
I
was
like
what
I
did
I
actually
I
was
like
tonight.
Actually
I
cut
my
access
keys
out
on
me
on
the
screen
here
and
yes,
you
can
see
here.
It
says
example.
At
the
end
negotiation
example
key
here,
so
it
looks
like
they
probably
started
out
as
real
ones
and
then
we're
changed
somehow
and
are
now
fake
ones.
A
Panel
here
we
can
see
that
mini
cube
is
now
running
and
is
just
starting
kubernetes
for
us
Leonardo
says:
maybe
use
AWS
vault
not
to
expose
credentials
they're
from
the
AWS
dots
yeah
they're,
all
from
the
Amazon
Docs
anyway,
I
see
what
I
see
what
they're
saying
that
Amazon
Doc's
are
where
those
were
generated
from
I
was
talking
to
chuck
about
this
before.
Sometimes
it's
like
people
say
things
and
it's
like
ten
seconds
ago
and
my
my
world.
So
it's
like
what
what
are
we
talking
about?
A
Oh
yeah,
this
is
just
from
like
ten
or
fifteen
seconds
ago
anyway,
so
many
cube
is
up
and
running,
and
the
cluster
cuddle
command
is
gonna,
be
smart
enough
to
actually
look
and
see
if
the
mini
cube
is
up
and
running.
If
we
pass
a
flag
to
it
and
we'll
just
try
to
use
an
existing,
keep
communicating
with
this
mini
cube,
and
so
we've
modified
the
cluster
cuddle
command
a
little
bit
and
I've
added
that
command
here
right
here,
cluster
cuddle
command
is
the
one
command.
A
This
is
like
the
Big
Mama
command,
we're
gonna
be
using
later.
That
should
actually
get
our
cluster
up
and
running
so
anyway.
Go
me
back
here.
A
This
is
now
set
up
and
what's
cool
is
you
can
actually
do
a
mini,
cube
SSH
and
you
can
see
like
we
get
some
cool
like
terminally,
ASCII
art
here
and
you
can
actually
see
if
we
go
to
Etsy
kubernetes
we're
on
a
virtual
machine
now-
and
here
is
our
admin
comp,
which
is
a
really
really
great
sign
that
this
kubernetes
cluster
has
in
fact
been
bootstrapped
with
mini
cube.
So
if
we
exit
out
of
here,
it
should
have
written
the
mini
cube
context
directly
to
my
cute
config.
A
So
to
validate
that
I
can
do
hang
it
knows
and
if
it
works,
it
should
just
say
mini
cube.
Yeah
name
is
equal
to
many
cube,
so
we're
good
to
go
and
we're
running
version,
1.9
point
four
and
let's
just
go
ahead
and
test.
If
we
do
take
it
Poe
no
resources
found
yet
so
now,
let's
form
you
formulate
our
cluster
command
and
let's
talk
a
little
bit
about
that.
Also
I'm
like
losing
my
voice
again,
so
here
we're
gonna,
run
cluster
cuddle,
create
cluster
we're
gonna.
A
Just
pump
the
verbosity
all
the
way
up
to
four
right
out
of
the
gates,
we're
going
to
tell
it
provider
equal
to
AWS.
This
is
another
gripe
of
mine
for
the
cluster
API
folks.
If
we're
going
through
the
trouble
of
writing
implementations
that
are
unique
to
a
cloud,
why
do
we
need
to
pass
in
provider
equals
AWS
here?
A
A
So,
let's
just
only
CD
into
this
directory
and
let's
remove
this
out
directory
and
let's
go
back
and
now
we
can
paste
our
full
command
here
generate
without
SH
and
you
can
now
go
back
to
where
we
were
and
we
can
go
into
out
Oh.
A
A
Ok,
there
we
go
now
we
can
go
into
out
and
we
can
see
that
we
have
these
three
EML
files.
Let's
see
what
folks
are
saying,
Duffy
says,
plus
one
on
it's
awesome,
cool,
so
yeah.
So
let's
take
a
look
at
this
cluster
Y
animal.
So
remember
earlier
in
the
episode
I
talked
about
how
we
would
be
to
find
all
these
different
yamo
objects
and
then
applying
them.
This
is
the
actual
declaration
of
our
kubernetes
cluster
that
we
want
to
create
in
amazon.
A
So
if
we
cat
out
cluster
DML,
it's
very
simple:
we
give
it
a
name
which
is
test
one.
We
define
this
clustering
network
field
here
and
then
we
have
services
and
we
give
the
sign
your
blocks
and
it
looks
like
this
one
is
the
I
think
it's
calico
default
cider
here,
one
nine,
two,
one,
six,
eight
with
a
slash.
Sixteen
we
give
it
the
service
domain,
cluster,
local,
which
I
think
the
service
domain
is
like
the
the
DNS
bits
for
a
cluster,
in
this
case
we're
doing
cluster
local.
A
So
we
don't
need
DNS,
and
then
we
have
this
provider
config
here
now
the
provider
config
is
exciting,
because
this
is
the
part
that
is
actually
unique
to
Amazon.
Gke
is
gonna,
have
its
own
specific
type
of
private
provider.
Config
Amazon
will
have
its
own
type
of
provider
config,
and
these
are
going
to
be
all
of
the
the
unique
Amazon
bits.
In
this
case
we
have
an
API
version.
We
have
a
kind.
This
is
for
the
the
API
records
that
we're
defining.
A
So
you
could
have
a
provider
config
with
a
nested
value
of
a
certain
API
version,
which
is
defined
here
in
here
so
we'll
be
able
to
plug
in
different
API
is
here
for
different
reasons
and
that's
how
we,
as
a
group,
have
decided
to
solve
the
the
abstraction
as
a
whole
for
creating
a
unified
cluster
object,
while
still
being
able
to
find
bits
that
are
specific
to
your
cloud.
Duffy
says
own
the
odd
the
URL.
A
Doesn't
the
URL
doesn't
Parrs,
plus
one
on
github.com,
slash,
99designs,
AWS
vault
managed
to
URL
to
fix
okay,
Duffy
I
have
no
idea
what
you're
saying
Shawn
looks
like
it
worked
for
him,
so
whatever
you're
doing
looks
like
it
worked
for
Shawn
anyway,
coming
back
here,
that's
clustered.
Let's
now
look
at
machines
animal,
so
here
I
think
we
really
only
need
to
focus
on
this
part.
A
So
here
we
have
a
set
of
items
which
this
is
a
kubernetes,
a
little
level
construct
that
allow
you
to
define
like
a
set
of
resources,
and
here
we
have
an
API
version
which
is
K
dot,
IO
dot
v1
alpha
one,
it
kind
is
equal
to
machine.
So
this
is
a
depth.
This
is
like
the
actual
API
part
of
the
cluster
API,
and
this
actually
represents
a
single
in
ec2,
and
here
you
can
see
that
we
have
a
generate
name
and
we
have
some
labels
that
we
can
use
for
arbitrary
reasons
downstream.
A
And
then
we
have
like
our
very
familiar
spec
for
a
machine,
and
in
here
you
notice.
We
have
versions
and
underneath
versions
we
have
cubelet
and
control
plane.
So
this
is
the
actual
version
of
the
cubelet
systemd
unit
that
we
want
to
run,
and
this
is
the
actual
version
of
all
of
the
kubernetes
control,
plane,
components
we
want
to
run,
and
then
we
see
the
provider
config,
as
we
saw
in
cluster
that
actually
defines
this
specific
machine
type
information.
So
here
we
give
it
a
key
name
as
well.
We
tell
them.
A
We
want
to
use
the
I,
am
instance
profile
that
we
generated
earlier,
and
we
want
to
tell
it
that
it's
a
t2
medium
and
the
whole
point
of
this
is
for
it
to
be
declarative
so
that
you
would
be
able
to
come
in
and
change
this
T
to
medium
to
T
to
large,
either
yourself
or
from
like
an
auto
scaler
there's
a
perspective
or
some
other
piece
of
software,
and
then
the
controller
would
go
and
reconcile
that
moving
forward.
So
those
are
our
machines,
my
animal
and
last
but
not
least,
somebody
asked
about
this
earlier.
A
A
We
have
these
deployment
here
and
you
can
go
through
and
you
can
deploy
various
parts
of
the
the
provider
components
here
for
generating
your
cluster
anyway.
I'm
gonna
clear
that
before
anything
else,
Chuck
says
I
need
to
drop
a
good
luck
with
the
rest
of
the
demo.
Thanks
for
giving
a
cluster
API
provider,
AWS
has
been
so
Chuck
like
we're,
gonna
cat,
your
creds
out
on
screen
and
then
we're
just
gonna
mic
job
thanks
Chuck.
So
anyway,
let's
go
ahead
and
let's
go
and
let's
run
our
cluster
cuddle
create
command.
A
So
we'll
do
this
and
we
won't
run
the
command
because
we
still
have
to
get
our
mini
cube.
Config
so
go
up
on
this
and
let's
do
cluster
cuddle,
create
cluster
yeah,
yeah
I
know
Jason.
If
you
have
access
to
AWS,
can
you
nuke
those
creds
in
about
five
minutes
when
I
get
done?
That
would
be
helpful
things
so
anyway,
yeah
try
to
think
what
we
want
to
do
here.
Okay,
so
yeah,
we're
gonna
run
this
cluster
cuddle
command.
A
We
want
to
run
our
help
again,
so
I
can
get
the
name
of
this
really
long
flag,
which
is
this
one
here.
This
existing
bootstrap
cluster
cube
config
string
and
we're
going
to
pin
that
to
the
end
of
this
command
here
and
all
we're
gonna
do
is
set
that
equal
to
a
user's,
slash,
Nova
slash
just
my
default
cube,
slash
config,
okay,
so
this
should
do
a
cluster
cuddle,
create.
What
did
this
say
for
us?
A
A
What's
going
on
so
what
we
have
here
is
our
controller
logs,
so
let's
do
K
logs
cluster
API
controllers
will
follow
them.
It's
going
to
yell
at
us
and
tell
us.
We
need
to
define
a
container
so
we're
gonna
get
let's
go
ahead
and
do
the
cluster
controller,
and
then
we
can
look
at
the
machine
controller
as
well.
So
as
we
do
that
we're
to
do
minus
C
for
container
the
cluster
controller
and
we
can
actually
go
and
see-
and
please
don't
let
there
be
secrets
in
this
log
as
well.
A
A
You
know
here
we
go
not
gateways,
ok,
so
test
one
that
is
pending.
So
if
we
go
back
to
our
logs,
we
can
see
that
yes,
we're
creating
this
nat
gateway
here,
which
is
what
we
just
saw
in
the
Amazon
console.
So
this
controller
is
actually
going
and
setting
up
the
cluster
level,
infrastructure
and
I
bet.
If
we
go
in
and
we
can
do
AWS
cluster
controller,
it
changes
this
to
AWS
machine
controller.
A
We
can
actually
go
and
see
that
we
have
a
whole
separate
controller.
That's
actually
reconciling
machines
for
us,
Jason
I
know
I'm
picking
on
you
Jason.
There
are
no
secrets
in
the
logs
here
which
is
good
to
know.
Okay,
so
this
is.
This.
Is
the
machine
controller
that's
actually
going
and
creating
machines
for
us
in
Amazon?
So
the
whole
pattern
here
is:
we
have
this
mini
Q
cluster?
That's
creating
this
new
cluster
from
scratch
in
Amazon
for
us.
A
So
if
we
go
to
ec2,
we
can
actually
that
we
should
have
some
instances
coming
up
any
moment
now,
let's
check
our
logs
and
see,
what's
going
on
here,
unable
to
create
machine,
AWS
control,
plane,
req
in
one
minute
and
network,
not
ready.
So
this
is
this
is
our
reconciler
in
action
and
I
guess.
This
is
a
really
great
example
of
why
the
reconciler
pattern
is
so
handy
for
things
like
this,
because
we
can
start
to
bacon
things
like
automatic,
retries
and
reconciliation
detection
deltas
and
we
can
come
in
and
say.
A
Oh,
you
know,
let's
really
do
a
nice
advanced
audit
of
the
systems
in
place
and
go
and
make
sure
that
you
know
our
reconciler
is
smart
enough
to
go
and
ensure
whatever
infrastructure
we
defined
is
actually
in
existence
and
making
you
changes
as
needed.
So
you
can
build
in
all
kinds
of
cool
and
exciting
logic
here
for
checking
over
and
over
and
over
again
and
actually
reconciling
that
to
make
it
work.
A
A
Those
gateways
can
be
pretty
pesky
when
it
comes
time
to
actually
creating
them
and
the
whole
reason
we
need
to
create
them
in
the
first
place
is
because
we're
doing
this
sort
of
private
topology
implement
implementation
where
we're
creating
a
private
sub
subnet,
and
we
need
to
actually
not
into
it.
So
we
need
that
nat
gateway
in
place.
So
let's
go
back
to
our
V
PC
and
see
if
we
can
find
any
more
information
here.
Internet
gateway.
A
Okay,
so
it
does
say
it's
available,
so
I'm
wondering
if
our
controllers
are
just
magically
gonna
start
to
kind
of
recap
themselves
after
their
sweep
of
I.
He
was
one
minute
is
what
we
saw
earlier
happens,
so
that
would
be
really
handy.
I'm
not
going
to
and
see
I
wasn't
even
thinking
about
deleting
the
pods
and
we
kicking
it
manually,
but
I'm
just
gonna.
Let
the
controllers
do
their
thing
so
that
we
can
demonstrate
that
they
will
actually
come
and
create
infrastructure
as
needed.
A
A
You
ran
cluster
cuddle,
from
which
remember
we're
running
cluster
cuddle
here
in
this,
and
this
is
sort
of
hanging
for
these
components
to
come
up
and
get
get
running
in
an
Amazon
and
it'll
want
to
when
it's
finally
done
will
echo
out
this
cute
config
that
will
allow
us
to
actually
go
and
interact
with
the
cluster
in
Amazon.
So
we
have
this
cube
config
which
lives
in
this
directory
on
my
file
system,
and
then
we
also
have
the
cube
config
that
lives
in
this
directory.
A
Now
this
one
is
the
one
that's
talking
to
mini
cube,
and
this
one
is
the
one
that
it's
actually
going
to
talk
to
Amazon
after
we
get
up
and
running,
which
is
where
the
whole
KTX
in
cube
context.
Ething
comes
into
play
as
we're
going
to
be
looking
at
different
clusters.
Also,
it's
important
to
note
that
cluster
cuddle,
if
we
didn't
do
the
mini,
keep
start
on
our
own
and
actually
pull
up
these
logs.
A
So
we
can
see
what
was
going
on
the
cluster
cuddle
command,
would
ephemerally,
create
a
mini,
cube
cluster,
deploy
to
it
and
then
delete
it
afterwards.
Upon
that
delete,
we
would
actually
lose
all
the
log
information
that
we're
looking
at
here
to
debug
this
whole
system
to
keep
it
up
and
running.
So
it
doesn't
look
like
these
are
kicking
off
just
yet.
Let's
see
if
I
can
give
him
a
quick
nudge
here,
let's
kkp.
A
Oh,
this
is
me
being
malicious
I'm
going
to
nuke
this
pod
in
the
so
REE
kicking
off
what
it
needs
to
do
and
restarting
the
controller
logic
so
that
we
don't
have
to
wait
for
it
to
riku,
so
how
we
want
to
do
that
is
Kay
delete,
P,
oh
the
name
of
the
pod
BAM,
and
now
we
can
do
our
K
get
P
Oh
again,
grab
the
new
pod
name
and
we'll
actually
grab
our
logs
here
as
well.
For
let's
do
the
cluster
controller!
A
You
cluster
controller
attempt
you
to
acquire
leader
lease.
So
this
looks
pretty
good
and
this
all
this
code
that
we're
actually
looking
at
the
logs
for
is
what's
running
in
the
cluster
cuddle
or
the
cluster
AWS
API
provider
repository
as
well.
Okay,
so
yeah,
you
can
see
that
we've
been
able
to
Riku
it
manually
and
it's
actually
coming
in
and
making
a
lot
of
the
mutations
that
we
wanted
it
to
and
creating
a
lot
of
the
infrastructure
for
us.
So,
oh
we're
still
getting
this
error.
A
A
Jason
says
I
think
the
cluster
actuator
is
hitting
a
bug
related
to
easy
ordering
I,
don't
think
it'll
recover,
oh
interesting,
so
it
looks
like
we
found
our
first
bug,
which
is
exciting.
I
do
love
finding
bugs
so
yeah.
We
still
have
our
not
gateway
available
here,
which
is
cool,
so
I
wonder
if
we
can
actually,
let's
try
to
kick
this
off
one
more
time
and
see
I'm
going
to
Nick
this
see.
If
we
can't
sort
of
wreak
this
off
and
redeploy
it
and
see
what
happens
breather
way.
A
I
just
really
wanted
to
get
folks
looking
at
this
code
today
and
understanding.
Why
we're
doing
what
we're
doing
and
seeing
the
complexities
of
the
different
controllers
and
operators
and
understanding
the
whole
point
of
them
moving
forward?
It
would
be
really
rad
if
I
can
retake
this
really
quick,
so
I
can
actually
show
you
that
that
controllers
then
get
deployed
to
the
cluster
in
Amazon
as
well.
So,
let's
see,
if
I
can
do
this
in
like
the
next
five
or
so
minutes.
A
So
what
we
want
to
do
is
we
want
to
do
a
mini,
cube,
delete
and
we
want
to
come
back
here
and
we
can
clear
that
screen
and
then
we'll
do
a
mini
cube
starch
and
then
why
we're
doing
that
we're
gonna,
go
and
just
manually
delete
a
lot
of
our
infrastructure
here.
So
the
whole
point
of
the
cluster
API
declarative
infrastructure
is
to
sort
of
solve
this
problem
that
we're
actually
looking
at
right
now,
which
is
abandoning
resources
in
a
cloud.
A
A
And
let's
see
you
don't
have
permission,
don't
you
change
to
different
role,
maybe
I'm
hitting
some
sort
of
permission
thing
now,
but
anyway,
the
whole
point
of
cluster
API
is
so
that
we
can
bake
a
lot
of
this
retry
logic
in
so
once
these
controllers
are
healthy.
That
they'll
be
able
to
come
and
make
sure
that
this
type
of
stuff
doesn't
happen
for
us
moving
forward.
So
it's
just
a
matter
of
dialing
in
the
initial
creation
just
right
for
us
to
get
a
lot
of
the
value
for
having
this
reconciliation
loops
running
over
time.
A
So,
let's
see
we
don't
have
any
security
groups
here,
let's
go
check.
Our
V
PC
I
feel
like
the
V
PC
and
Amazon
is
always
like
the
one
thing
that
like,
if
you
can
get
rid
of
the
V
PC
you're,
pretty
much
getting
rid
of
everything.
So,
let's
see
what
happens
when
we
try
to
delete
this
I
bet
it's
gonna
yell
at
us,
because
I'm
in
that
gateway-
let's
see,
maybe
not,
we
might
get
it
first
right.
That
would
be
really
rad
if
we
did.
A
A
A
I
think
we
I
think
I'm
using
the
wrong
am
I
in
the
the
cluster
cuddle
provisionary
amyl,
which
I
don't
want
to
go
refund,
read
to
find
that
anyway,
because
those
secrets
are
in
there,
so
I
don't
want
to
pull
those
back
up
but
either
way.
I
think
this
is
a
good
stopping
point
for
the
episode.
We
can
look
a
little
bit
more
into
the
actual
pivot,
from
the
mini
cube
cluster
up
and
running,
but
I
just
wanted
to
sort
of
explain.
A
The
clustering
API
demonstrate
that
we're
getting
this
thing
up
and
running
in
Amazon
and
give
folks
an
opportunity
to
start
thinking
about
what
it
would
take
to
write,
machine
and
cluster
controllers
and
operators
to
actually
go
and
get
a
lot
of
this
stuff
done
and
I
just
wanted
to
really
poke
our
engineers
and
get
the
stuff
up
and
running
and
see
see
what's
been
going
on,
I'm
one
of
the
organizers
in
the
the
cluster
API
working
group
and
I
don't
get
to
spend
nearly
as
much
time
writing
code
as
I
used
to
so
I've
just
been
really
antsy
to
take
this
for
a
spin.
A
A
A
Paying
me
if
you
have
any
questions,
sweet,
cool
and
I'm
sure
this
is
not
the
only
TG
I
can't
we're
going
to
do
on
this
I
think
after
we
get
a
lot
of
the
declarations
stuff
up
and
running,
it's
going
to
be
really
exciting
in
other
parts
of
kubernetes
to
to
the
cluster
api
as
well.
I
think
the
one
that
I'm
personally
most
excited
about
is
hooking
up
an
auto
scaler
to
the
cluster
API,
where
we
are
actually
going
creating
machine
EML
files
as
needed.
A
Programmatically
we're
not
actually
going
and
doing
all
this
stuff
manually
like
we
were
doing
here,
Felipe,
says
on
a
similar
subjects.
Next
month,
I'll
present
that
velocity
comes
an
operator
I'm
working
on
to
deploy
CD
in
clusters.
I
hope
to
see
you
there
I
very
much
will
be
at
velocity
next
month,
so
I
would
love
to
get
together
and
learn
more
about
your
your
priority.
There
Philippe
that
sounds
really
rad
and
I
always
love
a
good
operator,
honestly
they're
just
fun
to
it
to
tinker
with
so
anyway,
thanks
for
joining
everybody.
A
A
You
could
have
controllers
that
check
CPU.
You
can
have
controllers
that
do
arbitrary
business
logic
or
you
know.
Maybe
we
need
to
run
this
one
cron
job
and
create
a
couple
of
GPU
instances
or
or
whatever,
and
this
all
becomes
now
more
of
a
programmatic
exercise
instead
of
an
actual
exercise
of
creating
nodes
and
dealing
with
like
the
Bashi
implementation
detail
that
this
whole
project
we're
looking
at
today,
attempts
to
solve
I
think
the
other
exciting
part
of
cluster
api.
That
a
lot
of
folks,
don't
realize
is
the
ability
to
test
infrastructure
afterwards.
A
If
we
live
in
a
world
where
everything
is
programmatic
and
we
can
create
everything
through
kubernetes
native
objects,
it
actually
becomes
super
easy
to
run
it
tests
against
it
and
then
get
a
status
back
and
validate
that
status.
Is
we
what
we
wanted?
So
actually
writing
infrastructure
unit
tests
for
the
first
time
is
now
possible
due
to
thanks
to
programs
like
the
one
we're
looking
at
today.
A
So
this
is
all
really
exciting
stuff
for
me,
as
an
infrastructure
engineer
and
a
lot
of
the
stuff
here
that
we're
looking
at
is
in
my
book
cloud
native
infrastructure
as
well.
So
this
is
all
really
really
great,
and
thanks
for
thanks
for
joining
everyone,
let's
see,
looks
like
people
in
the
chat.
Have
things
to
say,
Justin
says
bye,
Josh
says
thank
you
have
a
great
weekend.
Marco
says
thanks
for
an
awesome
episode
have
a
great
weekend.