►
Description
Due to technical problems sequences of the lecture may be missing! We apologize for this.
I'll be talking about Cluster API provider for MAAS from Canonical. It's a fully open-source provider allowing users to bridge Cluster API with an open-source bare-metal provisioning tool.
Website: https://www.spectrocloud.com/
Organized by @Microsoft @kubermatic7173 @SysEleven
Thanks to our sponsors @CapgeminiGlobal, @gardenio, @sysdig, @SUSE, @anynines, @redhat, nginx, serve-u
A
So
the
way
how
it
describes
so
there
are
some
examples
of
these
are
specifically
examples
for
eks
I'm
going
to
be
talking
about
mass
in
a
minute,
but
it
sort
of
resembles
what
puppets
work.
Terraform,
unstable
Chef
do
for
for
your
infrastructure
bits.
So
that's
a
description
in
yaml
files.
What
how
you
would
Define
your
kubernetes
clusters?
Okay,
okay,
so
what
is
it
you
produce
with
it?
You
create
a
cncf
compliant
clusters:
okay
and
the
probably
the
most
important
thing
is
not
to
create.
The
thing
here
is
not
to
create
a
cluster
right.
A
You
could
create
it
with
menu
tools
out
out
there,
but
it's
a
day
two
operations.
So
that's
management.
That's
upgrades,
that's
installation
and
or
ad
in
do
these
Sim
class
translations
of
monitoring
logging.
All
these
Beauty
things.
Okay,
now
it
uses
kubernetes
style
API,
and
it's
built
on
abstractions,
so
I'm
not
going
to
go
deep
into
cluster
API
and,
quite
frankly,
if
you
want
to
use
it,
you
don't
really
need
to
have
a
deep
understanding
of
it.
You
could
just
benefit
from
the
configuration.
A
So
if
you
want
to
start
it's
just
quite
quick
now
it
has
some
controllers
and
it
uses
CRT
behind
the
scenes.
So
I'm
sure
most
of
you
are
familiar
with
crd,
so
there
is
nothing
specifically
complicated
behind
it.
So
that's
cluster
API
now
the
second
part
is
mass,
so
first
of
all
I
just
wanted
to
cover.
Why
do
you
would
you
be
interested
in
creating
a
cluster
kubernetes
cluster
directly
directly
on
bare
metal?
There
may
be
a
couple
of
reasons.
A
Most
of
the
real
world
scenarios
I've
seen
is
where
you
will
deploy
a
new
infrastructure
or
a
new
version
of
your
infrastructure,
and
when
you
didn't
inherit
this
from
2010s,
when
there
is
a
huge
state
of
virtual
machines,
you
could
directly
Deploy
kubernetes
on
bare
metal.
Just
because
all
of
this
all
of
this
problems
that
virtualization
solves,
like
you
know,
highly
available
virtual
machines,
migration,
storage.
To
some
extent,
this
has
been
now
sold
by
kubernetes.
A
So
therefore,
if
you're
deploying
a
new
state
or
if
you're
deploying
a
next
version
of
your
current
state,
you
may
just
skip
virtualization
whatsoever
now
it
might
not
be
easy
to
attack
straightforward,
but
there
is
certainly
an
interest
behind
it.
Specifically
when
you
want
to
gain
in
performance,
when
you
don't
really
have
the
skills
to
manage
your
virtualization
State
again,
it
depends
on
what
your
virtualization
provider
is
and
then
again
you
would
probably
wouldn't
pay.
You
would
wouldn't
want
to
pay
if
you
have
a
expensive
subscription
with
some
of
the
virtualization
vendors
out
there.
A
Now
how
deployment
of
kubernetes
on
bare
metal
looks
like
today
today,
if
you
have
a
number
of
physical
servers
machines,
they
would
be
installed.
You
would
need
to
install
a
protein
system
on
that
now.
This
problem
has
been
solved
many
times
in
the
in
the
past
years,
but,
as
you
know,
we
all
like
to
reinvent
the
wheel
with
kubernetes,
but
how
it
works.
Today,
you
would,
you
would
install
operating
system
on
your
bare
metal
boxes
and
then
separately,
you
would
install
kubernetes
now
both
of
these
problems
have
been
more
or
less
solved,
especially
the
bottom.
A
One
with
the
operating
system,
but
the
problem
we
which
we
are
trying
to
solve
and
which
we
are
trying
to
tackle
here
within
Spectra
Cloud
environment
here,
is
the
discrepancy
between
the
two
right.
So
there
is
a
kubernetes
life
cycle
management
tools.
There
are
plenty
of
those
you've
probably
heard
already
some
about
some.
You
probably
will
probably
hear
some
time
or
later
today
or
tomorrow.
There
are
obviously
many
various
tools
of
meant
for
managing
your
operating
system
estate.
That
also
exists.
A
The
problem
is,
they
are
not
connected
between
each
other
okay,
so
what
we
are
trying
to
solve
is
to
treat
the
bare
metal
as
a
kubernetes
cluster
using
cluster
API.
So
this
is
why
I'm
talking
about
these
two
technologies
and
that
we
we
bridge
the
gap
between
the
two
so
cluster
API,
it's
a
tool
that
allows
you
to
manage
your
kubernetes
clusters
in
a
declarative
way.
Mass
is
an
operating,
is
a
tool
that
allows
you
to
manage
servers.
A
So
let's
go
straight
into
must
right,
so
that's
a
tool
that
allows
you
to
connect
to
your
service
state,
let's
say
to
idrac,
to
ILO
to
any
server
management
to
boot.
Up
the
server
install
operating
system,
and
do
it
all
also
in
the
controls
in
a
declarative
way?
Okay,
so
what
we
are
trying
to
do
here,
we
are:
we
have
written
a
provider
for
cluster
API
that
is
published
on
GitHub
that
Bridges
these
two
technologies
together.
So
once
again,
there
is
a
cluster
API
that
creates
kubernetes
cluster.
A
There
is
mask
that
installs
operating
system
on
on
your
physical
bare
metal
boxes.
Now,
within
our
Labs.
Well,
we
say
for
different
numbers
with
the
quality
assurance
lab
have
tested
it
as
I
heard
in
low
thousands,
so
a
number
of
devices.
This
is
why
I'm
talking
about
about
deployment
on
scale,
just
because,
when
it's
declarative,
it's
easy
to
test
on
the
huge
numbers,
so
both
of
the
Technologies
are
open
source.
The
cluster
provider
we've
written
is
also
open
source.
So
that's
mostly
what
I
wanted
to
to
tell
you
about
so
two
technologies
provided
between
themselves.