►
From YouTube: Keynote: Tales from the Kube: Contributions by the... - Brian Grant & Tim Hockin & Clayton Coleman
Description
Join us for Kubernetes Forums Seoul, Sydney, Bengaluru and Delhi - learn more at kubecon.io
Don't miss KubeCon + CloudNativeCon 2020 events in Amsterdam March 30 - April 2, Shanghai July 28-30 and Boston November 17-20! Learn more at kubecon.io. The conference features presentations from developers and end users of Kubernetes, Prometheus, Envoy, and all of the other CNCF-hosted projects
Keynote: Tales from the Kube: Contributions by the Community - Brian Grant, Principal Engineer & Tim Hockin, Principal Software Engineer, Google & Clayton Coleman, Architect, Kubernetes and OpenShift, Red Hat
While most people are familiar with how Kubernetes became an open-source project, there are other stories about its early days that you haven’t heard. This presentation will shine a long-overdue light on the invaluable contributions of a number of contributors who helped shape what the project has become.
To Learn More: https://sched.co/GswX
A
He
was
threatening
to
dance
good
evening.
Everybody
I
hope
you
guys
had
a
great
time
today.
I
know
I
did
we've
only
got
20
minutes,
so
let's
get
right
into
it.
This
is
cube
con.
What
is
our
favorite
topic
at
cube
con?
We
love
to
talk
about
community.
The
community
is
really
what
is
at
the
heart
and
soul
of
cádiz,
but
but
what
does
it
mean?
What
does
community
really
mean?
My
take
community
means
friendly
collaboration
right.
It
means
relying
on
people
trusting
them
to
do
the
right
thing.
A
It
means
bringing
real
commitment
to
the
project
and
building
a
shared
vision,
and
sometimes
you
know
you
disagree
on
stuff
and
your
passions
come
through
and
you
can.
You
can
argue
about
things
and
then
you
get
past
that
and
you
work
forward.
Our
community
is
made
up
of
many
people
from
many
backgrounds
who
bring
really
diverse
experiences
and
backgrounds
and
perspectives,
and
that
adds
up
to
a
result
that
no
one
person
could
possibly
have
created.
The
three
of
us
we've
been
here
with
the
project.
A
Basically,
from
the
beginning,
this
is
our
fourth
cube
con
or
at
least
fourth
cube
economy.
In
America
we've
seen,
we've
been
been
witness
to
tons
of
crazy
ideas,
things
that
came
out
of
left
field
that
we
never
could
have
thought
about
new
perspectives
that
just
illuminated
opportunities
that
we
never
would
have
thought
about,
or
just
contributions
from
people
who
just
stepped
up
and
made
things
better
people.
These
things
start
with
people,
people
who
step
up
and
say
I
have
an
idea
or
have
you
thought
about,
or
you
know
we
can
do
better.
A
A
Early
early
early
kubernetes,
we
drew
on
Google's
experience
with
work
before
kubernetes
in
cubelet
we
Borg
and
Borg
lit
and
Borg
and
Borg
lit,
evolved
into
a
project
that
we
called
oh
make.
Ax
and
omelette
omelette
and
Omega
led
into
kubernetes
some
of
the
people
who
are
working
on
omega
and
and
Borg.
A
For
example,
Victor
Marmol
fish,
new
Canon
and
Rohit
Jenna
Gaul.
They
worked
to
open-source
Google's
internal
container
tools.
We
called
it
LM
CTF.
Why?
Let
me
contain
that.
For
you,
it
was
a
very
different
model
from
docker
very
different
UX,
but
then
2013
does
docker
thing
came
out
and
man
it
took
off
right.
We
saw
that
it
had
wings
and
those
three
guys
they
went
and
they
shifted
their
energy
and
they
focused
on
contributing
to
docker
and
then
lib
container
and
eventually
run
C
and
OC
I.
This
thing
that
now
powers
our
entire
ecosystem.
A
This
also
spawned
the
C
Advisor
project,
which
brings
monitoring
and
stats
into
community,
has
been
a
really
important
piece
of
the
community's
project.
You
know
some
of
the
ideas
have
carried
forward,
Borg
had
allocations,
Omega
had
scheduling,
units
and
communities
has
pods
and
some
things
are
wildly
different.
Networking
and
storage
they're
not
even
close,
and
you
know.
Interestingly,
not
a
lot
of
people
know
this
I
wanted
to
name
kubernetes.
Let
me
schedule
that,
for
you
think
about
that.
A
The
next
time
there's
a
debate
about
naming
stuff,
so
I
got
involved
in
kubernetes
really
early
Google
announced
they
were
going
to
open
source
this
project
and
it
was
a
docker
Con
2014
in
San
Francisco,
and
there
was
this
I'm
also
once
they
was
like
an
era
of
excitement,
people
who
came
to
docker
Khan,
who
were
building
and
running
systems
that
use
docker.
There
were
people
who
had
already
started.
A
There
are
people
who
starting
new
things,
but
every
single
one
of
those
people
was
incredibly
excited
and
unfortunately,
unlike
Brian,
I,
don't
have
a
photographic
memory,
and
so
it's
very
difficult
for
me
to
remember
the
names
of
everybody
that
day,
like
I,
remember,
being
Brian
and
I
was
like
wow.
This
guy's,
really
smart
but
like
that
day
was
so
exciting
and
that
excitement
carried
over
and
you
know
as
a
as
a
community
it.
A
It
was
interesting
in
the
early
days
because
there
was
a
lot
of
focus
from
people
who
are
trying
to
build
systems
on
top
of
docker.
They
wanted
to
build
applications
at
scale
and
run
those
applications,
and
so
our
initial
set
of
people
who
are
very
interested
were
less
about.
You
know,
maybe
the
the
individuals
and
more
about
people
who
had
a
problem
to
solve,
and
we
had
a
theme
which
was
you
know
we
all
kind
of
agreed
we
didn't
want
to.
We
don't
learn
from
the
lessons
of
the
past.
A
We
didn't
want
to
make
any
of
the
same
mistakes,
we're
gonna,
learn
from
Heroku
and
we
wanted
to
embrace
ideals
that
docker
came
up
with
and
it
was
really.
It
was
almost
like
this.
We
thought
anything
was
possible
and
I
was
like
really
exciting
here,
because
you
know
five
years
later,
this
was
possible.
So
all
of
those
teams
kind
of
started
with
a
little
bit
of
a
bias.
A
They
were
bigger
and
they
asked
this
question
was
kind
of
an
important
question,
which
was:
how
do
we
split
this
up
among
a
set
of
teams
among
a
set
of
people,
so
we
should
have
come
up
with
a
code
word
like
chop,
okay,
so
one
of
the
first
things
that
we
did-
and
this
is
like
the
the
seed
of
multi-tenancy-
was
he
said.
Well,
we
want
to
split
these
things
up.
A
We
had
a
bunch
of
really
clever
and
overly
complex
ideas
and
I
really
want
to
give
credit
to
derrick
karr
and
to
Eric,
toon
and
daniel
smith.
Those
guys
iterated
in
the
open
in
this
community
from
derek
was
from
red
head
eric
and
daniel
had
worked
on.
It
worked
at
Google
for
a
long
time
these
taking
these
like
disparate
worlds,
open-source
enterprise,
the
way
they
do
things
at
Google
and
they
brought
them
all
together.
We
went
through
three
or
four
iterations.
A
Actually,
the
first
iteration
was
labels
and
we
couldn't
figure
out
how
to
make
that
work.
So
we
scratch
that
idea,
and
then
we
said
well,
maybe
we
don't
need
namespaces
at
all
and
they're
like
yeah.
Probably
gonna
need
something,
and
the
interesting
thing
is
in
the
end,
the
focus
that
really
was
about.
How
do
we
do
this
simplest
thing
that
could
actually
work
that
we
could
explain
to
someone,
and
so
we
did
this.
A
You
know
bucket,
then
you
put
stuff
in
and
we
couldn't
come
up
with
any
better
names
and
there
were
these
long
drawn-out
arguments
where
nobody
can
really
convince
everybody
to
get
it.
But
we
ended
up
with
the
name
of
name
space,
because
we're
unoriginal
like
that
and
this
name
space
actually
I
read
funnily
enough.
There's
all
these
places
in
the
code
still
that
if
you
pass
an
empty
namespace,
we
were
going
so
hard
for
usability
that
we
passed
in
default
instead
and
I.
A
B
Yeah
so
along
the
lines
of
trying
to
be
easy,
you
know,
while
we
Google
and
Red
Hat,
we
were
committed
to
making
kubernetes
work
well
for
all
of
our
users.
We're
also
fortunate
to
have
many
users
of
kubernetes,
even
early
on,
who
are
enthusiastic
and
intent
on
making
Kearney's
work
well
for
themselves
and
for
the
whole
community.
One
memorable
example
of
that
was
Sam
codes,
who
is
evaluating
kubernetes
a
box
around
artists
2014,
which
was
only
two
months
after
we
open
sourced
kubernetes.
So
there
are
a
lot
of
rough
edges.
B
Sam
wanted
to
improve
the
usability
of
our
command-line
tool,
which
at
the
time
was
cute
config,
but
it
was
basically
implemented
as
a
couple
of
big
switch
statements.
So
Sam
proposed
that
we
refactor
it's
making
more
modular,
testable
and
easier
to
contribute.
It
soon
became
clear
that
actually
a
redesign
was
needed,
not
just
a
refactor,
so
all
of
us
had
a
lot
of
input
on
the
design
that
Sam
brought
is
perspective
as
a
user
and
as
a
contributor
to
that,
and
one
of
the
more
significant
changes
Sam
proposed
was
to
actually
change
the
verb.
B
B
The
cue
config
to
you
was
the
working
name
for
the
tool
we
considered
a
number
of
other
names
such
as
cube
and
keep
it
so
but
keep
control
one,
but
we
still
can't
agree
on
the
nuances.
The
whole
design
and
development
effort
took
about
six
weeks
for
the
basic
truck
commands
like
traits
get
delete,
cute
controlled
describe,
was
born
out
of
this
initial
proposal
and
implementation,
but
the
more
complicated
commands
like
rolling,
update
and
apply
took
about
another
year
to
complete,
and
that
was
it
started
around
kubernetes
zero
point
four.
A
So
anybody
who's
used
kubernetes
may
have
heard
that
networking
is
a
little
complicated.
We
have
some
opinions
about
networking.
It's
not
always
easy
to
implement
those.
It
was
described
to
me
as
where's
my
floppy
drive
sort
of
moment,
but
there
was
a
bunch
of
smart
people
that
I
had
some
really
long
conversations
with
early
on
in
communities,
folks
like
Eugene,
yeah,
Cobo
and
Alex
Paulette
and
Christopher
Lilian
stopa,
and
did
Casey
Davenport,
who
helped
me
and
convinced
me
that
this
actually
wasn't
insane.
A
A
So
this
work
on
the
cni
project
started
in
core
OS
rocket
started
by
guys
like
John,
Jonathan,
Bowl
and
Eugene
Jakob
ovitch
again
in
late
2015,
we
merged
C&I
support
into
kubernetes.
That
was
folks
like
Raja,
Chopra
and
Paul
tbody.
These
guys
made
this
CNI
interface
in
committees,
which
is
now
the
the
standard.
This
is
what
everybody
uses,
and
now
there's
tons
of
CNI
plugins
out
there,
commercial
and
free,
there's
again
a
whole
secondary
ecosystem
of
vendors
who
are
producing
these
things
and
it
has
its
own
community.
A
It's
used
in
other
projects
and
in
fact
some
of
the
CNI
maintainer
z'
are
also
big
kubernetes
contributors.
Brian
poram
is
here
Casey
Calandra,
ello,
Dan
Williams.
These
guys
are
very
active
in
the
Signet
workspace.
Beyond
that
we
needed
an
HTTP
API.
There's
a
lot
of
demand
for
people
to
how
to
solder
I
set
up
my
my
HTTP
load
balancer.
So
we
invented
this
thing
called
ingress
and
it's
an
API
without
an
implementation.
He
said:
here's
here's
what
she
did
decode
to
almost
overnight.
We
had
implementations.
A
So
thanks
to
manuel
alejandro
de
Brito,
fountas
or
la
DBF
and
Prashant
Bala
Subramaniam,
who
implemented
some
of
the
very
first
ingress
implementations
I
touched
we
talked
early
about.
We
were
trying
to
ship
right.
It
was
really
important
for
us
to
be
able
to
deliver
something
and
I.
Remember
at
the
first
cube,
con
Tim
was
up
on
stage.
We
had
just
released
one.
A
Oh
and
Tim
was
taking
heat
from
the
crowd
about
this
ridiculous
claim
that
kubernetes
only
supported,
100,
nodes
and
Tim
was
trying
to
like
defend
this,
which
was
like
we
approach
this
with
respect
of
us.
What's
more
important
software,
that
is
nice
and
theoretical,
or
software
that
actually
ships,
and
we
said
we
want
to
ship
it.
We
set
goals,
and
we
said
you
know
we're
gonna
respond
to
API
requests
within
a
second
and
with
that
for
a
hundred
nodes,
but
there
was
so
much
feedback.
A
A
Sig
scalability
was
a
group
of
users
and
programmers
and
people
who
are
obsessed
with
performance
and
they
in
the
course
of
four
releases
executed
this
massive
shift
and
how
kubernetes
worked.
We
optimized
the
living
daylights
out
of
everything,
and
so
there's
plenty
of
folks
on
that
on
that
committee.
But
I
wanted
to
thank
folks
like
Tim
st.
Clair,
who,
at
the
time,
was
at
the
Red
Hat
performance
lab
for
really
being
a
key
driver,
a
data-driven
driver
of
that
community.
A
Folks,
like
America,
browse
key
and
voytek
Shanon
ski
I,
screwed
up
that
name
I
can
do
it,
I'm,
sorry,
I'm,
sorry,
boy,
tech,
I
apologize
so
Wojtek
and
america
came
from
the
google
labs
help
kind
of
set
in
place.
The
framework
that
we
use
for
testing
scalability
today
and
both
of
them
stuck
to
their
guns.
I,
remember
one
release
where
we
almost
didn't
ship,
kubernetes,
1/7
I
think
it
was
because
we
weren't
passing
a
scalability
result.
A
That's
how
seriously
on
the
sig
took
it
bob
wise
and
aaron
quicken
berger,
we're
very
involved
from
very
early
on
is
key
as
they
wanted
to
use
kubernetes
at
samsung,
and
they
were
very
motivated
to
do
that.
Those
guys
really
built
a
community
Erin
went
on
to
become
a
steering
committee
member
and
for
everybody
out
there.
Erin
has
a
tireless
advocate
for
just
doing
things.
He
runs
a
whole
sig
by
his
beard
and
so
and
as
well.
A
You
know
during
this
process
we
didn't
just
change
kubernetes,
but
the
sed
project
which
kubernetes
was
based
on
Shengli.
Was
the
team
lead
at
the
time
and
Derek
Carr
and
Shengli
had
a
huge
set
of
discussions
around
at
CD
2,
where
we,
we
gave
a
bunch
of
feedback
about
things
that
could
make
it
to
be
better
in
the
sed
team.
A
The
community
and
the
folks
at
core
or
less
came
together
and
built
at
CD
3
around
the
requirements
and
the
needs
of
kubernetes,
and
it
was
a
huge
shift
in
terms
of
how
we
could
run
crazy.
So
in
3/4,
sorry,
1/4
I
think
we
claimed
5,000
nodes
may
was
1/5,
but
we
went
from
a
hundred
to
5,000
and
we've
sustained
that
and
even
as
we
add
features,
six
scalability
has
really
been
in
there.
So
thank
you
to
everybody
on
that.
A
B
Going
to
talk
about
some
more
users
that
made
significant
contributions
to
kubernetes
from
the
beginning,
we
wanted
communities
to
work
well
everywhere,
but
Google
didn't
have
the
expertise
to
add
support
for
other
clouds.
So,
within
the
first
few
months,
several
cloud
providers
were
added
by
the
community.
B
However,
not
all
of
these
data
up-to-date
with
the
rapid
development
of
the
project
in
those
early
days,
a
more
sustained
effort
was
needed.
Justin
Santa,
Barbara,
start
started,
contribute
around
2014
and
provided
exactly
that's.
He
started
by
me
the
Cuba
provisioning
script,
more
reliable
on
one
of
the
other
major
clouds.
B
He
then
moved
on
to
implement
missing
features
like
volume,
support
and
even
set
up
testing
to
prevent
regressions.
Once
he
started
to
work
on
new
features
like
high
availability
and
multi
zone,
support
and
upgrades
of
the
clusters,
he
quickly
hit
the
limits
of
what
was
seen
to
do
in
a
shell
script,
so
he
built
cops,
cops
the
since
become
one
of
the
most
popular
open-source
cluster
lifecycle
tools
for
kubernetes
and
not
only
is
justin
a
cop's
maintainer,
but
he
also
became
the
owner
of
that
cloud
provider.
B
Implementation,
one
of
the
top
contributors,
the
project
overall,
we're
fortunate
that
we
have
a
number
of
contributors
like
Justin,
who
are
very
passionate
about
the
project,
while
other
clouds
were
an
obvious
obvious
target
environments
for
kubernetes
and
about
mid-2015.
An
environment
emerged
that
none
of
us
that
it
hadn't
occurred
to
any
of
us.
B
He
initially
did
that
in
his
own
repository
kubernetes
unarmed,
and
then
he
started
to
work
to
upstream
the
changes.
His
first
PR
was
a
one-line
PR.
It
was
approved
by
Tim
I
think.
Eventually,
these
20
PRS
were
required
to
get
it
completely
working.
The
solution
was
built
on
our
multi
node,
so
called
getting
started
guide,
which
was
a
guide
intended
for
small,
simple
in
use
cases
and
to
be
completely
portable
but
she's
as
well.
So
Lucas
started
to
improve
that.
A
So
we
we
call
that
a
bunch
of
people
by
name
today,
but
there's
no
way
that
we
could
possibly
name
everybody.
We
only
have
20
minutes.
So
instead
we
put
the
names
of
everybody.
Who's
made,
kubernetes
a
special
project
onto
this
slide.
So
a
little
hard
to
read,
but
I
promise
that
everybody's
name
is
in
there
and
to
wrap
up
today
the
AV
guys
flipped.
When
we
put
a
black
slide
up,
we
want
to
wrap
up
we'd
like
to
share
just
sort
of
one
anecdote,
each
that
sort
of
sticks
with
us.
A
I'm
really
highlights
for
us
the
project
so
I'll
start
I
have
a
very
clear
memory.
It
was
an
early
meetup
in
Seattle.
There
was
about
1518
people
in
Seattle,
we're
hanging
out
in
the
bar
room.
Upstairs
the
the
meeting
room
upstairs
and
I
was
complaining.
That
I
was
having
trouble
getting
my
NGO
development
environment
to
work
and
I
was
having
some
trouble
getting
some
things
to
work
and
Sam
goads
said.
Let
me
help
and
he
sat
down
at
my
laptop
good
god.
A
B
Know
who
he
was
and
we
started
talking,
we
realized
that
at
least
I
thought.
We
had
a
lot
of
shared
operational
experiences
and
then
once
we
realized
we,
we
were
gonna
work
together.
I
had
a
lot
of
confidence
that
the
relationship
was
going
to
work
really
well,
and
then
we
could
really
rely
on
Clayton
to
make
good
design
decisions,
so
that
was
I
previously
worked
on
just
internal
infrastructure
at
Google
and
hadn't
really
worked
with
the
open
source
community.
So
that
was
a
really
big
shift
for
us,
so
I.
A
Had
an
anecdote-
and
this
morning,
I
actually
had
an
epiphany,
so
I'm
gonna
go
a
little
bit
off
script
here,
so
I
at
the
LES,
yes
Ari
at
the
last
round.
We
are
anyway
I
got
the
mic.
I
realized
at
last
cute
kind
of
askew
con,
but
I
don't
really
make
it.
The
sessions
I
end
up
sitting
outside
in
the
hallway
I
talk
to
people
and
at
first
I
thought.
It
was
because
we're
trying
to
solve
problems
and
get
stuff
done
and
I
actually
realized.
These
people
are
my
friends.
These
people,
I
trust
them.
A
I,
see
them
like
once
a
year,
sometimes
like
I,
haven't
seen
you
guys
since
last
December
and
I
realized
that
you
know
this
community.
The
set
of
people
folks
like
Tim
and
Brian
folks
and
the
various
SIG's
who've
been
around
since
the
beginning
and
folks
who
just
started
like
every
single
one
of
these
people
is
someone
that
I
like
working
with,
and
it's
not
the
technology
that
keeps
me
here.
It's
these
guys,
and
so,
like
I
kind
of
wanted
to
say,
like
you
know,
we
as
a
community,
we
do
a
lot
of
stuff
for
technology.
A
A
A
About
2018
that's
this
year,
so
every
if
you've
seen
a
pole,
so
there's
two
things:
I
want.
First,
this
community
works
because
every
one
of
you
works
with
other
people
make
a
friend
build
that
friendship,
friendship
and
Trust
are
the
things
that
keep
this
community
going.
If
someone
walks
out
of
here-
and
you
don't
know,
I'm,
say
hi,
that
would
make
me
happy,
because
I
really
hope
that
we,
you
know
the
community
can
commute
us
together.
Second
thing:
oh,
we
don't
like!
So.