►
From YouTube: OpenCrowbar Design Meeting 2014-6-25
Description
Design meeting discussing Ready State Wizard, IPMI design and changes, networking, CI workflow
A
Hello,
everybody
and
welcome
to
the
jun
25th
former
design
meeting
we've
got
should
a
lot
of
exciting
stuff
talk
about
so
I
know
it's
not
everybody
consenting
the
whole
time.
So,
let's,
let's
jump
in
sorry
any
any
additional
items
dad.
A
A
B
The
big
problem
with
Adam
actually
no
Adam's,
not
a
problem,
Oh
stickin
being
able
to
run
a
red
hat,
but
we
could
squeeze
Red
Hat
or
been
tossed
onto
an
atom.
A
A
A
A
A
A
I
got
this
is
a
dev
system,
so
it's
got
bogus
nose
in
it
in
weird
States.
This
is
good
love,
yes,
but
I
got
really
frustrated
with
doing
demos.
Where
do
a
demo?
You
actually
have
to
go
in
to
create
a
network
if
grid
I
meant
have
to
assign
the
nodes
into
the
deployment
using
both
get.
It
is
to
pick
them
and
then
and
then
go
into
the
deployment,
commit
the
deployment
or
add
the
roles
commit.
A
You
be,
we
have
a
barrack
so
from
the
wizard,
which
I
think
we
would
put
under
cover
the
nodes
menu.
You
get
to
pick
up
the
deployment
that
you
want
to
add.
So
we're
going
to
call
this
community
call
community.
You
pick
the
conduit,
you
want
it
attached
to
ones.
You
want.
Usually
it's
safe
to
fault.
You
pick
the
range
that
you
want,
so
this
is
probably
range
public.
Is
of
these
good
one
for
this.
A
So
it
doesn't
ask
you
the
question,
but
it
will
do.
Is
it's
going
to
create
a
network
name,
the
same
thing
as
the
deployment
attached
to
the
conduit?
It's
going
to
create
this
range
with
these,
with
these
eye,
peas
in
it,
I'll
show
you
that
have
all
that
stuff
working
very
nice.
So
there
are
enough
information,
yep.
B
A
B
Yeah
I
mean
the
idea
is
we
want
to
be
able
to
find
networks
and
ranges
and
whatnot
without
necessarily
automatically
giving
systems
addresses
on
those
ranges.
So
the
only
ranges
that
systems
will
automatically
get
addresses
down
to
on
discovery
is
host
and
you
really
shouldn't
and
that
what
it
should
pretty
much
stay
in
place.
Okay,.
A
B
B
B
A
Thank
You
Victor,
so
that
the
idea
here
is
I'm
trying
to
minimize
the
number
of
things
that
people
have
to
do.
This
is
this
was,
as
far
as
I
could
tell
the
smallest
number
of
choices
that
you
could
make
to
come
up
with
something
reasonable.
Technically,
you
don't
need
a
network
to
do
this
first
step,
but
we
were
always
creating
a
network,
so
I
added
it
and
then
very
simply,
it
gives
you
a
check
box
of
all
the
non
admin
so
this
environment,
so
it
doesn't
automatically
check
it.
A
I
like
it
basically,
this
list
shows
all
the
nodes
that
are
in
the
system
deployment,
so
they're,
the
ones
that
are
available.
If
you
I'm
like
this
will
do
two
of
them
because
it
will
reveal
about,
and
then
this
is
a
dev
system,
so
it
doesn't
have
any
operating
systems.
I
need
to
fix
that
it
would
let
it
will.
A
So
what's
going
to
lose
a
dev
system,
so
I
don't
have
have
that
option
right
now
and
then
you
can
just
so
from
that
you
go
and
commit
your
nodes
are
added
and
all
that
and
then,
if
you
go
into
the
network,
slips
here's
the
network,
I
just
added.
So
those
things
are
here
and
I
can
check
to
see
the
community
deployment.
All
those
things
are
set,
and
here
is
the
oh
okay
hosts
v6.
It
does
not.
Thus
I
have
a
bug.
A
C
B
A
C
C
B
A
Goal
is
to
so,
if
you're
dealing
with
a
just
a
simple
discovery
system,
I'm
trying
to
eliminate
all
of
the
stick
steps
or
confine
all
of
six
steps
you
that
we
always
seem
to
do
to
handle
a
new
system
discovery
model.
So
if
I
was
going
to
bring
up
just
install
bring
up
six
node
system
to
install
the
OS
is
to
ready
state,
then
this
one
screen
would
basically
let
me
take
them
all
and
install
the
OS
can
go.
Very
I
can
imagine
in.
B
A
B
B
B
C
B
A
Around
it
is
strictly
for
doing
that
that
initial
discovered
provision-
okay,
yeah
I,
don't
I,
don't
wanna
I,
don't
want
a
la
belle,
the
whistle
on
it
I
included
state
because
people
you
should
you
should
do
this.
Actually
you
don't
have
to
I,
don't
even
sure
I
need
state
because
it
doesn't
matter.
If
the
note
is
ready.
Degree
is
already
check
on
hardware
discovered
anymore.
B
A
B
What's
what's
what
is
the
goal?
I
mean?
What
is
the
goal
here
once
what
use
case
for
creating
a
separate
networks
at
the
same
time,
because
I
pretty
much
only
ever
do
that
for
the
sets
for
the
test
deployment
depended
on
the
workloads.
If
I'm
not
throwing
your
particular
workload
on
it,
then
I
don't
know
what
network
crashes
or
what
network
a
network
I
should
be
creating
for.
A
That
was,
that
is
the
question
that
had
me
stumped
a
while
I'm
doing
this
at
all.
It's
part
of
a
demo,
it's
sort
of
part
of
a
demo
process
that
I've
seen
for
this
and
fine,
creating
screens
that
help
a
demo
or
not
adding
functionality.
But
the
simplest
proof
point
we
have
that
this
work
is
that
it's
that
it
created
and
provisioned
the
network
and
from
that
perspective,
giving
it
a
host
aim
for
the
range
and
having
it
provision
this,
these
IPS
would
actually
be
really
useful.
Yeah
I
think
it
dramatically
improves
the
impact.
A
If
someone
successfully
gets
IP
addresses
on
their
host,
they
can
go
into
them
and
verify
the
OSS
there.
It
dramatically
weakens
it
if
they
picked
a
different
range
name
and
the
IP
addresses
aren't
working
as
expected
right.
Well,
though,
they'll
get
I
piece
from
the
admin
network.
This
would
be
a
second
secondary
network.
B
Yeah,
because
the
admin
network
is
really
designed
for
just
you
know
in
in
band
management
and
provisioning,
not
really.
A
And
then,
when
this
runs
through
you're
going
to
crew
going
to
take
the
we're
going
to
take
the
second
Nick,
the
second
one,
gig
Nick,
we're
going
to
attach
these
IP
addresses
to
set
up
the
interface
an
attached
to
the
other
challenge
that
this
is
going
to
require
a
second
Nick,
because
the
1g
one
network
is
Nick's.
Do
I
could
do
this
as
a
pulldown.
A
B
Yeah,
there's
a
bunch
of
there's
a
bunch
of
logic
that
I
want
to
add
to
basically
everything
conduit
related,
because
once
we
pass
setting
up
the
admin
that
work
on
a
box,
I'm
sledgehammer,
we
know
what
Nick's
it
has
and
we
know
how
they'll
and
you
know
how
they
will
map
the
conduits.
So
we
should
be
able
to
cut
so.
The
back
end
should
be
my
word
to
have
awareness
about
what
is.
It
is
invalid
for
any
given
system,
but
that's
that's
a
lot
of
plumbing
dad
in
the
back
end
as.
A
Well
and
there's
some
general
generalities
that
would
be
interesting.
It
be
I
love
the
idea
of
pull
up
well.
This
is
back
to
the
family
screen
really
in
chrome
are
one
we
built
but
never
exposed
this.
This
concept
of
looking
at
nodes
in
a
group
and
grouping
things.
What
you're
describing
is
saying
all
right
in
this
deployment.
All
of
these
are
the
conduits
are
available.
These
are
the
these.
Are
things
are
common
across
all
the
infrastructure?
A
A
What
I'm
hoping
to
do
is
probably
another
pull
request
or
two
in
coming
for
this,
but
what
I
would
ask
is
to
accept
it
actually
or
even
accept
it
sooner
rather
than
later,
so
that
we
can
incrementally
make
sure
it's
not
breaking
other
stuff
and
trying
to
keep
it
very
self-contained
and
I've
been
building
BBD
tests
that
validate
the
behavior
of
the
UI
and
the
logic.
So
it
should
come
in
with
with
pretty
strong
capabilities.
I'd
love
to
see
it
in
in
by
the
next
planning
session.
But
sorry,
that's.
A
B
B
D
B
D
D
Hey
I've
done
some
research
in
that
area.
There
are
some.
There
are
some
very
compact
power
supplies
that
you
can
run
those
many
IT
exports
from
12,
volts,
single
single
rail,
12,
volts
and
power
up
the
entire
board.
So
it
might
be
an
option
to
run
a
single
large
power,
12
volt
supply
and
then
just
stack
a
bunch
of
these
with
a
minimal
chassis.
D
E
Yeah
sure
so
did
a
little
bit
of
work
on
Travis
actually
robbed
in
most
of
it
I
just
kind
of
looked
over
her
shoulder
but
effectively.
E
What
we
have
is
we
have
Travis
actually
hooked
up
to
the
repositories.
So
what
is
possible
is
for
us
to
use
that
as
a
CI
gating
mechanism
on
your
own
private
branch
prior
to
prior
to
the
to
the
actual
push
requests
or
pull
requests
into
the
into
the
main
line.
So
you
should
be
able
to
once
we
hit
the
rest
of
that
stuff.
E
So
that's
the
way
it
looks,
I
think
the
work
that
we
have
to
do
is
you
know
effectively
document
that
appropriately
and
make
sure
people
have
the
right
files.
You
need
a
Travis
configuration
file
inside
of
your
side
of
your
pieces
and
you
know
make
sure
that
that
works
appropriate
for
appropriately
for
us
in
the
in
the
travis
vm
environment,.
A
And
so
I
was
having
a
conversation
victor
about
the.
There
were
two
choices
on
this,
which
was
to
either
add
make
a
single
command
that
ran
the
deficit,
that
fool,
the
dev
environment
and
the
BDD
in
one
thing,
in
sort
of
implied
the
container
to
over
run
the
dev
container
and
then
run
BTW
XD
external
to
it.
I've
been
leaning
towards
the
second,
because
I
didn't
want
to
add
Erlang
to
our
our
docker
container.
A
To
run
chrome
are
so
that's
sort
of
a
sort
of
where
things
lined
up
it's
possible
that
we
could
actually
run
the
dev
mode.
Actually,
you
know
might
be
simpler.
Victor
I'd
be
interested
in
opinion
on
this
running.
We
could
actually
have
Travis
just
run
the
rails,
app
as
a
rails,
app
use
attached
to
post
grass
without
without
the
container,
because
the
depth
of
the
dev
mode
doesn't
need
to
do
all
the
setup.
It
really
just
runs
as
a
rails
app.
E
That
actually
brings
up
a
good
point
to
bring
up
on
the
call
you
know
doing
it
in
dev
mode
actually
contains
the
deployment
and
testing
infrastructure
into
the
you
know
into
that
docker
container,
but
effectively
simulates
the
bring
up
of
additional
systems.
You
know
a
which
is
a
good
first.
First.
A
E
E
E
A
A
A
A
B
Garment
head
and
basically
rewritten
the
hardware
workload
to
look
a
lot
like
the
OpenStack
workload,
where
it's
prepped
for
now,
it'll
start
off
as
just
a
a
collection
of
puzzles
of
essentially
crowbar
20
bar
clamps
in
the
same
workload
and
for
the
ipmi
workload.
I've
got
it
so
that
it
is
up
and
running
and
functional.
On
the
poweredge
t320
that
I
duck.
We
could
go.
B
Ok,
so
what
working
means
is
that
we
can
detect
that
the
ipmi
controllers
there
and
we
can
configure
it
to
use
our
username
and
password
and
we
can
configure
it
to
use
the
we
can
configure
to
the
IP
address
that
we
wanted
to
run
on
top
of
along
the
way
I
went
and
stripped
out
a
bunch
of
code
that
was
really
there
to
work
around
some
issues.
That
I
feel,
like
rule
doesn't
add
any
more
there's
been
a
fair
amount
of
refactorings
done
there
for
the
ipmi
roles
themselves.
B
D
B
Be
written,
ipmi
jig
that
will
add
capability
to
the
notes
to
figure
out
a
band
power
on
power
off.
You
know,
check
the
identification
and
provide
hooks
to
do
for
the
detection
to
see
if
the
IP
my
controller
running
on
as
an
ID
rack,
in
which
case
you
move,
do
other
stuff,
biases
paging
to
it,
using
rack
Adam,
detecting
to
see
if
it
has
tibias
man
capabilities
and
on
so
forth.
A
B
From
the
architectural
side,
I
went
ahead
and
did
something
a
little
unusual
and
that,
if
you
have,
if
the
ipmi
workload
is
there,
then
all
notes
will
get
an
IP.
Mi
discover
rolls,
and
if
that
discovers,
that
there's
an
IPO,
my
controller,
that
we
can
configure
or
out
of
phantom
management's,
it
will
go
ahead
and
add
so
it
will
go
ahead
and
allocate
an
allocate
an
IP
address
for
the
IP
my
controller
and
find
the
fisa
corresponding
network
role
and
the
ipmi
could
figure
role.
B
B
B
D
B
B
B
Of
profits
simply
too
abstract
that
out
too
for
things
like
vm
if
we're
not
controlling
the
hypervisor
abstract,
add
up
into
like
an
in-band
power
management.
Singie
united
talked
a
little
bit
about
them,
they're
being
sometimes
being
a
separation
of
concerns
between
a
possible
separation
of
terms
between
jigs
and
out
of
that
management.
For
a
note.
A
B
Jason
just
note
management
stuff
in
general,
because
the
jig
job
right
now-
and
it
probably
should
be
going
forward-
is
just
to
do
things
to
notice.
Based
on
what
roles
tell
us
do
and
there's
space
there
for
another
concept
of
up.
We
want
to
be
able
to
do
out-of-band
things
to
notes
themselves
that
aren't
necessarily
controlled
by
any
specific
role.
Like
turn,
the
note
on
and
off
that
sort
of
thing
right.
B
Yeah
like,
for
instance,
right
now,
once
we
once,
we
discover
a
node
in
its
car
managed
node.
We
just
leave
the
note
of
running,
but
if
we've
along
the
way,
we've
discovered
that
it
has
a
beast
that
it
has
an
IP
mi
controller
and
that
weaken
diggit.
One
thing
we
can
do
is
after
we
discovered
that
we
can
manage
a
node.
We
just
turn
it
off,
and
then
we
wait
until
someone
finds
a
new
set
of
role
and
convince
us
to
turn
the
note
back
on
for
us
to
start
doing
things.
A
Right
which
any
weed
power
cycle
note
anyway,
after
the
discovery,
step
and
start
the
next
things
know
it.
If
somebody
had
said
the
trick
is
in
discovery,
we
don't
set
the
drives
or
the
bios.
We
just
wait,
but
if
we're
doing
out
a
band,
if
we
have
band
capabilities,
we
can
do
those
without
the
OS
being
on
anyway
Marco,
but.
B
B
B
We're
not
going
to
you
know,
since
we're
not
going
to
do
all
of
our
grade
and
BIOS
configuration
out
of
the
band
the
long
term
I
mean
there
are
boxes
that
their
IP,
my
controllers,
they
don't
expose
WS
man,
they
don't
have
the
capability
of
doing
class
configuration
and
raid
configuration
not
a
fan.
They
have
to
do
it
using
up.
B
Yeah
and
the
other
thing
I've
done
with
the
ipmi
is
right
now
use
DHCP
still
present.
It
may
even
work,
but
I
haven't
tested
that
at
all,
but
I'm
instead,
as
I've
made
the
role
smart
enough
to
the
definitive
text
that
if
the
IP
on
my
controller
already
has
an
actress
in
the
right
range
that
it's
being
used,
then
we'll
just
allocate
that
address
port
instead
of
picking
a
new
one.
So
if
we're
testing.
B
B
A
A
A
E
B
Right
now,
this
is
the
initial
commit
of
functionality.
We're
open
forever.
I
haven't
been
tracking
issues
at
all.
Okay.
E
B
A
E
Yeah
I
mean
there's
two
or
three
pieces
to
it
right.
I
understand.
You
know
Victor
that
you're
at
discuss.
You
don't
pull
your
hair
out,
that's
actually
really
good
stuff.
I
love
it.
The
question
is,
is
you
know,
is
there
any
other
input
that
we
want
to
have
into
the
stuff
before
it's?
You
know
before
it's
actually
committed
in
code
and
how
are
we
going
to
track
the
stuff
and
make
sure
that
we
put
it
in
the
appropriate
documentation,
places,
pain.
E
B
A
There
I
would
encourage
us
to
be
more
chatty
on
the
open,
crowbar
list
with
some
of
these
things,
the
design
specs
for
the
supermicro
ITX
board
as
an
experiment
system,
and
if
we
find
cases
and
be
worth
just
at
least
saying,
hey,
I'm
considering
this
from
doing
that
actually
had
a
good
back
and
forth
on
when
we
didn't
circulate
it
to
the
list
on
the
bench,
the
milestone
flag
for
no
opera
but
yeah.
Look
at
these
things
on
Valeska.
It's
really
helpful
to
the
community.
A
B
B
So
pretty
much
right
now
with
ipv6
the
address
space
is
so
huge.
That's
you
know
we
don't
have
nearly
as
much
call
or
needing
to
tear
out
specific
ranges
for
our
network.
Oh
right
now,
just
if
you
create
a
new
network,
then
everything
that
is
on
that
network
automatically
gets
an
ipv6
address
in
it,
and
we
assume
that
it's
just
one
big
flat
address
space.
B
Ok
and
so
at
some
point,
we'll
need
to
have
a
discussion
on
you
know
whether
or
not
that's
that's
something
that
customers
are
interested
in.
E
B
B
B
B
That
chef
will
work
fine
in
ipv6,
only
environments,
all
the
corpse
in
all
these
the
grub,
our
web
server
works,
fun
and
I
can
be
signaling
environments.
So
the
only
thing
that
we,
aside
from
the
only
things
we
absolutely
have
to
have
ipv6
RIPD
for
right
now
for
is
doing
initial
pixie
booth
and
with
and
talking
Drew's
quit
everything
else.
We
run
all
of
our
basic
all
of
our
admin
stuff.
Well,
ni
p.m.
I
require
that
he
fishing
for
right
now
because
of
my
team.
B
B
B
A
little
bit
I
mean
most
people.
I
don't
expect
they're
running
they're
out
of
bed
management
that
works
over
anything
other
than
a
private
ipv4
address
range
anyways,
except
if
they're,
unless
they're
totally
insane
in
our
fine
with
exposing
their
out
of
baggage,
which
in
the
internet
which
some
people
are
so
people
are
crazy,
like
that.
A
A
The
goal
very
briefly
on
what
that's
supposed
to
be
about
is
I,
want
to
show
how
to
use
docker
containers
as
a
development
environment
to
do
cycle
testing
on
a
workload
the
template,
the
template
we
have
created,
new
workload
is,
is
actually
locally
out
of
date
and
we'll
have
to
fix
that
I
know.
Mike
and
I
have
talked
about
the
deep
workloads
so
I'm
you
can
look
at
what
the
hardware
workload
is
spike
in
the
fact
port.
Those
changes
in
simply
it's
one
time.
Work
just
needs
to
get
done.
B
Well,
for
the
hardware
workload
specifically
I'm
just
doing
the
things
that
I
did,
which
is
according
to
crow
bar
to
bar
clams
over
because
they
its
design
I,
designed
the
parklet
and
forth
up
to
be
able
to
do
that
with
minimal
changes.