►
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
Intro: Chaos Engineering BoF - Sylvain Hellegouarch, ChaosIQ
In this session, we will cover the work performed by the community since we made the proposal for a CNCF Chaos Engineering WG at KubeCon Copenhagen. We will review the WHITEPAPER and future direction of the group.
To learn more: https://sched.co/Grcr
A
Thank
you
for
making
it
upstairs.
I
know
it's
like
the
far
end
of
the
building.
I
guess.
So
thank
you
for
making
the
effort
we
hate
today
to
talk
about
curse
engineering.
So
this
is
the
first
talk
of
two
talks.
Actually
just
the
first
talk
of
two
sessions
and
tomorrow's
talk
is
going
to
be
a
to
most
demo.
So
two
more
today
is
not
demo.
Sorry
don't
run
tomorrow
will
be
demo,
so
that
is
a
you
know.
Waiting
your
appetite.
We're
going
to
talk
about
curtain
hanging,
try
to
sort
out.
A
You
know
what
it
means
where
we're
going
with
this,
and
you
know
what
what
exactly
would
you
could
actually
learn
from
it?
Quick
round
who's
done
some
form
of
calendaring
who's,
familiar
with
principles
and
practices.
What
quite
a
few
people
that's
good,
so
my
job
is
done.
Thank
you
alright.
So
so
my
name
is
Savannah
gouache,
just
called
me
Silvana.
If
you
can
that's
probably
easier
and
the
chaos.
Thank
you
CTO.
It's
a
company
around
calendaring
and
we're
more
known
for
the
kills
toolkit
and
kills
platform
products.
Alright,
that's
enough
of
me
what's
interesting.
A
Is
this
talk?
So
it's
going
to
be
a
talking
three
acts.
The
first
act
would
be
setting
a
scene
and
then
another
one
would
be
the
characters,
the
community
itself
and
the
last
one.
The
plan
so
I've
got
alpha.
Now
attraction
go
through
those
three
acts
and
I'll
try
to
stick
to
that
with
enough
time
for
questions,
but
otherwise
we'll
ask
questions
outside
so
that
if
there
is
another
talk
after
that,
they
are
free
to
go
for
it
right.
A
A
Nonetheless,
we
need
to
set
the
scene
like
I
said,
so
we're
going
to
illustrate
that
with
what
we're
talking
about
with
case
study-
and
that
case
you
know
that
case
study
we're
going
to
take
is
near
loss
and
recovery
of
America's.
First
space
station
I
didn't
come
up
with
title,
but
I'm
really
proud
of
using
space
station.
In
my
talk,
you
know
I've
sort
of
made
it
it's
it's
a
case
study
made
by
night
Nessa,
because
apparently
they
do
those
sort
of
things.
You
know
space
station
and
it's
a
fantastic
resource.
A
You
should
go
on
those
links
on
Christy
case
studies.
They
are
fairly
shot
to
read.
It
doesn't
take
a
long
time
and
they
go
well
beyond
you
know
just
space.
They
come
back
on
earth
and
actually
do
case
studies
for
videos
as
well,
and
it's
a
fantastic
read
because
they
tell
you
what
they
learned
from
new
things
and
that's
really
extraordinary
right.
So
what
what's
the
context
here?
What
are
we
talking
about?
Well,
a
pond
in
the
early
70s,
the
u.s.
tried
to
do
a
first,
their
first
space
station
for
science.
A
You
know
and
they
defined
to
launch
that
after
years
of
design
and
following
fully
you
know
the
the
Apollo
program.
So
what
I'm
saying
seeing
that
as
a
context
here
what's
important
here
is
the
years
of
design
and
the
fact
it
was
not
the
first
time
they
did
it,
not
the
space
station.
It
was
the
first
one,
but
at
least
you
know
launching
stuff
into
space
because,
as
we're
going
to
see,
it
actually
played
little
against
them.
So
what
happened
here?
Well?
They
suffered
the
loss
of
a
shield,
a
Sun
radiation
shield
during
the
launch.
A
That's
pretty
bad,
because
if
you
put
a
shield
is
you
expect
things
to
go
wrong
if
you
don't
have
and
heat
to
go
very
high,
very
quickly,
so
losing
that
wasn't
a
good
good
thing.
Luckily,
that
was
the
first
launch,
wasn't
meant
otherwise
it
would
have
been
really
bad
for
them.
So,
like
I,
said
temperature
when
I
the
first
thing
they
did
was
recovery
as
engineers.
A
Usually
that's
what
you
do
if
you
on
call
we're
going
to
try
to
recover
the
system
as
best
as
you
can
first
right
before
doing
any
learning,
any
anything
you're
going
to
first
recover
as
best
as
you
can
delete
them,
the
first
they
change
the
angle
of
Space
Station's,
like
Lee,
said
the
Sun.
You
know,
Sun
radiation
would
be
here
strong
sand.
A
The
temperature
would
go
down
a
little
now
that
in
itself
was
obviously
quite
tricky
same
time
that
to
keep
it
in
a
certain
angle,
so
that
it
could
actually
still
control
it
and
bear
in
mind
it
was
like
40
years
ago,
45
years
ago.
No,
no,
listen,
computer,
a
normal
by
things
like
that.
It
was
it's
quite
an
achievement.
What
did
you
and
they
decided
that
they
still
needed
to
control
the
temperature?
So
that's
what
we
think
very
quickly
out
of
the
box,
how
to
do
that,
and
that's
what
we
see.
A
A
So
what
was
the
findings?
That's
going
to
be
a
long
paragraph
here,
I'm
going
to
try
to
go
through
them
quickly
there,
except
from
best
case
studies
itself,
so
go
back
and
with
them.
But
what's
interesting
here,
like
I
said
earlier:
they
they
were.
You
know
the
base
of
the
Apollo
program.
What
they
realize
is
that
thing.
The
shield
itself
was
not
part
of
the
Apollo
program.
They
just
added
it,
and
but
some
some
things
like
the
control
management
were
essentially
the
same
as
your
program.
A
So
I
wouldn't
say
they
were
careless
or
anything
like
that,
but
they
met
assumption
about
the
fact
that
it
worked
well
and
it
should
work
well
again
and
they
didn't
realize
that
adding
a
new
thing
was
actually
a
change
quite
complex
in
itself,
not
just
a
new
thing
that
they
did
well.
If
we
talk
about
our
industry
doesn't
ring
a
bell
like
you
know,
I
did
that
in
the
past
and
it's
just
a
small
chance.
You
know,
let's
push
it
in
production,
we've
all
been
there.
We've
done
that
well
quit.
A
A
So
they
were
smart
and
sharp
engineers
and
you
know,
probably
the
best,
but
it's
that
they
they
were
confused
by
the
confidence
in
in
in
their
design,
and
that
provoked
the
error
and
the
last
binding
that
I
find
interesting
to
earth
is
their
design.
There
are
lots
of
regal
lots
of
formalism.
Lots
of
of
you
know
of
documentation.
Now,
don't
get
me
wrong.
Do
that
do
a
lot
of
peer
reviews,
documentation,
testing,
whatever
you
know,
all
those
things
are
perfect,
but
they
realized
them.
It's
not
enough.
We
still
need
to
have
our
intuition
as
engineers.
A
We
can't
lose
that
side
of
that.
We
need
to
realize
that
as
engineers
we
have
a
hunch,
maybe
a
bit
too
strong
of
a
world,
but
at
least
experience
that
tell
us
well,
maybe
we
should
pay
more
attention
to
do
this
or
that
and
that
it's
really
interesting
because
I
realized,
even
though
we
spends
you
know,
we
send
Space
Station.
We
still
need
those
that
sort
of
intuition
and
whether
realize
they
really
need
to
have
trust.
You
know
to
basically
fertilize
the
old
scientists.
A
So
it's
a
different
mission
but
same
take
in
my
opinion
here
is
on
the
mass
and
on
the
rover
mission
sent
to
Mars.
They
realized
that
you
have
the
cards.
You
want
right
what
you
try
to
have
as
constraints
what
you
design,
where
you
think,
but
then
you
arrive
in
on
mass
and
then
you
know
orbits
are
basically
things
are
different.
Now,
I
understand
that
those
are
extremely
far-fetched,
maybe
to
you
examples,
but
there
relate
to
or
delete
you
know
to
a
domain
as
well.
It's
always
the
same.
A
So
basically,
what
they
told
us
is
be
ready
not
to
be
ready
in
the
sense.
Where
keep
keep
in
mind
that
things
will
go,
you
know,
go
somewhere
and
you
don't
want
them
to
we.
We
are
not
mathematicians
here,
we're
not
demonstrating
through
formalism.
Only
you
know
the
laws
of
nature
nature
actually
impact
us
and
I
mean
that
really
when
you're,
you
know
when
your
hard
drive
goes
down
and
you
lose
a
volume
of
something
you
know.
That's
that
sort
of
thing
that
you
don't
really
expect.
A
You
know,
but
the
prime
here
with
what
we're
going
to
discuss
today.
What
is
the
impact
of
that?
It's,
not
necessarily
that
we
know
that
something
is
going
to
happen,
but
what
is
what
are
the
consequences
of
losing
this
or
that
this
is
what
kills
entering
is
going
to
bring
you
first
of
all,
what
to
2018
we've
improved.
We
are
better
at
it
right.
The
inside
mission
is
successful
and
it's
fantastic
in
my
opinion,
but
a
system
reliability,
especially
in
our
industry,
goes
we.
We
still
have
to
a
lot
to
improve
I'm.
A
So
sorry,
it's
in
French,
but
when
I
took
the
screenshots
I
didn't
realize.
I
was
actually
using
that
later
on,
but
I
gave
you
you're
familiar
with
a
screen
of
certificate.
That
has
expired.
This
one
comes
from
actually
a
French
public
service.
Now
I
know
we
don't
work
a
lot
about
the
and
things
like
that
in
France,
but
we
do
tend
to
be
careful
about
those
things,
but
this
one
is
minor
service
someone.
You
know
in
the
chain
that
you
know.
Nobody
really.
You
know
cared
for
I,
guess
and
that
certificate
expired
after
nine
days.
A
I
just
had
to
use
that
service
and
I
couldn't
I,
wasn't
the
only
one.
So
I
did
what
everyone
does
goes
to
Twitter
and
complain
and
again.
If,
if
Twitter
is
your
first
alerting
system,
you
don't
have
a
lot
in
system
so
try
to
be
before
you
know:
Twitter
that'd
be
nice,
but
I
didn't
complain,
I,
guess!
No.
A
You
can't
read
that
anyway,
and
luckily
I
wasn't
the
only
one
and
some
people
who
actually
said
the
same
thing
and
I
have
no
connection
that
I
do
and
they
fix
the
thing
and-
and
it
was
all
good
but
sometimes
impacts
are
a
stronger.
If
you
were
in
the
UK
last
week
or
if
you
are
a
British
citizen,
you
may
use
the
o2
Network
mobile
network.
A
Well,
a
similar
thing
happened
on
the
infrastructure
that
Ericsson
provides
apparently
the
first
analysis
Holden
there,
but
said
that
italia,
TLS
certificate
did
expire
as
well,
but
the
ripple
effect
for
that
was
that
the
old
mobile
network
went
down
for
the
better
of
two
days,
but
there
was
no
network
at
all
and
it's
actually
quite
quite
stressful,
because
my
co-founder
and
my
friend
is
on
that
network.
I
tried
to
reach
him.
He
was
on
the
road
and
I.
Couldn't
so
I
got
worried
now.
A
Luckily,
it's
not
you
know
it,
don't
it's
not
much
for
us,
but
limiting.
If
you
need
to
talk
to
your
mother
or
she
needs
to
talk
to
you,
but
you
know
to
mess
you
you
know
so
those
sort
of
urgency
right.
So
even
a
small
thing
that
you
forgot
about
will
actually
impact
users
and
we're
in
a
real
life
and
that's
what
you
something
you
want
to
actually
experiment
on
before
they
actually
hit
the
problem,
a
quick
tip
issue
of
Tyria
certificate
certificate
and
no
matter
in
monitoring
at
all.
A
Just
use,
google
and
gender
or
something
you
know,
sets
set
an
alarm
somewhere
saying,
oh
in
one
months
by
where
this
really
expired,
because
what
happened
in
the
previous
example
was
the
buddha
to
get
three
other.
Such
you
get
three
or
five
years
ago,
and
they
probably
just
problem-
you
know
forgot
about
it
or
the
person
just
left
so
just
leave
somewhere
somehow
its
body
alright.
So
we
need
more
reliable
systems
and
really
truly
I
mean
that
we
do
need
more
reliable
systems.
A
I
think
we
need
more
reliable,
more
reliable
system
than
we
need
more
pictures
these
days,
that's
more
critical
for
our
users
that
our
systems
are
reliable
because
they
do
trust
us
to
provide
a
feature,
a
service,
it's
more
important
to
me
that
your
systems
are
reliable
than
there
are
shiny
new
features
in
them
so
that
the
end
of
Act
one
we've
stood
this
machine.
We
need
reliability
and
by
reliability,
I
mean
any
form
of
variability.
It
can
be
resiliency
can
be.
Security
can
be
just
pick
one
it's
an
umbrella
term.
A
A
A
A
So
I
think
it's
a
strong
signal
that
the
scene,
if
the
since
you've
you
know,
cares
for
that
enough
to
actually
create
a
working
group
that
it
matters
to
the
cloud
native
system
at
large.
It's
not
a
little
thing.
What
are
the
deliverable
and
challenge
that
we
we
wanted
to
you
know
yet
we
had
two
things.
So
what's
paper
and
landscape
I'm
going
to
start
with
a
white
weight
paper
so,
first
and
foremost,
white
paper,
the
sense
EF
is
very
strong
on
that
it's
not
an
RFC,
it's
not
a
specification.
A
It
doesn't
tell
you
what
to
do.
It
gives
you
a
good
starting
point.
It's
really
a
starting
point
for
your
journey
into
chaos
engine.
So
it
will
give
you
sorry
ground
about
at
least
a
definition,
a
set
of
tools.
You
know
the
use
case.
Is
you
know
so
that
you,
you
can
figure
out
what
what
it
means.
I
think
it's
a
good!
It's
a
good,
it's
good
starting
point,
so
why
Kevin
joining
fine
Paul
from
the
white
pepper?
We've
got
a
few
use
case
that
way,
I
think
not.
A
That
would
be
yes,
that
will
be
the
one
after
well.
We
want
to
unless
improve
system,
reliability,
that's
a
first
and
for
and-
and
you
know,
first
one
chaos
enjoying
is
to
actually,
unless
your
system
reliability,
it
benefits
the
cloud
native
system,
because
cloud
native
has
all
the
properties
that
will
help.
That
will
be
quite
easy
for
case
and
drying.
You
know
during
the
your
your
software
development
cycles,
you
see
that
and
it's
basically
bringing
the
operational
practices
into
the
development
world.
A
Alright,
so
some
use
cases
like,
obviously
to
trying
your
network
CPU
so
far,
while
trying
to
increase
your
CPU
suddenly
how
your
system
copes,
we
know
that
communities
will
manage.
You
know,
resources,
request,
limits
and
stuff
like
that.
But
have
you
tried?
Have
you
pushed
your
service
to
the
limits
and
see
how?
Basically,
they
are?
You
know
changed
from
node
to
another
node.
What
are
the
consequences
of
that
that
change?
A
If
your
system
takes
like
one
minute
to
start
and
suddenly
it's
been
moved
from
machine
to
machine
to
machine,
but
not
quickly
enough,
then
maybe
you
start
having
a
ripple
effect.
That
is
not
very
good
for
your
system
overall,
so
it's
really
a
way
to
push
your
system
into
places
that
it
doesn't
like
to
be,
but
you
do
that
in
a
controlled
fashion,
you're
observing
the
system.
At
that
time,
you
controlling
the
experiment
will
do
that
multi-cloud,
McRae
migration
or
cloud
migration
altogether.
A
It's
interesting
because
let's
say
you
want
to
move
to
a
different
cloud
or
of
ivory
ivory
cloud,
you're,
not
sure
where
to
start
right,
but
the
worst
part
is.
You
know
exactly
the
challenges.
You've
got
on
your
system,
but
if
you
move
to
different
clouds,
yes,
you
know
probably
or
provide
the
same
features
at
higher
level,
but
you
don't
know
the
consequences
from
the
run
time
perspective.
Maybe
it's
Laura.
What's
impact
and
honestly
calendaring
can
be
used
from
the
process
and
you
mine
perspective.
It's
a
really
socio
technical
and
oral
practice.
A
A
Maybe
you
can't
even
actually
reach
the
machine
because
she's
the
only
one
who
knows
the
password
and
for
some
reason
you
week
he
hasn't
been
updated
for
six
months.
That
would
be
good.
So
that's
the
sort
of
thing
that
kills
entering
will
help
you
very
technical.
Very
you
know
don't
worth
you
know.
Let's
try
to
prevent.
You
know,
remove
all
you
know
the
disc
or
prevent
like
like
simulate
a
failure
and
volume
and
you've
got
a
stateful
application.
What
happens?
How
does
it
impact
my
data
right?
The
second
deliverable
and
that's
we've
got
those
two.
A
Only
the
landscape,
your
fail.
You
know
aware
of
the
massive
landscape
of
the
CN
CF.
Well,
we've
got
a
tiny
place
for
it
right
now,
it's
fairly,
not
populated
enough.
In
my
opinion,
we've
got
those
projects
all
delivered
know
either
forty
injections
orchestration
around.
You
know
chaos
injuring,
but
there
are
so
many
more.
The
prime
is
right
now
the
tooling
that
you'd
find
mostly
come
from
some
companies
who
have
done
some
form
of
reliability
testing,
but
they
are
very
specific
to
to
a
specific
need
for
that
company.
So
we've
got
segmented,
fragmented
net.
A
You
know
a
tooling
space
here,
so
it's
going
to
grow,
but
it
will
be
at
some
time
so
on
communities,
we've
got
quite
a
few
things,
but
on
some
other
areas
we
don't
have.
We
don't
have
much
you'll
have
to
be
patient
or
come
with
your
own
tool.
That's
right
so
and
difficulty
as
well
as
as
a
tool
provider.
I
may
want
to
actually
organize
my
myspace
in
this
way.
So
in
the
landscape,
I'd
like
to
be
organized
in
that
in
that
place,
but
that
we
need
to
think
in
terms
of
users
as
a
user.
A
How
do
you
look
for
those
things?
You
may
look
for
chaos,
engineering,
but
you
still,
you
know
better.
You
don't
know
where
to
start
right.
You've
got
many
tools
and
you're
back
to
square
one,
which
one
should
I
pick.
That's
one
of
the
prime
of
the
landscape.
It
has
so
many
solutions
for
some
specific
needs,
some
time
that
you
don't
know
where
to
start,
and
it's
overwhelming.
So
we
need
to
decide
the
right
way.
As
a
user
are
two
guide
you
towards
the
right.
You
know
right,
Lee,
I,
think
it's
important.
A
So
the
challenge
is
it's
a
new
practice.
So
what
you
draw
the
line?
Either
we
go
too
far
in
meaning
we
start
almost
becoming
like
how-to
or
dogmatic
about
it,
and
we
don't
want
that
or
we
don't
go
far
enough
and
you
you're
probably
coming
to
it
and
saying
I,
don't
know
where
to
start.
It
doesn't
really
help
me
and
that's
tricky
to
be
honest
as
a
whitepaper,
it's
very
tricky
to
find
the
right
balance.
I'll
do
better
engage
the
community.
A
We've
got
a
good
community,
as
you
know,
in
terms
of
number
of
people,
maybe
you're
part
of
that
slack
channel,
but
it's
very
quiet,
I
think
people
right
now,
more
consumers-
and
you
know,
actors
on
that.
So
we
need
to
engage
and
what
we
could
do
is
already
share
your
resiliency
and
reliability
stories.
The
failures
it's
fantastic
to
share
what
this
fight
is
for
me,
because
we're
going
to
discuss
about
that
and
from
that
maybe
we
can
start
drawing
ideas
towards
what
kills
and
drawing
experiments.
We
could
actually
come
up
with.
A
So
if
you
are
really
really,
if
you
can,
obviously
sometimes
you
know
sensitive
data
and
stuff,
but
if
you
can
share
your
stories,
I
think
it's
a
very
good
starting
point
right.
So
that
would
be
the
milestone
that
our
community
is
facing
right
now,
complete
the
white
paper.
Please
come
and
join
and
and
NPR
and
find
a
white
way
to
actually
organize
such
an
the
landscape,
put
more
tools
in
there
so
that
you
know
where
to
start
and
hopefully
submit
the
widget,
the
the
working
group
to
the
cnc
of
GOC.
A
So
it
becomes
a
true
working
group.
So
we
need
to
make
a
stand
back
to
what
we
were
saying
earlier.
We
need
to
make
a
stand
about
reliability.
I
know.
Future
velocity
is
so
big,
it's
sometime
difficult
to
fight
back
against
the
product
product
product
teams,
but
they
need
to
understand
that
the
more
reliable
we
are
we'll
be
it's
it's.
A
The
only
way
to
get
there
if
we
are
reliable
and
we
trust
in
our
country-
and
we
are
confident
in
our
systems
and
ourselves-
then
we
can,
you
know
we
can
deliver
as
fast
as
we
can
and
that's
a
community
thing.
An
active
rest
act
is
opening
up.
You
know
having
an
outlook
on
on
thing.
I
said
it
won't
here,
because
we
are
in
this
together,
it's
not
just
in
CF
working
group
or
just
me,
or
you
know
a
few
here
and
there
we
all
have
a
way
to
actually
contribute
to
the
idea
of
reliability.
A
So
we
must
not
reduce
you,
know:
bag,
shading
and
stuff,
like
that,
we
can't
reduce
chaos.
You
know
into
definition
and
truly
I
would
be
foolish.
Who
thinks
that
chaos
and
running
kills?
Is
chaos
monkey
right?
Well,
no,
it's
because
a
little
beyond
that,
hopefully,
chaos.
Monkey
is
a
terrific
tool,
but
it
is
a
tool.
It
is
only
a
mean
it's
one
right,
so
we
need
to
go
beyond
those
things.
A
It's
a
deliberate
practice.
So,
first
of
all,
it's
deliberate
you
do
it
because
you
decide
you're
going
to
do
it.
It
must
not
be
imposed
you're
doing
it
as
an
experiment,
and
you
want
to
explore
what
you
don't
know.
Don't
you
know
you're
using
humanities?
Maybe
so
many
things
you
don't
know,
you
have
no
idea
how
things
are
going
to
react.
Change
your
policy
somewhere.
You
know
this
morning.
Liz
spoke
about
the
variability
control
policies
of
you
know,
trying
a
new
thing
on.
A
You
know,
trying
to
implement
control
policies
so
that
you
admission
policy
is
viable.
That's
good!
You
can
test
that
see
that
works,
but
you
really
don't
know
what
are
the
consequences
of
the
system
at
all,
because
if
some
at
some
point
one
of
your
service
tries
and
is
denied
that,
how
does
it
react?
What
are
the
you
know?
What
is
a
ripple
effect
of
that
on
to
the
Earth's
overall
system?
A
That
will
find
the
kind
of
thing,
because
you
don't
know
you're
going
to
try
that,
where
we
skills
in
doing
so,
you're
going
to
build
an
experiment,
say
what
I'm
going
to
try
to
set
and
we'll
see
what
happens
so
yqf
enjoying,
because
reliability
is
strategic
to
everyone.
However,
people
say:
that's
projection
system
chaos,
engineering.
It's
really
strategic
to
you
because,
like
Oh
to
the
o2
Network
I'm,
not
sure
that
they
did,
you
know
that
they
had
a
good
end
of
week.
We're
not
actually
the
old
network
went
down.
A
So
at
this
stage,
I'd
like
to
start
opening
up
a
bit,
we've
said:
we've
got
chaos
and
ring
to
open.
You
know
to
run
experiments,
whether
stage
we
need
actually
to
collaborate
together
like
I'll
say
later
on.
Why
do
you
come
to
conference
because
you
want
to
discuss
with
others?
You
want
to
see
what
others
do
in
other
companies.
You
want
to
collaborate
and
that's
the
only
place
where
you
can
actually
go
and
talk
to
your
fellow
you
know
engineers
and
say:
how
do
you
do
that?
How
do
you
cope
with
that
thing?
A
But
really
you
need
the
platform
that
platform
is
very
wide
world
here.
It's
just
say
you.
We
need
a
space
that
is
neutral
where
we
can
exchange
and
talk
and
collaborate
together.
Let's
have
a
shot.
The
two
before
going
on.
I,
don't
know
if
you're
using
Google
cloud,
but
if
you
are,
they
provide
very
awesome
before
Black
Friday,
they
add
long.
Article
looks
pretty
good
about
getting
ready
for
that.
If
you
are
using
their
platform,
they
even
mentioned
is
a
store,
recovery
and
chaos
monkey
again,
but
there
was
nothing
actionable.
A
So
you
got
a
lot
of
content.
You've
got
a
lot
of
thoughts,
but
nothing
you
can
action,
nothing.
You
can
run
yourself
and
to
me,
Google
cloud
or
any
other
provider
should
give
you
something
that
you
could
run
against
their
own
platform
and
say:
if
you
do
run
this
you
should
you
will
learn
those
things
right.
That
should
be
actually
extremely
important.
You
can't
just
bet
that
they
tell
you
all
trust
us.
We've
got
a
good
SLA,
that's
not
enough!
A
A
It's
a
short
document,
basically
where
folks
from
Netflix
and
other
places
have
gathered
and
and
basically
you
know
stated
what
it
means
to
do-
chaos,
injury
and
they
gave
us
some
vocabulary.
It's
very
nice
to
read.
You
should
actually
go
in
and
have
a
look.
It's
worth
the
shot
read.
You
start
with
asking
question
write
a
question
like
what
happens
is
my
tailor?
Certificates
died,
expire
or
what?
A
A
So
once
you've
done
that
you
are
asked
you
have
got
your
question,
you
build
an
experiment
that
experiment
simply
says
I'm
going
to
look
for
the
system.
When
sorry
I'm
going
to
look
at
the
system,
I'm
going
to
perform
one
action
that
will
degrade
or
change
the
conditions.
Some
people
do
breaking
some.
You
know,
but
it's
not
necessary.
What
you
do
is
simply
change
the
condition
of
your
system
to
try
to
highlight
whether
or
not
your
first
question
is
actually
valid
or
not.
You
observe
you
observe
your
system.
A
A
You
know
the
science
method,
you
ask
you
a
question:
you
form
an
iPod
cities
and
you
run
an
experiment
and
you
find
out
what
gives
pretty
cool
because
you
can
now
say
I'm,
engineer
and
scientist
and
you
explore
your
system
reliability,
you
capacity
to
sustain
some
failure
and
some
thought
and
because
chaos
engineering
is
now
science.
It
should
be
open
science,
that's
where
we're
coming
back.
You
know
if
you
do,
that
on
your
own
you're,
still
not
learning
from
other
systems
and
I
would
be
so
lovely
and
so
powerful.
A
If
you
could
just
go
to
a
place
where
you
get
those
findings
and
those
experiments
and
you
run
them
and
you
learn
from
that.
You
really
want
other
people
to
actually
tell
you.
Oh
you're,
using
issue
I'm
good
with
this.
You,
maybe
you
run
that
because
honestly
I'm
not
an
expert
in
issue,
so
I'd
love.
A
If
some
expert
could
actually
give
me
the
basics
and
just
make
sure
that
I'm
confident
with
those
things
that
I've
set
up,
so
we
must
strive
to
share
experiments
in
finding
now
I
understand
that
it's
not
easy
because
of
sensitive
data
and
things
like
that,
but
we
can
find
a
way
to
she
share
enough
that
you
know.
Other
people
will
find
that
valuable.
A
So
we
need
to
work
obviously
on
open
standards
and
API,
so
that
goes
without
saying
Kuban.
It
is
paved
the
way
here
who
would
have
thought
that
five
years
six
years
ago,
so
many
big
players
who
come
to
together
and
actually
come
with
one
API
and
one
thing
we
keep.
We
need
to
keep
really
hard
as
a
community
of
engineers.
That
thing
is
really
really
important.
A
We,
if
we
do
that,
if
we
don't
do
that
again,
we'll
fail
and
we're
going
to
isolate
and
create
silos
company
silos
on
each
on
the
bricks
and
we
bring
on
top
of
that.
Do
we
have
to
be
careful
yet
serverless.
The
world
7s
working
group
is
a
very
good
example
of
that
they
are
doing
a
good
job
at
federating
across
the
industry
again,
and
what
that
leads
me
to
is
the
open
chaos
initiative.
A
So
we've
got
the
principal
we've
got
the
principal
the
principal's
give
you
a
vocabulary
that
states
where
we
are
the
open
key
initiative
is
more,
but
we
need
to
share
that
together.
We
need
to
work
together
on
that,
because
if
we
don't,
then
we
can't
learn
from
each
other.
So
we'll
be
back
obviously
to
conferences
and
we'll
ask
you
no
questions
but
it'd
be
nicer.
If
we
could
go
to
places
and
say,
can
you
send
me
your
experiment
that
you
run
for
that
specific
use
case
and
I'm
going
to
run
that
again?
A
A
A
This
is
really
important.
You
know
this
goes
across
the
industry.
We
need
to
get
better
at
this,
because
no
one
wins
when
the
company
goes
down
because
I
had
a
security
breach
or
reliability
issue.
No
one
wins,
so
we
need
to
get
better
at
it
and
that's
why
you
know
we're
building
codes
and
doing
siobhan
that,
thank
you
very
much.