►
From YouTube: Kubernetes on BOSH for the Enterprise
Description
Kubernetes and BOSH are key open source technologies in Pivotal Container Service (PKS). Merlin Glynn, Technical Product Manager at VMware, gives an overview of KuBO and explains how operators can deploy and manage Kubernetes clusters on vSphere at scale with BOSH.
https://cloud.vmware.com/pivotal-container-service
A
Hi
I'm
Roland,
Glen
I'm,
a
technical
product
manager
to
VMware,
and
this
lightboard
session
we'd
be
talking
about
Kubo,
which
is
an
acronym
stand
for
two
key
components:
kubernetes
and
Bosch.
So
we're
talking
about
kubernetes
powered
by
Bosch
and
we're
really
going
to
put
an
emphasis
on
that
being
enterprise-grade.
A
A
Software
system
like
kubernetes,
in
addition
to
the
CPI
Bosch,
has
another
component
called
a
Bosch
agent,
and
the
agent
is
something
that's
going
to
be
embedded
in
every
VM.
That
Bosch
is
going
to
deploy
to
instantiate
our
kubernetes
system,
and
this
is
gonna.
Allow
us
to
do
a
couple
of
things
we'll
talk
about
when
you
get
to
the
day
two
side
of
this
so
day,
1
I
need
to
deploy
kubernetes
and
that's
typically
gonna.
A
Be
done
by
a
platform
operator
now
this
is
this:
is
who's
going
to
be
responsible
for
deploying
kubernetes
and
keeping
it
functional,
so
the
developers
and
the
application
teams
that
need
to
publish
applications
on
their
on
their
kubernetes
clusters
can
always
have
that
available.
So
platform,
ops,
our
platform
operator.
A
Now,
the
way
the
platform
operator
is
gonna
actually
deploy
kubernetes
new
Bosch
he's
got.
It
he's
got
to
give
two
things
to
that:
boss,
vm
two
key
things.
The
first
of
those
key
things
is
called
a
deployment
manifest,
and
it's
usually
in
yeah,
Mille
or
actually
isn't
you
know,
and
that
deployment
manifest
is
gonna
tell
bosch
exactly
what
it
has
to
deploy
or
describe
how
it
has
to
deploy
kubernetes,
how
many
instances
of
the
kubernetes
components?
What's
the
kubernetes
cluster
name?
A
A
Now
a
Kubo
Bosch
release
is
going
to
include
things
like
the
actual
kubernetes
components,
so
the
software
to
be
deployed
and
some
instructions
on
how
to
deploy
it,
and
it's
usually,
it's
usually
distributed
in
a
tgz
like
a
tarball
format,
and
so
when
you
take
the
Bosch
release
and
the
deployment
manifest,
the
platform
operator
is
given
to
Bosh.
Bosh
can
now
via
the
CPI.
Go
ahead
and
create
a
kubernetes
cluster
for
us,
so
we'll
just
call
it
Kate's,
hey
in
our
non-descriptive
Manish,
manifest
in
our
Y
amel.
We
could
tell
it
things
like
hey
Bosch.
A
We
want
to
have
multiple
master
nodes.
You
know
we
want
to
have
some
high
availability.
We
don't
want
to
have
a
single
man,
single
master,
node
scenario.
We
also
want
to
have
multiple
NC
Dino's,
because
sed
is
the
key
value
pair
that
really
lets
kubernetes
function
and
scale
its
requirement.
So
we
might
want
to
deploy
three
of
those
because
it's
a
majority,
Norg
majority
node
quorum
set
means
we
need
to
have
an
odd
number
and
multiple
to
make
sure
that
it's
available
and
finally
the
workers.
A
You
know
the
guys
that
are
actually
gonna
be
running
the
cubelets
in
running
that
could
the
pods
in
the
containers
we
might
want
to
deploy
three
workers
in
our
cluster.
So
all
these
things
we
started
told
boss
how
we
wanted
to
do
this
or
that
the
variables
that
we
wanted
to
do
inside
of
a
deployment
manifests
and
Buzz
boss
used
the
Kubo
release
to
have
all
the
pieces
and
components
to
be
able
to
deploy
kubernetes
on
top
of
our
given
I
ass.
A
So
that's
a
good
day.
One
story
and
discussion
of
what
would
Bosch
can
give
the
enterprise
platform
operator
to
make
sure
he
can
deploy
kubernetes
at
scale.
But
let's
talk
about
day
two.
So
after
I've
got
one
of
these
things
deployed
what
are
some
of
the
day
to
ops
that
I
might
need
to
take
into
consideration.
So
one
of
those
is
gonna
be
rebuild
or
repair.
A
And
this
is
the
notion
that
if
something's
unhealthy
and
my
kubernetes
cluster
I
want
to
be
able
to
recover
it,
I
want
to
be
able
to
make
sure
that
it
stays
up
and
available
so
that
my
applications
are
running
on
my
developers
accessing
it
can
maintain
access.
Now,
remember
we
mentioned
the
boss
agent,
the
boss
agent
gets
packed
into
each
one
of
these
VMs
and
one
of
the
key
things
that
it
does
is
it
reports
the
health
status
of
all
all
of
the
VMS
inside
of
our
kubernetes
deployment.
A
So
if
one
of
these
VMs,
it's
like
a
master
VM
were
default
for
whatever
reason,
the
agents
actually
gonna
make
Vash
aware
that
and
Bosch
will
try
to
automatically
repair
that
master
agent
either
by
trying
to
repair
the
software
I
the
cubelets
running
on
the
nodes
or
actually
rebuilding
the
nodes,
so
boss,
she's
gonna,
do
this
automatically
for
us.
So
that's
a
really
great
day
to
capability
of
keeping
the
kubernetes
cluster
is
healthy.
A
Another
day
to
op
that
we
might
run
into
the
scale
and
scale
can
take
two
cotton.
You
know
two
ways
of
viewing
this
and
operate.
It
may
need
to
scale
a
given
kubernetes
cluster
up
because
of
the
workload
demands
on
it
are
increasing,
so
in
that
case,
platform
operator
could
just
simply
modify
the
manifest
to
tick
the
instances
of
workers
from
three
to
five.
A
Tell
boss,
hey
here's,
my
new
manifest
for
that
deployment
and
boss
will
actually
upgrade
in
place
at
deployment
and
scale
the
workers
up
for
so
it
gives
us
a
pretty
nice
way
and
consistent
way
to
do
our
scaling
inside
of
a
kubernetes
cluster.
Another
way
to
look
at
this
is
what,
if
I,
have
more
teams
at
any
two
on
board,
so
I
may
need
to
deploy
another
K,
its
cluster
or
multiple
K.
Its
clusters
and
I
also
want
to
partition
multiple
tenants
in
them.
A
So
I
want
to
use
namespaces,
which
is
a
kubernetes
construct
to
give
us
multi-tenancy,
so
I
might
add
a
namespace
one
in
namespace
two
and
therefore,
when
I
want
to
on
board
the
teams.
Here,
I
don't
have
to
have
to
give
a
dedicated
kubernetes
instance.
I
can
actually
have
teams
that
are
working
on
shared
micro
service
projects,
share
the
same
kubernetes
api.
A
A
A
What
changes
who
committed,
what
changes
from
a
platform
operating
perspective
to
make
our
kubernetes
cluster
do
this
scale
or
create
in
an
edition
boss,
will
actually
log
all
these
operational
tasks
that
are
being
taken
place
so
that
we
can
get
some
level
of
compliance
and
auditing
to
see
who's
actually
taking
operator
actions
against
these
kubernetes
clusters.
So
really,
when
you
look
down
or
look
at
it,
boss
isn't
just
a
tool
to
deploy
or
do
day1
activities,
but
it's
very
much
a
tool
to
do
day2
activity
performing
all
of
these
activities
across
both.