►
Description
Because we're apparently still incapable of producing a stream of perfect quality, this week's audio issue comes to you sponsored by Matt Oswalt, whose voice lives in a bucket. Sorry about that - Matt tried to stay quiet as much as possible.
See the UK NRE Meetup here: https://www.meetup.com/Network-Reliability-Engineering-Meetup-UK/events/259893162/
Broadcasted live on Twitch -- Watch live at https://www.twitch.tv/nrelabs
B
A
A
So
welcome
to
Anna
Reed
I
live
it's
the
NRE
labs,
weekly
stream,
with
hosts
myself
cloud
toad
and
mirrored
and
Matt
I
am
Derek
winkworth.
Otherwise,
no
one
is
cloud
on
Twitter.
You
can
follow
me
there
and
my
co-host
is
Matt
Oswalt
otherwise
known
as
mirrored
and
mie
RDI
n.
A
So
if
you're
a
network
engineer-
and
this
is
something
automation-
is
something
you'd
like
to
learn
about,
then
this
website
is
for
you
and
we
do
this
stream,
which
is
you
know,
is
about
any
labs
every
single
week.
We
cover
different
issues,
we
have
different
guests,
etc,
etc,
and
this
week
we
have
David
G.
It
was
all
the
way
on
the
other
side
of
the
pond
in
the
UK,
and
he
has
he's
done
some
pretty
interesting
things.
So,
let's
just
jump
right
into
it,
Dave.
What
do
you
got
for
us?
A
C
Am
I
what
I
haven't
got
I've
got
everything
for
you
guys
so
terraform
on
Juno's.
What
4g
knows
I
think
this
is
relatively
uncharted
for
networking.
As
we
know
it
today,
this
isn't
a
control
thing.
This
isn't
a
build
it
destroy
it
watch
and
feed
it
thing.
We
can
use
terraforming
some
really
really
cool
ways
for
traditional
networking
and
a
little
bit
on
the
NRI
UK
meetup
as
well.
So
we've
got
some
interesting
things
going
on.
C
Second
NRI
meter
is
the
18th
of
April
in
burton-on-trent,
otherwise
known
as
beer
town,
nice
yeah.
So
we've
got
a
three
to
three
and
a
half
hours.
We've
got
some
guests
come
along
I'll,
probably
try
and
limit
about
30
minutes
worth
of
time
to
talk
about
something
I
built
recently,
not
terraformed.
We've
got
some
guests
coming
home
as
well,
so
anybody
else
wants
to
come
along.
Please
come
along.
We've
got
unlimited
tea
and
coffee
beer,
chili
and
nachos,
as
well
with
rice,
soaked,
uncommon
commentary,
chili
yeah
in
the
UK.
B
B
B
C
A
Yeah,
so
we
have
there's,
there's
a
there's,
a
company
here
who
is
really
interested.
They
have
they've
hired
some
engineers
and
they're
very
interested
in
contributing
to
energy
labs
and
math,
flew
out
so
that
he
and
I
could
go
meet
them
in
person
and
that's
why
he
was
here
or
here
in
Wisconsin,
all
the
way
from
Portland,
which
is
a
four-hour
flight
for
four
and
a
half
hour
flight,
and
he
decided
we
decided
when
he
was
at
my
house
to
watch
a
movie
and
it
was
called
won't
be
two
in
the
shadows.
A
B
A
B
B
Alright,
then
so
so,
what
about
so
I
guess,
let's
start
out
I'm
curious
to
see
like
if
we
could
dive
into
the
specifics
and
like
what
makes
terraform
different,
because
it
is
very
different
from
what
like
most
Network
autumn.
Eaters
may
expect
with
something
like
an
ansible
which
is,
like
you
know,
a
workflow
kind
of
tool.
Do
this
then
do
that
change?
This
variable
add
these
integers
together
that
kind
of
stuff
like
what
what
let's
before
we
dive
into
like
the
integration
that
you
build.
You
know.
So,
let's,
let's
talk
about
like
what
is
the?
C
I
think
it's
it's
all
down
to
the
fact
that
the
engineer
can
let
go
instead
of
worrying
about
what
the
configuration
looks
like
and
how
to
mutate
it
to
a
point.
We
just
ask
the
terraform
module
for
what
we
want.
Give
me
a
port.
Give
me
a
VLAN.
Give
me
an
interface.
Give
me
a
BGP
session.
We
still
have
to
feed
variables
in
on
what
terraform
takes
care
of
not
only
the
order
of
creation
for
things
which
is
pretty
cool,
and
we
don't
have
to
worry
about
kind
of
pesky
switches
for
features.
B
C
B
C
So
terraform
is
is
all
around
the
declarative
aspects
which
I
think
makes
it
a
different
beast,
but
I
think
one
of
the
most
beautiful
things
about
terraform
is
it
takes
care
of
the
destruction
as
well
so
site,
it's
a
whole
crud
friendly
platform.
So
you
take
something
like
ansible.
How
many
people
do
you
know
that
that
will
also
build
the
back
out
planners
as
opposed
to
the
so
the
forward
mutation,
as
opposed
to
reverse
mutation?
C
You
you've
defined
your
let's
say:
terraform
config
'lets
or
your
you've
defined
to
find
your
resources
and
you
do
a
terraform
plan,
while
the
same
plan
can
be
used
to
remove
all
of
the
the
kind
of
ghosts
of
the
past
instead
of
having
to
them
as
an
engineer
figure
out
how
to
remove
all
this
crap
I
think
that's
something
that
there's
I
love
really
love
about
it.
To
be
honest,.
B
B
It
has
its
roots
in
like
the
clout
so
like
it's
commonly
used
for
describing
like
network
resources
and
things
like
you
know,
virtual
machines
in
the
cloud
you
can
see
that
kind
of
stuff
and
I
got
really
addicted
when
I
first
started
using
it,
because
you
can
very
easily
create
these
things
just
using
simple
text
files
which,
again,
if
you're,
if
you're
using
instable,
you
might
say
like
well
I,
can
create
those
with
ansible
playbook.
But
the
thing
with
terraform
is
like
you
don't
have
to
talk.
B
You
don't
have
to
there's
no
there's
no
time
dimension
to
this
look
with
ansible
there's
a
time
you
mention
you
have
to
like
go
step
by
step
by
step
by
step
and
for
simple
day.
That's
fine,
but
I,
think
for
very
complicated
setups.
It
becomes
difficult
to
describe
both
what
it
is
that
you're
trying
to
create
and
also
how
you
try
to
create
it
and
I
think
what
you're
saying
David
is
spot-on.
B
Everything
is
built
on
top
of
terraform
by
the
way,
with
that,
with
the
key
infrastructure
behind
every
labs
and
I
got
really
addicted
to
to
like
constantly
rebuilding
stuff,
because
I
was
trying
different
things
and
wanted
to
configure
things
differently
and
I
was
partnering,
so
it
was
just
really
funny
they're,
just
like
totally
blow
it
away
and
I
think
I
tweeted
this
where
it's
like.
Can
you
imagine,
like
even
five
years
ago
in
the
world
where
no
people
still
doing
things
in
the
data
center?
B
C
All
what
do
you
say
that
I've
kind
of
got
a
little
bit
addicted,
so
even
when
I'm
kind
of
building
things
or
pocs
or
whatever
I'm
now
finding
ways
to
terraform
eyes
it
it's
just
like?
Yes,
this
is
so
easy.
This
is.
This
is
just
cool.
It
makes
life
brilliantly
easy.
It's
fun
as
well,
as
these
kind
of
is
brought
some
joy
back
because
it
kind
of
worked.
It
just
works.
It
doesn't
kind
of
go
wow.
I'm,
not
gonna.
C
Do
this
today,
for
whatever
reason
or
this
version
around
horse
light,
do
you
think
I
don't
know
it
just
feels
a
lot
more
feels
nice
feel
stable.
You
don't
have
to
worry
about
templating
and
all
these
kind
of
nonsense
and
I'm
learning
more
and
more
about
it
as
well.
I
mean
in
order
to
actually
build
this
stuff
of
Acts
and
look
at
the
source
code
quite
heavily
and
I
spend
like
two
or
three
weeks,
just
poking
and
filling
around
with
things,
a
lot
of
a
lot
of
thought,
and
it's
still
by
no
means
clear.
C
Yet
I
came
across
a
bit
of
a
gotcha
last
week
in
terms
of
deployment,
which
is
something
is
probably
with
interesting
to
talk
about
so
I.
Don't
I
don't
to
talk
about
Jason
happy
too
much
here,
because
some
of
this
is
kind
of
related
to
that.
But
when
you
take
terraform
and
you
want
to
do
certain
things-
and
you
also
want
to
test
for
certain
things-
you
have
to
kind
of
step
out
of
the
box
a
little
bit
and
think
how
does
a
device
view
the
world
and
not?
How
do
you
view
the
world?
C
I
can't
lead
ascribe
this.
So,
for
instance,
let's
talk
about
an
access
switch.
Let's
just
talk
about
provisioning
access
pause
and
you
want
to
provision
an
access
port.
Well,
how
do
you
know
when
you
provisioned
a
port?
Do
you
go
well?
I've
changed
a
description
field
or
I'll
check
that
I
don't
know
the
interface
is
open,
I'm,
seeing
MAC
addresses.
What
do
you
do
with
that,
and
so
from
a
terraform
perspective,
how
does
terraforming
know
its
provision
to
port?
C
This
is
gonna,
get
deep,
any
beer
coffee's,
not
good
enough.
So
why
do
we
automate
stuff?
Well,
we
automate
human
gadgets
and
human
levers.
All
right,
not
necessarily
true.
We
June
Austrian
us
is
different,
but
we
take
it.
I,
say
an
average
network
operating
system.
What
we're
actually
doing
is
were
automating
the
things
that
a
human
would
otherwise
do.
If
we
didn't
automate
the
human
things
we're
gonna
be
programming,
it
so
we'll
probably
switch
over
to
p4
OpenFlow,
something
and
I.
C
A
A
C
So
what
you
have
to
do,
there
were
some
mandatory
things
to
do
when
you're
building
the
modules-
and
there
are
four
functions
there
is
create,
read,
update
and
delete,
and
the
idea
is
that
you
have
to
fill
in
all
four
of
those
kind
of
like
boilerplate
methods
or
the
kind
of
art
that
you
can
build
it
in
different
ways
of
guess.
But
I've
built
miners
methods
on
a
on
struct
I'm
still
reading
go
so
yeh
super
super
nice
enough
to
work
in
any
other
kind
of
less
desirable
language.
I'm,
not
gonna,
mention
any
names.
C
As
long
as
you
fill
those
fields
in
then
it's
down
to
you
is
the
operator.
We
kind
of
you
build
you
plan
and
then
terraform
keeps
a
local
state
which
should
match
up
with
the
state
of
the
downstream
device.
Although
I
always
kind
of
call
it
like
the
I
guess.
The
control
plane.
A
A
A
Have
to
I
mean
you
can
make
CR
you
D,
then
you
can
make
create,
read,
update
and
delete
methods,
I,
suppose
and
and
and
then
you
can,
all
those
can
be
called
from
the
command
line
and
terraform
we've
provided
terraform
file
for
the
right.
You
know
for
the
input
for
the
right,
but
that
has
the
input
for
the
right
method
and
then
it'll
go
and
do
whatever
it
is.
You're
asking
yeah.
C
And
you
can
do
some
very
simple
to
some
very
complex
things.
So
terraform
builds
a
cyclic
graph
from
all
the
dependencies,
so
you
could
you
can
create
an
access
port.
Let's
say
you,
let's
say
you
take
the
traditional
come
to
view.
This
is
not
the
three-tier.
You
take
an
access
port,
you
take
a
VLAN,
you
take
a
layer.
3
interface
bind
that
to
the
VLAN.
C
C
So
there
are
a
number
of
equivalent
I
guess:
I'm
gonna
call
them
tie
breakers,
but
there
are
a
number
of
rules.
So
let's
say
you
write
a
real
bad
function
and
for
whatever
reason
it
crashes
out
and
what
will
happen
is
terraform
will
try
and
leave
it
in
a
certain
state.
So
there
are
rules
around
what
it
does
with
IDs.
If
something
doesn't
happen,
it
might
update
an
ID,
my
not
update
an
ID,
nothing.
C
Half
the
problem
when
I
was
first
kind
of
playing
around
with
the
with
the
code
was
trying
to
figure
out
exactly
what
terraform
will
do
under
certain
situations.
They
really
know
if
they
kind
of
build
on
that
one
or
not
I
guess
what
I'm
trying
to
say
is
that
the
court
operations
providing
that
you
write
solid
code,
that,
with
some
basic
error
checking,
is
some
handling
in
there
it's
very
hard
to
kind
of
screw
it
up
once
you've
got
your
plan
of
attack
in
place
in
terms
of
how
these
things
supposed
to
work.
C
I
think
the
hardest
problem
I
had
was
I'm
dealing
with
the
connection
to
the
to
the
southbound.
So
what
you
do?
Do
you
open
the
socket,
you
kill
the
socket,
do
you
leave
it
like
half
lazily
close?
Do
you
pass
around
the
same
Handler
and
all
that
kind
of
thing,
or
do
you
find
ways
to
batch
updates
together
or
sequentially
update,
update
there,
the
southbound
device
I
mean
that
was
a
bit
of
a
bit
of
a
brain
challenge
at
one
point,
mm-hmm
interesting.
A
So
it's
my
the
last
time
I
saw
a
terraform
file
which
wasn't
that
long
ago
it
looked
an
awful
lot
like
Gamal
when
I
was
looking
at
it
and,
of
course,
ansible
looks
like
yeah
Mille,
so
I
guess
what
is
the
primary
difference
between
between
those
two
off
the
top
of
your
head?
I'm
gonna
put
you
on
the
spot.
B
C
A
C
A
A
A
C
I'm,
just
I'm
trying
to
find
the
right
screen.
I've
got
millions
of
things
and
what
I've
been
doing
for
the
record
by
the
way
as
well,
every
time
I
do
work
on
the
provider
I'm
also
creating
example,
resources
in
terms
of
how
to
use
them,
because
one
of
my
biggest
gripes
in
life
is
just
nothing's
ever
documented.
Nothing
ever
works
right,
okay
and
I.
Try
and
do
this
chef
never
done
this
before
so
turn
on
the
screen
share.
C
C
Doing
the
wrong
thing,
so
let
me
do
this.
So
what
we
have
this
is
a
resource
to
find
in
HCl
or
Hasek
or
configuration
language.
This
is
a
curly
brace
up
and
Matt
was
on
about.
So
we
have
a
keyword
that
says
here's
a
resource
in
quotations.
We
have
the
genus
qfx.
Now
this
is
a
reference
to
a
module,
in
which
case
what
we
call
something
like
I,
think
it's
terraform
provider.
In
fact,
I
mean
this
is
all
a
bit
live.
Sorry,
guys
nothing
like
looking
at
something
for
real,
so
is
HCl.
A
C
All
by
itself,
or
is
it
just
yes,
language
dedicated
to
terraform,
so
the
documentation
is
actually
really
quite
cool
for
HCl
I
mean
you
can
do
some
really
powerful
things.
So
a
little
bit
answer
boy.
You
can
go
off
and
do
some
ginger
ginger
to
type
stuff,
but
you've
got
the
ability
to
bring
in
data
sources.
So
Aaron
ansible,
you
look
like
your
host
FAR's
or
your
player
files
directory
with
HCl.
You
can
go
off
and
bring
external
data
sources,
and
so
you
can
run
a
script
to
get
data
to
build
some
data
sources.
C
It's
really
comprehensive
and
really
really
well-thought-out
on
with
the
latest
version
of
HCl
as
well.
They've
improved
things
like
loop
logic
as
well,
which
was
a
little
bit
a
little
bit
missing
previously.
But
it's
it's
really
improving.
Sorry,
I'm,
just
gonna,
try
and
find.
C
C
Yep,
okay,
so
this
is
the
file
name
of
the
provider.
So
you
see
this
telephone
provider,
two
nos
qfx.
So
then
what
happens
in
the
actual
HCl
define
resources?
You
see
June
osteo
effect.
Well,
that's
the
provider
name,
and
then
we
have
the
kind
of
resource
that
the
provider
is
going
to
create,
which
is
real,
an
access
port.
Then
we
have
a
name
now
this
name.
It
isn't
programmatically
significant.
C
It's
only
significant
to
this
actual
HCl
blob
of
configuration
and
what
we
can
use
that
to
pull
in
reserve
to
pull
in
variables
from
other
to
pull
in
information.
Sorry
using
it
is
variable
for
the
things
so
I've
got
a
resource
name
which
is
programmatically
significant
and
I'll
go
into
details
of
why
that
is,
and
then
we've
got
the
actual
kind
of
primitive
information
that
we're
interested
in.
So
we've
got
a
port
number.
C
We've
got
a
description
and
you
can
see
here
I'm
doing
some
variable
interpolation
and
the
idea
is
that
that'll
be
replaced
with
a
VLAN
number
that
comes
out
of
the
VLAN
resource.
So
if
I
go
off
to
the
VLAN
and
go
well,
it's
got
a
name:
VLAN
4,
4,
2,
X,
Y
Zed,
and
it's
got
a
VLAN
name:
VLAN
4,
4
2.
So
if
I
go
here,
VLAN
4,
4,
2,
X,
Y
Zed.
C
That
name
is
basically
this
one,
but
in
the
in
the
VLAN
resource,
which
is
significant
to
the
resource,
but
not
significant
to
the
actual
implementation
of
the
kind
of
module
which
took
me
some.
It
took
me
some
figuring
out
to
try
and
get
my
head
around
that
these.
These
kind
of
things
are
split
and
they're.
Not
we
they're
not
related
to
each
other
necessarily
well.
This
is
a
little
bit
of
HCl
and
now
this
is
an
implicit
reference
as
well
here.
So
this
is
used
in
the
way
that
terraform
builds
a
graph.
C
Now
there
is
a
keyword
you
can
actually
use
this
as
depends
on
so
I
could
actually
go
down
here
and
go
something
like
oops.
How
come
is
it
you
can
never
type
any
streaming
live.
Is
the
pons
depends
on
and
then
you
could
do
the
same
kind
of
thing.
You
could
go
off
and
call
a
caller
variable
out
there,
which
are
create
an
explicit
dependency
instead
of
an
implicit
dependency.
C
So
if
you
don't
really
want
to
kind
of
mess
around
with
it
the
mess
of
variables,
that's
one
of
the
way
to
do
things
and
terraformed
builds
a
plan
from
that
and
it's
some
people
and
having
worked
with
terraform
now
for
quite
a
bit
I
kind
of
use,
other
tools
and
get
frustrated
at
the
lack
of
visibility
and
how
things
are
going
to
run.
You
only
ever
see
things
we're
kind
of
just
in
time
or
when
execution
or
when
you
instantiate,
but
with
terraform.
C
You
can
build
a
graph
and
you
can
say
well
actually
I
want
to
know
what
my
plan
looks
like
or
I
want
to
know
what
my
apply
looks
like
or
I
know
what
my
destroy
looks
like,
and
actually
you
can
graph
these
things
and
I
think
that
is
really
really
cool.
So
what
I
could
do
is
I
could
do
a
terraform
in
it.
This
might
not
work
by
the
way.
I've.
Not
done
anything
with
this
for
a
while
terraform
plan
and
what
we
might
have
yeah.
C
Anyway,
that's
not
necessarily
a
problem.
There
are.
There
are
ways
of
seeing
this
soon,
hopefully
on
the
NRE
labs.
But
the
idea
is,
you
can
do
a
plan
and
then
you
can
see
either
a
vector
output
or
a
PNG
of
the
actual
plan
that
terraform
is
going
to
run
through.
So
you
end
up
being
able
to
actually
see
what
terraform
sees
and
then
kind
of
use
their
own
logic
to
figure
out.
C
If
it's
the
right
thing
to
do
so,
it's
pretty
cool,
it's
cool
in
so
many
ways,
and
also
that's
not
even
getting
on
to
the
concept
of
like
back-end.
So
you
can
store
a
lot
of
these
things
in
console.
We
said
early
wrong,
you
you
store
like
a
local
version
of
state
and
if
you
destroy
that
state,
then
you
have
a
significant
problem,
but
the
idea
is
I've
done
the
wrong
thing.
C
I've
never
done
this
before
so
yeah.
Forgive
me
on
that
one
yeah,
so
you
can
or
terraform
information
on
a
different
set
of
backends.
It
doesn't
have
to
be
in
your
own,
your
own
file
server
or
on
your
local
machine,
and
the
idea
is
then,
especially
something
like
console
is
really
quite
powerful
for
an
organization,
and
you
can
deal
with
high-level
concepts
like
locking
as
well.
I
could
go
on
for
hours
about
this
stuff.
Tara
forms
just
kind
of
really
change
the
way,
I
think
about
life.
C
A
So
you've
you
created
some
Juno's
modules
or
were
there
Judas
modules
out
there
for
a
terraform?
No.
C
I
created
them
and
I'm
still
we're
not
quite
at
an
alpha
release,
yet
I'm
still
thinking
about
the
logic
way
that
these
things
should
work.
So
what
happens
in
order
to
treat
Juno's
kind
of
I?
Guess
you
to
blir
immutably
terraform
actually
builds
out
configuration
groups
on
top
of
journals,
so
the
idea,
then
we
create
a
VLAN.
C
We
actually
create
a
configuration
group
which
isn't
applied
to
join
us
and
then
the
configuration
trees
we
built
when
after
a
commit
and
that
weighs
well,
we
can
change
the
contents
of
those
configuration
groups
and
providing
that
the
resource
name
hasn't
changed.
Terraform
can
still
blow
them
away
a
later
date
without
about
them
so
where
engineers
go
off.
So
if
we
take
a
an
Ansel
approach
and
say,
we've
built
a
template,
we've
put
some
variables
in
and
then
later
down
the
line
somebody
comes
along
and
manually
changes.
C
Those
things
there's
no
way,
I
mean
answer
walls,
just
a
dumb,
dumb
interface
for
something
else.
There's
no
way
it's
gonna,
pull
that
information
back
and
remove.
You
know
that
the
horrors
of
an
engineer,
but
we
terraform
because
of
the
way
it's
been
based
in
groups.
It
can
just
blow
the
group
away
so
everything's
like
self-contained
in
its
own
stanza.
That's
the
way
we've
done
it.
I'm
gonna
make
a
few
minor
changes
at
the
minute.
Their
global
figuration
groups,
but
I'm
gonna
be
applying
them
now.
The
actual
resource
level,
so.
C
C
That
means
now
you've
got
a
workflow
before
your
workflow
in
order
to
prepare
the
device
for
automation
and
one
of
the
other
things
I
wanted
to
do
was
have
terraform
almost
populate
all
interfaces
with
a
generic
description
field,
and
you
can
do
that
through
a
configuration
group.
But
then
the
problem
was
is
there's
an
order
of
inheritance.
I
have
to
happen.
C
So
what
I've
kind
of
done
now
is
I've
said
well,
I,
think
what
needs
to
happen
is
all
of
the
kind
of
preps
the
prep
work
needs
to
be
in
the
global
space
and
then
all
of
the
actual
the
doing
has
to
go
into
the
the
correct
stanza.
So
the
inheritance
works
properly
within
Julis
itself,
and
this
shall
be
a
nightmare,
I
think
inhaler
operating
system
other
than
doing
us.
It's
just
doing
us
is
really
well
structured
for
this
kind
of
thing.
Yeah.
A
C
C
C
B
Funny,
it's
like
it
sounds
like
you're
having
to
deal
with
this
trade-off
of
like
going
like
making
things
happen
in
a
safe
way.
Like
you
want,
you
want
to
respect
the
inner
the
natural
like
interdependence,
II
of
config
snippets
within
Juno's,
but
by
doing
that,
like
you
have
to
commit
pretty
much
every
time
you
make
a
change
and
of
course
that's
gonna.
Like
add,
potentially
minutes
to
a
do,
you
know
to
a
change
in
Juno's
depending
on
the
device
and
how
big
its
configures
absolutely.
C
And
this
is
another
reason
why
I
need
some
of
the
configuration
items
from
the
global,
the
global
standard
down
to
the
actual
kind
of
feature
stanza,
because
it
reduces
that
kind
of
global
tree
rebuild
and
I.
Think
it's
been
quite
nice,
because
engineering
Jim
juniper
engineer
had
been
really
supportive
of
this,
and
it's
kind
of
it's
invoked
a
lot
of
conversation
around
how
kind
of
configurations
are
managed
internally
and
how
the
trees
are
built.
C
So
it's
it's
been
a
real
challenge
to
even
get
this
far,
but
even
now,
having
used
it
a
little
bit
and
thought
about
it,
there
are
optimizing
and
not
optimizations.
There
are
clean
changes.
I
need
to
make
to
try
and
get
the
best
out
of
this
I'm.
Just
looking
at
kodachi
right
now
and
trying
to
figure
out
I
can't
remember
where
I
do
the
do
the
commits,
if
I
can
find
it
I'll
show
you
doing
these
things
alive,
he's
always
a
man,
because
you
know
it'll
happen.
The
second
you
give
up
you'll
find
it.
B
It
would
be
just
as
a
side
note.
It
would
be
really
cool
if,
like
we,
we
I
think
juniper
take
putting
the
Juniper
hat
on
a
little
bit,
have
a
lot
of
open
source
out
there
and
it
to
say
that
it
it
comes
with
varying
levels
of
support
is
kind
of
an
understatement,
but
we
have.
We
have,
like
you
know.
Pi
Z
is
our
Python
library,
like
that's
about
as
well,
supportive
in
terms
of
in
terms
of
open
source
products,
you're
going
to
get
from
Juniper
like
ever.
B
We
build
so
much
on
top
of
Pi,
easy
like
if
everybody
in
juniper
knows
about
it.
That
does
anything
with
respect
to
automation
versus
like
an
open
MTI,
which
is
more
or
less
I.
My
understanding
is
that
project
is
more
or
less
dead,
so
you're
gonna,
you're
gonna
encounter
varying
levels
of
support.
It
would
be
kind
of
cool
if
we,
if,
like
the
terror
home
provider,
was
like
the
new
AZ,
you
don't
I
mean
like
it
like.
B
C
C
So
what
I've
kind
of
done,
I've
almost
reached
parity
between
dating
configurations
and
deleting
configurations
and
the
committee's
implicit?
So
some
of
those
calls
so
I'll
send
a
raw
blob
of
config,
so
yeah.
So
what
I
do
is
I
I'll
do
some
locking
on
there
on
the
datastore
and
that's
just
a
secondary
a
lot
to
make
sure
if
anybody
has
even
VOC
stuff
through
terraform
that
it's
not
gonna
crap
all
over
itself,
the
sockets
opened
up.
Gonna
dial
happens
that
send
the
raw,
so
I
actually
form
XML.
C
In
the
background,
so
I've
got
a
load
of
data
structures
which
are
serialize
into
XML.
There
is
no
sidechain
checking
with
the
Amal
here.
What
I
tend
to
do
is
I,
try
and
match
up
the
version
to
the
version
of
June
us
and
the
day.
Stroke
just
come
straight
off
the
devices
themselves,
so
that
one-to-one
in
terms
of
priority.
Suddenly
you
roar,
XML
and
I
also
send
a
commit
and
I
close.
C
A
driver
unlock
the
kind
of
warning
data
store
and
then
return
from
the
function
which,
which
cause
out
to
do
nos
and
I,
couldn't
really
find
a
cleaner
way
of
doing
that,
and
that
might
be
just
down
to
me
not
being
overly
familiar
with
terraform
at
the
time.
But
everything
with
terraform
feels
very
flowy,
and
what
I
didn't
want
to
do
is
assume
that
I
could
hold
like
a
go
routine,
opening
one
of
the
implications
of
a
provider
to
make
one
candidate
change.
C
It
was
he
just
wasn't
gonna
happen:
it's
not
the
way
that
terraform
works,
so
they
recall-
and
this
is
why
the
graphing
is
so
important-
it's
not
like
with
anthem,
but
you
have
to
create
the
template
and
then
commit
the
whole
thing
with
terraform.
You
can
make
these
kind
of
micro
transactions
and
terraform
takes
care
of
the
ordering
and
okay
on
a
large
device.
You
know
some
of
the
commit
times
can
be
can
be
fairly
high.
C
I
mean
I've
gone
from
anywhere
from
handfuls
of
seconds
to
about
30-40
seconds
on
some
larger
devices,
but
then
maybe
at
some
point,
we'll
find
a
different
way
of
doing
that,
and
one
of
the
ideas
I
had
at
some
point-
and
this
is
not
even
out
in
general
general
availability.
Yet
was
what
if
we
could
actually
do
terraform
for
a
femoral
configuration.
C
So
we
do
terraform,
2g
RPC
binding
and
manage
manage
that
I
mean.
Then
we
wouldn't
have
the
commit
prop
right.
But
then
obviously
you
give
the
world
a
nuke,
you
kind
of
say:
hey,
you
want
to
go
mess
things
up,
it
wasn't
terraform
providers
and
it's
like
I,
don't
know,
I,
don't
know
how
I
feel
about
that.
One
dangerous.
C
First,
then,
more
than
likely
they're
MX
extensions
for
things
like
managing
bgp
sessions
and
l3
VPNs
and
then
SRX
and
obviously
we're
the
security
aspect.
There
is
a
crap
ton
of
modules
that
need
to
be
built
and
also
data
structures
as
well,
but
I
don't
want
to
build
these
things
by
hand,
so
I'm
actually
right
now,
building
a
tooling
pipeline
which
builds
all
the
resources
automatically.
C
So
the
idea
is
I
feed
in
some
XML
the
data
store,
which
is
a
spat
out
and
then
I'll
have
to
do
very,
very
minimal
wiring
up,
which
means
and
the
rates
of
development
can
be
much
faster.
So
some
of
these
at
the
minute
have
all
been
hand,
hand
waved
but
I've
been
playing
around
with
data
structure.
Generators
which
is
going
really
quite
well.
I
also
played
around
with
go
testing
as
all
to
make
sure
that
I
wasn't
do
anything
stupid
in
terms
of
delaying
the
actual
processing
time
so
everything's
fairly
researched
at
this
point.
C
That's
an
egg
I,
don't
really
want
to
smash
at
the
moment,
so
what
I'm
doing
is
fairly
tactical.
So
if
I
think
about
hydrogen
vironment
one
of
our
customers
touching
the
most
some
things
you
can't
figure
wants
and
walk
away.
Some
things
you're
always
doing
it's
a
little
little
little
things.
That
kind
of.
If
you
envision
configuration,
is
like
a
pyramid
what's
at
the
wide
chunk
at
the
bottom,
what
are
all
those
toil
bit
so
I'm
starting
there,
and
the
idea
is
for
every
task
that
somebody
wants
to
do.
C
I'll
create
an
example
spit
the
XML
out
right
down
at
that
object.
Layer
instead
of
the
whole
thing,
so
I
think
we
could
spend
I
could
spend
a
long
time
trying
to
model
the
whole
of
junus
to
create
those
resources
and
I
think
it
would
be
a
never-ending
task.
I
probably
need
a
team
of
people
to
do
that,
and
that's
probably
me
talking
from
a
lack
of
other
no
experience
on
that
kind
of
scale,
but
it
certainly
feels
that
way
at
the
moment,
nice
bit
of
a
challenge.
A
Well,
I'm,
definitely
looking
forward
to
it,
I'd
like
to
give
my
hands
out
and
play
with
it,
I'm
actually
using
ansible
now
to
get
V
SRX
spun
up
in
AWS,
and
it
I've
been
told
by
everyone
that
terraform
would
be
a
much
better
way
to
do
that,
just
spin
it
up
and
get
it
configured.
So,
if
I'm
looking
forward
to
it
well.
C
A
B
B
C
Course
so
I
think
this
is
probably
the
coolest
thing,
and
this
this
is
changed.
Everything
we've
got
an
NRI
lab.
This
we've
got
a
lesson
coming
up
so
lesson.
31
is
a
terraform
lesson,
so
when
this
is
published,
anybody
can
spoil
one
of
these
environments
up
within
like
120
seconds.
There's
a
five
part
lesson
where
you
can
learn
all
the
basics
of
terraform.
You
can
invoke,
you
can
do
the
units
you
can
do
a
plan.
C
You
also
have
part
of
one
of
these
stages
in
the
lessons
is
to
change
the
contents
of
one
of
the
resources
in
HCl,
and
you
can
actually
once
you've
gone
through
the
basic
lessons.
It's
free
flow.
You
can
play
around
change.
Anything
you
want
now
I've
been
fairly
risk-averse
with
this
I've
not
put
any
documentation,
for
whatever
else
may
be
in
the
module
at
the
moment
is
what
you
see
is
kind
of
what
you
get
in
terms
of
the
modules.
C
But
what
we
do
is
we
take
the
normal
and
Aria
labs
triangle
of
QF
X's
and
then
create
some
basic
resources
on
riku
effects.
One
I
think
it
is
to
finish
off
some
of
the
configuration
and
then
do
some
testing
and
then
it's
also
I
think
it
exercises
a
graph
pile
as
well,
which
is
probably
one
of
my
favorite
parts
and
you
get
to
blow
it
all
the
way
and
destroyed
at
the
end.
So
that's
coming
out
whenever
you
push
the
button
map.
Yes,.
B
And
quick
update
on
that,
so
so
David's
terraform
lesson
is
merged.
It's
all!
It's
all
part
of
the
master
branch
of
the
curriculum,
so
it
will
absolutely
get
in
the
next
release.
Obviously,
because
it's
a
master,
we
are
currently
working
on
some
image
related
things.
There
are
two
other
lessons
that
were
hoping
to
be
part
of
this
release,
one
of
which
is
on
open
config
and
the
other
one
is
on
the
Junos
sdk,
which
is
jet,
both
of
which
require
a
version
of
Juno's.
B
That's
like
a
little
bit
more
full-featured
than
the
image
that
we
have
been
using
traditionally
in
the
image
of
the
tariff
unless
Jesus.
So
we
are
currently
ironing
out
some
wrinkles
on
that
one.
It's
it's
moving
along
we're
hoping
to
we're
hoping
to
have
the
next
version
release
sometime
this
week.
That's
that's
the
goal,
if
not
then
very
early
next
week,
but
very
soon.
It's
we're
just
really
waiting
on
this
image
stuff
to
get
worked
out.
We're
working
on
it
all
the
time.
B
So
as
a
quick
status,
update,
so
yeah
I
think
depends
and
so
and
and
no
it's
awesome
and
now
and
I.
Remember
what
I
saw
was
so
you
and
I
started
talking
about
this
lesson
when
we
were
in
Vegas
together
we
were
at
one
of
Juniper
scales,
kick
house
and
then
I
believe
it
was
a
week
or
two
later
I
saw
a
PR
come
in.
B
I
saw
that
you'd
written
like
he
wrote
like
five
labs
for
this
lesson
so
like
in
the
drop-down
that
you
can
see
on
an
areal
on
an
energy
labs
lesson,
you
can
see,
there's
a
drop
down
for
like
different
like
chapters
and
labs.
They
wrote
five
so
like
this
isn't
just
like
some
like
you
know,
you
know,
here's
terraform
run
terraforming
it
and
see
that's
kind
of
cool.
There's
some
green
text,
Bobby
love
like
no.
This
is
like
this
is
legit.
This
is
like
a
five,
a
five
lab
out
of
the
gate.
B
C
Off
the
gate,
I
mean
just
to
show
a
sign
of
commitment
as
well.
I
actually
wrote
more
functionality
ID
for
this,
because
BGP
wasn't
there
and
I
was
like
damn
unless
there's
nothing
to
complete.
Without
this,
so
I
ended
up
I,
literally
like
coffee
after
coffee
kind
of
I'm
gonna
finish,
this
I'm
gonna
push
this
code
to
them.
Yeah.
It
was
kind
of
cool
to
see
he
was
neat
so.
B
What
you're
telling
me
is,
you
is
you,
you
drank
a
lot
of
coffee
and
then
real
quick
wrote
some
code
to
get
some
features
in
last
minute
because,
like
you,
just
absolutely
had
to
do
it,
and
can
you
sacrifice
sleep
today?
You
know
see
I,
just
don't
identify
with
that.
I
don't
know
if
that's
about
I
have
nothing.
I
have
nothing
to
say
that
relates
to
that
at
all.
Yeah.
A
C
A
A
C
A
B
B
The
target
for
the
next
release,
which
again
David's
lesson,
has
already
merged.
It's
already
live
on
our
PTR
site.
So,
like
that,
that's
dumb
deal
next
release
is,
unless
is
gonna
be
a
the
only
thing
we're
working
on
right
now
is.
Is
there
are
two
other
lessons
which
are
totally
reviewed,
they're
not
merged,
yet
because
we're
making
sure
that
the
image
that
were
trying
to
work
on
a
DQ
affects
a
whole
DQ
affects
image.
B
It's
pretty
beefy
actually
is
is
ready
to
go
for
those
lessons
and
then
once
that
is
done,
all
of
those
will
get
merged
and
we'll
just
kick
off
the
release
for
flow.
Just
like
that,
nice,
so
yeah,
the
target.
The
target
for
all
of
that
to
happen
is
sometime
this
week
and
honestly,
if
it
doesn't
happen
in
the
next
like
two
or
three
days,
that
it
probably
won't
be
because
I'm
traveling
at
the
end
of
the
week
so
I'm
gonna
be
that's
all
I
want
to
be
working
on
in
the
next
few
days.
B
C
C
You
we
hate
things
and
the
reason
is
once
you
start
building
content
and
taking
somebody
on
a
journey
through
actual
stuff
that
you
can
feel
and
see
your
own
stuff
just
feels
crap.
Your
own
platform
feels
like
you're
just
missing
something,
and
it
can't
describe
it
quite
as
completely
change
around
approaching
things
now,
I
can't
it's
completely
rewired.
My
brain
insight
now
I
want
an
en
are
just
yeah
I
can't
I
can't
describe
it.
He's
changed
things
yeah.
A
I,
what
I
like
about
NRI
labs,
I've
written
a
lot
of
blog
posts
as
well
and
I,
wonder
if
and
I
I
look,
especially
ones
that
are
more.
You
know
that
are
speaking
more
technical
things,
I've
done
a
bunch
of
well
a
bunch
of
those
and
I
and
I
go
back
and
I.
Look
at
him
and
I!
Think
man,
if
only
we
had
any
labs.
Oh,
we
could
sorry
I,
don't
know
why
I'm
yawning
I
stop
for
like
ten
hours
last
night,
but
if
only
we
had
entery
labs
right.
A
C
A
For
a
yawn,
oh,
how
awesome
in
are
we
love?
Oh
yeah,
I!
Think
man,
if
I
could
blog
if
I
could
make
every
one
of
my
technical
blogs
have
some
tangible
demonstration
of
what
I'm
talking
about
that
I'd
have
like
that's
just
the
way
it
ought
to
be
done.
I
mean
it's
that
way,
and
you
don't
just
have
to
read
about
it.
A
You
can
immediately
log
into
a
device
and
look
at
it
config
or
run
up
or
look
at
the
state
of
some
protocol,
and
you
can
like
it
just
immediately
get
a
tangible
sense
of
what
it
looks
like
on
the
command
line
right,
which
is
for
a
lot
of
Engineers.
You
know
it.
That's
what
paints
the
picture
so
I
I.
B
So
you
can
share
a
YouTube
video
by
just
sending
the
link
right
same
thing
applies
to
another
real
ABS
lesson
to
be
managing
with
them,
but
with
YouTube,
but
with
YouTube
videos.
You
can
also
share
via
embed,
so
you
can
so
like
I'd
like
to
share
this
and
embed
it
in
another,
another
page
so
like.
If
you're
writing
a
blog
post
like
you,
don't
necessarily
want
to
just
like
link
people
to
NRA
labs
you'd
like
to
keep
them
in
your
site
and
have
like
a
little
widget
there.
B
B
C
C
A
B
C
I,
wouldn't
go
I
mean
I.
Think
you
pay
to
come
out
with
a
virus.
It's
not
good.
Sneaking
populists
is
what
you
do,
go
upstairs
and
run
through
the
doors,
we're
back
in
a
minute
and
then
run
off
with
a
cookie.
That's
what
I
do
I'm
not
condoning
that
blocks
a
way
to
do
it.
You
patron
the
business
they're,
fine,.
A
B
C
A
B
C
B
I
should
mention,
as
with
everything
I
mean.
This
is
one
of
the
one
of
the
one
of
the
things
that's
very,
very
important
to
me,
and,
and
all
of
us
involved
with
the
project
is
that
everything
is
totally
aboveboard.
There's
no,
like
you
know,
there's
no
like
preparing
things
in
a
back
fricking
like
meeting
room
and
then
like
being
like
here's
the
content,
so
everything
is
done
in
the
open
and
so
the
jet
lesson
and
the
the
open
config
lesson
both
are
currently
at
they
they're.
Currently
in
unmerged
open,
pull
requests
on
the
antidote
process.
B
You
can
take
a
peek
Daren
at
any
time
and
and
they're
there
they're
not
merged
yet
because,
like
I
said
we're
trying
to
figure
out
some
of
the
image
stuff,
but
a
lot
of
them
content
is
is,
is
up-to-date,
I
they've
already
both
been
through
like
a
full
review
cycle,
so
they're,
you
know
they're,
basically
ready
to
merge
as
long
as
the
image
stuff
gets
worked
out,
so
the
contents
done
it's
just.
It
just
needs
to
get
the
image
work
in
the
demerger
I'm.
B
Remember
I:
remember
you
doing
this
because
we
were
talking
about
maybe
packaging
like
a
Todd
agent
pioneering
and
putting
that
into
jobs
yeah.
That
would
be
really
cool
yeah,
especially
if
we
could,
especially
if
you
could
make
that
repeatable
thing.
So
basically,
I
think
you
did
this
actually,
where,
like
somebody
says
like
hey,
I,
have
a
go
project
and
I
want
to
run
it
on
Juno.
It's
like
how
do
I
do
that
yeah
only
big
boxes
yeah
it.
C
Totally
exists:
I
can't,
for
legal
reasons,
opens
but
I'm
totally
planning
on
trying
to
I'm
gonna
work
with
the
whole
process
and
just
get
that
done,
because
I
worked
for
a
quite
an
amount
of
time
to
get
that
thing
built
I
think
it's
pretty
cool,
actually,
most
most
of
the
time
engineers
I'll
say
you
can
take
C,
C++
and
Python
and
package
that
you
can
package
eyes
it
and
put
it
on
to
do.
Nas
technically
is
anything
that
will
run
on
FreeBSD.
C
B
When
he
was,
we
were
talking
about
this
yeah
again,
I
looked
over
a
year
ago,
originally
and
I.
As
you
know,
one
of
my
currently
old
projects
that
I'm
trying
to
revive
and
rebuild
is
is
Todd
the
whole
distributed
testing
thing
and,
and
one
of
the
things
that
I
have
so
consistently
been
asked
for.
That
project
is
like
hey.
This
is
cool,
but
like
I'm,
not
on
the
server
team,
and
they
don't.
Let
us
spin
up
servers
and
I'm
like
that
sucks,
but
what
they
wanted.
B
What
they
want
to
do
is
run
these,
like
the
network.
Engineers
would
want
to
run
tools
like
this,
but
they
don't
have
servers,
they
got
their
switches,
and
so,
if
we
can
run
these
tools
as
small
as
they
are
just
on
a
router
or
on
a
switch
like
that
makes
a
world
of
difference,
and
so
this
is
very
interesting
to
me,
especially
if
you're
talking
about
a
language
like
go,
which
can
you
can
cross,
compile
very
very
easily
to
different
architectures
from
any
source
machine
like
I.
B
Do
this
for
a
syringe
which
is
which
what
powers
NRI
laughs,
I,
I,
compile
for
like
Windows
and
Darwin
and
Linux
just
all
from
one
machine?
And
so
you
combine
these
things
together
and
and
you've
got
a
really
powerful
way
to
bring
a
lot
of
these
new
ish
tools
into
the
world
of
where
network
engineers
can
like
actually
put
them
into
production.
You
know,
I,
think
it's
I,
think
this
powerful.
No,
it's.
B
A
A
C
A
I
think
we'll
wrap
it
up
then,
and
next
next
week,
we'll
I
think
we'll
cover,
maybe
in
a
little
more
detail.
Some
of
the
you
know
like
we
usually
do
every
time,
issues
and
stuff
like
that
PRS
that
are
open
and
just
have
a
look
at
what's
going
on
in
the
project
that
way,
but
I
think
we've
used
up
a
bit
of
our
time.
This
time,
plus
I
want
to
eat
lunch
before
my
next
call,
because
I'm
very
hungry,
so.
B
A
A
So
please
please
do
and
you'll
probably
he'll
be
the
first
to
let
everyone
know
when,
when
the
terraform
stuff
goes
public
and
don't
and
we'll
announce
when
his
lessons
are
live
on
the
interview
labs
website,
as
matt
said
sometime
this
week.
Hopefully,
if
something
goes
horribly
wrong,
then
early
next
week,
but
we'll
just
follow
at
enery
labs
on
twitter
and
you'll
you'll
you'll
be
you'll,
know,
you'll
know
when
it's
ready,
so
I'm
at
Klout,
toad
and
Matt
Oswalt
is
it
mirrored
in
mie,
RDI
n
and
thanks
for
watching.