►
From YouTube: speedrun_gitlab_runner_fargate_codebuild
Description
This video is an overview of the setup steps in the blog post "How to build containers with the AWS Fargate Custom Executor for GitLab Runner and AWS CodeBuild."
https://about.gitlab.com/blog/2020/07/31/aws-fargate-codebuild-build-containers-gitlab-runner/
A
Hey
everyone,
daniel
here
product
manager,
for
get
live,
runner
hey
so
for
today's
speed
run
we're
just
going
to
quickly
follow
up
on
the
speed
run
we
had
from
last
month.
If
you
recall,
we
talked
about
using
the
github
runner
fargate
custom
driver
on
aws
fargate
in
order
to
to
run
your
ci
jobs.
Obviously
it's
an
mvc
product
that
we
released
pretty
recently
this
year
and
it
was
meant
to
help
specific
issues.
So
today
I
want
to
kind
of
build
on
that.
A
Discussion
is
a
very
recent
blog
post
that
was
released
actually
on
friday
july
31st.
It
talks
about
how
to
build
containers
with
the
aws
file,
customer
executive
for
gitlab
runner
and
aws
on
cobo,
and
we
just
pop
up
with
that
that
doubles
and
that
blog
post
is
written
by
elliot
rushjin
and
ricardo
mendez
and
susha.
A
quick
shout
out
to
those
two
to
those
two
chaps.
A
We
did
some
great
great
work
with
within
this
blog
post
done
for
us,
and
so
what
I
wanted
to
cover
today,
really
fast
is
just
to
go
over
a
few
key
points.
A
the
first
thing
is
just
kind
of
a
brief
overview
of
a
few
of
the
options
specific
to
amazon
web
services
and
and
hosting
around
amazon
web
services.
A
Then
the
other
pattern
is
auto
scaling
the
runner
using
the
github
runner,
auto
scanning
pattern
that
uses
docker
machine.
So
in
this
example,
as
a
diagram
shows
here,
you're
now
hosting
a
git
camera
on
an
ec2
instance,
we
typically
call
this
in
this
scenario.
Typically
called
render
manager
and
then
you're
using
a
docker
machine
executor,
that's
configured
for
that
github
burner,
then
you're
able
to
spawn
additional
these
two
instances
with
github
runner
for
executing
the
ci
jobs.
So
that's
one
auto
scaling
pattern
on
ec2
and
specifically
the
duct
machine,
auto
scanning
pattern.
A
A
The
difference
between
this
pattern
and
the
one
above
it
where
we
are
auto
skilling
on
ec2
is
that
this
gitlab
run
is
now
configured
to
use
the
custom
executor
and
then
that
customer
executor
is
then
you
is
using
the
figure
to
use
the
new
gitlab
5k
driver,
and
so
some
folks
maybe
have
been
following
this
very
closely.
The
customer
executives
the
means
by
which
we
enable
us
to
extend
the
gitlab
entity
to
be
to
be
used
with
platforms
that
we
don't
support.
A
A
So
then,
in
this
panel
you
can
see
the
run
manager
on
aws
82
instance
configured
with
the
customer
executor,
here's
your
phone,
your
driver
and
then
it's
pawning,
your
task
on
on
aws
file,
gates,
200
jobs.
So
that's
just
kind
of
the
first
thing.
Instead
of
this
high
level
overview
just
wanted
to
recap
again
for
customers
that
are
looking
to
use
aws
to
host
their
owners,
multiple
various
options
that
we
support
today
to
try
to
get
that
that
problem
solved
right.
A
I
don't
necessarily
see
a
lot
of
value
and
moving
to
a
file
gate
solution,
yeah
spot
on
for
those
types
of
customers
or
users
that
are
pretty
comfortable
using
eks.
That's
a
partner
that
can
continue
down
continue
using
no
issues
there.
What
we've
heard
from
other
customers,
though,
and
and
this
is
what
kind
of
resistor
drove
our
our
need
to
develop?
A
Second
bullet
point
here
that
community
members
are
rightly
pointing
out,
so
the
current
run
on
families
has
a
couple
of
key
limitations,
and
this
this
is
kind
of
relates
back
to
the
purpose
of
the
blog
post
and
our
brief
brief
overview
today
I'll
try
to
keep
it
brief
and
and
on
one
of
those
limitations.
The
key
one
really
is
that
doctrine.
Doctors
has
a
known
limitation,
that's
probably
the
right
technology
to
use,
but
I'm
just
using
it
as
a
chord.
A
Is
not
supported
on
foggy
and
we
can
all
surmise
the
number
of
reasons
why
that
is
one
community
member
was
able
to
get
pretty
close
with
using
kaneko
and
the
five
way
driver,
but
that
did
not
succeed.
So
you
know
a
key
cic
use
case
for
most
folks
in
a
very
well
versus
ci,
cd
and
devops.
It's
really
this
documentary
use
case
right.
You
want
to
build
a
container
as
part
of
the
part
of
the
pipeline
by
the
process,
your
deployment
process.
A
A
So
everything
was
completely
serverless
and
then
having
that
fargate
task
for
the
runner
manager
be
initiated
using
web
hooks
and
and
without
the
function
so
really
truly
kind
of
that
first
floor
into
a
100,
serverless
ci
cd
build
solution
using
aws
constructs
and
one
community
member
rightly
asked
the
question:
hey.
Why?
Why
are
you
guys
spending
all
this
time
investing
into
something?
That's
somewhat
bespoke
I'm
on
aws
today
and
why
does
not
use
native
aws
tooling
to
get
the
same
sort
of
functionality,
things
like
gold,
build
gold
pipeline
and
so
on?
A
And
that's
an
interesting
point
I
mean-
and
I
think
generally
speaking,
if
you're
a
customer
using
your
own
aws,
that's
right
approach
to
use-
and
you
know
you
can
use
aws
for
your
entire
devops
for
your
for
a
lot
of
your
use,
cases
related
to
devops
and
there's
a
lot
of
tooling
right
now
that
aws
is
ruling
out
that's
going
to
be
helpful
in
those
patterns.
But
if
you
like,
most
of
all
github
users
and
customers,
you
probably
have
different
patterns
of
support.
You
have
on-premise
deployments.
A
You
probably
are
supporting
multiple
clouds
you're
looking
for
a
different
solution
in
the
entire
devops
life
cycle,
and
so,
if
you're
needing
that
flexibility,
which
of
course,
we
all
know
that
that
provides
then
you're,
probably
not
going
to
be
locked
in
and
looking
to
be
locked
into
all
of
the
aws
crosstalk
and
you're.
Looking
for
flexibility,
especially
when
it
comes
to
hosting
your
cnc,
build
servers
or
agents
on
aws,
and
so
that's
gonna.
A
So
that's
kind
of
the
reason
why
we
have
spent
a
lot
of
time
trying
to
complete
solutions
to
make
it
as
easy
as
possible
for
end
customers
and
users
to
host
their
owners
in
a
cost.
Effective,
scalable
easy
to
manage
very
frictionless
through
an
aws
and
we'll
continue
down
that
path
in
the
future.
We're
gonna
be
investing
heavily
in
and
making
the
maintenance.
A
Again,
that's
the
brand
new
blog
post
that
we
just
came
up
with
on
july
31st
just
last
friday.
Then
there
are
some
other
blog
posts
that
are
already
out
there
on
the
gitlab.com
site.
There's
how
to
create
kubernetes,
cluster
amazon,
eks
and
github
is
a
great
poster
very
easy
to
follow
very
easy
to
set
up
and
get
going.
There's
also
a
blog
post
by
community
member
best
practices
to
keep
your
communities
runners
moving.
A
A
Let
me
close
that,
and
so
this
is
the
brilliant
blog
post
again
was
written
by
elliot,
rushton,
ricardo,
mendez
of
cint,
and
it's
how
to
build
containers
with
the
aws
5k
classroom
executor.
So
for
today's
speed,
when
I'm
not
going
to
go
through
all
the
steps,
when
you
first
read
the
blog
post,
there
are
a
few
steps.
There
are
a
couple
of
steps
in
here
to
follow
at
first.
A
You
see
it's
pretty
involved,
but
you
realize
that
it's
pretty
straightforward,
I'm
just
going
to
touch
on
those
that
may
require
a
quick
highlight
just
that
need
a
bit
of
highlighting
so
in
terms
of
overall
architecture.
Again,
this
builds
on
the
pattern
that
I
just
showed
earlier
and
this
pattern
we're
keeping
the
construct
pretty
straightforward
and
pretty
similar.
So
it's
not
overwhelming.
A
A
Then
we
have
our
build
containers
ready
to
go
as
task
if
you
will
on
aws
funding.
So
what
happens
here
is
when
a
job
comes
in
and
the
run
on
the
ec2
instance
will
say:
okay,
great
I've
got
a
job.
The
family
driver
will
initiate
the
creation
of
these
additional
tasks
or
containers
just
for
the
life
of
that
job.
A
But
what
we
have
done
here,
what
the
folks
have
done
here,
is
you
get
this
additional
construct
now?
Where
there
is
this
ci
coordinate,
02,
that's
the
purpose
of
this
blog
post.
That's
also
enabling
you
now
to
take
whatever
you
need
in
terms
of
your
your
docker
construct
and
build
your
docker
file
and
push
it
to
aws
code,
though
so
you
can
get
sort
of
a
document.functionality
using
aws,
vocal
with
local
shots.
A
A
You
would
have
installed
the
firmware
driver
and
you
want
to
make
sure
that
you
have
in
terms
of
permissions,
your
the
right,
aws
immune
submissions
to
allow
you
to
create
and
configure
s3
build
resources,
and
that's
those
steps
here
in
step,
one
which
are
pretty
straightforward,
but
for
for
most
folks
that
are
familiar
with
aws
and
the
steps
are
as
later
here
enable
you
to
do
this
manually
again.
You
can,
of
course,
automate
all
these
steps,
so
there's
nothing
much
to
call
out
here.
You're.
A
Basically,
creating
an
s3
bucket
and
you're,
giving
it
a
name
in
step.
Two,
you
simply
create
an
aws
code,
build
project,
nothing
too
much.
Okay,
here
the
the
only
thing
it's
not
typically,
when
you're
doing
a
creating
a
business
called
the
project
in
aws
they're,
just
typically
a
few
more
steps.
I
can
just
pop
up
with
that
console,
so
you
can
take
a
look.
A
You
create
new
projects
and
you
can
see
there's
like
things
like
the
project
name
descriptions,
some
other
fields
that
you
typically
have
to
fill
out,
and
so
that's
why
popping
back
over
to
the
blog
post,
there
are
a
few
more
steps
in
the
blog
post
and
again,
it's
very
straightforward,
you're,
basically
just
following
everything
here
and
putting
out
the
right
configuration
settings
to
get
your
code
built
projects
set
up
in
order
to
complete
the
the
tutorial
in
the
blog
post
step,
three
is
building
a
ci
container
image.
What
does
this
mean?
A
So
just
to
recap,
I'm
going
to
pop
back
over
here
just
to
the
diagram,
and
so
the
ci
container
image
is
what's
going
to
be
used
to
create
the
task
on
fungi.
The
thing
that
actually
builds
your
jobs
right
now
executes
your
jobs
on
file
gate.
That's
one
thing
with
fungi
and
with
the
foggy
construct,
there's
a
bit
more
involved
up
front
in
that
you
have
to
create
your
cip.
A
You
have
to
complete
this,
yet
you
have
to
create,
add,
build
containers
and,
at
the
moment,
there's
just
no
way
of
getting
around
that,
because
we've
got
to
do
some
things
by
in
terms
of
injecting
ssh
keys
and
doing
some
things
with
the
runner.
So
you
have
to
build
that
container
in
order
to
use
file
gate
in
this
construct,
and
so
that's
what
this
is
calling
out
here
and
step.
Three
is
building
the
the
cn
container
image
and
it's
basically
giving
you
an
example
of
what
you
need
to
add
to
your
dot
profile.
A
In
order
to
do
so,
so
you
build
your
docker
image
and
you
add
this
construct
and
you're
good
to
go
what
I've
done
and
when
I'm
doing
my
testing
and
folks
can
have
different
ways
of
addressing
this.
In
my
testing
I
have
a
bit
lab
repo
here,
I'm
just
popping
over,
so
you
can
take
a
quick
look
and
this
given
rico
repo.
You
can
see
that
I
have
a
docker
file.
It's
got
some
stuff
in
there
and
then
for
the
purpose
of
this
demo.
A
A
I
actually
use
gitlab.
In
some
cases.
I
push
the
the
image
direct.
I
use
gitlab's
repository
in
some
cases.
I
push
you
to
aws,
but
you
can
see.
Then
I
use
gitlab
to
go
ahead
and
limit
my
image
build.
It
have
the
stuff
that
I
need
in
this
case
I'm
actually
pushing
it
directly
to
ecs
or
ecr.
In
this
case
sorry,
so
it's
just
pretty
straightforward,
so
you
create
a
repo
for
your
docker
images
that
you're
going
to
be
using
for
your
cn
build
containers.
A
That's
really
the
easiest
way
to
do
it
to
have
these
stats
version
controller
as
well
again,
following
the
same,
I
guess
infrastructure
is
called
paradigm,
so
everything
is
in
tow,
I'm
so
pretty
straightforward.
I'm
getting
a
lot
of
flexibility
and
I
can
come
in
here.
I
can
change
the
branch
if
I
want
to
push
it
to
a
different
repo.
I
can,
if
I
want
to
add
something
to
the
darker
image
I
can
so
even
though
at
first
you
might
be
thinking
well,
I've
got
some
maintaining
stuff.
A
A
Four,
you
add
a
cold
codeball
configuration
to
the
repo
and
specifically,
in
this
case-
and
I
I
would
be
remiss
by
not
calling
out
here
that
later
on
in
the
in
the
blog
post,
there's
actually
a
link
to
an
example
repost
you
can
just
link
it
in
an
example.
Example,
repo
and
pull
in
the
code,
but
what
it's
talking
about
here
is
that
in
the
project
that
you're
using
for
testing
on
this
demo,
this
is
what
you
would
want
to
add.
This
ghost
color
gamma
file
is
what
aws
code
will
be
looking
for.
A
Yeah
and
you
can
see-
and
I
copy
this
directly
from
the
refund,
so
it's
very
straightforward,
straightforward
file,
so
you're
creating
a
repo.
You
can
focus
example
that's
later
on
in
the
blog
post,
whatever
makes
sense
for
you,
but
that's
basically
what
that's
going
on
yeah
and
then
finally,
you
set
up
your
ci
you'll,
get
website
job
right.
Everything
there's
really
nothing
much
to
call
out
here.
A
The
one
thing
is
that
this
at
first
looks
pretty
gnarly,
but
again
all
of
the
example
scripts
are
included
in
the
in
the
link
repo
below.
So
it's
pretty
straightforward,
so
you
don't
have
to
change
any
of
this
code
to
get
going.
You
just,
as
I
said
mention
earlier.
You
just
swap
the
repo
set
up
your
gitlab
yama
file
and
you
should
be
good
to
go
and
then
you
add
a
job
to
the
resulting
container.
A
So
once
everything
is
set
up,
you
have
all
the
steps
to
complete
the
source
repository
as
the
blog
post
is
mentioned.
When
you
structure
just
follow,
you'll
have,
as
you
would
expect,
to
get
that
memo
file
the
build
spec
that
yama
the
code
build.sh
file,
which
is
referenced
here
again,
you're,
not
greenish
yourself,
you
don't
have
to
change
it
to
get
going,
just
copy
and
paste
this
and
then
your
docker
file
and
then
the
final
step
is,
of
course
you
add
a
job
to
the
github
game,
and
this
is
the
job
you
can
see.
A
A
Then
it's
going
to
use
the
aws
cli
to
configure
default
region
and
copy
the
resulting
zip
file
over
to
cobo
so
by
and
then
finding
droppers.
It
talks
about
how
you
actually
set
up
your
aws
credentials,
I'm
not
going
to
cover
that
that
so
much,
that's
pretty
straightforward!
So
now,
step
seven
is
it's
show
time
so,
but
once
you
finish
steps
one
through
seven
you're
ready
to
test
this
this
pattern
and
ready
to
test
this
configuration
so
just
to
highlight
just
to
recap
right.
A
A
What
you'll
notice,
if
you've
set
this
up
properly.
I
I
don't
have
a
callback
project
to
the
plugin,
but
I
do
still
have
my
s3
bucket
so
I'll
pop
back
over
here,
I'm
going
to
s3
bucket.
Let's
skip
that
download
build.
In
my
case,
I
actually
came
out
actually
that
one
jiggle
from
understood-
and
so
you
can
see
the
builder
zip
file-
was
uploaded
as
part
of
the
end-to-end
process.
So
so,
just
to
recap,
today's
code
bill
and
kind
of
the
purpose
of
it
is
to
kind
of
summarize
here
one
more
time.
A
So
this
this
blog
post
again
just
came
out
john
31st.
It's
talking
about
how
to
build
containers
with
the
aws
private
customer
executive.
You
get
web
runner
and
aws
cobra,
and
the
whole
purpose,
as
the
intro
to
the
and
the
blog
post
mentions,
is
to
try
to
address
that
limitation
or
on
front
gate
of
not
supporting
the
outcome
doctor,
and
so
we
hope
that
this
is
somewhat
helpful
and
that
the
community
members
can
take
this
time
forward.
Again.
We
recognize
that
this
is
there's.
A
So,
but
for
right
now
we
just
wanted
to
cover
this
and
that's
the
last
thing
to
wrap
up
today's
speed
runner.
I
know
I
wasn't
so
happy
to
demo
today.
I
just
want
to
again
call
out
some
of
the
other
blog
posts,
so
this
is
a
blog
post
that
was
on
the
31st,
the
one
I
called
up
earlier,
how
to
create
a
kubernetes
classroom,
adapt
amazon,
eks,
a
brilliant
blog
post.
A
A
I
can't
remember
their
favorite
one,
but
this
is
really
solid
for
folks
that
actually
there
is
there's
really
some
other
folks
that
are
getting
have
gotten
a
little
bit
more
advanced
in
exploring
kubernetes
runners
kubernetes
on
aks,
and
then
this
was
just
the
universal
entry-level
blog
post
that
when
we
first
released
the
aws
5g
driver
just
kind
of
announcing
that
and
telling
you
how
to
get
things
set
up.
So
just
in
closing
into
recap.
A
So
again,
if
you're,
a
user
customer
with
github
and
you're,
trying
to
figure
out
a
way
to
host
your
runners
on
aws,
we
have
a
number
of
different
constructs
and
that
we
support
today
and
we're
building
additional
capabilities
into
some
of
those
constructs.
We
know
we
have
to
do
when
it
comes
to
getting
you
to
a
long-term
100
silver
solution,
but
we
think
we
have
some
great
building
blocks
in
order
to
do
so.
So
to
recap:
we've
got
support
for
ec2
very
basic
pattern.
We
have
support
for
eks.
A
We
have
support
for
all
skating
using
our
docker
machine
pattern,
and
we
also
have
support
for
auto
skating
using
fungi
on
elastic
lithium
services.
As
the
newest
panel
in
the
nbc
panda
was
just
released
next
time
around
we'll
try
to
find
something
very
interesting
for
speedruns
until
until
next
time
talk
to
you
soon
cheers.