►
From YouTube: Kubernetes graph based visualization - Jack Greenfield
Description
Recorded on Feburary 25th, 2015 at the Kubernetes Gathering in San Francisco, CA, USA
A
So
my
name
is
Jack
Greenfield
I'm,
an
engineer
at
Google,
and
my
interest
has
been
for
a
long
time
and
still
is
application
configuration.
So
one
of
the
themes
at
the
conference
here
has
been
that
kubernetes
is
about
raising
the
level
of
abstraction
you're
working
with
applications
you're,
not
just
thinking
about
machines
right.
A
You
guys
like
to
put
together
some
UI
for
a
little
demo
down
in
San
Francisco.
So,
three
weeks
ago
we
set
out
from
scratch
and
the
kiss
mattock
team
built
the
chrome
and
the
dashboard
pieces
that
you've
seen
and
our
team
built
this
graph
and
like
he
said
it
was
very
fast-paced
and
a
lot
of
fun.
In
fact,
the
designer
who
came
on
board
initially
didn't
have
much
time
to
give
to
our
project.
A
He
was
fully
involved
elsewhere,
but
he
got
so
excited
by
the
speed
of
development
that
he
started
working
around
the
clock,
and
so
he
was
able
to
give
us
lots
of
great
feedback.
That
said,
we
still
have
a
lot
to
do
so.
Please
bear
with
us
on
the
any
sort
of
rendering
or
styling
issues
needs
work,
but
overall,
what
you're
looking
at
is
a
graph
control
using
d3.
Again,
it's
the
angular
material
design,
the
angular
framework,
material,
design,
look
and
feel,
and
what
this
is
all
about
is
visualizing
your
clusters.
A
So
this
is
actually
running
against
live
data
and
I'm
actually
going
to
try
and
hit
the
internet
here,
despite
all
the
smart
phones
in
the
audience
and
see
if
we
can
fetch
some
live
data
and
it
doesn't
look
like
we
can.
But
that's
okay,
because
I
actually
have
canned
data
that
I
can
use
instead
and
when
I
fetch
the
data
rear
Enders
the
the
graph
forming.
A
But
the
interesting
thing
about
this
is
that
I
can,
if
I
want
to
play
with
this,
and
so
one
of
the
things
we're
thinking
about
is
how
do
people
want
to
visualize
things?
How
do
they
want
to
understand
their
cluster?
So
here
we
have
the
cluster
in
the
center
we
have
a
number
of
nodes
which
are
these
larger
red
ones.
We
have
replication
controllers,
which
are
these
guys.
We
have
services
which
are
these
guys.
We
have
pods,
which
are
these
guys.
A
We
have
containers
which
are
these
blue
guys,
and
then
these
are
the
processes
that
are
running
in
the
container.
Okay,
what's
kind
of
cool
about
this,
is
you
can
sort
of
see
this
in
the
legend
and,
if
you
wanted
to
say,
get
rid
of
the
clusters
and
the
nodes,
and
maybe
the
images
you
can
see
here,
the
images
that
all
the
containers
are
running
right?
Well,
not
so
interested
in
the
images.
So
let
me
just
focus
on
the
things
that
are
really
exciting:
containers,
pots,
services,
replication
controllers
and
keep
the
processes
around
for
good
measure.
A
What
we
can
do
now
with
this
visualization
is,
we
can
select
things
and
look
at
them.
So,
for
example,
we
can
zoom
in
on
a
particular
piece
of
this.
Now
we're
still
playing
with
this,
but
one
of
the
things
we
thought
we
would
do
is
take
the
selection
and
throw
away
everything
else
right
or
just
keep
the
selection
just
give
you
ways
of
exploring
the
the
structure.
That's
here
right.
A
The
other
interesting
thing
that
we
decided
to
do
was
give
you
a
way
to
zoom
in
on
details.
So
when
you've
selected
something
you
can
see
the
details
of
the
selected
node
over
here
on
the
left
hand,
side
and
these
details
are
sort
of
a
digest.
If
you
will
oftentimes
you're
going
to
want
to
see
the
whole
shooting
match,
so
you
can
come
over
here
and
scroll
on
down
and
get
into
all
of
the
information
that's
available
about
the
node,
which,
for
example,
for
a
container
includes
all
the
environment
variables.
A
A
Interesting
thing
about
this
data
is
a
lot
of
it.
It's
easier
to
get
your
head
around
your
cluster
when
you
can
sort
of
see
it
as
a
picture
if
you
will
and
explore
it,
and
so
we'd
like
to
also
turn
this
into
more
of
an
application
oriented
view
by
labeling.
So
one
of
the
things
I
did
want
to
show
you
guys
is
how
we
get
this
information,
so
it
turns
out
that
we're
actually,
first,
let
me
go
over
here
and
show
you
the
raw
feed
okay.
A
So
it
turns
out
how
many
people
here
have
actually
tried
to
pull
information
programmatically
from
kubernetes
clusters
to
do
stuff
with
it
yeah?
Okay,
you
know
it's
not
easy.
Yeah
there's
a
number
of
different
api's
and
kubernetes
and
docker
are
sort
of
different
sources
of
information,
and
you
sort
of
have
to
reason
over
the
data
to
correlate
the
stuff
that
you
get
back
from
each
of
these
api's.
A
And
so
we
have
a
team
that
built
the
backend
for
this
gentleman
named
Vaz
Bala
and
another
gentleman
named
Aaron
Gabbar
built
a
container
that
you
can
drop
into
a
cluster
which
pulls
the
metadata
from
all
the
different
api's
synthesizes
it
into
a
single
graph
and
feeds
it
up.
It's
very
cool.
So
what
we
did
was
we
took
that
and
we
wrote
a
nice
little
JSON
path
and
JavaScript
a
filter,
and
this
filter
is
very
simple,
but
it
allows
you
to
project
the
data.
A
A
So
it
was
a
collaborative
effort,
a
lot
of
fun.
You
had
a
great
time
with
kis
Maddock.
If
anyone
is
interested
in
dropping
in
a
tab
or
getting
at
the
data
using
the
container
that
I
talked
about
or
doing
anything
else
with
these
technologies.
Please
look
up
me
or
Patrick
Reilly
from
kis
madacorp
BC
Brassard
from
cosmetic
afterwards.