►
From YouTube: Chaos Debugging: Finding and Fixing Microservice Weak Points - Idit Levine & Mitch Kelley, Solo.io
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
Chaos Debugging: Finding and Fixing Microservice Weak Points - Idit Levine & Mitch Kelley, Solo.io
Distributed microservices introduce new challenges: failure modes are harder to anticipate and resolve. In this session, we present a “Chaos Debugging” framework enabled by three open source projects: Gloo Shot, Squash, and Loop to help you increase your microservices’ “immunity” to issues. Gloo Shot integrates with any service mesh to implement advanced, realistic chaos experiments. Squash connects powerful and mature debuggers (gdb, dlv, java debugging) to your microservices while they run in Kubernetes. Loop extends the capability of your service mesh to observe your application and record full transactions for sandboxed replay and debugging. Come to this demo-heavy talk to see how together, Squash, Gloo Shot, and Loop allow you to trigger, replay, and investigate failure modes of your microservices in a language agnostic and efficient manner without requiring any changes to your code.
https://sched.co/MPXi
B
A
A
B
A
A
A
B
A
C
C
C
And
just
like
that,
the
debugger
took
control
of
my
screen
as
and
we
can
step
right
through
it
and
inspect
variables,
just
like
we
would
in
a
local
developer,
a
debugger
CUV
I've
got
two
here
and
five,
just
like
we
specified
looking
at
a
head,
I
see
that
we're
about
to
call
our
back-end
service.
So
let's
set
a
breakpoint
just
on
the
other
side
of
that
and
let's
go
ahead.
Let
it
run
the
code
is
executing
and
it's
it's
flowing
now
over
to
the
backend,
and
we
should
hit
that
other
breakpoint.
Oh
there.
C
C
Think
we're
dealing
with
a
little
bit
of
latency
of
it.
You
can
inspect
the
variables
here
as
well,
so
here's
two
just
like
we
saw
before
and
let's
let
it
run
again,
execution
flows
back
to
the
the
front-end.
But
let's
jump
back
into
your
app.
The
code
still
hasn't
propagated
forward,
but
I'll
just
release
that
let
it
run
through
and
and
there
you
go-
that's
the
process
has
continued
to
flow.
C
B
B
A
Says
understand
was
just
happening,
so
basically
what
is
cosh
because
I
don't
know
if
it
was
clear.
So
let
us
explain
that
first,
so
first
of
all,
squash
is
a
debugger
orchestration.
Ok,
so
in
the
end
of
the
day,
what
we
did
here,
look
all
those
you
know
squash
easily.
Show
me
all
the
port
that
running
in
my
cluster.
They
so
show
all
the
the
container
in
the
poll
that
I
chose
and
then
actually
I
could
actually
attach
whatever
debugger
I
wanted.
A
So
what
squash
is
actually
doing
here
is
doing
all
the
piping
right,
because
it's
kind
of
like
a
complicated
problem
to
start
attach
them
to
each
now.
You
can
attach
the
Java
debugger
to
the
java
microservices.
They
go
debug
after
the
government
to
the
dlv
to
the
government,
micro
services
and
just
basically
one
and
walk
through
your
code,
because
it's
a
debugger
you
can
just
change
it
in
memory
and
so
on.
So
it's
kind
of
like
very
easy,
simple
experience.
Now
in
squash
itself,
there
is
two
mode.
The
first
one
is
the
default
one.
A
That's
basically
a
CLI
right,
I
mean
this.
Is
there
is
no
really
more
than
this?
It's
not
doing
more
than
this
never
explain
explain
in
a
in
a
second
and
the
second
one
is
the
secure
mode.
So
let's
look
a
little
bit
of
our
it's
executing
in
the
end
of
the
day,
I
ever
known,
I
have
name
space
of
my
application
and
what
I
also
have
is
the
name
space
of
squash
in
the
default
version.
It's
pretty
simple,
I'm
just
attached.
A
It
basically
deployed
a
debugger
in
that
case
guilty
because
I'm
debugging
code
and
just
attach
it
to
the
today
running
and
micro-services
ironic
process
basically,
but
the
problem
that
I
really
have
is
to
understand
to
which
process
I
should
actually
attach
in
the
namespace
of
the
operating
system
PID
and
that's
basically,
the
magic
of
squash.
What
we're
doing
is
we're
trying
to
do
the
piping,
but
if
we
are
in
the
US
level,
the
PID
is
not.
We
don't
know
what
is
the
period
of
action
of
this
process,
so
we're
doing
some
game
here.
A
What
are
we
doing
with
basically
going
to
create
the
decree
object
of
kubernetes
and
actually
running
a
command
inside
the
container
of
the
micro
services,
the
application
itself?
What
we're
doing
there
is
basically
looking
for
the
namespace
and
they
are
not.
That
is
unique
to
this
specifically
a
process.
So
now
we
know
which
process
on
the
first
level,
we
should
actually
connect
it,
and
then
we
just
connecting
it
now
in
the
secure
option.
It's
a
little
bit
different
because
I
want
to
make
sure.
So
it's
like
if
I'm
running
locally
and
my
coop
come.
A
You
know
what
a
mini
comb
I,
don't
care.
I
can
I
can
actually
debug
everything,
but
if
it's
a
shared
cluster,
you
probably
don't
want
people
to
start
actually
debugging
your
processes,
and
for
that
one
we
need
to
make
sure
that
they
have
a
permission
to
do
that
and
that's
pretty
simple
what
we're
doing
is.
First
of
all,
now
we
need
a
server
right,
because
something
need
to
watch
something.
It's
not
a
see,
a
light
kind
of
like
experience
anymore.
A
The
second
thing
that
we're
doing
we,
basically
when
you
actually
ask
for
us
to
actually
debug
the
micro
services
we
putting
an
intense
C
or
D
in
the
name
space
of
the
application.
Now,
if
it
works
its
main,
you
have
permission
if
you
have
permission,
we're
letting
you
go
ahead
and
do
it.
If
not
you
not
not
be
able
to
actually
debug
these
micro
services
and
then
it's
really
simple.
By
the
same
thing,
we
realize
what
is
the
PID?
We
spinning
up
the
debugger
and
attach
it
so
it's
pretty
pretty
simple.
A
So
squash
is
open-source,
it's
actually
really
really
cool,
it
should
go
and
try
it
look.
We
are
so
low
doing
service
smashes,
distributed
system,
API
gateway
all
our
product.
The
reason
we
actually
released,
that
is
because
we
using
it
internally
constantly.
This
is
why
we
build
it
for
yourself,
so
I
really
really
recommend
you
just
go
ahead
and
use
it.
It's
open
source
and
it's
free.
But
that
brings
us
to
the
next
question.
This
is
great,
we're
actually
doing
it
in
production
and
there.
A
So,
of
course
not
humans
will
be
crazy
enough
to
actually
attach
a
debugger
to
a
running
process
in
production.
No
one!
That's
don't
do
that
right
so
now
the
question
is:
what
can
we
do
about
it?
So,
let's
look
at
the
simple
request
that
coming
to
the
class,
all
right,
I've
pulled,
I
have
cluster
request
is
coming
it's
going
for
all
this
micro
services
going
maybe
to
a
database
going
maybe
to
s3
continue.
That's
great,
but
I
want
to
know
what's
going
on
there.
So
what
are
we
doing?
We
need
logs
right
we're.
A
Never,
you
know
debugging
anything
anymore,
it's
all
about
troubleshooting,
so
you
need
the
logs,
but
here's
the
problem
that
you
have.
There
is
a
lot
of
logs
right.
There's
a
lot
of
production,
running
they're
all
spread
it
around.
How
do
you
do
it
so
because
they
were
spreader
on
this
problem
is
already
fixed
right.
It's
called
open,
I
mean
it
was
calling
up
and
tracing
then
open
consensus,
not
open
telemetry,
but
this
is
a
tracing
system.
That's
simple!
A
The
problem
that
we
have
with
that
is
that
there
is
a
lot
of
logs
and
you
can't
start
flying
logs
over
your
network.
It
will
kill
it.
So
what
should
we
do?
So
that's
the
real
problem
that
we
have
the
open,
tracing
and
open
tracing
open
telemetry.
Now
what
they
doing
is
the
basically
of
two
of
two
way
to
actually
go
about
fixing
it
number
one
they're.
Only
taking
the
errors,
that's
all,
so
the
body
is
something
that
you
not
get.
A
The
second
thing
is
that
the
sampling
sampling
is
the
most
important
word
in
the
open.
Telemetry
word:
you
can
take
everything
so
now
you
need
to
figure
out
what
you
can
take
and
you
missing
some
stuff
right.
I
mean
it
is
what
it
is
is
a
trade-off
in
the
end
of
the
day
and
seriously
the
if
you
think
about
the
challenge
of
physical
address
intact
with
child
simply
which
one
to
sample
how
much
to
sample
it's
an
art.
A
You
don't
want
to
swap
your
network,
you
don't
want
to
get
the
data,
it's
kind
of
like
a
port
okay.
So,
that's
that
that's
great
so
now
service
fish,
I'm
a
big
fan.
This
is
what
we're
working
on
all
the
time.
These
technologies
come
about
to
actually
help
us
write
today
when
you're
actually
using
STL.
A
Maybe
you
know
I
think
linearly.
Yes,
maybe
you
basically
get
yes
look
at
this
well
you're,
basically
getting
all
these
open,
tracing
and
stuff
out
of
the
box,
it's
already
wired
and
it's
way
because
they
want
to
give
you
observability,
it's
all
about
observability
and
security.
So
that's
great
right!
It
will
look
exactly
the
same
thing
right
now.
Instead
of
actually
going
directly
to
the
application,
it's
always
going
to
go
to
a
sidecar
right.
So
basically,
that's
Endocyte
or
is
a
proxy,
and
this
proxy
in
most
of
the
cases,
is
envoi
and
sometime.
A
It's
linker
dip
fork
see
right!
So
that's
the
idea.
That's
great!
But
again
it's
not
solving
me
any
real
problem.
Right
I
still
need
to
put
it
in
offer
intuitive
and
if
you
think
about
it,
what
I
need
to
do
is
still
have
the
same
problem
of
I
need
to
sample.
I
can't
get
it
all
right,
but
everybody
it's
logging.
You
need
to
log
it
in
your
application,
you're
equally
important
library,
it's
quite
a
lot
of
mess
around
ok,
so
we
looked
at
it
unless
we
said
we're
working
service
for
two
years
right.
A
If
the
announcement
of
SMI,
it
was
basically
our
announcement
together
with
Microsoft.
So
you
know
we're
very
passionate
about
this
space
and
we
said,
look
service,
which
is
great.
It's
open
three
different
problems
right
today,
growling
right,
that's
the
first
one.
The
second
thing
is
MPLS
security
policy.
That's
the
you
know
to
be
able
to
actually
do
Microsoft's
to
security
community
to
each
other
and
the
last
one
is
basically
observability.
A
But
the
question
is:
is
that
all
it
can
do
and
no
that's
basically
why
the
company
exists,
because
we
believe
that
the
interesting
stuff
will
come
on
top
of
those
measures.
That's
why
SMI
is
so
important.
Okay!
So
let's
look
at
what
we
can
do
so
do
we
really
need
all
the
data?
Do
I
really
want
all
the
lock
directives
in
the
system
and
my
opinion
vitalik.
Let
me
think
about
it.
Do
I
care
of
the
one
that
actually
was
successful,
maybe
for
auditing,
but
I.
Don't
really
care
right,
it's
a
successful
transaction.
So
what?
A
If
I
will
be
able
to
actually
solve
this
big
data
problem
by
actually
taking
only
what
I
really
care
of
so
let's
do
this.
So
what
basically
we're
working
with
the
eyelift
actually
and
we
extended
the
top
filter
that
mod
client
did
we're
working
together?
Actually
because
that's
something
that
they
need
internally,
so
we're
building
that
for
them
and
what
we're
doing
it
basically
came
with
this
decision
of
what.
If
we
can
actually
record
record
everything
that
happened
in
your
request.
A
That's
a
data
data
world
and
now
we'll
send
it
on
the
enter,
but
I
will
be
able
to
do
that
body
and
the
error
and
I
will
be
able
to
do
by
the
way
because
of
the
server
smash
everything
that
happening
inside
my
application.
Even
if
my
application
is
went
to
a
database
or
to
s3
and
bring
me
the
data
back
I'm,
getting
the
snapshot
of
what
I
got
back,
which
is
huge
right.
A
It's
exactly
snapping
the
data
on
your
production,
which
I
think
it's
great
and
we're
calling
it
loop
right
loop,
because
we
think
that
what's
special
about
top
is
the
ability
to
actually
replay
these
things
after
it
and
that's
why
we're
doing
now
live
that's
a
little
bit
different
use
case
and
I
will
share
that
with
you.
What
lipid
is
doing
is
basically,
they
have
no
system,
and
the
system
is
that
you
know
someone
calling
because
I
don't
know
they're,
not
managing
to
invite
you
to
order
a
lift
and
what
they
doing
it
are
basically
calling.
A
This
is
this.
You
know
support
system
and
they
have
a
support
system
and
they
wanted
to
be
able
to
actually
click
a
button
and
record
everything
that
related
to
this
user.
So
that's,
basically
what
we're
building
for
them
and
then
they're
engineer
can
go
and
actually
explore
with
it.
So,
basically,
if
you
think
about
what
we
just
did,
we
recorded
the
audio
problem
in
production,
including
the
data
running
in
the
pipe.
C
We
just
fixed
that
calculator
app,
but
apparently
we
introduced
a
new
bug
and
I.
Don't
know
what
it
is
so
I'm
just
going
to
attach
loop
and
see
if
we
can
catch
any
500s
we
have
loops.
Etl
makes
that
easier.
We
can
list
any
any
messages
that
have
been
found,
but
actually
first
we
we
need
to
tell
loop
what
to
look
for
I've,
set
up
a
filter
to
look
for
five,
hundreds,
it's
pretty
simple.
This
is
just
a
basic
filter.
C
C
B
C
C
C
And
and
it
caught
in
the
debugger
as
we'd
hope,
let's
jump
forward,
see
what's
going
on,
I
think
there
might
be
something
and
compute
some
so,
let's
play
through
and
we're
down
there
and
again
we
can.
We
can
inspect
the
values
and
it's
it's
actually
with
the
inputs
that
produce
the
error.
So
I
find
that
really
helpful.
Let's
step
into
compute
Sun,
let's
scroll
down,
ok,
looks
like
it's
doing
the
math
correctly
of
it
I
think
it.
Somebody
put
an
easter
egg
in
here
anytime.
The
sum
is
between
500
and
600.
B
A
So
so
we
really
passion
about
it.
We
think
it's
great.
It's
going
to
be
open
source
very,
very
soon.
The
reason
it's
not
open
source
yet
is
because
you're
on
that
top
filter,
we
want
to
put
it
upstream
and
void
because
that
way,
we're
basically
getting
it
in
every
service
match
that
using
endpoint
was
very,
very
important
to
us
that
you
will
be
able
to
seamlessly
work
with
this
with
sto
and
up
mash,
with
console
connect
and
so
on.
So
now
it's
bringing
us
to
this
question,
that's
great!
A
So
if
there
is
a
bug
we
can
actually
fixed,
can
we
proactive
prevent
bugs-
and
this
is
a
good
question
right-
and
this
is
exactly
what
chaos
engineering
is
all
about,
so
think
about
it,
like
a
vaccine
right
I
mean
in
the
end
of
the
day,
when
you're
going
and
taking
a
flu
shot
right.
Why
are
you
doing
this
because
you
can
actually
inject
something
to
your
body
that
will
be
actually
dealt
with
and
you
know
really
would
basically
be
be.
A
They
basically
mimic
if
you
want
an
attack
on
your
body,
but
you
controlling
how
strong
the
attack
is
and
therefore
when
they
actually
they
actually,
the
attack
is
coming.
Your
body
already
knows
how
to
deal
with
it.
So
that's
basically
the
idea.
This
is
all
about
girls,
engineering,
game
resolution.
It's
really
really
simple!
A
Engineering
is
not
you
think.
Right,
I
mean
it's
exists.
It's
there
there's
companies
working
on
a
net,
fixity,
just
add
and
so
on,
but
there
is
two
thing
in
today's
system
that
you
have
to
do
in
order
actually
using
it
and
the
first
one.
You
need
a
library
right
like
everything
we
need,
a
library,
we
need
a
library,
it's
mean
that
you
it's
only
going
to
work
with
a
specific
language
which
is
pretty
annoying
if
you're,
using
something
that
it's
not
exist.
The
second
thing
it's
mean
that
you
need
to
modify
the
code.
A
If
you
actually
want
to
import
the
library
now
you
just
need
touch,
and
this
is
exactly
what
service
much
trying
to
prevent
you
just
detach
they
actually
configuration
or
management
or
whatever
you
want
to
call
it
with
the
business
of
big
publication
log.
If
the
library
you
need
to
upgrade,
you
know
whatever,
because
the
provider
provide
a
break,
you
need
to
redeploy
the
micro
services
and
so
on,
not
really
healthy
experience.
But
now
we
have
service
wish
right.
What
did
we
do
with
service
mash?
We
have
a
cluster.
We
had
micro
services.
A
We
somehow
managed
to
abstract
the
network
on
every
traffic
that
coming
note
to
sort
that's
exactly
what
maybe
I
get
way
stirring
right.
We
have
a
glue,
but
there
is
a
lot
of
out
out
out
there,
and
the
second
thing
is
that
we
also
need
some
service
mesh
right
to
take
care
of
all
the
east-west
traffic
right.
A
That's
basically
what
we're
doing
so
in
the
end
of
the
day,
every
time
that
a
traffic
coming
to
the
cluster,
it's
going
through
this
proxy
that
the
API
get
way
or
the
ingress
is
representing
and
every
time
that
the
traffic
going
across
you
know
aim
between
microservices
in
the
cluster,
already
they're
going
to
go
through
the
sidecar
that
your
micro
service
is
using.
Now.
Why
why
it's
important
it's
important,
because
the
proxy
configuration
that
you're
giving
to
a
cycle
should
not
be
like
the
one
you're
giving
on
the
edge
it's
different
right.
A
I
mean
it's
more
secrets.
Maybe
it
should
be
more
secure,
maybe
right
limiting
interesting
in
the
edge,
but
I
don't
care
it
inside
the
cluster
and
so
on.
Okay,
so
now,
when
we
actually
obstructed
the
network,
we
can
actually
inject
on
that
level.
So
we
don't
need
to
change
your
application.
You
can
just
inject
fault
and
latency
on
the
proxy
level,
and
now
you
can
actually
do
curse
engineering
on
the
proxy
level.
You
are
not
changing
your
application,
you
don't
care
which
language
it
is
it's
just
going
to
work.
We
will
be
passionate
about
that.
A
So
again,
it's
all
about
the
control
experimenting.
It
eventually
I
mean
you
need
to
schedule.
You
know,
usually
what
people
doing
today.
They
are
scheduling,
an
experience
that,
if
I'm
going
I,
don't
know
every
Friday
at
9,
let's
say
and
then
what
you
need
to
do
is
to
watch
the
actually
metrics
to
make
sure
that
you
don't
taking
all
your
system
down
right.
A
You
want
to
make
it
control,
like
you
know
like,
like
an
experiment
that
you
can
actually
control
the
result
and
not
taking
or
your
system
down,
and
that
way
you
actually
basically
that's
exactly
what's
going
to
happen
if
it's
growing
to
a
threshold.
That
is
that
that
you
know
it's
taking.
You
know
that
you
defined
that
it's
too
warm
warning
your
placate,
your
system,
the
experiment
would
stop,
and
that's
the
beautiful
that's
all
about
the
control,
so
we
introduced
blue
short,
blue
shot
is
basically
a
chaos
engineering
on
top
of
any
service
mesh.
A
We
will
talk
about
that,
but
it's
all
on
service
match
because
of
the
SMI
announcement
that
we
make
two
days
ago,
right,
we'd
make
Microsoft
together
and
actually
and
Boyan,
it's
very,
very
important
to
us
to
build
stuff
and
extend
the
ability
of
service
mesh
on
the
level
of
their
of
their
or
you
know
the
stuff
coming
on
top
of
it.
So
again
it's
pretty
simple
right.
You
can
create
an
experiment,
you
can
say
which
services
you
want
to
inject
a
message:
delay
and
network
is
already
going
and
watching
actively.
A
C
To
demonstrate
glue,
shot,
we'll
use
Geo's
book
info
app,
just
a
quick
overview
on
that
we
have
a
product
page,
it's
the
sort
of
the
landing
page
and
that's
aggravating
data
from
a
few
different
services.
We
have
this
review
service
that
calls
out
to
a
radiance
service,
but
we
want
to
know
what
happens
to
our
system
if
there
is
a
fault
in
the
radiance.
What
if
the
radiance
service
is
not
available.
So
let's
set
that
up
with
an
experiment.
C
C
Alright,
let's
just
verify
that
the
service
is
working
before
we
put
in
the
the
chaos
experiment,
so
I'll
connect
to
it
and
we've
got
kevanna
for
reference.
So
here's
the
landing
page
and
from
that
diagram
you
see,
we've
got
the
reviews
content
over
here
and
then
ratings
and
yep.
The
app
is
loading.
Fine
has
every
load
it.
The
funnel
metrics
will
populate
in
a
time
cycle.
C
Okay,
so
that's
that's
live
and
oh
no,
we
lost
all
of
our
content
from
the
from
the
reviews,
not
just
the
ratings.
So
that's
that's
a
vulnerability
but
notice
that
as
I
continue
to
refresh
the
page,
we
triggered
more
and
more
failures
and
we
we
met
our
Prometheus
thresholds
to
terminate
the
experiment
before
it
cost
too
much
damage.
And
you
see
the
the
rise
of
500s
triggered
the
the
cancellation
of
that
experiment.
And
if
we
continue
to
reload
this
page
it'll
just
be
two
hundreds
going
forward.
C
C
C
The
the
reviews
content
still
shows
up,
even
though
we're
missing
that
radiance
data.
This
is
preferred
behavior,
refreshing,
not
getting
any
more
500s,
because
the
500
metric
is
is
measured
between
the
reviews
and
the
product
page,
and
the
review
service
has
not
not
produced
a
cascading
failure
when
it
was
lacking
the
radiance
data
continuing
to
review,
and
we
see
that
it
timed
out
on
a
timeout
basis
and
this
music
experiment
succeeded.
Our
app
is
now
more
resilient.
A
Now
the
question
is:
how
do
you
is
the
best
way
to
start
with
all
this
project?
So
yesterday
we,
for
instance,
sorry.
We
announced
their
semi
the
service
mesh
interface.
This
is
the
ability
to
actually
unify
all
the
service
mesh
functionality
in
a
very
simple
API
and
it
will
be
shared
between
every
provider.
The
fact
that
we
managed
to
bring
all
those
community
around
with
us
and
the
great
Microsoft
to
stand
blind
at
the
switch
that
it's
great
for
us
for
the
community
and
the
old
innovation
that
will
come
on
top
of
it.
A
So
let
me
show
you
what
what
we
built
on
the
last
year
and
what
we
you
know
when
we
came
with
this
SMI
and
what
we
think
that
should
be
done
to
start
with
those
applications.
So
I
will
show
you
announced
last
week
something
called
service
Mushaf
service.
My
shop
is
the
ability
to
play
with
service
right.
It's
there.
A
If
you
want
service
match,
that's
the
place
for
you
and,
as
you
can
see,
I
already
have
a
two
measures
install
one
up
mash
and
one
and
one
link
your
D,
and
what
you
can
see
is
that
I
can
easily
easily
install
another
one.
It's
pretty
simple,
I'm,
just
giving
it
a
name
I'm,
actually,
basically,
just
choosing
there,
the
namespace
that
I
want
to
deploy
in
and
which
version
I
want
to
festial
and
I'm
just
clicking
it
right.
A
It's
pretty
simple,
the
installation
of
of
Sto,
but
you
can
still
get
that
in
a
second
sto
is,
will
even
get
install
and
what
you
can
see
that
we
also
discover
every
service.
Is
that
running
and
managed
by
this
Sto?
So
you
can
see
it's
discovered,
and
now
what
we
can
do
is
actually
go
and
extend
it,
because
service
match
is
all
about.
A
The
functionality
of
these
technology
is
giving
us,
but
you
can
see,
there's
quite
a
lot
of
out
there
already
flag
your
key
ally
and
also
everything
that
we
did
and
you
can
basically
like
an
app
store,
just
go
Android,
so
just
click
and
download
it
to
your
mesh.
Now
it's
smart
because
you
know
everything
about
your
mesh.
You
know
everything
about.
You
know
the
primitives
that
you
using,
if
you
have
security,
knows
what
you
can
and
cannot
install
on
your
mash
and
it's
just
going
to
go
ahead
and
install
it.
A
So
this
is
a
very
simple
example:
it's
key
ally,
right
coming
from
Doretta
already
fact.
Folks
will
do
the
same
thing
with
look
right.
It's
pretty
simple!
Just
go
ahead:
it's
an
application!
It's
extending
it's
going
to
work
seamlessly!
Now
it's
there.
It's
free,
just
service
match
how
that
IO
go
play
with
that.
It's
a
finish
in
a
second,
but
what
you
can
see
is
that
it's
pretty
pretty
straightforward
and
simple.
A
A
A
Once
again
summarize
what
you
saw
today,
so
that
was
a
very
short
session
35
minutes
and
look
how
much
we
actually
did.
So
we
talked
about
squash,
open
source
code
right
seriously,
no
relation
to
service
mess.
This
is
how
we
develop
our
application,
go
ahead
and
try
that
open
source
again
free.
We
on
the
slack
come
ask
us
question:
if
there
is
any
power,
but
it
should
be
so
simple.
Now
loop
coming
up
soon,
with
the
lift
guys
again,
very,
very
simple:
it
will
be
open
source,
so
go
fix.
A
It
go,
go
and
basically
capture
the
problem
that
you
have
in
production.
The
last
one
is
blue
shirt.
Blue
shirt
is
a
chaos.
Engineering
already
open
source
go
make
your
application
a
little
bit
more
resilient
and
the
service
mesh
hub
again
also
free,
just
go
and
play
with
meshes.
We
have
more
portaxx,
I
will
just
say
if
you're
already
there.
Maybe
we
should
take
a
look
at
out
little,
which
is
an
API
gateway.
A
You
might
feel
the
best
that
exists
in
the
market,
CRD
natively
on
it
on
a
on
a
kubernetes
and
also
as
his
name.
It's
gluing
those
environment
of
monolithic
micro
service
and
several
s.
Scoop
is
a
gruff
clothes
ever
built
on
top
of
android,
as
well
as
glue
by
the
way
built
on
top
of
Android
and
super
glue
is
basically
the
SMI.
It's
basically
SMI
we
discovery
and
a
the
ability
to
group
meshes
together
and
treat
them
as
one
big
match.
A
Flooding
the
networking
cost
them
go,
try
to
play
cool
and
that's
it
that's
all
we
have
so
we
are
so
low
we're
care
which
we
are
hiring
Cambridge
Massachusetts.
We
are
hiring
those
look
at
our
hub.
There's
a
lot
of
stuff
open-source.
There
I
think
you
will
benefit
from
it
immediate.
This
is
me
following
us
on
Twitter
and
go
build
very
resilient
application
thanks.
So
much.