►
From YouTube: Kubernetes SIG Service Catalog 20170619
Description
- Demo of API aggregation
- Demo of Atlassian Smith
- Questions about secret parameters
A
All
right
all
right
well
welcome
everyone
to
the
June
19th
edition
of
the
sixth
Service
Catalog
weekly
meeting
I'm
going
to
be
emceeing
today
and
today
we've
got
two
demos,
starting
with
Morgan
on
aggregation,
API
aggregation
in
our
back
and
then
I
am
sure.
I'm
pronouncing
this
name
incorrectly,
so
guide
I
think
it's
meal
term,
Atlassian
demoing
the
Smith
tool
that
he'd
built
it
looks
like
very
cool
stuff
and
then
we're
going
to
get
to
some
questions
and
comments
after
the
two
Devils
are
done.
A
B
Can
you
see
my
screen
now:
yeah,
okay,
yeah,
so
I
just
prepared
some
short
demo
for
the
for
our
project
with
Service
Catalog
use
case.
So
this
is
the
example
of
application.
We
can
run
using
Service
Catalog
resources,
provisioning
provision
through
Service
Catalog
like,
for
example,
have
a
just
a
bunch
of
running
instances
and
resources
like,
for
example,
for
reading
in
Amazon,
RDS
and
Q,
and
obviously
we
have
some
load
balancer.
B
So
to
implement
this
kind
of
application
in
communities
and
service
catalog,
we
need
a
bunch
of
objects
like
the
one
I
like
drawn
with
the
graph
here.
So
we
need
ingress
and
service
for
a
twelve
balancing.
We
need
a
bunch
of
cords
running.
We
need
Quatro
set
to
inject
secrets
as
environment
variables
in
support
and
into
like.
B
Obviously,
just
by
saying
like
hey,
we
first
create
instances
then
create
bindings
and
so
on,
but
the
better
way
of
doing
this
is
probably
to
declare
the
graph
of
dependencies
in
some
way
and
then
in
an
automated
way,
just
to
watch
for
the
states
of
all
these
objects
and
stop
creating
the
objects
depending
on
them.
So,
like
a
also
on
this
diagram,
you
can
see
that
there
are
two
independent
streams.
Why
cancel
the
Symbionese
could
be
Parisian
in
parallel,
while
what
cross
F
depends
on
like
both
resources.
B
This
format-
and
we
have
also
a
graph
of
dependencies-
just
described
else,
depends
on
whatever
like
binding
depends
on
instance,
what
process
depends
on
two
bindings
deployment
depending
on
proper
set
and
so
on.
So
this
depends
on
attribute
implicitly
defines
the
order
in
which
these
objects
will
be
created,
and
also
we
have
support
for
completing.
But
it's
not
necessarily.
We
can
obviously
like
if
we
know
all
of
these
values,
we
can
just
copy
paste
them
between
the
objects,
because
we
know
that
like
this
is
the
name
of
the
secrets
to
be
created
and
so
on.
B
A
B
C
A
B
A
B
B
B
B
B
I
should
see
that
winding
status
is
changing
over
time
and
over
the
subject
should
appear
in
particular
order,
which
is
important,
the
most
important
part
here.
So
it's
really
quick
here.
So
it's
already
shows
that
it's
ready
and
there
is
a
deployment
of
two
ports
running
I-
can
show
you
here
in
the
lock
what
is
actually
happening
behind
the
scene.
So
every
time
any
event
is
happening
to
the
bundle
itself
like
in
the
specification
of
the
bundle
or
to
any
of
the
object
defined
in
the
bundle.
B
We're
abeam
reveal
the
graph
and
try
to
make
the
actual
stage
reflects
the
desired
State.
So
we
build
a
sorter
graph
here
and
then
we
visit
every
single
object
defined
in
this
specification
and
check.
If
it's
ready
or
not
and
the
state's
ready
it
could
be
different
like
for
the
built-in
objects,
it's
just
call
it
inside
Smith,
but
for
the
TPS
there
are
annotations
supported,
so
you
can
define
any
field
in
your
llamo
or
JSON
to
represent
the.
What?
B
What
does
it
mean
for
this
type
of
object
to
be
ready
and,
like
you
can
show
you
like,
for
example,
here
you
can
see
that
the
point
of
appointment
is
required
by
resource
service,
but
it's
not
ready,
which
means
that
until
the
deployment
will
be
will
have
a
status
ready.
So
this
one
we
created
and
one
deployment
status
changes.
We
revealed
it
again
and
eventually.
B
B
So
is
it
not
just
fields
what
the
graph
once
we
can
actually
edit
something
here,
and
you
can
see
that
it
will
rebuild
the
whole
graph
and
we'll
update
the
corresponding
object?
If
you
remove
the
object
from
the
specification,
it
will
drop
them
and
so
like,
for
example,
let's
change
the
deployment
to
have
three
replicas
instead
of
two
and
let's
update
the
version
of
nginx
here.
B
B
Okay,
I
think
I
have
removed
this
debug
logging
doesn't
matter,
but
you
can
see
that
all
the
other
objects
remains
the
same.
They
weren't
recreated
or
anything,
but
the
corresponding
objects
here
were
recreated
or
not
to
really
updated,
and
so
potentially
this
could
be
used
for
any
update
in
the
graph.
B
But
for
now
there
are
limitations
like,
for
example,
if
you
update
secret
I,
don't
think
that
port
will
be
recreated
and
other
cases
like
that,
but
if
they
disappoint
for
these
changes
will
be
added,
it
will
be
whichever
just
I'm,
taking
only
applying
the
changes
you
need,
not
in
all
the
objects
or
already
starting
or
running
the
opponent.
You
can
update
any
object
in
the
specification
and
then,
if,
if
the
smooth
strokes
to
make
it
happen,
so
writing
the
square.
That's
it
I
cannot
answer
some
questions.
D
B
A
Neal
I
was
actually
yeah,
I
can
explain
a
little
bit,
but
also
I
was
actually
going
to
ask
how
you
guys
are
using
this
internally
and
Atlassian
too,
but
it
yes,
as
far
as
I,
can
see
I'm
not
super
familiar
with
this
tool.
Yet,
but
as
far
as
I
can
see,
this
deals
a
lot
more
with
dependencies
than
helm.
Currently
does.
A
B
A
B
Good
to
look
at
just
an
example,
so
we
support
deployments
and
other
types
of
objects
and
also,
as
I
said,
we
support
support
resources.
So
you
can
define
your
own
object
types
and
you
don't
have
to
add
their
supporting
sneeze
itself.
You
just
apply
annotation
I,
don't
think
I
have
I,
don't
think.
I
have
an
example
right
now,
but
it's
just
which
we
define
the
JSON
path
to
the
field
inside
your
object,
which
represents
the
state,
and
then
you
define,
what's
the
value
of
it,
to
define
that
it
is
ready
like,
for
example,
for
bondo.
B
It
will
be
a
path
to
the
condition
with
the
type
ready
and
the
value
should
be
true.
That's
the
way
it
works.
Yes,
you
correct
that
it's
all
about
the
dependency
graph
here
and
also
we
store
it
in
communities,
so
we
use
reconciliation,
loop
mechanism
of
communities
to
make
it
to
make
it
happen.
Yeah.
D
So
I
should
clarify
it's.
It's
obvious
to
me
that
this
does
you
know
that
this
understands
a
dependency
graph
that
that
helm?
Doesn't?
You
know,
understand
there,
there
is
nothing
in
hell,
but
that
can
do
that.
Currently,
it's
it's
a
feature.
That's
been
requested,
so
I'm
curious
that
there's
like
a
lot
of
overlap
here,
right
that
this
deals
with
bundles,
whereas
helm
deals
with
the
chart.
But
fundamentally
it
seems
to
me
they're
both
the
same
thing,
they're
both
different
words
for
for
package.
D
E
C
C
E
Trying
to
hollow
out
an
API
such
that
you
could
plug
in
things
like
this
and
app
controller,
to
be
able
to
do
your
dependency
resolution
for
you
and
then
the
annotations,
and
it
changes
to
manifest
that
you're,
showing
you
can
still
template
these
data
tom.
So
I
think
these
things
are
not
I.
Think
the
racially
exclusive
basically.
B
E
E
B
So
in
the
future,
I
think
that
we
can't
stop
using
if
you
aggregate
about
the
base,
support
of
Cisco
was
added
before
so
it's
not
even
released
or
yet
right.
So
we
for
now
we
just
say
which
really
support
the
Service
Catalog
URL,
and
we
know
that,
for
instance,
in
binding
sites
with
this
group
and
version
Service
Catalog
is
the
server
owning
these
type
of
resource.
B
Can
take
a
look
at
this
here?
What
they
keep
differences
here
and
yeah
if
it
makes
sense
to
put
something
upstream,
we
can
obviously
do
something
I'm,
not
sure
if
it
makes
sense
on
all
that.
For
us,
it
makes
sense
that
the
question
was
like:
how
do
you
use
it?
Actually,
the
Service
Catalog
example
is
usually
one
of
the
examples
which
will
be
used
in
our
case,
but
it's
so.
This
is
one
of
the
use
cases
in
general.
B
We
we
will
I
think
we
will
use
it
whenever
we
need
to
create
some
objects,
depending
on
other
objects,
with
the
particular
order
and
like
currency
for
service
catalog.
If
you
want
to
deploy
application,
do
you
have
to
define
how
to
charge
for
your
application?
Is
it
the
way
you
expect
it
to
work?
Now,
there's
there's
two.
B
A
So
what
we
do,
what
service
catalog
does
is
kind
of
just
tries
to
submit
everything
at
once
and
soon
when
it's
when
the
aggregator
is
running
in
1/7,
you
would
be
able
to
submit
all
that
at
once
without
regard
to
the
ordering
via
a
helm
chart,
and
then
the
expectation
would
be
that
you
know
the
the
binding
would
fail
until
the
instance
is
completed
and
the
app
would
fail
to
start
until
the
binding
is
complete.
So
yeah
it
is
Service.
Catalog
is
not,
inter
instance,
or,
inter
instance,
binding
aware
dependency
aware.
A
A
E
The
other
thing
that
is
part
of
this,
like
problem
space,
which
we
haven't
talked
much
about
yet,
is
that
a
long
term
when
you
create
a
pod
preset
any
deployment
or
controller
whose
pod
template
spec
matches
the
labels
of
that
odd
percent
should
automatically
get
recreated.
A
new
generation
of
pods
should
be
created.
E
That's
not
already
implemented,
but
it's
one
of
the
things
that
I
would
like
to
discuss.
Four
criteria
to
take
pod
presets
of
data
is
that
that
you
should
be
able
for
any
in
controller
that
creates
generations
of
pods
I.
Think
if
you
have
a
new
pod
preset,
that's
created
or
one
that's
deleted,
you
should
recreate
a
new
generation
of
pods,
so
you
automatically
get
the
newest
stuff.
Okay,.
B
B
This
could
be
also
a
solution
for
this,
because
I
the
difference
here
is
that
it's
not
just
like
a
specification,
it's
more
like
if
you
have
like
a
template
of
the
theme,
and
you
need
to
inject
some
variables,
which
will
be
available
only
while
other
objects
are
created
and
these
various
are
available.
This
is
the
way
we
could
solve
it
by
this
templating
yeah
yeah.
A
So
one
one
piece
that
we
had
discussed
I
think
in
our
first
face-to-face,
which
was
now
I,
don't
know.
Maybe
that
was
over
nine
months
ago,
or
maybe
more
I
don't
remember,
and
that
was
if
an
instance
has
a
dependency
on
a
binding
or
another
instance
in
that
is
that's
something
that
I
see
this
Smith
tool,
handling
and
Neal
knew
and
I
talked
I
thought
I.
A
Think
I
talked
with
you
and
three
of
your
co-workers
last
week,
Neal,
and
that
was
that
was
one
thing
that
I
thought
was
very
interesting
that
you
guys
handle
this
case
because
Neal
I
don't
know.
If
you
can
talk
about
some
of
your
internal
use,
cases
that
it
lasts
in
there
was
I
think
there
was
one
that
you
hinted
at.
That
was
that
was
fairly
interesting.
B
A
B
We're
building
the
internal
costs
for
running
arbitrary
application
and
some
of
the
applications
could
require
databases,
resources
others
could
require,
like
other
applications,
to
run
sidecars
and
so
on.
So
it
would
surely
associated
completely
generic
tool,
which
can
just
create
objects,
independent,
like
in
dependency
order,
so
I'm
not
sure
if
I
can
like
I,
can
give
you
an
example
which
can
clarify
this,
but
I'm
not
sure
what
can
I
bring
here
to.
E
B
F
What
was
the
so
I
kind
of
miss
Aaron?
Think,
like
you,
were
saying
something
about
findings
between
instances
and
then
we
were
kinda
like
teeing
it
up.
For
me,
you'll
say
something
special,
so
maybe
you
can
just
go
ahead
and
believe
the
suspense
and
say
what
were
you
after?
What
did
he
want
you
to
say:
I,
don't
know
I
I.
A
F
A
A
F
B
Yes,
one
of
the
use
cases
which
are
very
different,
I
guess
from
the
other
source
code.
All
these
cases
is
that
we
don't
only
want
to
run
communities
applications
to
manage
applications
running
on
communities,
but
we
also
want
to
use
the
same
set
service,
catalog
solution
for
running
applications
in
amalgam,
for
example.
B
So
we
can
like
reuse
the
same
resources
provision
through
Service,
Catalog
and
random
as
easy
to
in
Amazon,
not
just
as
containers
in
communities
and
that's
an
example
of
whether,
where
you
need
to
have
dependencies
between
like
you,
can
you
need
to
provision
resource
you?
Can
you
need
to
generate
credentials
using
binding,
but
then
you
need
to
inject
these
credentials
into
the
other
instance
of
like
application
type
or
something
like
that
computer
type
whatever
so,.
B
For
now,
yes
upon,
our
I
think
it's
just
one
instance
of
Service
Catalog
running
in
and
managing
not
only
two
units
cluster
or
like
electing
results
into
Cuba
native
culture,
but
also
just
like
I.
Actually,
we
are
like
building
as
I
said,
ours,
which
needs
to
store
the
state
of
the
system
and
then
orchestrating
creation
of
all
the
resources
deployment
and
so
on.
So
we
want
to
use
a
log
both
for
running
applications
in
communities,
using
our
resources
provision
in
cloud
somewhere
and
also
we
want
to
use
it
for
deploying
applications
themselves
outside
of
communities.
F
G
G
Okay,
to
start
out,
let
me
just
show
what
we've
got
running,
so
we
can
hopefully
see
that
there's
sort
of
nothing
up
my
sleeve
I'm
running
a
that
brantas
cube
D&D
cluster,
saying
which
sets
up
a
couple
of
nodes,
and
you
know
that's
not
important,
but
basically
that
explains
why
there's
there's
three
proxies
running
and
there's
big
chasm.
Look
at
a
poll
cluster,
so
we
got
basically
sort
of
the
standard
cube
set
up
going.
G
I've
got
tiller
running
because
you
know
movies
now,
charts
that
we
have
I'm
going
to
show
you
the
API
server
config
here,
just
just
to
say
right
now,
I
don't
have
our
back
on
zai
it's
on,
but
it's
in
you
know
permissive
mode.
This
is
not
the
way
you
would
run
it,
but
you
know:
I
got
to
run
it
right
right
now
to
before,
until
I
can
figure
out
exactly
how
what
the
what
my
hang-up
is
in
with
it
with
the
patch
that
I
have
so
I'm
gonna.
G
Visit
there's
a
you
know,
sort
of
a
readme
in
the
PR
for
this.
Basically,
what
you
need
to
do
is
you
need
to
have
a
an
API
registration
endpoints
and
create
an
API
service
and
into
that
API
service.
You
shove
certificates
that
you
end
up
also
giving
to
the
API
server
itself
that
basically,
they
authenticate
to
each
other
and
the
API
service
is
the
thing
that
puts
the
our
API
into
the
aggregator
and
so
it'll
be
visible
to
an
external
clients
so
which
we
start
with.
G
That's
not
the
one
I
wanted,
but
that's
that
stuff.
Those
are
the
API
service
objects.
These
is
all
the
basically
the
api's
that
are
running
and
are
sort
of
registered
internally
with
kubernetes.
What
I
wanted
to
show
you
was
it's
not
yet
it's
just
a
bedrooms.
Okay.
So
to
start
with
the
first
thing
you
have
to
see,
if
you
have
to
see
this,
this
API
registration
API,
and
that
is
where
all
the
API
services
get
registered
and
if
you
don't
see
that
is
even
going
to
work.
G
Eventually,
later
we'll
see,
I'll
run
the
same
command
and
we'll
see
the
Service
Catalog
in
here
fingers
crossed
automated
all
of
the
setup
into
into
a
script
and
yet
see
which
branch
and
okay
I'm
in
the
right
branch,
I
put
it
in
contribu
I,
would
like
to
figure
out
how
to
shove
this
into
the
helm.
Chart
later
and
I
forgot
to
do
this
again.
G
Now
it's
called
something
different.
Now
it's
called
that,
so
it
might
actually
be
executable,
but
this
is
going
to
run
through
and
it's
going
to
generate
a
CA
and
it's
going
to
generate
a
certificate
and
again
a
private
key
and
okay.
Now
we've
got
a
bunch
of
stuff
in
here.
We've
also
got
a
bunch
of
environment
variables,
set.
G
G
The
authentication
is
so
that
the
things
the
pieces
talk
to
each
other
and
the
API
server
once
it's
installed
behind,
will
will
actually
ask
the
the
aggregator
what
it
can,
what
it
can
do
and
that's
what
this
enables
such
that
works
and
the
aggregator
is
we're
telling
it
okay
you're
going
to
be
installed
behind
the
aggregator,
and
this
basically
sets
just
a
flag
on
the
standard.
Api
server,
there's
no
special
magic
there.
It's
we'll
be
using
the
API
machinery.
G
G
It
knows
that
if
there's
authentication,
it's
got
to
ask
for
permission
to
do
certain
things
right
now.
We
have,
you
know
permissive
mode,
enable
so
that
that's
always
going
to
be
true
and
then
basically
we
set
the
three
search.
These
three
parts
of
the
certificate
bundle,
so
we
set
a
CA.
We
set
the
you
know
the
serving
certificate
key
underneath
the
private
key,
and
then
all
of
this
is
just
done
sort
of
in
line
by
base64
it
and
shove
it
in
there
debug.
G
G
E
G
E
A
G
A
A
G
E
A
D
G
A
B
B
E
I
can
speak
to
that
as
so.
The
issue
now
is
asking
that
is
passing
secret
data
as
parameters
we
got.
Derek
got
started
on
that
and
we
got
some
guidance
from
the
sig
off
saying
both
that
we
should
not
use
secrets
for
things
that
do
not
target
pods
and
that
also
we
should
straighten
out
all
of
the
things
that
already
use
secrets
that
don't
cogs
to
avoid
using
watch,
and
the
reason
for
that
last
part
is
that
the
the
best-known
component
as
an
example
of
something
that
does
this
is
the
cubelet.
E
E
Notifications
for
those
resources,
a
single
watch
channel,
so
the
TLDR
there
is
basically
that
we
were
thrown
into
confusion
by
guidance
from
say
god.
I
actually
went
in
attendance
to
go
up
to
act,
asked
about
this
and
we
decided
that
use
of
secrets
is
okay
in
this
API
and
basically
I.
Think
since
then,
Derek
has
been
distracted,
okay,
I
think
work
items
by
he
is
okay.
Sorry,.
B
Go
ahead,
I
was
just
going
to
say
that
there
were
the
cigars
meeting
where
they
said
that
they
claimed
that
secret
shouldn't
be
ever
used
for
any
use
case
outside
polls
like
they
do.
They
said
that
it's
not
the
case,
and
there
are
limitations
like
you
can't
set
up
global,
which
is
on
all
the
changes
of
secret
and
also
they
like
they
try
to
I
think
they
are
considering
to
line
up
the
changes
for
secrets
for
the
future
to
separate
the
actual
sensitive
data
from
the
watching
the
metadata
around
it
like
you
like.
E
That's
that's
accurate,
that's
that's!
The
meeting
that
I
was
referring
to
so
yeah
I.
Think
the
TLDR
is
that
we
were
thrown
into
confusion.
I
Derek
had
some
other
duties
that
I
took
I,
took
his
focus
off
of
the
secret
stuff.
I
might
pick
that
back
up
this
week
or
he
might
pick
it
up.
He
was
going
to
be
taking
some
PTO
this
week.
It's
it's
something
that
I
personally
want
really
bad
and
I
have
a
requirement
for
it.
I
think
that
basically,
we
just
became
confused
and
then
got
distracted,
but
okay
I
can.
B
E
I
think
that
we
had
consensus
on
it,
but
everything
had
a
big
question
mark
asked
after
it.
After
we
got
this
guidance
which,
for
the
record
was
inconsistent,
prince2
got
because,
like
I
said
they
said
they
said
both
that,
oh,
you
shouldn't
use
secrets
for
things
that
don't
argot
pods,
and
also
that
all
the
listing
things
that
you
secrets
that
don't
target
pods
of
which
there
are
several
I,
should
be
migrated
to
use
a
pork
watch
of
the
time
that
we
described
here.
So
it
wasn't
consistent.
E
B
So
you
can
take
a
look
at
the
notes
from
the
missus
mayor
thing
meeting
from
cigar
and
I
think
they
meant
it
there
that,
like
they,
don't
say
that
it
can
be
used
for
any
other
use
case,
it's
more
like
for
the
future.
They
want
to
separate
the
concerns
by
separating
the
data
from
metadata
and
stuff
like
that,
and
also
they
explicitly
said
that
you
should
never
watch
on
the
on
the
secret
type
of
object.
You
should
watch
on
the
particular
instance
of
secret
and
that's
I
think
that's
fine.
E
For
our
use
case
right
so
that
that
watching
a
particular
instance
of
the
secret
is
the
book
watch
facility
that
I
was
talking
about
so
like
the
use
case
for
cubelet,
is
that
cubelets
should
only
watch
the
secrets
that
are
consuming
by
pods,
scheduled
on
to
the
node.
The
cube
it
manages
the
cubelet
will
do
is,
as
it
gets,
update
events
and
add
events
on
delete
events
or
new
pods,
scheduled
on
to
the
node
and
pods
that
have
gone
away.
E
It
will
add
and
remove
watches
for
those
individual
secrets
and
then
there's
a
special
authorizer
that
in
the
api
server
says
yes,
you're
a
secret
or
really
cubelet,
and
it
walks
the
reference
chain
of
like
establishing
that
there's
actually
a
pod
scheduled
on
to
the
node
that
cubed
manages
and
that
the
secret
being
watched
is
referenced
by
one
of
those
pods.
So
eventually
we
may
have
to
do
something
similar,
I'm
not
sure
if
the
authorizer
portion
of
this
is
going
to
be
stable.
E
It's
it's
apparently
very
expensive,
surprise,
surprise
to
to
implement
that
authorizer.
But
as
far
as
the
future
scenarios,
where
we
support
parameter,
updates,
I
think
that
we
would
definitely
use
the
board
watch
facility
for
we
we
would
go
one
by
one
watch
of
individual
secrets.
Add
them
to
the
boat
watch,
get
update
events
for
the
metadata
when
the
metadata
is
changed,
and
then
there's
going
to
be
a
special
sub
resource
of
Secrets.
That
actually
gives
you
the
secret
data,
so
that
the
watch
channel
doesn't
carry
the
secret
data.
Okay,.
A
So
Morgan
it
looks
like
Morgan
is
able
to
demo
the
aggregation,
I
think
I
would
sort
of
like
to
see
it
I
think
a
lot
of
us
would
as
well.
So
we
I
know
what
I
did
wrong.
Okay,
I'd
like
to
move
on
to
that
and
then
Neil
I
know
you
have
a
couple
other
questions.
If
you
could
go
into
the
Google
group
and
pose
them
there,
that
would
be
very
helpful
and
we
can
all
kind
of
jump
on
and
address
them
as
we
can.
Okay.
B
G
Okay,
I'm
just
scrolling
backwards
to
show
you
like
what
happened.
It
wasn't
exporting
the
variables
so
I
was
just
trying
to
do
a
bunch
of
blank.
All
these
things
were
blank
because
I
forgot,
you
have
to
source
it,
not
not
just
run
it
and
then
I
still
had
debug
on,
and
so
it
printed
out
all
the
stuff
but
I
installed.
It
is
this.
Is
it
being
installed
and
it's
setting
all
of
the
all
of
the
junk
and
that's
not
terribly
important,
but
you
can
see
at
the
end.
You
know
we
got
the
standard
stuff.
G
G
G
G
But
you
know
there's
only
one
yeah,
it
is
only
one
thing
in
queue:
config
I
didn't
set
a
context,
so
we're
using
the
default
context
to
ete
tests,
run
that
that
creates
a
broker
and
deletes
a
broker,
and
that
needs
the
shell
variable
set
and
all
I've
got
set
is.
Is
this
and
you
can
see
they're
set
to
the
same
thing?
This
is
the
one
for
the
Service.
Catalog
is
one
for
the
cube
config
and
it
works,
and
if
you
got
some
animal
for
me
to
slam
in
there,
real
quick
I
can
do
that.
A
G
A
G
A
G
G
G
A
G
E
E
E
A
Now
so
who
did
it
awesome
to
see
all
right?
So
we
have
two
minutes
left
I,
don't
think
it's
worth
starting
up
on
a
new
topic
be
given
the
time
constraints.
So
I
would
like
to
say
thank
you
for
everyone
for
coming
today.
Neal
again,
thank
you
so
much
for
your
demo
as
well
post.
The
remaining
questions
that
you
have
in
Google
group
and
also
VA
and
Neil
I,
would
love
to
see
you
guys
have
that
conversation
that
I
had
to
cut
you
off
and
open
on
the
Google
group
as
well.