►
Description
Join us for Kubernetes Forums Bengaluru and Delhi - learn more at kubecon.io
Don't miss KubeCon + CloudNativeCon 2020 events in Amsterdam March 30 - April 2, Shanghai July 28-30 and Boston November 17-20! Learn more at kubecon.io. The conference features presentations from developers and end users of Kubernetes, Prometheus, Envoy, and all of the other CNCF-hosted projects
Bringing Cloud Native Buildpacks to Kubernetes - Dr Nic Williams, Stark & Wayne
Applications deployed to Kubernetes will live for next 10 years. The Dockerfile might be useful for an initial developer to get their application compiled into an OCI and deployed to Kubernetes, but over the next 10 years many different parts of the organisation will want to fix, upgrade, and audit the applications.
Instead of Dockerfile, we will investigate the Cloud Native Buildpack (incubator) project that is based on the last 10 years of success at Heroku, and hundreds of Cloud Foundry installations around the world.
Buildpacks provide automatic discovery of your applications' dependencies, the latest security patches, and are a homogenous method for all your organisation's application to be built into OCI/docker images over the next 10 years, and for all the running images to be upgraded without developer intervention.
Please, no more bespoke, unmaintained Dockerfiles.
A
All
right,
this
looks
like
an
excellent
crowd.
This
is
exactly
the
group
of
people.
I
was
hoping
over.
There's
more
people
no
didn't
want
those
people
excellent.
Now
my
name
is
dr..
Nik
Williams
I
am
Australian,
which
normally
is
more
meaningful
when
you're
traveling
the
world,
so
I'll
be
more
specific.
I
come
from
Brisbane
and
but
I
do
want
to
introduce
briefly
who
I
am
and
and
and
then
we're
gonna
talk
about
build
packs,
which
are
super
important
and
I.
A
Look
forward
to
you,
hopefully,
learning
something
new
and,
more
importantly,
starting
to
use
them
in
your
organization.
About
seven
years
ago,
linkedin
tells
me
that
I
started
stock
and
wine.
We
did
Cloud
Foundry,
which
has
now
become
you,
know,
kubernetes
cloud
foundry
runs
and
kubernetes.
Obviously,
everyone
here
is
excited
about
communities
and
we
yeah
so
by
the
American
European.
We
do
a
little
bit
over
in
Australia
I
up
in
Brisbane
I've
got
excited
by
turning
all
my
old
raspberry
pies
into
something
useful.
A
So
if
you
come
and
meet
the
guys
at
Rancher,
they
have
a
thing
called
K
k3s,
which
is
like
it's
it's
full
kubernetes,
but
made
into
a
really
small
footprint,
both
a
disk
and
Ram,
really
cool
and
and
that
that
is
that
picture
is
growing
as
I
start
to
find
all
the
other
things
I
ever
bought
that
set
in
boxes
and
becomes
this
cluster
and
Angus
Lee's.
Put
me
onto
this
idea.
A
Personally,
it's
not
the
talk
I'm
going
to
give,
but
the
idea
of
having
a
product
owner
who
owns
the
outcome
of
an
experience
for
all
the
developers
is
really
valuable
in
something
we've
take
market
with
Cloud
Foundry.
For
last
you
know,
half
a
decade
and
and
I
really
hope
that
as
a
community,
we
come
to
that.
An
understanding
that
it's
not
just
toys
then
look
how
these
all
actually
work
at
all
that
we
want
to
own
an
experience
for
the
rest.
A
The
organization
and
one
of
the
things
that
to
segue
that
that
came
out
of
that
thing
with
Cloud
Foundry,
was
the
build
system
of
how
we
want
to
encourage
developers
to
build
things
which
led
to
what
we
now
have
in
the
CN
CF
incubation
as
cloud
native
bill
packs.
So,
let's,
let's
start
with
the
problem
and
then
we'll
come
back
to
what
is
Bill
Paxson
all
right.
So
every
app
over
you
know
10
years
has
this.
A
Obviously,
the
dependencies
the
application
uses
both
language
dependencies
and
OS
dependencies
they're
going
to
change
the
base
OS
that
you
build
on
whether
it's
CentOS
RedHat
Ubuntu
pick
another
one
that
will
change,
maybe
even
you'll
switch
from
one
to
the
other
because
they
become
less
important
over
time.
As
you
know,
scratch
containers,
it
may
be
the
ultimate
OS
for
most
of
us.
A
How
you
do
deployment
will
change.
I
know
there
are
people
here
still
using
Jenkins
for
good
reasons.
Many
others
have
gone
from
no
CI
to
some
CI
you'll,
keep
changing
what
is
CI
might
be
inside
companies
might
be
outside
communities,
it
will
keep
changing
and
the
entire
group
of
people
you
work
with
will
no
longer
work
here.
A
That's
the
next
ten
years
and
and
obviously
the
CVS
are
everywhere
and
they
get
cool
logos.
Now,
that's
what's
new,
we
always
had
CVS.
Now
they
have
cool
names
and
cool
logos,
which
I
think
is
perhaps
the
most
endearing
and
valuable
aspect
of
modern
day.
Cvs
is
their
awareness
that
you
haven't
rolled
them
out,
yet
you
didn't
know
they
existed
and
you
didn't
know
whether
you'd
own
anything
and
now
you
know,
there's
a
cool
logo
and,
and
you
have
to
roll
them
out.
A
But
how
so
we
want
to
separate
these
things
out
and
if
we
separate
these
things
out
in
our
organization,
it
makes
them
a
lot
easier
for
people
to
know
that
it's
my
responsibility
to
roll
out
that
CVE
and
not
the
nodejs
guy,
which
leads
the
problem.
So
one
way
we
might
separate
concerns
is
will
leave
the
developer
team
and
their
product
owner
in
charge
of
the
real-world
problem
changes.
Maybe
we'd
have
a
language
specific
group,
the
the
java
group,
the
figuring
out.
A
This
is
a
super
challenging
problem,
and
if
teams
going
to
need
to
look
after
this,
and
so
I
really
I
really
appreciate
zero.
Sharing
that
story,
amongst
all
the
other
keynotes,
perhaps
are
more
more
likely
is
the
developer
team
will
need
to
worry
about
programming,
language
things
that
make
sense.
I
mean
they
chose
libraries,
they're,
probably
going
to
be
the
ones
that
have
to
keep
track
of.
Do
I
upgrade
a
new
version
of
rails
and
new
versions
of
spring.
A
That's
gonna
be
on
their
to-do
list
and
the
rest
of
it
I
think,
is
more
likely
that
that
idea
of
a
platform
team
is
super
valuable
and
as
far
as
possible
from
the
dev
team,
the
dev
team
might
have
got
you
into
a
kubernetes
problem,
but
now
it
needs
to
be
extracted
away
and
and
and
someone
to
own
that
it
should
be
their
responsibility.
But
that
still
leaves
us
with
the
build
aspect.
A
A
This
is
the
simplest
one
that
attempts
to
build
a
ruby
application
which
I
pick
in
part
because
I
kind
of
like
Ruby
and
B
you've
never
seen
it
before.
Those
people
are
not
Ruby
developers,
but
they
all
everyone's
looks
a
bit
like
this
from
Ubuntu
I,
don't
know
which
one
you
used
at
the
time
that
you
built
this
image.
This
Draca
file
no
longer
tells
me
what's
in
production.
It
just
tells
me
what
might
have
been
used
when
it
was
built
whenever
that
happened.
A
Similarly,
apt-get
update
and
apt-get
install
Ruby
really
depends
on
what
the
first
line
meant
depends
on
what
you
get
next
bit
of
a
segue
friends:
don't
let
friends
and
stall
programming
languages
from
packaged
systems?
You
know,
you've
got
some
good
friends
out
there,
just
guide
them
away
to
the
light.
Now
so
you'll
see
a
lot
of
docker
files.
Well,
they'll
they'll
build
their
programming,
language
from
source
and
it'll
be
check,
summed
and
there
that's
equally
that's
wonderfully
awesome
and
probably
highlights
more
the
problem
of
now
whose
responsibility
is
it?
A
Congratulations,
you've,
copied
and
pasted
from
Stack
Overflow
how
to
build
Ruby
from
source
who's
going
to
find
the
instructions
you
know
who's
going
to
own
and
operate
that
over
time
and
and
then
later
on.
There's
the
other
things
about
you
know,
ruby,
gems
and
and
and
your
programming
language
I
have
a
version
of
this.
That's
specific
and
the
problem
I
have
with
this
is
that
of
separation
of
concerns.
A
problem
I
have
is,
it
seems
more.
A
Obviously,
this
is
the
developers
problem
when
all
through
this
docker
file
is
an
entire
organizations
set
of
problems
that
that
hole.
All
these
issues
unfortunately
meet
in
that
docker
file
and
and
I
have
a
concern.
So
another
version
of
the
concern
is
in
my
mind,
this
is
what
docker
hub
looks
like.
A
You
go
to
docker
hub.
You
can
get
your
own
artisanal
docker
image,
sure
you
got
it,
but
you'd
have
no
idea
how
to
build
it
exactly
the
same
error
again,
and
so
what
I
think
I
found
is
docker
was
wonderful.
Docker
brought
a
whole
new
way
of
thinking
about
infrastructure,
about
deploying
about
ownership
and,
and
it
generated
at
least
two
problems.
One
was
scheduling
like,
which
is
what
we're
here,
we're
here
to
celebrate
a
solution
to
the
docker
problem
of
how
we
get
these
containers
running
in
places
and
everything
layered.
A
On
top
of
that
that
solution,
unfortunately,
wasn't
Dockers
and
now
dr.,
as
a
company,
did
not
benefit
from
for
much
of
what
they
sort
of
spawned
and
I.
Probably
shouldn't
laugh.
Oh
good
said
the
other
problem
they
created
was
the
docker
file
with
which
I
just
introduced,
and
everyone
here
is
excited
about
solving
the
first
problem.
There's
no
conference
for
solving
what
do
we
do
about
docker
files?
I,
don't
know
that
the
problems
well
understood.
I,
don't
think.
A
A
We
updated
everything,
host
machines,
containers,
images,
etc,
and
how
are
you
going
to
promise
to
do
that
for
next
decade
and
so
cloud
natty
bill?
Pacs
is
an
incubator
project
inside
the
CN
CF?
It
is
one
of
the
ways
you
might
like
to
do
it
I'm,
partly
an
academic
side,
not
really
into
selling
the
thing.
That's
like
I'm
gonna
tell
you.
This
is
the
only
solution.
A
I
care,
a
lot
about
you,
understanding
the
problem
and
being
aware
of
it
I'd,
like
you
very
much
to
try
cloud,
maybe
bill
packs
and
then
toolchain,
and
to
start
to
know
that
there
are
solutions
and
then
you
can.
You
know,
as
an
organization
will
figure
out
the
path
you
want
to
take.
Nothing
is
easy,
but
I
do
care
about
you
being
aware
that
this
problem
exists.
If
unattended,
there
was
a
lot
of
content
on
the
internet
about
cloud,
maybe
Bill
packs.
Fortunately
my
voice
is
not
the
only
one
you
can
have
to
listen
to.
A
There
is
a
lot
of
the
people
that
work
on
climate
bill,
packs,
Heroku
and
pivotal,
especially
I've
general,
a
lot
of
content.
It's
a
lot
they're
feeder
to
to
learn
after
this,
but
I'll
give
you
a
bit
of
that
the
rundown
and
then
I'm
going
to
do
some.
Some
super
demos
and
that's
sort
of
tautology.
All
demos
are
super,
whether
they
work
or
don't
work
just
watching
a
demo.
A
It's
exciting,
so
I
want
to
give
you
that
that
thrill
of
live
demos,
so
the
gist
of
buildpack,
says
it
as
a
as
a
product
as
a
project
is
three
parts.
What
is
the
bill?
Pacs,
which
is
kind
of
in
the
name,
so
a
bill
pack
is
an
isolated
set
of
functionality.
That
knows
how
to
do
something
possibly
to
your
application.
So
your
application
might
be
a
ruby
application.
It
might
be
Java
source
code,
it
might
be
a
pre
compiled
Java
jar
file.
A
A
This
cloud
native
bill
pack
is
the
future
of
Cloud
Foundry
and
probably
the
future
of
the
pivotal
VMware
go-to-market
around
our
build
build
systems.
The
next
part
is
the
Builder.
The
Builder
is
is
how
we
pull
this
together
and
what
are
the
selection
of
build
packs
we
use
in
our
organization,
so
you
actually
get
to
new.
Have
a
nuanced
selection
of
bill
packs
that
your
company
is
going
to
use.
A
It
also
means
that
all
your
Java
applications
can
have
a
build
pack
that
no
one
else
is
using,
so
it
allows
you
to
across
the
board
make
sure
all
the
next
versions
of
images
that
go
out
have
some
mutually
agreed.
You
know
monitoring
agent
security
features
whatever
now
without
having
to
go
and
find
all
those
docker
images.
Those
Dockers
and
the
lifecycle
is
the
behavior
that
binds
us
all
together
and
life
cycle.
A
Is
that
so
libraries
I'm
gonna,
show
you
it
wrapped
up
in
a
couple
of
tools
and
and
you'll
get
the
gist
of
what
that
is,
and
what
we
get
in
the
end
is
still
a
docker
image.
Well,
what
we
start
to
refer
to
as
an
OCI
just
to
make
a
little
bit
more
vendor-neutral
that
it's
pretty
hard
to
say
OCI,
without
also
wanting
to
say
docker
image.
Just
so
we
all
agree
what
we're
building
the
benefit
of
the
produced.
A
Docker
image
is
that
because
we
know
how
it
got
built,
we
know
what
we
can
do
to
it,
independently
of
who
built
it.
You
can't
say
that
about
a
normal
docker
file,
because
the
docker
file
might
be
from
the
bun
too,
but
then
it
might
do
some
crazy
things
that
aren't
what
we
call
ABI
compatible,
and
so
you
don't
know
that
you
can
just
replace
the
layers,
which
is
something
you
can
do
by
the
way.
I
should
start
with
that
in
a
docker
image,
it's
layers,
but
they're,
not
hard-coded.
A
A
If
you
know
how
they
were
built,
then
you
have
infinite
confidence
that
you
can
a
lot
more
confidence
that
you
can
make
that
change
and
we
call
it
rebasing
and
it's
something
that
Heroku
and
cloud
foundry
organizations
have
been
doing
for
almost
a
decade
before
docker.
This
idea
of
layers
sitting
on
top
of
the
base
image
and
us
being
artists
replace
that
base
image
whenever
we
want
without
ever
meeting
the
developers.
A
We
have
been
doing
this
for
seven
or
eight
years,
so
that
idea
is,
is,
is
well
tested
and
and
what
we're
doing
now
is
bringing
that
to
docker
images.
Alright-
and
this
is
this
idea
of
rebasing,
where
a
new
version
of
what
we
call
the
run
image,
which
is
you
know,
the
base
image
comes
out
and
that
platform
operators
have
the
ability
to
recreate
every
image
with
confidence
and
rewire
them
at
the
registry
level.
A
Sorry,
that's
the
features
that
operators
get
you
get
build
packs
that
they
can
curate
and
share
with
their
entire
organization.
We
get
the
ability
to
rebase
images
so
that
they
can
push
out
new,
secure
versions
of
base
operating
systems,
but
now
the
rest
of
it.
What
I
want
to
show
you
is
the
shiny
things
that
developers
get
to
use,
and
you
can
start
to
think
how
you
might
use
these
in
your
organization,
so
we're
gonna,
look
at
two
things
and
they
both
wrap
up
that
lifecycle.
A
One
is
CLI
called
pack
and
the
other
one
is
a
running
system
that
you
can
run
on
your
kubernetes
and
it's
called
k
pack,
and
it
does
this
as
a
service
for
you
all
right.
So
we're
going
to
leave
the
slides,
but
I
put
everything
in
the
slide
so
that
you
can
read
the
slides
and
remember
the
good
times
that
we
all
had
together
and
and
if
the
demos
go
terribly,
I
can
just
come
back
to
the
slides
harrangue
leave
quickly.
At
the
end,
sorry,
let's
have
a
look
at
our
app.
A
You
might
never
have
seen
an
odious.
They
all
look
like
this
they're
about
four
lines:
long,
it's
production
quality,
it's
it's
it's
good
stuff
and
we
got
ourselves
some
hollow
world,
and
so
this
this
is
my
app.
You
can
service
it
just
to
make
sure
that
you
know
where
I
am
nothing
up.
My
sleeves
unrelated,
if
you're
on
Netflix
there
is
a
new
magic
magic,
the
human
season,
god
I,
love
that
guy
he
makes
me
believe
in
magic.
This
is
just
he's
wonderful,
anyway,
back
to
back
to
not
magic.
A
So
we
have
this
pack
CLI
and
it
has
some
functions.
I,
the
one
that
I
I
have
already
run
is
set
default,
build
up
I
mentioned
before
a
builder.
Is
the
collection
of
build
packs
so
I'm,
using
one
from
from
the
Cloud
Foundry
pivotal
group
of
people
you're
allowed
to
build
your
own
builders,
and
they
will
also
we
make.
A
Builders
are
more
of
a
dynamic
thing
that
just
has
happened
automatically,
but
the
interfaces
is
this
builder,
and
the
builder
includes
lots
of
build
packs,
build
packs
that
my
organization
is
happy
to
use
and
send
out
into
into
production,
so
I've
set
a
builder
and
what
else
we
got
we
got
to
build
this.
It's
really
that
simple
I'm
going
to
give
it
a
name.
A
So
I've
got
a
the
like
thing
at
the
end
is
is
whatever
you
know
if
you're
using
GG
cRIO,
you
do
something
like
that.
If
you
were
wherever
whatever
it
is,
that
your
image
is
going
to
be
hosted
on
so,
as
we
probably
all
know,
without
saying
anything,
it
means
the
the
docker
I/o
or
docker
hub
and
flag.
That's
novel
is
the
publish
flag.
The
publish
flag
is
not
the
same
as
docker
push
literally.
A
This
is
going
to
happen
against
the
registry,
which
I
don't
think
a
lot
of
people
know
is
the
thing
you
can
kind
of
do
that
you
can
actually
just
interact
with
the
registry,
so
I'm
going
to
build
this
and
then
I'm
going
to
pull
it
back
down
as
if
it
was
never
on
my
laptop
to
start
with.
So
let's
give
this
a
go,
you
know
what
I'm
going
to
give
it
a
tag,
because
it's
pretty
already
cached
so
we'll
go
forum.
A
A
A
A
It's
faster,
and
now
we
can
curl
and
there's
that
wrap
okay,
so
we
took
our
Java
ad
know
JSF
and
we
turn
it
into
a
docker
image
with
all
the
beautiful
attributes
that
are
described
on
the
slides
of
an
operator
can
come
along
and
rebase
this
every
day
or
week
that
a
new
base
OS
comes
out
now
what
I
want
to
show
you
was
the
layers,
these
the
stages
on
the
left.
You
can
see.
These
are
the
annotations
of
detector
restorer
analyzer
builder.
A
These
are
the
stages
of
the
lifecycle,
and
this
is
essentially
the
functionality
of
the
cloud
native
buildpack
project.
Now,
what's
not
shown
here
is
all
the
other
build
packs
that
were
not
used
in
the
Builder,
and
there
are
lots.
There's
like
a
12
different
ones
related
to
Java
and
Java
was
a
real
winner
here
in
this
to
cloud
native,
build
packs,
because
there's
a
lot
of
different
ways
that
people
want
to
put
together
a
java
wrap.
A
A
A
And
then
more
layers,
so
we
I
guess
this
is
the
most
important
part.
Look
we
actually
download
a
Java
a
node,
so
we
downloaded
node
set
node
up.
How
do
we
do
that?
Who
knows?
Who
cares?
I?
Don't
care,
you
don't
care
unless
you're
from
the
node
core
team
and
you're
interested
in
what
flags
are
set,
you
don't
care,
it's
all
open
source.
You
can
go
and
care
all
you
like,
but
you
didn't
have
to
copy
and
paste
something
from
Stack
Overflow
that
you're
now
responsible
for
because
you
put
it
in
your
docker
file.
A
So
then
we
move
on
to
the
yarn
bill
pack,
which
means
we
look
at
the
yarn
dot,
lock,
file
figure
out
what
the
dependency
is
download.
Those
dependencies
set
up
all
that
and
and
then
we're
good
to
go.
It
defines
how
this
container
will
be
started,
which
is
all
encoded
inside
the
container
and
super
nice
aspect
is
all
this
metadata
of.
A
What's
in
this
image,
so
clothes
is
tagged
in
the
the
metadata
of
the
image
itself,
a
little
hard
to
read
because
it's
it's
Jason
stringify
does
Jason
inside
Jason,
but
it
means
that
you
can
go
and
look
across
all
your
images
and
find
out.
What's
in
them.
Is
there
a
bad
version
of
nodejs
in
a
version
of
all
your
images,
you
can
now
look
and
that's
a
really
valuable
thing
from
an
ordering
perspective.
A
Alright,
what's
next
what's
next
is
we
will
run
this
on
kubernetes
because
we're
at
a
Cuban
news
forum,
I,
think
it'll
be
pretty
sneaky
of
me
to
walk
away
and
never
haven't
run
anything
on
Cuban
days
and
I.
Won't.
Do
that
to
you?
Sorry,
let
us,
let's?
What
do
we
want
to
do?
We
will
watch
tube
cuddle
get
hard
service.
A
Let's
start
there,
there's
obviously
nothing
going
on
interesting.
Oh
there's
that
default
one
all
right,
all
right,
coop
cuddle
run
and
our
image,
or
is
our
image
called
there?
It
is
yeah
I'm,
not
gonna,
type
cool
anymore.
You
know,
I
hope
everyone's
a
list.
It's
2k
if
you're
still
talking
to
coop
cuddle,
cut
that
out.
Alright
go
and
find
self
a
set
of
good
aliases
get
Auto
completions
your
life
will
be
better
you'll
you'll.
Thank
me
and
everyone
else
that
ever
uses
it.
A
Alright
we're
going
to
run
this
image,
give
it
a
name,
we'll
call
it
forum
and
we'll
call
it
sample,
because
that's
what
I
call
my
other
demos
and
then
we'll
run
that
that
will
create
a
deployment.
A
deployment
will
create
a
replica
sets,
replicas,
ok,
pods
and
the
pods
will
appear
because
cube
news
is
awesome.
None
of
this
has
got
anything
to
do
bill
packs.
This
is
just
me
making
sure
that
you
like
kubernetes
and
I'd
hate
you
to
get
to
the
end
of
the
day
and
no
one
actually
ran
any
commands
in
real
life.
A
How
it'd
be
really
sad
if
you're
a
new
person-
and
you
came
back
from
the
forum-
it's
like
I-
never
saw
anyone
use
kubernetes,
it
was
the
weirdest
day
ever
so
you
can
tick
that
box.
So
you
actually
saw
someone
use
the
goddamn
thing.
So
here
we
have
our
pod.
We
don't
have
a
service,
yet
we
we
can
expose
their
deployment
I'm,
not
advocating
typing
these
components
versus
using
amble
files,
but
it
you
know
it
does
demo
better
expose
deployment
sample.
A
A
A
A
And
watch
me
type:
that's
pretty
cool
such
a
good,
typist
alrighty.
We'll
do
this
again
of
note,
as
we
watch
this
happen
is
that,
whilst
in
cloud
foundry
and
Heroku,
we
the
the
build
and
the
deployer
coupled.
So
if
I
do
a
see
if
push
it
will
go
through
this
staging
process,
we
call
it
staging
or
the
build
pack
process.
It
will
also
know
that
I've
implicitly
given
two
permission
to
push
that
app
out
in
this
pack
command.
We
don't
implicitly
have
that
relationship.
A
It
will
produce
the
docker
image
on
my
registry
and
for
which
I
will
be
happy.
It
is
my
subsequent
responsibility
to
decide
if
and
when
and
where
that
should
now
go
off
somewhere
and
I
will
do
that
using
the
cube
kernel
commands,
you
might
do
it
using
some
other
CI
system
that
you've
got
good
because
you
might
take
that
document
and
first
test.
It
locally,
perhaps
bundle
some
tests
inside
it
and
then
perhaps
it
after
staging
communities
and
then
through
a
pipeline.
A
But
what
we
get
is
this
immutable
URI
of
what
it
is
so
now
it's
to
cuddle
set
image
there
we
go.
That's
gonna!
Leave
that
there
so
I
can
remember
what
I'm
typing,
because
you
know
okie
dokie
and
now
we
can
see
because
we're
watching
we
updated
the
deployments.
The
deployment
creates
a
new
replica
set,
the
replica
set
says
well,
I've
got
no
pods
I
need
pods,
it
goes
off
crates
pods.
Then
it
gets
rid
of
the
old
replica
set.
A
That
means
it
doesn't
need
those
old
pods
and
what
we
see
is
pods
coming
and
going,
and-
and
we
now
see
that
our
app
has
been
rolled
out
so
build
packs,
helped
us
make
the
image
without
ever,
making
a
docker
file
without
ever
worrying
about
how
we
do
node
and
what's
the
right
way,
how
do
we
do
it
in
our
company
right,
and
you
asked
that
question
enough,
every
or
every
team
that
asks.
How
do
we
do?
Java
here
means
we've
got
another
bespoke
docker
file,
because
they're
best
you're
gonna
copy.
A
What
some
other
teams
done
at
that
moment
in
time,
and
now
everyone's
organization
Forks
and
wanders
off
and
creates
a
new
docker
file,
so
there's
no
way
that
they're
all
going
to
look
end
up.
Looking
the
same,
so
we
get
this
a
additional
benefit
that
we're
always
going
to
be
going
through
the
same,
build
cycle
and
that
we
know
every
image
that
was
just
created
went
off
the
same
builder.
All
right.
A
We
did
that
it
was
awesome.
There's
a
Java
one.
You
get
the
gist
of
like.
There
is
on
a
very
simple
Java
wrap.
It
pulls
in
five
six
seven.
Well,
actually
it
says
seven
out
of
14
build
packs.
There
are
14,
Java
build
packs
and
it
used
seven
of
them
and
you
can
give
it
clues
as
to
what
you
want
that
that's
for
Java
people,
it's
a
super
valuable,
perhaps
less
valuable
for
some
of
the
other
ecosystems
where
we
have
our
different
and
perhaps
are
already
building
go
apps,
just
stuffing
them
into
scratch
containers.
A
We
went
through
this
demonstration
of
deploying
our
pods
and
exposing
them
as
a
service,
and
then
we
updated
it
and
the
crowd
went
absolutely
bananas.
We
looked
at
at
how
it
was
the
node
picked
the
web
process.
If
it
was
a
package
Jas
lock
file,
then
it
would
have
used
NPM
in
a
different,
build
pack,
all
right
to
finish
off.
A
A
Is
that
process
that
life
cycle
process
as
a
service
and
running
on
kubernetes
and
as
justified
through
the
slides
as
a
khabees
thing,
its
primary
experience
is,
we
are
going
to
create
some
some
resource
instances
of
custom
resource
definitions
and
it
will
then
add
a
controller.
It
will
do
all
the
work.
I'm
gonna
show
you
the
work,
we're
not
going
to
get
too
much
into
the
you
know
how
to
write
operators
and
things
so
I'm
going
to
expand
this
now
to
I
want
to
look
at
I
think
called
an
image
I've
already
installed
Capac.
A
It's
really
simple!
You
set
up
some
credentials
for
your
registries
and
for
your
git
repository,
you
install
the
the
operator,
it's
a
helmet
chart
and
what
else
some
builders,
and
that
was
it,
and
so
now
I
and
I'm
gonna
show
you
what
one
looks
like
you
know
what
that
looks
a
little
ugly,
let's
make
it
colorful
and
there
it
is
so
this
is
invisible
on
my
black
screen.
Let's
not
do
that.
Let's
put
on
a
white
screen,
that's
much
better!
So
we're
saying
hey!
There's
a
builder
I
want
to
use
which
is
got
my
bill.
A
Packs
in
it.
I
might
have
multiple
builders
in
my
organization,
but
this
one
I
know
it
knows
how
to
do
no
js'
apps
I
wish
to
have
a
disk
for
caching
so
that
my
builds
get
faster
over
time
and
stops
downloading
node.
Every
time
node
is
kind
of
small
open.
Jdk
is
kind
of
large
and
caching
starts
really
kick
in,
and
what
I'd
like
it
to
do
is
to
continuously
watch
my
git
repo.
So
every
time
there's
a
new
commit
on
that
branch,
fire
off
a
build.
All
that's
going
to
happen.
A
Is
it's
going
to
build
docker
images?
It's
then
up
to
you
to
decide
whether
you
wish
them
to
go
into
production
or
not
all
right.
Let's
do
this.
Kaf
is
cuckoo.
Cuddle
apply
minus
F
because
by
now
I
told
you
earlier
to
install
all
your
aliases,
so
I
assume
you're
up
to
speed
on
those
aliases.
I
assume
you
can
all
do
two
things
at
once
now
down
here
we
have
our.
We
have
our
image,
so
we've
created
that
image,
that's
what
I
just
manually
did
and
what
it
did
is.
A
It
started
a
new
pod,
that's
doing
the
bills,
and
you
can
see
that
it's
primarily
made
up
of
init
containers.
Let's
find
out
what
it's
up
to.
There
is
a
little
CLI
we
get
for
free
that
we
can
install
and-
and
so
what
this
CLI
is
doing,
is
pretty
much
watching
the
output
of
those
init
containers.
What's
in
the
init
containers,
each
stage
of
the
lifecycle
of
cloud
native
bill
packs
they
just
sort
of
do
the
one
after
each
other.
A
It
is
borrowing
the
concept
of
emit
containers
to
do
a
sequential
series
of
steps,
and
it
does
exactly
what
we
did
before.
Except
you
know,
it's
spat
out
a
tagged,
docker
image
and
you
don't
have
to
run
locally
just
commit
and
off.
It
goes
so
now.
You'd
have
a
second
system
that
watched
for
these,
but
we
will
do
it
here
manually.
A
A
And
and
off
it
will
go
I'll,
add
one
more
instance
or
resource
to
our
little
thing,
so
we
can
watch
so
the
way
it
works,
wake
up
how
it
works.
It's
watching
my
git
repo,
because
the
image
tells
it
to
and
when
it
finds
a
new
commit.
It
first
creates
a
new
build
instance.
So
that's
what
we're
waiting
for
now.
Is
it
for
it
to
to
do
that
polling
or
whatever?
Actually
it's
polling,
I
didn't
set
up
a
web
books,
that's
definitely
polling,
it
will
create
a
new
build.
The
build
will
create
a
pod.
A
A
Interesting
KPAC
as
a
thing,
the
way
they
make
the
debates,
images
that
they
use
to
run
kpac
on
my
cluster
is
built
using
the
PAC
CLI
I
thought
that
was
wonderfully
dogfooding,
so
five
points
to
them
for
eating
dog
food.
That's
not
the
phrase!
You
don't
actually
eat
dog
food
for
anyone.
That's
immune
to
that!
So
I
highly
recommend
you
you
start
to
play
with
this.
They
have
their
own
slack
channel.
A
That
allows
you,
as
a
platform
team,
to
be
able
to
go
off
and
do
your
job.
We
need
to
just
try
to
encourage
people
to
not
go
straight
to
docker
files.
The
moment
something
doesn't
work,
and
maybe
you
have
to
go
forward
and
find
all
the
teams
using
docker
files,
encourage
them
to
migrate
away.
And
let's
finish
with
the
the
grand
finale
of
of
seeing
our
app
roll
out,
so
we
got
our
new
pod.