►
From YouTube: OpenCrowbar Docker Nodes [GOOD VERSION]
Description
This OpenCrowbar using Docker video is actually a demo/tour recorded. The focus is setting up Docker nodes for deployment, but it also shows off many of the new features of OpenCrowbar and explains some differences.
A
Cool,
thank
you.
So
we'll
do
the
tour
for
bringing
up
docker
containers
against
our
doctor
admin
using
open,
crowbar
and
I
already
have
the
docker
admin
running.
So
what
I
did
to
start
all
this
stuff
was
I
ran
the
doctor.
Admin
I've
already
done
that
stuff.
It's
really
not
that
exciting.
What
it
does
is.
It
starts
with
a
base
centos
65
container
that
we've
installed
all
the
prereqs
for
so
the
postgres
93
server
Ruby
to
the
rails,
a
lot
of
gems.
A
So
the
first
thing
we
do
is
we
mount
our
working
directory
into
the
container
and
then
the
container
installs
from
that
mounted
share,
and
then
so
it's
actually
just
as
if
you
installed
in
rpm
and
then
ran
through
the
full
install
even
down
to
installing
the
database,
and
then
in
this
one
it
uses
a
script
to
actually
hit
the
api's
to
create
the
admin
note
and
then,
when
it's
done,
it
gets
the
crowbar
converge.
That
is
the
end,
and
now
I've
got
the
crowbar
system.
A
So
this
is
all
the
node
roles
that
are
deployed
in
by
the
annealer
which
look
over
here
like
this,
and
so
what
you've
got
is
this
admin
node
is
deployed
all
of
the
roles
necessary
to
run
crowbar
admin.
It
actually
has
this
no
opera.
This
milestone
role
doesn't
do
anything,
but
we
use
it
as
a
break
point.
So
we
found
it's
really
really
helpful
to
have
these
coalescing
points
that
don't
do
anything
but
the
the
work
ties
to
them
and
that's
really
important
because
of
the
way
we
had
to
deal
with
docker
and
I'll.
A
A
A
So
VMs
VMS
are
like
physical
machines
right
they're,
going
to
pick
C
boot
I'm
going
to
have
a
network
in
them
I'm
going
to
have
control
of
the
OS.
So
every
single
thing
inside
of
the
vm
is
basically
mine
to
control.
All
I
really
have
to
do
to
automate.
Vms
is
talk
to
the
API
and
say
give
me
a
machine
based
on
this
disk
image
and
it
goes
midnight
then
I
can
do
whatever
I
want.
I
can
bond.
The
Knicks
I
can
do
teaming.
I
can
set
the
D
the
dns
entries.
A
I
can
set
my
mind
records.
I
have
a
lot
of
control
inside
of
a
container
and
I
can
install
an
OS.
I
have
to
install
an
OS
if
I
do
it
with
a
container
I.
Don't
have
that
flexibility
and
what
what
happens
is
that
of
the
roles
that
I
can
actually
do
to
bring
up
that
they'll
actually
set
up
once
this
node
finishes,
booting
it'll
go
through
and
start
a
kneeling.
You
know
setting
up
the
clients
setting
up
a
whole
bunch
of
these
pieces.
The
the
containers
don't
need
to
do
most
of
that
stuff.
A
It's
all.
It's
already
baked
in
based
on
your
host
configuration
so
bringing
up
docker
slaves
requires
your
host
to
have
a
degree
of
configuration
that
works
within
the
system,
boundaries
and
then
even
more
tricky
is
it
requires
you
to
have
two
paths
through
your
configuration
deployment
infrastructure,
and
so
what
that
means
is
we
have
to
create
this
new
concept
in
in
crowbar.
A
So
the
idea
here
is
every
one
of
these
things
as
a
role
that
has
to
be
set
up
for
you
to
be
successful
with
a
boot
and
there's
a
role
in
here
called
crowbar,
managed
node
that
is
installed
when
crowbar
breaches.
It's
a
it's
a
no
op
miles
to
enroll,
there's
also
one
for
installed
OS
so
will
boot.
The
discovery
boot
also
goes
all
the
way
through
managed
node,
because
that's
the
point
at
which
you
can
control
it
installed.
A
A
A
But
these
little
wrenches
show
that
you
have
proposals
and
in
the
proposals
you
could
actually,
if
there
were
configurable
items
most
of
the
discovery
stuff
doesn't
have
configurable
items,
you
could
actually
go
in
and
mess
with
it.
Oh
that's
something
I've
show
in
another
another
video
where
you
actually
picked
the
OS,
but
you
know
what's
going
on
right
here,
is
we're
literally
moving
through
its
installed
the
chef
agent
on
this
machine.
It's
actually
got
put
the
Knicks
in
the
right
modes.
A
It's
in
setting
up
the
network
at
the
admin
network
right
now,
and
it's
just
going
to
keep
happily
blowing
through
each
one
of
those.
The
other
thing,
that's
really
fun
is
because
we're
the
way
we're
using
shift
so
low.
We
actually
these
are
logs
of
each
run.
So
when
you
come
back
and
look
at
the
system
you're
going
to
see,
this
runs
not
done
yet
just
about
to
finish.
A
You
can
see
it
actually
doing
the
getting
notifications
as
it
completes
the
work,
so
you
can
actually
drill
down
and
see
exactly
what
happened
on
that
node
and
how
things
are
going
and
what
happened
and
if
there's
a
problem,
you
can
fix
the
error
and
hit
retry
and
get
some
new
data
and
here's
all
ipv6
stuff
we
set
up,
so
it
sets
up
by
pv6
by
default.
Oh
the
other
thing
you'd
be
interested
to
see.
This
is
the
the
new
version
of
network
Jason.
A
So
if
I
want
a
new
network,
I
can
set
up
my
new
network
I
can
attach
it
to
a
default,
so
the
network
is
in
a
scoped
boundary
based
on
deployments
set
up
ipv6.
For
me,
this
is
going
to
be
1g
one
and
add
the
network,
and
then
in
here
you
can
actually
go
through
I.
Think
we're
cleaning
the
screen
up
a
little
bit,
but
you
can
set
the
team
bridge
all
this
stuff
and
then
add
in
whatever
range
is
you
want
and
it
basically
is
building
the
same
stuff.
A
A
A
So
what
what
we've
done
here
is
there's
a
tools,
soccer
slaves
and
then,
if
you
provide
how
many
slaves
you
want
Victor's
testing
with
a
hundred
to
flush
out,
lock,
contention,
issues
and
scale
issues,
I
don't
have
enough
RAM
to
do
that.
Six
is
usually
I.
Usually
my
machines
are
am
constrained
enough
that
it's
6
I'll
start
to
get
some
some
failures,
so
I
hit
enter
and
what
it's
doing
is
its
using
team
ox,
which
is
like
screen
to
bring
up
all
of
the
containers.
So
I'm
gonna
have
to
bring
up
find
another.
A
Yet
another
alright,
so
black
background
is
my
admin
container
purple
background
is
my
six
docker
container,
stacker
slaves
and
blue
is
going
to
be
my
dev,
my
original
devon
fire.
So
if
I
say
dr.
PS
strange,
oh
it
takes
time.
So
one
of
the
things
the
note
in
this
is
that
each
one
of
these
guys
are
the
scripts,
do
not
create
the
docker
containers.
A
Crowbar,
creates
the
docker
containers,
the
scripts
use
the
API
use,
crowbars
API
and
tell
tell
crowbar
so
what
we
did
was
when
in
the
in
the
crowbar
API,
where
you
create
a
node,
it's
not
a
single
action.
We
broke
it
into
pieces,
so
you
create
a
node
and
then
you
actually
go
and
assign
the
role
that
you
want
that
node
to
take.
A
So
in
this
case,
in
a
discovery
node,
you
say,
oh
I'm
want
to
be
a
crowbar
manage
note,
and
that
tells
it
to
bring
up
a
regular
discovery
process
and
sledgehammer
has
all
that
baked
into
it.
The
docker
slave
script.
Doesn't
it
says,
create
note
for
me
but
use
the
docker
Grove
our
docker
node
role,
and
that
actually
has
a
different
dependency
chain
and
installs
different
roles
and
crowbar
knows
how
to
handle
that
milestone
role
differently.
Then
the
crowbar
managed
role,
node
role,
and
but
they
are
both
equivalent.
A
So
there's
a
concept
that
we
introduced
called
provides
where
you
can
say
all
right.
If
I'm
going
to
I'm
going
to
assign
crowbar
docker
roles
to
this
node
and
it's
the
system
is
going
to
treat
it
as
if
it
was
a
crowbar
managed,
node,
actually,
I
think
a
crowbar
installed
OS
role
and
that
lets
us
have
different
system
behaviors
different
role,
graphs,
but
downstream
from
that
treated.
Exactly
the
same.
A
So
if
I
wanted
to
pay,
if
I
want
to
build
a
deployment,
automation
against
OpenStack
or
Hadoop
or
SEF,
or
something
like
that
and
I
want
to
test
with
docker
I,
don't
I
just
build
the
automation
as
if
it
needed
installed
OS
role
and
then,
when
I
bring
up
the
docker
admins.
The
docker
rolls
for
test.
It's
going
to
say:
oh
well,
these
could
do
the
same
thing.
It's
like
an
alias.
A
You
know
I'm
bringing
up
six
nodes.
I
could
do
a
six
node
deployment
and
it's
going
to
take
me
about
three
minutes
start
to
finish.
It
took
me
three
minutes
just
to
get
through
the
discovery
process.
For
a
single
note,
I
haven't
even
stalled.
The
OS
I
can
tell
you
when
I,
when
I
install
the
OS
on
this
VM,
it's
going
to
take
another
ten
minutes
to
complete,
and
this
is
a
is
a
machine
with
16
gigs
of
RAM.
A
A
A
Definition
was
also
another
piece,
because
if
you
look
at
the
graph
that
we're
setting
up
here,
you
know
there's
a
whole
bunch
of
roles
that
are
essential
for
a
vm
of
a
physical
note
that
you
just
don't
even
need,
and
so
we
had
to
have
an
infrastructure
that
could
deal
with
the
fact
that
these
are.
These
are
very
different
workflows
at
the
system
deployment
level,
but
then
would
be
identical
workflows
at
the
application
deployment
level
and
that
that's
a
big
part
of
what
this
ends
up
being.
B
A
A
It
probably
could
use
a
login
client
that
would
that
would
be
something
we
could
add
it
for
now.
We've
we've
kept
them
pretty
minimal
from
a
test
deployment
perspective
and
to
add
it,
you
would
just
make
make
the
logging
client
a
prerequisite
for
one
of
these
other
roles
and
it
would
it
would
just
throw
it
in
okay.
The
thing
that
it
really
doesn't
like
to
have
is
it
is:
it
doesn't
like
to
have
you
set
bind
dns
clients,
there's
certain
things
that
you're
not
on
you
know
the
NT
ntp
isn't
necessary
in
here.
B
A
This
make
sense
yeah,
and
you
know
that
was
one
thinks
crowbar.
You
clever
one
especially
relied
on
pixie.
It's
the
discovery
mechanism,
most
direct
tools
out
there
do
right
razor
cobbler
form,
Foreman's
yeah.
It
does
actually
rely
on
pixie
boot
to
get
images
in
place.
If
you
take
pics
eboot
away
as
an
image
deployment
thing,
you're
you're
actually
going
to
have
trouble
finding
an
alternate
way
to
get
that
image
on,
especially
when
you
you're
not
actually
installing
the
image
you're,
just
you're
running
a
command
that
creates
that
image.
A
A
So
if
you
want
to
have
teamed
nick's
to
support
your
container
and
then
plumb,
your
your
your
docker
bridge
to
those
team
Nix,
that's
that's
where
we're
focused
right,
it's
the
under
it's
the
under
plumbing
of
of
this
and
then,
if
you
give
whatever
docker
management
infrastructure,
you
want
to
add
which
would
be
like
a
cloud
foundry
or
an
open
shift,
pass
yeah
that
you
install
that
layer
and
let
it
let
it
manage
the
containers
coming
and
going.
That's
that's
their
job.
A
Not
yet,
but
we
could
that's,
that's
the
goal,
so
what
we
would
do
is
we
create
a
new
deployment
here
and
then
say:
come
in
I'll
bulk,
edit
them,
so
they
come
they
go
through
and
we
still
we
haven't,
worked
yet
to
optimize
workflows
for
crow
vartha.
A
lot
of
these
things
are
still
using
pretty
basic,
workflows,
okay,
good
and
then
so.
I
can
come
into
my
my
default
deployment
over
here
and
I
could
add
in
say,
Seth
and
then
what
would
happen
is
now
I'd
actually
be
deploying
my
workload.
A
Yeah
we
haven't,
we
don't
have
the
workload
yet
for
for
me
to
demo
what
you
would
install
and
docker,
but
the
idea
the
idea
is
pretty
straightforward:
you
would,
you
would
take
an
application
and
then
you
would
bind
it
clicking.
These
buttons
binds
the
application
in
to
the
dr
node
and
then,
when
you
commit
the
deployment
ultimate
build,
a
new
diploma
will
show
you
what
it
looks
like
from
an
OS
perspective
from
all
right.
So
we
had
a
vm
oops
yeah.
B
A
Couldn't
deploy
chef
because
I
didn't
have
a
SEF
network
on
such
a
dingbat,
so
the
SEF
actually
requires
several
in
the
in
the
system.
So
I
need
to
have
a
roll
named
set
a
network
name
staff
to
create
the
SEF
roll
and
it
has
to
be
in
the
appropriate
scope
boundary,
which
is
this
deployment,
and
I
can
call
it
1g
one.
That's
fine.
B
A
A
I
haven't
had
time
yet
to
give
the
round
trip
Ajax
feedback
on
airs
yet,
but
so
bye.
By
doing
this,
I'm
literally
binding
the
network,
the
SEF
network
in
sefton
fig
into
the
environment
when
I
hit
commit
in
here
it
actually
goes
and
starts
the
annealer.
So
you
get
this.
This
is
the
work
queue
of
what's
going
on.
B
A
I
can
see
you
know
anything,
that's
an
error,
the
things
that
are
there's
ten
worker
nodes,
we're
using
too
late
job
whoops,
that's
already
done
because
it
didn't
have
much
to
do,
and
so
it
literally
goes
in
does
transition
to
do
items
or
things
that
could
be
done,
but
we
just
don't
they're
blocked
based
on
the
the
way
we
we
execute
work,
they're,
not
really
blocked,
it
could
be
done,
but
there
other
things
are
in
transition
and
blocked
items
have
a
upstream,
no
upstream,
no
troll.
That
is
in
to
do.
A
If
that
makes
sense,
right
and
then
you'll
know,
a
lot
of
the
stuff
is
pretty
familiar.
Oh
you
can
actually
see.
I
didn't.
I
didn't
create
an
ipv4
address.
Space
I
only
created
ipv6
address
space,
but
that's
okay
right.
So
it's
not
it's
going
to
say
not
set
because
there's
no
ipv4
address.
If
I
come
over
here,
I'm
going
to
do
an
IPA.
A
What
you'll
see
is
yeah
see
this?
Is
it
put
it
put
the
admin
address
in
here
and
then
it's
also
created
the
ipv6
addresses
that
it's
going
to
use
and
you
can
see
that
if
I
click
over
in
I'm
in
dr.
for
dr.
for
so
this
is
this-
is
the
admin
address
that
we've
we've
installed
and
so
you've
got
both
the
ipv6
address
and
the
ipv4
address.
So
we
we
assigned
it
and
then,
since
chef
is
also
installed
in
here,
see,
should
be
installed.
No
I,
don't
think
we
haven't
done
anything
that
requires
chef.
A
Yet
I
can
tell
you
yep,
so
the
chef
agents
installed
so
we've
actually
installed.
The
chef's
agent
in
here
should
be
able
to
just
run
I'm,
not
able
to
run
chef.
This
is
Sol
chakra
chefs
all
his
shifts.
Oh
gosh
yeah,
so
yeah
literally
the
systems
come
in
and
it's
set
up
chef
solo
and
then,
whenever
we're
still,
we
don't
have
to,
but
we
have
so
much
working
code
and
chef
that
we
just
ported
the
all
the
configuration
and
set
up
into
chef
solo,
but
there's
no
chef
server
with
us.
A
We
just
execute
the
code
that
needs
to
be
executed.
We
SSCP
it
in
and
do
the
execution.
So
if
you
wanted
to
turn
crowbar
off,
you
just
turn
it
off
walk
away,
and
you
know
it
doesn't
need
to.
It
doesn't
need
to
stay
embedded
in
the
environment.
There's
no
agent
overhead.
From
that
perspective,
sir,
like
Ansah.
B
A
A
A
If
we
need
it,
what
will
happen
is
if,
if
the
system
is
powered
off,
the
annealer
will
fail
at
that
one
step
and
that's
and
then
we'll
actually
have
it
blocked
and
you'll
see
that
actually
very
clearly,
and
that's
one
of
the
nice
things
right
anything
that
goes
wrong
in
the
annealer.
You
end
up
with
very
quick
feedback
that
there's
something
that's
happened.
Oh
I
was
going
to
actually
do
the
vm
deployment
over
here
I
added
to
OS
install.
A
This
is
what
it
would
look
like
to
actually
set
the
OS
it's
set
by
default
to
centos
I'll
switch
it
to
a
bunt
too,
and
it
automatically
builds
these
pics
images.
So
you
you
install
the
ISO,
you
provide
the
ISO
and
for
RedHat
Ubuntu
or
centos.
We
automatically
convert
that
into
a
bootable
pixie
image.
So
it's
really.
A
To
have
multiple
os's
and
then,
if
I
go
back
to
this
deployment
I'm
in
proposed
mode,
so
I
can
do
editing
I
can
commit
here.
This
is
always
fun
to
watch.
It's
funny,
Kovar
one!
It's
just
even
more
fun
in
crowbar
too,
because
you
can
see
the
in
dealer.
So
it's
going
to
come
over
here
and
in
the
second
it
will
reboot
it'll
yep
there
you
go
reboots
the
system
so.
A
A
Although
with
the
ability
to
do
jigs
in
worker
queues,
out-of-band
we're
switching
to
use
I'd
rack
direct
api's,
instead
of
having
to
run
them
in
that
weird
loop
through
the
sledgehammer
image.
So
with
that,
we
could
actually
power
systems
down,
because
the
ID
rack
is
still
addressable.
Even
with
the
system
powered
down.
Yeah.
A
A
There's
there's
a
interesting
loop
that
we
can
do
with
I'd
rack
discovery
where
it
would
actually,
you
can
set
up
a
dhcp
server
that
tells
the
ID
rack
to
talk
hit
this
web
server
and
it
gets
its
credentials
from
that
web
server.
It's
a
secure
sort
of
loop
and
it'll
do
that
process
when
the
power
is
when
the
server
is
plugged
in
even
before
it's
powered
on
so
we'd.
Have
we
have
the
ability
to
take
a
rack,
discovered
rack?
A
Have
it
have
the
power
on
servers
off
and
actually
have
crowbar
discover
them,
and
because
we
don't
rely
on
the
sledgehammer
to
actually
be
running,
to
create
the
api's?
You
actually
create
the
systems
and
you
could
then
go
and
say:
oh
I
want
these
to
be.
You
could
do
all
sorts
of
configuration
fry.
A
But
the
doctor
nodes
are
basically
in
here
and
because
I
have
this
screen
going,
I
can
jump
back
and
forth
two
different
things
and
if
I
want
to
experiment
with
my
configuration,
so
I'm
deploying
OpenStack
against
six
nodes
and
I
want
to
bounce
back
and
forth
I
could,
when
she's
going
to
get
really
slow
with
all
this
going
on
it
I
could
go
back
and
say:
here's
dr.
one
grab
its
IP
address,
hey
too
good,
so
I
could
SSH
route.
A
Oops
then
I
always
pick
up
the
64,
so
I
you
know
I
can
I
can
always
do
what
I
expected
to
do.
This
is
a
little
bit.
Mind-Blowing
I
am
in
a
centos
docker
container
on
my
death
machine
sshe
into
and
a
bunt
to
container
running
on
my
same
machine
and
so
I
could
sit
need
to
be
in
the
right,
fought
and
the
right
path
to
make
this
easy.
I
can
touch
food
text
over
here
and
then,
of
course,
if
I'm
get
back
to
my
doctor
container
number
one
we're
here.
A
There's
my
food
tax
right,
it's
a
saint,
just
ssh
into
the
machine,
but
you
know
now,
if
I'm
actually
troubleshooting
and
working
on
a
deployment
trying
to
get
everything
right,
one
I
don't
have
I,
have
a
six
node
system
without
six
VMs
which,
even
on
our
32
gig
lap,
32
gigs
of
ram
laptops,
is
pretty
onerous
and
heavy
to
do.
But
I
set
up
a
six
node
system
in
a
couple
minutes
less
than
that,
and
then
I
have
really
super
easy
access
to
the
consoles
of
all
the
systems.
A
A
So
I
touch
bar
text
this
time
this
guy
hears.
I'm
back
to
my
admin
note.
So
let
me
exit
now
I'm
back
to
my
admin
note.
So
I
CD
outfit
opt
up
and
crowbar
core
and
you
will
see
I'm
going
to
put
it
or
bar
text.
Oh
sorry,
this
is
a
different
machine.
This
is
my
actually
be
in
the
right
you
got
to
be
in
the
right
system.
Make
this
stuff
work.
A
A
A
I
didn't
switch
into
core
there.
We
go
so
here's
bar
text,
so
if
I,
if
I
make
changes.
So
if
I
go
into
chef,
tart
cookbooks,
and
so
if
I
was
to
go
into
some
out
when
we
added
Burke's,
which
is
really
cool,
so
Burke's
lets
us
pull
code
in
from
repos,
so
that
every
time
when
you
run
the
code,
you
can
actually
have
very
fine-grained
control
over
which
cookbooks
you
use
to
do
the
deployments.
It's
not
all
bundled
into
crowbar
anymore.
A
That
was
a
recent
recent
change,
but
if
I
come
into
the
network,
se
hace.
This
isn't
a
good
example
because
its
core,
but
if
I
go
into
the
recipes
here
and
start
messing
if
I
changed
the
default
recipe
here,
the
switch
config
recipe
here,
the
next
time
it
runs
that
node
roll.
It
actually
picks
that
change
up
right
away.
So
you
don't
have
to
upload
it
or
anything
like
that.
It
I'm
editing
in
my
dev
environment.
A
A
So
there's
no
upload
or
sink
or
anything
like
that.
You
write
code,
you
save
it,
you
retry
the
node
role
and
it
runs
at
it.
So
between
all
these,
these,
these
you
know
docker
containers
and
some
of
the
ways-
and
this
is
a
dot-
is
a
aspect
to
dr.
containers
being
able
to
do
this
fast
mounting.
We've
really
tried
to
optimize
the
development,
the
DevOps
development
experience.