►
From YouTube: Kubernetes UG VMware 20200305
Description
March 5, 2020 meeting of the Kubernetes VMware User Group. This meeting covered using Cluster API for Kubernetes deployments on vSphere, with a demo. Also covered revised plans for user group activities at KubeCon Europe (delayed).
A
A
B
Sure
so,
as
I
was
saying
before
we
started
the
recording,
we
normally
do
this
thing
at
the
start
of
the
user
group,
where
anyone
that's
new
to
the
user
group
introduces
themselves.
You
know
who
you
are,
what
you
do,
what
you're
interested
in,
and
you
know,
maybe
what
you
want
to
get
out
of
the
user
group.
That
kind
of
thing
so
I'm
gonna
pick
up
John,
because
I
work
with
John,
so
John
you're
up.
First,
all.
C
D
E
H
Yeah
thanks
about
thanks
thanks
mom,
so
Joe
man,
I,
am
a
staff
cloud
solution.
Architected
VMware
I
support
the
global
cloud
provider
program
with
the
focus
on
the
modern
application
product
portfolio
at
VMware,
been
working
with
cap
V
over
the
last
couple
months
and
specifically
interested
in
learning
a
little
bit
more
about
the
shift
from
V
1
alpha
2,
V,
1,
alpha
3,
and
what
that
means
for
kind
of
the
provisioning
and
management
workflow
of
Cathy.
I
A
A
B
K
K
A
So
the
first
agenda
item
was
that,
just
in
case
people
weren't
aware
of
it,
we
were
looking
forward
to
having
a
user
group
face-to-face
at
Q
con
Europe
later
this
month,
but
the
CNC
F
canceled
it
because
of
the
corona
virus.
They
didn't
cancel
it.
It's
tentatively
delayed
until
July
or
August,
is
the
current
suggestion
and
I
put
a
link
in
the
agenda
notes
document.
If
you
want
to
learn
more
about
that,
one
thing:
I
did
want
to
reach
out
Robert
from
Amsterdam
I.
A
Think
you're
on
and
I
just
wanted
to
check
whether
I
know
I
and
some
others
RSVP'd
for
the
V
bears
face-to-face
social
event
at
Q.
Khan.
Are
you
okay,
with
just
presuming
that
we
have
canceled,
even
if
we
put
in
an
RSVP
or
I
suppose
you
locals
could
have
that
event
anyway?
If
you
wanted
to
but
I
just
wanted
to
yeah.
L
A
Okay
and
yeah,
and
so
we
look
forward
to
it
being
tentatively,
July
August
but
I-
think
that's
up
to
the
CNC
F
and
we'll
just
have
to
see.
Let's
move
on
then
to
cluster
API,
so
you
seen
is
with
us,
but
just
in
case
there
are
people
on
the
call
who
aren't
familiar
with
the
concept
of
the
cluster
API.
A
It's
not
a
cloud
provider
abstraction
that
still
exists,
but
this
does
not
replace
it.
However,
there
is
a
plug-in
called
a
cluster
API
infrastructure
provider
that
does
adapt
it
to
whatever
it's
managing
and
what
we're
going
to
talk
about
today
is
using
cluster
API
to
deploy
kubernetes
cluster,
specifically
on
vSphere
the
cluster
API,
as
a
concept
is
targeted
at
working
with
deployment
tools
like
cube
ADM,
to
deploy
kubernetes
to
any
target.
You
know
whether
it
be
vSphere
AWS
either.
A
You
know
whatever
public
cloud
and
with
that
said,
as
an
intro,
oh
and
I
did
post
a
link
to
a
deep
dive
deck
on
cluster
API
that
was
presented
at
cube
Con
in
November
20
19.
So
it's
reasonably
fresh.
If
you
want
to
go
into
the
details,
but
I'll
turn
it
over
to
you
seen
to
specifically
cover
cluster
API
as
it
relates
to
vSphere
infrastructure.
E
E
Cap
V
basically
helps
you
manage
to
Banaras
clusters
on
top
of
vSphere
using
the
kubernetes
api
style.
So
I
have
a
quick
demo
that
I
can
show
you
later,
but
I
wanna
first,
but
I
wanted
the
first
to
go
on
the
walkthrough
and
like
the
bare
requirement
to
run
cap
V
on
your
vSphere
installation.
So
for
the
information
that
we
need
for
cap
V,
we
obviously
need
the
credential
that
you're
going
to
use
for
your
setup
when
we're
gonna
need
the
an
endpoint,
a
V
Center
endpoint.
E
We're
going
to
also
need
some
specific
information,
such
as
the
data
center.
The
data
store
networks,
the
resource
for
where,
where
we're
going
to
learn
the
VMS,
the
template
that
we're
gonna
use
to
create
machines
out
of
it
and
a
template
for
what
we're
calling
the
H
a
proxy
load
balancer,
which
is
basically
VMs
that
hosts
a
Shiprock
HT
proxy
instances
that
are
going
to
load,
balanced
traffic
to
the
API
server
and,
finally,
an
SSH
key
that
you'll
love
that
it
will
help
you
to
access
the
machines
once
they're
provisioned.
E
So
with
that
I
have
a
quick
recording
that
I
made
today.
So
first
first
thing
that
you'll
need
is
what
we
call
a
kind
code.
Oh
might
need
to
refresh
this
so
yeah
as
I
said.
First
thing
that
you'll
need
is
what
we
call
kind,
which
is
urban
areas
in
docker
turbine
Edison.
Docker,
helps
you
to
create
kubernetes
clusters
with
nodes
that
are
docker
containers.
E
So
first
thing
is
you
create
the
kind
cluster
I'm
gonna
move,
so
once
you
create
the
kind
cluster,
we
have
a
command
a
CLI
which
is
claw
circle.
Claw
circle
has
a
command
which
is
in
it
that
helps
you
install
all
of
the
provider
components
into
the
kind
cluster.
So
it's
gonna
install
basically
all
of
the
controllers
of
cluster
API
and
cluster
API
for
vSphere.
So
if.
A
E
As
you
said,
yeah
the
kind
cluster
is
for
bootstrapping.
Basically,
you'd
need
to
have
a
kubernetes
cluster,
where
you
install
all
of
the
controllers
that
are
going
to
reconcile
the
CR.
These
are
clusters,
machines
and
whatnot.
So
you
this
is
going
to
be
used
specifically
for
bootstrapping.
So
once
we
specify
the
versions
we
want
for
cluster
API
and
for
cap
V,
which
is
that
infrastructure.
E
Cluster
Karl
is
going
to
fetch
all
of
the
providers
and
install
them
directly
to
your
kind
clusters.
So
it's
gonna
fetch
things
like
like
certain
manager
since
we're
using
it
for
certs
and
for
managing
what
works.
Certs.
It's
gonna
install
all
of
the
CR
DS
and
all
of
the
deployments
that
hosts
the
controllers.
E
So
once
once
it's
installed,
it's
gonna,
it's
gonna,
prompt
you,
the
first
command
that
you
want
to
use,
which
is
cluster
called
config
cluster.
This
basically
help
well,
this
basically
fetch
a
llamó
file
that
we
publish
usually
at
each
release.
So
if
I
go
to
the
CAD
via
repo
and
I
go
to
a
release,
we
publish
some
assets.
One
of
the
assets
that
we
publish
is
the
cluster
template.
E
So
if
I
open
this
so
it
basically,
the
cluster
template
looks
like
this,
which
is
a
bunch
of
me
llamo,
which
represents
things
like
cluster
h8,
proxy
load,
balancer,
the
vSphere
cluster
and
all
of
the
info
that
are
needed
to
bootstrap
the
cluster
and,
as
you
see,
everything
is
like
using
variables
so
close
circle.
Once
you
are
going
to
use
this
Yammer
with
classical
classical
is
going
to
automatically
replace
these
dis
with
specific
values.
So.
E
E
E
E
B
E
E
For
the
template
currently
are
published,
and
only
one
to
be
specific,
we
were,
we
were
attempted
at
some
point
to
publish
a
second
flavor,
so
flavors
means
that
yeah
for
for
classical
we
have
the
notion
of
flavors.
Flavors
are
basically
different
workload,
cluster
templates.
So,
if
you
have
let's
say
a
load,
balancing
solution
is
based
on
a
cheap
proxy
you'll
have
a
flavor
that
is
a
cheap
proxy
and
if
you
are
are
for
example,
on
you
have
added
a
controller
that
creates
I,
don't
know
another
load.
E
Balancing
solution
such
for
example
as
AWS
ELB
you'll,
have
another
flavor.
So
for
now
we
have
only
one
and
the
second
one
that
we
were
tempted
to
provide
is,
instead
of
using
the
new
cube,
ATM
control
planes.
Crd,
we
were
planning
to
add
a
second
flavor
which
was
called
cubed,
which
is
basically
not
based
on
cube,
ATM
control,
plane,
but
just
tbdm
config.
The
difference
is
that
when
you
use
a
cube,
ATM
config,
you
need
to
provide
a
machine
too.
So.
E
We
didn't
want
to
provide
this
because
it'll
mean
that
you
it
mean
that
you're
gonna
run
on
some
issues,
such
as.
How
do
you
upgrade
your
cluster?
How
do
you
you
know?
How
do
you
do
how
checks?
So
this
is
the
one
of
the
reasons
that
we
didn't
want
to
provide.
Another
flavor,
but
I
think
that
in
the
future,
my
f
ba,
probably
we're
gonna,
provide
some
other
load-balancing
solution
other
than
H
a
proxy,
so
so
for
for
this
year,
we're
gonna
provide
provide
other
templates
based
on,
for
example,
NSX
G.
E
If
you're
talking
also
about
VM
templates,
you
can
definitely
use
other
ones.
So
if
I
go
to
the
readme,
we
have
currently
supported
Centro,
s7,
Ubuntu
and
photon.
So
we
like
the
community,
is
provided
through
the
image
builder
these
three
and
we
are
you
constantly
updating
the
versions
so
that
you
can
ship
the
latest
kubernetes
version.
So
currently
we
have
117
and
once
118
is
out
we're
gonna.
Add
the
two
yeah.
I
So
if
you
want
to
build
your
own
images,
there's
a
separate
project
called
image
builder.
So
if
you
just
swap
that
for
image
build-
and
that
is
a
bunch
of
packer
and
antipas
scripts
to
build
your
own
images
and
it
which
you
can
build
your
OVA
template
to
that.
So
if,
when
you
specify
a
communities
version,
you
need
to
have
a
OVA
uploaded
into
your
recenter
with
the
match
and
and
they
need
to
match
so
in
the
future,
we're
thinking
of
writing
or
automated
controller.
I
E
I
M
I
I
I
Created
a
haitch,
a
proxy
load
balancer
for
which
we
are
provided
a
template,
I
believe
the
scripts
for
that
I
in
the
hack
directory.
At
the
moment
we
need
to
find
a
final
resting
home
over
time,
but
for
now
yeah
we
provide
run
upstream
image.
You
are
able
to
build
your
own
and
you
need,
and
you
will
def
if
you
want
to
make
use
of
a
stateful,
high
availability
control
plane.
You
need
to
use
that
policy
as
well.
Well,.
M
E
E
H
Just
just
a
quick
question,
so
I'm,
looking
at
the
back
in
the
cluster
init
command
I'm,
trying
to
like
I,
said
initially
trying
to
understand
the
change
between
alpha
2
and
alpha
3,
these
core
bootstrap
control,
plane
infrastructure
flags
that
you're
defining.
Where
are
those?
Are
those
file
names
locally
on
your
machine
and
how
are
those
generated
into
fine
so.
E
E
So
if
I
go
here,
it's
gonna
go
and
fetch
specifically
this
file.
So
this
is
the
file
that
is
going
to
be
fetched,
so
are
so
classic
all
is
fetching
the
core
components
which
contains
if
I
open
this
it's
gonna
contain.
All
of
you
know
the
namespace
definitions,
the
CR,
these
different
web
hooks.
So
it's
gonna
fetch
this
and
install
it
for
you.
So.
A
E
Have
there's
a
last
time
I
used
this.
It
was
a
script
that
is
called
under
hack.
It's
called
over
add
of
write
PI.
Basically,
it
overrides
the
files
that
we
want
to
use
and
it's
gonna.
Basically,
if
you
drop
a
file
under
close
the
cluster
API
directory.
So
if
you
drop
a
file
under,
for
example,
let's
say
vSphere
and
you
drop
the
the
components
file
here-
it's
gonna
fetch
it
from
here.
Ok,
so
this
was
originally
used.
E
H
H
I
E
There's
a
gotcha
that
you,
you
need
to
be
aware
of
currently,
if
I
nadir,
correct
me
if
I'm
mistaken,
but
currently
closed
circle,
do
not
allow
having
empty
variables.
So,
for
example,
if
you
don't
want
to
specify
an
ssh
authorized,
key
you'll
end
up
with
an
error.
So
you
need
to
you
to
specify
some
value
here.
C
I
Yeah
one
thing
you
might
want
to
do
so
because,
plus
the
cartel
supports
environment
variable
so
as
a
moment
go
said,
don't
do
that
because
it's
in
your
battery,
but
if
you're
using
something
like
pass,
you
could
have,
you
could
have
the
environmental
set
as
a
output
of
a
command
that
fetches
it
from
encrypted
at
rest.
Source
temporarily
makes
it
available,
as
in
a
moment
variable
prior
to
execution
of
cluster
castle.
So
there's
probably
different
ways:
you
could
scrape
this.
E
E
I
Boy,
it's
it's
full
user,
so
tossed
API
has
no
need
to
SSH
into
your
machine.
So
it's
purely
a
break
glass
mechanism
for
operators
to
get
into
the
machines
if
they
need
to.
We've
included
this
in
a
template
that
if
people
don't
want
it,
there's
a
casing
made
that
we
don't
put
in
the
default
template
at
all.
I
F
Correct
me,
if
I'm
wrong
linear,
but
essentially
with
cluster
API,
construct,
there's
like
a
cloud
in
it
profile
that
gets
created
by
the
bootstrap
provider
and
the
infrastructure
provider
when
it
goes
to
create
an
actual
node
for
either
the
control
plane
or
for
the
worker
node.
And
it
injects
that
SSH
key
in
there,
so
that
you
have
a
way
to
SSH
into
either
your
control,
plane
or
worker
notes
that
are
generated.
Freezing
close.
I
Tightly
yeah
and,
for
instance,
some
that
a
SS
party
we've
taken
position
to
remove
the
ability
to
SSH,
if
you
say
which,
and
that
will
probably
people
so
my
T's,
that
you
can
always
customize
the
output
of
the
cluster
cutter.
Whatever
class
gold
race,
you
can
always
customize
the
output
before
you
apply
it
to
the
either
the
kind
class
or
whatever.
What
we
call
a
management
cluster
where
these
objects
and
ants.
E
So
when
you
apply
the
EML
that
was
completed,
the
first
thing
that
you'll
do
is,
for
example,
you
get
the
machines
and
they're
gonna
show
up
as
pending.
So
these
are
the
machine,
the
worker
machines
that
are
generated
by
the
machine
deployment.
If
I
get
the
cue,
medium
control,
plane,
I'm
gonna
find
that
there
is
a
cube,
ADM
control
plane
that
was
applied
with
all
of
the
values
that
were
templated
such
as
the
SSH
key,
so
I
think
one
of
the
things
that
I
might
show
you,
since
I,
am
going
to
apply
the
mo
anyway.
E
E
E
E
So
cluster
is
a
generic
resource
that
is
going
to
hold
some
generic
informations
that
are
going
to
be
true
for
every
provider.
So
for
every
provider
it's
gonna
need
a
pot
cider.
It's
gonna
need
a
reference
to
the
to
a
control
plane.
It's
gonna
need
a
reference
to
also
two
and
an
infrastructure
resource
that
are
that
is
going
to
hosts
specific
informations
about
the
providers
such
as
you
know,
the
V
Center
endpoint,
the
credential
D,
all
of
the
configure
the
bits
of
configuration
that
are
specific
to
your
provided.
So
what's.
A
E
We
like
with
closed
circle.
Config
we've
provided
an
easy
button
experience,
so
you
can
not
worry
about
fetching
all
of
these
informations,
but
if
you
want
to
you
know,
if
you
want
to
construct
your
own
workload,
cluster
template
I
would
say
that
the
Cappy,
the
cat
pee
book,
is
the
right
place
to
start
I.
Don't
know
if
he
didn't
know
if
it
was
updated
for
v1
or
three.
I
E
F
Sorry
go
ahead.
Yeah
I
was
just
going
to
say
that
a
lot
of
the
configuration
around
specifics
of
cluster
bootstrapping,
the
bootstrap
provider,
which
the
I
think
the
primary
in
use
today
is
cube.
Adm,
if
you
are
have
any
existing
experience
with
making
cube
ATM
configurations
and
whatnot
the
majority
of
those
configurations
options
if
not
the
entirety
of
them
are
exposed
or
available
here
so
customizing.
You
know,
you
know
your
cubelet
configuration
and
your
hot
cider,
server-side
or
stuff
like
that
is.
It
should
be
very
similar
but
yeah
definitely
a
few
gaps
in
the
documentation.
F
E
Yeah,
if
I
show
an
example
of
community,
you
should
be
able
to
I
would
say
extensively
change
the
the
the
spec
here,
because
we
since
we're
using
the
cube,
ADN
types.
So
there's
no
reason
to
all
that
you
wouldn't
be
able
to
reuse
any
configuration
that
you've
had
previously
with
cube
ADM
and
that
you
will
use
it.
E
E
With
that,
I
can
show
also
a
bit
about
the
other
resources
that
I'll
get
back
to
the
demo.
So
for
the
class,
as
I
said,
we
had
a
cluster
which
is
provided
by
Kathy.
We
have
machine
deployments,
which
is
this
basically
based
on
the
same
concept
as
the
kubernetes
deployment
we
have
machine
sets
for
replicas
sets
and
we
have
machines
that
matches
the
concept
for
pouts.
So
these
are
the
main
series
that
you
interact
with.
E
E
E
Let's
see
if
you
get
any
machines,
I
think
my
connection
is
laggy,
so
I'm
gonna
mostly
rely
on
the
demo
that
I
recorded,
so
we
have
on
v-103
something
that
is
called
vSphere
VM.
So
vSphere
VM
is
a
construct
that
hosts
all
of
the
information
to
create
a
virtual
machine.
It's
not
it's,
not
something
that
is
going
to
be
a
kubernetes
node.
It
can
be
just
a
VM
to
create
you
know
a
cheap
proxy,
so
we
introduced
this
specifically
to
be
able
to
manage
easily
H
a
proxy
VMs.
E
A
E
So
yeah,
as
I
said,
the
didn't
like
the
main
changes
for
this.
This
version
is
the
introduction
of
a
check
proxy
as
a
load,
balancing
solution
plus
v1
alpha
3.
So
once
once
you
get
like
a
hand
around
hold
around
these
specific
changes,
there
are
like
there
are
no
major
updates
in
terms
of
of
the
way
Cathy
operates
from
v1
of
the
two
to
v1.
Also,
three
so
accept
the
changes
of
the
control,
plane,
endpoint
and-
and
this
you
should
be
good
to
go.
E
I
think
on
other
thing
that
we
need
to
mention
is
that
we
still
like
we
plan
to
provide
for
the
community
a
document
that
outlines
the
things
there,
that
you
want
that
you
that
you'll
want
to
do
to
move
from
a
cab.
We
close
tab.
We
cluster
that
was
bootstrapped
under
v1.
Up
at
you
to
see
are
these
that
are
V
1
alpha
3,
so
this
is
gonna,
come
a
bit
later
after
the
final
release.
E
E
E
I
You
think
add
is
this
hecho
parts
of
work
pain
is
a
sort
of
experiment
that
we're
doing
in
cluster
a
guy
around
load
balancers.
So
we
induce
very
cluster.
We
actually
have
something
called
the
load
bounce
of
reference.
Now
it
could
reference
a
different
implementation,
so
I
speaking,
some
people
cloud
nation
London
yesterday
and
there's
a
lot
of
people
here-
want
something
that's
more
like
an
MIT
based
solution,
something
like
people
like
D,
there's
nothing
to
stop
people
from
writing
that
implementation,
using
that
in
its
Ted.
M
I
E
Probably
so
yeah
it's
just
to
wrap
up
I'll
just
go
further
on
the
demo.
So
once
once
you
get
the
your
your
first
control
plane,
node
that
has
a
provider
ID
and
that
is
in
a
run
stage.
The
the
other
machine
deployments
generated
machines
are
going
to
start
and
provision.
So
this
way
we
ensure
that
you,
like
the
workers,
are
going
to
join
once
the
image
is
completed,
so
we'll
have
all
of
the
machines
that
are
going
to
be
in
a
provision
stage.
A.
E
E
E
E
B
J
Okay,
yeah
I'll
go
ahead,
then,
so
the
wonderful
folks
at
SAT
I
wanted
to
contribute
to
the
vSphere
clan
provider,
and
so
they
had
the
initial
support
for
creating
an
excess
tea
load
balancers
in
front
of
you
conveyance
clusters
using
the
service
API,
and
so
this
is
a
pretty
new
feature
it
merged
early
this
week
and
it's
it's
an
alpha
features.
So
we
are
looking
for
early
feedback
folks.
J
Our
folks
have
at
least
year
and
NSSC
installation,
we'd,
love,
we'd
love
for
you
to
reach
out
and
give
this
try
and
yeah
so
I'm,
starting
to
put
out
there
on
to
see
if
anyone's
going
to
try
it.
Please
don't
use
it
in
production
because
it
is
alpha
and
next
week
or
not
next
week,
but
the
next
user
group
meeting
I
will
try
to
prepare
like
a
really
quick,
10
minute
demo
on
how
to
use
it
and
what
to
do
that'd.
A
F
Just
one
thing
from
me
for
more
so
of
like
a
potential
future
topic
as
I
know,
there's
a
lot
of
conversations
and
and
stuff
going
around
on
the
integration
of
some
of
the
avi
networks.
Components
into
you
know:
vSphere
and
the
service
mesh
area
things
around
the
like
nsx
service
measure,
spec
I,
think
that'd
be
really
cool
to
do
as
a
potential
topic
for
a
future
meeting.
A
A
Okay,
if
there
are,
if
there
are
no
other
nominations,
I'll
call
this
to
a
close,
but
that
agenda
doc
is
open.
I'll
carry
the
service
mesh
asked
over
and
put
it
on
the
agenda
for
next
month's
meeting
and
we'll
see
who
we
can
bring
miles
and
I
had
a
presentation
plan
for
cuke
on
Europe.
That
will
be
about
the
timeframe
of
the
next
one.
So
maybe,
if
we
need
more
content
well,
the
two
of
us
will
put
together
the
deck
we
had
planned
for
Q
Khan
and
put
it
on
here.