►
From YouTube: TGI Kubernetes 030: Exploring Skaffold
Description
0:00:45 - Hi everyone!
0:02:42 - Heptio sponsors Cloud Native Infrastructure
0:03:55 - SPIFFE
0:05:37 - Amazon IOT Kubernetes Quickstart Button (Tim Carr)
0:06:29 - Comparisons of dev tool blog
0:07:40 - kubed-sh (Bash for Kubernetes)
0:08:21 - Click from Databricks
0:09:40 - Working group around dev tools in Kubernetes upstream
0:11:50 - Gloo toolset
0:14:30 - Introduction of skaffold
0:17:35 - Comparing to Flux
0:22:08 - Getting started with skaffold
0:23:00 - Installing Skaffold
0:24:00 - Goreleaser (awesome tool for managing Go releases)
0:28:06 - Joe uses GoLand
0:32:50 - Inspecting the Dockerfile
0:35:00 - Inspecting the Skaffold YAML
0:37:00 - Porting the YAML to Amazon
0:42:14 - First skaffold dev command
0:43:09 - Hello world
0:47:09 - Skaffold vs Ksync latency
0:48:10 - Inspecting the container registry
0:48:50 - Skaffold run
0:53:00 - Exploring helm
0:54:00 - Possible integration with ksonnet
0:56:45 - Looking at tagging strategies for the container images
1:01:03 - Going beyond Dockerfiles
1:05:22 - How skaffold detects deltas
1:06:00 - Microservices example
1:06:44 - Kubernauts Parameterizer
1:08:30 - Creating new repository in ECR
1:10:35 - Running the microservice example
1:12:04 - Fixing the leeroy app
1:12:27 - Learning how skaffold manages YAML delta
1:16:36 - Launching a kuard
1:18:20 - Examining the service and round robin DNS
1:19:10 - Port forwarding
1:20:50 - Tweaking the HTTP get request in Go (Turning of Keepalive)
1:28:20 - Working example!
1:31:00 - Looking at CRDs with skaffold
1:32:00 - What's next?
* Skaffold: https://github.com/GoogleCloudPlatform/skaffold/tree/master/docs
* v0.3: https://github.com/GoogleCloudPlatform/skaffold/blob/v0.3.0/CHANGELOG.md
* Cloud Native Infrastructure e-book: http://go.heptio.com/cloud-native-infrastructure
* SPIFFE joins the CNCF: https://www.cncf.io/blog/2018/03/29/cncf-to-host-the-spiffe-project/
* Heptio Quick Start for Kubernetes IoT button: https://twitter.com/timmycarr/status/979393239429107718
* Comparison for dev tools: https://blog.hasura.io/draft-vs-gitkube-vs-helm-vs-ksonnet-vs-metaparticle-vs-skaffold-f5aa9561f948
* kubed-sh: https://github.com/mhausenblas/kubed-sh
* Click from databricks: https://databricks.com/blog/2018/03/27/introducing-click-the-command-line-interactive-controller-for-kubernetes.html
* K8s WG around Dev Tools? https://groups.google.com/forum/?utm_medium=email&utm_source=footer#!msg/kubernetes-dev/YcjXRDrCdbI/msPgV3tBBgAJ
* Draft v0.12 released: https://github.com/Azure/draft/releases/tag/v0.12.0
* solo.io Gloo function gateway: https://www.solo.io/
* Weave Flux for gitops deployment: https://github.com/weaveworks/flux
* Goreleaser: https://goreleaser.com/
* Kail -- Kubernetes tail: https://github.com/boz/kail
* Kubernauts Parameterizer: https://github.com/kubernauts/parameterizer
Come hang out with Joe Beda as he does a bit of hands on exploration of Kubernetes and related topics. Some of this will be Joe talking about the things he knows well. Some of this will be Joe exploring something new with the audience. Ask questions, comment and help decide where things go.
Deploying applications to Kubernetes is one thing. But what about developing on Kubernetes? How do you use Kubernetes to run your application in an environment that is similar to production?
Last week we looked at ksync. This week we are taking on Skaffold from Google. We'll follow up in coming weeks with Draft and Telepresence.
https://github.com/GoogleCloudPlatform/skaffold
A
All
right
welcome
episode,
30
of
TGI
kubernetes
I'm
Joe
I'm
your
host
for
the
day.
This
is
a
weekly
issue,
I've
stream,
where
I
go
over
interesting
stuff
going
on
in
the
world
of
Cooper
Nettie's
and
then
do
a
deep
dive
on
some
topic,
whether
it
be
a
new
tool
or
a
feature
or
component
of
kubernetes,
and
really
dig
into
that.
I
am
the
CTO
of
a
startup
called
hep
tio
we're
focused
on
bringing
kubernetes
to
the
enterprise
and
so
yeah,
so
I
like
to
start
out
by
first
of
all
say
hi
to
everybody.
A
B
A
The
Vic
is
in
Santa
Barbara
Sean
good.
To
see
you
all
right.
We
have
tons
of
folks,
let's
see
o
Minneapolis,
oh
I'm,
sorry,
it's
still
cold
there,
Oh
bina,
Joe
and
London,
say
Eden
London
Matt
in
San,
Francisco
Google,
San
Francisco,
that
is
a
beautiful
office.
The
the
cafeteria
there
in
the
Google
San
Francisco
office.
Has
this
amazing
view
of
the
Bay
Bridge
and
at
least
I
think
they
have
a
couple
cafes,
but
the
one
they
had
like
you
could
borrow
like
binoculars
and,
like
you
know,
go
look
out.
A
A
So
with
that
in
mind,
I'm
gonna
jump
right
in
so
the
first
thing
I
wanted
to
call
attention
to,
and
let
me
make
the
font
here
a
little
bit
bigger
is
here
at
hefty.
Oh,
we
sponsored
the
cloud
native
infrastructure
book
that
Chris,
who
you
know,
sits
right
next
to
me,
works
for
hefty,
oh
and
then
Justin
garrison
who
works
for
Disney
down
in
LA.
A
B
A
We
sponsored
the
cloud
native
infrastructure
book
and
this
is
the
full
version
of
the
book.
It
turns
out
that,
depending
on
how
a
Riley
prices,
these
things,
sometimes
we're
able
to
just
do
an
excerpt
like
what
we
did
with
the
kubernetes
up
and
running,
but
with
this
one
we
were
able
to
do
the
the
full
book,
and
so
I'm
really
excited
about
that.
And
so
all
we
all
we're
asking
is
a
little
bit
information
and
we'll
try
and
make
sure
that
we
use
email
responsibly.
A
So
if
you
want
to
get
a
free
ebook
copy
of
that
now's,
the
time
I'll
put
all
the
all
the
links
in
the
description
after
the
after
the
live
stream.
The
next
thing
I
want
to
talk
about
is
the
CNC
F
is
hosting
some
projects,
some
new
projects
and
into
the
sandbox,
which
is
sort
of
the
very
early
sort
of
sponsorship.
A
Tier
and
one
of
those
is
this
project
that
I
started
before
I
started,
hefty
o
called
spiffy,
and
so
it's
being
developed
and
driven
by
this
company
called
Scytale
and
they're
doing
a
great
job
with
it
and
spiffy.
If
you've
ever
done,
any
work
with
managing
kubernetes
you'll
know
that
there's
this
huge
mess
around.
How
do
you
actually
get
the
right
certificate
authorities
and
the
right
certificates
moved
around
and
and
plugged
in
the
right
way?
A
And
then
once
you
have
that
done,
how
do
you
actually
get
certificates
between
workloads
that
are
actually
running
on
that
infrastructure?
Whether
it's
running
on
ladies
or
not,
and
so
spiffy
really
is
working
to
be
a
common
framework
for
getting
the
right
certificates
in
the
right
place
at
the
right
time
securely
in
it.
You
know,
and
it
would
be
a
it's
complementary
to
something
like
vault,
because
vault
can
play
the
role
of
the
CA,
but
vault
can
also
be
this
hash.
A
Core
vault
can
also
be
a
secret
store,
we're
going
to
be
working
overtime,
to
make
sure
that
you
can
use
Biffy
to
be
able
to
access
a
vault,
so
it
ends
up
being
your
bootstrap
credential
that
you
can
then
use
to
get
other
information
so
exciting,
very,
very
exciting,
to
see
spiffy
move
forward
there
and
my
second
project-
that's
in
the
CNC.
Yes,
so
there
you
go
all
right,
so
LR
good
to
see
you
again,
Abraham
in
Saudi,
Arabia,
good
to
see
you
all
so
dung
conks
in
Cork,
I'm,
sure,
I
or
not
I'm.
A
Sure
I
am
messing
up
names
here.
I
apologize
marko
from
belgrade,
so
awesome
good
to
see
you
all
all
right.
So
the
next
thing
this
is
super
cool.
I'm
totally
going
to
build
one
of
these
tim
car
and
he
put
up
a
tutorial
here.
I
haven't
dug
into
it.
He
grabbed
one
of
the
the
amazon
iot
buttons.
These
are
like
you
press
a
button
and
it
sends
a
signal
and
and
and
when
you
press
a
button,
it
launches
one
of
the
the
kubernetes
quick
starts.
So
oh
yeah
tim's
online
here
good
to
see
you
too.
A
A
We
need
to
create,
like
a
better
label
for
it,
to
replace
the
the
you
know,
the
the
taped
on
one
there.
Alright.
The
next
thing
I
wanted
to
point
out-
and
I
haven't
had
a
chance
to
read
all
this
yet
but
shot
hid
and
I
apologize
if
I'm,
if
I'm
messing
up
his
name,
wrote
a
blog
post
that
goes
over
comparison
of
a
bunch
of
different
tools.
Oh
and
apparently
it's
on
it's
on
the
orange
site.
Right
now,
let's
see
and
there's
some
good
good
commentary
there.
A
I
haven't
had
a
chance
to
to
dig
into
this
right
now,
but
it
goes
over
some
of
the
pros
and
con
of
these
things
and
I.
Think
one
of
the
things
that
I
assume
is
is
is
coming
out
of
this
is
that
each
of
these
things
will
do
like
slightly
different
different
things
and
so
like.
If
you
compare
sort
of
like
draft
versus
helm
versus
case
on
it,
the
overlap,
there's
some
overlap
between
these
different
things,
but
you
know
same
thing
with
with
scaffold
or
meta
particle,
but
I
don't
think
we
were.
You
know.
A
It's
kind
of
like
a
bash
for
kubernetes,
where
you
you
can
launch
stuff
in
and
it
automatically
creates
a
pod
around
it.
So,
like
you're
kind
of
you
know,
whenever
you
hit
a
command
that
actually
causes
something
to
happen
on
your
kubernetes
cluster,
so
it
kind
of
is
an
interactive
way
to
launch
stuff,
and
then
apparently
this
also
does
some
file
syncing
stuff,
similar
to
to
the
tool
that
we
covered
last
week.
A
Is
that
you're
like
dealing
with
a
particular
context,
an
object,
and
you
end
up
having
to
copy
and
paste
stuff
a
lot
and
repeat,
like
the
namespace
again
and
again
and
again
and
switching
contact
all
that
stuff
is
really
really
a
pain.
This
thing
is
essentially
carrying
a
bunch
of
state
along
and
putting
that
inside
of
a
repple,
so
it
looks
really
really
interesting.
Also,
and
so
like.
The
the
thing
here
is
that
you
can
say
show
me
all
the
pods
select
pod
number
two
describe
it.
You
know
show
me
the
events.
A
A
So
with
that
in
mind,
there
is
some
talk
around
creating
a
working
group
in
kubernetes
around
developer
tools,
a
lot
of
the
stuff
that
I've
been
covering
in
this
series
you
know,
is
being
brought
up
here.
Things
like
you
know,
scaffold
and
draft
and
telepresence
we
flux,
which
is
really
a
deployment
tool
mini
cube.
All
these
things
are
are
I,
think,
there's
a
lot
of
interest
around
this
stuff.
There's
some.
If
you
and
I'll
put
this
thread,
this
was
on
the
kubernetes
dab
mailing
list.
I'll
put
this
in
the
notes.
A
Also,
there's
some
discussion
of
whether
this
should
be
a
separate
working
group
or
whether
this
should
be
something
that
sig
apps
should
take
on.
Sig
apps
is
already
pretty
overloaded
with
a
bunch
of
stuff.
So-
and
this
is
definitely
you
know-
touches
on
a
bunch
of
different
a
bunch
of
different
areas.
So
it's
interesting
to
see
some
of
this
stuff
start
to
boot.
Up
also,
there's
definitely
a
lot
of
interest
around.
What
is
it
like
to
develop
applications
while
you're
against
kubernetes
right,
and
so
that's
yeah,
all
right
and
then
so
draft?
A
And
cover
next
week
is
draft.
There
are
some
really
big
changes
with
draft
that
I
think
are
super
interesting,
specifically
it's.
It
has
a
fully
client-side
mode
now,
which
I
think
makes
it
much
more
comparable
apples
to
apples
with
scaffold
in
some
ways
and
in
doing
so
makes
it
much
easier
to
deploy
draft
into
and
use
it
in
a
bunch
of
different
situations.
So
things
like
you
know
they
listed
here.
You
know
you
can
deploy
to
any
installment
of
kubernetes
I.
A
Don't
like
the
word
distribution
here,
but
regardless
of
where
and
how
you
running
district
kubernetes,
you
can
use
it.
You
know
things
are
vastly
simplified.
The
security
model
ends
up
being
better
and
then
over
time,
as
we
thinks,
like
things,
see
things
like
image
or
build
whatever
we
can
start
integrating
with
those,
so
I
think
it's
great
to
see.
Drafts
start
scoping
down
a
little
bit
and
maybe
maybe
introduce
some
of
the
plug
ability
that
we're
looking
at.
A
A
Don't
sure
exactly
what
that
means
when
it's
built
on
top
envoy
Matt
Klein
pointed
me
at
what
these
folks
are
doing,
and
it
looks
really
interesting
where
this
ends
up
being
sort
of
a
feature
full
proxy
that
you
can
put
in
front
of
things,
whether
they
be
you
know,
regular
old
rifle
services
or
whether
it
be
you
know,
server
list
function
as
a
service
type
of
stuff.
So
it's
really
interesting
I
think
to
start
to
see,
people
use
the
sort
of
fundamental
building
block
that
is
envoy
in
new
and
interesting
ways.
A
You
know,
there's
been
folks
who've
been
asking
hey,
can
you
do
an
episode
on
envoy
and
I've
been
a
little
bit
hesitant
to
do
that?
Just
because
envoy
really
is
not.
You
know
from
an
end-user
point
of
view,
it's
it's
going
to
be
rare,
for
people
to
use
envoy
directly,
I
think
a
lot
of
times.
What
we'll
see
is
that
there'll
be
a
bunch
of
sort
of
control,
planes
that
are
doing
interesting,
things
that
are
using
envoy
under
the
covers,
but
what
I
do
want
to
do?
A
And
I'm
going
to
do
a
little
bit
of
tease
here,
Matt's
here
and
see
out
all
the
the
engineer
who
started
envoy
and
I'd
love
to
have
him.
Do
a
guest
show
at
some
point
where
we
maybe
have
him
in
the
office
standing
here
with
his
laptop
and
we
get
him
up
and
running
and
he
can
do
a
little
bit
of
an
interactive
demo
with
envoy.
A
So
I
think
that
would
be
really
fun
all
right,
so
I'm
burning
through
these
and
then
finally
we're
gonna
get
to
scaffold,
and
just
this
morning
I
saw
the
announcement
that
scaffold,
V,
0.3
was
released
and
so
I
think
there's
some
interesting
stuff
going
on
there.
There's
some
breaking
changes
there,
I
think,
and
so
a
lot
of
this
is
to
is
to
really
make
sure
that
you
know.
We
know
that
you
know
when
you
see
stuff,
that's
this
or
early
you're
gonna
see
some
churn
you're.
A
You
know
when
you're
working
on
this
stuff,
you
as
a
developer,
you
really
want
to
still
have
some
freedom
to
get
it
right
and
break
off
some
of
the
break
off
some
of
the
the
compatibility
there.
So
so,
there's
a
huge
caveat
when
you're,
when
you're
dealing
with
tools
that
are
really
this,
this
early
right,
so
so
yeah
so
before
I,
move
on
to
jumping
into
scaffold
and
I
feel
like
I
went
through
a
lot
in
15
minutes
there,
but
I
still
feel
guilty
about
blowing
15
minutes
at
the
start.
A
A
Don't
the
time
delay
here
is
tough,
so
all
right!
Well,
maybe
if
stuff
comes
up,
we'll
dig
into
it
as
we
go,
but
yeah
let's
go
ahead
and
get
started
with
scaffold.
It
looks
super
interesting
and
I
have
not
done
a
lot
of
preflighting
here.
I
haven't
had
time
this
week,
and
so
all
I
have
here
is
a
bare-bones
cluster
on
AWS.
This
is
the
the
hefty
Oh
QuickStart
cluster.
If
you
want
to
do
something
similar
and
and
we're
going
to
try
and
make
scaffold
work
on
that
scaffold
is
this
is
the
crazy
world.
A
We
live
in
it's
a
project
that
is
coming
out
of
Google
Google
cloud
platform,
and
so
my
guess
is
that
they
haven't
done
a
lot
of
testing
on
AWS,
because
that's
just
how
these
things
go,
and
there
was
nothing
talking
about
running
on
AWS
anywhere
in
the
instructions
but
I'm,
confident
that,
because
kubernetes
is,
is
relatively
down
pat.
That
will
actually
have
a
good
experience
here,
we'll
see,
but
the
interesting
thing
here
and
I
started
reading
through
this
there's
some
features
here.
A
It's
a
it's
a
workflow
for
development,
where
you
can
actually
sort
of
you
know,
build,
have
it
automatically
detect
and
then
do
another
build
and
push
and
reload
stuff,
which
is
really
cool,
Oh,
Matt,
you're,
good,
so
you're
the
lead
on
scaffold,
and
so
we
can
blame
you
if
it
doesn't
work
well,
we're
gonna,
find
issues
and
we'll
file
them
and
with
something's
like
a
V
0.3,
we're
gonna
scale
our
expectations
appropriately.
So
don't
worry
about
it.
Matt
we'll
be
nice
to
you
as
long
as
we're,
not
exposing
you
know
dashboards
to
the
Internet.
A
We
should
have
a
good
time
learn
a
lot
so
so
it
looks
like
there's
like
there's
like
double
duty
with
scaffold
and
I
think
this
is
one
of
the
things
that
I
want
to
dig
into.
On
the
one
hand,
there's
this
issue
of
running
that
sort
of
I
save
code.
I
have
something
running
in
kubernetes
and
really
being
able
to
turn
that
crank
as
fast
as
possible,
and
so
in
that
way,
I
think
you
know.
Scaffold
is
very
much
part
of
that
compile,
run
debug
loop
and
that
we
see
going
on.
B
A
It
going
now
it's
good
to
see
ya
and
then,
and
then
there's
the
other
thing,
which
is.
You
can
also
use
this
to
drive
deployments
when
you're
in
a
CI
CD
pipeline
or
when
you're
actually
pushing
stuff
to
production,
so
I
really
want
to
one
of
the
things
I
want
to
get
out
of.
This
is
understand
that
do
dual
role.
A
How
do
you
actually
automatically
deploy
that
in
a
consistent
in
automatic
way,
and
so
I'm
gonna
be
interested
to
see,
if
maybe
there's
a
little
bit
overlap
between
flux
and
scaffold?
In
that
way?
Obviously,
scaffold
does
the
development
you
know
the
development
loop
also,
which
is
definitely
not
in
scope
for
flux,
yeah.
All
right
so
so
feature
is
no
server
side
component.
So
there's
no
overhead
on
the
cluster
and
I
think
you
know
this
is
probably
putting
this.
A
In
contrast
to
the
previous
version
of
draft,
but
but
it
looks
like
draft
is
in
a
similar
position
now
detect
changes
in
your
source
code
and
then
automatically
run
the
they'll
push,
deploy,
managed
image
tags.
So
this
is
really
interesting.
Is
that
a
lot
of
times
when
you
do
a
manual
version
of
this?
You
end
up
like
having
like
a
make
file
and
you
all
that
and
then
you
end
up
getting
lazy
and
you're
like
I'll.
A
Just
use
the
latest
tag
right
and
then
you
forget
to
and
then,
when
you
write
your
yeah
mole
for
kubernetes,
you
forget
to
actually
do
image,
pull
always
right
and
then
and
then
you're
like
I,
like
you
know,
redone
this.
Why
didn't
it
show
up
and
you
waste
15
minutes,
figuring
out
why
the
heck
you
aren't
seeing
your
changes,
show
up
after
you're,
like
you
know,
45
minutes
into
a
hacking
session
and
so
managing
the
image
tag
so
that
each
time
you
actually
do
a
build,
you
get
a
unique
image.
A
A
Understanding
is
I
love
the
idea
that
we
can
take
different
tools,
start
using
them
in
new
ways
and
really
build
up
a
toolbox
approach
versus
a
sort
of
this
is
the
one
true
way
to
do
things
and
then
deploy
regularly
when
saving
files
are
run
one-off
deployments
using
the
same
configuration
okay,
so
you
can
either
run
it
in
a
in
a
you
know,
implicit
mode
hit,
save
stuff
happens
or
an
explicit
mode
of
like
deploy.
Do
it
now
all
right,
Olly
from
Sweden
good
to
see
you
and
thanks
for
joining
Mike
all
right?
A
So
and
I
think
this
diagram
does
a
good
job
here
is
that
it
looks
like
scaffold
really
thinks
about
things
in
terms
of
build
push
and
deploy,
and
you
can
configure
it
with
different
schema.
So
if
you're
doing
development,
let's
say
against
mini
cube
the
oak,
that's
not
what
I
wanted
I
didn't
want
to
click
on
that
you
build
with
your
local
daemon.
A
You
go
ahead
and
you
push
you
don't
have
to
do
a
push,
because
you
know
you're
doing
you're
using
the
same
local,
daemon,
etc,
etc,
and
then
you
can
do
a
deploy
directly
and
so
that's
the
local
flow,
whereas
the
remote
flow
might
be
something
like
you
know.
You
use
the
Google
container
builder,
which
is
a
service
that
Google
Offers
or
we
could
swap
in
other
ways
of
building
in
the
future,
and
then
you
push
to
say,
GC
or,
and
then
you
deploy
with
with
helm.
A
You
know
changes
to
take
to
really
boot
up
with
yeah
and
so
Shawna
and
and
matter
are
talking
in
the
chat
about
docker,
desktop,
plus
cube
control
and
and
yeah.
So
so
the
edge
versions
of
docker
and
I
don't
have
the
edge
version
of
docker
installed.
I'm
running
the
mainline
does
have
support
for
kubernetes,
and
this
is
part
of
the
really
sort
of
you
know,
nascent
support
for
kubernetes.
A
That
docker
is
starting
to
show,
and
so
it's
a
it's
interesting
that
that
that's
being
positioned
as
sort
of
a
a
you
know,
side
by
side
with
mini
cube
in
this
in
this
particular
situation.
Alright,
so
and
then
there's
scaffold
dev,
which
is
a
continuous
deployment
where,
as
scaffold
run,
is
that
sort
of
one-off
trigger
going
on
alright
and
here's
a
demo,
we
can
watch
a
video
how
do
I
restart
it,
because
it's
a
Jif,
these
videos
are
actually
like.
Oh
no.
What
is
that?
Oh
here
we
okay,
these
these,
maybe
maybe
there
we
go.
A
B
A
B
A
Doesn't
end
up
taking
a
huge
amount
of
space
all
right,
so
what
we'll
need
so
we
need
to
download
the
OS
X
1.
So
this
is
what
I
like
seeing
is:
there's
no
magic
curl
to
bash
bruise,
even
better,
but
that's
ok!
What's
going
on,
oh
okay,
so
this
is
something
that
ok.
So
what
we're
doing
here
is
I've
already
done
a
get
clone
a
scaffold,
and
so
I
got
an
error
there,
because
I
was
trying
to
create
a
regular
old
file.
That
was
the
same
way.
A
B
A
A
B
A
Says
he's
trying
to
use
a
first
capital
yeah?
So
if
you
need
help
with
that,
go
ahead
and
look
at
it,
one
of
the
things
if
you
use
the
caisson,
it's
tough
I
haven't,
had
an
update,
a
chance
to
update
it.
Yet
is
that
before
it
would
automatically
try
and
actually
create
stuff
and
do
your
submit
for
for
the
brew
thing
because
it
doesn't
get
submit
but
we're
using
a
PR
flow,
and
so
now
they
have
a
mode
there.
A
A
A
Yolo
on
saw
the
update,
while
we're
doing
other
stuff
Oh
370
megabytes
holy
moly,
so
this
is
docker
on
the
Mac.
Here
you
go.
Okay,
now
watch
this
download
alright
and
then
the
docker
image
registry.
We
have
to
have
setup,
and
so
this
has
says
you
know
so
we
could
use
the
docker
1
itself,
which
would
be
easy,
or
we
can
lean
on
some
of
the
stuff
that
we
did
before
and
try
and
use
the
the
Amazon,
the
ACR
stuff.
A
A
There
we
go
so
it's
upgrade
because
I
think
that's
probably
more
realistic,
I
think
anybody
who's
really
doing
this
they're
either
gonna.
Do
it
locally
with
like,
say
the
mini
cube
flow
or
they're
gonna
be
working
against.
You
know
some
sort
of
production
infrastructure,
they're
gonna,
want
to
be
using
a
private
repo
and
the
the
you
know.
If
you're
running
in
the
cloud,
you
probably
want
to
use
the
repository
the
the
docker
image
repository
that
that
is
paired
with
that
cloud
that
ends
up
being
the
fastest
thing
there.
So.
B
A
Set
that
up
I
gotta
remember
how
to
do
this.
Cuz,
it's
been
a
while
I
believe
this
is
under
container
service.
Even
though
it's
not
part
of
it,
then
we
have
repositories,
let's
get
started.
Let's
call
this
thing:
I,
don't
know
we'll
see
what
we
had
need
to
call
it
as
we
go.
Okay,
all
right
and
I've
already
gone
through
and
cloned
that.
So
here
we
go
and
we'll
see
a
seedy
scaffold
see
the
examples
CD,
let's
see
so,
which
one
do
it.
We're
gonna
try
getting
started
getting
started.
Okay,
so.
A
Oh
and
the
other
thing
that
I've
been
playing
with
is
using
goal
and
as
my
IDE
here
and
let's
see
how
that
works,
because
I
just
updated
that
so
there's
a
little
bit
of
Yolo
going
on
here,
you
guys
are
all
laughing
about
curl
the
sudo
bash,
exactly
import.
So
oh,
this
is
the
first
time
I've
run
the
latest
ones.
A
Let's
we're
doing
it
live
Oh
license
a
value
I'm
still
in
the
evaluation
period
here,
so
I'm
probably
gonna
end
up
buying
it,
but
I
haven't
yet
Chris
Chrissa
encouraged
me
to
go
ahead
and
do
this
I,
don't
know
what
it's
gonna
show
on
screen
right
out
the
gate
here.
So
let's
see
all
right
cool
there
we
go.
So
that's
that's!
Golan
and
and
unregistered
VCS
go
routes
all
right.
So
we're
gonna
ignore
all
this
stuff
because
I'm
not
sure
where
all
it
means
all
right.
So
what
do
we
got?
A
So
we
got
a
docker
file
here,
which
is
doing
the
the
go
Alpine
dockerfile
thing
sitting
the
work
directory
to
the
whole
thing:
copying,
Maine,
dot,
go
and
then
go
ahead
and
do
that
yeah,
so
I
haven't
played
with
so
supposedly
the
latest
version
of
goal
and
which
was
just
released,
has
support
for
kubernetes
manifest
so
I'm
really
looking
forward
to
that.
That
should
be
interesting.
A
A
A
A
B
A
A
A
A
So
here
we
have
the
dockerfile,
and
so
this
is
a
pretty
darn,
simple
dockerfile,
which
is
essentially
saying:
hey,
we're
gonna,
do
nothing
more
than
copy
a
single
file
over
and
then
do
a
go
build
on
that
this
is
not
doing
like
the
multistage
docker
images
right
so,
which
is
something
that
I
think
might
be.
That
I
would
look
to
be
doing
in
a
situation
like
this.
A
The
other
thing
to
keep
in
mind
here-
and
this
is
a
downside
to
this
approach-
is
that
we're
not
going
to
be
like
when
I
do
a
go,
build
if
I'm
doing
a
big
project.
This
actually
may
do
a
lot
of
stuff
it
may,
if
I
have
a
lot
of
dependencies
and
I
do
a
go,
get
as
part
of
this
and
I'm
bringing
all
this
stuff
in
there.
I
want
to
make
sure
that
I'm
not
recompiling
the
world
every
single
time
and
so
I
think
we'll
see.
You
know
the
speed
that
we
see
here.
A
Every
time
we
do
a
build.
Loop
may
not
be
indicative
of
a
significantly
sized
project
that
has
some
real
dependencies
and
such
so
that's
something
to
keep
in
mind,
also
the
let's
see
and
then
here's
main,
and
then
we
have
two
things
yeah,
because
right
now,
docker
file
doesn't
actually
cache
stuff
from
from
one
run
to
the
next.
Here's,
the
simple
scaffold
in
presentation
mode,
dual
tool:
windows,
project,
okay,
so.
A
So
Vic
is
saying
that,
like
once
you
get
into
this
mode,
you
find
ways
to
optimize
your
docker
file.
You
can
chase
this
down
really
far
like,
because
I
did
a
bunch
of
stuff
that
I,
probably
shouldn't
have
with
the
kubernetes,
build
around
sort
of
cashing
in
or
me
stuff
so
that
we
get
a
doctor
eyes
build
as
fast
as
possible.
It
was
really
painful
to
especially
if
you
start
doing
like
random
cross-platform
stuff,
like
what
we
do
in
kubernetes.
A
Okay,
so
Chris
is
saying
in
presentation,
mode,
fuel
tool,
windows
project,
oh
when
I'm
in
presentation
mode.
So
let's
try
this
again
we're
gonna,
try
and
make
the
presentation
mode
stuff
work.
So
this
does
fullscreen
I,
don't
want
to
do
fullscreen
I
just
want
to
do
that.
Sighs
and
then
you
said
view.
A
A
A
Let's
see
it
matt
saying
we're
planning
support.
Nan
docker
build
previously
worked
on
some
of
the
basil
rules
to
build
docker
images.
I
think
that
will
help
have
much
faster,
builds,
all
right,
very
cool,
there's
some
other
startups.
That
I
don't
want
to
name
names
because
they
might
still
be
a
little
bit
stealthy
that
are
doing
some
really
really
interesting
stuff.
Around
continuous
builds,
also
so
yeah,
okay,
so
and
then
we're
let's
see
for
building
it's
saying.
Well,
here's
the
name
of
the
artifact
and
the
workspace
is
the
current
thing.
A
A
Ok,
so
there's
a
glob
here
going
on
when
you're
using
the
Google
cloud
builder,
then
ok,
so,
instead
of
local,
what
you
do
is
you
actually
say:
Google,
Cloud
Builder,
and
this
says
project
ID
and
so
I.
Think
one
of
the
interesting
things
is:
how
is
this
extended?
Are
these
plugins
built-in?
Is
there
a
discovery
mechanism?
It
makes
doing
something
like
parsing
the
Shamu
file
that
much
harder.
If,
if
you
really
truly
make
this
stuff
extensible.
A
Alright,
but
the
first
thing
we're
going
to
need
to
do
is
we're
going
to
need
to
actually
change
the
image
name
since
we're
going
to
be
doing
our
own
thing
here,
and
so,
if
we
go
back
to
Amazon
we're
gonna
call
this
scaffold.
Example,
ok
and,
and
so
now
what
we
can
do
is
you
know?
Gotta
love
me
some
amazon.
This
is
the
image
that
we're
building
now,
and
so
this
ends
up
being
your
account.
We're
building
and
hosting
us
west
to
DKR
is
probably
left
over
from
a
previous
version
of
the
the
ECR.
A
This
is
probably
docker
container
registry
or
something
like
that.
I
don't
know,
but
they
when
they
came
out,
I,
don't
know
what
DKR
DKR
and
then
East.
It's
very
confusing,
I'm,
not
sure
what
DKR
stands
for
and
then
and
then
this
is
the
name
of
it
and
then
oh
I,
guess
we
probably
don't
need
both
of
these.
So
there
we
go.
So
that's
the
name
of
the
image
workspaces
there
I
probably
need
to
go
through
and
change
this
also,
so
we
say:
here's
the
image
there,
so
we
got
that
going.
A
Once
we
get
this
stable
enough,
Sean
says
docker
makes
sense,
just
have
any
options
worthy
yeah,
one
of
the
things
that
you
know
we
have
been
playing
around
with
the
hashey
core
plug-in
stuff
where
it
launches
us
something,
and
then
you
use
G
RPC
to
talk
to
it
for
arc
is
using
that
at
the
hefty
outside.
So
we
started
using
that,
and
it's
pretty
interesting.
A
Let's
see
nadar
says:
ECR
also
supports
pulling
images
in
OCI
format,
probably
of
a
different
end.
Point
yeah,
interesting,
okay,
so
yeah.
Maybe
that's
what
it
is
right.
So
maybe
this
is
saying
I'm
doing
the
docker
version
of
it
versus
some
other
version
of
that
endpoint,
okay
and
then,
instead
of
then
what
I
can
do
is
oh
I'm
gonna
do
PowerShell.
We
run
the
docker
login
command
that
was
retirement.
I.
Don't
want
to
hear
I'm
going
to
do
this.
A
B
A
And
I
get
I.
Do
okay,
give
me
a
second
here:
okay,
I'll,
do
this
it'll
work
it'll
be
fine,
so
what
I
have
is
AWS
actually
I
would
give
me
a
second
here
when
I
switch
back
from
my
face.
I
don't
want.
I
got
to
be
careful
because
I
don't
like
whenever
I'm
doing
this
stuff,
I
don't
want
to
flash
credentials
on
the
screen,
because
then
I
have
to
go
rotate
them
later
and
that's
a
total
pain
in
the
butt
so
and
I
just
want
to
make
sure
that
they
can
fix
that
I
have
setup.
A
A
You
export
AWS
profile
equals
demo,
so
what
you
can
do
is
you
can
set
up
different
profiles
with
the
AWS
tool
if
you
have
multiple
accounts
or
if
you're
doing
roles
into
an
account
and
then
I
can
run
where,
where
did
that
and
then
I
should
be
able
to
run
this
command
here,
and
this
is
gonna,
ask
me:
oh
no,
it
didn't
ask
me
for
anything.
Sometimes
I
ask
for
two-factor
off
and
then
okay
I'm
not
going
to
worry
about
people,
copying
this
and
then
I.
A
Take
this
and
I
run
this
and
Dockers
like
whoa
you're,
passing
a
password
over
that,
but
it's
fine,
okay,
so
so
now,
I
should
be
able
to
a
docker
push
of
that
thing
up
to
the
to
the
AWS,
the
ACR
stuff.
Now
this
particular
credential
I
believe
is
good
for
like
20
hours
or
something
like
that.
So
if
you're
using
this
flow,
you're
going
to
have
to
do
this
about
once
a
day,
Sonique
says
to
scaffold,
support
supplying
registry
credentials,
or
does
it
just
require
you,
the
dock
or
a
login,
to
push
the
images?
A
My
guess
is
that
and
I'd
love
to
get
Matt
to
take
on
it.
My
guess
is
that
the
build
stage
here
is
really
build
and
deploy
and
or
maybe
I,
don't
know
our
built,
build
and
push
right.
It's
probably
here
but
I
think
well,
we'll
dig
into
that.
Well
understand,
what's
happening,
I
may
have
some
bad
assumptions.
We're
gonna
see
how
things
work.
The
other
thing
that
I
that
I'm
interested
in
is
that
nowhere
here
did
we
deal
with
the
image
tag
right
and
so
and
so
like.
A
If
scaffold
builds
an
image
and
gives
it
a
gives
it
a
tag.
Will
this
the
this
pod
llamó?
How
will
we
actually
specify
what
the
tag
is
there,
and
you
know
one
thing
it
could
do-
is
do
a
little
sort
of
sneaky
search
and
replace
along
the
way
or
it
could
be
something.
That's
more
explicit,
I
think
that
that's
something
that's
going
to
be
interesting
or
maybe
that's
part
of
the
deploy
options
that
you
can
specify
it.
A
A
A
You
guys
know
kale
I
talked
about
this
before
this
is
a
fun
tool.
This
is
a
kubernetes
log
viewer
that
lets
you
specify
a
whole
bunch
of
stuff
for
for
doing
it.
So
so
it
has
a
kale
like
thing
built
into
it.
Alright,
so,
okay,
so
scaffold
has
done
the
following.
For
you
build
an
image
from
the
local
source
tag
it
with
its
sha-256
sets
that
image
in
the
kubernetes
manifest
to
find
a
scaffold,
yamo
and
then
a
cube
control
apply.
A
A
Why
is
it
under
shell
I?
Don't
know
splits
horizontally
boom?
Okay,
so
we're
gonna
watch
out
we're
gonna
go
a
CD,
T
GI
k
and
then
I
have
my
fun
like
thing
here
that
automatically
quickly
sets
it
up
for
us
cube
control
get
pods.
So
we
see
we
have
the
one
pod
going
on.
There
cube
control
get
getting
started.
Oh
yeah,
mall
and.
A
A
And
then,
and
then
going
through
and
I
believe
this
ends
up
being
the
pod
name,
then
that
this
thing
is
actually
talking
to
okay.
So
let's
go
and
make
a
change.
Hello,
TGI,
K,
we're
gonna,
do
that
so
suppose
I
haven't
hit,
save
yet!
Oh
did
it
autosave
for
me,
I,
don't
know
I,
don't
think
I
hit,
save,
maybe
I,
it's
just
a
reflex,
but
it
looks
like
it's
automatically
boom
and
then
it
relaunched
it.
So
I
think
that
is
that
is
pretty
darn
good
in
terms
of
of
being
able
to
relaunch
something
super
quickly.
A
A
Did
I
miss
something
or
I'm
gonna
hit
control-c
and
started
over
again?
Ignore
it
I,
don't
know
what
happened
there
Matt
it
seemed
like
it
kind
of
got,
confused
and
hung
all
right,
so
hello,
a
TGI
Kate's
goal
and
our
goal
and
or
goal
in
or
how
you
pronounce
it
yeah.
Okay,
so
cool
I
mean
so
that's
the
flow
and
that's
actually
I
think
really
interesting.
I.
Think
if
we
compare
this
no
worries.
I
want,
like
you
know,
zero
point.
A
A
I
have
one
of
the
things
that's
a
little
bit
confusing
to
me
here
is
that
we
have
the
image
tags
which
are
a
shot
256,
and
then
we
have
the
sha-256
of
the
resultant
built
image
and
so,
and
these
things
end
up
being
different-
and
my
guess
is
this-
is
a
show
off
to
256
of
that
directory.
It's
not
clear
what
that's
actually
what's
what
that
is
actually
iterating
over
for
that.
So
so
that's
something
that
I
want
to
understand
a
little
bit.
I
won't
understand
it.
A
How
exactly
do
we
get
that
tag
in
there
and
actually
modify
the
yeah
Mille
I'd
like
to
play
with
this
and
do
something
like
a
replica
with
three
replicas
and
see
how
that
works
too?
So
I
think
that's
probably
something
that
we
can
play
with
all
right
and
then
all
right,
and
so
now
we
can
go
through,
and
if
we
do
this,
we
can
also
do
scaffold
run,
which
says:
hey
just
turn
the
crank
once
so
it
says
so
it
doesn't
build
and
then
it
tags
it.
A
Yeah
so
there's
a
lot
of
hashes
going
around
here,
so
there's
there's
successfully
built
something-
and
this
is
maybe
a
random
and
then
tagged
it
with
this
and
then
tagged
it
with
this
thing
here
which
looks
like
that's,
maybe
the
get
hash
that
it's
using
to
actually
do
that
run
defaulting
to
the
git
commit
tagger.
Oh
okay,
I
want
to
understand
a
little
bit
about
tagger,
taggers,
okay,
so
Michael
is
asking.
Do
you
plan
to
do
some
sort
of
dark
magic
to
auto-generate
that
configs
I
would
view
that
generating?
So
this
is
I
think
Michael.
A
I
want
to
understand
about
this,
so
the
first
one
being
okay,
so
we
can
understand
the
helm
stuff.
So
how
does
how
does
helm
look
differently?
So
there's
a
so.
We
have
a
helm
deployment
example
and
to
do
this,
you
have
to
sort
of
like
bootstrap
and
create
your
own
little
helm.
Chart
I'm
not
going
to
dig
into
that
right
now,
but
we
can
take
a
look
at
it
and
I
don't
have
helm
installed
in
this.
Yet,
let's
see
so
so
here
we
have
scaffold
yeah
moe
and
what
we
end
up
here
is
this
okay.
A
A
The
image
parameter
that
you
actually
want
to
send
through
this,
and
so
this
assumes
something
about
the
the
values
that
you
have
in
your
helm
chart.
So
let's
look
at
the
hell
on
chart
that
they
have
here
and
then
also
I.
Don't
know
like
what
is
this
building?
Does
it?
Actually?
The
dockerfile
here
is
nothing
but
nginx,
okay.
So
it's
just
a
sort
of
a
super,
simple
type
of
thing,
but
then,
if
we
look
at
scaffold
here,
there's
the
values
diamo,
okay:
this
is
the
nginx
roof
or
n
engine
x1
and
the
the
valley.
A
Oh
wait.
No.
This
is
I'm
confused
all
right,
so
this
is
the
full
nginx
chart
that
got
pulled
in
here,
or
at
least
a
significant
one.
Oh
no!
It's
not.
This
is
a
cut-down
version
of
the
nginx
chart.
That
includes
a
bed
that
includes
ingress
and
deployments
and
all
that
okay
chart
release
heritage
so
yeah.
So
this
is
theirs
and
then
it
looks
like
these
values
are
the
defaults
for
that
and
then
you're
essentially
saying.
Okay,
the
so
one
of
the
interesting
thing
is
that
we
have
the
repository
and
the
tag
which
are
separate
here.
A
Okay,
so
so:
okay,
Wow,
okay.
So
so
there's
there's
some
assumptions
going
on
here,
and
some
of
this
is
based
on
sort
of
the
the
conventions
for
how
helm
works.
But
if
we
look
at
the
scaffold
gamal
here,
it's
same
values
as
images,
but
images
is
really
a
object
that
actually
has
three
values
under
it.
That
includes
the
repo
the
tag
and
then
the
pull
policy,
and
so
it's
probably
assuming
the
three
sort
of
sub
values
there,
which
is
really
a
convention
type
of
thing.
A
Let's
see
we're
hoping
it
favored
higher
level
application
representations
other
than
cube
control
and
helm.
Yeah
I'd
love
to
see
this
integrated
with
caisson
it
definitely
something
that
I'm
interested
in
seen
caisson
it
is,
is
really
built
to
be
driven
by
tools
like
this,
because,
again,
it's
all
client-side
and
I,
like
the
idea
that
you
can
pass
some
of
the
values
that
are
going
on
there
with
case
on
it.
You'd
probably
pick
the
environment
that
you're
working
against.
A
Let's
see
the
tag
policy
is
how
you
configure
how
the
build
step
will
tag
the
image
so
I
want
to
look
into
the
different
options
there.
Oh
and
then
Matt
says
that
there's
an
annotated,
scaffold
IMO
in
the
example
structure.
Something
could
definitely
go.
Look
at
that
and
then
mix
is
the
helm,
create
nginx
the
default
started.
Okay,.
A
A
So
it
expects
the
fully
specified
image
or
the
broken
out
thing.
I
think
that's
something
that's
worth
looking
at
all
right,
but
let's,
let's
go
through
and
we'll
look
at
the
example
and
here's
the
annotated
scaffold.
So
this
looks
really
interesting
and
I'm.
Sorry
I
keep
forgetting
to
do
this.
You
guys
got
to
call
me
out
when
the
font
is
too
small.
Okay,
so
tag
policy
determines
how
scaffold
is
going
to
tag
your
image.
We
provide
a
few
strategy
here,
although
most
likely
you
won't
need
to
care.
A
A
A
Are
those
layers
not
config
familiar
with
ECR
Thomas,
says
I
blew
my
foot
off
using
the
gate,
commit
tag
and
yeah.
You
can
get
conflicts
on
that
because
it's
not
super
specific
when
you
have
dirty
stuff
I
believe,
let's
do
a
little
bit
of
examination
here.
Let's
do
docker
images
make
this
big
and
the
one
that
we
talked
about
here
that
I
just
built.
Let's
try
and
examine,
what's
going
on
here,
I,
don't
think
they're
layers,
but
this,
but
the
latest
one
that
I
built
was
oh.
A
This
is
interesting,
I
would
I
mean
one
of
the
things
that
I
think
would
be
interesting
here
is
to
have
some
sort
of
tagging
strategy
that
embeds
the
time
and
date
that
the
thing
was
built
along
with
some
unique
thing,
because
then
they're
gonna
sort
well,
whereas
right
now,
if
I
sort
by
image
tags
I
believe
then
you
know
these
things
are
gonna
actually
sort
out
of
order.
So
so
the
time
that
the
thing
was
built,
maybe
in
UTC,
might
be
a
useful
tagging
strategy.
A
A
A
There
are
two
different
ways
to
actually
address
a
docker
image,
so
one
of
them
is
via
an
image
tag
which
you
can
essentially
rewrite
and
repair
it,
and
so
there's
no
guarantees
with
it.
The
other
way
is
with
the
sha-256
way
of
doing
this
stuff,
and
so
this
I
believe,
like
like
docker
image,
256
I,
believe,
instead
of
like
the
way
you
specify
this,
is
that
there's
a
different
way?
I'm.
Sorry
I'm,
just
digging
into
this,
because
I
don't
remember
this
so.
A
A
What
you'll
see
here
is
that
we
both
have
the
image
here,
but
then
we
also
have
the
image
ID
here,
which
is
fully
resolved,
and
so
so.
This
is
one
of
the
things
that
the
cubelet
I
believes.
Write
writes
back
in
the
status.
So
so,
if
you
say
you
know,
if
you
do
do
something
silly
like
reuse
latest
again
and
again
and
again,
this
is
where
you
can
see:
okay,
which
image
am
I
actually
using,
because
this
sha-256
here
can't
be
faked,
and
so
one
way
to
do
this
is
you
could
do
you
could?
A
Can
we
actually
move
those
things
together
and
so
I
think
if
I
was
going
to
give
you
advice
on
the
schema
here,
we
have
both
the
artifacts,
but
then
have
a
build
strategy
or
have
something
like
okay,
we're
doing
docker
under
that
and
there's
you
know,
the
docker
build
is
one
build
strategy,
but
if
we're
using
image
or
builder,
they
may
have
different
ways
of
describing
stuff
yeah,
so
matt
says
that's
exactly
what
we're
thinking
about
now
yeah.
So
how
does
this
look
for
something
like
a
mature
cluster
builds
and
I?
A
Think
the
the
and
then
and
then
it's
the
difference
of
like
how
do
you
specify
how
you
build
the
artifact
versus?
How
do
you
specify
you
know
what
you're
using
to
build
the
artifact
right
and
maybe
like
the
local
versus
geez?
You
know
versus
the
Google
cloud
builder.
Are
those
things
actually
like?
Maybe
this
should
be
collapsed
up
into
the
artifact,
but
if
you're
building
multiple
images,
you
don't
want
to
have
to
repeat
yourself.
So
there's
a
little
bit
of
a
you
know,
fully
blown
out
denormalized
thing
to
this
scheme
of
versus
hey.
A
Do
we
want
to
make
this
thing
be
more
normalized
and
human-friendly
so
yeah,
something
to
think
about
there?
Okay,
so
there's
a
local
strategy
which
defers
to
everything
in
docker
config
for
for
authentication
information
for
things
like
push,
but
then
in
that
we'll
use,
docker,
credential,
helper
and
all
that
stuff.
But
oh
and
you
can
do
skip
push
true
there
when
you're
doing
that
this
doesn't
the
annotation
here
doesn't
mention
the
Google
Cloud
builder
I
assume
that's
pr2
local,
but
it
probably
has
its
own
own
ways
of
doing
stuff
and
then
deploy
here.
A
A
There's
there's
probably
an
injection
injection
method
here,
and
maybe
you
look
for
the
existing
image
name
and
you
just
replace
that
something's
super
duper
dumb
and
simple,
like
some
sort
of
sentinel,
like
you
know,
handlebars
type
of
thing,
but
then,
if
you're,
not
careful
that
starts
to
that
starts
to
you
know,
replace
templating
solutions
here.
So
so
that's
something
that
I
want
to
dig
into
here.
Also,
okay,
so
cool!
So
that's
a
little
bit
more
annotation
on
what's
going
on
there,
then
we
have
one
that
has
no
manifest.
What
is
no
manifest?
A
Look
like
so
this
says,
deploy
manifest
you
just
don't
specify
anything,
and
it's
like
I
got
nothing
to
do.
Okay,
scuffles!
It's
careful!
This
will
deploy
simple,
okay,
okay,
if
you're
using
keep
control,
deplore,
you
can
deploy
images
without
and
company
kubernetes
manifests
simple
deployment
template.
Oh
okay,
so
it
just
assumes
that
it
has
a
deployment
template.
A
That's
nice
and
then
microservices
is
going
to
be
something
we
got
leroy
web
and
app.
How
do
we
actually
do
multiple
I'm
gonna
read
the
readme,
because
that's
how
I
roll
deploy
multiple
applications
with
scaffold
only
rebuild
and
redeploy
the
artifacts
that
have
changed?
Okay,
so
there's
a
so
Oh
since
scaffold
probably
says:
hey
I
see
the
image
tag
show
up
in
this
yeah
mo,
so
I
only
need
to
redeploy
that
yeah
mole
versus
actually
redeploying
the
other
stuff.
So
scaffold
Deb
now
in
terminal
hit
that
Leo
it
Roy
a
web
in
point.
B
A
A
B
A
A
B
A
So
we're
I'm,
confused,
cuz,
there's
also
helm.
Template
parameter
is
a
command
line
tool
and
kubernetes
operate
for
Henley
application,
lifecycle
management.
So
this
looks
like
oh.
This
looks
like
I
wonder
how
much
like
yet
another
sort
of
like
way
to
specify
and
sort
of
put
stuff
in
there.
Okay!
Well,
alright!
So
that's
an
interesting
thing:
I'm
going
to
save
that
in
a
tab,
we'll
definitely
put
it
in
the.
A
Going
to
be
a
gazillion
opinions
on
on
templating
languages,
here,
okay,
so
this
says
so
we're
gonna
go
ahead
and
do
this
and
this
one
assumes
mini
cube
also,
but
we're
not
doing
mini
cube,
so,
let's
go
through
and
see
if
we
can
fix
it
up
for
non
mini
queue.
So
we
have
a
scaffold.
So
this
says:
okay,
so
I'm,
building
two
things
here
and.
A
A
A
B
A
Michael
says,
it
allows
the
a
proper
use.
Any
language
like
Helmer
kisana
came
too
late.
All
it
does
is
automate
the
transformation
steps
that
seems,
like
maybe
I,
don't
know,
you've
got
to
think
about,
because
something,
because
we
shouldn't
assume
that
all
that
inputs
are
always
going
to
be
like
a
JSON
sort
of
values,
file,
I
think,
there's
always
there's
going
to
be
more
interesting
ways
to
capture
this
stuff
in
the
future.
It's
an
interesting
thing
all
right.
A
All
right
and
let's
look
at
these
things,
so
this
one
is
they're
both
go
files
ones
calling
from
golang
10.
So
that's
why
we're
actually-
and
this
is
the
full
golang
10.
This
is
not
the
like
Alpine
version,
so
we're
pulling
down
a
heavy
duty
base
image
here.
But
obviously
this
is
a
one-time
thing.
You
saw
that
before
that
this
stuff
gets
gets
so
for
Amazon
with
ACR,
because
you
have
to
explicitly
go
through
and
specify
what
all
the
images
are.
The
image
repositories
that
you
want.
A
It's
a
pain
in
the
butt,
because
you
have
to
sort
of
be
very
explicit
upfront,
but
Amazon
is
much
more
flexible
than
GCR.
With
respect
to
being
able
to
set,
you
know
individual
permissions
for
access
to
different
repositories,
and
so
with
with
GCR.
You
can
essentially
have
one
policy,
for
you
know,
essentially
your
entire
all
your
images
in
a
particular
GCP
pop
project
with
ACR,
it's
much
more
flexible
about
being
able
to
specify
access
policies
on
a
registry
by
registry
basis.
Alright,.
A
App
server
ready.
We
see
that
now.
If
I
go-
and
this
is
both
oh
and
this
show
somebody
forgot-
I'm
gonna
fix
that.
Let's
fix
that.
So
that's
a
Leeroy
app
server
ready,
but
it's
not
the
app
server.
It's
the
web
server.
So
let's
go
this
and
I
bet
you
there
with
some
copy
and
paste
going
on
here.
We're
gonna
go
app
server
ready
boom.
Let's
fix
it!
It's
rebuilding
its
rebuild,
pushing
there
we
go.
Oh
I
used
leroy
app,
I
screwed
up.
Oh,
this
is
gonna,
be
interesting
here
if
I
say,
if
I
change.
A
On
the
web,
when
I
look
at
this,
if
I
change
my
yeah
mole,
does
it
automatically
actually
detect
that
as
part
of
it,
Leeroy
web
go
ahead
and
do
that,
but
they
did
screw
up
and
actually
say
this
is
Leroy
Webb,
okay,
okay,
what's
going
on
here,
starting
the
deploy
watching,
did
it
already
happen?
I,
don't
know.
A
A
No
okay,
that's
the
images!
Okay,
here's
the
images
here
so
they're
both
running
out,
so
something
didn't
recognize.
My
change
to
the
web
server
here.
I
think
we
got
scaffold
wedged
again,
I'm
sure,
there's
something
silly
going
on
there.
So
all
right.
So
this
is
doing
the
redeploy,
and
now
we
have
web
and
app
okay,
so
we're
up
and
running
all
right,
and
so
let's
see
okay.
A
Which
is
okay
because
I
think
what
you'll
find
is
that,
like,
let's
say
that
you're
doing
no
development,
you
know
developers
understand
that
hey,
you
know.
My
regular
flow
is
I'm
doing
stuff
will
work
fine,
but
there's
going
to
be
certain
things
that,
if
I
change
something
that's
really
sort
of
fundamental
to
my
web
server
or
whatever
I
really
need
to
go
ahead
and
restart
the
web
server.
That
type
of
thing,
if
you
look
at
something
like
Hugo,
also
it'll,
also
automatically
do
like
a
reload
type
of
thing.
A
But
if
you
change
some
of
the
core
parameters,
you
really
need
to
just
sort
of
restart
the
server
daemon.
So
that's
not
I
mean
one
of
the
interesting
things
you
know.
Matt,
that
might
be
a
cheap
way
to
deal
with.
This
is
to
essentially
say:
hey.
I
noticed
something
changed
that
I
can't
handle.
You
probably
want
to
restart
scaffold.
Dev
actually
have
a
deal
with
that.
Okay,
now
it
says
Leroy
webserver,
ready.
A
B
A
What
happened
there?
Let's,
let's
look
at
the
ammo
Leroy
app
cluster
IP,
none!
Okay!
So
it's
type
equals
cluster
IP,
but
saying
don't
give
me
a
cluster
IP,
interesting
I
believe
what
happens
with
this
is
that
you
then
get
you
get
individual
DNS
entries.
That's
a
headless
service
yeah!
It's
only
supposed
to
be
called
by
Leroy
Webb
and
then
what's
it
like
okay.
A
So
this
actually
says
it's
up
to
the
client
to
do
whatever
load,
balancing
makes
sense
and,
and
normally
what
happens
here-
is
that
the
order
of
these
things
gets
rotated
by
the
DNS
server.
Every
time
you
do
it
I,
don't
know
exactly
which
DNS
server
we're
running.
Then
then,
one
nine
four
mini
cube
or
nine
for
hefty
Oh
QuickStart,
but
but
it's
not
doing
this
rotation
here.
So
that's
something
that
is
unfortunate.
A
A
Yeah,
so
you
get
get
to
see
the
logging
across
all
of
these
things
as
the
web
server
comes
up
and
then
the
other
ones
we're
left
running.
Okay.
So
now
this
is
the
cool
thing
that
I'm
super
excited
about.
If
I
do
a
cube,
control,
port
forward
and
I
can
do
deployment,
slash,
Leeroy
web
and
then
I
want
to
what
port
is
it?
It's
it's
8080
and
I.
Do
that
it's
like!
A
B
A
A
Leroy
app
and
we'll
do
s,
and
then
what
is
it?
Os,
hostname
well
go
ahead
and
we'll
hit
that
that's
gonna
do
a
rebuild
here,
which
is
what
we
expect.
Oh,
oh,
it's
mad
at
me,
okay,
so
it's
mad
at
me.
So
this
is
interesting.
So
what
happens
when
you
see
an
error?
It
aborts
and
it
doesn't
do
the
push
or
the
run.
So
that's
actually
so
we're
going
to
be
doing.
That's
a
good
part
of
the
development
process
here
is
that
you
can
actually
see
this
now,
I.
A
Think
if
I,
if
you
were
doing
like
a
scaffold
dev
in
like
a
terminal
window
in
your
browser,
then
you
can
go
through
and
probably
just
go
ahead
and
click
on
this
thing
and
that'll
be
nice.
You
know
if
we
see
tools
like
this
get
better
and
better.
What
you
may
find
is
that
there's
a
continuous
build
going
on
in
the
background
and
those
errors
get
integrated
with
your
IDE
in
a
good
way
where
they
actually
show
up
automatically
so
that'll
actually
be
really
cool.
So
now,
I
want
to
do.
A
A
A
But
now
it's
why
watching
for
changes?
It
says.
Okay,
now,
the
all
the
App
servers
are
ready,
they've
pulled
it
I
go
here
and
we're
always
getting
the
first
one
which
is
xjm
to
G
and
if
let's
go
through
and
in
Leroy
web,
what
it's
doing
is
it's
actually
going
through
and
just
doing
a
get.
So
it's
not
doing
anything
special,
it's
using
the
resolver,
that's
built
into
going,
and
so
I'm
gonna
go
through
I'm
going
to
be
like
we
don't
want
none
for
cluster
IP
here.
A
B
A
A
A
And
so,
okay,
so
now
this
thing
deployed
so
like
that's
not
what
we
would
expect
here
so
actually
sort
of
like
you
know.
So
that's
a
test
case
for
cubature
will
apply.
Is
that
if
I
actually
have
cluster
IP,
none
and
I,
actually
it
it
doesn't
actually
turn
that
back
into
a
default.
The
way
apply
it
works.
Is
it
thinks
that
you're
actually
setting
it
to
a
null
value?
So
there
we
go
so
now.
A
A
A
B
A
A
A
A
A
No,
that's
like
the
a
transport.
How
do
I
get
the
transport
alright
I'm,
totally
down
a
rat
hole
here,
okay,
Vic
is
trying
to
trying
to
type
stuff,
I,
think
I.
Think
in
the
transport
you
can
do
disable
keep
alive
true
right
but
like
we
don't
have
a
transport
and
we're
just
doing
the
simple
HTTP
GET.
Alright,
let's,
let's
I'm
down
a
rat
hole,
I'm
gonna
find
my
way
out
to
the
other
side.
A
B
A
A
Think
that'll
get
us
there
now
we
as
we've
been
doing
this
okay.
So
so,
if
we
were
developing,
we
would
expect
that
scaffold
in
the
background
would
be
like
building
it
by
the
time
I
switch
back,
I'd
have
an
up-to-date
one,
that's
been
built
and
pushed,
but
but
scaffold
got
wedged
so
pretend
that
happened
and
all
this
stuff
was
happening
while
I
was
typing
in
the
other
window,
then
I'm
like
okay,
now,
oh
I,
director,
oh,
my
port
forward
got
lost.
Okay.
A
A
But
we
wouldn't
have
noticed
the
issue
here,
necessarily
where
we
actually
have
ourselves
pinned
to
a
single
back-end
right
and
so
by
actually
doing
development
in
a
environment.
That's
as
close
as
possible
to
our
production
environment,
which
is
kubernetes,
then
we're
able
to
actually
test
this
thing
out
in
a
realistic
way,
which
is
actually
really
cool,
all
right,
so
I'm
gonna
leave
it
there
because
I'm
an
hour
and
a
half
in
I'm
really
excited
about
scaffold.
I
think
it
looks
really
interesting.
I
love
the
the
focus
on
providing
a
framework.
A
I
guess,
that's
the
name,
scaffold
the
focus
of
providing
a
framework
here
for
how
you
can
actually
think
about
this
cycle
and
and
then
having
plug
ability
in
terms
of
the
tools
that
you
use,
whether
it
be
a
local,
build
or
cloud
builder,
or
some
fancy
new
thing
that
comes
along
in
the
future.
I'm
also
interested
to
see,
like
you
know,
are,
is
there
room
to
actually
specify
things
that
move
beyond
docker
files
being
able
to
use
a
queue,
control
versus
the
versus
the
helm
stuff?
That
seems
really
interesting.
A
A
B
A
Now
you
do
that
in
a
blind
way
like
I
guess
what
I'm
trying
to
say
is
that
if
I
had
a
CR
D
with
my
own
custom
controller,
because
I'm
like
hey
I,
want
to
have
a
super-duper,
cron
job
or
I
want
you
know
or
super-duper
deployment
like
let's
say
that
I'm
using
the
Dometic
controller,
with
its
Bluegreen
deployments,
er
D
and
I
want
to
mix
that
with
scaffold
now
I'm,
actually
using
a
type
of
object
that
that's.
That
is
not
that
scaffold
probably
doesn't
know
about,
because
it's
a
CR
D.
A
Will
you
look
for
that
string
that
shows
up
anywhere
or
will
you
actually
look
for
that
in
a
very
specific
place
and
I
think
that's
some
of
the
you
know
it
might
be
worthwhile
to
think
about.
Okay,
you
have
this
default
way
where
you
recognize
the
stock
kubernetes
image,
but
maybe
having
a
build
step
there,
where
you
could
run
some
sort
of
like
program
that
says
hey,
you
know:
here's
a
temporary
directory
where
we've
staged
all
the
ammo
files.
Why
don't
you
go
ahead
and
run
whatever
program?
A
You
want
to
that'll,
go
ahead
and
find
the
right
thing,
and
you
know
and
substitute
in
the
image
there.
So
that
might
be
another
extensibility
point
to
think
about,
because
I'd
love
to
see
this.
You
know
working
with
something
like
the
meta
controller
and
a
Bluegreen
deployment
to
really
start
to
see
how
these
things
might
actually
work
together
with
CDs
all
right.
So
that's
it.
Oh
and
I
was
yeah,
so
I.
Let
me
switch
back
here,
so
it
does
look
like
there's
some
good
documentation,
that's
actually
being
built
up
here,
which
is
good.
Okay.
A
A
I
know
a
lot
of
folks
who
are
joining
me
from
from
faces
that
are
far
places
that
are
far
from
flung
have
already
started
their
weekend
and
I
will
be
back
next
week,
I'm
going
to
draft
the
week
after
that,
I'm
going
to
be
on
vacation
me
and
the
family
are
going
to
be
climbing
down
into
the
Grand
Canyon
and
climbing
back
out
and
if
I
survive
that
the
week
after
that
I
think
I'm
going
to
be
looking
at
telepresence.
So
thank
you
very
much
and
I'll
see
you
all
later
good,
bye.