►
From YouTube: Tutorial: Deploying Windows Apps with Draft, Helm and Kubernetes - Patrick Lang & Jessica Deen
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
Tutorial: Deploying Windows Apps with Draft, Helm and Kubernetes - Patrick Lang & Jessica Deen, Microsoft (Limited Seating Available - See Description for Details)
In this session, we'll go through two key workflows:1) How to deploy and manage a Windows-based application on Kubernetes We'll start with an already-built Windows application in a container, and walk through a simple deployment on Kubernetes. Along the way, we'll go over some of the key considerations such as handling secrets, required node selectors, and other common differences you may encounter in clusters with Windows & Linux nodes. This section does not require access to a Windows machine, only a browser, ssh & kubectl will be required.2) Building an end-to-end workflow with Draft & HelmOnce you understand how to deploy a Windows application with Kubernetes, we can move on to setting up the full developer workflow. We'll create a new app using .Net Core on Windows, scaffold and deploy with Draft. This requires Windows 10 (laptop or VM running on Azure or VMWare Fusion)
To Learn More: https://sched.co/GrRU
A
A
So
if
you
follow
me
online,
I'll,
probably
post
something
along
those
lines,
no
relation
to
James
Dean,
that's
a
big
one,
because
for
30
years
of
my
life,
people
have
consistently
misspelled
my
last
name
with
EA,
and
if
you
want
to
find
anything,
especially
the
resources
for
today's
session,
you're
gonna
need
to
make
sure
you
spell
it
with
double
e's.
So
two
E's
in
my
name
and
can
find
me
online
at
twitter,
insta,
instagram
and
github.
B
B
But
you
know
outside
of
work,
love,
music
travel,
but
especially
like
and
stuff
I
make
coffee.
You
know
what
brew
that
from
the
beans
up
beer
wine
me,
like
you,
know:
woodworking,
it's
just
I
like
I,
like
building
stuff,
and
so
anyway,
I
like
sharing
that
I'm,
not
on
Twitter.
There
someone's
been
squatting
on
my
name
with
a
bot
for
a
long
time
and
I
can't
get
them
to
give
me
that
address
back.
B
A
Before
we
begin
just
a
brief
disclaimer,
this
is
obviously
tutorial.
We
were
actually
emailed
about
three
weeks
ago
saying
that
originally
it
was
expected
that
we
were
potentially
gonna
be
over
capacity,
so
this
was
intentionally
designed
to
be
a
more
conducive
learning
environment,
which
is
why,
obviously,
it's
in
a
seated
room
as
such.
A
lot
of
the
content
not
only
is
available
in
a
lab
space
format
here,
but
we
have
also
replicated
it
and
it
will
be
going
live
during
the
session.
A
Actually,
on
my
blog,
it
was
intentionally
designed
to
get
you
thinking
and
to
show
you
where
we're
at
with
kubernetes
with
windows,
it's
kind
of
just
to
give
you
a
bird's
eye
view,
but
also
more
of
like
a
hands-on
experience
for
it.
A
lot
of
the
stuff
that
we're
doing
is
new.
It's
still
in
development,
including
the
labs
that
we're
using
the
lab
viens
that
we
do
have
on
hold
are
technically
using
a
preview
service
which
is
called
as
your
labs
and
we're
going
to
put
them
to
the
test
today.
A
So
what
could
possibly
go
wrong
with
80
people
in
the
room?
We're
also
going
to
test
the
Wi-Fi
network
of
today's
conference.
Now
again,
all
resources
will
be
available,
including
a
blog
post
that,
like
I
said,
will
go
live.
We
will
add
the
video
of
this
recording
to
that
blog
post
after
that
video
is
available.
So
for
whatever
reason
you
can't
finish
the
lab
today.
Don't
worry
next
slide.
All
right
now
we're
gonna
level
set.
A
This
is
normally
the
slide
and
I
think
everyone's
deck
or
timing
of
their
presentation,
or
they
would
start
using
more
buzzwords
like
digital
transformation,
digital
disruption,
transformation
technology.
This
is
why
I
use
something
Star
Wars,
because,
like
I
mentioned,
I,
do
love,
Disney,
Star,
Wars,
so
of
course,
something-something
api's,
something-something
transformation.
The
ultimate
idea
is
to
talk
about.
We
already
know
that
our
industry
is
changing
technology's
changing.
A
So
the
big
demand
from
the
window
side
of
things
is
how
can
I
keep
up
with
this
change
in
the
same
way
that
Linux
containers
and
orchestrators
are
going
so
one
of
my
favorite
quotes
is
Socrates
and
it's.
The
secret
of
change
is
to
focus
all
of
your
energy,
not
on
fighting
the
old
but
building
the
new.
A
So
we
know
that
we
have
this
newer
technology,
but
rather
than
fighting
the
old
one
of
the
things
that
Patrick
and
his
team
has
been
working
on
is
finding
a
way
to
still
make
that
old
work
with
the
new.
So
the
content
we
put
together
today
is
Windows
containers,
hybrid
scenario,
so
you
have
Windows
nodes
and
Linux
nodes
side
by
side,
a
Windows
ASP
lift
and
shift
maybe
more
traditional
application,
relying
on
maybe
a
Linux
database.
A
This
is
content
that
I
have
been
asked
about
personally
several
times
over
the
last
year,
I
was
asked
about
it
last
year
at
cube,
con
I'm
sure
everyone
in
the
room
is
curious
about
that.
That
is
one
of
the
demos
is
to
take
not
fighting
the
old
but
bringing
it
in
with
the
new
stuff
today
and
then,
of
course,
the
more
new
modern
scenario
is,
if
you
have
refactored
and
you
are
using-
maybe
dotnet
core,
but
still
windows-based-
how
you
can
have
that
still
side
by
side,
your
Linux
applications.
B
All
right,
so,
let's
talk
a
little
bit
about
how
we
got
here.
So
if
I'm
mateus
is
incorrect,
Michel,
please
let
me
know,
because
he's
even
year,
long
and
I
have
so
containers
on
Windows
started
during
the
halfway
through
the
Windows
Server
2016
development
cycle.
So
server
was
in
a
really
long
release
cycle
and
this
big
disruptive
technology
picked
up
called
docker,
and
so
we
looked
at
it.
B
Is
it
really,
you
know,
creates
a
space
where
each
application,
you
know
effectively
thinks
it
has
its
own
Shayne,
and
so
we've
had
the
same
capabilities,
mostly
there
in
Windows,
and
so
we
actually
worked
early
on
to
make
containers
work
with
docker
on
Windows
Server
2016,
so
that
was
available
in
the
technical
previews
and
shortly
after
Windows,
Server,
20,
2016
shipped.
Of
course,
docker
had
a
commercial
offering.
While
this
was
in
development,
a
number
of
other
companies
got
together
along
with
this
kubernetes
community
and
created
cig
windows.
B
So
you
know,
Michael
had
a
prenda
as
well
as
some
folks
from
cloud-based
Ticketmaster
docker
in
well.
We
all
looked
at
this
early
and
said
you
know
what
we
think
we
want
to
have
one
platform
for
Windows
and
Linux,
and
so
they
started
the
ball
rolling
there
and
so
over
the
next
couple
of
years.
What
we
did
was
we
had
engineers
on
the
windows
team
working
with
the
people,
work
that
were
working
on
crew,
Bernays
and
said.
Well,
what
do
we
need
to
do
to
make
Windows
on
kubernetes
really
viable?
B
This
meant
we
needed
to
get
things
like
Network
namespaces
working.
We
need
to
make
improvements
to
the
storage
system
and
then
we
needed
to
start
really
building
partnerships
in
order
to
make
this
work,
and
so
we've
got
teams
that
are
working
on
thing
on
the
cni
plug-ins
for
things
like
calico
with
two
gerra
flannel,
and
then
we've
also
got
some
open-source
plugins
available
so
like
over
time,
we've
been
gradually
innovating,
both
the
close
source
code,
that's
in
Windows,
as
well
as
the
open
source
code
in
parallel.
B
So
we
can
make
all
this
possible-
and
this
is
all
really
building
up
to
you
know
the
Windows
Server
2000
19
release
that
just
went
out
a
couple
weeks
ago,
and
so
also
during
this
time,
as
more
and
more
stuff
has
come.
Online
we've
had
a
lot
of
vendors
pre-announce
and
even
start
previews
based
on
kubernetes
technology.
So
this
is
something
that
is,
you
know,
it's
super
active
development
and
we're
working
hard
to
get
the
last
test
cases
and
things
stabilized
so
that
it's
seen
as
a
stable
feature
within
the
kubernetes
community.
B
So
that's
what
you
know,
what
our
main
focuses
on
cig
windows
right
now
and
we'd
love
to
have
more
views
there
and
more
people
helpin
out,
so
that
we
can
make
sure
that
you're
that
we're
making
everything
work.
The
way
you
want
it
to
work.
So
so,
how?
How
are
we
going
about
doing
this?
So
when
we've
got
a
kubernetes
cluster
today?
B
Typically
we're
running
in
view
of
the
controller
manager,
scheduler
API
server,
all
the
stuff
you're
familiar
with
this
runs
on
a
couple
Linux
nodes,
and
then
you've
got
a
couple
of
Linux
agents
attached
to
that.
What
we're
doing
in
cig
Windows
is
no
different,
except
for
we're
just
simply
adding
Windows
nodes
into
that
same
mix.
So
you
could
start
with
a
cluster
where
you
just
have
entirely
Linux
nodes,
and
then,
when
you
want
that
Windows
capability,
you
just
take
a
Windows
Server
2000
19
machine
install
the
cubelet
install
the
cube
proxy.
B
Your
C&I
plugin
configure
it
the
same
way
where
you
in
a
linux,
node
and
now
you've
got
a
cluster.
That's
got
the
capability
of
being
able
to
manage
both
both
Windows
and
Linux
nodes
side
by
side.
The
kubernetes
api
is
the
same.
There's
just
a
few
differences
in
terms
of
how
you
actually
configure
it,
because
you
know
you're
running
a
Windows
app,
not
a
Linux,
app
slashes
are
facing
the
wrong
direction
and
there's
a
few
more
things
like
that
that
we'll
get
to
over
the
course
of
the
demos
but
you're.
B
The
key
thing
here
is
this
is
really
about
building
consistency
and
making
sure
that
we
can
get
the
same
code
able
to
run
in
both
places.
We're
not
asking
people
to
change
everything
to
support
Windows.
What
we're
doing
is
we're
making
Windows
work
in
the
deployment
you
already
have,
and
so
once
this
is
actually
up
and
running,
it's
just
kubernetes
as
usual.
If
I,
if
you
go
and
run,
you
know,
your
typical
commands,
like
Cube
control,
get
node
I
chop
this
down
to
fit
the
slide.
B
But
you
know
the
OS
image
and
version
number
are
listed
there
just
like
anything
else,
and
the
windows
nodes
are
just
additional
agent
notes,
not
no
big,
surprise
there,
except
for
when
I
took
a
screenshot
I
was
running
alpha.
Don't
do
that
and
then,
when
I
go
and
look
at
the
rest
of
the
pods
are
running
in
cube
system.
B
Those
are
all
still
you
know,
continuing
to
run
the
same
way.
They
were
on.
The
next
note,
unfortunately
left
that
column
off
there,
but
because
I
was
running
an
alpha.
I
got
this
nice
added
restart
here,
which
is
fun,
but
hopefully
that
won't
affect
things
later,
and
so,
when
you
go
to
actually
deploy
things,
of
course,
you
can
just
create
deployments
that
are
going
to
contain
pods
on
you
know
different
nodes.
You
can
use
node,
selectors
and
they'll
all
just
land
on
the
right
place
and
run.
B
So
this
is
a
sort
of
a
screenshot
of
what's
going
to
come
later,
when
we're
actually
running
workloads
that
are
spanning
both
sequel
on
Linux
nodes,
as
well
as
some
asp.net
framework
apps
running
on
the
windows
nodes,
so
nope,
no
big
surprises
there
from
it
for
an
admin
experience
and
so
now's.
The
part
where,
if
you've
got
a
laptop
with
you,
I've
got
80
VMs
that
are
already
set
up
and
deployed.
So
you
can
get
your
hands
on
this.
B
C
B
C
C
B
Does
everybody
everybody
got
that
good,
okay,
so
over
on
this
page?
What
we
have
is
the
first
thing
is:
if
you
don't
have
a
Windows
remote
desktop
client
I've
got
a
link
to
one
that'll
work
there.
The
next
thing
that
you
need
to
do
first
is
go
ahead
and
click
this
link
here
I'm
going
to
copy
that,
because
I
need
to
open
it
in
private.
B
But
when
you
click
this
other
link
here
for
the
lab.
What's
going
to
do
is
take
you
to
a
registration
page
and
what
that'll
do
is
you
can
use
any
email
address?
You
want
I'm,
just
just
a
warning
if
I
scroll
through
the
list
of
running
VMs,
the
if
you
user,
name
and
email
you
put
in
there
may
be
on
the
recording.
B
So
if
you
don't
want
it
on,
there
then
use
a
different
one,
but
if
you
get
disconnected
you
can
just
keep
refreshing
this
site
and
then,
when
you
hit
connect
here,
it's
actually
going
to
go
ahead
and
connect
to
your
lab
VM.
So
I'll
go
ahead
and
do
that
so
go
ahead
and
and
get
that
started
and
then
once
you
go
to
actually
login
I've
got.
You
know
a
nice
super
secure
password
here,
seven
handled
wheel
with
a
bang.
At
the
end,
it's.
B
B
B
B
So
once
you
connect
the
next
dangerous
step
that
you
need
to
do
is
I've
got
a
big
cluster.
That's
already
provisioned
and
I
wrote
a
you
know:
super
secure
non
non
HTTP
web
service.
That's
going
to
give
you
a
cube
config.
So
if
you
run
this,
which
I've
already
run
at
once,
what
it's
going
to
do
is
actually
give
you
a
cube,
config
with
a
random
namespace.
So
that
way
you
can
go
ahead
and
access
this
cluster
that
has
both
the
Windows
and
Linux
nodes
running
on
it.
B
Okay,
I'm,
apparently
wad
user
use
the
nice
dictionary
that
was
there
under
on
Ubuntu
and
pick
random
words,
and
so
that's
what
that's
what
you
get
so
I've
got
the
namespace
here
and
then
the
username
which
you're
going
to
need
for
for
tiller
a
bit
later.
So
now,
I've
got
my
cube
config
here
and
that's
all
good
to
go
so
it's
time
to
go
ahead
and
start
deploying
it
an
app.
B
C
C
C
B
B
A
B
B
Getting
the
cube
config,
the
other
thing
that
you
need
to
do
at
the
same
point
is
also
set
your
registry.
That's
going
to
be
used
in
the
draft
lab,
and
so
you
can
just
copy
and
paste
that
command
here
and
now.
I've
got
a
registry.
That's
going
to
be
there
when
I
use
draft
to
scaffold
and
deploy
an
app
later
and
then
I
already
run
the
docker
login
step
on
these
VM.
So
they
should
be
good
to
go.
B
B
Be
running
home
a
bit
later,
so
now
that's
ready
and
then
I've
got
a
link
to
the
next
section
here,
which
is
the
kubernetes
for
windows,
tutorial,
get
that
open,
and
so
the
dispersed
github
repo
that
we
went
through.
That's
just
a
set
of
steps
for
today.
You
shouldn't
need
to
refer
to
that
afterwards,
because
they're
all
going
to
be
dead
links,
I'm
gonna,
shut
those
clusters
down
and
throw
away
because
they're
they're
not
secured,
but
this
repo
has
step-by-step
walkthroughs
that
you
can
use.
B
You
know
on
your
own
VMs
and
everything
that
I
installed
on
those
VMs
is
all
automated
here
in
this
same
repo.
So
if
you
want
to
setup
at
your
own
lab
for
experimenting
or
doing
classes
all
the
resources
you
need
are
here
and
hopefully
I'll
be
able
to
put
into
a
blog
post
later
all
right.
So
let's
go
ahead
and
get
into
the
the
first
deployment
that
I'm
gonna
go
at
that
we'll
be
doing.
B
Okay,
so
the
first
deployment
we're
going
to
do
is
going
to
be
running
an
asp.net
4
app.
This
is
a
fabric,
ham,
fiber,
it's
something
that
was
designed
for
teaching,
dotnet
classes,
I,
don't
know
5
10
years
ago,
I'm,
not
sure
exactly
when
it
started,
but
I
like
using
it,
because,
it's
sort
of
you
know
a
reference
app
where
you've
got.
B
This
is
going
to
be
using
sequel,
server,
developer
edition
on
Linux,
and
so,
if
you
might,
if
you've
got
an
existing
app
that
you're
still
you
know
maintaining
this
is
this
is
a
great
way
to
run
it,
because
now
you
can
just
use
a
containerized
sequel
server
load
in
a
real
database
with
real
data
and
then
use
that
as
part
of
your
test
passes.
So
every
time
you
want
to
test
a
change
or
you
could
test
it
with.
You
know
empty
data.
You
can
load
old
data.
You
could
test
your
sequel.
B
B
B
Okay,
so
I'm
going
to
clone
that
repo,
and
so
I've
got
all
the
source
code
here,
but
before
I
actually
do
the
deployments
I
want
to
show
what
I
had
to
change
to
make
this
work.
So
your
typical
asp.net
apps
are
almost
always
going
to
have
a
web
doc
config
file.
That
actually
includes
things
like
your
database
connection
strings,
and
so
what
I
added
here
is.
B
We've
got
a
new
new
dotnet
library
that
you
can
they
it's
a
new
NuGet
package
that
you
can
use
called
configuration
builder
and
the
one
I'm
going
to
use
for
this.
One
is
the
environment
builder,
so
this
is
all
using
dotnet
dependency
injection.
You
don't
have
to
rewrite
your
app
to
be
able
to
use
this.
You
just
have
to
reference
it
in
the
project,
build
it
and
then
you
can
go
ahead
and
start
using
that
feature.
B
B
The
data
source
here
with
a
couple
references
that
are
going
to
be
filled
in
for
environment
variables,
so
that
way,
I
can
get
my
service
host
my
service
support
and
then
scroll
right
here
a
bit
and
then
you
know
also
get
the
password
that
are
in
there
as
well.
So
the
service
addresses
will
be
filled
in
based
off
of
the
kubernetes,
that
is
referencing
the
sequel
server,
but
the
password
is
going
to
come
from
a
kubernetes
secret.
So
just
like
you
can,
with
linux
pods
you
can.
B
You
can
use
a
combination
of
secrets
plus
database
plus
service
references
so
that
dns
lookups
and
all
that
stuff.
You
know
just
work
so
I've
already
built
this
in
a
container
and
pushed
it
on
to
docker
hub
so,
rather
than
going
through
the
whole
build
process.
I'm
just
going
to
skip
straight
to
the
deployment
here
so
over
here
I've
got
the
number
of
different
deployment
files
and
so
I'll
just
go
through
them
in
order
and
deploy
them
here.
B
Okay,
just
keep
keep
moving
forward,
so
cute
control
create
so
next
I'm
going
to
do
is
go
ahead
and
deploy
the
sequel
server
Linux
instance,
and
so
this
is
going
to
be
referencing
that
secret
that
I
already
created
and
the
image
on
this
one
is
the
existing
one.
That's
there
for
the
Microsoft
sequel
server
and
because
I
haven't
added
a
separate
product
key
on
it.
It's
automatically
and
default
to
the
developer
edition,
and
then
we've
got.
You
know
the
same
resource
constraints
that
they
recommend
using
either
as
a
minimum
and
then
what's
really
important.
B
B
B
B
B
Because
when
I
go
to
create
the
sorry,
it's
not
on
this
file
in
draft,
you
have
to
specify
the
name
if
space
is
going
to
deploy
too.
But
here
it's
just
going
to
use
my
current
one.
So
this
one
I'll
just
call
it
FFF,
demo
online
and
so
I'm
using
an
ingress
router
so
that
we
don't
have
to
wait
for
the
time
it
takes
to
get
each
IP.
Sometimes
that
happens
a
few
seconds.
Sometimes
it
takes
a
few
minutes
just
depending
on
which
data
center
you're
running
in
and
and
that.
C
B
The
one
thing
I've
seen
in
this
particular
claw
there
we
go
in
this
particular
setup
is
there's
something
in
this
one
where
sometimes
it
it's
running
on
the
node,
but
it
takes
kubernetes
like
an
extra
30
or
45
seconds
to
start
up.
I
think
that's
something
that
was
fixed
after
the
build
that
I
deployed
here,
because
if
I
do
cube
control
cluster
info,
this
is
actually
still
running
on
one
of
the
alphas.
Oh
wow
I
have
to
go
to
another
her
strict
of
this
user
from
being
able
to
run
that.
B
B
But
if
I
do
a
cubes
control,
log
I
feel
look
at
the
database.
We
can
actually
see
that
it
is
trying
to
connection
right
now.
The
first
login
fails
because
it's
looking
for
a
database
that
doesn't
yet
exist
this
apps
using
nmt
frameworks.
So
it's
just
going
to
fix
that.
So
the
next
time
I
go
and
look
at
the
logs,
we
can
see
that
it
went
ahead
and
started
up
the
database.
It's
running
so
at
this
point
the
database
is
running.
The
app
should
be
already
running
and
yeah
there.
B
It
is,
and
so
now
I've
got
this
application
using
dotnet
framework
for
on
Windows
Server,
2000
19
running
in
kubernetes,
with
the
sequel,
server
back-end,
and
so
this
is
a
pretty
common
pattern
that
I
see
a
lot
of
customers
doing
because
they
want
to
shut
down
these
VMs
that
are
each.
You
know
running
just
a
couple.
Apps
like
this
and
just
move
in,
you
know
within
their
within
their
kubernetes
cluster
with
Windows
nodes.
So
that
way
they
can,
you
know,
consolidate
them
and
make
them
a
lot
easier
to
manage.
B
So
that's
an
example
of
one
of
the
first
people.
We
call
lift
and
shift
workflow,
so
the
next
step
from
there
would
be
if
I
were
to
actually
run
this
fully
in
the
cloud.
Then
the
next
step
is
rather
than
actually
running,
sequel
server
in
a
container
without
persistence.
What
I
recommend
doing
is
actually
go
ahead
and
use
open
service
broker,
because
then
you
can
just
request
a
database
from
the
cloud
that
you're
running
on.
B
B
So
the
bottom
line
shows
that
horizontal
pod
scalars
is
is
forbidden,
so
what
happens
is
I
created
a
whole
bunch
of
namespaces
and
each
one
actually
has
its
own
its
own.
Our
back
policy
applied
to
it
and
so
I
left
all
the
core.
Api
is
in
there,
but
I
forgot
to
add
into
API
group
auto-scaling,
and
so
this
is
actually
our
back
doing
what
it
should
I'm
not
doing
a
scaling
demo
today.
B
B
A
So
this
is
also
a
good
example.
Remember
that
Windows
and
where
we're
at
with
Windows
and
kubernetes
is
still
in
development
like
a
lot
of
this,
isn't
something
that
would
be
recommended
for
production
use.
But
now
you
can
get
kind
of
hands-on
and
experience
with
it.
But
when
there's
new
releases
coming
quite
frequently,
issues
can
happen.
Yep.
B
B
B
You
know
right
now
we're
running
everything
on
the
Linux
VM,
but
you
could
take
your
cube.
Config
drop
it
on
a
Linux.
You
know
FreeBSD,
whatever
anything,
that's
got
a.
You
know
the
cute
kubernetes
clients
is
just
going
to
work
on
it.
It's
not
going
to
change
anything
there
and
you
know
in
terms
of
what
we're
installing
on
the
machines
it's
pretty
consistent
with.
What's
there
on
Linux,
you
ever
run
the
cubelet
cube
proxy
CRI
plug-in
to
see
and
I
plug
in,
and
then
the
supporting
infrastructure,
like
the
container
registry
that
I
have
set
up.
B
You
know
it's
exactly
the
same
thing
that
works
on
both
Windows
and
Linux,
first
or
in
the
containers,
so
what's
different.
They
need
to
consider
so.
The
first
thing
is
that
when
you're
actually
deploying
the
windows
applications,
we
need
to
make
sure
that
they
were
actually
run
on
a
Windows
node
and
so
the
that
that's
where
the
node
selectors
come
into
place.
B
B
So
if
Windows
is
going
to
be
too
minority
of
your
deployment,
then
that's
probably
going
to
make
sense
to
go
ahead
and
put
the
taint
on
the
nodes
and
then,
when
you
deploy
the
Windows
apps,
you
just
put
a
node
selector
and
the
Toleration
on
it.
So
in
your
Windows,
apps
will
run
on
Windows
nodes
and
everything
else
will
just
default
to
Linux.
So
that'd
be
an
easy
way
that
you
can
do
it
across
the
board,
and
then
that
gives
you.
B
You
know
time
to
go
back
and
update
your
helmet
arts
as
needed
to
have
the
node
selectors
in
there.
The
other
thing
that
I
want
to
call
out
is
that
the
resource
consumption
for
Windows
containers
is
higher.
That's
because
Windows
every
time
you
start
a
Windows
container,
a
bunch
of
background
services,
they're
needed
to
do
things
like
you
know,
run
the
users
security
database
be
able
to
handle
background
processes
like
the
web
servers,
all
those
need
to
be
replicated
for
each
container.
B
B
B
B
B
B
Yes,
so
what's
elevated
in
the
container
has
no
privileges
on
the
host,
it's
a
completely
different
security
database,
and
so
things
like
the
file
permissions
are
resolved
within
the
context
of
that
container,
not
on
the
host
so
like
the
recommend.
Best
practice
on
Linux
have
often
seen
is:
don't
use,
don't
use
root
in
your
container,
because
if
you
break
out,
you've
got
the
same
user
ID
and
now
you've
got
privileges
being
resolved
based
on
that
UID
on
the
host.
B
B
And
so
just
to
clarify
on
the
Colonel
thing,
because
right
now
they
all
share
the
same
Colonel
there's
a
lot
of
Windows
user
mode,
binaries
like
ntdll
and
user
32,
where
you,
for
whatever
reason
when
Windows
was
developed,
they
took
a
lot
of
api's
and
actually
implemented
them
in
kernel
mode.
That's
a
problem
for
containerization,
because
the
user
mode
binary
is
just
a
shim.
That's
just
going
to
do
a
context,
switch
and
then
operate
it
in
kernel
mode
and
what
that
means
is
so
those
api's
need
to
match
the
kernel
that's
running
on
the
host.
B
So
if
you
build
a
container
on
Windows,
Server,
2000
19,
you
need
to
run
it
on
Server
2019
and
likewise,
if
you're
using
another
version
like
1803
or
2016,
the
same
applies
there.
The
way
that
we're
working
on
addressing
that
is
using
hyper-v
isolation,
because
with
hyper-v
we
can
actually
boot
up
the
match
and
kernel
that
you
need
for
that
container.
So
that
way
the
user
kernel
boundary
matches
again,
and
then
you
also
get
that
extra
layer
of
isolation.
B
That's
something
that
is
always
done
on
Windows
10,
like
if
you're
using
docker
cee
on
a
Windows,
10
laptop
every
time
you
take
docker
run
it's
actually
starting
that
lightweight
VM
behind
the
scenes
actually
could
start
from
a
clone,
so
you're
not
waiting
on
the
full
boot
cycle.
It
will
run
your
container
and
then
once
that
container
terminates
that
partitions
thrown
away,
and
so
we've
got
that
enabled
through
an
alpha
annotation.
That's
there
today,
but
our
plan
is
to
replace
that
with
the
runtime
class
in
kubernetes.
So
that
way
you
get
an
experience.
B
That's
consistent
with
you
know.
If
you're
going
to
use,
you
know
kata
or
one
of
the
other
runtimes
that
has
different
layers
of
isolation.
So
that's
work,
we're
probably
going
to
start
in
either
in
version
14
or
15
to
get
that
consistent
with.
What's
with
what's
done
in
the
next
world
and
just
for
for
reference,
their
runtime
classes
are
actually
still
an
alpha
feature
in
13,
so
we're
getting
in
that
early,
and
so
that
way
you
can
have
the
isolation
or
the
back
backwards.
Compatibility
with
your
hyper-v.
B
B
So
today,
no,
you
could
manually
put
the
label
on
the
windows
nodes
before
you
add
them,
but
the
problem
is
that
we've
got
a
there's,
a
restricted
list
in
the
cubelet
of
what
labels
it's
allowed
to
set
and
right
now
with
the
kernel
version,
is
not
one
of
the
ones
that
it
can
that
the
cubelet
can
ask
the
API
server
to
set
on
its
behalf,
so
I
think
that's
something
that
we're
probably
going
to
need
to
propose
and
and
get
done.
So
that
way
we
can
handle.
C
B
Going
to
be
supported
for
for
it's
a
long
term
support
release
if
you
were
developing
on
18:03,
that
release
is
going
to
expire
in
November,
so
my
recommendation
would
be
you
know
at
this
point
just
stick
with
2019
and
that's
what
we're
gonna
move
all
of
our
stream
or
sorry.
All
of
our
testing.
That's
happening
on
test
grid,
we're
about
to
move
that
on
the
Windows,
Server,
2000,
19
and
just
sort
of
focus
on
that
being
you
know
the
version
that
we're
going
to
support
through
sig
windows.
B
Okay
and
good
time
to
call
this
out,
there's
a
web
page
here,
that's
linked
in
the
slides,
but
Microsoft
actually
has
a
page
that
lists
every
single
Windows
server
and
Windows
10
release
along
with
the
kernel
version.
So
for
whatever
reason,
the
version
is
10
on
all
of
these,
but
the
kernel
version
is
something
that
changes
for
each
major
release.
So
my
laptop
is
running,
looks
like
I'm
running
Windows.
18:03
kernel
version
is
one
seven
one,
three
four.
So
this
is
what
I'm
running
the
Windows
10
equivalent
of
of
Windows
Server
version
1803.
B
Here,
the
Windows
10
and
the
Windows
Server
in
kernels
they're
version
the
same
they're
just
tweaked
a
little
bit
differently,
but
then
the
last
one
that
I
have
asterisks
here
is
actually
the
patch
level.
And
so,
if
you're,
if
you've
got
a
question
of
you
know,
does
this
note
have
a
Windows
patch
applied
to
it?
You
can
go
look
at
that
web
page
and
it
will
give
you
the
kernel
version
number,
along
with
the
corresponding
patch.
If
you
want
to
go
so
you
can
see
which
version
which
patch
level
is
there
and
then
update
it.
B
Okay,
so
so
the
question
was
for
the
cluster
that
we're
running.
Basically,
how
did
we
deploy
so
I'll
cover
that
part?
First,
so
I'm
using
an
open
source
project
called
ACS
engine.
This
is
where
we
are
basically
prototyping
new
features
that
will
eventually
be
released
through
the
azure
kubernetes
service
IKS
brand,
and
so
we
use
ACS
engine
for
doing
daily
testing
of
the
kubernetes
upstream
project.
That's
where
how
we're
getting
our
results
on
test
grid
and
then
we're
gonna
have
results
on
GCE
very
soon.
You
know
yuju
and
peter
are
hard
at
work
on
that.
B
So
that's
going
to
be
coming
soon
and
then,
as
far
as
a
roadmap,
I
can't
give
a
specific
date
because
one
hasn't
been
announced
yet,
but
you
know
we
do
an
ACS
engine
first
in
the
open
source
and
then
once
once
we're
comfortable
with
the
quality
of
it.
Then
it
goes
out
through
HS,
previews
and
eventually
general
availability.
A
I'm
sure
everyone
in
the
room,
obviously
we're
all
familiar
with
some
of
the
release,
automation,
tools,
draft
and
helm.
We
know
that
kubernetes
is
hard.
It
can
be
very
complex,
obviously
even
dealing
with
Windows
deployments.
There
were
quite
a
bit
of
files
and
configurations
that
we
had
to
make
in
the
first
traditional
lab
or
the
more
lift
and
shift
kind
of
model,
but
I'm
sure
one
of
the
big
questions
that
I've
been
asked
and
everyone's
probably
thinking
is,
can
we
use
draft
and
helm
with
Windows?
A
The
short
answer
is
yes,
so
we
do
have
ways
that
we
can
do
that
for
those
in
the
room
who
may
not
be
familiar
with
draft
draft
simplifies
helm.
Helm
is
like
the
package
manager
for
kubernetes
they're,
open
source
tools
that
Microsoft
has
a
vested
interest
in
the
best
helm
itself
is
the
best
way
to
find
and
use
software
that's
built
for
kubernetes.
A
Not
core
Patrick
has
a
draft
pack
for
Windows,
and
so
that
command
has
to
be
run
on
Windows
same
thing,
for
if
you're
doing
any
of
the
built
as
Patrick
talked
about
the
versions,
as
you
start
to
think
about
adding
DevOps
to
the
pipeline
and
taking
this
outside
of
a
lab
and
how
you
would
actually
apply
it
to
your
enterprise.
That's
where
you
start
thinking
of
okay.
How
do
I
handle
the
automation
parts,
whether
it's
draft,
whether
it's
the
build
aspect?
A
All
of
that
you
would
run
those
commands
still
from
a
windows-based
system,
so
it
gives
you
the
files
or
the
draft
pack
accordingly
and
then
builds
gives
you
the
appropriate,
build
dll's
or
solutions
that
you
will
actually
end
up,
deploying
and
checking
in
accordingly
for
your
pipelines.
So
I
actually
have
two
demos
at
the
end
that
I'll
show
you
with
either
a
DevOps
or
code
fresh,
two
different
tools,
just
to
kind
of
show
you,
the
CI
CD
part
of
it.
With
both
of
these
projects.
A
You
could
also
use
Jenkins
or
teamcity
any
CI,
CD
platform
that
you
want
when
it
comes
down
to
DevOps.
The
key
things
to
consider
and
same
thing
for
using
draft
and
helm
is
remember.
When
we
talked
about
versioning,
the
build
node
is
going
to
have
to
match
the
cluster
part
of
it.
So
if
you
do
setup
Jenkins,
it
has
to
be
Jenkins
on
Windows
to
handle
the
build
steps
and
then
technically
the
helm
release
doesn't
have
to
run
on
Windows.
You
would
just
need
your
build
stage
to
be
part
of
Windows.
A
A
That
would
give
you
the
ability
to
do
node,
selector
and
then
edit,
that
from
your
values
file
accordingly
and
then
again
for
your
ingress,
you
want
to
make
sure
that
you're,
if
you're,
deploying
an
ingress
that
it
obviously
gets
specified
accordingly
and
so
again
with
Jeff
the
same
way
that
draft
works
using
draft
packs
on
Linux
there's
an
additional
draft
pack
that
I
believe
do
we
add
it
separately.
Yeah.
A
A
And
the
entire
purpose
of
this
is
to
kind
of
show
you
the
same
thing
right:
the
same
scenarios
as
Windows
as
being
the
base,
whether
you
have
a
more
lift
and
shift
model,
because
you
still
have
the
more
monolithic,
traditional
applications
in
your
enterprise.
But
then
you
also
want
to
still
kind
of
scale
for
current
modern
growth
with
dotnet
core.
You
can
still
use
the
same
tooling,
either
in
either
scenario.
Yeah.
B
Okay,
so
going
back
to
the
same
repo
that
had
the
previous
tutorial
steps,
the
ones
just
accrue
Bernays
for
Windows
tutorial.
The
second
tutorial
on
here
is
actually
using
draft
and
how
so
click
on
this,
and
so
when
you're,
using
draft
draft
and
helm,
the
first
thing
you
need
to
do
is
actually
run
helm
in
it
down
here.
Ok,
I'll!
Get
to
that.
A
second
ok,
so
starting
a
new
app
with
drafted
or
with
draft
and
dotnet
core
is
super
straightforward,
so
I'm
going
to
go,
create
a.
B
We've
got
an
inception
problem
there:
okay,
so
let's
go
ahead
and
make
make
a
new
one,
I'll
call
it
Wario
and
then,
if
I
do
dotnet
new
MVC,
that's
just
going
to
create
a
new
dotnet
core
Model,
View
control,
sample
application
and
then
I'll
open
that
up
in
Visual
Studio
code,
and
so
your
dotnet
makes
it
real,
really
easy
to
start
up
a
simple
project:
that's
ready
to
build
if
you've
used
yeoman.
It's
you
know.
Dotnet
has
something
very
similar
to
that
built
in.
B
B
So
if
you
do
a
draft
pack
list,
that'll
show
you
all
the
packs
that
are
installed,
and
these
are
things
that
our
pluggable,
you
can
just
download
additional
ones
using
this
pack
repo
tool-
that's
built
in
so
on.
These
machines
I've
already
added
my
reg
repo,
which
is
this
Windows
draft
pack
1,
and
that
gave
me
two
different
ones
that
I
can
choose
from
here.
I'm
going
to
use
this
one
which
is
tuned
to
running
c-sharp
applications
using
dotnet
core
on
windows.
B
B
B
The
last
thing
I
need
to
do
is
I
need
to
update
a
few
things
before
I
deploy.
This
first
thing
is
that
I
need
to
update
the
namespace
to
the
namespace.
I
actually
have
privileges
to
use,
and
then
I'm
also
going
to
set
this
up
to
use
an
ingress
controller
which
is
not
the
default
normally.
If
you
were
going
to
run
this
on
your
local
machine,
you
wouldn't
you
wouldn't
need
that,
but
in
this
case,
since
I'm
running
out
of
the
remote
cluster,
I
want
to
create
an
errand
that
route
to
that.
A
B
So
I
update
the
namespace
to
the
one
that
I
have
access
to
as
this
user
and
then
I
added
the
ingress
rule
and
now,
if
I
do
draft
nope
I'm
not
ready
for
that.
Yet
I
should
read
my
own
doc.
So
next
thing
I
do
is
make
sure
helm
is
ready,
and
so
this
is
where
I
need
to
go
ahead
and
run
helmet
in
it.
And
what
that's
going
to
do
is
install
tiller,
which
actually
manages
the
helm
based
deployments.
B
A
And
so,
by
the
way
for
anyone
who
is
new
in
helm
is
new
right
now.
The
reason
why
he's
doing
that
is
different
typos.
So
the
reason
why
he's
doing
this
right
now
is
because
again
helm.
Helm
itself
is
a
local
client
binary
and
then
communicates
with
tiller
until
their
lives
on
the
cluster.
So
because
everyone
has
their
own
namespace,
you're,
deploying
tiller
into
that
namespace
on
the
cluster,
so
tiller
can
handle
the
helm
deployments
and
pass
them
off
to
the
kubernetes
api
on
the
cluster
for
the
lab
yeah.
B
A
And
so
what
draft
us
cuz
draft
is
designed
to
simplify
helm
draft
created
the
packs
for
you,
but
it's
also
going
to
manage
your
deployment,
so
it
created
the
docker
file,
the
charts,
everything
that
you
would
use
for
helm.
But
when
you
do
draft
up,
it's
now
going
to
take
the
whole
lifecycle
of
your
deployment
process.
So
it's
going
to
build
your
docker
image
and
then
it's
going
to
end
up
releasing
that
into
your
kubernetes
environment.
So
it's
going
to
deploy
it
to
kubernetes
using
helm.
But
draft
is
the
one.
A
That's
actually
handling
that
connection,
and
this
can
take
some
time
because,
especially
when
it
has
to
pull
the
image
and
go
through
and
build
after
it
does
build
the
first
few
times
the
layers.
Of
course,
if
your
docker
file
would
be
cached,
so
that
would
allow
it
to
work
much
faster,
but
the
initial
steps
is
going
to
take.
A
It
can
take
several
minutes
and
I've
seen
that,
even
when
you
apply
like
CI
CD
to
that
that's
another
thing
to
consider
if
you
are
using
a
CI
CD
service
that
has
a
hosted
agent
for
any
reason,
and
it's
not
something
that
you
own
that
you
manage
then
each
time
it
rebuilds
unless
that
hosted
agent
allows
you
to
have
persistence.
It's
gonna
have
to
rebuild
that
image
each
time
and
that
can
take
several
minutes
at
times.
Yeah.
C
B
B
The
next
step
here
is:
if
I
do
a
helm
list,
helm
shows
that
the
deployment
was
created.
Ii
do
helm
status
or
you,
and
it
will
show
me
everything
that
was
deployed
here.
So
I've
got
the
service
deployment
ingress.
The
pod
is
running
and
then
the
last
step
here
is.
It
even
gives
me
a
nice
helpful
of
a
link
to
where
this
would
deploy
hey.
It
worked
alright,
so
not
the
applications
up
and
running.
You
know
it's
all.
It's
all
there
and
if
I
wanted
to
go,
you
know
make
some
simple
changes
at
this
point.
B
Let's
change
that,
so
the
message
on
the
about
page
is
different,
but
if
I
do
draft
up
again,
it's
not
going
to
redeploy
everything
is
just
going
to
rebuild
the
container
and
then
redeploy
what
changed
and
that's
something
that
that's
really
important
about
home
because
it
tracks
all
those
are
the
resources
that
were
there
and
because
they
didn't
change
in
a
chart.
It's
not
gonna.
It's
not
going
to
change
them.
So
that
way,
I'm
not
going
to
break
things
like
the
ingress
rules
and
all
of
that
yeah.
A
Actually,
even
in
one
of
the
demos
that
I
did
when
I
first
deployed
the.net
version
of
the
application
through
CI
CD,
the
service
type
was
cluster
IP.
So
obviously
that
makes
it
very
hard
for
me
to
show
you
as
a
demo.
Of
course,
I
could
coop
cuddle
port
forward
proxy,
but
if
I
wanted
to
actually
have
a
load,
balancer
I
just
changed
the
name
of
the
service
type
to
load
balancer
in
the
helm,
chart
redeployed
through
committed
it
ultimately
to
the
repo
and
let
the
pipeline
redeploy
and
then
it
automatically
updated
the
release.
A
B
And
so
now,
going
over
to
this
page,
we
can
see
that
I've
already
got
to
change
live,
so
that
shows
you
know
a
really
quick
little.
You
know
loop
of
a
workflow,
that's
great
for
when
you're
developing
stuff
from
scratch
and
that
little
command
I
had
everybody
run
to
download
the
cube
configs.
Well,
that
was
something
I
wrote
about
a
couple
hours
ago
and
I
just
get
started,
got
done,
got
it
live
with
draft
on.
You
know
less
than
an
hour.
So
let's
see
what
that
looks
like
actually
running
a
at
a
larger
scale.
Ok,.
A
Switch
over
to
my
computer
here
so
like
I,
said:
I
have
two
different,
see
ICT
tools.
Oh
here.
Well,
actually,
recap
what's
happening
first,
so
in
the
demos
I'm
going
to
show
you
is
we
have
DevOps
either
to
lift
and
shift
and
then
DevOps
to
modern.
So
in
the
lift
and
ship
fabric
in
fiber
demo,
the
Patrick
had
you
run
through
the
whole
way
that
that
application
works
is
it's
compiled
outside
of
a
container
initially
right,
you
actually
use
new,
get
restore
as
a
command
out
side.
A
So
in
your
DevOps
pipeline
you're
going
to
consider
that
as
part
of
your
pipeline,
those
are
tasks
that
you
would
run
manually
and
you
just
set
up
so
in
the
demo
that
I
have
the
code
that
I'm
using
is
just
a
forked
version
of
Patrick's,
so
that
I
could
set
it
up.
In
my
own
CI
CD,
tooling
I
am
using
a
private
agent
with
Azure
DevOps,
and
the
only
reason
why
was
because
it
was
its
Microsoft.
A
It's
very
easy
for
me
to
get
Visual
Studio
on
it
and
set
it
up
just
the
way,
I
need
it,
but
because
that
application
was
designed
for
18:09.
That
meant
that
the
server
I
deployed
for
the
build
server
is
also
1809.
I
had
to
do
prerequisites
of
everything
you
would
have
to
do
on
the
same
system.
The
same
way
you're
running
the
lab
means
that
the
build
server
has
to
be
set
up.
It
has
to
have
docker
installed,
has
to
have
Visual.
Studio
has
to
have
helm
any
of
these
commands.
A
I
would
install
previously,
and
so
I
have
code
and
github.
It's
going
to
talk
over
to
my
CI
CD
tool,
and
then
it's
gonna
run
a
series
of
tasks
once
it
finishes
the
building
tasks,
part
of
it
and
it
gets
down
into
the
image,
build
it's
gonna
build
with
docker.
It's
gonna
push
to
a
private
registry.
In
this
case
ACR
the
same
way.
The
demo
did
it's
gonna
package
up
through
helm
and
it's
gonna
ultimately
end
up,
deploying
it
out
to
kubernetes
on
Azure.
A
It's
a
kubernetes
cluster,
exactly
like
what
we
used
in
the
labs
today,
so
we'll
take
a
look
and
actually
see
through
CI
CD.
The
same
way
that
the
lab
had
a
windows
is
application,
relying
on
an
in
cluster
or
Linux
database,
fully
deployed
through
CI
CD.
Now,
as
we
move
to
a
more
modern
approach,
I
setup
using
a
different
tool
show
diversity.
Again,
you
could
use
Jenkins
I,
just
didn't
have
time
to
set
up
a
Jenkins
Windows
build
server.
I
took
dotnet
core,
but
it's
dotnet
core
for
Windows.
A
The
difference
is,
is
the
same
type
of
demo.
That
Patrick
just
showed
you
with
a
brand
new
dotnet
application,
where
the
dotnet
restore
happens
inside
the
docker
container.
We've
now
consolidated
some
of
those
tasks.
So
now
that's
going
to
happen
inside
docker,
it's
still
gonna
push
to
a
private
registry.
We
still
have
how
package
and
deploy
what's
unique,
is
again
in
the
deploy
stage
or
the
release
stage
to
whatever
environment.
A
C
A
I'll
actually
switch
over
here,
and
you
could
kind
of
see
it
in
action
and
I
have
this.
This
isn't
something
that's
part
of
your
lab.
This
is
resources
that
I
published
that
you
can
go
and
play
with
after
cuz
we're
getting
pretty
close
to
time
anyway,
but
true
end-to-end
is
giving
you
resources
that
you
can
use
outside
of
just
a
hands-on
lab
tutorial
right,
so
you
can
see-
and
as
your
dev
ops
right
here,
I
have
a
build
pipeline.
You'll
see
if
I
zoom
in
and
have
to.
A
Happen
on
stage
live
conference
this
year,
so
if
I
zoom
in
you
can
see,
the
agent
pool
says
default,
that's
just
the
default
agent
pool
that
I
use
with
my
build
server.
To
give
you
an
idea
of
what
that
looks,
like
mia
zoom
out
one
more
time,
I,
like
my
handy
mic,
stand
I
actually
have
RDP
running
in
the
background.
So
this
is
my
build
server.
I
started
the
script
for
setting
up
the
private
agent.
It's
not
anything
production-ready!
It's
just
to
give
you
the
visual,
but
you
can
see
each
time.
A
I
had
a
job
run
for
the
various
build,
so
switching
back
over
you
can
see
that
I
have
github.
If
I
zoom
in
one
more
time,
you
can
see
very
briefly
that
I
do
have
a
different
branch
that
I'm
using,
but
it
is
helm
2019
for
Windows,
my
sequel,
Linux.
That's
everything
is
set
up
to
already
run
that
way
and
then
I
just
do
my
standard
tasks
right,
I'm
prepping,
my
build
server
to
make
sure
nuga
is
install.
If
it's
not,
it
will
install
it.
A
If
it
is,
it's
gonna,
be
it's
just
gonna
move
on
and
then
I
do
in
new
get
restore
command
the
build
solution,
everything
that
I
kind
of
explained
in
the
in
the
overshadow
and
then
it
goes
through
and
builds
an
image.
I,
don't
have
to
do
anything
special
aside
from
build
an
image
use
the
default
context,
give
it
a
build
image
name.
This
is
all
very
standard,
I
make
sure
helm
is
set
up
and
then
it
does
a
package
command.
A
The
final
command
is
actually
going
to
be
publishing
that
helm
tarball
file
to
my
release
stage.
So
when
I
actually
run
those
builds,
you
can
see
a
variety
of
different
builds
that
went
through
and
you
could
see
that
some
of
these
happen
from
checking
them
into
github,
which
is
why
I
have
different
pictures,
but
if
I
actually
see
I
added
pull
secrets
to
the
appointment,
so
it
had
I
modified
the
helm
chart
to
make
sure
that
I
could
specify
a
home
secret
for
a
secret
I
created
in
the
namespace
I
plan
to
release
too.
A
Now,
if
you
scroll
down,
you
see
all
the
commands
that
it
runs
through.
You
still
get
the
same
insight
if
I
were
to
check
on
any
of
them.
I
still
get
the
same
insight
where
I
can
see
logs,
it
all
works.
The
same
way
a
Linux
container
pipeline
would
work,
even
though
it's
for
more
of
a
lift
and
shift
model.
I
am
connected
off
of
a
hotspot
right
now,
so
it
might
be
a
little
slow,
but
we
can
see
it's
still
running.
It's
a
visual
studio,
build
it's
running,
Visual
Studio
on
the
private
agent.
A
Now,
when
I
go
over
to
release
for
the
release
stage,
again
still
successful,
but
if
I
go
in
and
edit
the
actual
tasks
it
doesn't
matter
what
the
agent
is
I'm
no
longer
running
on
default,
I'm
now
running
on
just
a
hosted
agent
I
could
change
the
to
Linux.
It's
going
to
operate
the
same
way
and
I
know
that,
because
the
next
pipeline
I
have
is
Linux,
so
the
only
thing
that
I
did
under
the
home
upgrade
command
when
I
scroll
down.
A
Of
course,
I
have
to
add
in
my
kubernetes
cluster,
which
is
2019
I
specified
the
a
DevOps
namespace
I
gave
it
that
package
that
I
handed
off.
That's
a
Windows
package
and
then
the
only
other
thing
I
did
was
I
set
an
image
name,
so
that
way,
I
make
sure
that
that
image
I
override
the
values
file,
that
image
is
gonna
always
be
updated
to
whatever
the
build
ID
was
in
during
the
build.
It's
the
only
other
thing
that
I
that
I
did.
That
was
different,
so.
A
I'm
now
connected
to
my
terminal
and
I'm,
actually
I
deployed
this
out
to
an
1809
cluster.
So
I
set
up
a
function.
That'll
actually
run
a
query
against
it,
so
we
can
see
coop
cuddle
get
pods
wide,
and
so
you
could
see
that
the
two
different
nodes
I
have
a
Linux
node
and
then
I
have
the
Windows
node
right
beneath
it
for
the
web
application
on
Windows
and
then
the
sequel
on
Linux
and
then,
if
I
also
wanted
to
confirm
coop
cuddle.
A
A
This
is
everything
I
already
deployed
before,
because
we
don't
have
time
to
go
through
the
pipeline,
but
it's
a
working
pipeline
now,
as
you
get
over
into
modern
and
a
dotnet
core
on
Windows
application
again
just
chose
to
use
a
different
tool,
but
you'll
notice
that
the
build
settings
and
the
steps
are
nearly
the
same.
The
only
thing
that
we
removed
was
essentially
the
nougat
and
the
build
of
Visual
Studio,
because
we're
now
having
that
as
part
of
docker,
but
it's
still
Windows
so
fun
fact
about
this
particular
demo.
A
For
those
who
don't
know
code
fresh
is
just
another
CI
CD
tool,
they're
designed
for
kubernetes.
So
it
gave
me
some
insight
into
certain
things,
but
I
had
to
make
sure
that
the
build
server
that
they
were
using,
obviously
matched
the
cluster
I
deployed
tail
and
they
upgraded
the
build
server.
Last
night,
so
I
got
to
rebuild
this
this
morning
and
all
I
had
to
do
to
do
that
was
I.
Could
change
under
on
the
deploy
step?
I'll
show
you,
but
I
changed.
A
Just
the
cluster
I
deployed
it
to
and
had
it
rebuilt
now,
because
the
the
pipeline
itself
had
cached
the
layers
I
actually
had
to
make
sure
that
any
cached
was
killed
because
initially,
when
they
upgraded
the
build
server,
the
cache
that
it
had
in
memory
was
still
for
2016.
It
was
still
for
1803.
So
when
it
tried
to
rebuild
you
using
that
cash
on
a
2019
node,
it
would
fail
because
there's
a
mismatch.
So
it
is
one
thing
to
also
consider
as
you.
A
A
C
A
Build
a
docker
image
and
it's
gonna
push
the
docker
image,
and
this
is
the
only
step
or
pipeline
I
should
say
if
I
scroll
back
up
and
I
move
down
to
the
runtime
environment
I'm
using
a
Windows
runtime
environment.
Now,
if
I
go
over
to
deploy,
which
is
not
deploy
right
now,
but
if
I
go
over
to
deploy
right
here,
you
can
see
the
runtime
environment
now
is
Linux.
So
this
is
where
I'm
going
through
I
specify
a
coop
context.
A
So
that's
why
I
changed
it
to
1809
for
the
version
when
I
had
to
update
it.
This
morning
the
chart
name
stays
the
same.
The
namespace
I
isolated
to
stays
the
same
I,
give
it
a
release,
name
just
standard
at
helm.
Tasks
I
can
set
triggers
accordingly,
so
I
can
set
up
like
for
open
a
PR
close
a
PR
merge
into
this
branch.
A
A
Can
actually
see
under
helm
releases
I
have
a
little
releases
tab
over
here.
I
can
see
the
release
that
I
made,
which
is
a
shop
CF
I
only
did
it
one
time
this
time
because
I
had
to
redeploy
this
morning,
but
when
I
click
into
it
and
I
go
to
the
history,
I
can
see
the
history.
I
can
also
see
any
of
the
values
that
were
used
post-processing,
and
so
we
can
see
that
the
get
commits.
Alright
here
is
2
6,
9
6.
A
A
Have
to
make
sure
I'm
on
the
right
branch
I'm
using
this.
This
is
all
in
the
documentation,
but
I
am
using
a
the
branch.
We
can
see
696
zoom
in
right
there
it's
deployment
and
then
you
can
actually
see
the
application.
That's
deployed
right
here.
It's
a
full,
more
modern
application.
If
I
go
back
over
to
my
terminal-
and
we
change
this
to
code
fresh
it'll,
it
should
switch
the
hold
on.
A
A
C
B
Other
thing
I
want
to
mention
real
quick,
is
that
if
you've
got
more
questions
about
Windows
containers
and
cout
Bernays
development
in
general,
we
have
a
session
tomorrow
as
well
along
with
Michael,
and
so,
if
you
want
to,
you
know,
meet
more
the
people
in
cig
windows.
You
know
share
what's
going
well
or
you
know
what
direction
you'd
like
to
see
the
project
going.
You
know
please
come
by
that
session
tomorrow,
that's
at
11:42
1250
and
I
think
it
is
actually
in
the
same
room
if
I
remember
right.