►
From YouTube: Rust and Skaffold for Iterative Development on Kubernetes - Gastón Kleiman & Gilbert Song
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
Rust and Skaffold for Iterative Development on Kubernetes - Gastón Kleiman & Gilbert Song, Mesosphere
Iterative development requires to be able to quickly build/push/deploy. These tasks aren’t easy in a cloud-native environment, but automating them is worth the effort. Being able to see changes reflected as they are being done leads to a tremendous increase in developer velocity. In this talk we will show how automating building and pushing a service out to the a Kubernetes cluster remotely allows for quick iteration. We will develop a REST service from scratch using Rust. The service will be deployed to Kubernetes, and continuous deployment will be achieved by using Skaffold. This demonstration will be presented in steps, starting with a simple web server without persistence, then adding a DB to the Kubernetes deployment and using Diesel to connect with it. All the while Skaffold will be used to seamlessly watch our changes “show up” on the K8s cluster as we’re working.
To learn more click here: https://sched.co/FuLo
A
A
B
B
A
B
Today
we
will
be
talking
about
how
to
use
some
tools
to
do
interactive
deployment,
sorry
and
development
you
see
in
coordinators,
so
we
will
first
explain
what
iterative
development
is
all
about.
Then
talk
a
little
bit
about
CI,
CD
and
emitter
infrastructure,
some
of
the
challenges
at
this
close
half
or
this
middle
middle
metallurgist
half,
and
then
we
will
present
some
tools
that
we
will
use
in
on
a
demo
to
do
iterative
development
on
kubernetes,
mostly
a
trust
coordinators,
of
course,
and
scaffold.
B
B
B
But
you
don't
do
it
only
once
you
do
an
iteration
that
is
designed
development
and
testing
and
deployment,
and
then
you
do
an
hour,
iteration
and
then
another
iteration
and
on
each
iteration
you
add
more
value
to
your
product,
and
since
you
are
going
to
do
a
lot
of
iterations,
then
each
iteration
has
to
be
quick
and
cheap.
To
do
and
on
this
talk
we
will
show
you
how
to
do
that
quickly,
how
to
yeah
iterate
it
faster.
B
As
I
said,
if
you're
going
to
do
iterative
development,
then
you
totally
need
to
have
some
CI
CD
tooling.
You
need
to
automate
how
you
will
do
testing,
because
you
will
do
it
very
often,
and
you
also
need
to
automate
how
you
are
going
to
deploy
your
applications
and
containers
are
great
tools
to
be
able
to
do
it,
because
you
can
just
make
your
CI
CD
tools
create
a
container,
and
then
you
can
deploy
that
locally,
while
you're
testing
a
non
production
afterwards
without
having
to
change
anything.
B
B
Basically,
if
you're
studying
an
application-
and
you
have
very
few
users,
then
you
have
a
few
servers
and
you
love
your
servers
and
you
give
each
of
your
servers
a
very
cute
name
like
a
pet
name,
but
us,
and
if
one
of
those
servers
has
a
problem,
then
you
take
care
of
that
server.
It's
like
bringing
it
to
the
veterinary
and
you
try
to
fix
it
and
you
give
it
all
your
love,
but
that
doesn't
scale.
If
you
have
a
hundred
system
of
servers,
then
you
cannot
do
that.
B
You
have
to
start
trading
your
servers
like
cattle.
You
don't
give
them
any
cute
name
anymore.
You
basically
give
them
boring
names
like
sour
one,
two
three
BB
one,
two
three
and
stuff
like
that,
and
then
you
need
some
configuration
management
tools
in
order
to
be
able
to
maintain
those
servers,
and
that
also
scales
up
to
a
point.
But
it
gets
complicated
as
you
have
many
services
and
that's
where
UW
infrastructure
comes
into
play
and
where
you
can
define
your
infrastructure
in
code,
deploy
it
somewhere
and
then
have
some
tooling.
B
B
You
need
to
put
some
effort
there
at
the
beginning
and
if
you
have
giant
binaries
or
containers,
it
takes
a
long
time
to
do
the
building,
pushing
and
deploying
could
be
minutes
to
hours,
which
is
nice
and
it
still
doesn't
solve
the
problem
of
safety,
like
you
still
have
to
make
sure
that
your
applications
are
safe
both
when
we
were
talking
about
concurrency
and
also
memory
safety.
So
the
tools
that
we
will
introduce
today
help
you
deal
with
all
of
these
problems
and
yeah.
We'll
talk
about
them
a
little
bit.
A
A
You
might
have
like
a
lot
of
like
contributors
which
can
modify
the
same
codebase
and
you
want
the
test
to
be
quick
efficient
and
it
could
be
iterative
quick
enough
to
make
sure
like
Oh
I
mix
a
mistake
from
one
thing
and
it
should
post
enough
to
merge
to
the
master
branch
and
which
might
potentially
impact
the
other
team.
So
we
we
aim
for
I
improved
the
velocity
of
the
CI
CD,
with
VI
CIC,
deterring
for
your
own
development.
A
So,
right
now
it's
like
a
quick
introduction
about
like
rust,
so
we
used
R
Us
with
scaffold
on
kubernetes
and
it's
created
by
Mozilla,
which
please
like
safe
and
then
it
handle
like
concurrent
concurrency,
and
then
it
is
like
a
practical
language.
So
it
is
also
a
low-level
language
and
performance
is
also
similar
to
C
press
press
yeah.
This
is
some
base.
Information
we
use
rust
and
we
so
we
do
have
some
challenges
with
rust
here.
So
if
we
look
at
like
some
project,
we
people
talk
about
in
this
morning's
keynote.
A
A
So
the
kubernetes
people,
like
a
lot
of
part
about
kubernetes
kubernetes,
like
they
have
focused
on
continual
constriction,
the
Defiler
user
friendly
api,
and
they
have
a
lot
of
submissions
like
operator
tooling,
for
your
matrix
alerts,
servers
and
help
you
to
do
like
easy
upgrade
so
and
then
a
lot
of
like
ecosystem
around
kubernetes,
including
machine
learning
and
many
other
application
like
there's
a
service.
So
we
expect
to
have
like
kubernetes
to
manage
continuous
across
different
clusters,
and
you
could
expect
like
it's
supposed
to
be
like
highbrow
cross.
A
Like
you
hack,
you
can
launch
the
cluster
on
your
on-prem
machines,
physical
machines,
and
then
you
could
still
have
your
servers
on
public
rao.
This
enable
you
to
use
like
immutable
infrastructure
and
then
it
becoming
a
standard
to
defy
your
distributed
application,
no
matter
it's
like
foresee,
ICD
or
any
other
application,
so
yeah
and
the
solution
here
we
can
represent
in
in
the
demo.
Guess
don't
gonna
show
you
guys.
A
It
is
based
on
major
sphere
DCs
be
launched
like
kubernetes
service,
and
then
we
have
the
scaffold
running
on
top
of
kubernetes
for
the
IDs
see
ICD
papers
and
a
quick
introduction
like
in
from
these
years,
we
didn't
have
like
one
button
launch
of
multi
kubernetes
cluster,
so
from
the
UI.
You
can
just
like
one
button
to
launch
multiple
clusters
simultaneously
and
which
is
efficient
and
easy
for
users
to
use.
A
This
is
more
meaningful
for
company
which
have
the
expert
like
they
have
multiple
teens
share
the
same
cluster
and
they
expect
multi-tenancy,
and
it's
really
up
to
the
user
right.
They
want
like,
for
example,
they
want
the
front
end
to
use
one
single
kubernetes
cluster,
and
then
they
extract
they're,
like
CI
CDT
news
and
other
scalpel
on
kubernetes
as
a
separate
cluster.
So
you
can
all
do
all
those
things
together
with
TCS
yeah.
A
A
A
That's
the
goal-
and
here
is
a
quick
photo
about
like
the
base
structure
of
the
scalpel
I,
will
not
go
to
the
architecture
detail
here
and
basically
from
the
user
perspective,
because
we
only
care
about
the
user,
so
weak,
less
careful
to
build
image,
push
it
to
the
registry
and
then
get
the
deployment
down,
and
all
of
this
should
be
configured
by
one
single
file
and
easy
and
quick,
so
user.
We're
gonna
unify
all
the
process
for
user
user.
Don't
know
the
no
longer
necessary
to
handle
all
those
like
operator
responsibilities.
A
Here
is
a
quickly
simple
like
if
you
run
scalpel
with
on
kubernetes
locally,
we
don't
have
pushed
so
we
just
build
it
with
talk
a
team
and
build
the
image
and
then
scalpel
gonna
do
the
deployment
for
you.
Just
you
just
need
to
configure
a
quick
config
file.
So
if
you
want
it
to
be
distributed
and
then
have
the
scuffle
running
on
capillaries
remotely
in
your
cluster,
so
yeah
you
can
define
any
image
builder,
you
warned
it
could
be
docker,
it
could
be
like
Google
container
builder,
and
you
just
need
whatever
register
you
warned.
A
So
it
could
be
talking
up.
You
could
epic,
it
could
be
the
registry
from
Alibaba
cloud
or
it
could
be
the
harbor
project
from
VMware,
which
is
was
introduced
this
morning
and
then
deployment
is
also
easy
on
kubernetes
cluster.
So
for
here
we
can
yeah,
for
example,
MKE
on
this
year's,
which
is
the
multiple
kubernetes
cluster
on
our
TCOs
data
center
operating
system
yeah
using
skaffa,
we
expect
like
we
could
discover.
A
We
expect
like
scrapper,
could
handle
all
those
like
botton
layer
things
for
you,
sir,
and
we
know
the
user
no
longer
like
have
to
each
of
the
the
developer
strength-based
like
setup
new
development
environment.
So
they
just
focus
on
the
tombow
file
and
that's
it.
That's
it
and
scrap
we're
gonna.
Do
the
rest
for
you,
yeah
I.
Think
in
the
next
example
guess
Tonkin
at
mo
we're
gonna
show
like
how
to
create
a
complex,
manifest
and
then
define
us
careful,
and
then
you
can
have
your
database
service
running
running
on
top
of
this
infrastructure.
A
Okay
yeah:
this
is
a
quick
highlight
for
users.
Users
just
need
to
save
your
config
and
then
defy
your
application
and
trigger
the
scuffle
on
kubernetes.
That's
it
and
you're
supposed
to
have
like
your
application,
running
yeah
I
think
we
can,
because
this
demo
gonna
take
kind
of
long
and
we
don't
have
like
VPN
access,
so
we
running
it
locally,
so
we
gonna
get
started
right
now
for
the
demo.
Okay,
I
can.
B
Hold
the
mic
for
you:
that's
great
I
notice.
What
I
was
talking,
that
I
will
need
both
hands
to
type
so,
as
you
were
said,
we're
doing
this
all
locally
here,
so
we're
going
to
use
mini
cube
and
local
docker
registry,
because
I
didn't
want
you
to
trust
the
network
here,
all
the
demo
space
on
a
github
repository
that
is
linked
in
the
slides.
B
So
you
can
take
a
deep
look
at
the
code
if
you
want
later
what
I'm
going
to
do
right
now
is
just
go
through
all
the
steps
quickly
and
show
you
how
you
can
get
build
your
first
classification
very
quickly
and
then
use
scaffolding
coordinators
to
deploy
it
very
fast
and
how
you
can
iterate
and
do
changes
and
see
at
reflected.
So
the
first
step
of
that
obviously
is
to
create
your
base
Russification.
B
So
we
will
start
by
creating
the
skeleton.
The
first
step
is
to
create
a
cargo
tamil
file,
where
you
specify
some
information
about
your
application.
To
start
with,
it
will
be
the
name
version
and
then
the
dependencies
we
will
be
using
diesel
here
for
what
this
is
yeah
here
for
connecting
to
post
resumed
machine
so
just
to
a
database,
and
then
we
will
use
frog
it
as
a
webserver
and
that's
to
add
first
step.
B
Once
we
have
the
double
file,
then
we
will
create
a
very
basic
web
server
that
always
returns
the
same
stuff,
always
rocket
web
server.
When
you
query
that
that's
enough
it's
a
very,
very
little
code,
then
we
will
create
a
docker
file
to
tell
dogger
how
to
build
this
application.
This
looks
very
long
because
I'm
doing
some
magic
here,
so
it
caches
the
dependencies.
B
B
B
B
B
But
the
next
step
like
this
is
it
was
the
basic
stuff
just
running
stuff
locally.
The
fun
part
is
now
running
things
on
kubernetes
and
I'm,
hoping
that
that
will
work
perfectly.
The
first
step
is
to
create
your
coordinators
resources,
so
we
will
create
a
service
and
a
deployment
that
tells
Schoolnet
us
how
to
deploy
this
Raths
application
that
we're
building.
B
So
we
have
to
create
a
ml
file
there.
The
next
step
is
to
configure
scaffold,
so
it
uses
that
file
to
to
your
deployments
in
your
builds.
So
the
scaffold
compression
is
very
short.
We
will
tell
it
what
the
image
is,
that
we
should
build,
that
it
should
be
used,
augur
to
build
it
and
everything
should
be
done
locally
and
that
you
should
use
this
Yama
file,
which
was
created
in
a
previous
step.
To
do
the
deployments.
B
A
B
Luckily,
this
deployment
worked
so
the
local
one
didn't
work
using
docker,
but
we
just
managed
to
deploy
an
application
and
Cornett
is
just
creating
two
general
files,
and
the
next
step
is
to
see
how
easy
it
is
to
change
your
application
and
how
they
deployed
again
so
I'm
going
to
modify
this
web
server
and
just
change
the
message
from
rocket
web
server
to
scaffold,
updated
me
and
it's
just
a
matter
of
changing
the
file,
so
I
just
change
the
file,
and
then
you
will
see
up
here.
Let's
couple
the
deck,
sad
and
envious
it
again.
B
We
can
see
a
scaffold
updated
at
so
basically
I
just
change
one
file
and
scaffold,
build
it
again
and
redeployed
it.
So
now
that
we
have
our
basic
application
ready
on
rust
and
we
have
scaffold
configured
to
build
our
application,
build
a
container
and
deploy
it.
We
will
use
rest
on
some
rest
frameworks
to
make
this
application
talk
to
a
database,
of
course
with
database
and
then
write
some
code.
So
we
have
a
REST
API
to
modify.
B
This
is
a
will,
be
a
pretty
boring
applications
just
a
service
at
less
human
than
a
list
of
employees
and
add
and
remove
them.
But
you
can
use
this
as
an
example
and
then
do
something
more
fancy
in
real
life.
So
the
first
step
is
to
deploy
Postgres
process
eagle
and
I'm
going
to
do
that
in
a
very
huggy
way,
just
running
it
as
a
single
container
without
any
persistent
storage
or
anything
like
that,
it's
just
something
for
the
demo.
I,
don't
recommend
doing
that
in
production.
B
So,
in
order
to
do
that,
we
will
basically
add
Posterous
service
and
a
deployment
here
and
we're
just
going
to
use
the
ball
stress
image
and
a
hard-coded
username
and
password
again,
you
shouldn't
do
this
in
production.
This
is
not
a
che
at
all,
not
Haley,
really
highly
available
at
all,
but
it
works
fine
for
demos.
B
A
We
find
a
valuable
because,
like
when
we
are
working
on
container
with
continuous
solutions
like
CSI,
container
storage,
interface,
CNI
and
kubernetes,
as
well
as
like,
like
Jenny,
mentioned
this
morning,
so
this
infrastructure
solution
it
might
be
boring,
but
it
really
make
operators
its
life
easier,
and
hopefully
we
we
expect
like
this
could
change
like
unified
process.
How
operator
work
with
cluster.
B
So
now
we
can
see
that
we
can
connect
to
the
process
database
and
now
we
have
to
just
create
the
right
end
tables
and
that
stuff
so
that
web
application
can
use
it.
I
will
assume
that
you
have
already
installed
diesel
I
have
already
installed,
but
it
it
provides
a
nice
CLI
that
you
can
use
to
create
your
tables
and
then
to
rama
creations
and
rerun
them
and
stuff
like
that,
and
the
first
thing
that
you
have
to
do
is
create
an
environment
file
telling
people
how
to
connect
to
your
database.
B
So
this
created
to
my
retrieve
files,
one
is
the
up
sequel
script
and
the
IO
and
sit
down
the
up
one
is
created
to
is
trying
to
create
all
your
tables
and
the
done
one
will
destroy
everything.
We
will
create
a
very
basic
table,
just
a
single
one
for
employees
with
just
a
first
name
last
name
agent
title.
B
B
Let's
use
this
o
to
round
out
and
create
a
table
and
okay
now
we
have
posters
running
on
coordinators
and
we
use
these
to
create
a
tables.
The
next
step
is
to
make
our
application
connect
to
this
database
that
we
created.
In
order
to
do
so,
we
will
create
a
new
file,
a
rust
file,
and
that
has
a
method
that
will
give
you
a
post
res
connection.
B
B
The
next
step
is
to
create
a
matching
model
for
this
table
that
we
created
it
is
a
very
simple
class
were
struck
with
the
same
fears
as
a
table,
and
then
this
a
little
has
some
magic
that
will
do
the
mapping
before
between
the
table
and
this
struct.
In
order
to
do
so,
we
have
to
call
this
tool
that
wouldn't
generate
the
code,
and
basically
now
we
have
a
rust
application
that
has
all
the
moles
and
connection
logic.
To
talk
to
you,
your
courses,
posters
database.
B
We
have
a
rocket
web
server
already
set
up
and
we
have
the
database
running.
So
we
have
to
bring
everything
together
for
step.
To
do.
This
is
to
just
insert
default
user
or
sorry.
Employee
will
be
very
boring,
will
be
called
some
person,
that
is
a
software
engineer
and
twenty
five
years
old,
and
then
we
will
list
that.
So
we
do
that
manually
and
yeah.
We
see
that
something
was
created.
B
The
next
step
is
to
add
that
secret
tune
to
cornetist,
so
it's
possible
to
for
our
application
to
find
the
URL
and
be
able
to
connect
with
other
ways
this
bit
here.
Does
that
this
characters
that
you
see
here
is
basically
this
text
here,
which
has
the
password
and
username
and
the
host
name
encoded
using
base
64.
B
Again,
you
can
see
I
just
added.
I
I
appended
to
the
existing
file
with
the
resources
and
then
scaffold
automatically
takes
care
of
creating
a
secret
for
me,
so
it
will
watch
this
directory
yang
apply
all
the
changes
on
the
on
the
cluster.
The
next
step
is
to
update
the
current
deployment
for
application
and
make
it
use
the
secret
that
I
just
created.
B
B
We
wait
for
it
to
be
deployed,
it
was
deployed.
We
can
see
it
here.
So
when
you
do,
when
we
get
the
index
page,
we
should
see
the
information
of
that
user.
So
we
have
some
person
that
we
created
so
yeah.
We
basically
now
half
of
our
application
connected
to
the
database
and
the
final
step
is
to
implement
arrest
methods.
So
you
can
add
list
remove
an
update
employees
from
from
your
database.
B
B
Then
we
need
to
do
add
some
boring
classes,
so
we
can
handle
errors
and
input
from
from
users.
So
basically
these
uses
here
rocket
forms,
so
we
can
get
employees
as
JSON
again
just
a
handful
of
lines
of
code.
Russ
makes
it
really
easy,
and
finally,
here's
where
all
the
magic
is
I
wish
I
could
go
into
details,
but
I
only
have
500
minutes
left.
B
So
if
you
want
to
see
this,
you
can
take
a
look
at
the
repo
that
I'm
using
here
and
but
this
is
the
only
implementation
for
the
HTTP
get
put
post
and
edit
methods
it's
less
than
200
lines
of
code.
So
it's
really
short
and
concise,
and
language
is
going
to
replace
the
previous
fight
with
this
one
and
change
our
main
file.
So
it
uses
this
method
is
implemented
and
that's
basically
a
final
step.
Now
scaffold
redeploy
will
replay
that
mediate
very
fast.
B
A
B
B
B
B
B
A
So
that's
the
last
s
the
last
for
this
talk
I
want
to
summarize
like
so
you
guys
might
be
curious,
like
oh
you
guys,
like
this
missus
people.
Why
are
you
attending
like
kubernetes
kubernetes
Kong?
What
are
you
presenting
the
work
like
supporting,
kubernetes
and
how's?
The
competition
between
like
Macy's
and
kubernetes
yeah
I
got
a
lot
of
those
questions
for
the
time
being
so
I
just
like.
A
Maybe
this
is
a
good
chance
to
make
like
clarification
here
like
we
don't
have
like
competition
relationship,
and
then
we
do
in
our
product
these
years
we
do
have
like
support
kubernetes
on
our
product
and
we
have
kubernetes
running
on
top
of
mesas
easily
and
it
works
well
in
production.
So
this
two
component
they
focus
on
on
different
layers
and
also
it's
totally
up
to
user,
like
you
want
more
on
resources,
resource
allocation,
resource
management
or
you
want
more
on
like
continued
orchestration,
you
want.
A
Do
you
want
like
single
scheduling
strategy
or
you
one
like
the
fairness
resource
arrangement
on
your
tasks,
so
it's
totally
up
to
the
user
and
on
our
platform,
our
users
and
customers.
They
could
like
either
run
mesas
task
or
kubernetes.
Job
is
totally
up
to
the
user
and
most
of
them
they
just
run
both
on
their
clusters,
but
it's
totally
up
to
like
what
the
application.
It
is.
What's
the
purpose
and
yeah
yeah.