►
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: Everything Worked Before Kubernetes - Vicki Cheung, KubeCon + CloudNativeCon North America 2019 Co-Chair & Engineering Manager, Lyft
https://sched.co/UdQE
A
Hey
good
morning,
please
bear
with
my
voice
because
I
have
a
cold
and
have
completely
lost
it,
so
everything
was
working
before
kubernetes,
but
not
me,
I,
guess:
okay,
so
I'm
Vicki
I
work
at
lift
I've
been
running
kubernetes
for
different
types
of
workloads
in
production
for
about
four
years,
which
I
think
makes
me
think
I'm
kind
of
an
early,
adopter
and
I've
hurt
the
sentiments
of
everything
worked
before
kubernetes
quite
a
bit
over
the
last
four
years.
A
A
Actually
many
things
you
start
running
into
limitations
on.
How
do
you
even
run
the
stack
on
your
laptop
anymore,
like
how
do
I
even
serve
all?
Luckily
other
limitations
might
be
with
your
cloud
provider.
You
might
have
trouble
loading
your
cloud
providers
web
console
because
you
just
have
too
many
things.
A
Here's
another
example
from
maybe
machine
learning,
so
maybe
you
start
doing
machine
learning
experiments.
You
have
one
researcher
running
training
on
their
laptop
and
everything's
fine,
and
then
you
start
training,
bigger
and
bigger
models.
So
now
you're
like
okay,
that
doesn't
fit
on
my
laptop
anymore.
So
let
me
train
on
the
beefy
box
somewhere
else,
maybe
in
a
Colo,
maybe
on
a
cloud
and
then
hopefully
things
are
going
well,
so
you
again
start
hiring
more
and
more
people
onto
your
team.
A
So
now
you
have
a
fleet
of
these
beefy
boxes
and
then
you're
like
what
even
is
running
anymore
I,
don't
know
how
many
people
have
worked
with
machine
learning.
Scientists
who
have
like
literally
lost
their
experiments
somewhere
because
they
forgot
where
they
were
running
it.
A
So
so
I
think
in
this
model,
infrastructure
teams
are
often
the
bottleneck,
because
you
have
to
figure
out
how
to
manage
and
version
control,
this
fleet
of
beefy
boxes,
and
you
have
to
manage
how
to
upgrade
them.
How
to
destroy
the
audits
and
yet
also
have
to
make
sure
that
you're
actually
using
using
your
resources
effectively,
and
maybe
it's
because
cost
savings
or
whatnot
and
the
team
that
manages
this
environment
might
be
your
same
infrastructure
team
or
maybe
a
different
team
entirely.
A
So
meanwhile,
hopefully
your
business
is
booming,
so
you
have
all
these
new
requirements.
Your
team
keeps
growing.
You
have
all
these
data
scientists
and
machine
learning
researchers,
and
why
and
meanwhile
you
have
to
move
fast
and
also
do
all
these
things
like
multi-region
sprinkle,
some
cheap
heaters
into
your
fleet,
maybe
to
multi
cloud.
A
A
A
So
we
are
where
we
are,
because
we
turn
to
open
source
and
started
building
agreements
on
what
abstractions
we
want
it
kind
of
like
how
we
started
converging
on
Linux
as
the
base
OS.
We
agreed
on
containers
for
empowering
developers
and
moving
fast,
and
we
agreed
on
communities
for
orchestrating
these
containers
at
scale
by
leveraging
this
huge
community
effort,
infrastructure
teams
can
provide
flexible
building
blocks
for
their
end
users
and
and
it
clearly
that
resonated
with
a
lot
of
people,
because
I
mean
we're
all
here
and
yeah.
A
A
Well,
I
think
people
make
fun
of
how
much
yam
well
they're
right,
I.
Think
partially
it's
a
testament
to
how
straightforward
and
extensible
the
kubernetes
a
API
is
and,
more
importantly,
how
user
centric
it
is
for
non
infrastructure
engineers.
They
can
do
a
lot
by
just
manipulating
these
yellow
files.
A
A
That's
not
all
awesome,
yet
it's
not
rainbows
and
unicorns
yet
so
we
still
have
some
ways
to
go.
We
took
on
this
super
ambitious
task
of
empowering
our
businesses
and
developers
to
meet
all
their
needs
and
kubernetes
gives
us
a
path
to
achieving
all
that
and
scale
with
the
complexity.
But
the
complexity
didn't
really
fully
go
away.
We
are
still
supporting
all
these
things
for
the
infrastructure
team.
A
A
So
one
path
for
it
that
the
community
is
investing
heavily
in
is
open
policy
agents.
This
is
an
example
of
how
a
simple
snippet
like
this
can
be
used
to
restrict
what
images
can
be
run
by
the
containers
on
your
cluster,
and
there
have
been
a
number
of
talks
at
this
conference
digging
deep
into
OPA.
So
we
can
show
the
interest
from
the
community
in
this
direction.
A
Another
question
I
hear
often
is
simply
how
do
I
develop
like
now
that
we've
moved
into
this
world
of
immutable
containers
running
on
the
cloud?
How
am
I
supposed
to
develop
my
application?
How
do
I
get
to
a
workflow
where
I
can
use
my
debugger
or
iterate
as
fast
as
I
used
to
just
running
Python?
On
my
laptop
and.
A
There
are
other
people
who
have
asked
this
very
same
question
and
there
are
a
lot
of
answers
out
there.
There
isn't
a
clear
one
winner,
but
these
are
just
some
examples
of
open
source
projects
that
are
tackling
this
very
problem.
There's
so
many
awesome
options.
They
all
take
a
slightly
approach
or
slightly
slightly
different
approach
in
solving
this
problem.
So
your
pick
might
be
different
depending
on
the
workflow
that
your
organization
is
used
to
I
want
to
give
a
shout
out
to
the
applications
development
track
where
some
of
these
projects
are
getting
covered.
A
Related
to
being
in
a
mutable
world,
this
is
by
far
the
most
common
question.
I
get,
which
is
how
do
I
D
bug
do
I
have
to
just
keep
adding
print
statements
and
rebuilding
my
images
and
deploying
them
and
then
looking
at
the
print
statements
and
then
just
keep
going
until
I,
like
figure
out
what's
wrong,
which
is
why
I
think
a
lot
of
people
do
it
doesn't
have
to
be
that
way.
It
doesn't
have
to
be
that
painful
and
in
particular,
I'm
very
excited
about
this
new
feature
in
116,
which
I
covered
yesterday.
A
A
So,
for
example,
you
could
have
my
running
pod,
those
or
maybe
crashing-
or
maybe
it's
already
running,
but
it's
not
doing
what
you
want
instead
of
adding
print
statements
or
redeploying
the
pod.
Until
you
get
it
right,
you
could
just
attach
another
container
with
all
the
debugging
tools
you
need
to
this
running
pod
and
just
use
other
tools
they're,
not
in
your
images,
so
I
think
one
useful
example
for
this
is
a
lot
of
times.
People
have
maybe
misconfigured
their
credentials
or
maybe
misconfigured
their
network.
A
And
overall,
if
I,
look
at
the
theme
of
the
common
challenges
that
we
just
went
over,
I
think
that
we're
we've
devolved
to
a
place
where
we're
iterating
on
user
experience.
We
started
with
this
huge
problem:
space
kubernetes
provided
the
foundation
for
orchestrating
containers
and
the
building
blocks
to
power
our
workloads.
A
Now
that
the
building
blocks
are
in
place,
we
have
focused
on
building
a
higher
level
or
we
can
focus
on
working
on
higher
level
of
problems,
to
simplify
and
improve
the
user.
Experience
and
I
think
that
again,
everything
works
before
communities,
but
it
was
unclear
that
it
was
going
to
keep
working
as
well
as
they
they
did.
Given
all
the
new
requirements,
they're
asked
of
us
and
I
think
that
you
know
thanks
to
the
community
and
everyone
here,
we're
getting
there
with
kubernetes.
Thank
you.