►
Description
Join us for Kubernetes Forums 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
Filling the Gaps in Kubernetes Test Coverage - Hippie Hacker, ii.coop
Are the Kubernetes behaviors your applications actually require well tested and guaranteed to be available on all cloud providers?
In this session, you will learn how to ensure your Kubernetes API surface area usage is exercised by tests all Kubernetes Certified Service Providers must pass.
We will cover:
- the e2e test suite
- automation that runs the suite before code is merged into Kubernetes.
- the API surface area covered by these tests
- the API surface area required by several popular applications.
- Identifying the untested API surface area your applications require
- Contributing tests that increase API surface coverage
- Promoting tests to Conformance
A
You
are
here,
thank
you
for
being
exactly
who
you
are
and
for
choosing
to
do
so
with
me
today.
Each
of
you
has
something
special
to
contribute
and
share
with
the
world
and
I
want
to
live
in
a
world
where
you
are
able
to
do
just
that
I'm.
So
glad
you're
here,
in
addition
to
you
being
here,
I
am
here:
I'm,
hippie,
hacker
and
I,
invite
you
to
bring
your
best
ideas
and
hack
the
planet
with
me,
because
we
won't
see
what
is
humanly
possible
unless
we
invite
everyone.
So
don't
forget
to
invite
your
neighbors.
A
So
that's
us
now:
let's
talk
about
I
I.
There
is
a
special
atmosphere
when
two
or
more
of
us
put
our
minds
together
and
I
think
there's
something
quite
divine
in
that
you're
and
I
I'm
and
I,
and
we're
putting
our
minds
together.
That
means
I
I
is
here,
there's
a
viral
generosity
that
happens
when
by
sharing
something
with
others
that
is
meaningful
to
them.
They
feel
empowered
and
compelled
to
share
it
with
others,
just
as
I
feel
compelled
to
share
with
you
what
others
have
freely
shared
with
me.
I
think!
That's!
A
Why
we're
here,
but
to
bring
this
a
little
more
down
to
earth.
Software
is
basically
executable
ideas
and
we
are
reaching
a
point
globally,
where
very
complex
ideas
can
be
orchestrated
across
all
our
neighborhoods
and
Nations.
It's
important
that
we
have
a
shared
understanding
of
what's
required,
to
execute
and
orchestrate
these
incredible
ideas
it's
going
to
have
to
go
beyond
those
of
us
in
the
room
of
tech
to
provide
meaningful
expressions
of
innovation.
Our
communities
need
always
buy
local.
If
you
can,
it
helps
create
sustainable
ecosystems
of
innovation.
A
A
Local
infrastructure,
developed
and
run
by
locals
everywhere
allows
participation
from
all
people
and
all
nations.
What
will
your
community
be
known
for
free
software?
Five
bucks
I
used
to
sell
CDs
at
first
Saturday
hacker
farmers
market
in
Dallas,
Texas
I
did
quite
well
some
weekends
I'd
make
bucks
just
walking
around
and
giving
people
what
I
had
downloaded
off
of
the
internet
on
a
CD
that
I
could
physically
hand
them
in
exchange
for
money.
A
We
don't
use
CDs
much
anymore,
but
the
international
standards
organization
had
a
ISO
nine
six,
six
zero
four
CD
ROMs.
That
specification
is
still
used
by
nearly
every
operating
system
that
is
distributed
on
the
internet
from
imagination
to
reality.
Our
executable
ideas
and
production
workloads
are
going
to
require
a
new
standard
for
distribution
and
orchestration.
A
A
Innovation
occurs
best
when,
with
lots
of
freedom
to
recombine
and
improve
on
existing
ideas,
we'll
need
to
have
shared
customs
and
protocols,
in
addition
to
the
local
ones,
for
connecting
those
new
ideas
with
the
infrastructure
that
can
run
them.
However,
this
is
gonna
require
some
global
cooperation
and
some
shared
agreements.
A
A
A
A
It's
used
to
ensure
the
API
is
for
infrastructure
globally,
that
the
it's
used
to
ensure
the
API
is
for
infrastructure
globally,
provide
the
behaviors
we're
coming
to
depend
on
this
sunburst
graph
from
API
snoop
represents
the
surface
area
of
kubernetes.
That
middle
ring
with
the
alpha
beta
and
stable
API,
is
represented
by
the
red,
blue
and
green.
A
If
we
zoom
in
and
look
at
the
green,
those
generally
available,
api's
should
provide
a
very
stable
surface
area
for
us
to
deploy
our
ideas
on.
They
will
take
years
for
those
api's
to
be
deprecated
and
we're
currently
only
testing
about
30%
of
the
surface
area
for
all
of
our
CN
CF
certified
kubernetes
providers.
A
A
A
We
go
forward
to
1972.
One
of
the
most
famous
selfies
of
all
of
us
was
taken
from
the
crew
of
the
Apollo
17
module
in
1990,
the
probe
Voyager
1
traveled
230
degrees
above
the
ecliptic.
Looking
back
at
our
solar
system
and
sent
a
picture
back
to
earth
that
is
referred
to
as
pale
blue
dot,
you
probably
can't
see
it,
but
in
that
third
red
band
about
halfway
down
you'll
see
a
picture,
a
pixel,
that's
slightly
blue!
That's
that's
us!
A
By
2001
we
had
a
lot
more
satellites
and
we
began
to
get
a
full
picture
of
the
earth
over
time
and,
finally,
a
map
with
New
Zealand
on
it,
but
often
gets
cut
off
by
2002.
We
had
enough
satellites
to
get
a
clear
picture
of
the
entire
earth
in
one
day
over
time.
This
allowed
us
to
see
the
entire
earth
without
clouds
in
New
Zealand.
There's
a
maori
proverb.
There
reads
what
is
the
most
important
thing
and
the
response
is
almost
automatic.
It's
the
people,
it's
the
people,
it's
the
people.
A
So
all
of
this
understanding
the
surface
of
the
earth
and
the
coverage
and
is
mapping
somewhat
to
our
journey
with
kubernetes
ken
amici,
wrote
the
first
software
that
parsed
our
logs
just
the
logs
of
Emir
and
tests,
and
this
exposed
our
initial
11%
API
test
coverage.
Thank
you,
Ken
for
contributing
that
and
being
part
of
our
community
forward
to
1:11
and
Rowan
Fletcher
wrote
the
initial
API
snoop.
It
was
a
man-in-the-middle
proxy
with
a
sidecar
and
our
initial
audit
logger
parser.
A
He
also
implemented
the
sunburst
graph
and
the
logo
that
API
Snoop
is
known
for
thanks
heaps
road
in
1.12
song,
zoo,
added
support
for
the
API
server
for
recording
which
software
was
making
the
API
request
the
kubernetes.
This
allowed
us
to
update
the
ete
framework
to
include
the
name
of
the
tests
with
each
request
before
1.12.
We
had
no
idea
if
endpoints
were
being
tested
and
whether
they
were
conforming
or
not.
A
We
got
the
114
that
we
had
a
real
critical
thing
happen.
Prior
to
this
features
were
getting
promoted
from
beta
to
generally
available
to
the
public
that
didn't
have
any
tests,
and
in
order
for
us
to
get
coverage,
we
needed
to
stop
that
bleeding
Aaron,
Creek
and
Berger,
while
serving
on
the
114
release
team
as
a
lead.
Had
the
foresight
to
implement
a
policy
that
GA
features
must
have
conformance
tests.
I
agree,
that's
a
big
deal.
Thank
you.
A
So
much
Aaron
for
helping
us
keep
our
standards
high
in
115,
Zack
Mandeville
performed
a
major
rewrite
of
our
manual
process
of
looking
at
Python
and
JSON
into
a
beautiful
Postgres
and
graph
QL
API.
This
allowed
us
to
start
asking
questions
about
our
coverage
and
the
and
answering
them
without
needing
to
write
more
tooling.
We
could
just
update
queries
and
get
a
lot
of
dynamic
information
to
help
us
comprehend
the
full
API
surface
area
and
prioritize
what
we
need
to
be
testing
next
Thank
You
Zach.
A
116
Jordan
Liggett
improved
our
invisibility
by
ensuring
we
see
every
generated
event
from
any
application.
There
were
some
some
bugs
and
he
found
those
and
I
really
appreciate
that
Jordan
1:17
we
just
had
a
release
and
we
still
have
gaps.
This
is
where
you
come
in.
We
need
to
not
only
measure
the
surface
area
of
kubernetes,
but
we
need
you
to
be
involved
in
prioritizing
what
it
is
that
you
do
with
kubernetes
and
making
sure
that
we
have
conformance
tests
for
that.
I
have
a
bit
of
a
demo
that
we'll
walk
through.
A
B
A
A
In
order
to
access
that
information,
you
need
to
get
into
your
cluster
to
connect
to
the
Postgres
and
the
hysteria
boards.
So
here's
some
commands
that
you
would
need
the
type
to
forward
that
I've
taken
the
time
of
doing
that
in
advance
that
we
don't
have
to
wait,
and
we
also
have
port
forwarding
for
Hazara.
A
When
you
load
API
snoop
inside
your
clusters,
it
will
go
and
retrieve
the
latest.
Ci
builds
from
test
grid
that
raw
audit
event
table
at
the
bottom
that
that's
around
four
hundred
Meg's.
That
includes
all
of
the
events
that
were
part
of
the
conformance
run.
The
swagger
includes
all
of
the
definitions
of
the
API
and
we
combine
those
into
generating
material
for
us
to
query.
A
If
we
look
at
our
current
coverage
from
our
with
an
IP
I
snoop
we're
able
to
see
that
our
percent
of
conformance
coverage
is
still
around
30%
as
of
less
than
two
hours
ago,
we'd
love
for
that
to
be
higher.
Obviously,
you
can
run
the
query
against
API
to
see
what
the
latest
route,
which
trow
result
or
cut.
You
were
retrieving
as
well
as
a
link
to
that
ik
job.
A
Our
workflow
within
the
conformance
test,
writing
team
is
to
try
to
identify
what
api's
need
to
be
have
tests
written
for
them.
Next
part
of
our
flow
is
to
use
queries
into
API
snoop
to
define
those,
but
it
would
also
be
great
if
people
would
run
within
their
cluster
their
own
applications
to
help
prioritize.
What
API
is,
they
think
are
important.
A
This
is
the
result
of
one
of
those
queries.
Here's
some
operation,
IDs
that
have
zero
conformance
test
coverage
once
you've
found
an
API
that
you
care
about.
Finding
the
documentation
would
probably
be
the
next
step
to
create
a
test
for
that
or
look
at
how
your
applications
use
that.
So
we
have
a
good
story.
A
The
next
step
is
to
actually
write
a
test
and
writing
your
test.
We
have
examples
as
far
as
how
do
we
test
particular
portions
in
a
different
language,
the
EDD
test
framework
is
written
in
ingo
and
uses
ginko,
but
as
far
as
exploring
what
the
api
is,
the
our
team
uses
several
different
languages.
Here's
one
ingo
that
we
run
here's
another
one
where
we
have
tests
and
python
and
we
also
have
I
think
tests
in
JavaScript
as
well
after
we
ran
those
tests.
A
See
that
the
components
are
healthy,
we'll
come
back
in
a
moment
when
we
update
and
look
at
the
live
events
inside
of
your
cluster.
At
that
time
we
can
see
that
we
have
a
table
called
endpoints
hit
by
new
test,
pretty
simple,
and
if
we
look
at
all
of
those
endpoints
and
and
specifically
look
at
the
user
agents,
we
can
see
that
the
tests
we
written,
which
we
set
the
user
agent
to
live
component
test,
hits
the
two
endpoints
that
we're
interested
in
all.
A
So
we
also
have
a
metric
for
what
would
be
the
increase.
So
if
we
look
at
our
test
coverage
over
here
to
the
right,
we
can
see
that
we
increased
our
coverage
by
two
points.
There
there
are
around
1,400
endpoints
in
kubernetes
in
total.
As
far
as
the
number
of
operations
you
can
do,
approximately
400
of
them
are
in
unstable
and
those
are
the
ones.
Those
are
the
only
ones
that
are
part
of
the
conformance
efforts.
A
Now
that
you've
kind
of
seen
the
process
of
deploying
API
snoop
into
the
cluster
running
some
tests
or
applications.
The
next
question
is:
how
can
you
help?
Well,
it's
running
it
in
your
own
clusters,
identifying
gaps
in
coverage
that
you
depend
on
and
work
with,
the
API,
snoop
and
Kate's
conformance
teams
to
fill
the
gaps
in
coverage.
A
I'd
love
to
help
help
you
get
these
things
up
and
running
or
help
in
writing
tests
or
have
our
team
help
to
write
the
tests
because
there's
lots
of
empty
space
out
there
in
the
endpoints
in
our
coverage,
and
this
is
what
we
need
to
increase
our
coverage
on.
So
that's
all.
As
far
as
the
demo
and
the
presentation
I'll
open
up
the
floor
for
some
Q&A
and
I
think
there's
a
microphone
up
here.
If
you
wouldn't
mind
coming
up
and
retrieving
the
mic.