►
From YouTube: Using Services Outside Kubernetes from Kubernetes with the Service Catalog - Ara Pulido, Bitnami
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
Using Services Outside Kubernetes from Kubernetes with the Service Catalog - Ara Pulido, Bitnami
Combining Kubernetes with external services available from public cloud providers can be a powerful way to deploy cloud native applications. Developers can focus on deploying their applications to Kubernetes, while delegating things like database management to a public cloud managed service. The Service Catalog bridges these two worlds by allowing developers instantiate services outside Kubernetes directly from the Kubernetes cluster. In this demo led talk we will show how the Service Catalog allows users to build hybrid applications that use best of breed cloud services and custom made Kubernetes applications running in a Kubernetes cluster. We will demo some public cloud ready Helm Charts and how we can manage these services from Kubeapps, a Kubernetes application dashboard.
To learn more: https://sched.co/GrRR
A
Okay,
we
are
going
to
get
us
started
thanks.
Everyone
for
coming.
My
name
is
eric
koleda.
I
work
at
a
company
called
vietnamese
engineering
manager
and
for
the
next
few
minutes.
What
I'm
going
to
be
talking
about
is
the
service
catalog
and
how,
as
an
app
developer,
you
can
use
it
to
connect
cloud
managed
cloud
services
with
your
communities
applications.
A
So
the
way
I'm
going
to
do
this
thing
is
I'm
going
to
spend
about
10
minutes
explaining
very
briefly
with
the
Service
Catalog
is
and
how
it
looks
like
and
then
I'm
going
to
do
a
demo
that
I'm
going
to
hopefully
show
four
different
examples
on
applications.
I
use
the
Service
Catalog
to
connect
to
those
services.
A
The
examples
are
using
home
charts
home
is
a
package
manager
for
kubernetes.
But
if
you
are
not
familiar
with
home,
the
format
is
basically
go
templating
on
llaman,
so
you
you
can
follow
along
because
it's
basically
llamo
but
first
who
we
are
Vietnamese.
So
we
are
leaders
in
application
packaging.
We
have
more
than
400
applications
packaged
for
the
different
cloud
providers.
A
So,
basically,
if
you
go
to
AWS
marketplace
or
DCP
marketplace,
and
you
want
to
spin
up
a
cloud
image
running
WordPress,
for
example
out
of
the
box,
that
image
may
be
ours
and
that
catalog
of
publications
we
also
for
us
containers
and
and
home
chart,
so
we
actually
maintain
many
of
the
community
repo
home
charts
and
our
containers
are
updated
on
a
daily
basis
or
rebuilt.
So
there
they're
up
to
date
for
security
issues
on
this
application.
A
Well,
we
also
do
a
lot
of
open
source
projects
on
the
application
space
in
communities
like
cubes,
which
is
a
curious.
That's
where
application
dashboard
for
your
cluster,
the
communities,
production,
runtime,
so
secrets,
cublas
and
other
stuff
I'll
be
sharing
a
little
bit
cube
ops
and
during
this
talk,
but
first
why
we
need
manage
services.
So
we
all
know
that
curious
is
great
to
run
your
applications
that
run
containers
in
production.
A
But
one
of
the
great
benefits
of
being
really
cloud
native
is
that
you
can
use
also
these
services
that
this
cloud
provides
and
they
manage
themselves.
So
you
really
can
focus.
You
can
focus
on
what
makes
your
application
your
application.
What
provides
value
to
your
business
rather
than
having
to
provide
all
this
plumbing
for
your
applications?
So,
for
example,
if
your
application
needs
a
database
or
a
pub
sub,
Q
or
storage,
a
packet
storage,
and
instead,
obviously
you
can
create
that
yourself.
A
Profissional
yourself
in
your
cluster
maintainer,
doing
backups,
I'd,
say
all
these
things
or
you
could
use
one
of
these
services
coming
from
these
clouds,
and
they
do
all
of
that.
For
you
they
can
offer
multi-region
auto
scaling.
They
can
offer.
Obviously
backups
for
for
your
databases,
and
the
problem
is,
is
disconnect
between
your
kubernetes
application
and
those
services.
So
maybe
you
have
to
go
to
this
cloud
you
ice
and
provision,
one
of
these
services
or
using
the
CLI
or
using
the
API
or
using
something
like
terraform.
A
You
can
use
as
well,
but
then
you
have
to
really
connect
that
to
your
application.
You
have
to
create
your
secret
synchronise
yourself
to
to
connect
to
that
app
or
use
external
service.
So
the
thing
is
that
this
breaks
the
declarative
model
in
communities
where
you
can.
The
goal
is
that
you
can
describe
your
application,
how
it
has
to
look
like
and
Cabrini's.
That's
the
other
thing
for
you
on
the
a
reconciliation
loop,
so
the
service
catalog
basically
tries
to
solve
that.
A
So
it's
a
declarative,
kubernetes
api
extension
to
discover
and
manage
services
that
implements
the
open
service
broker
API.
So
the
open
service
broker
API
is
standard
schema.
It
was
really
created
for
the
Cloud
Foundry,
but
now
he
has
implementations
also
for
Cuban,
eighties
and
OpenShift,
and
it's
super
super
basic
API.
It
has
doesn't
have
that
many
objects,
so
this
is
more
or
less
how
it
looks
like
at
the
bottom.
A
You
have
the
service
broker,
which
is
the
abstraction
of
the
service
provider,
so
you
can
have
a
service
broker
for
DCP
one
/
AWS
one
for
rushar
or
even
for
your
own
cluster.
So
if
you
want
to
as
part
of
your
infrastructure,
you
want
to
offer
services
that
are
on
your
infrastructure
using
the
same
API,
you
can
basically
implement
a
service
broker
yourself.
A
On
top
of
that,
you
have
different
service
classes,
which
is
the
representation
of
this
vyses.
Of
example.
You
can
have
service
class.
Could
be
sure
Postgres
or
DCP
storage
or
AWS
RDS?
All
of
that
are
the
classes
and
for
each
of
the
classes
you
a
broken
is
to
implement
one
or
many
plans
so
and
the
plan
tries
to
abstract
the
idea
of
the
different
tiers
of
these
services.
A
So,
for
example,
you
can
have
a
basic
plan
that
it
also
have
a
very
only
one
instance
and
it's
very
small,
with
very
little
CPU
and
very
little
ram,
and
then
you
can
obviously
with
a
low
price
point
and
then
going
up
to
a
bigger
instance
or
mania.
So
many
reasons.
So
all
of
that
are
the
service
plans.
A
Up
to
that
level.
The
service
broker
service
class
and
service
plan
are
cluster
level
level
objects.
So
it
has
to
be
a
cluster
admin
who
deploys
that
you
to
the
cluster
and
they're
available
cluster
wide,
and
then
we
have
service
instance
and
service
bindings.
So
those
two
objects
are
named
spaced
and
usually
a
developer
who
wants
to
use.
One
of
this
service
is
what
they
would
do
is
to
create
one
of
these
services
as
objects
and
the
service.
A
Catalog
controller
will
talk
to
the
service
broker
and
provisioned
an
instance
for
them,
and
so
this
service
binding
is
the
tries
to
describe
the
connection
between
your
application
and
a
service
instance,
and
that
takes
the
form
of
a
secret,
recruiting
a
secret
that
then
you
can
use
to
in
your
application.
So
a
very
simple
API.
A
So
these
are
couple
of
examples
just
to
see
that
these
objects
really
look
like
any
other
community
objects.
This
is
the
class
of
service
broker
for
DCP
and
a
sure
you
have
your
API
version,
your
kind
some
metadata,
and
then
these
super
simple.
You
just
have
an
endpoint,
a
URL
that
talks
to
to
that
service
provider
and
some
of
obviously
you
have
to
have
your
credentials
somewhere
to
to
basically
to
your
project
in
DCP
with
your
subscription
in
usher.
But
you
can
see
that
these
are
very
different
implementations.
A
If
you,
if
you
realize
the
DCP,
you
are
always
pointing
to
Google
itself,
so
it
runs
the
broken
ranks
on
your
project
in
disappear.
In
the
usher
case,
you
can
see,
is
it's
a
service
running
on
your
cluster?
So
if
you
want
to
provide
your
users,
your
cluster
users,
you
want
to
provide
sure
you
need
to
create
a
deployment
and
the
broker
will
bring
a
support
in
your
cluster.
A
So
when
the
class
reporter
creates
one
of
these
resources,
basically
what
it
happens
is
in
it
connects
to
the
the
controller.
Will
talk
to
that
end
point
and
if
you
have
permissions
to
the
so
it
will
retrieve
a
list
of
available
services
and
plans
and
it
will
create
all
these
cluster
service
classes,
resources
and
classes
service
plans
again
at
the
cluster
level,
and
this
is
a
couple
of
examples
of
of
these
classes.
One
is
Google
storage,
the
other
one
is
the
usher
database
for
Postgres
sequel
and
one
of
the.
A
A
Another
way
to
interact
it
is
I'm
going
to
be
showing
during
the
demo
a
couple
of
tools,
so
I
want
to
show
them
here.
One
tool
is
sv
cut
it's
in
the
same,
very
positive
service,
catalog,
basically,
an
official
CLI
tool
for
the
service
catalog,
it's
great
for
testing,
so
it
basically
allows
you
to
do
things
like
add
brokers,
get
classes
gets
instances
all
those
things.
A
Obviously
you
can
do
with
with
cube
CTL
as
any
other
object,
but
the
formatting
of
the
output
is
a
little
bit
better,
but
also
it
allows
you
to
do
provision
your
instances
without
having
to
be
declarative.
So
you
it's
a
it's
a
good
way
to
try.
If
you
want
to
test
that,
you
can
actually
provision
ELISA
that
you're
giving
it
the
right
parameters.
All
of
that,
it's
a
very
quick
way
to
give
you
a
try
and,
as
I
said,
we
are
going
to
be
using
cube
ops
during
the
demo
as
well.
It's
an
application.
A
A
Expects
Wi-Fi
Wi-Fi
to
work.
So,
let's
hope
that
happens
so
I'm
running
I'm
running
a
5,
node
GPU
cluster
I
have
the
service
catalog
running,
so
you
have
the
API
server
and
then
the
controller.
As
you
can
see,
it's
an
API
server.
So
it's
not
a
seer
deem
which
may
not
may
or
may
not
be
a
good
idea.
But
that's
how
it
is.
I
have
Q
apps
running
with
a
lot
of
Singh
jobs
and
I.
A
Have
a
couple
of
brokers
deploying
my
cluster,
so
the
DCP
broker
and
the
open
service
broker
for
Usher
both
are
on
our
ready
status.
So
that
means
that
I
should
have
some
classes
available
from
those.
So
you
can
see
that
there
are
a
lot
of
things
that
I
can
use
in
my
Cuban
Aires
application.
So
I
have
my
sequel
from
both
providers.
Pops
up
have
BigTable.
Even
all
these
things.
A
A
Okay,
so
this
is,
this
is
more
or
less
kebabs.
This
is
the
these
are
the
applications
that
I
have
deployed
in
in
my
cluster
I
can
check
a
catalog
of
home
charts
available
from
different
repos
I
can
add
my
own
private
repo
if
I
want
most
of
them
coming
from
the
community
of
repo,
some
of
them
from
the
bean
ami
one,
some
of
them
from
Usher
and
I
can
also
check
the
available
service
instances
that
I
have
as
part
of
a
service
catalog.
A
In
my
in
my
cluster
again,
this
is
namespace,
so
this
is
checking
to
the
full
name
space
right
now
so
on.
This
is
not
going
to
be
a
fully
left
Emma,
because
these
services
take
a
while
to
provision.
So
I've
do
some
work
before
the
first
example
is
going
to
use
is
going
to
use.
Wordpress
is
going
to
deploy
the
helm
chart
for
WordPress,
but
instead
of
normally
the
home
turf
for
WordPress
is
dependent
on
the
Maria
DB
chart.
A
A
A
It's
it's
basically
dependent
on
the
implementation
of
the
broker.
So
there
is
no
standard,
so
it
depends
from
the
broker.
It
depends
on
the
class
or
in
this
case,
in
the
plan.
So
you
have
to
know
beforehand
through
the
commentation
from
the
brokers,
how
those
are
going
to
look
like
in
order
to
be
able
to
connect
your
applications
to
those.
So
in
this
case
is
how
it
looks
like
for
a
sure,
my
siculus.
A
A
A
A
If
we
now
check
the
application
should
be
deployed,
we
have
our
WordPress
up
and
running
using
our
my
sequel
when
you
create,
by
the
way,
when
you
create
those
instances,
for
example
using
cube
ops.
If
you
want
to
create
a
new
one,
you
can
select
a
plan
and
the
different
plans
have
a
JSON
schema
of
the
parameters
that
it
expects.
A
So,
let's
put
a
name
so
and
that
we
read
that
JSON
a
schema
over
here
so
again,
cubes
is
also
a
good
way
to
do
testing
of
these
things,
because
documentation
is
not
great
and
reading
G,
so
my
schema
yourself
could
be
a
little
bit.
So
this
is
a
good
way
to
know
what
are
the
parameters
that
this
service
is
expecting?
A
Okay,
so
that's
for
the
first
example.
The
second
one
is
very
similar.
It's
a
little
bit
same
way.
That
imperative
is
not
declarative
in
this
case
is
char
museum,
but
I
wanted
to
add
it
here
quickly
because
it
uses
TCP
and
just
instead
of
Usher,
so
we
can
have
a
couple
of
brokers
of
examples,
so
charm
use
iam.
A
If
you
don't
know
what
it
does,
it's
an
various
way
to
run
a
chart,
a
helmet,
our
repository
on
your
community's
cluster,
and
it
allows
you
to
use
different
backends
for
storage
like
local
storage
or
AWS
or
DCP,
that
we
we
just
we
just
did
so.
In
this
case
we
did
the
same
thing,
so
we
created
an
instance
for
disappear
storage.
So
back
a
storage,
it
was
provisioned
successfully.
A
We
created
a
binding
and
the
good
thing
about
this
binding
and
therefore
the
secret,
and
that's
why,
in
this
case,
I
didn't
have
to
fork
anything
I
just
created
the
folds.
The
default
chart
museum,
chair
Reaper,
which
is
in
the
stable
community
of
Reaper,
is
that
well
it
has
a
bucket
ID
and
then
has
a
JSON
file
in
the
K
private
key
data
that
has
the
default
format
for
authentication
to
DCP
services.
So
this
is
a
JSON
file
that
people
connect
in
to
DCP
services
suspects
the
same
format
all
the
time.
A
So
that's
why
the
charm
is
iam
ones
are
very
brave
for
it,
so
this
is
already
deployed.
The
recent
ways
of
very
deployed
is
because
I'm
using
derma
C
and
this
one
too,
as
a
Reaper
for
my
fork,
so
I
had
to
to
have
it
already
there,
but
if
I
click
here
and
upgrade
I'm
not
going
to
upgrade
it,
but
it
allows
you
to
show
the
different
values
that
I
provided
so
basically
I
said:
ok
for
backing
storage
use
Google.
A
This
is
the
name
of
the
bucket
I
could
have
used
the
key
instead
of
just
copy
and
pasting.
By
should
that
they
Chari
is
not
ready
for
that,
so
it
doesn't
have
a
reference.
I
would
need
to
fork
that
as
well,
if
I
wanted
to,
but
then
on
the
at
the
very
bottom-
and
this
is
ready
from
the
upstream
shark
and
say:
ok,
DCPI
authentication,
it's
in
a
secret.
This
is
the
name
of
the
secret
and
the
key
that
contains
the
JSON
to
connect
to
it.
It's
called
private
private
key
data.
A
So
that's
another
example,
but
that's
not
the
best
way
to
use
the
service
catalog,
so
the
best
way
to
use
the
service
catalog
it's
using
the
corrective
description
of
your
of
your
application.
That's
that's
the
main
goal
of
the
Service
Catalog,
so
we
are
going
to
see
quickly
a
couple
of
examples
that
do
that
the
first
sample
is
going
to
be
a
mean
Stax,
Mongoose,
press,
angular,
etc.
A
That
is
it's
going
to
use
our
manga
in
the
backhand
from
cosmos
TV.
So
in
this
case
these
are
two
different
example.
The,
for
example,
the
mean
one.
Is
that
the
case
where
you
have
already
have
an
instance,
a
MongoDB
instance
in
Usher
and
thanks?
Obviously,
you
can
use
it
for
four
different
applications,
so
you
don't
have
a
binding
yet,
and
the
binding
is
something
that
you
want
to
to
show
as
part
of
your
like
to
create
as
part
of
your
application,
so
the
min
chart
from
benami
it's
already
ready
for
this.
A
It's
going
to
start
deploying
application
and,
as
part
of
my
application
is
going
to
create
the
service
binding.
So
if
we
go
here
to
the
MongoDB
we
refresh
here,
you
can
see
that
is
how
created
binding
as
part
of
the
description
of
of
my
application,
and
probably
this
already
deployed
waiting
for
the
load
balancer.
A
While
this
deploys
I'm
going
to
check
how
that
works,
this
museum
one.
This
is
a
Minton
plate,
as
you
can
see
as
part
of
the
templates
now
I
have
a
binding
again.
This
is
an
object
in
kubernetes
as
part
of
the
description
of
my
application,
but
it
has
a
service
binding
that
I
basically
say.
Ok,
the
instance
is
this
one
that
I
put
them
and,
and
then
it's
going
to
do
the
same
thing
like
you
see
in
the
different
keys
that
I
know
the
name
of
to
connect
to
to
that
database.
A
So
if
that's
applied
so
now,
I
have
my
mean
start
to
start
creating
my
application,
backed
up
in
usher
and
the
final
example.
It
is
also
the
clarity
of
is
how
you
should
do
things
here.
It's
basically
a
ghost
block
that
it's
back
up
by
Usher.
This
is
a
chart
created
by
assure
themselves
as
a
as
an
example
of
these
things.
It's
a
fork
from
the
original
ghost
chart
and
it's
it's
under
get
Ripper
as
well.
So
let's
try
to
find
that
one
and
this
one
is
going
to
do
the
full
thing.
A
A
A
You
can
see
this
provisioning
I
didn't
do
anything.
It's
perfectly
application
description,
so
this
is
bit
takes
a
while
to
provision
one
of
these
instances.
It
depends
on
the
broker.
It
depends
on
the
service
for
this
particular
one.
It
takes
about
10
10
minutes,
10
50
minutes.
So
obviously
we
are
not
going
to
show
everything,
but,
as
you
can
see
here,
because
it's
part
of
the
declaration
of
the
application,
this
has
created
the
object
for
the
instance,
the
object
for
the
binding,
very
Vasily,
the
binding.
A
It
says
error
instance
not
ready
and
what
it's
going
to
do.
Obviously
this
is
a
reconciliation
loop.
It's
going
to
keep
trying
and
at
some
point
the
provision
will
be
successful
and
then
we'll
create
the
binding
and
that
binding.
If
you
can
see
now
the
deployment
here,
so
you
can
see
that
the
ghost
pod
is
failing
its.
A
So
you
can
see
that
the
configure
is
giving
is
that
the
secret
that
I
referenced
on
the
deployment
is
not
found.
The
reason
why
is
not
finds
because
the
binding
hasn't
been
successful
yet,
which
is
the
object
that
creates
the
secret,
but
this
is
all
on
the
reconciliation
loop,
so
you
can
basically
have
your
application.
As
part
of
you
see,
ICD
it's
going
to
deploy
everything
and
Cabrini's
will
keep
trying
until
the
incident
has
been
provision.
A
Okay,
so
these
so
quickly,
this
is
the
the
home
chart.
Pro
ghost
again
it's
in
the
assure
github
Reaper.
So
you
can
see
it
creates
a
service
instance
as
part
of
it
with
a
type
of
class
or
type
of
plan,
some
fiber
rules.
This
comes
from
the
Jason
is
came
out.
I
was
talking
about
of
the
required
values.
It
will
create
a
binding
as
well,
and
then
it
will
do
the
same
thing
of
using
the
different
keys
from
from
the
secret.
A
A
Basically,
what
I
would
I
would
give
us
takeovers
is
that
the
open
service
broker
is
a
standard
specification
to
deliver
services,
a
serving
the
service
Castle
with
an
implementation
as
a
Cabrini's
api
extension
for
it,
and
basically
that
in
general,
declarative
is
always
better
for
communities.
That's
one
of
the
key
concepts.
So
hopefully
the
service
catalog
is
still
under
heavy
development,
but
what
this
solution
of
may,
the
solutions
that
are
coming
up
will
solve
this
issue
of
bridging
the
gap
between
the
external
service
system
and
Canaries,
and
that's
it.
Thank
you.