►
Description
Join us for Kubernetes Forums Seoul, Sydney, 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
Tutorial: Cloud-Agnostic Serverless - Sebastien Goasguen, TriggerMesh
Serverless design patterns have grown in popularity because they allow developers to move faster by focusing on business logic without worrying about the underlying infrastructure where their code is run. However, many enterprises use diverse infrastructure including on-premises servers and multiple clouds. In this tutorial, we will leverage Knative, Google's Kubernetes-based open source platform to build, deploy, and manage modern serverless workloads. We will push serverless functions and apps to production on any cloud of choice and switch the provider as necessary. We will leverage GitLab and TriggerMesh technology in the tutorial and also share how developers can use other options.
https://sched.co/MPgx
A
B
A
I
went
to
bed
at
2:00
a.m.
because
something
broke
in
this
tutorial
and
I
should
fix
it.
So
I'm
a
little
bit
more
stressed
than
usual,
but
I
think
it
will.
It
will
do
fine
all
right
so
we're
here
to
talk
about
clock
that
agnostic
server
life
with
Kay
native
and
we've
done
this
tutorial.
You
know
I'm
the
founder
of
trigger
mesh
and
we've
done
it
in
partnership
with
github
the
development
trigger
mesh.
We
worked
actively
together
and
we've
developed
the
integration
of
K
native
inside
github
alright.
A
So
what
we
want
to
show
in
this
tutorial
is
that
you
can
use
you
know
if
you
have
code
in
in
get
lab,
you
know
specifically,
but
potentially
the
other
places.
You
can
actually
use
K
native
primitives,
a
native
API
to
deploy
server
lights,
applications
on
any
cloud,
any
cloud,
meaning
basically
anywhere
where
you're
going
to
have
a
cube
unit,
is
Questor
on
which
you
install
k
native
right.
A
It's
a
tutorial!
So
I
have
some
slide.
You
should
have
gotten
the
the
URL,
but
we're
gonna
try
to
do
things
hands-on
right
and
that's
why
the
folks
from
gitlab
have
given
you
some
little
pieces
of
paper
with
the
login
password
to
test
instance
right.
So,
if
you
didn't
get
the
credentials,
raise
your
hand
and
we'll
get
you
the
credentials
all
right.
C
A
I'm
Sebastian
question.
Helping
me
today
is
coolant
Taylor
with
Killian
all
right.
There
raise
your
hand,
killing,
doesn't
hear
me.
Okay,
he
actually
wrote
a
lot
of
the
tutorial.
Chris
right
here,
raise
your
hand,
Chris
actually
develop.
The
integration
of
K
native
inside
the
github
priyanka
was
supposed
to
be
with
me
here
today
from
get
lab,
but
she
was
with
used
exit
from
the
US.
That's
that's.
A
I've
been
working
on
cavities
for
quite
a
while.
Almost
since
the
beginning,
I
developed
with
other
people
like
Chris
things
like
compost,
K
machine
cube,
less
one
of
the
first
survey,
less
solution
on
top
of
Cuba,
nineties,
cabin
and
so
on,
and
right
now
created
trigger
mesh
I.
Also
do
a
bunch
of
things
with
O'reilly.
A
Maybe
some
of
you
have
taken
some
of
the
O'reilly
courses
that
I
give
online
okay,
so
just
prereq
for
this
tutorial,
the
little
sign-in
card
you
have
a
username
workshop
user,
1,
2,
3
200,
you
have
the
password,
everybody
has
the
same
password,
don't
freak
out,
will
scratch
everything
in
two
hours
and
everybody
can
log
into
HTTP
GET
lab
10
key
that
post
all
right.
That's
a
good
lab
instance.
We're
gonna
connect
to
this
github
instance
and
from
there
we
are
going
to
try
to
deploy
service
applications
or
functions
alright.
A
So
that's
the
goal
now,
of
course,
you're
welcome
to
do
it
and
if
you
just
want
to
you,
know,
watch
this
under
the
hood.
There
are
things
that
we're
going
to
use
that,
hopefully
we're
you
know,
are
fairly
transparent
and
the
goal
of
service
is
to
make
all
the
tooling
and
all
the
infrastructure
transparent,
but
we
do
use
a
Force
Q
by
T
Square,
sir.
There
is
a
little
bit
of
cube
control.
A
C
C
A
And
the
lab
content,
what
we
try
to
do
so
it's
quite
challenging
in
an
hour
and
a
half
an
hour,
20
minutes
to
try
to
do
you
know
what
we're
gonna
try
to
do,
okay,
but
with,
but
you
know
we
would
try
to
do
for
lab.
Okay.
If
everything
works,
fine,
it's
going
to
be
super
smooth.
Okay,
that's
my
hope.
So
get
lab
come
get
lab
workshops!
Survey!
Let's
workshop!
Okay!
This
repo
is
actually
cloned
in
the
github
instance,
where
you
can
log
in
so
you're
going
to
log
in
and
to
those
posts.
A
You
got
the
card
when
you
log
in
you're,
going
to
see
a
project
or
repository.
If
you
come
from
Ghita
blend
and
you're
going
to
actually
see
the
content
of
those
labs
right,
we're
going
to
try
to
do
all
the
labs
here
right
now,
but
of
course,
when
you
can
go
home,
you
can
go
back
to
those
labs
and
try
to
do
it
on
your
own
good,
I.
A
Hope!
That's
why
you
came
here
to
do.
You
know
hands-on
stuff
right
so
agenda
I'm
going
to
do
try
to
keep
it
short
introduction
about
server
less.
This
is
not
a
presentation
about,
you
know:
server
less
per
se,
it's
really
tutorial
and
labs
and
then
we're
gonna
try
to
do
the
labs,
hands-on
self-paced.
If
you
go
really
fast,
go
really
fast,
you
can
leave
we're
going
to
do
lap,
one
two
three
try
to
demo
them
live
and
then
hopefully
a
lot
of
you
gonna
be
successful.
A
B
A
Cool
try
to
keep
it
short
of
little
introduction
about
server
less,
so
it's
very
challenging
right
now,
at
least
that's
my
that's
my
personal
view.
We
have
a
bunch
of
IT
landscapes
and,
as
professional
it's
very
difficult
to
navigate
those
landscape.
Even
the
CNC
at
landscape
is
becoming
very
challenging
to
navigate
all
right,
so
we're
being
bombarded
with
new
tech
every
day
and
Kay
native
is
no
exception.
You
may
have
heard
about
tentative.
You
may
have
heard
about
server
less
you're
being
bombarded
with
those
new
paradigm
and
the
question:
is
you
know?
A
What
do
you
do
with
those?
How
can
you
adopt
them?
Are
they
going
to
help
you
solve
issues
and
so
on
right?
So
we're
being
bombarded
with
this
I
try
to
find
a
nice.
You
know
picture
to
bombardment,
but
a
apart
from
planes,
you
know
didn't
find
anything,
and
then
the
CNC
F
landscape
is
quite
overwhelming
right
and
you
see
the
growth
here
in
the
conference.
It's
been,
it's
been
quite
nuts,
the
last
you
know
four
years,
but
this
is.
A
This
is
very
difficult,
and
the
problem
also
is
that
you
know
even
ourselves
in
our
community.
We
tend
to
say
hey
to
the
future,
and
there
is
this
great
blog
I,
don't
if
you
know
it
from
circle
CI
that
tells
you
you
know
it's.
It
goes
basically
through
the
cloud
native
technology
and
it's
one
person
who
just
wants
to
grow
a
blog
and
the
other
guy
says.
Oh
well,
you
know
it's
easy.
A
Just
take
docker
and
you
know,
put
a
load
balancer
in
front
and
then
you
need
a
fleet
of
machine
and
you
need
a
container
optimized
operating
system
and,
at
the
end
of
the
day,
that
the
person
who
just
wanted
to
deploy
a
website
ends
up
having
to
deploy
a
whole
slew
of
things
right
so
really
check
this
check
out
this
blog.
It's
very
funny,
but
it's
actually
quite
true,
and
sometimes
you
know
we
we
end
up
in
this
type
of
description
and
the
truth
with
that
landscape.
Is
that
it's
very
complicated.
A
You
know
you
end
up
creative
cluster
in
the
cloud
install
a
container
runtime
install
an
Orchestrator,
install
an
app
package.
Er
then
install
two
or
three
of
those
other
systems
using
hair.
On
top
of
your
Orchestrator
now
deal
with
the
net
networking
and,
of
course,
install
a
service
mesh,
you
know
and
then
finally,
maybe
you'll
get
to
actually
deploy
in
your
app,
and
you
know
it
will
be
broken
down
in
lots
of
micro
services.
Right,
if
you
do
all
of
this,
you
can
benefit
from
scale
you're
going
to
benefit
from
automation.
A
You
know
and
and
velocity
that's
the
hope
all
right,
but
the
truth
is
that
this
is
actually
super
complicated.
So,
of
course,
when
we
have
completely
a
complication
like
this,
the
solution
is
to
say:
hey:
we
need
new
abstraction.
We
need
a
new
and
we've
been
doing
this.
You
know
for
a
long
time,
and
hopefully,
with
those
new
attractions,
we
get
simplicity,
so
serve
relax.
A
You
know
it's
basically
trying
to
abstract
a
lot
of
that
complexity
of
managing
the
infrastructure,
so
that,
obviously
you
can
focus
on
your
application
and
deploying
that
those
application
right
main
component
of
server
less
even
driven.
So
there
are
events
flowing
through
the
system
and
you
use
those
events
to
trigger
little
micro
services
of
functions.
It's
using
a
lot
of
cloud
provider
services,
maybe
storage
system,
machine
learning
system,
you
name
it
and
then
the
monetization
aspects
that
you
only
pay
when
you
pull
your
your
functions
right.
A
A
B
A
Type
of
application
coming
from
a
WS
lot
of
pictures
in
storage
bucket
in
s3
automatically
triggers
thumbnail
creation,
triggers
PDF
creation,
if
you're
putting
in
Word
documents
and
then
put
them
in
another
bucket.
You
know,
for
example,
file
processing
stream
processing.
You
have
lots
of
messages
coming
on
a
an
event:
stream,
kinases
or
Kafka.
You
name
it
triggers
functions,
stick
something
into
a
database.
You
know,
keep.
A
You
insert
things
into
a
database,
you
know
dynamo
DB,
but
could
be
also
on-prem
database
triggers
a
function,
do
some
processing
and
then
archive
the
the
result.
All
right.
All
these
comes
straight
out
of
AWS,
okay
and
at
a
high
level,
all
these
are
application
pipelines
that
are
fairly
simple,
but
the
services
that
are
being
used
for
the
storage
for
the
eventing,
the
database
and
so
on.
All
those
are
services
that
are
that
can
scale
and
that
are
very
resilient,
all
right.
A
So
what
has
happened
is
that
in
terms
of
server
less,
you
know
again,
AWS
is
the
leader
all
right.
They
are
the
leader
with
the
AWS
lambda
solution,
all
right.
What
happens
if
you're,
not
an
AWS
user?
Okay,
what
happens
if
you
want
to
build
that
ETL
pipeline?
If
you
want
to
build
that
IOT
application,
you
know
in
a
very
simple
manner.
You
want
that
you
want
to
make
sure
that
it
scales
and
things
like
this,
but
maybe
you
want
to
do
it
on
as
your.
Maybe
you
want
to
do
it
on
Google
Cloud.
A
A
So
the
challenge
you
know
is:
how
do
you
do
this?
How
do
you
build
those
apps
without
too
much
lock-in
and
that's
why
we're
using
open
source?
And
we
are?
We
are
those
companies
potentially
with
some
services
that
are
on
Prem
right
and
then
limit
the
operational
cost
limit,
the
the
view
of
the
infrastructure
for
your
developers,
your
users,
things
like
that,
make
sense,
I'm
rushing
a
bit,
but.
A
Recently,
Google
announced
cloud
run,
which
is
you
know,
one
of
their
new
service
offering
and
it
runs
on
top
of
their
internal
board.
Maybe
it
runs
on
on
humanity's
I,
don't
know,
but
most
likely
it
runs
on
their
internal
board
system
right.
But
if
we
want
to
do
something
like
cloud
run,
but
in
a
different
cloud
and
maybe
unfriend,
it
makes
total
sense
to
try
to
build
that
on
top
of
your
kubernetes
infrastructure.
How
can
you
do
this?
You
do
this
with
Ken
ativ.
What
is
K
native?
A
It's
an
extension
of
Cuba
natives
extension
of
the
communities
API
that
gives
you
things
like
automatic
scaling,
better
deployment
strategies,
traffic
splitting.
You
know
lots
of
benefits.
You
know
you
get
them
from
the
cognitive
abstractions
all
right,
so
you
create
a
Kenny
TVPA
object.
Just
one
and
automatically
you're
gonna
get
a
route
for
traffic
splitting
you
can
I
get
an
original
pod
autoscaler.
A
You
know
you're
going
to
be
able
to
yeah.
Do
it
do
automated
builds
and
then
you're
going
to
be
able
to
declare
how
you
want
events
to
flow
and
things
like
this
all
right.
So
that's
the
hope
of
this
QA.
This
extension
is
to
give
you
all
the
the
primitives
so
that
you
can
build
those
server,
less
pipelines
and.
A
A
You
can
customize
the
qat
api,
so
I
talked
about
CR
DS.
You
know
quite
a
bit
in
in
different
settings,
but
let's
say
that
you
want
to
create
a
new
object
of
the
database.
You
can
actually
write
a
manifest
that
declares.
What
is
a
database
object?
You
create
this
object
and
suddenly
kubernetes
is
aware
of
a
database.
Okay,
just
from
the
API
standpoint,
once
you've
extended
the
API,
what
you
need
to
do
is
create
controllers.
Okay,
they're
also
called
operators.
So
that's
why
you
hear
things:
maybe
you
build
there?
A
Yesterday
there
was
a
workshop
on
queue
builder.
There
was
a
workshop
from
reddit
on
operator
framework,
so
once
you've
extended
the
community's
API
with
crd,
you
need
to
write
a
controller
right.
That's
actually
going
to
do
things
right.
You
create
a
database.
Now
database
object.
Now.
What
do
you
do?
Well,
you
need
to
go
somehow
launch
my
sequel
or
lounge
post
bread
somewhere
right.
So
you
need
to
write
the
brain
of
that
deal
with
your
CRD.
So
when
you
install
K
native,
you
get
a
bunch
of
CR
DS
if
you're
a
developer.
A
You
don't
really
need
to
know
about
this,
but
I'm
giving
this
to
you
as
if
you
are
more
the
operator
persona,
so
that
you
know
what's
happening
a
little
bit
under
the
hood.
Okay,
even
though
we
want
to
abstract
from
it.
So
you
install
can
a
tip.
You
get
lots
obscurities
and
with
those
theories
you're
going
to
be
able
to
build
those
service
applications
all
right.
So
we've
extended
communities.
A
B
A
So,
just
like
any
new
paradigm,
people
come
up
with
names,
and
then
people
are
new
about
the
name.
So
people
argue
that
serverless
is
a
really
bad
name
right,
because
they
are
servers
behind
to
run
your
process
right,
but
there
is
a
somebody
in
your
Patrick
Dubois
with
one
of
the
instigator
of
the
DevOps
movement.
He
gave
a
talk
couple
years
ago
and
instead
of
talking
about
service,
he
talked
about
service
fool,
which
is
that
you
know
it
was
actually
developing
application
in
still
developing
applications.
Cloud
natively.
A
So
in
the
cloud
right
and
his
problem
is,
the
I
mean
is
that
is
using
a
lot
of
services
is
using.
You
know,
RDS
is
using
x3
is
using
kinases,
so
he's
using
a
bunch
of
services.
So
it's
application.
You
say
you
know
it's
not
service,
it's
service,
full
all
right
and
to
me,
that's.
You
know
it's
very
true.
It
resonates
because
what
you're
trying
to
do
when
you
build
those
pipelines,
your
trunk
you're,
trying
to
write
glue
so
that
you
can
connect
all
those
services
so
service
full.
It
comes
back.
A
A
If
you
don't
have
the
card,
you
can
work
with
your
neighbor
we're
going
to
log
in
to
this
get
lab
instance
and
we're
just
going
to
run
a
pipeline
and
get
that
pipeline
and
what
that
pipeline
is
going
to
do
we're
going
to
look
at
it.
That
pipeline
is
actually
going
to
talk
to
a
K
native
installation
which
is
the
trigger
mesh
in
the
trigonis
cloud
and
inside
that
trigger
mesh
cloud.
It's
going
to
use
the
kinetic
API
to
build
a
Python
function
and
make
it
available
right.
A
C
A
A
A
A
And
you
notice
that
this
function,
so
very
simple,
Python,
just
one
function
and
point
returns,
hello.
The
current
time
is
and
prints
the
time
you.
If
you
look
at
this
function,
it's
exactly
like
what
you
can
do
in
a
double
J's
lambda.
There
is
no
HTTP
server.
There
is
no
nothing!
All
of
that
bring
a
comes
in
when
you
deploy
the
function.
Jaka
native
okay,
so
that's
just
simple
function.
So
what
we're
going
to
do
is
just
go
to
see.
Icd
pipelines.
A
A
A
Defines
stages
so
different
baby
steps,
let's
call
them
like
this
and
then
you
have
deploy
hello
function
and
you
see
that
in
this
deploy
hello
function,
kind
of
stage
what's
being
used
is
an
image,
a
docker
image.
This
docker
image
comes
from
us
and
that
docker
image
has
a
client
that
can
talk
to
the
K
native
API,
okay,
and
then
it
does
TM.
A
Anybody
here
familiar
with
the
server
less
framework,
all
right.
So
it's
it's,
not
the
server
less
framework,
but
it's
supposed
to
be
fairly
compatible
and
it
resembles
the
service
framework.
Okay.
So
what
you
have
in
the
in
this
here
is
that
you
have
the
list
of
functions
that
you're
deploying
so
here
you
see
an
L
o
function.
A
It
tells
you
where
the
source
of
the
function
is
so
lap,
one
hello,
and
then
it
tells
you
a
runtime,
which
is
what
is
going
to
basically
package
that
function
and
in
the
runtime
you
end
up
with
like
an
HTTP
server,
so
that
we
can
call
the
function
and
and
so
on,
put
a
little
description
and
then
some
some
build
arguments,
because
this
little
function
needs
to
somehow
be
injected.
You
know
somewhere
and
ends
up
in
a
container
so
that
you
can
call
that
that
little
bit
of
Python
alright.
B
A
Let
me
try
so
technically.
This
is
self-paced
so
now
that
you're
logged
in
on
the
instance
I'd
like
you
to
go,
try
to
go
through
the
labs.
Okay,
try
to
do
the
lab
one
and
I'm
going
to
try
to
see
what's
going
on
here
and
then
we're
gonna
try
to
do
lap.
Two!
Okay,
let's
take!
Let's
take
ten
minutes
to
do
this.
All.
A
C
A
A
C
A
So
here
I'm
logging
in
into
the
cloud
that
you're
targeting
and
I'm,
seeing
all
the
pods
right.
So
while
we
do
all
of
this,
we
end
up
with
humanity
spot
running
that
should
be
exposing
your
functions
right.
So
I
see
that
I
see
some
of
the
pods
and
when
I
look
at
these
objects,
services,
I'm
going
to
explain
what
that
is.
If
I
click
on
this
I.
A
Okay,
let's,
let's
give
it
a
little
bit
of
time,
let's
see
if
you
know
the
system
catches
up
somehow
so
that's
you
know
the
basic
thing
which
obviously
didn't
work
out.
You
know
exactly
out
of
the
box.
What
happened
you
know
at
the
end
is
that
khane,
a
tip
should
have
created,
what's
called
a
kinetic
service
and
a
tentative
service.
Basically,
under
the
hood
is
going
to
create
a
community's
deployment,
you
can
I
get
a
pod,
but
you
also
get
things
like
an
auto
scaler
and
you
get
a
route
which
does
you
know?
A
Did
it
be
an
equivalent
to
an
ingress
object
right?
So
in
one
API
object,
you
encompass
several
things
that
give
you
a
to
scaling
scale
to
zero
deployment
strategies
and
so
on
to
create
those
those
service,
those
can
achieve
services
you
have
in
the
connected
serving
namespace.
You
know
you
have
a
controller,
so
you
have
the
brain
that
you
know
looks
at
those
can
ative
CR,
DS,
k
native
exposes.
A
You
know,
as
I
said,
services
routes
configuration
revision,
so
those
are
new
object,
new
construct
that
you
have
in
your
communities,
cluster
and
it
as
I
said
a
higher
level
than
the
basic
UNC's
deployment.
One
thing
that
is
interesting
is
that
you
have
these
red
visions,
which
gives
you
immutable
snapshot
of
your
apps,
plus
your
configure
right.
So
every
time
you
make
you
know,
potentially
an
update
to
your
your
function.
You
make
an
update
to
the
variable
you
give
to
that
function.
A
What
under
the
hood?
What
does
a
can
achieve?
Service
looks
like
it
looks
like
this:
that's
its
most
basic
form
and
at
the
beginning
it
was
a
little
bit
confusing,
because
what
you've
seen
this
manifest
that
you
see
a
container.
So
there
is
a
container
called.
You
know
this
year,
iokay
native
samples,
hello
world.
So
there
is
no
concept
of
functions
here.
You
know
at
least
at
the
beginning.
A
What
you
have
is
that
you
have
a
container
okay,
that
you
have
pre-built,
but
you
deploy
it
with
tentative
a
service
and
you
get
the
auto
scaling
scale
to
0
you
get
the
routing
and
so
on.
That's
exactly
what
Google
cloud
run.
Does
ok
and
again
in
the
gitlab
CI,
when
you
do
this
TM
deploy
it
creates.
You
know
nothing
else
than
just
disconnected
service
object
and
post
it.
On
the
on
the
end
point
that
service
object
is
generated
from
server.
That's
the
general
manifest
which
is
more
familiar
to
people
who
do
lambdas
alright.
A
E
A
E
A
A
A
You
know
the
idea
here
is
that,
yes,
if
we
expose
container
it's
not
much
of
a
function,
but
in
the
end
you
know
anything
is
going
to
be
represented
as
a
as
a
container
container
is
a
great
artifact
to
to
deploy
right.
So
if
you,
if
you
look
Kennedy,
has
a
system
called
the
Kennedy
build
that
allows
you
to
actually
build
a
docker
container
inside
acuities
cluster.
All
right,
you
can
do
build
I
put
the
cat
here,
just
as
a
side
note
that
kennedy
build
actually
gave
rise
to
something
called
Tecton
CD.
A
Yet
another
project
in
the
landscape-
and
you
know
Tecton
CD-
is
now
part
of
the
cloud.
Continuous
delivery
foundation.
Okay,
so
we
pass
on
that,
but
because
we
have
gitlab
CI,
we
can
actually
build
containers
directly
in
gitlab.
Okay,
so
get
lab
is
very
good
at
this
and
the
way
github
does
the
does
it
is
that
it
uses
connect.
Oh
right,
have
you
heard
about
Kanak?
Oh
yeah.
A
Kaneko
is
a
great
tool
which
allows
you
to
build
a
docker
image
without
having
a
docker
demon
right
and
without
having
privileged
containers
that
may
have
to
mount.
You
know
docker
socket
and
things
like
this.
So
if
you
use
Kaneko,
you
can
actually
build
a
docker
image
within
another
docker
container.
That
means
that
you
can
build
a
docker
image.
We
then
a
curate
this
pod,
so
that's
extremely
extremely
powerful.
All
right
get
lab.
A
A
So
clearly
there
is
something
wrong,
but
when
I
go
back
and
that
I
cannot
hit
it,
you
know
it
may
actually
be
good
behavior,
because
if
you
don't
hit
your
application,
there
is
a
scaling
to
zero
right.
So
if
you
don't
hit
it
scales
to
zero,
no
pods
are
serving
that
function.
First
time
you
hit
it
that
should
trigger
a
wake-up
and
the
deployment
should
scale
up
to
one
all
right.
So
well,
here
it
doesn't
respond.
You
know
it's
not
always
a
bad
sign.
It
may,
actually,
you
know,
come
up.
A
A
So
if
you
look
at
lab,
there
are
lots
of
snapshot
that
tell
you
how
to
how
to
set
it
up,
and
you
end
up
having
to
write
a
new
get
lab
CI.
You
know
file
which
does
this.
You
know.
Oh
yeah.
Okay,
there
are
two
functions
so
first
you
deploy
the
function.
I
just
want
to
show
you
the
Kenny
Coby
out
here,
so
that
step
actually
does
the
Kanaka
build
alright.
So
you
have
an
image.
A
A
What
is
it
going
to
build?
It's
going
to
build
this
here
lab
at
dockerfile,
so
there
is
actually
a
docker
file
in
the
in
the
project
and
that's
a
little
bit
controversial,
because
some
people
say
if
you're
truly
server
lies.
You
shouldn't
have
to
expose.
You
know
docker
file
to
users
right
this
doctor
file
is
pretty
simple
from
golang.
You
know
it's
going
to
copy
the
files
you
find
an
entry
point
and
so
on.
So
when
you
run
the
pipeline,
it's
going
to
do
run.
A
This
can
occur,
build
and
docker
eyes
this
app
and
store
it
in
the
github
container
registry.
Alright,
the
other
function
will
still
be
running
and
then,
when
you
execute
the
pipeline,
you
know
they
will
actually
talk
to
each
other.
Alright,
so
cooling
was
telling
me
will
leave
the
instances
up
so
that
you
can
actually
we're
going
to
fix.
What's
what's
broken
here,
so
that
you
can
still
do
the
lab.
You
know
until
Thursday
or
Friday.
A
The
lab
3
I'm,
going
to
try
to
I'm
gonna,
try
to
do
it
for
you,
so
here
the
hope
was
that
you
will
be
able
to
deploy
functions.
You
know
transparently,
okay,
but
you
can
actually
do
this
on
your
own
with
your
own
kubernetes
cluster
right,
so
that
one
I'm
going
to
try
to
do
it
from
scratch.
To
show
you
what
you
know
what
we
mean
and
that
one
may
actually
work.
If
we
get
I'm
going
to
go
to
you
know
the
hosted,
get
lab
directly
and
I'm
going
to
create
a
new
project.
A
So
I'm
creating
a
new
project
on
the
on
the
hosted,
get
lab.
I
have
my
new
project
and
let's
say
this
is
where
I'm
going
to
host
my
functions
now.
I
want
to
be
able
to
deploy
those
functions,
so
I
need
to
have
an
endpoint
that
I
can
I
can
target
right,
so
I'm
going
to
go
into
here,
operations
server
less!
So
that's
the
integration
that
we
did
with
and
it
tells
me
hey.
A
You
need
to
install
Kay
native,
a
quick
and,
of
course,
before
I
can
install
Kay
native
I
need
to
have
a
QA,
miss
coaster,
so
I
add
a
QA.
This
question:
this
is
the
greatest
integration
in
get
lab
and
I
can
add.
A
cluster
using
gke,
google,
pal
or
I
can
bring
my
own
by
just
putting
in
the
credentials
and
the
URL
of
my
communities
cluster,
so
I'm
just
going
to
do
gke
here,
sign-in
with
Google
I'm,
going
to
call
it
chicken
demo
demo
to
one
of
my
projects.
A
A
A
Yeah,
so
here
your:
if
you
do
this,
you're,
definitely
paying
for
the
VM
instances
that
your
teammates
is
cluster
is
going
to
use
right.
You
won't
pay
for
the
functions
because
you
are
going
to
deploy,
functions
and
manage
those
functions
right.
It
is
server
less
in
the
sense
that
you
have
the
operators
of
that
cluster.
You
know
internally
at
your
company.
That's
that's!
What
they're
going
to
do
they're
going
to
create
the
foster
they're
going
to
manage
the
kinetic,
but
the
developers
they
are
not
going
to
see
this.
A
A
Okay
communities
cluster
now
is,
has
been
successfully
created
to
be
able
to
install
K
native.
If
you're
familiar
with
kubernetes
and
helm,
then
you
need
to
install
helm
as
a
package
manager
and
that's
what
I
mentioned
at
the
beginning
right
is
that
it's
complicated
you
need
to
install
it
faster.
You
need
to
install
an
app
package
er,
you
know
it
gets.
It
gets
complicated.
That's
where
you
know
service
is
supposed
to
shine.
Is
that
technically
you
don't?
A
You
know
you
don't
want
to
have
to
be
able
to
do
this
or
your
users
shouldn't
not
should
not
have
to
do.
You
know
all
of
this,
so
here
we
installed
chiller
helm
and
then
we
come
down
here
and
now
we
install
K
native
alright,
so
I'm
going
to
say
what
cubic
on
the
trigger
mesh
that
IO
I
give
it
a
DNS
domain
that
I
want
to
use
to
be
able
to
expose
my
function
and
I
install
K
native.
D
A
A
Okay,
so
let's
give
it:
let's
give
it
a
bit
of
time
while
it
does
the
beam,
so
we
go
back
to
it
and
the
fourth
plant
that
you
can
still
do
you
know,
will
fix
the
the
ingress
in
who
is
inventing.
So
we've
talked
about
deploying
functions,
maybe
having
to
do
a
build,
which
you
can
do
on
lab,
CI
and
then
deploy.
You
know
more
service
container.
You
can
install
everything
from
scratch.
A
If
you're
a
good
lab
user,
you
can
go
to
github.com
if
you
are
using
gke
to
do
what
I'm
doing
now,
which
is
the
full
install
from
scratch
and
then
the
fourth
lab
is
actually
to
start
dealing
with
events.
Okay,
that's
actually
the
very
interesting
part
deploying
a
function
itself.
You
know
it's!
It's
not
super
interesting,
but
starting
to
link
events
and
services
together.
A
That's
the
the
exciting
thing
so
lab
4
is
all
about
eventing
and
in
even
thing
the
ideas
that
you're
going
to
consume
event
from
whatever
sources,
and
then
map
I
need
an
event
to
in
our
consumer,
and
you
can
do
this
in
K
native
with
some.
You
know
new
abstractions
that
they've
developed,
channel
and
subscription
are
some
of
the
objects
that
you
have
and
now
there
is
also
broker
and
trigger.
A
So
these
are
API
objects,
just
like
you
have
incubated
deployment
services
and
so
on.
These
are
extensions
of
the
communities
API
that
you
can
use
in
Kennett
--iv
and
with
those
you're
going
to
be
able
to
say,
hey,
my
event
should
go
to
this
broker,
and
then
you
know
once
they're
on
this
broker
I'm
going
to
be
able
to
trigger
this
function
using
a
particular
event
type
and
things
like
this
right,
so
you
can
filter
your
events
and
you
can
send
them
to
you
know
specific
endpoints,
so
that
gives
you
this.
A
You
know
event
to
function.
Mapping
all
right
same
thing,
then,
with
serving
what
you
would
see
in
your
cluster.
If
you
ever
install
it,
you
know
from
scratch
and
manage
that
cluster
yourself,
you'll
see
a
candidate
event
in
time.
Space
you'll
see
a
controller
in
that
namespace
you'll
see
things
like
channel
controllers,
which
are
you
know
under
the
hood,
how
messages
are
being
managed
or
handled.
A
So
you
see
this
this
event
in
control
and
then
you
have
to
define
event
sources
so
typical.
You
know
with
the
traditional
type
of
event
source,
it's
just
a
cron
job
right,
so
a
cron
job,
you
define
it
and
you
regularly
meet
em,
it's
something
all
right.
So
here
you
have
a
cron
job
source
and
you
see
that
on
schedule
it
emits
hello
world
and
in
that
object
you
see
that
there
is
a
sink
being
defined,
so
that
particular
way
of
defining
the
the
sink
is
making
kind
of
a
hard
binding
between
the
event
source.
A
A
Instead
of
pointing
to
a
service
as
a
sink,
you
can
just
send
this
to
a
channel
or
a
broker
all
right,
and
then
the
events
will
be
redirected
to
their
endpoint
function.
Endpoints
thanks
to
the
the
trigger
definitions
right,
so
you
can
do
this
in
lab,
4
and
I'll.
Show
you
I'll
show
you
a
little
bit
how
this
looks
like.
E
E
A
C
B
B
B
C
D
A
A
C
A
C
C
A
Okay,
so
that
was
you
know
another
thing
to
install,
so
if
we
look
just
at
the
eventing
I
just
want
to,
you
know,
show
you
at
least
an
example
of
the
the
gitlab
source
here,
so
we've
created
an
event
source
for
gitlab.
There
is
also
in
Kennedy
when
event
source
for
github.
There
is
an
event
or
AWS
sqs
we've
created
lots
of
other
event
sources
for
AWS
services,
but
here
you
see
a
manifest
that
basically
declares
hey.
A
Here,
you're
declaring
this
event
source
and
you're
doing
that
event
function.
Mapping
notice
that
there
is
a
need
of
a
secret
because
Kay
native
is
going
to
need
to
be
able
to
talk
to
the
github
API
so
that
it
can
register
Web
books
right.
So
in
some
ways
this
is
a
way
to
automate
the
creation
of
gitlab
Web
books.
A
A
There's
going
to
be
some
red,
but
don't
panic,
Wow?
Okay,
it's
ready!
So
here
you
see
the
URL
demo
cube
con
one
two,
four
four:
four:
that's
my
cutest
name
space
that
has
been
created
by
gitlab
in
creative
cluster
cube
country,
grammar
that
I.
Oh,
that's,
the
DNS
domain
that
are
registered
in
cloud
DNS
and
this
is
the
web
app
that's
being
served
by
this
container.
Gc
r
io
cloud
run
hello.
All
right!
A
A
A
A
A
D
A
One
so
I
refresh
this
one
if
you
that
scale
to
zero,
it's
gonna,
take
you
a
little
bit
of
time
to
come
back
up
and
see
it
spinning,
because
since
I
didn't
hit
it
when
it
scaled
down
to
zero
and
now
what
I'm
seeing
is
cold
start.
So
that's
the
code
start
of
restarting
the
container,
depending
on
the
infra
that
you
have
the
cold
start.
Is
you
know
bad
or
acceptable
or
whatever?
Here
it's
you
know,
did
loan
right,
measured,
cold,
start
of
K
native.
It
is
roughly
2.5
seconds.
A
I
mean
performance
test
which
you
know
is
not
really
great
right
now,
but
it's
going
to
get
it's
going
to
be
improved.
Eventing
allows
you
to
trigger
function.
So
once
you
have
everything
set
up,
then
you
can
start
actually
thinking
about
that
application.
And
then
you
know
we
declare
that
application
through
a
set
of
event,
sources
triggers
and
and
functions
right.
So
the
idea
definitely
is
that
can
it'll
give
you
this
portability
and
that
you
can
go
service
even
on
pram,
but
you
know
in
any
pub
using
Tomatoes
faster
and
then
the
point
entity.
A
You
know
clearly
not
great
didn't
work
as
expected
here.
We're
gonna
leave
the
instances
up
so
that,
if
some
of
you
are
highly
motivated,
you
can
still
try
to
do
the
labs.
They
should
work.
We've
gone
through
them
here
looks
like
you
know:
the
the
load
was
not
too
good
or
something
went
off
with
the
ingress,
but
please
you
know
try
to
try
to
do
it.
Reach
out
to
us.
You
know
Twitter,
whatever
LinkedIn,
and
if
we
can
help
you,
you
know,
go
through
them
and
and
test
everything.
We'd
love
to
answer
your
your
questions.
A
Sep
go
Sebastian,
Cuban
right
there
Chris
right
there
he's
the
one
who
actually
did
the
creative
integration
in
get
lab
which
actually
worked.
You
know,
that's
the
good
thing
is
that
you
see
I
demoed
it
from
get
lab
calm.
You
know.
So
this
is
life.
You
can
use
it.
It's
a
variable.
Any
get
lab
user
can
I
can
use
this
so
great
job
Chris
works
well
right.
So
thank
you
very
much.