►
Description
Automating Infrastructure Deployment for Kubernetes - Alena Prokharchyk, Rancher Labs, Inc. & Brian Scott, The Walt Disney Company
"Many organizations run Kubernetes clusters in a single public cloud like GCE or AWS, and as a result have reasonably homogenous infrastructure needs. In these situations deploying Kubernetes clusters is relatively straightforward. Some organizations, however, have diverse infrastructure needs and as a result need to automate infrastructure deployment for Kubernetes across multiple clouds and data centers. This gives rise to a few challenges:
1. How to ensure Kubernetes clusters in different clouds and data centers can communicate with each other, or in some cases even have a single Kubernetes cluster span multiple data centers.
2. How to automate the deployment of multiple Kubernetes clusters.
3. How to incorporate the new Kubernetes multi-zone clusters (f.k.a. Ubernetes) into multi cloud and multi datacenter deployments.
4. How to manage the health of Kubernetes cluster itself, including, for example, how to detect and recover from etcd node problems.
5. How to automate the upgrade of Kubernetes master and Kubelets, and how to handle multiple versions of Kubernetes clusters that exist in a single deployment.
In the last 6 months, we have worked with several enterprise IT organizations to solve these problems. We will share our experience on how to automate and simplify Kubernetes deployments."
About Alena Prokharchyk
Alena Prokharchyk is a Principal Software Engineer at Rancher Labs, where she develops container and microservices frameworks including Kubernetes. At Rancher Labs and previously as a developer at Citrix and Cloud.com, Alena has worked on cloud and container production deployments for large enterprise organizations world-wide. You can find her latest thoughts on containers and microservices on her blog (http://rancher.com/author/alena/) or by following @Lemonjet
Join us for KubeCon + CloudNativeCon in Barcelona May 20 - 23, Shanghai June 24 - 26, and San Diego November 18 - 21! Learn more at https://kubecon.io. The conference features presentations from developers and end users of Kubernetes, Prometheus, Envoy and all of the other CNCF-hosted projects.
A
Hello,
everyone
thank
you
for
joining
us
today.
My
name
is
Alena
I'm,
a
software
engineer
at
venture
labs
Rancher
is
an
open
source
product
that
natively
supports
and
manages
kubernetes
Matheson's
form
clusters,
and
here
I
have
Brian
Scott
the
Walt
Disney
Company,
one
of
our
favorite
Kudrin
areas,
users
and
today
we're
going
to
talk
about
automating
infrastructure
deployment
for
kubernetes,
but
first
we
want
to
start
with
a
short
demo,
a
trailer
actually
something
that
Disney
does
for
its
for
its
presentations.
B
A
We
try
we
try
to
make
the
rest
of
a
presentation
as
fun
as
this
preview.
So
kubernetes,
we
all
know
kubernetes
is
awesome.
It
automates
your
application
deployment.
It
provides
the
DNS
and
service
discovery
for
for
the
services,
it
monitors
the
application
health
and
it
recovers
it
on
the
fly,
and
it
also
provides
the
easy
way
to
upgrade
your
application
so,
but
who
does
the
same
for
kubernetes
cluster
II?
A
Because
if
you
think
of
it
do
native
system
cluster
is
nothing
else,
but
the
set
of
services
they
talk
to
each
other,
that
they
have
a
database
and
the
house
has
to
be
monitored
and
upgrades
should
be
handled.
The
same
way,
there's
many
great
tools
in
the
market
that
provide
all
that.
The
Kuban
iam
ansible
chaos,
boot
cube,
terraform
and
there's
my
favorite
source
Kelsie
Hightower
thanks
a
lot
to
him
for
his
dogs
and
demos.
A
His
article
kubernetes,
the
hard
way
actually
makes
it
very
easy
to
understand
the
internals
behind
that
application,
cluster
deployment
and
all
of
these
tools.
They
kind
of
try
to
solve
the
same
set
of
challenges
and
the
biggest
of
all
them
is
probably
the
cloud
infrastructure
management
because
thrown
your
kubernetes
cluster.
You
first
have
to
provision
the
host,
whether
it's
the
bare
metal
or
somewhere
in
the
cloud,
and
then
you
have
to
set
up
a
networking,
so
the
horse
can
talk
to
each
other.
A
A
And
we
started
with
laying
the
foundation,
we
wanted
everything
like
the
network
in
the
DNS,
the
story,
drivers
to
works,
cross
cloud,
and
we
didn't
want
to
maintain
the
different
set
of
scripts
for
AWS
GCE
digitalocean.
We
wanted
the
experience
for
the
user
to
be
cloud
agnostic,
so
the
user
comes,
he
picks
the
AWS
cloud
deploys
his
kubernetes
cluster
and
when
he
decides
okay,
I
want
to
switch
to
some
other
cloud.
A
The
experience
shouldn't
be
any
different
and
the
other
interesting
thing
is
like
I've
spoken
to
some
people
today
and
they're,
saying:
okay,
we
want
to
just
go
straight
to
the
kubernetes
application
deployment.
We
don't
want
to
deal
with
all
the
setups
like
we
don't
want
to
learn
about
how
to
set
kubernetes.
We
want
to
just
focus
on
the
application
deployments
in
communities
and
that's
something
we're
in
Rancheria
makes
it
easy
for
the
user
and
I
want
to
talk
a
little
bit
about
Rancher
architecture.
A
So,
as
you
can
see
here,
you
register
your
hardware
with
Rancher,
whether
it's
a
bare
metal
or
virtual
machine
in
the
cloud,
and
we
build
the
infrastructure
services
on
top
of
that.
All
the
infrastructure
services
are
continued
rise
in
Rancher.
We
don't
think
of
a
container
as
the
extra
layer
on
top
of
the
storage
networking
stack.
We
think
that
container
redefines
all
the
layers
and
everything
should
be
deployed
in
the
container.
So
we
run
load
bonds
from
the
container
networking.
Then
storage
database
DNS.
A
So
once
this
infrastructure
services
layer
is
set,
we
deploy
that
application
orchestration
platform,
whether
it's
a
kubernetes
swarm
on
masses,
it's
all
containerized
and
built
with
just
one
click
so
and
then
after
we
can
here
eyes
and
deploy
these
two
layers,
user
can
come
and
actually
start
deploying
his
applications
and
the
application
management
is
not
something
that
we
do.
We
delegated
to
the
orchestration
provider
of
the
user
choice
Granero
swore
masses
of
key.
A
A
We
enable
all
the
framework
services
on
the
in
the
cluster
to
manage
the
kubernetes
system.
Services,
like
the
rancher
health
check,
that
micro
service
will
manage
the
kubernetes
system,
services,
health
and
recover
them.
If
they
go
south,
then
the
rancher
dns.
It
provides
internal
DNS
for
kubernetes
system
services
to
talk
to
each
other,
and
then
we
have
a
range
of
metadata
like
easy
to
style
metadata
that
any
service
in
kubernetes
cluster
can
access.
A
A
Once
all
this
is
picked,
the
template
gets
deployed
with
just
one
click.
So
here
you
can
see
all
our
choices,
orchestration,
kubernetes,
the
framework
services,
the
the
networking
and
here's
a
scattering
schedule
is
actually
it's
not
the
scheduler
for
kubernetes
application.
It's
rather
the
scheduler
for
a
kubernetes
System
stack.
A
Besides,
building
the
kubernetes
deployment
template,
we
also
let
user
to
pick
the
orchestration
of
orchestration
for
the
kubernetes
cluster
deployment.
So
we
have
three
popular
choices.
One
is
a
standalone.
It's
one,
your
kubernetes
system,
services,
ETD
control,
plane
and
user
applications
get
deployed
on
a
single
host
like
let's
say
your
wanna
create
something
really
quick
and
you
don't
care
about
the
resiliency
and
stability.
So
we
go
with
this
choice.
Then
there
is
a
setup
where
you
can
aggregate
your
each
CD
and
control
plane
on
one
set
of
cost
and
user
applications
on
another
set.
A
Of
course
it
is
better.
It's
it's
more
resilient,
but
adders
like
potential
performance
issue
with
the
in
CD
and
Control
playing
components
and
then
there's
a
third
option.
Is
the
resilient
separate
planes,
its
high
performance,
because
it's
CD
and
donated
capi
server
and
other
control
services
are
deployed
on
different
set,
of
course,
and
the
user
application
is
on
the
third
set.
Of
course
this
is
production
ready
tested
by
our
customers.
A
So
once
your
applicator,
once
your
kubernetes
cluster,
is
up
and
running,
we
monitor
its
health
all
the
time.
So
you
shouldn't
really
see
like
red
and
yellow
colors
on
this
picture.
Everything
should
be
green
in
a
perfect
world,
but
yeah
here
we
go.
You
see
that
one
of
the
services
became
unhealthy.
It
means
that
it
started.
It
stopped
reporting
to
the
health
check
and
their
can
visit
the
great
estate.
A
We
also
have
a
different
health
check
strategies
based
on
your
application,
so,
for
example,
the
regular
one
just
freely
it
replace
the
container.
You
shouldn't
worry
about
anything
like
if
it's
dad
is
going
to
be
replace,
but
when
it
comes
to
it,
services
like
it
CD
we're,
maintaining
the
quorum
is
crucial.
That's
where
we
create
an
quorum
strategy
is
very
useful,
don't
recreate
unless
your
quorum
is
maintained
once
your
quorum
is
lost,
just
leave
the
application
around.
Don't
do
anything
because
you
can
Mouse
without
it
CV
data.
We
leave
it
up
to
you
to
recover.
A
A
We
suggest
that
you
have
like
three
hosts
in
your
setup,
but
again
like
the
quorum
can
be
lost
even
in
the
most
stable
environments
like
your
network
can
go
south
or
your
host
can
go
down
and
it's
very
important
to
configure
a
backup,
the
remote
backup
in
case,
something
like
all
the
hosts
go
down.
So
in
Rancher
we
recover
from
majority
of
the
ETD
failures.
We
do
the
disaster
recovery
and
we
also
allow
you
to
configure
remote
backups
and
then
you
can
restore
from
the
backups
if
you
lose
all
the
hosts
in
your
system.
A
So
let's
say
you're
a
grenadiers
cluster
is
up
and
running
it's
functioning.
Well,
it's
it's
being
health
check,
one
it
fit
and
recover
it
as
it
goes.
But
there's
always
time
where
you
want
new
features.
New
kubernetes
features
or
bug
fixes
and
kubernetes
is
pretty
good
with
its
release.
Processes
like
the
religious
happened
really
fast,
and
we
in
Ranchi.
We
follow
CUDA
latest
release
and
test
every
major
and
almost
every
minor
release
once
tested
we
published
in
played
through
our
infrastructure,
catalog
and
the
users
who
have
the
kubernetes
deployments.
They
get
a
notification.
A
A
A
A
Nice
welcome
to
Rancher,
so
you
have
a
choice:
whether
you
want
to
allow
to
collect
anonymous,
tats
I'm,
just
gonna
I,
just
going
to
disable
this
option
for
now
got
it
alright.
So
here
is
the
rancher
UI
and
the
the
first
thing
you
start
with
is
with
building
your
environment,
template
and
creating
an
environment.
So
what
is
the
environment
in
Rancher?
A
A
A
You
can
also
enable
access
control
for
your
cluster.
It
can
be
like
we
support
various
various
access
control
modes.
It
can
be
a
github
or
LDAP
or
anything
else,
I'm
just
going
to
go
with
all
the
access
control
for
now
so
once
I
click
create
so
here's
my
genetics,
environment
and
I
can
choose
it
here,
but
nothing
is
going
to
happen
yet
just
because
this
environment
doesn't
have
any
hosts,
so
we
have
to
add
a
host.
At
least
one
is
fine,
so
I
go
to
the
infrastructure
page
and
here
I
have
an
option.
A
A
A
Dns
health
check,
IPSec
and
here's
my
coronaries
system,
clustering
and
you
can
see
there
is
a
great
available
option
just
because
I've
chosen
that
are
not
the
latest
version
of
the
kubernetes
to
deploy
from
here.
You
can
actually
see
the
graph
with
your
computer's
application,
so
you
can
see
the
each
CD.
The
kubernetes
api
server
and
all
other
components
talking
to
the
kubernetes
api.
Everything
is
green.
Everything
works
so
far.
You
can
also
see
the
template
version
of
the
rails
cluster.
A
We
try
to
avoid
introducing
a
new
configuration
language
for
configuring,
the
kubernetes
system
clusters,
so
we
went
to
the
standard
docker
compose,
so
here
is
just
just
regular
dr.
commands,
plus
some
scheduling
labels
that
we
use
to
orchestrate
the
kubernetes
system.
Cluster
ii.
If
you
go
back
I'm
just
going
to
test
an
H
here,
really
quick
I'm
going
to
do
it
by
killing
one
controller
manager
container.
A
Go
please
load
oops,
okay,
so
it
is
loaded.
I'm,
just
gonna,
delete
it
right
here
and
see
what
was
going
to
happen
here.
So
here
immediately
oops
the
service
went,
unhealthy
and
I
goes
fish
from
the
yeah.
So
what
is
initializing
state?
So
we
treat
the
containers
like
the
startup
process.
We
divide
into
two
pieces.
One
is
the
docker
state
which
is
like
running
or
stopped
and
then
there's
an
application
speed,
because
the
docker
running
doesn't
mean
that
your
application
is
up
up.
A
Your
application
is
up
only
when
it
starts
listing
a
specific
board
that
health
check
is
shaken
against.
Only
then
we
consider
the
service
as
active
and
yeah.
You
can
see
you
can
see
the
details
of
any
service
in
your
cluster.
Like
you
can
see
the
controller
manager
continue
the
IP
address
assigned
to
it.
You
can
see
the
boards,
you
can
see
the
log.
This
is
a
service
log.
A
A
All
right,
so
if
you
go
back
to
the
infrastructure
stacks
okay
upgrade
available,
let's
try
it
out
great.
So
here
you
can
pick
okay
I'm
going
to
pick
the
one
for
five,
because
that's
the
latest
you
just
going
to
agree
in
my
code.
Renee
disclosed
three
okay,
so
it
starts
without
grading
the
API
server,
and
you
can
see
that
the
bunch
of
services
like
a
ingress
controller,
goes
on
healthy
because
it
talks
at
a
pacer
and
now
the
connection
is
lost.
For
for
a
second,
it's
going
to
be
recovered.
We
listen.
We
go
okay.
A
If
my
take
a
bit
longer
just
because
it's
downloading
the
new
coronaries
image
to
the
to
the
host,
but
it
shouldn't
take
a
lot
of
time-
I
hope,
I'm,
not
afraid
of
having
a
failing
down
more
or
a
controller
manager
compared,
is
stuck
in
starting
state.
You
can
actually,
let's
see
okay,
so
it
back.
I
can
use
logs.
I
can
see
that
it's
it
it
connected
to
the
kubernetes
api
server
secure
for
successfully
so
should
be
green
again.
Now
go
back
to
the
infrastructure
stats,
okay.
So
it's
still
spinning
okay
scheduler!
A
A
A
B
Alright,
hey
guys
so
I'm
Brian,
I'll
kind
of
go
into
inch
a
little
bit
and
it's
a
bit
so,
as
you
know,
Disney's
a
very
large
organization
with
many
different
silos
and
companies
underneath
it
we're
so
far.
The
big
lumpy,
a
company
out
there
with
a
pretty
big
brand
to
kind
of
keep
up.
So
with
that
we
have
many
different
teams
within
Disney
all
doing
different
things.
B
So
we
literally
like,
like
use
almost
every
orchestration
framework
out
there,
and
it's
actually
pretty
good,
because
we
get
a
lot
of
feedback
from
both
external
and
internal
teams,
so
I'm
going
to
kind
of
quickly
show
a
quick
intro
or
a
quick
demo
of
a
app
deployment.
So
again,
I
support
both
Walt
Disney,
Studios
and
Imagineering.
So
I
have
a
embedded
team
embedded
in
both
of
these
segments
and
our
job
is
to
assist
that
business
unit
with
whatever
they
whatever
their
hosting
needs
are.
B
So
here's
all
our
challenges.
Right
is
the
infrastructure
setup
of
kubernetes,
because
we
are
a
multi
cloud
as
well
as
on-prem
our
back
access.
So
again
we
do
have
to
have
strict
rules
around
which
members
can
access,
what
which
LDAP
groups
have
access
to
which
environments,
especially
for
QA
and
dev
teams,
and
then
we
have
to
be
able
to
develop
to
be
like
multi
tenant,
as
well
as
a
multi
cloud
within
Walt
Disney
Studios.
B
We
have
30
different
dev
teams
that
we
have
to
to
support
and
we
have
to
deploy
our
infrastructure
across
both
Amazon
to
Google
and
on
print,
which
is
basically
a
cloud
like
like
cloud
stack
and
and
OpenStack
and
again
we
have
to
be
a
self-service
as
possible
because
we
are
a
small
team
of
five
within
Studios
servicing
30
different
teams.
We
have
to
be
able
to
provide
as
much
self-service
and
automation
as
possible
out
to
our
various
teams.
B
So
again,
right
now,
we've
been
on
rancher
for
about
I
would
say
six
to
eight
months.
We
have,
since
the
India
we
have
about
roughly
60
applications
running
within
venture
all
polygon
allocations.
So
no
GS
Java
spring
go
no
js',
you
name,
it
is
probably
in
there
the
two
main
orchestrations
that
we've
chose
is
cattle
and
kubernetes
and
then
again
we're
across
two
different
types
of
clouds.
So
we
are
able
to
scale
our
applications
across
any
infrastructure
that
they
like.
We
need
that
we
seem
fit
to
our
major
applications
that
we
are
running
within.
B
Committees
is
Disney
movies,
anywhere
calm.
If
you
haven't
tried
it
out.
Basically,
if
you
buy
a
Disney
movie
and
let's
say
iTunes,
that
entitlement
automatically
flows
into
Amazon
and
or
Google
Play
so
allows
you
to
buy
the
movie
once
you
get
access
to
access
to
it
almost
anywhere
and
again,
we
just
launched
the
Doctor.
Strange
missiles
was
one
of
the
applications
that
we
launched
within
rancher.
We
literally
got
the
code
Sunday
night
and
had
to
be
live
for
a
announcement
that
following
morning
so
with
that
I'll
go
into
a
quick
demo.
B
So
this
is
the
rancher
infrastructure
that
I
spun
up.
Digitalocean
was
very
nice
enough.
To
give
me
some
notes
to
kind
of
play
with
so
I
went
ahead
and
launched
a
five
note
cluster.
So
this
cluster
is
pre
running
with
committees
already
on
it.
If
we
go
ahead
and
we
go
to
deployments,
you
can
see
that
everything
is
all
0
it
out.
We
have
a
predefined
stack
already
here
that
I
deployed
via
the
catalog,
so
I'm
going
to
go
ahead
and
blow
that
away
real
quick.
B
So
let's
go
to
services
and
we're
just
going
to
blow
away.
Actually
I
should
probably
make
sure
that
deep
implication
controllers
are
actually
offline,
yeah,
otherwise,
they'll
bring
everything
right
back
up.
It's
like
you
can't
kill
anything
in
kubernetes,
yeah
that'd
be
enough
cool
all
right,
so
that's
all
gone
a
little
bit
back
over
to
pods
and
I'm
going
to
delete
these
pods.
Let's
get
rid
of
these
who
needs
WordPress
anyway,
get
rid
of
this
okay,
so
that
should
be
going
away.
B
B
B
Is
it
already
launched
my
sequel,
which
is
pending
and
our
web
front-ends
are
pending
and
the
host
IP
is
already
here
so
obviously,
if
I
had
any
type
of
DNS
hosted
with
ID
net
simple
or
something
I'll
go
ahead
and
add
this
to
my
DNS
and
the
site
will
be
life
so
we're
going
to
go
and
let
that
bake
for
a
little
bit,
let's
go
ahead
and
deploy
our
own
application.
So
I
have
a
violin
application
here.
B
So
if
I
go
into
configs,
I
have
a
service
and
a
deployment
so
we'll
go
ahead
and
deploy
this
out
create
up
Wi-Fi
come
on
Wi-Fi
there
we
go,
never
do
a
demo
on
hotel
Wi-Fi
alright.
So
if
we
go
back
to
deployments,
we
can
see
that
on
microwatts
here,
let's
go
ahead
and
see
what
IP
it
currently
has.
So
we're
gonna
go
back
over
here
to
services
and
we
can
see
that
the
my
MicroBot
is
running
on
this
port.
So
let's
go
into
it.
We're
go
ahead
and
do
a
downgrade.
B
B
B
There
you
go
so
now,
it's
back
to
v1
and
then,
if
we
want
to
go
ahead
and
bring
this
back
up
to
v2
again
without
we
can
always
do
it
via
like
cube
control,
but
I'm
just
going
to
go
ahead
and
show
you
through
the
UI
and
we're
gonna
go
ahead
and
bring
this
back
up
back
up
tight
up
to
the
v2
again
as
far
as
host
you
can
see.
I
have
five
hosts
running
here.
I
have
both
the
eliminated
stack
deployed
on
here,
as
well
as
my
application
and
as
well
as
WordPress.
B
So
if
I
go,
if
you
guys
want
to
see
like
the
WordPress
stack,
we
can
go
back
to
the
kitchen
controllers.
Oops
sorry
set.
Where
is
it
services
there?
We
go
and
I'm
going
to
click
on
this.
Here's
the
WordPress
site,
I'm
go
ahead
and
click
next
give
your
test.
Admin
will
get
and
copy
this
install,
WordPress
I
would
did
my
gmail
address.
B
Login
I've
been,
and
now
we
have
our
Harlem
follow
WordPress
site,
so
we
have
two
applications
now
deploy
to
our
equity's
cluster
via
both
the
catalog
and
be
like
you
control.
So
again
you
still
get
that
same
interaction,
but
by
using,
like
you
control,
you
know
the
really
change
like
change.
Like
your
workflow
again,
we
have
a
full
solid
pipeline,
so
we
use
Jenkins
for
almost
everything.
So
once
a
developer,
does
a
git
commit
to
either
get
lab
into
a
github
Deacon's
takes
over
compiles
those
those
containers
pushes
up
to
our
registry.
B
Then
it
goes
ahead
in
it
scans.
It
does
a
security
scan
of
those
container
images
like
images
to
make
sure
that
there's
no
lab
run
abilities,
because
if
there
is
it
stops
to
build
and
then
those
containers
get
deployed
into
Rancher
using
Rancher
les
CLI.
So
that's
our
full
pipeline
for
literally
every
single
application
that
we
deploy
in
here.
So
the
same
app
gets
treated
the
same
way.
Every
app
gets
scanned
in
the
exact
same
fashion
and
then
we're
actually
also
populating
some
of
our
applications
into
the
catalog.
B
So
if
a
developer
needs
access
to
go
ahead
and
deploy
a
dev
environment
or
a
borate,
let
predefined
elke
stack
into
an
environment
for
their
logs
than
good,
and
do
that
and
everything's
pretty
much
all
self-service.
So
from
operations
perspective,
we
are
technically
we're
technically
just
platform
maintainer,
so
we
just
maintain
the
platform
and
assist
in
onboarding
those
dev
teams
into
onboarding
their
applications
into
your
Rancher.
B
It
usually
all
starts,
can
get
so
we
literally
just
give
them
a
template
like
skeleton
of
what
they
need
via
a
git
repo
and
often
there
they
can
get
fully
started,
don't
even
even
need
to
access
Jenkins
to
get
added
into
the
pipeline.
We
use
like
jenkins
files,
docker
compose
file
and
a
rancher
compose
file,
and
they
can
literally
just
on
board
their
application
without
having
to
access
any
of
our
any
of
our
existing
tools.
So
it's
full
self-service
and
we've
been
able
to
move
pretty
fast
again.
B
We
have
60
apps
already
in
there
and
we've
been
I
would
say
any
new
site
already
new
application
that
that
comes
online
does
get
put
into
into
containers.
We
don't
hardly
use.
You
know
the
only
the
old
methods
of
deploying
apps
onto
on
TBMs.
We
still
have
a
number
of
applications
that
we're
trying
to
figure
out
what
to
do
with
so
again
we're
waiting
until
a
Microsoft
inductor
kind
of
solve
that
problem
for
us
and
then
hopefully
we
can
give
that
a
try
any
I
guess
they
need
questions.
B
A
B
So
I
didn't
show
you
guys
this
once
you
get
my
committees
deployed,
you
can
go
here
and
generate
the
config
for
your
local
laptop.
If
you
want
to
use
IQ
control
on
like
your
local
machine
or
use
it
in
Jenkins
or
you
can
just
go
ahead
and
use
this
and
go
ahead
and
run
the
same
commands
right
in
the
browser.
So.
C
A
Have
we
have
a
small
micro
service
called
WebSocket
proxy?
It
actually
authorizes
authorizes
your
request
and
then
just
send
them
to
the
to
the
API
server
when
once
authorized.
All
the
sir
is
Asian
is
down
on
the
ranger
side.
What
we
are
thinking
about,
integrating
with
a
kubernetes
native
authorization
components
by
implementing
the
external
provider.
C
A
B
B
Oh
so
so
like,
for
example,
so
if
you're
within
Amazon,
we
have
we
use,
we
use
a
the
Amazon
auto-scaling
groups
where
we
pass
in
a
user
data
script
into
the
node
and
it
automatically
bootstraps
it
based
on
a
predefined
image
that
we
have
that's
Disney
approved,
and
then
we
go
ahead
and
just
bootstrap
it
with
the
wrench
rage.
And
then
it
just
shows
up
within
rancher
for
that
like
for
that
particular
tenant.
B
If
it
is
not
within
Amazon,
then
we
look,
then
we
buy
utilize
a
terraform
for
that
flock
of
those
buildups
I
will
be
pushing
out
Mike.
My
terraform
config
out
to
my
github
to
allow
you
guys,
if
you
guys,
want
to
use
it
to
stand
up
Rancher
as
well
as
a
communities
within
within
my
digital
ocean
I'll.
Be
posting
that,
right
after
this
talk.
A
No
specific
any
any
image
that
any
format
that
community
supports
the
litigation
controller
llamo
service
llamo,
so
you
just
publish
it
to
the
catalog
and
you
can
actually
yeah.
If
you
click
on
this
add
catalog,
you
can
add
your
custom,
get
people
catalog
just
for
your
users
and
you
can
publish
applications
there.
We
also
accept
the
user
applications
to
our
community
catalog
and
in
the
future
we
are
planning
to
integrate
with
home.
C
A
Actually,
yeah,
that's
a
good
question,
so
we
started
with
maintain
our
own
Fork
of
kubernetes,
just
because
our
network
was
kind
of
embedded
and
hard-coded.
Then
at
some
point
we
moved
the
rancher
Network
into
the
CNI
plugin.
So
there
is
no
real
reason
for
us
to
maintain
the
fork
anymore
and
we
are
currently
in
a
in
a
process
of
up
streaming,
our
entry
cloud
provider.
So
it's
going
to
be
just
the
community's
official
version.
A
B
A
A
One
is
like
we
are
going
to
start
working
on
the
kubernetes,
the
de-rated
control
plane
and
integrating
with
our
UI
and
we're
going
to
be
going
to
make
our
dns
more
pluggable,
because
today,
Rancher
internal
DNS
is
the
default
choice,
you're
going
to
add
sky
DNS
as
one
more
plug
in
the
same
for
load
balancer
like
today,
when
you
configure
ingress,
we
create
a
red
shoulder
that
is
backed
up
by
HP
proxy,
that
transit
container
and
we're
going
to
add
more
providers
like
traffic
engine
acts.
It's
all
again,
pluggable
model.