►
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
Intro: OpenStack SIG - Chris Hoge, OpenStack Foundation & David Lyle, Intel
SIG OpenStack coordinates the cross-community efforts of the OpenStack and Kubernetes communities. This includes coordinating improvements to and documentation of the OpenStack cloud provider implementation in Kubernetes as well as supporting efforts to deploy OpenStack itself using Kubernetes. Attend this session to learn more about the SIG's mission, recent accomplishments, and future plans.
To Learn More: https://sched.co/Grbk
A
A
So,
just
to
start
off
with
how
many
of
you
are
familiar
with,
OpenStack
I
mean
presumably
you're.
So
you
don't
know
you
don't
really
need
to
see
the
architecture
diagram
of
what
we're
talking
about
here,
but
sometimes
it's
nice
to
just
you
know.
Talk
about
OpenStack
is
programmable
infrastructure
and
there
are
a
bunch
of
different
projects
that
allow
you
to
accomplish
that
and
they
all
kind
of
work
together.
A
Okay,
so
with
that
introduction
of
kind
of
what
is
sig,
what
is
OpenStack
about
what
is
sig
OpenStack
about,
and
this
is
actually
an
organization
that
spans
both
both
the
kubernetes
organization
and
CN
CF
and
the
OpenStack
foundation
in
the
OpenStack
foundation.
We
would
call
this
cig
kubernetes
but
they're
the
same
group,
but
the
same
leadership
with
all
of
the
same
goals
and
our
goal
is
to
coordinate
the
cross
community
efforts
of
the
OpenStack
in
communities
communities.
A
How
can
we
make
that
the
best
experience
for
you
possible
as
a
service
provider
for
kubernetes,
an
example
of
this
is
you
might
want
to
have
multi-tenancy
and
some
sort
of
authentication
that
is
backed
by
that
is
backed
by
Keystone
or
you
might
have
a
large
number
of
storage
systems
available,
and
you
want
to
be
able
to
plug
you
know
to
plug
those
storage
systems
and
the
communities
using
cinder
without
pulling
in
all
of
the
rest
of
the
OpenStack
infrastructure.
This
is
what
we
mean
by
saying
that
OpenStack
is
a
service
provider
to
kubernetes.
A
We
can
also
talk
about
it
as
a
collection
of
application
to
run
on
top
of
kubernetes,
and
so
for
those
of
you
who
are
familiar
with
the
SE,
the
OpenStack
helm
project.
This
is
this
is
a
project
that
starts
with
a
kubernetes
at
the
infrastructure.
Layer
has
some
OpenStack
services
sitting
next
to
that,
like
Keystone,
for
example,
and
then
it
uses
helm
to
deploy
a
full
OpenStack
cluster.
On
top
of
that,
all
of
this
information
is
available
within
our
within
the
kubernetes
community
gear
repository
along
with
all
with
information
about
the
other
six.
A
So
last
year
we
published
a
paper
that
kind
of
talks
about.
You
know
that
that
tries
to
give
a
visualization
of
what
this
looks
like
and
and
how
this
works
in
practice.
And
so,
if
we
start
at
the
left
of
the
diagram
and
then
we
move
our
way
over
to
the
right
of
it
at
the
base,
let
the
base
layer,
you
can
see
an
open,
a
small
seed,
node
running
OpenStack
ironic,
which
you
can
use
to
control
your
bare
metal.
A
Once
you
have
that
initial
seed
established,
you
can
use
OpenStack
to
install
the
either
OpenStack
or
kubernetes
or
with
on
top
with
your
bare
metal
right
there,
and
this
is
being
done
actively
right
now
with
projects
such
as
such
as
airship,
which
is
working
towards
using
OpenStack
ironic
as
a
spare
metal
provisioning
layer
to
install
both
OpenStack
and
kubernetes
services.
And
the
interesting
thing
is,
is
that
you
know
you
can
actually
have
the
seed
node
be
OpenStack
installed.
You
know,
through
a
bunch
of
different
container
projects.
A
There
are
any
number
of
upstream
projects
that
allow
you
to
run
OpenStack
and
kubernetes
or
in
containers
in
general,
such
as
container
run
times.
You
know
in
docker
and
Alexei
container
orchestration
using
you
know
things
like
ansible
or
kubernetes,
and
then
container
images
such
as
those
built
by
Cola
or
the
Loki
project,
or
the
LXE
images
both
for
OpenStack
ansible
and
so
right
there
at
the
infrastructure
layer
layer.
Once
you
have
your
bare
metal
in
place,
you
can
drop
in
at
OpenStack
and
you
can
drop
in
and
OpenStack
installation.
A
And
then
you
have
your
cloud
on
the
far
right
and
you're
running
all
of
your
OpenStack
services.
There
plugging
into
that
is
kubernetes
sitting
on
top
of
it
and
it's
using
an
OpenStack
cloud
provider
that
lets
the
kubernetes
installation
talk
to
the
underlying
cloud.
Much
in
the
same
way
that
you
could
talk
to.
A
You
know
an
azure
or
a
Google
or
a
or
an
Amazon
Cloud,
all
the
same
api's,
and
it
gives
kubernetes
introspection
on
things
that
are
running
inside
of
that
are
running
inside
of
your
cloud.
So
it
has
more
information
about
nodes.
It's
able
to
it's
able
to
allocate
notes
directly
through
cloud
requests.
It's
able
to
allocate
load
balancers
from
services
like
Octavia.
You
can
attach
storage
to
it
through
cinder.
You
know
you
kind
of
like
you
get
all
of
these
things
and
they're
built
in
you
know
you
know
versus
without
the
cloud
provider.
A
A
It's
within
the
kubernetes
organization-
and
this
is
part
of
a
larger
effort-
that's
in
collaboration
with
cig
cloud
provider
where,
where
we're
trying
to
manage
the
cloud
controller
manager
and
the
clouds
that
offer
providers
to
plug
into
that,
so
that
we
can
create
a
consistent
experience,
no
matter
what
sort
of
cloud
you're
running
so
so
open
stack,
just
isn't
interested
in
providing
you
the
best
experience
for
running
kubernetes
on
your
cloud.
We're
actually
deeply
invested
in
making
sure
that
you
can
run
kubernetes
and
have
a
great
experience
on
any
cloud.
A
We
also
have
a
standalone
keystone,
authentication
and
authorization
webhook
if
you're.
If
you
are,
if
you
want
to
have
keystone
act
as
a
front-end,
kubernetes
are
back.
This
is
a
really
nice
way
to
plug
that
in
and
have
that
work
and
kind
of
you
know
get
some
light
multi-tenancy
with
in
kubernetes
the
great
thing
about
this
is
you
know
you
get
a
light
multi-tenancy
there.
A
The
way
that
we
expect
it
to
do,
we
expect
it
to
one
of
the
interesting
things
is
OpenStack
was
one
of
the
first
cloud
providers
that
was
reporting
and
and
results
with
the
cloud
provider
back
to
back
to
the
kubernetes
test
grid,
and
it
was
one
of
the
first
providers
to
kind
of
meet
the
minimum
called
qualifications
that
if,
if
changes
to
kubernetes
broke
OpenStack,
then
that
would
say
hold
up
a
release.
You
know
so
this
is
you
know
it's
pretty
important
and
OpenStack
has
actually
been
hard,
has
been
a
perfect
cloud
provider
for
kubernetes.
A
You
know
for
a
very
long
time,
which
brings
us
to
the
next
point,
which
is
the
other
one
of
the
other
major
things.
We're
working
on
right
now
is
removing
the
entry
provider
from
kubernetes
kubernetes,
and
so
the
cloud
provider
actually
got
its
start
along
with
with
the
number
of
other
cloud
providers
as
entry
code,
that's
running
inside
of
kubernetes,
and
and
and
provides
some
implementation
of
some
different
interfaces
inside
of
there,
but
that
actually
presented
a
fairly
large
problem.
A
We're
changing
code
if
you
ran
into
a
problem,
became
harder
because
you
have
to
go
through
the
process
of
the
larger
communities
community
validating
code
that
actually
only
impacts,
one
provider
and
the
experts.
Who
would
be
there
to
review
that
code
may
not
actually
be
there.
They
may
not
be
able
to
provide
a
proper
review
to
ensure
that
the
code
works
by
pulling
this
code
out
of
tree
and
having
an
external
provider,
we're
able
to
focus
the
efforts
of
the
sig
on
maintaining
that
with,
without
in
some
ways,
dragging
the
larger
critics
community
into
this.
A
So
this
is
a
project,
that's
being
managed
by
the
cig
cluster
life
cycle
group
and
it's
it's
meant
to
be
a
kubernetes
like
API,
for
you
to
manage
the
cluster
that
your
crew
vanadis
runs
on
top
of,
and
so
it
will
allow
you
to
specify
kubernetes
cluster
to
drop
on
a
cluster
with
the
number
of
control
nodes
and
the
number
of
worker
nodes
and
manage
the
installation
and
life
cycle
and
eventually
deletion
of
that.
And
so
that
is
that
work
is
very
much
in
its
alpha
stage.
A
Cube
spray-
you
know
to
you
know
to
you
know
to
have
like
you
know,
you
know,
you
know
a
a
consistent
way
to
deploy.
Kubernetes
clusters
with
the
communities
like
API
is,
is
pretty
exciting
and
I.
Think
it's
some
of
the
most
important
work.
That's
going
to
be
happening
in
the
community
communities
community
in
the
next
year
and
I'm
really
happy
that
we
have
the
beginnings
of
a
provider
for
that
in
place.
A
A
A
So,
to
get
set
up
with
using
this
and
using
the
external
provider,
you
have
to
basically
provide
a
configuration
for
this
and,
at
its
simplest
level,
the
configuration
is
shown
in
the
global
parameters
is
a
username,
a
password
tenant
domain
IDs
and
an
auth
URL,
pretty
standard
way
of
authenticating
to
OpenStack
clouds.
You
know
the
this
is
the
basic
set
that
gets
you
in.
A
This
is
taken
straight
from
the
docks,
the
dot,
dot,
dots,
they're,
just
that's
a
location
somewhere
on
your
disk,
where
the
where
the
binary
is
for
it.
You
set
the
external
crowd
provider
to
be
true,
you
set
the
name
for
it.
You
say
where
your
cloud
config
is,
and
you
and
you're
bringing
up
your
cluster.
There
is
a
note
that
that
you
do
need
to
create
cluster
role
in
cholesteral
bindings
so
that
the
provider
can
talk
to
the
API
and
there's
code
upstream.
That
allows
you
to
do
this.
A
So
moving
on
to
the
cluster
API
provider,
OpenStack
and
right
now,
it's
more
of
a
comple,
concrete
implementation
and
a
true
provider.
Part
of
this
is,
is
the
the
api's
are
being
sorted
out
right
now
and
so
there's
there's
a
bit
more
velocity
on
changing
the
api's
and
testing
implementations
for
each
provider
to
kind
of
you
know
kind
of
own
that
implementation
and
how
it
works.
A
A
So
the
note
there
says
that
it's
still
in
the
prototype
stage,
while
we
get
feedback
in
the
API
types
and
it's
there
to
experiment
that
with
the
API
and
demo
its
abilities,
but
expect
this
to
stabilize
pretty
quickly
I
this
morning
they
they
had
a
meeting
about
this
and
they're,
basically
trying
to
shake
out
the
last
details
of
the
API
they're,
not
quite
ready
to
call
it
stable.
Yet
because,
as
many
of
you
know,
once
you
once
you
have
stated,
an
API
is
stable.
A
So
here's
a
little
bit
of
an
overview
of
which
is
taken
from
the
the
cluster
API,
the
the
cluster
API
repository.
You
know
that
just
that
just
kind
of
shows
you
the
structure
of
how
this
works.
What
do
you
exciting
things
about?
This
is
because
it's
a
plug-in
to
the
kubernetes
api
you
can
actually
through
the
application,
cube
cluster
and
a
cube
control.
You
canyou
can
query
the
state
of
this
and
you
can
and
you
can
you
can,
you
know,
update
it.
You
can
expand
it.
You
know,
there's
a
there's,
a
lot
of
management.
A
A
We've
had
full
four
full
releases
of
the
external
provider,
and
these
are
synced
up
with
the
kubernetes
releases
now,
so
you
can
actually
go
and
download
the
binaries
and
start
using
them
right
away
if
you're,
interested
and
and
they're
matched
up
to
the
kubernetes
releases,
and
so,
if
you're
running
a
1.11
cloud,
you
can
download
the
provider
for
that.
If
you
are,
if
you've
just
stood
up
1.13
cloud,
you
can
run
the
provider
for
and
in
and
that's
all
ready
to
go.
A
A
Clouds
like
next
hosts
they'll
stand
up
a
criminalities
installation
for
you
on
your
on
your
cloud
without
any
trouble
and
Magnum
is,
is
the
the
project
that
CERN
is
using
right
now
to
manage
all
of
their
research
kubernetes
clusters.
Last
last
time,
I
talked
to
the
researchers
there.
They
had
over
300
tenant,
isolated,
Corona,
DS
installations,
doing
research
for,
for
you
know,
for
high
energy
physics
today,
and
the
number
keeps
growing.
So
it's
been,
it's
been
a
really
fantastic
demonstration
of
the
power
that
you
get
running.
A
You
know
combining
up
a
stack
in
kubernetes
soon,
we've
also
published
an
open
stack
in
containers
white
paper.
You
can
go
read
that
that
was
a
lot
of
work
done
by
this
cig.
Many
of
you
who
were
who
have
been
part
of
the
sig,
contributed
to
that
and
I
wanted
to.
Thank
you
for
that
and
that's
something
that
we're
gonna
be
maintaining
in
the
coming
months.
Also
and
updating
as
the
state
of
the
art
changes.
A
A
These
are
in
progress
and,
in
fact,
will
probably
be
one
of
the
first
cloud
providers
taken
out
of
the
tree.
This
is
in
part
because
we
went
we've
written,
really
clean
code
where
there
are
interfaces,
we're
we,
we
use
interfaces
where
they're,
where
they
are,
and
we
don't
depend
upon
a
lot
of
internal
api's
or
have
a
lot
of
internal
constants
and
other
things,
and
so
we're
pretty
close
to
being
able
to
pull
a
lot
of
this
out.
There's
just
some
work
that
needs
to
be
done
within
the
kubernetes
codebase
itself.
A
What
that
means
is
anything
that's
in
kubernetes
staging
our
API
is
within
current
IDs
that
won't
be
changed.
Anything
non
staging
is
an
internal
API
that
the
current
ID
scheme
feels
like
that
they
can
change
without
impacting
a
project
and
so
identifying
those
and
moving
those
moving.
Those
in
a
way
that
doesn't
break
the
project
is
it's
in
progress,
and
this
is
this.
A
That
is
maybe
looking
at
creating
a
cinder
light
which
will
remove
some
of
the
dependencies
like
on
rapid
mq
and
will
and
will
give
you
a
way
to
access
all
of
the
drivers
that
cinder
has
without
you
know
having
you
know
through
kind
of
a
lighter-weight
installation,
and
so
that's
some
interesting
work
that
people
have
been
talking
about
them.
I
build
out
the
robust
cluster
API
implementation
and
also
build
a
common
credential
library,
as
we
mentioned
before
so
in
general.
A
If
you
want
to
get
involved,
we're
involved
on,
we
have
the
kubernetes
lac,
cig
OpenStack,
many
many
of
you
are
in
that
and
it's
where
you
can
get
information
on
things
that
are
happening
right
now
and
you
can
also
ask
questions
and
you
know,
and
collaborate
with
folks.
We
also
have
a
Google
Group
kubernetes
SIG
OpenStack,
which
is
a
low
traffic
mailing
list.
A
B
C
C
B
So,
with
respect
to
the
API
project,
I
can
clearly
envision
multiple
clusters
built
on
VMs,
all
of
which
have
access
to
OpenStack
services
kind
of
makes
sense.
You
know
nice
tenancy
model.
Do
you
also
intend
or
think
about
supporting
multiple
disjoint
bare
metal
clusters
carved
out
of
the
larger
OpenStack
deployment
question
so.
A
That's
I
think,
that's
it
I
think
that's
a
great
question,
and
so
the
answer
is
I.
You
could
do
that
right
now
with
caveats
and
I
like
ironic
as
a
driver
for
Nova,
because
when
you
use
it
as
a
Nova
driver,
you
get
all
of
the
benefits
of
a
virtualized
cluster.
You
get
them
multi-tenancy
you
get
network
control,
you
get
control
of
your
images
that
you
deploy
so
I.
Think
it's
a
very
nice
solution.
A
The
main
caveat
is
VMs
behave
differently
than
bare
metal,
largely
from
the
point
of
view
of
resiliency
and
and
capacity
we're
in
a
VM.
You
in
many
ways
can
assume
you
have
infinite
capacity
and,
and
your
capacity
is
limited
by
how
much
you're
willing
to
spend
on
your
credit
card
we're
on
bare
metal.
You
don't
you
have
your
data
center
is
the
size
that
it
is
and
and
if
a
node
goes
down,
you
don't
necessarily
want
to
shut
down
the
entire
cluster.
A
Because
of
that,
you
know
you
may
not
be
able
to
expand
the
way
you
want,
and
so
there
may
need
to
be
additional
information
inside
of
the
provider
to
tell
you
that
you're
running
on
bare
metal
and
to
account
for
the
instances
where
it's
okay
to
not
have
a
full
cluster.
That's
that's
standing
up
where
it
so
you
know
it
deals
with
the
resiliency
and
it
may
be
a
different
way.
A
From
all
over,
it's
it's
I
mean
OpenStack
is,
is,
is
fairly
widely
deployed
within
the
kuru
Nettie's
world,
we're
seeing
a
lot
of
effort
coming
from
the
telecom
side,
and
so,
if
you
look
at
the
OpenStack
airship
project,
which
is
one
of
the
top-level
projects
within
the
OpenStack
foundation,
that's
eight!
That's
the
project
that
18
T
is
looking
at
using
to
deploy
their
5g
networks
and
and
they're,
using
a
hybrid
OpenStack,
kubernetes
infrastructure
layer
that
will
then
drive
their
OpenStack
installation
and
other
application
installations.
A
On
top
of
that,
you
know
and
we're
seeing
you
know,
places
like
SK,
Telecom
and
Ericsson,
who
are
also
interested
in
this
project.
So
that's
one
place
that
we're
growing
out
into
scientific
computing
is
a
place
that
we've
been
in.
We've
we've
been,
you
know,
we're
in
a
lot,
but
really
we're
we're
in
industry
all
over
the
place.
It's
you
know,
and
it
you
know
when
you
want
to
run
hybrid
workloads,
it
winds
up
being
a
pretty
good
solution.
So.
B
A
A
D
Give
a
chrome,
AT&T
question
around
the
cadence,
so
kubernetes
releases
come
a
lot
faster
and
let's
say
you
have
an
open,
stack
release
and
you're.
It's
relatively
old.
You
probably
keep
it
stable
in
that
sense.
What
are
the
challenges
with
these
new,
like
the
cloud
provider
aspect,
for
example,
in
terms
of
compatibility
where
kubernetes
is
moving
so
much
faster
and
OpenStack.
A
So
so
that's
a
good
question.
You
know
regarding
you
know,
with
the
cadent,
the
kubernetes
cadence,
that's
moving
so
much
faster
than
the
OpenStack
released
cadence.
So
one
of
the
one
of
the
guiding
principles
of
the
OpenStack
community
has
has
been
to
only
deprecated
and
remove
API
is
when
it's
absolutely
necessary,
and
so
you
actually
haven't
seen
a
lot
of
churn
and
how
the
OpenStack
API
is
work.
We
made
me
add
new
features
as
time
goes
on,
but
one
of
the
very
basic
things
that,
like
you,
know,
control
of
nodes.
A
But
even
then
we
have
support
for
both
within
the
cloud
provider
and
within
with
it
within
gopher
cloud.
So
we
really
don't
see
unless,
unless
we
really
start
taking
advantage
of
advanced
features
like
say
Barbic
had
you
know,
key
secret
management,
there's
a
strong
possibility
that
we're
gonna
have
Barbican
backed
secret
management
for
kubernetes
through
the
cloud
you
know
through
the
cloud
provider.
That
instance
you
you.
A
If
you
want
those
things,
you're
gonna
need
to
be
running
into
our
cloud,
but
you
know
that
said
you
know
we
encourage
you
to
upgrade
your
OpenStack
clouds
and
that's
being
a
big
working
point
for
us
kind
of
once.
You
get
up.
I
think
to
Newton
upgrades
become
a
lot
easier
because
the
I
mean
that
was
kind
of
the
point
where
we
were
able
to
start
doing.
In-Place
upgrades-
and
you
know
you
know
really
supporting
that
from
a
community
standpoint
him.
So
if
you
can
upgrade
your
cloud,
do
it,
but
it's
it's!
It's
a
trick.
C
Can
you
expand
a
little
bit
more
around
the
authentication?
The
are
back
because
it
seems
like
you
have
our
back,
that
is
from
the
infrastructure
layer
that
needs
to
talk
to
OpenStack,
to
create
more
volumes
for
persistent
bombers,
etc.
But
then
you
might
not
the
same
users
might
not
be
the
same
that
to
interact
with
a
kate's
API
right
right.
You
might
want
to
have
like
two
different
authentication
methods
and
right.
A
That's
a
completely
different
mechanism
that
goes
through
a
web
hook,
and
so
you
can
actually
create
the
users
within
Keystone
that
you
want
to
map
to
your
Keystone,
our
back
controls
and
those
are
completely
independent
from
the
access
controls
that
are
happening
for
controlling
the
the
cloud
infrastructure
below
it
and
and
so
they're.
You
know,
and
you
can
actually
even
point
you
can
even
point
kubernetes,
two
totally
different
Keystone
server.
A
If
you
wanted
to
create
that
extra
layer
of
isolation
and
maybe
run
like
you,
know,
keystone
service
on
top
of
OpenStack,
so
so
the
the
answer
is
that
they're
strongly
decoupled
and
you
know
and
they're
you
can
pretty
much
use
them.
However,
you
want
to
the
cloud
provider,
doesn't
give
you
our
back
control
straight
out
of
the
box,
it
just
it
just
talks
to
Keystone
for
authentication
and
authorization
for
control.
The
cloud.
C
A
A
D
A
Right
now
we're
looking
for
volunteers
who
would
like
to
nominate
themselves
or
nominate
somebody
else
to
be
one
of
the
league's
one
of
the
one
of
the
co-leads.
If
you're
interested
in
doing
this,
you
can
respond
to
the
mailing
list
and
to
the
posting
that's
there
and
you
know
we
would
love
to
have
your
participation.
The
workload
is
pretty
much
running
meetings
and
you
know
running
things
like
this
and
just
kind
of
coordinating
efforts
through
across
the
twin
SIG's.
A
A
All
right,
well
with
that,
we
will
end
the
session
and
I
want
to
thank
everybody
for
joining
us
today
and
Owen
if
you're,
if
you're
interested
in
the
the
sig
cloud
provider
or
the
sig
cluster
lifecycle,
you
missed
the
intro
to
this
is
a
cluster
lifecycle.
This
today,
but
sig
cloud
provider
intro
is
tomorrow
and
deep.
Dive
is
on
Wednesday
and
CID
cluster
lifecycle,
deep
ciphers
long,
deep
dive
is
also
later
this
week.
So
thanks.