►
Description
Demo of Kubernetes on Windows Server - Michael Michael, Apprenda
Windows Server, and .NET, currently host 30-50% of enterprise workloads in the Global 2000. For Kubernetes to provide a single distributed application fabric to all enterprise workloads, it must have full integration with Windows Server 2016 and Windows Server containers. In KubeCon London 2016 Apprenda announced that it would lead this development effort and teamed up with Red Hat to extend Kubernetes to the Microsoft ecosystem. In this session we will demo Kubernetes on Windows Server 2016 and discuss its features.
About Michael Michael
Michael Michael (or M2 as he's known) is Apprenda's Senior Director of Product Management. Michael also leads the Kubernetes SIG-Windows which is tasked to bring Windows Server support natively to Kubernetes. He is an experienced, detail-oriented software engineer with problem-solving and program-management skills. Prior to Apprenda, Michael spent 12 years at Microsoft as a Principal Program Manager working on Virtualization, Azure, and Windows Server technologies. Michael holds a Master in Computer Science from Stanford University, a Technology M.B.A., and a Bachelor of Science in Computer Science from the University of Texas at Austin. Many Apprenda employees believe he could one day play ping-pong professionally. Follow him on Twitter at @MichMike77.
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
All
right,
so
so,
let's
get
started.
My
name
is
Michael
Michael
I'm,
the
senior
director
of
products
at
aprender,
a
lot
of
my
my
role
at
aprenden
cludes,
the
product
direction.
A
lot
of
the
investments
were
making
in
our
platform
as
a
service
as
well
as
working
with
customers
and
partners.
At
the
same
time,
I'm
also
the
sig
windows
lead,
so
I
work
with
a
team
of
folks.
A
A
lot
of
them
are
here
today,
so
we
have
a
g2
Alex
and
a
couple
of
others
are
heavily
involved
in
the
process
of
bringing
Windows,
Server
containers
and
Windows
server
support
into
kubernetes.
So
our
goal
from
the
beginning
was
there's
an
ecosystem
of
applications
and
servers
out
there
that
as
well
as
folks,
companies
have
made
investments
into
Windows
and
they
want
to
see
those
investments
into
kubernetes
and
be
able
to
take
advantage
of
the
larger
ecosystem
and
all
the
efforts
are
going
around
in
kubernetes.
So
so
we
are.
A
Who
are
we
so
most
of
the
team
in
cig
windows
at
least
the
active
team
is
comprised
out
of
aprender
members.
We
are
a
company
that
created
a
platform
as
a
service
solution
that
makes
it
very
easy
for
developers
and
IT
operators
to
create
a
self-service
platform
as
a
service
solution
within
their
enterprise.
So
you
have
your
own
hardware
installer
software
there
and
you
get
an
ecosystem
of
apps
and
services
that
allow
you
to
power
line
of
business
applications.
A
As
a
company,
aprender
who've
been
part
of
the
kubernetes
community
through
an
acquisition
of
mate
of
the
company.
Cookies
Maddock
and
we've
been
heavily
involved
into
software
and
solutions
into
the
community
since
early
March
of
this
year.
So
there's
some
of
our
goals
as
a
SSE
Windows
team.
You
know
like
I
mentioned
earlier.
Our
goal
is
to
bring
Windows
server
and
windows,
server,
containers
or
hyper-v
containers
into
the
ecosystem
of
kubernetes.
A
That
will
allow
kubernetes
to
be
able
to
expand
the
types
of
applications
that
you
are
supported,
so
that
gives
you
a
heterogeneous
hybrid
environment
between
Windows
and
Linux,
java.net
and
so
on
and
so
forth.
That
allows
us
to
kind
of
leverage
the
investments
that
Microsoft
as
well
as
docker,
are
making
into
this
ecosystem
and,
as
a
byproduct
of
that
were
able
to
loop
back
to
them
and
drive
enhancements
into
Windows,
networking
and
kind
of
going
to
deep
dive
into
that
later
on.
A
If
we
look
at
Windows,
Server
is
an
operating
system,
there's
a
ton
of
Windows
server
out
there
in
the
enterprise
there's
37
percent
Windows
server
operating
systems
in
traditional
enterprises.
Today,
if
you
look
at
a
shore-
and
this
is
a
little
bit-
it's
it's
not
a
representative
of
true
enterprises,
but
in
Windows,
Azure
80%
of
the
workloads
are
on
Windows
server
and
there's
almost
12%
of
all
the
web
servers
out
there
they're
running
IIS.
So
this
tremendous
need
for
for
supporting
workloads
running
on
Windows
and
dotnet
in
general.
A
In
a
survey
by
by
Gartner
they
say
that
85%
of
the
enterprise
software
in
the
global
2000
IT
organizations
will
still
run
Microsoft
and
Java
software
through
2017.
So
dotnet
is
not
going
away
anytime
soon.
So
I
kind
of
give
you
guys
a
little
background.
Why
we're
doing
this-
and
you
know,
there's
a
to
need
here
for
for
Windows
I'm,
assuming
that
a
lot
of
you
guys
are
here
because
you
care
about
Windows.
So
let
me
get
started
how
many
of
you
guys
have
tried
windows,
server,
containers
or
hyper-v
containers.
A
So
far,
so
a
few
of
you
very
cool,
so
looking
at
the
docker
architecture
here,
a
couple
of
pieces
that
we
had
to
modify
to
to
make
it
work
for
kubernetes
we're
mostly
Center
on
network
as
well
as
compute,
but
the
one
thing
that
made
things
simpler
for
us
as
a
cig
windows
team
was
the
fact
that
Microsoft
made
docker
its
de
facto
standard,
an
API
for
utilizing
windows,
server
containers.
So
what
I
make?
What
that
made
possible
is
to
have
an
infrastructure
on
the
left
side
that
has
your
doctor
engine
daemon.
A
The
images,
as
well
as
the
containers
running
on
the
Linux
host
and
be
able
to
talk
to
the
doctor,
API
and
the
doctor
client,
and
have
a
a
very
equivalent
API
said
on
the
windows
side
with
a
very
similar
architecture.
So
you
have
your
docker
images
that
are
coming
from
docker
hub.
You
have
your
engine,
that's
being
able
to
that.
The
API
can
talk
to
and
you
have
one
by
one
all
the
different
containers
that
you
can
run
on
top
of
the
Windows
Server.
Now
there's
two
types
of
containers
in
Windows
Server.
A
Today,
one
of
it
is
Windows
Server
containers,
which
is
a
shared
kernel
mode,
where
all
the
containers
share
the
kernel
and
there's
a
little
bit
more
secure
solution.
That
gives
you
better
isolation,
but
they
reduce
latency
of
starting
up
your
containers
and
and
basically
also
more
resource
utilization,
and
that's
called
the
hyper-v
containers.
The
API
side
is
the
same.
It's
just
a
simple
flag
that
you
change
when
you're
starting
up
your
container
for
most
of
the
work
that
we're
doing
today
and
the
demo
we'll
be
showing
you
is
all
about
Windows
Server
containers.
A
Taking
a
brief
look
at
a
production
cluster
example
of
kubernetes,
you
have
your
master
components
like
your
API
server,
your
scheduler,
the
controller
manager
load
balancer,
the
left.
You
have
the
cube
CTL,
you
have
your
HDD
cluster
up
on
the
top,
and
then
you
have
your
working
nodes,
which
are
the
different
nodes
that
you
brought
into
kubernetes.
So
you
have
the
couplet
cube
proxy
and
one
by
one,
the
different
pods
with
the
containers
running
into
them.
A
So
when
we
started
architecting
our
solution,
one
of
our
main
goals
was
we
didn't
want
to
modify
anything
on
the
HDD
or
the
master
components
we
wanted
to
concentrate
all
our
work
in
12
traction
layer
into
the
couplet
and
the
cube
proxy.
So
that
makes
it
easier
for
us
to
kind
of
advance
our
strategy,
independently
of
the
master
components
that
was
similar
to
what
the
rocket
did
to
introduce
a
port
for
rocket
into
kubernetes
as
well.
A
So
looking
at
what
we
did
here,
you
have
the
cube,
CTL
doesn't
changed,
you
have
your
master
components
that
are
unchanged
and
then
on
the
Windows
Server
2016
node.
You
have
modified
the
couplet,
the
cube
proxy.
We
still
run
docker
because
that's
the
API
layer
for
Windows
Server
containers
and
then
you
run
one
or
more
containers
in
the
pod,
along
with
the
infrastructure
container
on
the
front-end.
The
cube
proxy
talks
to
the
external
networking
interfaces.
A
In
order
for
us
to
add
the
Windows
Server
node
to
kubernetes,
who
had
to
make
a
few
changes
to
support
the
Windows
server
OS,
this
was
the
first
time
that
the
new
type
of
operating
system
was
added
in
kubernetes.
So
there
were
a
lot
of
assumptions
in
the
code
that
assumed
you're
running
on
a
Linux
OS.
You
use
Linux,
primitives
and
other
things
that
you
had
to
go
in
and
find
them
one
by
one
was
more
of
a
trial
and
error
process
until
the
point
that
it
changed
all
of
them.
A
But
the
bulk
of
our
changes
and
most
of
the
complexity
was
around
networking
and
we're
going
to
touch
those
points
in
a
second.
However,
at
the
end,
when
we're
able
to
finish
this,
we're
able
to
schedule
Windows
Server
containers,
you
current
I
is
workloads
in
kubernetes.
Today
you
can
run
asp.net,
you
can
run
dotnet
core
components
actually
does
Annette
core
allows
you
to
develop
an
application
running
on
dotnet
core
and
run
it
either
on
Linux
or
Windows.
A
A
Containers
within
a
pod
get
to
talk
to
each
other
in
Linux.
They
kind
of
shared
the
same
networking
namespace.
So
that
means
that
the
entire
pod
has
a
single
IP
address
and
all
the
containers
within
that
pod
can
communicate
over
localhost
in
Windows.
The
same
abstraction
did
not
exist,
so
it's
not
possible
for
you
to
use
container
more
networking
in
Windows
each
container,
who
have
its
own
IP
address
and
there's
different
networking
modes
that
they
allow
you.
So
they
have
an
odd
mode
where
you
have
and
not
on
the
container
host.
A
And
then
all
the
containers
can
pick
up
an
IP
address
from
the
nut
range.
You
can
have
a
transparent
network
in
which
point
every
container
in
the
host
gets
its
own
IP
address,
and
it's
directly
attached
to
the
physical
interface,
the
physical
virtual
switch.
They
have
also
L
to
tunnel
and
L
to
bridge
and
a
few
other
technologies.
We've
concentrated
on
transparent
mode
where
every
container
has
its
own
IP
address
and
that's
what
we
build
our
solution
on.
But
that
means
that
every
container
Gettys
on
networking
stack.
A
So
you
no
longer
have
the
ability
to
put
many
containers
in
a
pot
and
all
of
them
to
share
the
networking
stack.
Every
one
of
them
has
an
individual
thing,
so
that
does
not
allow
us
to
chip
parity
with
the
Linux
pod
abstraction
we're
working
on
how
to
rectify
that
in
a
future
release
on
the
inter
pod
networking,
that's
the
ability
of
containers
and
parts
to
be
able
to
communicate
with
other
containers
and
pods
on
either
the
same
host
or
other
host.
A
There's
no
Sdn
plugins
in
Windows.
So
you
can
do
things
like
what
colleague
owed
us
today.
So
there's
no
overlay
networks,
at
least
not
natively
in
Windows
Server
today.
So,
in
our
case,
we
couldn't
do
it.
You
couldn't
get
given
a
single
IP
address
to
a
pod.
That's
rather
ball
across
the
entire
cluster
that
doesn't
exist.
Each
container
has
his
IP
address.
That's
rather
ball
across
the
cluster.
A
So
when
we
continuing
down
the
networking
path,
cube
proxy
had
a
lot
of
networking
challenges
for
us
as
well,
so
there's
no
IP
tables
on
Windows
Server,
so
the
same
functionality
that
you
get
on
Linux
by
from
a
single
solution,
IP
tables
is
scattered
through
many
different
variants
on
Windows
Server,
and
some
of
them
is
not
even
existent.
So
what
we've
done
is
with
using
SSH
port
proxy
to
to
do
a
lot
of
the
work
around
redirecting
and
then
adding
the
network
packets,
but
that
doesn't
support
UDP.
So
that
means
a
DNS.
A
If
you're,
trying
to
do
a
DNS
query
from
within
a
container
won't
be
able
to
forward
your
packets,
so
then
it
relies
on
the
DNS
client
to
be
able
to
D
query
over
TCP
a
second
time
in
order
to
make
all
of
that
work.
So
we're
working
with
Microsoft
to
kind
of
rectify
this
as
well
and
make
sure
that
UDP
is
supported
on
port
proxy.
A
Then
another
thing
that
came
out
was
that
we
had
to
implement
the
proxy
user
space,
so
that
made
it
a
little
bit
simpler
for
us,
but
it
was
more
intensive
from
a
CPU
perspective.
In
all
our
tests,
we
didn't
find
that
to
be
prohibitive
from
a
resource
consumption,
but
that's
just
something
that
we
have
to
worry
about
them
and
keep
an
eye
on
in
the
future
and
obviously
the
the
Windows
Firewall
and
Windows
behaves
a
lot
different.
How
IP
tables
were
behaving
in
Linux?
Can
I
mention
this
earlier.
A
So,
let's
dive
into
a
demo
here
so
I'm
going
to
show
you
guys
an
environment
that
has
two
windows:
server:
nodes,
the
running,
Windows
Server,
2016
rpm
and
one
Linux
node
that
will
contain
all
of
the
kubernetes
core
components.
That's
the
master
components,
the
DNS
IP
server
and
all
these
things
for
the
application
that
you're
going
to
use
we're
going
to
use
the
guestbook
go
application.
That's
an
app
that
ships
with
the
distribution
of
kubernetes.
So
you
can
run
this
up
on
Linux
or
on
Windows.
A
A
The
first
thing
I
want
to
show
you
guys
is
what
my
environment
looks
like
on
Azure,
so
I'm
doing
this
demo
entirely
on
Azure
components.
So
we
have
a
cluster
of
five
nodes,
we'll
use
three
of
them
today
for
the
demo.
One
of
them
is
the
cue
bling
zero
one.
That's
my
Linux
machine
and
I.
Have
my
two
and
I
have
for
Windows
machines
here,
running
different
versions
of
Windows
Server
2016,
some
pre
RTM,
and
some
RTM
and
I
have
five
Ethernet
adapters
well
more
than
five
here,
but
five
of
them
very
amuse.
A
So
for
each
Windows
node
you
have
one
for
bridge
and
one
for
the
regular
network
traffic
and
in
Windows
Azure
the
way
that
of
kind
of
deployed
this
entire
environment
using
the
I/o
resource
manager
that
allows
you
to
template
designer
environment,
so
it
took
that
template
created
it
and
whatever
to
replicate
this.
This
entire
installation-
and
one
of
the
things
I
want
to
mention
here,
is
that
you
know
the
azure
cube
net.
A
Is
the
virtual
network
that
kind
of
connects
all
these
devices
together
in
terms
of
the
network
interfaces
across
the
different
nodes
in
my
environment?
And
the
last
thing
is
in
Windows
Azure
a
lot
of
the
routing
table
that
you
need
to
make
all
the
magic
happen
and
have
routable
requests
from
a
networking
perspective
across
the
different
nodes?
You
have
to
define
that
manually
within
your
environment.
So
we
have
our
routes
explicitly
defined
here
in
Azure.
A
I've
connected
using
paddy
to
my
kubernetes
master
components
on
my
on
my
server.
So
let's
go
ahead
here
and
show
you
all
the
different
of
the
different
nodes
that
have
running
here
so
I
have
two
nodes
there,
one
2700,
the
one:
that's
my
Linux
node
running,
all
the
kubernetes
master
components
and
the
10.0.
The
1.15
is
one
of
my
windows
nodes.
So
I
went
ahead
and
added
one
of
my
windows
nodes
ahead
of
time.
A
Just
so,
I
can
have
that
up
and
running
and
then
we'll
go
ahead
and
add
another
node
right
now,
so
both
of
them
have
been
running
for
about
five
hours
or
so
so,
let's
go
ahead
and
show
you.
How
is
it
is
not
to
add
one
more
Windows
node
into
my
environment
when
I
use
my
scripts
here
to
copy
a
couple
of
things.
A
A
lot
of
the
steps
of
initiating
the
both
the
couplet
in
the
proxy
are
a
little
manual,
but
we're
working
on
making
all
of
these
things.
Windows
services
with
an
installation
package,
so
you
can
say
I
want
add
a
new
node
into
my
environment.
You
run
the
installation
package
will
deploy
the
couplet
the
proxy
set
up
your
your
network
and
get
it
all
started.
So
now
my
my
cubelet
already
when
I
had
them
started,
but
I
want
to
show
you
guys
here.
A
If
I
did
docker
network
less
here,
I
get
to
see
that
the
my
net
network
that
I
just
indicated
earlier.
It's
a
transparent
network
so
like
I
mentioned
docker
on
Windows,
has
many
networking
modes
of
chose
transparent.
That
allows
us
to
have
one
network
interface
binded
to
the
physical
network
interface
per
container.
A
Now
it's
time
to
start
the
proxy
as
well,
very
similar
to
what
I
did
for
the
couplet
I
need
to
set
up
an
environment
variable
indicating
which
is
the
network
interface
I
want
to
use,
and
then
I
start
the
proxy
pointing
to
the
IP
server
to
the
API
server,
as
well
as
to
the
indicating
what's
this
machine?
What's
its
name
like
P
address
and
all
these
other
details.
A
All
right,
so
our
proxy
now
is
up
and
running.
So
if
I
go
back
to
my
kubernetes
command
and
around
the
cube,
CTL
and
I
get
the
nodes.
I
see
another
node
here,
the
dot
one,
seven
node
also
I've
been
running
so
I
have
a
secondary,
Windows,
Server
node
in
my
environment.
So
if
I
were
to
describe
one
of
these
nodes.
A
A
Every
other
aspect
of
the
labels,
the
config
of
the
nodes,
is
the
same
except
the
fact
that
two
of
them
are
Windows
and
one
of
them
is
Linux.
One
experience
across
both
ecosystems
of
Windows
and
Linux
and
kubernetes.
Our
goal
was
not
to
differentiate
here
the
experience
we
wanted,
all
of
it
to
seem
seamless,
well
integrated
across
the
board.
A
So
let's
go
ahead
and
create
some
ups
and
see
how
how
easy
it
is
to
both
create
them,
as
well
as
scaling
them
in
this
environment.
So,
like
I
mentioned
earlier,
I'm
going
to
go
ahead
and
create
the
Redis
application.
So
what
I
did
here
is
I
pasted,
four
commands
that
will
create
the
master
control
and
the
service
for
Redis
and
the
slave
controller
and
the
slave
service
for
ready.
So
what
does
that
look
like?
So,
let's.
A
Let's
cut
this
command
over
here,
you're,
seeing
a
very
standard
interface
or
a
JSON
command
here
that
you're
familiar
with
kubernetes.
You
have
your
replication
controller.
It's
already
slave,
in
this
case,
I'm
asking
for
two
replicas
and
one
of
the
most
important
things
here
is
the
the
image
here.
It's
going
to
come
out
of
windows
server
core.
So
that's
the
image
they're
using
here
for
our
for
installation
and
how
do
you
tell
it
to
use
Windows
as
a
target
operating
system
versus
with
Linux
use
a
node,
selector
and
windows
tag?
A
So
just
I
showed
you
guys
earlier
in
the
nodes,
Windows
and
Linux
is
differentiated
by
the
labels.
I
can
use
those
labels
to
define
words.
Each
of
Monomoy
workloads
needs
to
land
in
this
case,
I
want
them
to
land
on
Windows.
So
now,
I
can
actually
concubine
80s
again
and
ask
you
to
show
me
all
the
different
pods
that
I
have
on
my
nodes.
A
One
readies
master,
as
well
as
a
blog
engine,
pod
I,
actually
went
ahead
and
deployed
blog
engine
ahead
of
this
presentation,
just
to
have
that
I've
been
running
and
have
it
already
so
now
all
my
parts
are
running
notice
here.
The
note
which
is
are
running
they're,
all
running
the
dot
15
or
the
dot
17.
So
all
my
containers
are
running
all
my
containers
and
all
my
pods
are
running
on
Windows
Server
right
now,
and
each
of
them
has
a
unique
IP
address
across
a
cluster.
A
So
if
I
want
to
check,
has
my
Redis
master
and
the
slave
synchronized
I
can
use
standard
kubernetes
commands
to
you
check
that,
and
it
looks
like
the
synchronization
with
a
slave
has
succeeded.
So
now,
both
my
Redis
master
and
my
ready
slay
were
able
to
find
each
other.
More
importantly,
one
when
I'm
going
to
show
you
guys
the
application
a
little
bit
guestbook
go
rights
to
master
and
reads
from
slave
and
all
of
the
different
components
of
my
app
the
different
part,
the
the
slave,
the
master,
as
well
as
the
guestbook.
A
A
You
notice
that
again,
the
targeting
is
on
windows,
so
this
application
will
also
run
on
my
Windows
server.
So,
let's,
let's
see
if
all
our
pods
are
up
and
running
now
and
we
have
the
ghost
application.
Sorry,
the
guestbook
application,
also
up
and
running
so
let's
go
ahead
and
describe
our
service,
and
so
we
can
get
the
service
IP
address
and
be
able
to
call
it
and
see
that
work.
A
Now
this
is
telling
me
that
my
guestbook
application
is
running
on
10.0
92
IP
address
and
it's
on
port
3000.
So
let's
copy
that
here
really
quickly
go
back
to
our
windows.
Node
have
a
browser
here,
I've
been
running
and
I'm
going
to
visit
that
up
type.
Oh
sorry
now
visit
that
IP
address
and
there
you
go.
I
have
a
guestbook
application
running
on
Windows
I
can
go
ahead
and
enter
data
here,
so
I'm
going
to
call
it
demo
cubic
on
1
&
2.
A
So
so
we
have
some
data
that
went
back
into
the
red
is
back-end
if
I
want
to
click
on
the
environment
variables.
You
notice
that
most
of
these
are
very
specific
to
my
guestbook
application,
but
they
are
based
on
an
OS
of
Windows
and
T.
So
this
is
running
truly
on
Windows
and
I
can
even
click
into
the
info
here
and
look
at
some
of
the
already
specific
data.
So
we
have
Redis.
We
have
the
guestbook
application,
always
running
on
Windows
Server
containers
on
top
of
Windows
server
with
the
docker
API.
A
All
of
that
called
from
the
coop
rocks
in
the
cubelet
and
giving
you
a
very
similar
experience,
a
very
familiar
experience
to
what
you
get
in
linux.
You
also
can
take
advantage
of
all
the
features
that
kubernetes
provides.
So
we
didn't
have
to
do
anything
special
to
be
able
to
use
a
replication
controller,
be
able
to
scale
do
high
availability
on
our
container.
So
let's
go
ahead
and
try
that
out
right
now,
so
I
can
come
here,
an
issue,
a
command
to
the
cluster,
to
increase
the
replica
count
for
my
guest
book
to
3.
A
So
if
I
were
to
call
the
pods
again,
you'll
notice,
now
that
we
have
two
containers
are
creating
for
guestbook
and
we
just
click
of
a
button.
I
was
able
to
increase
the
replica
set
for
for
a
Windows
Server
application
when
I
demonstrate
to
you
guys
now
that
the
routing
actually
does
work
on
Windows,
so
go
ahead
and
grab
I'm
going
to
describe
my
service
one
more
time.
Second-Eldest
IP
address.
A
And
it's
done
92
and
I'm
going
to
call
an
API
about
20
times
from
one
of
my
windows.
Server
nodes
that
essentially
what
this
API
will
do
is
it
will
make
persistent
IP
address
calls
into
the
guestbook
application,
retrieving
the
data
that
you
have
on
Redis
and,
at
the
same
time
also
telling,
with
IP
address
that
I
was
able
to
heat.
So
it
will
show
the
the
node
per
load
balancing
that
the
kubernetes
offers.
For
me.
A
So
it's
looping
through
all
the
different
IP
addresses
doing
the
load
balancing
and
if
I
want
to
go
ahead
and
unplug,
one
of
my
nodes
or
delete
the
part
I'll
be
able
to
kubernetes
will
detect
that
and
be
able
to
stand
up
a
new
one
and
redirect
my
pods
keeping
my
replica
count
toward
I
had
defined
so
coming
back
here.
If
I
were
to
call
the
pot
again,
the
IP
addresses
for
guestbook
are
the
same
ones
as
the
ones
I
showed
you
earlier.
A
Now
guestbook
is
a
cool
application.
It's
a
multi
theory
has
their
Eddie's
bag
and
it
has
the
go
front-end.
We
also
wanted
to
show
you
guys
what
would
it
take
to
take
a
traditional
dotnet
application
run
it
inside
a
Windows
Server
container,
bring
them
into
kubernetes.
So
what
we
did
is
we
went
to
the
github
page
for
blog
engine.
We
downloaded
the
binaries
and
you
build
a
container
out
of
it
all
a
single
and
closed
container
of
the
blog
engine
application
and
published
it
so
I
went
ahead
and
created
a
pod
for
it.
A
A
And
visit
this
I
get
to
see
a
traditional
done
application
running
on
kubernetes.
We
didn't
have
to
do
anything
extra.
We
took
a
vanilla,
dotnet
application.
We
put
it
inside
a
container
and
we'll
to
run
this
now.
Not
all
applications
can
run
in
a
container
containers
and
windows
support.
Only
nano
server
and
windows
server
core.
A
So
if
your
application
can
run
on
either
of
those
two
OSS
and
one
of
them
has
a
limitation
on
what
dotnet
you
can
use,
so
nano
server
can
only
run
net
core,
but
as
long
as
you
can
live
with
those
limitations-
and
you
can
put
your
application
on
any
of
these
todo
SS-
you
can
package
it
in
a
container.
You
can
run
it
in
kubernetes
using
Windows
Server
containers.
So
this
is
a
fully
functional
lab.
I
can
go
ahead
and
log
in
here
I
can
see
the
admin
interface
I
can
do
Serge.
A
So
the
question
was:
how
about
persistent
storage,
the
volumes
and
storage
on
kubernetes
were
actually
working
on
that
some
of
the
functionality
is
is
going
to
be
available
for
you
when
we
go
to
alpha
with
version
1.5,
but
you
know
secrets
as
well
will
be
available
and
we're
also
going
to
work
on
getting
config
maps.
I've
been
running
for
for
Windows,
Server
containers.
A
So
the
question
was:
if
you
run
hyper-v
containers,
can
you
run
Linux
on
them?
Hyper-V
containers
still
have
the
same
limitation
in
the
West
that
it
has
to
be
the
nano
server
or
Windows
server
core.
So
you
cannot
run
Linux
containers
in
that
you
can
use
a
hyper-v
virtual
machine
if
you
wanna
run
a
Linux
OS,
but
then
it
would
be
a
virtual
machine.
You
would
not
be
a
container
yes.
A
So
we'll
I'll
touch
on
that
in
a
little
bit.
The
question
was:
if
DNS,
who
work
in
this
deployment,
so
DNS
from
inside
the
container
will
work.
If
your
DNS
client
is
a
little
bit
resilient
and
be
able
to
fail,
the
fails
request
will
come
over
UDP
and
that
will
fail.
If
you
reissue,
the
request
will
go
over
TCP
and
I
will
succeed.
However,
DNS
search
does
not
work.
So
when
you're
making
DNS
queries,
you
have
to
use
a
full
fqdn
of
the
resource
that
you're
trying
to
discover.
A
So
today
we
run
a
couple
of
containers.
We
run
the
infra
container
as
well
as
the
container
of
your
choice
in
the
pod
there.
It's
not
advised
to
do
that
today.
Just
because
a
lot
of
the
same
function
ADA
exists
in
Linux.
It
does
not
exist
in
Windows,
so
we
can't
share
the
stories
you
can
share
the
network
namespace.
Our
goal
is
to
get
to
the
point
that
you
can
give
you
a
full
solution
that
allows
you
to
run
multiple
containers
in
a
pod.
A
A
A
We
can
promise
that
we'll
be
better
than
because
a
lot
of
these
things
are
requiring
changes
into
Windows
from
a
networking
side,
as
well
as
certain
kind
of
pilot
projects
that
were
working
on
to
try
some
alternatives
to
supporting
that.
If
everything
goes
well,
we'll
definitely
be
in
better
with
1.6,
but
we
come
from.
Is
that
we'll
probably
have
more
clarity
within
the
next
month
or
so?
So
our
goal
is
releasing
this.
We
want
to
5.
A
The
better
release
will
include
the
love
the
networking
enhancements
that
were
working
on
as
well
as
monitoring
setup
and
configuration
so
make
it
easy
for
you
guys
to
deploy
that
get
a
brand
new
node
running
Windows
Server
get
docker
up
and
running
installer
then
couplet
as
well
as
aq
proxy
and
be
able
to
register
it
back
to
point
to
the
to
the
kubernetes
cluster
at
the
same
time,
or
heavily
working
with
the
windows,
networking
team
to
add
support
for
native
overlays.
So
we
can
do
as
the
end
type
solutions.
A
You
guys
are
really
interesting
about
the
call
to
action
come
and
get
involved.
We'll
have
a
slack
channel,
that's
fairly
active,
mostly
by
aprender
folks
right
now,
but
we
are
welcoming
anybody
else
to
come
and
join
us
and
below.
That's
that's
our
github
page.
That
includes
a
lot
of
the
code
that
we're
working
on
and
actually
both
of
our
PRS
got
merged
into
1.5
late.
Yesterday,
good.
A
So
the
actually
been
asked
that
question
a
few
times
so
far.
The
question
is:
is
the
kubernetes
master
running
on
Windows
in
the
roadmap?
There's
a
couple
of
pluses
and
minuses
into
that
that
the
positive
is
that
most
of
the
master
components
are
written
in
go
so,
ideally
they
could
be
portable
into
Windows.
However,
we
are
not
as
a
cig.
Investing
in
that
effort
for
a
couple
of
reasons.
The
first
one
is.
A
It
would
require
massive
undertaking
to
port
some
of
the
assumptions
that
the
code
is
making
that
is
running
on
a
Linux
operating
system,
so
kind
of
changing
all
those
assumptions
like
we
had
to
do
it
for
the
couplet.
It
will
be
a
massive
effort
and
the
second
thing
is,
you
have
to
test
all
of
that
so
right
now,
it's
working
very
well
on
Linux
I.
Think
that
keeping
the
hybrid
solution
where
you
have
Linux
and
windows
have
been
running.
A
A
So,
where
we
are
involved
in
terms
of
working
with
them
on
a
week-by-week
basis,
who
have
a
lot
of
conference
calls
with
them,
there
were
brainstorming
around
solutions
specifically
around
networking,
so
we
are
getting
support
from
Microsoft.
We
do
want
more
Microsoft
folks
to
come
and
join
the
SIG's,
so
they
are
actively
involved
in
the
day-to-day
operations.
I'm
hoping
we'll
get
there.
Maybe
this
one
releasing
in
alpha
1.5
is
exactly
what
Microsoft
needs
to
to
jump
on
the
ship
and
be
more
active
in
the
sea.