►
From YouTube: Metal3 Bare Metal Host Provisioning for Kubernetes - Feruzjon Muyassarov & Mael Kimmerlin (Ericsson)
Description
OpenShift Commons Briefing AMA Session on Metal³
Metal3 Bare Metal Host Provisioning for Kubernetes
Feruzjon Muyassarov & Mael Kimmerlin (Ericsson)
https://metal3.io
The Metal³ project (pronounced: Metal Kubed) exists to provide components that allow you to do bare metal host management for Kubernetes. Metal³ works as a Kubernetes application, meaning it runs on Kubernetes and is managed through Kubernetes interfaces.
In this Commons AMA session, Feruzjon Muyassarov and Mael Kimmerlin (Ericsson) will give an overview of the project, demostrate deploying with Metal3 and talk about the road ahead
A
All
right,
everybody
happy
monday
and
welcome
back
to
an
openshift
commons
briefing.
Today
we
are
going
to
talk
about
metal,
cubed
or
metal
3
or
whatever.
We
want
to
call
it,
but
I,
like
metal,
cube
the
best
and
it's
a
new
cncf
project.
We
have
two
folks
here
from
erickson
who
are
maintainers
on
the
project,
mel
and
farragh,
who
I'm
going
to
let
introduce
themselves
we'll,
have
live
q
a
at
the
end.
B
Hello
thanks
a
lot
welcome
everyone,
thanks
for
having
us
today
here
and
giving
us
a
chance
to
share
with
you
a
project
that
we've
been
working
on.
So
we
are
really
really
happy
to
be
here
and
today
we'll
be
talking
you
about
a
quite
a
young
project
called
the
metal
cube
which,
which
does
basically
provisioning
for
the
bare
metal
host
for
the
bare
metal
hosting
the
kubernetes
cluster.
B
B
Cool,
thank
you.
So
what
is
metal
cube?
What
problems
does
it
solve
and
what
does
it
really
offer
to
you?
So,
first
of
all,
it's
a
cncf
project,
sandbox
project.
In
fact,
it's
a
quite
young
project,
as
I
already
mentioned
it,
and
the
interest
to
the
project
has
been
increasing
so
far,
quite
a
lot
and
the
community
always
keeps
basically
growing
and
we're
seeing
more
and
more
people
joining
our
community
doing
all
sorts
of
contribution,
which
is
really
really
amazing.
B
So
now,
let's
see
what
is
metal
cube
stack
and
what
does
it
really
represent
for
us?
So,
let's
see
the
high
level
picture
of
what
would
you
get
from
the
metal
cube?
So
imagine
that
you
have
the
infrastructure
or
the
bare
metal
infrastructure
that
you
want
to
manage,
and
the
metal
cube
has
a
component
called
cluster.
Sorry,
the
bare
metal
operator
that
basically
takes
care
of
provisioning
and
then
the
provisioning
of
your
bare
metal
hosts.
B
The
ironic
it's
somehow
hidden
from
the
picture
so
and
the
bare
metal
operator
always
takes
care
of
managing
the
ironic
itself.
So
you
don't
have
to
do
the
management
of
the
ironic,
because
metalcube
will
do
it
for
you,
yeah
and
also
the
next
component
in
the
in
the
in
the
metal
cube
stack,
is
called
cluster
api
provider,
metal
cube,
which
is
also
another
big
component,
that
metal
cube
offers.
But
before
we
jump
into
it,
I
would
like
to
mention
that
parameter
operator
can
be
used
separately.
B
You
don't
have
to
use
any
other
components
of
the
metal
cube
to
do
the
management,
because
the
main
component
is,
let's
say,
bare
metal
operator.
But
if
you
want
to
integrate
your
cluster
management,
specifically
bare
metal
cluster
management
with
the
other
projects
like
cluster
api,
then
you
have
to
use
the
cluster
api
provider,
metal
cube,
which
is
basically
the
plugin
that
you
can
use
in
order
to
plug
your
management
into
the
cluster
api
project,
so
both
bare
metal
operator
and
then
the
cluster
api
metal
cube.
B
They
run
inside
the
kubernetes
cluster
so
which
would
be
in
this
case,
which
would
be
really
really
easy
to
manage
them,
and
you
don't
have
to.
As
I
already
said,
you
don't
have
to
have
some
extra
tooling
or
components
to
manage
the
clusters,
and
then
the
last
component
in
this
stack
is
the
cluster
api.
B
So
the
cluster
api
is
the
high
level
project
in
this
stack.
Let's
say
that
offers
some
machine
objects
and
then
cluster
objects,
and
then
those
objects
are
represented
by
different
infrastructure
provider
in
different
ways.
But
in
our
case,
for
example,
if
you
create
the
machine
object,
the
cluster
api,
which,
at
the
end
of
the
day,
represents
your
kubernetes
node.
That
machine
object
would,
for
example,
if
you
say,
create
one
machine
object
from
the
cluster
api.
B
Great,
so
so,
quick
overview
of
the
cluster
api
before
we
dive
into
the
metal
cube
project.
Cluster
api
is
a
is
a
kubernetes
sub
project
focused
on
the
cluster
lifecycle
management,
and
it
allows
you
basically
the
day
two
operations
or
the
manage
your
cluster
in
most
of
the
cloud
environments
or
in
most
of
the
cloud
providers.
B
And
then,
whenever
you
do
the
management
you
do
it
in
a
declarative
way
right.
So
you
do
you
use
the
kubernetes
native
apis
or
you
use
the
kubernetes
manifest.
So,
in
short,
you
will
be
able
to
deploy
and
and
manage
your
kubernetes
cluster
where
the
where
the
kubernetes,
so
that
means
all
the
components
and
then
the
the
building
blocks
of
your
cluster
api.
B
All
the
controllers
are
running
inside
the
kubernetes
cluster
and
then
those
controllers
actually
manage
your
target
clusters,
which
are
running
somewhere
in
the
cloud,
for
example,
or
in
the
in
the
environmental
infrastructure.
So
as
such,
we
always
need
a
cluster
to
start
with
cluster
api.
So
we
create
a
management
cluster
which
is
also
known
as
a
bootstrap
cluster
or
the
ephemeral
cluster,
which
you
see
on
the
left
side
of
the
slide.
B
And
then
in
that
formal
ephemeral
cluster.
We
start
installing
all
the
cluster
api
components
and
then
the
controllers,
and
once
you
have
a
bootstrap
cluster
up
and
running
with
all
the
cluster
api
components
and
all
the
necessary
controllers,
then
you
can
start
actually
creating
your
target
cluster
in
your
desired
cloud
environment,
whether
it's
gcp
or
aws,
or
azure
or
digital
ocean
whatsoever.
B
So
what
we
did
also
what
we
did
with
metalcube
is
that
we
somehow
extended
the
list
of
the
infrastructure
providers
by
adding
the
a
cluster
api
infrastructure
provider
for
bare
metal.
So
metalcube,
as
I've
already
mentioned,
allows
you
to
do
the
cluster
management
or
environmental
host
management
for
in
the
in
the
real
bare
metal
servers
or
in
the
real
environmental
infrastructure.
So
we
basically
added
one
of
the
info
providers
for
the
list
of
the
cluster
apis
infrastructure
providers.
B
B
So
once
you
create,
or
once
you
create
a
when
you
want
to
create
the
kubernetes
node,
so
you
can
use,
for
example,
customer
api
to
to
to
define
from
where
that
actual
node
should
be
created,
like
a
virtual
machine
or
the
actual
bare
metal
server.
So
the
node
is
transformed
into
the
machine
object
which
is
coming
from
the
cluster
api
project,
so
which
is
kind
of
the
generic
across
all
the
infrastructure
providers.
B
But
once
you
create
the
machine
object
from
the
cluster
api,
then
you
have
to
tell
in
which
infrastructure
do
you
want
to
actually
create
the
machine?
For
example,
you
can
say
that
I
want
to
create
the
machine
or
the
virtual
machine
in
aws,
or
I
want
to
create
a
machine
on
the
google
cloud
or
I
want
to
create
the
real
bare
metal
server.
That
would
represent
my
machine
and
that
will
be
basically
represent
the
kubernetes
node
right.
B
So
once
you
have
defined
it
from
where
you
want
to,
from
which
infrastructure
you
want
to
create
your
machine,
then
you
will
have
the
actual
cloud
infrastructure
that
will
be
taking
care
of
creating
virtual
machines
in
the
cloud
for
the
cloud
providers
or
the
actual
bare
metal
server
for
the
metal
cube,
for
example.
So
what
happens?
Is
that
in
most
of
the
cloud
providers,
the
process
will
end
up
in
in
in
the
in
the
cloud
infrastructure
itself,
so
that
all
the
cloud
providers
have
already
have
their
apis
to
manage
their
underlying
infrastructure.
B
But
in
our
case
since
we're
doing
the
the
bare
metal
node
thermometer
host
management,
we
we
didn't
have
any
cloud
provider
behind
and
we
had
to
create
our
own
api.
That
will
actually
do
the
the
management
of
the
reels
or
servers
or
the
physical
machines.
So
for
that
we
have
created
a
bare
metal
operator,
as
I
already
mentioned
a
bit
so
which
actually
does
the
management
of
the
underlying
infrastructure
or
your
bare
metal
servers.
B
And
then,
at
the
end
of
the
day,
it
will
basically
talking
to
your
the
the
bare
metal
server.
So
you
can
see
that
from
the
left.
Whenever
it's
moving
from
the
left
to
the
right,
we
have
the
node
object
that
is
referenced
by
the
machine
and
that
machine
is
referenced
by
the
infrastructure,
specific
machine
in
our
case,
metal,
cube
machine.
And
then
we
talk
to
bare
metal
operator
and
then
ask
hey.
I
want
to
create
a
one
server
in
this
data
center.
B
C
Yeah,
thank
you
very
much
yeah.
So,
let's
dive
a
bit
deeper
in
the
in
the
technical
details
of
the
of
the
metallurgy
project,
so
we're
going
to
start
with
this
kind
of
like
overview
of
the
different
elements
that
that
we
have
and
how
they
work,
how
they
work
together.
So
you've
already
heard,
like
all
these
kind
of
terms
like
damage
operator,
the
met
the
provider
metal
cube
for
cluster
api
cluster
api
itself.
C
So,
basically,
for
all
of
those
there's
like
there's
objects
representing
like
real
real
items
like,
for
example,
we
have
a
cluster
object
that
represents
the
the
kubernetes
cluster
and
with
its
like
equivalent
for
the
provider
that
is
here,
the
metal
cube
cluster
same
way.
We
have
the
machine
that
represents
the
the
kubernetes
node
and
the
metal
cube
machine,
that
is,
the
infrastructure
provider
equivalent
for
that
machine.
C
So
all
all
of
those
are
basically
referring
to
each
other,
the
like
for,
for
example,
the
cluster
points
to
the
metal
cube
cluster
like
telling
plaster
api
how
to
actually
like
deploy
that
cluster,
like
with
the
infrastructure
provider
and
the
machine
points
to
the
metal
cube
machine
like
telling
exactly
how
to
deploy,
deploy
that
kubernetes
node.
In
addition,
as
fellow
said,
there
is
the
birmingham
host
that
is
all
referenced
directly
from
the
metal
cube
machine.
C
To
tell
like
that,
you
want
to
deploy
this
kubernetes
node
on
this
specific
hardware,
so
the
different
controllers
are
represented
in
the
in
this
picture
and
they
interact
each
with
their
own
on
objects
and
there's
usually
a
dedicated
controller
for
each
of
the
each
of
the
objects.
The
controller
might
be
like
editing,
some
other
objects
as
needed
to
fulfill
its
role.
C
So
that
is
the
high-level
view
of
the
different
crds
that
we
have
in
the
in
the
metal
cube
project.
Now
we
are
going
to
like
dive
in
like
have
a
look
specifically
at
each
of
the
each
of
the
metal
cube
like
crds
that
we
have
so
the
first
one
is
the
metal
key
cluster.
It's
it
actually
consists
like
with
the
like
usual
elements
of
a
cr
of
cr.
C
It
needs
to
be
defined
beforehand
because
in
the
case
of
metal
infrastructure,
we
are
not
in
a
cloud
provider
environment
where
we
have
just
like
load
balancers
that
we
can
create.
Unfortunately,
we
have
to
handle
it
as
part
of
the
deployment,
so
you
need
to
give
it
beforehand
telling
like
this
will
be
the
the
end
point
of
the
cluster,
so
that
is
the
metal
cube
cluster.
Then
the
next
item
was
the
metal
cube
machine.
So
this
defines
like
gives
more
detail
about
like
how
the
how
the
kubernetes
node
would
be
deployed.
C
C
So
that
ironic
can
deploy
the
node
with
that
specific
image,
there's
also
a
host
selector
that
allows
you
to
choose
which
of
the
bar
metal
hosts
you
want
to
deploy
on,
and
then
you
have
a
couple
of
other
available
fields
like
data
template,
for
example.
That
allows
you
to
to
pass
templates
for
the
metadata
that
will
be
included
in
the
user
data
or
for
the
network
configuration
that
will
also
be
applied
by
ironic
through
cloud
init.
You
can,
of
course
you
don't
have
to
use
templates
for
that.
C
If
you
want
to
directly
give
the
metadata
or
the
network
data,
you
can
also
do
it
directly
via
where
the
metadata
network
data
fields,
so
this
basically
allows
you
to
configure
in
quite
fine-grained
level
of
details
the
deployment
of
the
kubernetes
node
that
you
will
have
it's.
It
doesn't
touch
anything
to
with
regard
to
kubernetes
itself.
It's
it's
more
like
if
you
want
to
do
some
customizations
on
your
on
your
node
or
like.
If
you
want
to
change
something,
deploy
a
specific
image,
and
all
of
this
is
in
this
specific
object.
C
C
So
the
first
very
important
one
is
the
bmc.
So
that's
where
you
will
have
the
the
details
of
the
of
the
management
interface
of
that
node,
so
you
can
have
either
like
well
in
that
case
a
little
bit
when
you're
using
some
virtualization,
but
if
you're
using
an
actual
server,
you
would
probably
have
something
like
redfish
or
ipmi
or
ilo,
like
depending
on
what
you
what
you're
using
so
a
lot
of
protocols
are
supported.
C
C
So
the
way
to
do
that
is
to
tell-
which
is
the
expected
mac
address
for
that
node
when
it
boots,
so
that,
like
ironic,
can
figure
out
that,
okay,
that
node
is
booting,
and
it's
this
I
I
know
that
it
will
be
like
this
parameter
host.
So
that's
how
it's
basically
matched
and
then
again
you
you
can
give
the
image
the
url
to
the
image
and
the
checksum
and
then
the
user
data.
If
you
have
any
that
you
want
to
to
be
given
through
cloud
for
cloud
init
run.
C
So
this
is
basically
the
the
three
core
crds
that
we
have
in
the
metalcube
project.
We
are
not
going
to
go
through
the
cappy
one,
because
there
are
already
like
a
lot
of
available,
webinars
and
other
things
related
to
this.
So
we
can
give
you
some
references
if
you
want,
let's
keep
instead
like
diving,
deeper
into
into
metal
human,
how
things
are
actually
working
here.
So
in
the
next
slide,
we
are
going
to
be
talking
about
the
barometer
operator
and
how
it
manages
the
node.
C
So
the
bamata
operator
itself
is
is
a
controller
that
interacts
with
the
the
cr,
the
bar
metal
host
crs,
and
it
can
do
a
couple
of
operations,
so
the
first
one
would
be
to
inspect
the
hardware
it's
able
to
like
boot
the
node
and
then
run
something
that
is
called
ironic
python
agent.
That
will
like
go
through
all
the
specs
of
the
of
the
hardware
and
report
it
to
to
ironic,
and
the
metal
operator
will
be
able
to
fetch
that
data.
C
C
The
third
operation
that
vermont
operator
can
do
for.
You
is
to
clean
the
disk,
and
that
usually
happens
during
the
provisioning
and
de-provisioning
of
the
of
the
diameter
operator.
There's
then
the
like
a
couple
of
let's
say
useful
capabilities
of
permit
operator.
It
can,
of
course
like
manage
the
power
of
your
node
so
like,
if,
if
you
need
to
reboot
it
to
power
it
on
power,
it
off,
like
you
can
do
it
through
the
crs
just
editing
the
diameter
host
vrs.
C
So,
let's,
let's
talk
a
bit
more,
maybe
like
in
in
more
fine
details
about
the
the
barometer
operator
and
the
ironic
interaction.
So
the
bare
metal
host,
the
the
cr
itself
points
also
to
a
couple
of
secret
and
those
secrets
contain
the
user
data
and
the
metadata,
for
example,
and
the
network
data.
All
those
are
for
cloud
in
it
and
when
you
give
it,
you
can
give
it
as
different
elements,
but
at
the
end
they
will
be
combined
into
a
config
drive,
and
so
ironic
takes
care
of
this.
C
And
then,
when
barometer
operator,
instructs
ironic
to
to
start
the
deployment.
Then
ironic
will
start
talking
to
the
bmc
to
turn
on
the
server
and
then
the
the
the
server
will
boot.
The
ironic
python
agent
ironic
will
then
directly
talk
with
the
with
the
python
agent
that
is
here
called
deploy,
ram
disk
and
ask
it
to
basically
download
the
image
from
whatever
web
server,
where
it
is
stored
and
write
it
to
the
disk.
So
that's
back
to
the
local
disk.
C
So
that's
how
the
magic
happens
now
like
if
we
want
to
go
even
deeper
in
the
details,
we
can
like
see
how
this
happens
under
the
hood.
So
when,
basically,
when
diameter
operator
registers
the
node,
what
openstack
ironic
does
under
the
hood
is
that
it
goes
to
talk
to
the
bmc
to
turn
on
the
server.
The
server
will
send
a
dhcp
query
when
booting,
because
it
tries
to
boot
over
pxe
the
dns
mask.
C
The
server
will
then
proceed
with
introspection
on
the
first,
the
first
time,
the
the
first
time
the
the
server
starts
with
epa.
It
will
always
do
the
introspection
and
then
send
the
report
to
ironic
that
diameter
operator
will
then
be
able
to
fetch
directly
from
from
ironic
and
once
the
node
is
introspected,
then
you
have
it
ready
and
you
can
start
like
provisioning,
a
node
and
like
deploying
something
on
top,
so
the
for
the
provisioning
so
same
thing.
C
The
metal
operator
will
start
talking
with
ironic
and
ironic,
will
like
put
the
epa
again
on
the
server
and
then
instructed
to
deploy
the
node.
So
the
epa
in
that
case
will
download
the
image
from
http
write
it
to
the
disk.
So
the
thing
is,
if
you're
downloading
the
image
like
in
a
in
some
specific
formats
like
qcar2,
they
will
need
to
be
uncompressed
and
like
adapted,
but
basically
at
the
end
it
will
write
the
raw
image
to
the
disk.
C
C
Now,
if
we
talk
a
bit
about
the
integration
of
the
cluster
api
provider,
metal
cube
and
the
different
functionalities
that
this
handles,
we
have
the
the
following.
So
this
is
really
the
the
integration
with
cluster
api
that
the
metal
cube
object
like
metal,
cube,
cluster
and
metal
cube
machine
are
actually
interacted
with
the
cluster
api
controllers.
C
So
we
have
the
different
elements
like
metal,
cube,
cluster
metal,
cube
machine,
but
there's
also
like
additional
ones
that
we
didn't
go
too
much
into
details
like
the
machine
template
that
allows
cluster
api
to
generate
metal,
cube
machine
based
on
this
lag.
You
can
think
about
it,
like
maybe
for
deployment.
The
spec
part
of
a
deployment
that
is
then
translated
into
actual
pods
or,
like
some
data
template
that
would
be
from
which
you
can
generate
the
user
data
and
like
including
also
the
network
data
for
the
for
the
node.
C
A
C
A
C
A
C
A
B
B
Thanks
so
we
have
done
the
recording
for
the
for
the
demo,
but
maybe
I
will
first
actually
show
something
else
before
we
jump
into
the
I
forgot,
the
actual
demo.
So
in
the
metal
cube
project,
we
have
created
a
special
repository
called
metal
cube
devens.
That
is
basically
responsible
for
that
contains
a
couple
of
scripts
that
you
can
use
in
order
to
test
metal
cube.
So
using
the
metal
kit
event,
you
can,
for
example,
deploy
cluster
api
provider,
metal
cube.
You
can
deploy
bare
metal
operator.
B
You
can
create
a
couple
of
virtual
machines
or
delivered
virtual
machines
and
manage
them
as
if
they
were
your
real
environmental
servers.
So
the
reason
that
we're
using
virtual
machines
is
the
first.
We
cannot
always
provide
private
metal
servers
for
testing
and
it
gets
very,
very
complicated,
but
thankfully
we
have
some
kind
of
tools
that
allows
us
to
really
really
replicate
the
real
world
scenario,
with
real-world
scenario,
with
the
with
the
virtual
machines.
B
So,
for
example,
instead
of
using
the
dmc
like
you
would
use
in
real
environmental
servers,
we're
using
virtual
bmc
that
will
basically
talk
to
the
management
of
your
delivered
virtual
machines.
So
what
happens?
What
will
happen
during
the
demonstration?
Is
that
we're
going
to
clone
the
metalcube
developer
repository,
go
into
the
path
and
then
run
the
make
so
in
the
place
where
we're
running
the
make.
So
it's
gonna
first
install
the
cluster
for
us
and
then
inside
the
cluster.
B
It's
gonna
install
a
couple
of
components
like
the
cluster
api,
which
I
already
talked
like,
which
is
a
high
level
that
they
decor
project,
let's
say
and
then
we're
gonna
install
it's
gonna,
install
the
cluster
api
provider,
metal,
cube
and
then
the
bear
model
operator
for
us.
So
all
these
components
are
running
inside
the
this
cluster
and
let's
call
this
cluster
as
a
source,
but
it's
also
known
as
fmrl
or
the
bootstrap
cluster.
So
once
they
all
these
components
up
and
running.
B
The
scripts
will
create
a
couple
of
virtual
machines
for
us,
our
library,
virtual
machines
and
then
it's
gonna
join.
It's
going
to
create
the
bare
metal
host
objects
reconciled
by
the
bare
metal
operator
and
those
bare
metal
host
objects
will
represent
the
virtual
machines
that
we
have
created
and
then
once
we
have,
the
the
virtual
machines
that
are
referenced
by
a
particular
parameter
host,
we're
going
to
start
the
provisioning
of
the
of
the
departmental
hosts.
B
So
the
provisioning
like
we're
going
to
install
some
operating
system
into
them,
then
inject
some
ssh
keys
and
then
go
into
the
inside
the
vm
and
see
if
the
cluster
is
running
and
if
the.
If
the,
if
the
bare
metal
host
or
in
our
case
libre
virtual
machine,
is
part
of
the
target
cluster
so
but
once
the
the
operating
system
is
installed
or
the
provisioning
is
done,
the
scripts
will
try
to
create
a
target
cluster
for
us
and
then
join
those
bare
metal
hosts
or
the
nodes
into
the
target
cluster.
B
B
So
we
will
be
running
on
the
bar
metal
on
the
virtual
machines
or
delivered
machines,
and
once
you
have
cluster
wrapper
running
with
those
nodes,
you
can
do
any
qctl
operation
crude
operation
create
delete,
update
whatever
on
the
bmh
object,
which
is
the
short
name
for
the
bare
metal
host
or
you
can
do
any
operation
on
the
top
level
object,
which
is
the
machine
coming
from
the
cluster
api
object
all
right.
Now,
I'm
going
to
switch
to
my
terminal
and
start
playing
the
recording
we
have
done
you.
B
B
So
what
happens
first
is
that
I
have
already
cloned
the
metalcubed
events
in
this
environment
and
now
I'm
exporting
a
couple
of
environment
variables
before
I
run
the
make.
So,
for
example,
here
I'm
specifying
the
container
runtime
to
be
used
in
the
metal
cube
development.
You
can
use
different
container
runtimes,
for
example,
you
can
use
docker
or
deployment
and
then
here
I'm
specifying
the
target
os
that
will
be
provisioned
that
will
be
used
to
provision
the
target
node.
B
So
I'm
specifying
ubuntu,
of
course,
I'm
not
telling
which
version
of
the
ubuntu,
because
we
have
made
it
in
the
screen
that
you
just
specified
version,
and
it
will
just
pick
the
right
one
for
you
and
then
we
have
the
another
environmental
called
ephemeral
cluster
which
specifies
what
tool
do
you
want
to
use
in
order
to
spin
up
the
source
cluster?
So
we
support
currently
the
kind
and
the
mini
cube
to
spin
up
the
source
cluster
and
then
the
sec.
B
The
la
the
next
variable
is
the
cap
m3
version,
which
is
the
cluster
api
provider,
metal,
cube
version.
We
have
different
versions
of
the
cap
and
three
so
in
this
case
we're
using
the
latest
one
viva
alpha
4
and
then
the
number
of
the
nodes,
which
represents
the
number
of
the
virtual
machines
or
delivered
machines
that
you
want
to
create
in
your
environment
right.
So
once
we
have
exported,
we
start
running
the
make
and
then
this
process
will
gonna
will
take
a
couple
of
minutes.
B
B
At
the
same
time,
we
have
four
ironic
nodes
that
represents
that
are
referenced
by
those
virtual
machines,
but
in
real
case
it
would
be
your
perimeter
server,
for
example,
and
then,
at
the
same
time
here
in
the
last,
you
can
see
that
we
have
four
bare
metal
host
objects,
not
zero,
not
one,
not
two,
and
not
three
in
the
ready
state.
So
ready
statement
means
it's
ready
to
be
provisioned.
B
It
has
no
operating
system,
so
you
can
start
actually
provisioning,
but
introspection
is
already
done
for
it
and
it
it's
also
registered
in
in
in
your
it's
also
registered
for
the
bare
metal
operator
right.
The
manageable
is
also
from
the
ironic
perspective.
That
means
that
you
can
like
start
provisioning,
those
ironic
notes.
B
So
then
we
in
the
in
the
same
environment.
We
have
a
couple
of
scripts
that
we
use
to
provision
the
bare
metal
environmental
hosts.
So
you
can
see
here
we
have
cluster
script,
we
have
control
plane
and
then
we
have
worker
script.
We
first
start
executing
the
clusters
dsh,
which
will
basically
create
the
cluster
object
and
apply
into
the
cluster,
and
then
it
will
create
also
a
metal
cube.
Cluster
object
which
is
metalcube
specific.
B
Then
the
control,
plane,
dot
stage
will
create
one
machine
object
and
then
metal,
cube
machine
object
and
then
the
bare
metal
host
object.
So
it's
kind
of
link
it
and
then,
at
the
same
time
it
will
create
another.
When
you
run
the
work
it
will
create
the
same
chain,
basically
machine,
then
metal,
cube
machine
and
then
the
bar
metal
host.
So
we
will
have
in
this
environment
for
now
two
bare
metal,
hose
two
metal
machines
or
two
machines,
or
two
delivered
machines,
virtual
machines
that
we
will
use
in
our
cluster.
B
So
one
of
the
machines
will
represent
the
kubernetes
control,
plane,
node
and
the
second
one
represent
will
represent
the
communities
working
out.
Let's
say
so.
I'm
going
to
skip,
however,
on
the
screen.
So
basically,
I
have
run
the
scripts
here
and
after
some
time
you
can
see
that
provisioning
has
started
so
provisioning
does
not
start
in
parallel.
It
will
start
it
will
do
the
provision
one
by
one.
So
you
can
see
one
of
the
provision.
B
One
of
the
nodes
has
started
provisioning
and
it
started
with
control
plane,
of
course,
and
then
the
the
the
virtual
machine
or
delivered
machine
that
represents
that
verbatim
host
star
is
up
and
running,
and
then
you
can
see
from
the
ironics
perspective
that
that
node
is
in
clean
way
state,
meaning
that
it's
actually
right
now
doing
the
cleaning.
As
my
earlier
mentioned
it,
and
it
basically
wipes
out
all
the
disks
that
are
available
on
that
delivered
virtual
machines
or
virtual
machines.
Sorry.
B
B
So
also
you
can
see
the
consumer
for
developmental
host
object.
You
can
see
the
consumer,
which
represents
the
the
metal
cube
machine
that
is
consuming
this
particular
machine.
Bare
metal
host
object
all
right
and
then
also
you
can
see
the
online
field.
The
online
field
is
set
to
true
for
both
of
these
verbal
hosts
that
represent
that
they
are
powered
on
right
now.
C
B
So
the
next
step
that
we're
going
to
do
is
that
see
the
chain
of
the
object
references.
So
we
mentioned
couple
of
times
that
we
have
cluster
objects,
some
objects
from
coming
from
the
cluster
api.
Some
objects
are
coming
from
the
metal
cube
and
then
they
are
referenced
by
each
other.
So
the
first
object.
The
core
object
coming
from
the
cluster
api
is
the
cluster.
B
You
can
see
that
it's
in
provision
state
and
then
we
have
infrastructure,
specific
cluster
object
that
is
referenced
by
this
top
level.
Cluster
object.
So
in
our
case
in
the
metal
cube,
it's
a
metal
cube
cluster.
So
you
can
see
that
name
is
test1
and
it's
basically
referenced
by
this
cluster
object.
Then
we
have
the
machines
object,
also
from
the
cluster
api.
So
we
have
two
cluster
api
machines
and
we
have
two
cluster
api
provider:
metal,
cube
machines
that
are
referenced
by
these
machines.
B
And
then,
at
the
same
time,
we
have
two
other
different
objects
from
the
cluster
api.
One
is
the
machine
deployment
and
then
second
is
the
kcp.
So
machine
deployment
is
like
basically
like
a
deployment
in
for
the
ports,
so
it
basically
you
can
use
this
object
to
manage
your
machines
and
then
you
have
the
tcp,
which
is
kind
of
the
similar
to
machine
deployment,
but
it's
specifically
meant
for
the
control
plane
nodes.
B
B
So
the
next
step,
what
we're
gonna
do
is
that,
if
you
remember
I
have
in
this
environment,
we
have
created
four
virtual
machines,
so
the
next
step
would
be.
We
will
try
to
basically
scale
up
those
machine
deployments
and
see
if
the,
if
the
metal,
kit,
machines
and
the
bare
metal
hose
gets
created
and
then
starts
provisioning.
B
But
before
that
I
would
also
like
to
show
you
that
now
we
have
two
parameter
host
provisions.
They
are
part
of
the
target
cluster.
One
of
them
is
the
control
plane
and
the
second
one
is
the
the
worker
so
and
then
we
can
see
we
have
the
cube
config
for
the
target
cluster,
and
here
you
can
see
that
qctel
get
ports
with
the
target.
B
B
B
But
if
you
install
the
calico
in
the
target
cluster-
and
you
do
the
proper
networking,
then
you
should
see
that
the
status
of
the
nodes
in
in
ready
state
so
and
then
the
last
check
is
that
if
we
do
kill
ctl
get
machines
from
the
source
cluster,
we
can
see
that
we
have
two
machines
with
the
exact
same
name
as
the
kubernetes
node
for
the
target
cluster.
And
that's
because,
as
I
mentioned
earlier,
we're
creating
two
machines
that
represent
at
the
end
of
the
day,
represent
to
your
kubernetes
nodes.
B
Okay,
this
is
the
first
part
now
we're
gonna
play
a
bit
with
machine
deployment
and
try
to
scale
it.
So
currently
we
have
one
replica
of
the
machine
deployment,
so
we
have
one
worker
node
and
then
we
will
try
to
increase
the
replicas
to
three,
because
we
have
two
other
delivered
virtual
machines
that
are
free
and
that
we
can
utilize
them.
B
So
you
can
see
that
it
says
it's
scaled
now
and
then,
if
we
check
the
status
of
all
the
corresponding
objects,
so
first
we
can
see
that
machine
deployment
is
now
scaling
up
and
then
it's
trying
to
have
two
three
replicas
of
the
of
the
machines.
And
then
you
can
see
that
here
we
have
machines
being
created
and
they
are
in
the
provisioning
state.
B
So
this
process
also
provisioning.
As
I
said,
it
takes
a
lot
of
time.
So
after
some
time
we
should
be
able
to
see
all
two
new
orbital
hosting
provision
state
and
two
lever,
virtual
machines
up
and
running
as
part
of
your
kubernetes
cluster.
So
you
can
see
that
now
they
are
up
and
running,
so
the
machines
are
running
and
then
the
metal
kit
machines,
they
are
also
ready
and
they
are
basically
the
other
two
virtual
machines
are
also
up
and
running,
and
we
can
also
see
the
ironic
node
status.
B
B
B
B
B
As
I
said,
metalcube
is
kind
of
is
young
project
and
but
we're
growing
really
fast
and
we
have
a
lot
of
contribution
from
from
different
companies
as
I,
as
was
listed
in
the
previous
in
the
previous
slides,
so
the
the
contribution
that
you
are
doing
it
can
be
basically
anything
so
you
can
do
any
documentations
a
contribution.
You
can
have
some
requests
for
the
new
feature.
You
might
have
some
found
some
bugs
and
you
might
have
to
fix
them.
B
If
you
want
to,
or
you
can
report
by
the
issues
you
can
participate
in
helping
creating
the
talks
or
presentations
like
this
or
the
writing,
some
blog
posts
we
have
in
the
metalcube.io
website.
We
have
a
lot
of
blog
posts
that
different
people
just
write
about
the
features
in
the
from
the
metal
cube,
trying
to
share
the
knowledge,
basically
from
with
the
community
and
outside
of
the
community,
and
then
any
questions
that
even
the
feedbacks
that
you
might
have
for
the
metal
cube
is
really
really
appreciated.
B
So
we
really
laugh
and
we
would
really
appreciate-
have
some
contribution
from
your
site
so
and
then
also
about
the
community.
I
mentioned
the
community,
so
we
have
the
metalcube
community.
It's
quite
diverse,
as
I
said,
we
have
different
contributions
in
different
time
zones,
so
the
the
the
the
github
page
is
the
metal
cube
door,
metal
cube
dash
io,
so
we
have
contributors
from
currently
at
dale,
eriksen,
fujitsu,
mirantis
and
then
the
red
hat.
B
So
if
you
want
to
reach
out
the
community
members
or
ask
any
questions
or
chat
with
the
community
members
or
the
contributors,
you
can
join
the
metal
cubes
for
the
cluster
api
bare
metal
on
the
kubernetes
slack
channel.
Or
if
you
have
some
questions,
you
can
also
reach
out
the
the
maintainers
through
the
cncf
mailing
list,
or
you
can
reach
out
the
communities
through
the
metalcube's
own
mailing
list
and
also
you
can
watch
some
updates
and
then
new
features
being
added
through
the
twitter.
B
So
we
have
the
community
meeting
that
happens
every
alternate
wednesday
at
1
pm,
utc
time
it
happens
on
the
zoom.
You
can
find
the
link
here
and
then
we
also
have
the
recordings
and
different
kind
of
demos
on
the
metal
cube
youtube
channel.
That
represents
different
features
of
the
metal
cube
and
then
some
discussions
that
we
have
during
the
community
meetings
that
you
might
be
interested
in,
and
that
is
the
last
slide
for
our
presentation
today.
A
Yeah,
no
thank
you
very
much,
mal
and
and
for
joining
us
today,
a
couple
of
things
pop
into
my
mind.
I
know
a
lot
of
the
people
who
are
watching.
This
are
probably
open
ship
users
so
they're,
I
believe,
there's
a
set
of
instructions
and
if
somebody
peter,
maybe
you
can
pull
up
the
center
strike,
there's
a
different,
a
slightly
different
deployment
approach
when
we
do
it
use
metal
cube.
I
think
the
bare
metal
deployment
on
there
so
just
not
to
confuse
the
two.
A
We
are
using
metal
cubed
for
the
openshift
provisioning
bare
metal,
but
there's
a
whole
set
of
documentation
on
how
to
do
it
using
openshift
so
and
that's
the
beauty
of
open
source,
so
it
gets
used
by
lots
of
people
in
lots
of
different
ways
and
we
all
get
to
collaborate
on
it.
So
this
is,
I
think,
it's
pretty
amazing.
It's
wonderful
that
you're
in
the
cncf
sandbox.
I
know
that
was
a
recent
event.
A
Was
that
a
couple
months
ago
I
don't
know
exactly
yes
yeah
a
couple
months
ago,
and
it's
a
pretty
healthy
community
that
you
have
around
around
metal
three
already.
I
think
it
was
it
filled.
It
definitely
filled
a
gap
in
the
pantheon
of
the
cncf
landscape,
which
is
amazing
to
think
that
there
was
a
gap
because
there's
so
much
stuff
in
that
landscape
diagram,
but
bare
metal
was
one
of
those
things
that
really
wasn't
being
addressed
very
well.
A
So
I
think
this
is
a
perfect
fit
for
for
the
cncf
sandbox
and,
hopefully
incubation
sometime,
not
too
distant
future.
Can
you
tell
us
a
little
bit?
I
guess-
and
let
me
see
in
the
chat
if
anyone
else
has
questions
besides
me,
peter
was
saying:
bmc
is
also
used
with
ocp,
but
it
bootstraps
all
the
typical
ocp
installation
process.
Yeah.
That's
definitely
there's
a
slightly
different
deployment
methodology
for
when
you're
using
this
on
openshift.
So
definitely
read
the
openshift
bare
metal
docs.
A
If
you're
watching
this
and
if
you're
watching
this
from
anywhere
else
doing
it
anywhere
else.
That's
you
know,
read
the
stuff:
that's
on
the
metal,
three
dot
io
website
and
contribute
your
feedback
to
that.
That
website
too,
I
think
that's
the
beauty
of
this
project
and
a
lot
of
other
projects
at
the
cncf
is
that
they
may
get
put
out
there
by
red
hat
initially
or
something,
but
then
they
get
adopted
by
erickson
and
at
t
and
mirantis
and
everybody's
collaborating
on
it.
A
Can
you
talk
a
little
bit
first
about
the
use
case
at
erickson?
You
know
what
made
it
so
important
for
erickson
to
get
involved
in
this
project
and
to
help
move
it
forward,
because
you
guys
two
of
you
maintainers
on
the
project.
Obviously
erickson's
got
a
big
commitment
to
using
this.
Can
you
tell
us
a
little
bit
about
that
and
how
you
got
permission
to
participate
so
actively
in
the
project.
C
So
we
were
looking
for
like
a
provider
for
like
them
at
all,
and
at
that
exact
point,
the
metalcube
project
popped
up
from
like
red
hat,
to
get
started
that
at
the
this
exact
time
it
took
us
a
bit
of
time
to
get
on
board
because
internally,
we
have
a
lot
of
ndas
and
stuff
like
this.
That
prevent
any
kind
of
open
source
contribution.
C
So
there's
actually
a
second
entity
that
was
created
just
for
open
source
contributions.
So
people
had
to
move
companies
to
to
be
able
to
to
do
the
the
open
source.
But
then
we
actually
like
really
got
on
board
and
started
like
contributing
as
much
as
we
could
to
and
and
now
metalcube
is
really
used
as
part
of
the
like
the
core
of
the
diameter
solution
for
ericsson's
kubernetes
distribution.
A
So
there
are
a
couple
other
questions
here
that
are
popping
in
that
are
more
related
directly
to
the
project.
One
is
asking:
what
are
the
prereqs
before
you
start,
the
installation
like
dns
network
availability,
bmc,
configurations,
etc,
and
it
seems
you
have
to
know
mac
addresses
and
other
details
about
the
physical
hardware
you're
provisioning
on
is
that
true.
C
There's
a
couple
of
things
you
need
to
have
done
before
you
can
start
any
any
deployment
is,
of
course,
the
networking
like
that
needs
to
be
in
place
like
on
the
physical
level,
but
also
on,
like
the
configuration
of
the
switches
and
everything
you
also.
So
then
that's
at
that
point.
You
should
know
already
the
mac
address
of
the
interface
that
boots
of
a
pxe
like
to
to
be
able
to
register
the
ironic
note
properly.
C
You
don't
need
to
know
too
much
more
about
the
the
hardware,
maybe
some
details
about
the
hard
drive.
If
you
want
to
select
one
specifically
for
the
for
the
installation,
the
the
and
with
regard
to
the
bmc,
it
needs
obviously
to
be
configured
and
reachable
so
ironic
needs
to
be
able
to
reach
reach
it
so
meaning
that
the
node,
where
the
ironic
pod
runs
in
the
cluster
needs
to
have
access
through
routing
or
direct
connectivity.
C
It
doesn't
matter
to
the
to
the
actual
bmc,
and
then
you
store
the
credentials
for
that
specific
nodes
bmc
in
the
in
the
secret
for
the
bare
metal
host,
so
yeah
there's
a
bit
of
work
to
be
done
beforehand.
Like
you
need
to
have
your
bmc
configured
and
the
credentials
there,
you
need
to
have
the
networking
done,
but
once
it's
kind
of
like
an
initial
setup
and
once
that
is
done,
then
you
can
deploy
whatever
you
need.
On
top.
A
I
think
it
did
and
definitely
I
these
have
been
great
questions
peter.
So
so
thanks
because
there
really
are
some.
You
know
a
lot
of
people
are
trying
out
this
project
for
the
for
the
first
time-
and
I
know
one
of
your
colleagues
or
on
the
project,
hamesh
is
working
on
doing
a
demo
of
deploying
metal
cubed
on
okd4,
so
he's
coming
to
the
okd
working
group
in
not
too
distant
future,
hopefully
to
demo
that
as
well.
So
there's
lots
of
lots.
There's
still
lots
to
do
in
this
project.
A
A
C
Get
anything
we
don't
have
talks
per
se,
but
we
have
office
hour
no
like
because
we
now
have
our
sandbox
project.
We
got
the
benefit
of
being
able
to
schedule
a
couple
of
office
hours
during
kubecon,
so
we're
gonna
have
to
one
of
them
is
in
the
17th
of
december
and
the
other
one
is
on
the
20th.
I
think,
but
I
don't
recall
the
hours
with
the
time
zone.
It's
anyway
such
a
mess.
I
wouldn't
be.
A
Summer's
far
far
far
away
and
kubecon
is
coming
much
faster
than
that.
It's
like
a
freight
train
coming
right
at
us,
so
definitely
look
for
those
office
hours
because
that's
on
the
17th
that's
coming
up
soon.
There's
a
couple
of
community
meeting.
I
think
there's
at
least
one
community
meeting
before
that
as
well
on
your
schedule.
If
I
looked
right.
A
Next
week,
so
there's
lots
of
places
where
you
can
participate
and
if
you
are
an
open
shift
user,
do
check
the
documentation
on
openshift
for
the
bare
metal
deployment,
because
it
is
a
slight
variation
on
this
and
you'll
probably
need
some
more
details.
But
this
is
a
great
way
to
to
showcase
how
someone
like
erickson
and
is
putting
huge
effort
into
stepping
up
and
contributing
back
and
collaborating
with
lots
of
other
people
to
make
something,
wonderful
that
we
all
get
to.
A
You
know,
get
to
use
and
take
advantage
of,
and
hopefully
contribute
back
to
so
peruse
and
mal
thanks
for
taking
the
time
today.
We
really
appreciate
this
and
keep
us
posted
we'll.
Have
you
back
when
you
get
your
next
release
and
you
can
tell
us
about
all
the
new
features
and
look
for
a
demo
by
hamesh
around
deploying
okd4
on
bare
metal
using
metal?
A
Three,
I
think
that's
one
of
his
channel
challenges
of
the
month
to
do
so,
and
there's
also
a
great
blog
post
series
of
blog
posts
on
the
metal,
three
dot
io
website
too,
to
check
out.
So
if
you
have
a
use
case
for
this
and
you're
looking
for
some
place
to
land
and
have
a
conversation
about
it,
definitely
reach
out
and
join
this
crowd.
So
thanks
again
everybody
and
have
a
wonderful
week,
it's
only
monday.