►
From YouTube: Infrastructure as Code: Manage your Architecture with Git, Danilo Poccia - Git Merge 2015
Description
With the AWS Cloud you have an on-demand, programmable infrastructure that you can manage using tools and practices from software development. You can create resources when you need and dispose of them when you don’t. Using Amazon CloudFormation you can describe your architecture in text files. To change your infrastructure, you edit those files. Having application and infrastructure code together in a single, robust, versioned repository like Git gives a lot of advantages. Using AWS Elastic Beanstalk you can link your Git branches to different infrastructure environments (e.g. test, production) and automate deployments. You can create test environments on-demand, even for a short time. Instead of continuously update your resources, you can recreate them quickly from scratch, simplifying lifecycle management and making deployments immutable. As a result, you have more time to focus on the unique features of your application.
B
Thank
you,
hi
everybody.
So,
for
the
last
presentation
of
the
day,
I
was
thinking
to
take
a
different
perspective
and
the
use
of
get
so
to
see
how
you
can
use
get
not
to
manage
only
the
source
code
of
your
application,
or
maybe
your
documents,
but
also
how
to
manage
your
architecture,
your
infrastructure
with
gates,
and
why
you
would
do
that
why
you
would
use
a
tool
like
get
to
manage
your
own
infrastructure?
B
Well,
it
oops
I
mean
in
the
past.
We
were
used
to
design
very
simple
architecture,
so
there
was
a
presentation
here,
some
business
logic
and
then
a
persistence
tier.
So
probably
this
architecture
is
not
difficult
to
manage,
but
what
happened
is
that
the
internet
is
scaling
and
growing
so
much
that
the
modern
architecture
they
really
don't
fit
into
a
multi
tier
diagram.
B
So
this
slide
from
Adrian
Cocker
off
to
the
former
cloud
architect
of
Netflix
shows
us
the
number
of
models
and
the
interaction
between
the
modules
that
compose
a
large-scale
scale
system
is
really
different
from
what
we
were
used
to
perceive,
and
so
we
cannot
use
the
normal
tools
to
manage
this
kind
of
architecture.
We
have
dependencies,
we
have
a
lot
of
modules.
Doesn't
it
not
resemble
the
source
code
of
an
application?
B
So
we
are
really
working
on
a
different
scale,
so
managing
the
infrastructure
as
code
can
really
help
it
can
help,
because
we
are
talking
about
writing
code
to
manage
your
configurations
and
to
automate
the
provisioning
of
your
infrastructure.
The
important
thing,
though,
is
writing
code
not
to
do
it
that
rural
web
interface,
because
when
you
write
code,
then
you
food
can
reuse
all
the
tools
and
the
best
practice
of
a
software
development
and
software
architectures.
B
So
when
you
have
your
infrastructure,
describe
it
in
a
in
some
text
files
like
you,
can
do
with
tools
like
puppet
chef
of
cfengine.
Then
you
can
put
the
source
code
of
your
infrastructure
under
version
control,
so
that
if
someone
is
changing
something
you
do
the
commit
on
the
code.
Send
the
Delta
two
to
your
infrastructure
platform
to
deploy
the
changes
as
required,
and
then
you
can
roll
back
to
a
specific
version
of
your
infrastructure.
B
In
the
same
way,
you
can
roll
back
to
a
specific
version
of
your
software
application,
because
you
know
that
that
version
was
still
working
and
it's
helpful
to
go
back
in
time.
Some
time
and
also
you
can
use
this
for
testing,
because
if
you
have
all
your
architecture
described
in
a
text
file,
you
can
use
a
disposable
architecture
like
cloud
architecture
where
you
can
create
resource,
use
them
and
destroy
them.
B
When
you
don't
need
any
anymore,
and
so
you
can
create
a
test
environment
on
the
man,
because
you
have
all
the
information
that
you
need
to
create
this
distant
vironment,
use
it
and
dispose
of
it
after
the
test
is
wrong
and
also
you
can
use
other
best
practice
from
software
development.
Like
small
deployments,
you
know
it's
much
easier
to
bring
in
production
a
lot
of
small
deployments
that
the
big
changes
that
probably
break
some
behavior
and
the
same
it
can
apply
to
to
your
infrastructure.
B
So
if
you
need
to
change
your
infrastructure,
do
this
in
step
and
bring
every
single
step
in
production
so
that
you
can
test
every
single
step.
It
is
affecting
some
other
part
of
your
of
your
overall
system,
and
so
in
general,
you
can
apply
all
the
design
partner
of
software
development
to
the
management
of
your
infrastructure,
and
here
I'm,
particularly
thinking
about
continuous
integration
and
continuous
deployment.
B
So
the
fact
that
you
can
continuously
adapt
and
adjust
your
infrastructure,
depending
on
the
changes
that
you
do
on
your
on
your
infrastructure
code,
and
you
can
also
have
multiple
stages
that
will
become
multiple
environment
of
your
s
or
your
infrastructure,
so
that
you
can
around
your
application
to
multiple
stage
in
a
pipeline
that
will
bring
your
application,
hopefully
in
production.
So
this
is
something
that
we
usually
have
with
a
continuous
integration
system.
B
So
you
have
your
application
code,
your
git
repository,
and
when
you
commit
your
changes,
you
in
some
way
talk
with
the
continuous
integration
system
that
will
deploy
your
application
owner
on
an
infrastructure
that
is
already
there.
But
if
you
have
also
your
infrastructure
code,
so
the
hope
describing
your
infrastructure
in
the
same
repository,
you
can
version
them
together,
so
that
version
1
of
your
application
option
is
using
this
infrastructure
version.
B
2
is
using
this
older
infrastructure,
and
the
continuous
integration
system
can
build
the
overall
environment
in
our
cloud
infrastructure
and
deploy
the
application
inside
this
to
test
it
or
to
use
it
in
production,
and
we
will
see
that
with
AWS.
There
are
two
services
that
allow
you
to
do
that
very
easily.
One
is
elastic
Beanstalk
that
is
very
easy
to
use,
especially
for
developers,
and
one
is
cloud
formation
that
is
more
powerful
because
it
can
help
you
to
describe
the
overall
architecture
in
a
group
of
text
files.
B
So,
let's
start
with
elastic
beanstalk
so
as
min-seok
is
very
easy
to
use
and
allows
you
to
deploy
a
web
application
or
a
web
service,
and
you
just
need
to
give
to
elastic
pistol.
Your
code
can
be
java.net
and
all
the
PHP
Python
Ruby
go
or
can
be
one
or
more
docker
containers,
and
this
package
of
the
application
is
given
to
be
in
stock.
That
will
create
a
whole
environment
and
auto
scalable
environment
or
a
single
host
environment
for
a
test
purpose.
B
It
is
dedicated
to
you,
it
will
do
log
collection
in
will
version
your
application
and
store
the
log
and
the
version
of
your
application
on
a
durable
storage
like
s3,
so
that
you
can
really
don't
care
about
how
to
implement
the
infrastructure.
But
miss
talk
is
a
little
bit
more
powerful
than
that.
When
you
create
an
application
with
Beanstalk,
you
can
have
multiple
application
versions
and
multiple
environment,
like
think
of
test
environment
production,
environment,
and
you
can
tell
Beanstalk
to
deploy
a
specific
version
of
your
application
into
a
specific
environment,
and
everything
is
done
automatically.
B
So
elastic
missile
can
integrate
with
get
so
that
you
can
map
a
git
branch
to
a
specific
environment.
So
the
test
branch
goes
to
the
test
environment.
The
master
branch
goes
to
the
production
environment,
for
example,
and
you
can
also
map
a
git
commit
to
a
specific
application
version
so
that
when
you
commit
something
that
you
want
to
deploy,
you
can
automatically
create
a
software
version
for
elastic
Beanstalk.
So
we
have
a
situation
like
this,
where
your
application
infrastructure
code
in
it,
you
can
have
multiple
environments.
B
I
did
just
a
couple
here
for
in
this
example
like
test
environment
and
production
environment,
and
then
you
can
link
the
developed
branch
to
the
test
environment
and
the
master
branch
to
the
production
environment.
Then
you
can
use
git,
commit
to
commit
your
code
and
it
be
deploy
to
deploy
this
committed
version
to
the
specific
environment
that
is
bound
to
the
branch
you
are
using
in
this
moment.
Eb
is
the
command-line
interface
for
elastic
Beanstalk.
It's
a
Python
package
that
you
can
install
on
any
system
with
Pippin
stalls,
so
it's
very
easy
to
to
install.
B
So
here
I
have
a
single
file,
and
this
is
not
version
at
right
now,
so
I
can
run
giving
it
get
to
create
and
initialize
the
git
repository,
and
this
file
is
a
very
basic.
It's
a
very
basic
web
server.
So
it's
just
answering
hello
world
any
kind
of
question
that
you
get
so
now
what
I
can
do
is
I
can
commit
this
version.
B
So
I
can
add
this
file
and
commit
as
the
first
release
and
now
I
can
initialize
with
EB,
and
you
see
that
the
syntax
is
very
similar
to
the
get
syntax
this
environment.
This
will
connect
to
my
account,
we'll
see
the
available
application
we
see.
There
is
already
on
one
application:
I
can
create
a
new
one.
I
use
the
application.
B
It's
already
there,
and
now
it's
asking
me
if
I
want
to
bind
this
branch
now,
of
course,
after
the
first
coma
time
in
the
master
branch
to
one
environment
and
I
want
to
link
it
to
the
production
environment.
Now,
if
I
do
git
git
branch
I
can
see
that
there
is
the
master.
If
I
go
to
B
list
I
see
that
there
are
two
environments
and
the
asterisk.
B
Let
me
which
environment
I'm
using
now
so
I
can
run
a
be
deploy,
and
this
will
deploy
this
version
so
that
it
will
be
packaged
and
sent
to
classic
Beanstalk
and,
as
you
can
see
now,
the
common
is
done.
I
can
interrupt.
This
wind
off
is
not
important
anymore,
because
everything
is
already
running
here
in
the
cloud
environment.
The
environment
is
updating
here,
I
have
all
the
events,
and
if
the
gods
of
the
Dame
are
not
looking
at
me
right
now,
the
environment
will
become
green
and
I
can
see
the
hello
world
here.
B
Maybe
again-
and
now
you
also
know
why
I'm
not
that
front-end
developer,
because
this
is
the
maximum.
So
what
I
can
do
now?
I
can
use
git
to
create
a
new
branch,
so
I
check
out
a
new
branch
that
we
can
call
test
in
this
case
and
if
I
do
a
belief
now,
I
see
that
I'm
not
bound
to
any
to
any
environment
and
in
no
more
so
I
can
edit
this
file
and
do
all
the
changes
that
I
need
to
do
to
run
into
a
test
environment.
B
B
B
You
check
out
sorry
and
I
can
also
create
a
new
environment
from
here
with
a
be
create,
and
of
course,
I
will
be
asked
all
the
usual
question
to
create
environment,
the
URL
to
use
the
DNS
name.
This
is
probably
taken,
I
can
call
it
hello
or
the
git
merge
and
so
on
now
the
event
is
creating
and
then
I
can
deploy
in
the
new
environment.
This
is
the
last
deployment
and,
of
course,
the
output
is
different,
so
the
magic
happens
so
going
back
to
to
the
deck.
B
This
is
the
first
way
to
use
the
infrastructure
of
code.
It's
very
easy,
especially
for
a
developer,
and
if
you
want
more
information,
that
is
this
presentation
by
coca-cola.
That
tells
you
how
they
are
using
elastic,
Beanstalk
store
the
deployment
or
their
web
infrastructure,
but
they
also
use
confirmation.
That
is
another
tool
that
allows
you
to
manage
a
collection
of
resources
in
a
single
text
file
or
in
a
collection
of
text
files.
B
B
So
the
syntax
of
this
text
files
is
JSON
and
there
are
four
main
syntax,
the
four
main
parts,
the
parameters
that
allows
you
to
reuse,
the
same
template
in
different
occurrences
like
you
can
change
the
password
of
your
database
or
the
number
or
the
size
of
the
width
or
machine
that
you
want
to
use
the
mappings
that
allows
you
to
use
this
template
in
different
regions.
We
have
regions
all
over
the
world,
then
you
may
have
different
codes
inside
these
regions.
The
resource.
B
This
is
the
declarative,
syntax
of
a
local
formation
that
will
ask
for
the
resource
to
be
deployed
and
the
output
that
can
be
the
HTTP
URL
of
your
website
after
you
have
been
created
by
CloudFormation.
So
this
is
an
example.
How
to
ask
for
a
relational
database.
I
want
my
sequel
engine
when
version
5.6
and
I
want
this
username
and
password.
Of
course,
as
I
said,
these
are
good
candidates
for
being
parameters
and
not
being
inside
the
sync,
the
file
of
the
template,
and
this
is
how
to
customize
a
virtual
machine
that
you
just
created.
B
You
use
the
package
manager
of
the
iStent
to
install
packages.
This
is
the
example
to
create
a
lamp
stack
and
then
download
this
source
bundle
from
s3
and
explode
it
under
the
the
part
where
Apache
will
will
look
for
the
source
code
of
your
application.
So
it's
fully
creating
and
deploying
a
lamp
stack
on
a
single
server.
B
So,
as
we
saw
the
template
and
the
parameters
will
create
an
implementation
that
we
call
a
stack
and
the
template
can
be
reusable
for
more
than
one
stack.
Then
you
can
map
a
different
stack,
for
example,
to
a
different
environment
like
we
did
before
so
a
test
or
a
production
environment,
and
the
parameters
are
usually
specific
for
a
single
stack
because
they
will
contain
maybe
the
password
of
your
database
or
they
will
contain
the
sides
and
the
environment
that
you
want
to
create.
B
So
the
parameters
are
good
candidates
to
be
stored
into
a
configuration
management,
and
let's
keep
this
in
mind
because
it
will
be
my
last
slide.
They
will
link
to
this.
We
also
have
a
tool
called
former
that
will
that
allows
you
to
look
at
the
resource
that
you
were
already
Ava
and
automatically
create
the
JSON
file
for
you
with
that
you
select,
so
you
don't
need
to
create
the
the
file
from
scratch.
B
In
this
case
for
the
very
short
demo,
we
will
use
this
simple
website.
You
have
an
elastic
load,
balancer
swallowed
balancer
is
balancing
a
group
of
ec2
instances
so
that
we're
talking
about
with
one
machine
with
a
very
basic
web
server.
You
can
imagine
from
the
previous
example
that
PI
skills
are
not
improved,
so
don't
expect
too
much
and
for
the
sake
of
simplicity,
we
will
have
a
single
web
server
into
this
group
and
we
will
use
text
editing
to
add
a
second
to
add
a
second
server.
B
So
let's
quickly
look
this
so
if
I
go
into
the.
So
let's
close
this,
this
is
the
CloudFormation
console
by
the
way
you
see
this
other
stack
here.
These
are
automatically
created
by
the
elastic
beanstalk,
because
the
elastic
beanstalk
is
using
hundred
the
hood
conformation
and
on
your
biast.
But
this
other
stock
here
called
infrastructure
is
a
stack
that
is
creating
the
resources
that
I
was
saying
so
the
elastic
load
balancer.
B
We
don't
see
here
the
elastic
load
balancer
the
security
group
that
is
the
firewall
protecting
each
of
these
instances,
the
configuration
for
launching
a
new
server
and
the
web
server
group
that
can
scale
depending
on
your
needs.
If
we
go
back
to
the
text
file,
we
can
detect
this.
So
this
is
my
infrastructure
description
file.
So
these
are
the
parameters
the
size
of
the
instance.
The
the
way
to
log
in
I
go
fast
here.
B
At
the
end,
we
have
the
web
server
group,
and
this
section
here
will
tell
the
minimum
size
the
maximum
size
of
this
auto
scaling
group
and
the
decided
capacity
since
I
didn't
put
any
automatic
scaling
policy.
This
table
scanning
group
is
static,
as
we
will
stay
with
one
instance.
So
let's
say
that
I
want
more
than
one
I
should
edit
this.
So
let's
do
getting
it.
B
B
B
Now
confirmation
is
receiving
the
new,
updated
file
is
made
looking
at
the
difference
and
it
will
run
the
up-to-date.
So
if
we
look
here,
the
update
is
in
progress,
and
here
we
get
all
the
events
that
are
happening
on
the
stack,
and
this
will
change
the
configuration
of
the
auto
scaling
group
and
in
a
few
minutes
the
distance
will
be
changed
from
1
to
2,
so
I'm
hoping
the
ec2
console
just
as
a
reminder,
so
that
we
can
have
a
look
at
it
after
so
you
see.
B
B
They
could
you
City
Rover
that
it
was
so.
The
NASA
is
using
confirmation
for
deploying
resources
that
allows
you
to
get
the
information
back
from
Mars
for
this
robot
and
if
you
want,
you
can
have
more
information
from
this
talk.
There
is
the
bbc's
or
the
british
broadcaster
that
tells
us
they
are
using
confirmation
for
management
and
deployment
at
scale.
So
in
the
past,
we
used
to
think
about
application
and
infrastructure
as
two
different
words:
two
different
teams
to
developers-
and
this
is
a
system
administrator
that
they
should
not
talk
together.
B
I
think
that
now
the
real
distinction
is
between
data
and
code,
and
there
is
data
in
the
application
and
in
the
infrastructure
and
code
in
both.
So
this
is
the
first
step
for
a
company
to
understand
and
to
embrace
the
develops
culture
in
my
perspective
and
what
I
think
is
still
missing
in
the
use
of
git,
and
this
is
a
question
that
I'm
making
to
you
is
how
to
manage
configuration
management.
B
So
you
have
configuration
management
for
the
application
and
the
infrastructure,
because
you
want
to
customize
passwords,
you
won't
customize
the
sides
of
your
infrastructure
depending
on
the
environment.
So
what
can
you
do
so?
The
best
idea
that
I
had
is
to
use
git
submodules,
so
you
use
a
different
repository
for
a
configuration
management.
This
also
is
good
for
security,
so
you
can
give
permission
to
access
this
configuration
management
repository
only
to
who
can
really
do
that
and
then
you
get
some
module.