►
From YouTube: OpenCrowbar Community Meeting 2014 04 09
Description
Demo of the Docker Slaves nodes and sprint planning
A
Hey
everybody
and
welcome
to
april
the
9th
2014
open,
crowbar
community
planning,
meeting
we're
planning
on
today
a
little
preparatory
chatting
and
then
reviewing
the
the
pull
requests
that
came
in
over
the
past
sprint.
Talking
about
what
we're
planning
on
doing
and
then
getting.
Hopefully
a
demo
and
some
discussion
on
the
docker
workload
nodes
that
are
now
one
of
the
core
features
of
opencrowbar.
A
I'm
trying
to
pull
up
here
a
list
of
the
prs
just
so
we
could
go
over
them
quickly.
What's
new
in
the
code.
A
Over
the
past
week,
we've
introduced
a
lot
of
two
or
three
very
important
features.
The
workshop
support
for
all
you
chefs,
who
are
writing
cookbooks
and
would
like
to
be
able
to
deploy
those
cookbooks
and
their
dependencies
with
crowbar.
It
now
makes
it
a
lot
easier
we're
still.
This
is
just
initial
support
for
the
quickbooks
that
our
project
is
using,
but
also
in
the
previous
sprint
and
the
upcoming
spring
beginning
to
integrate
upstream
cookbooks
into
our
merch
style
system.
A
There's
also
initial
support
for
puppet
up
at
the
jig
to
allow
us
to
connect
to
external,
to
use
actually
puppet
modules
and
deploy
workloads
with
puppet
and
also
on
the
public
client
role,
to
add
to
your
node
role
graph
to
actually
make
the
deployment
of
public
modules
and
the
use
of
the
public
client
habit,
and
we've
done
a
bunch
of
ui
cleanup.
A
There
were
a
bunch
of
bugs
that
we
that
we
killed
this
past
sprint,
mostly
around
the
ui
and
the
the
the
last
and
most
exciting
thing
is
that
we
can
bootstrap
docker
nodes,
so
we'll
be
demoing
that
a
little
later
we
can
bootstrap
enough
docker
modes
to
show
off
some
of
the
synchronization
things
that
we
still
need
to
fix
in
the
database.
So
yay.
A
A
I
tell
we're
going
to
be
going
into
our
sprint
planning
tomorrow
and
there'll,
be
a
few
things
that
we
can
already
anticipate
we're
going
to
be
working
on
continued
work
on
the
docker
stuff,
I'll,
probably
be
mostly
offline
due
to
chef
comps,
and
a
lot
of
folks
are
going
to
be
at
red
hat
summit
out
in
san
francisco
as
well.
A
A
Fixing
some
of
the
synchronization
issues
that
we're
having
in
the
database,
which
is
pretty
much
just
a
consequence
of
too
many
things,
trying
to
touch
the
same
room,
the
same
table
at
the
same
time
and
also
working
on
beefing
up
the
attribute
the
attrib
model,
so
that
when
you
change
the
trig
via
the
api
or
via
the
ui,
it
will
automatically
schedule
all
the
node
roles
that
either
provide
or
want
that
attribute
to
be
rerun.
With
the
new
value.
A
Yeah,
that's
a
really
important
feature
for
being
able
to
deploy
roles
different
well,
the
particulars
of
a
workload
that
are
dependent
on
each
other,
so
clients
can
know
the
correct
ports
for
services.
A
The
workshop
support
is
in
its
basic
form
right
now,
where
the
the
clickbooks
that
are
used
for
our
core
workload
now
resolve
some
of
their
dependencies
by
pulling
cookbooks
from
upstream.
A
I'm
looking
to
perhaps
enhance
this
to
make
a
clearer
and
more
flexible
work
chef
development,
workflow,
I'm
really
open
to
ideas.
I
started
documenting
some
of
it
out
on
the
docs
in
the
core
repository
of
robar
for
bulking.
B
A
It
seems
like
the
biggest
unresolved
question
in
development
is
kind
of
falls
into
two
lines.
First,
whenever
you're
just
hacking
on
a
cookbook,
how
well
will
berkshire
deal
with
needing
to
make
ephemeral
forks
like
forks,
that
need
to
last,
like
a
sprinter
three,
until
we
can
get
changes
merged
back
into
the
upstream
cookbooks?
A
If
you
have
any
kind
of
ditto
for
making
for
deploying
advent
nodes
with
local
forks
instead
of
using
up
streams
when
we
need
to
on
a
temporary
basis,
I
don't
like
having
to
go
in
to
edit
the
mercs
files
manually
whenever
we
have
to
do
that,
because
we
all
know
that
that's
going
to
lead
to
people
forgetting
to
change
things
back
and
hilarity
and
zooming
down
the
downstream
from
that
yeah.
I
haven't.
A
I
think
if
they
handle
that
off
the
top
of
my
head
yeah,
I
haven't
seen
any
capabilities
where
you
could
indicate
different
sources
for
the
same
cookbook,
but
at
different
versions.
So
version
0.1.5
is
local.
But
0.2
is
upstream.
A
Were
you
thinking
of
another
documentation?
Is
totally
hideous?
You
realize
yes,
yes,
no.
The
documentation
is
the
code
yeah,
because
I
mean
in
that
mind
I
from
that
point
of
view.
You
know,
frankly,
I'd
prefer
just
using
straight
git
to
bookshelf
for
that
sort
of
thing.
A
A
So
in
an
effort
to
not
reinvent
the
wheel,
we'll
see
what
burks
3
is
doing,
but
for
now
for
the
foreseeable
that
centralized
works
file
under,
say
your
workload
like
core
slash,
chef,
slash,
cookbook.
The
brooks
file
in
there
is
the
the
one
true
source
and
just
remember
that
when
burke's
install
gets
called,
which
is
really
a
barclay
at
install
time
and
works,
install
gets
called,
it
creates
a
works
file,
dot
lock,
and
it's
really
that
works
file.
Dot
lock.
A
That
is
determining
what
gets
shipped
out.
A
File
gets
shared
with
everyone
to
indicate
what
larger
dependency
relationships
there
are,
and
the
workfile.lock
is
is
unique
to
your
environment,
your
series
of
environments
or
the
environments
that
you're
collaborating
on
particular
people
with
it's
not
for
upstream
consumption.
B
A
Working
into
crowbar
a
relationship
between
deployment
and
workfile.log
files,
if
that's
desirable,
not.
A
For
which
which
dependencies
we're
going
to
resolve
and
they
make
available
in
the
workshop,
but
also
which
ones
we're
going
to
deploy
out
to
our
different
workload
deployments
yeah,
I
mean
what
kind
of
gets
me
is
this
is
this
is
the
this
is
the
same
sort
of
problem
that
I
that
I,
this
is
the
same
sort
of
problem
that
led
me
to
shove
the
idea
of
cry
of
remotes
having
numerical
priorities
into
the
dev
tool.
B
Yeah
and
but.
A
A
Available
in
first
three,
the
first
three
dependency
resolver
is
a
separate
process
that
manages
the
hierarchy
or
precedence
of
packets
of
cookbook
sources.
So
that's
definitely
it
for
berks
three,
and
we
should
look
at
that.
For
you
know
our
recommendations
once
perk
3
is
available.
A
Okay,
yeah,
because
if
you
have
that,
then
we
can
partially
solve
it
just
by
if
you're
not
making.
If
you
don't
have
a
if
you're,
not
making
hacks
locally
on
a
cookbook,
then
just
don't
have
a
local
copy
of
it,
and
if
you
don't
need
a
local,
if
we
don't
need
like
a
an
org
specific
fork
of
a
cookbook,
while
we
are
getting
bug,
fixes
or
features
push
up
stream,
then
we
just
don't.
Have
it
and
perks
would
just
automatically
fall
back
in
the
order
in
which
we
set,
where
we're
going
to
fetch
cookbooks
from.
A
A
We
can
also
make
a
lot
of
this
kind
of
thing
happen
with
the
burksfield.lock.
Indeed,
you
have
to
manage
it,
but
the
burkesfile.lock
will
say
okay
for
this,
my
particular
deployment
go
local
first
and
use
these
particular
cookbooks.
I
don't
care,
what's
in
the
bookshelf,.
A
Because
my
evidence
suggests
that
expecting
that
will
lead
to
all
sorts
of
fail.
A
All
right
is
there
any
other
topics
folks
want
to
dig
into
before
we
try
a
docker
container
demo.
A
Victor
imagine
you'd
be
driving
this
demo
because
I'm
not
set
up
for
it
yeah.
I
suppose
I
can
I'll
need
to
spin
up
an
admin.
Node
real,
quick
and
we'll
have
to
do
our
usual.
Okay,
song
and
dance.
A
I
can
do
skype
or
for
you,
it'll
have
to
be
skype.
Google
work,
you
know,
pepper
gmail
are
both
good
for
me.
A
A
Okay,
as
victor's
desktop,
comes
into
focus
and
folks
see
what
he's
doing
there.
A
You
can
tell
by
all
the
equal
signs
that
the
rake
and
the
rail
db
migration
tasks
are
okay
there
now
we're
pretty
well
resolved,
it's
not
as
lost
anymore.
A
There
was
while
we're
waiting
for
this.
There
was
an
excellent
podcast
that
I
listened
to
earlier
this
week.
On
my
way
up
from
new
york
city
called
the
ship
show
and
their
most
recent
hour
is
spent
with
three
relief
engineers,
a
couple
operations
guys
and
a
developer
talking
about
how
different
ways
to
use
docker
and
different
ways
to
be
using
docker,
and
they
were
kind
of
experimenting
with
ideas
and
tossing
them
around,
and
it
was.
It
was
really.
A
To
see
both
the
users
that
they
were
putting
it
to,
and
also
some
of
the
challenges
that
folks
have
understanding
that
docker
isn't
just
containers,
it's
not
just
creating
a
container.
It's
also
a
union
file
system,
which
is
a
layered
file
system,
where
you
can
actually
ship
around
layers
of
a
file
system
from
one
container
to
another,
if
they're,
if
they're
launched
from
the
same
image
share
diffs
from
them.
A
A
It
would
also
allow
the
os
to
not
necessarily
be
patched,
but
only
the
container
that
held
the
web
server
needed
to
be
patched.
You
could
really
overlay
the
new
ssl
libraries
on
top
of
the
old
ones,
without
changing
the
underlying
operating
system,
but
making
the
new
libraries
available
to
the
web
servers
reducing
your
need
for
potentially
rebooting
or
the
the
temptation
to
reboot
the
os
that
the
web
servers
are
running
on.
A
A
Hey
it
looks
like
the
admin.
Node
is
almost
up
once
it's
up
I'll
kick
off
the
the
docker
slaves.
A
More
than
20
I'll
start
showing
off
my
awesome
places
where
I
need
to
tweak
the
schema.
I
don't
know
if
our.
A
Is
is
prepared
for
your
steam
violation
or,
or
you
know,
transaction.
A
A
That's
ultimately,
that's
causing
the
issues
that
I'm
seeing.
So
it
is
a.
It
is
really
a
scheme,
a
tweak.
A
Mostly
because
the
row
in
question,
the
main
thing
that
I'm
updating
is
a
gigantic
blob
that
gets
pushed
such
a
gigantic
blob
of
json.
That
gets
pushed
straight
into
chef,
and
I
need
to
go
back
to
dynamically,
generating
it
on
the
fly.
But
in
such
a
way
that
doesn't
cause
the
exponential
slowdown,
the
more
nodes
we
get.
A
A
Yeah
doesn't
postpress,
have
features
that
will
do
this
automatically
for
you,
you
can
request
records,
request,
selects
to
be
returned
to
you
in
json
format,.
A
A
A
How
are
you
numbering
these?
I've
changed
the
admin
range
to
from
a
slash
24
to
22,
which
should
give
me
space
to
bump
into
the
1000
node
kind
of
design,
spec.
A
A
Okay,
so
these
don't
get
sledgehammer
on
them.
What
what
is
what's
being
put
in
these
containers
on
the
back
end-
and
I
just
switched
to
a
container
that
looks
like-
has
already
been
created
and
run
on
the
so
what
happens?
Is
I
wind
up?
A
Creating
I
create
the
node
via
the
api,
bind
a
role
called
docker
contain
or
crowbar
docker
node
to
that
node.
A
Then
I
commit
that
node
and
that
that
is
enough
to
tell
the
annealer
that
it
needs
to
do
stuff,
and
it
will
detect
that
I
it
will
detect
that
I
want
to
make
a
node
that
needs
to
transition
or
make
a
docker
node,
and
on
the
back
end,
there
is
a
role
called
provisioner,
docker
setup
that
goes
in
and
creates
a
custom
script
that
contains
both
the
necessary
indication
for
docker
to
wire
it
up
to
use
our
dns
server.
A
Make
sure
that
it
can
make
sure
that
the
annealer
can
or
make
sure
that
the
network
roles
can
add
and
remove
ip
addresses
to
it
and
make
sure
that
it
will
attach
to
the
right
bridge
and
then
and
then
also
it
also
contains
the
the
init
script
that
I
run
in
the
context
of
the
container
and
that's
what
you're
seeing
in
this
back
scroll
right
here.
A
And
then
it
just
and
then
the
the
docker
slave
scripts
just
waits
for
that.
Wait,
waits
for
that
script
to
appear
in
the
tftp,
boot,
slash,
node,
slash,
node
name
directory
and
when
it
does,
it
just
runs
it
and
that
creates
the
container
and
gets
everything
kicked
off,
which.
A
Which
image
are
you
using
to
create
the
container
a
custom
one
right
now,
based
on
the
ubuntu
image?
That
already
has
a
chef
solo
loaded,
so
we
don't
have
to
spend
so
much
time
loading
it
or
we
don't
have
to
spend
time
installing
it
right
great.
The
script
itself
doesn't
care
about
that.
But
if
you
don't
write
it
on
one
of
our
images,
it'll
be
forced
to
download
and
install
curl
and
chef
solo.
So
we
can
do
our
things.
A
And
it
looks
like
they
are
all
going
to
go
through
without
any
problems.
Most
of
them
have
already
finished,
setting
up
their
initial
roles
help.
Let
me
wait
for
skype
to
catch
back
up,
yeah.
A
And
there's
our
20
nodes
and
they've
already
checked
to
make
sure
there's
a
chef
there's
a
chef
solo
on
the
box,
they've
already
wired
up
their
admin
network
appropriately
right
now,
I'm
having
to
run
the
docker
containers
in
privilege
mode
to
do
that
future
releases
of
docker
will
allow
me
to
assign
no
assigned
container
capabilities
on
a
more
fine-grained
basis.
So
I
can
just
switch
to
letting
the
containers
have
cap
net
admin,
so
I
don't
have
to
give
them
the
full
power
the
privileged
implies
and
they've
set
up.
A
So
they
know
how
to
install
software
from
the
repositories
that
you
want
them
to
use
and
we
have
them
registered
as
a
crowbar
docker
nodes
so
that
other
roles
that
need
to
so
that
other
downstream
roles
that
need
to
target
a
managed
or
installed
system
can
use
docker
nodes.
A
A
So
these
are,
these
are
yeah
right,
so
they'd
be
talking
out
to
the
open
to
repositories,
tactic,
repos,
get
all
updated
and
ready
whatever
you
want
into
them.
Yeah
pretty
much.
The
responsibility
of
this
role
is
to
make
sure
that
the
node
has
the
right
gym.
Rc
the
right
proxy
settings
make
sure
that
everything's
set
up
to
use
the
proxies
on
the
admin
node
and
make
sure
that
the
node
has
all
the
references
to
the
right
in
cluster
and
upstream
package
repositories
that
we
wanted
to
have
access
to.
A
Right
now,
the
docker
stuff
only
works
when
you're
running
on
the
same
physical
machine
that
the
admin
container
is
running
on
or
the
admin
node
is
running
on.
But
that's
mostly
because
I
wrote
it
as
it
is
so
that
I
can
do
stress
testing
on
the
core
bar
core
data
structures
in
the
database
in
the
future,
we'll
flush
it
out.
So
it's
more
full-fledged
about
being
able
to
spin
up
docker
nodes.
A
If
we
need
that,
but
right
now,
it's
it's
mostly
around
just
as
a
proof
of
concept
and
a
development,
flash
testing
tool
yeah.
This
could
really
speed
up
development
and
testing,
especially
for
all
set
up
to
hit
external
repos.
B
A
That's
one
of
the
reasons
I
think
we're
at
a
state
now,
where
it's
reasonable
to
you
know,
restart
work
on
bringing
on
the
cloud
era
and
on
the
openstack
workloads
and
bring
that
under
open,
crowbar.
C
A
Docker
nodes
when,
with
docker
nodes
like
this,
the
most
I've
ever
tried
to
spin
up
is
about
150
and
the
only
reason
that
I
couldn't
do.
That
was
because
I
started
running
into
too
much
database
contention
on
those
on
the
dns
database,
the
dhcp
database
and
the
docker
database.
A
C
The
admin
node
my
first
attempt
was
pretty
good,
but
you
know
future
when
I,
when
I
actually
want
those
nodes
to
do
something,
I'm
going
to
need
a
little
more
headroom,
but
I
think
at
that
stage
we'll
have
docker
running
on
the
other
nodes
as
well.
Docker
containers
well
probably.
A
It's
definitely
suited
for
it's
definitely
well
suited
for
certain
workloads
that
it's
a
very
good
way
for
me
to
stress
out
the
core
work,
to
make
sure
that
it
can
scale
to.
You
know
having
to
being
able
to
manage
a
thousand
nodes
effectively.
A
Well,
one
of
our
use
cases
is
to
just
work
on
hadoop
deployment
and
not
necessarily
load
a
lot
of
data
into
hadoop
or
cloudera.
Aren't
docker
nodes,
the
moral
equivalent.
As
long
as
you
don't
need
a
special
file.
C
Yeah
yeah,
you
know
I
mean
I
I'm
looking
into
that
as
this
stuff
kind
of
matures,
but
not
sure
how
we're
going
to
implement
that
it
may
or
may
not
send
makes
sense
to
run
some
of
the
hadoop
workloads
inside
containers.
But
the
the
file
system.
A
Anything
that
doesn't
need
direct
access
to
physical
devices
like
hdfs
yeah.
We
could
probably
spin
it
up
to
run
in
a
docker
node,
no
problem,
but
it
you
know
anything
that,
like
like
a
storage
node
or
a
like
a
a
center
storage
node
or
a
or
sf
sfosd
node.
All
of
those
have
to
effectively
run
on
physical
hardware
because
they
need
to
essentially
be
able
to
take
over
a
hard
drive
and
you
could
work
around
that
by
pre-formatting
file
systems
and
by
mounting
them
into
the
container.
A
C
So
you're
getting
into
some
of
the
stuff-
that's
keeping
me
awake
at
night,
but
there
are
some
things
we
can
potentially
do
there.
I'm
more
focused
on
right
now,
kind
of
making
sure
the
basic
infrastructure
is
there,
but
mixes
of
virtual
containers
come
into
play.
C
C
A
There's
a
way
to
flag
roles
as
compatible
with
one
of
the
three
types
bare
metal,
one
or
more
of
the
three
cut
three
types
of
bare
metal,
container
or
virtual
machine
we
could
put
together.
A
You
give
us
the
freedom
to
experiment.
We
don't
have
that
yet,
and
I
don't
know
that
the
added
complexity
of
doing
that
for
a
data
release
will
be
worth
it
quite
honestly
sure
that's
one
of
the
things
that
I'm
not
going
to
add
until
I
have
until
there's
a
convincing.
You
know
business
driven
use
case
to
add.
C
Yeah,
I'm
thinking
much
further
down
the
road
so,
for
you
know
the
capabilities
we
have
right
now.
This
will
let
us
test
a
lot
of
the
stuff.
We
need
to
test
pretty
pretty
easily.
C
You
know
which
automation
is
a
big
part
of
the
focus
right
now,
just
to
be
able
to
spin
this
stuff
up
and
verify
that
it
works,
but
further
down
the
road.
There
are
some
interesting
things.
We
can
definitely
do.
A
Yeah
and
just
for
doing
smoke
test
workloads,
you
know
if
they
can
run
comfortably
in
a
docker
container
without
much
modification.
You
know,
there's
a
that's.
A
docker
is
a
pretty
good
way
to
do
that
sort
of
thing
like
if
we
could
spin
up
a
complete.
You
know
nova
cluster
inside
of
containers
without
without
having
to
modify
the
rest
without
having
to
modify
the
upstream
cookbooks
to
specialize
it
enough,
so
that
it's
not
really
an
accurate
test
anymore.
Yeah
it'll
be
great
for
that
sort
of
thing.
C
A
C
A
A
C
Looks
good
though
it's
coming
along
quite
a
ways
in
the
past
couple
of
weeks.
A
Yeah,
this
is
really
awesome
stuff,
I'm
having
a
particular
problem,
because
I
had
bandwidth
challenged.
A
What's
if
I
want
to
take
a
new
image
of
my
running
docker
admin,
node,
hey,
should
I
or
should
I
just
wait
for
10
minutes
every
time
and
and
b?
If
so,
how
like?
What
is
the
best
way?
I
thought
you
had
a
like
restore
pristine
or
something
script.
A
A
B
A
C
A
A
A
A
A
Yeah
yeah,
I
could
get
cable
or
bios
or
even
decent
cfl.
Where
I
live,
I
can't
even
get
dsl
up
too
far
from
the
co,
so
I
gotta
go
looking
again.
A
Yeah-
and
I
like
fried
bird,
you
know-
hybrid-
is
nice
I'll,
just
drop
off
the
roof
of
my
house.
Yeah
microwave
relays
are
great.
Also
this
point-to-point
wi-fi
works,
I'm
in
a
valley
and
I'd
need
to
set
up
a
tower
and
shoot
wi-fi
down
into
my
down
towards
my
house.
A
I'm
gonna
cut
off
going
amongst
your
neighbors,
who
are
in
similar
straits,
yeah
yeah.
No,
I
want
to
do
municipal
wi-fi,
but
I'm
afraid
of
isps,
because
the
the
incumbent
isps
really
have
a
stranglehold
and
if
you
there
have
been
recent
court
cases
where
municipalities
are
forbidden
from
competing
with
isps
even
if
the
isps
do
not
provide
service
in
that
area.
C
A
A
Not
to
immunity
wi-fi,
but
do
a
non-profit,
wi-fi
or
something
like
that.
Non-Profit
is
also
an
excellent
way
to
go
and
purchase
and
string
up
string
up
fiber
optics,
rent
space
on
the
power.
A
Telephone
poles
and
string
up
fiber
I'd
love
to
do
that.
We
have
in.
B
A
Area,
actually,
I
have
a
lot
of
retired
verizon
and
fellow
atlantic
people.
B
A
We're
way
a
field
crowbar,
but
those
are
so
if
you
are
bang
was
challenged
or
you
want
to
create
your
own
docker
images
of
our
admin,
node
that
those
are
kind
of
the
issues
that
you'd
be
running
into.
B
Thanks
everybody
for
climbing
on
board
my
docker
demo
and
reviewing
what
we're
up
to
and
we
look.
A
Forward
to
design
call
next
week,
actually
there
probably
won't
be
a
design
call
next
week
because
of
the
red
hat
summit
and
chefconf,
but
we'll
catch
up
the
week
following
on
april,
the
23rd
thanks
for
calling
in
alrighty.