►
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
Tutorial: Using Istio - Lee Calcote & Girish Ranganathan, SolarWinds (Limited Seating Available - See Description for Details)
In the adoption of cloud native technologies engineers find one of the greatest challenges is the management of services in distributed systems. Service management challenges include service discovery, load balancing, fault tolerance, end-to-end monitoring, dynamic routing for canary deployments and securing communication. Istio solves these problems by providing a layer of infrastructure between the services and network that allows service communication to be controlled outside the application code. During this workshop you will gain hands-on experience as we walk through deploying Istio alongside microservices running in Kubernetes. In a series of hands-on labs, we see:• Traffic Management, Resilient Communication and Load Balancing• Policy Enforcement and Rate Limiting• Telemetry, Monitoring and Reporting• Securing Communication• Canary Deployments
To Learn More: https://sched.co/Gram
A
A
A
All
of
the
above,
depending
upon
your
service
mesh,
which
in
and
of
itself
is
maybe
a
bit
of
a
you
know
like,
like
any
new
tech
term,
probably
needs
a
bit
of
a
definition,
a
bit
of
a
clarification
about
what
is
the
service
mesh?
What
qualifies
is
one
I
think
I
can't
blame
any
of
the
projects
or
of
the
vendors
for
wanting
to
take
that
super
buzzy
word,
and
maybe
just
sort
of
rub
it
up
against
their
project
and
things
and
cool
my
to
suggest
otherwise.
A
Large
it
attempts
to
address,
or
does
address
a
fair
number
of
concerns,
as
such.
It's
probably
an
interesting
place
for
vendors
to
extend
or
for
integrations
to
be
incorporated,
also
makes
it
very
powerful.
It
also
makes
some
workshops
like
these
very
interesting,
because
there's
just
there's
lots
to
talk
about
as
a
matter
of
fact,
we've
given
this
workshop
a
number
of
times
over
now,
we
will
not
make
it
through
all
the
material
today
I
was
anticipating.
A
This
was
more
of
a
three
or
four
hour
session,
and
and
guess
what
it's
gonna
be
a
super
jam-packed
90
minutes,
it's
gonna
be
exciting.
Some
of
you,
I'm,
seeing
have
already
gotten
a
jump.
The
gun
and
you've
headed
out
to
you've
probably
seen
a
tweet
from
me.
That
said,
hey
you
go
to
this
URL
I'm,
not
going
to
distract
you
now
with
what
that
is,
but
I'll
say
that
you
may
have
seen
some
other
tweets
from
me.
Maybe
Oh
at
the
last
time
that
what
was
it
well
with
dr.
A
Connie
you,
but
a
velocity,
London
and
I
couldn't
help,
but
just
had
a
notice
randomly
the
consistent
wearing
of
argyle
socks
that,
for
those
of
you
who
are
looking
for
a
singer
tidbits
about
my
particular
fashion,
senses,
apparently
its
argyle
socks.
That's
all
I
ever
seem
to
pack
there's
a
reason.
I
brought
that
up
today,
I
wanted
to
talk
about
another
piece
of
fashion,
and
it's
it
may
come
to
me
I'm
blank
with
that,
though,
you
know
to
hurry
up
and
get
into
some
of
the
material
to
hurry
up
and
hopefully
warm
up
our
fingers.
A
These
are
some
things
about
me:
I'm
wear
a
few
different
hats,
I'm
involved
in
a
few
different
ecosystems.
If
you
were
on
diverse
to
dad
jokes
and
corny
dry,
humor,
there's
more
of
that
there
really.
The
reason
to
show
this
slide
is
is
really
much
more
about
the
notion
that
not
only
this
talk
all
of
its
resources
and
a
number
of
other
talks
or
at
the
Cal
Cote
studios.com
slash
talks
URL.
A
So
if
you're
feeling
anxiety
about
the
fact,
we
may
not
make
it
through
all
of
the
talks
that
all
of
the
labs
today,
which
we
won't
and
or
some
of
the
material
gets
glossed
over
quickly.
Point
of
reference
I'm
joined
today
by
gireesh,
raga,
Nathan
I
agree.
She
reaches
that
full
time
focus
at
SolarWinds
folks,
familiar
with
SolarWinds
curious.
A
It's
an
interesting
to
take
a
survey
at
a
at
the
epicenter
of
cloud
native
now.
I.
Remember
why
it
is
that
I
brought
up
fashion.
That's
because
I'm,
not
I,
left
my
belt
at
home
today,
and
so
I
did
bring
my
argyle
socks.
But
if
you
see
me
doing
this,
it's
because
I'm
beltless
today
so
another
fashion
tip.
For
me,
those
are
free,
some
other
things.
I
want
to
talk
about
our
micro
services.
A
We
don't
have
to
talk
about
micro
services
or
rather
service
mesh
in
context
of
micro
services,
but
the
value
that
service
messages
provide
is
probably
most
prominently
seen
in
these
types
of
environments.
This
is
where
you're
gonna.
This
is
where
you're
going
to
come
to
understand
that
there
are
many.
You
know
this.
Isn't
the
micro
services
talk
but
you're
going
to
come
to
understand?
There
are
benefits
to
doing
things
in
this
architectural
style.
A
It
does
so
in
a
uniform
way
so
kind
of
to
the
second
point
that
you're
doing
micro
services
today,
if
you're
trying
to
overcome
I,
think
someone
said
or
instrument
or
implement
circuit
breakers
or
or
do
retries.
Someone
had
called
that
out.
These
are
things
that
you
can
do
with
client-side
libraries.
You
know
gokit
hysterics,
dotnet
libraries.
You
know
your
language
of
choice.
A
So,
while
there's
freedom
to
choose
your
particular
language,
who
are
the
best
instrument
for
a
given
service,
as
you
go
to
address
some
of
those
distributed
systems,
concerns
about
what
happens
when
one
service
can't
talk
to
another
or
when
what
happens
when
one
user
is
being
abusive
of
my
API,
how
do
I
rate
limit
them?
How
do
I
tell
them
to
back
off?
How
do
I
do
Canaries
in
a
controlled
way?
Sort
of
smashes
attempt
to
address
these?
A
So
some
of
the
monitoring
or
some
of
the
tooling
that
you
use
today
may
be
a
bit
outmoded.
Vendors
have
been
working
hard
in
the
space,
and
projects
have
been
working
hard
to
to
overcome
some
of
their
former
architectural
choices.
When
infrastructure
used
to
be
a
lot
more
static,
things
are
a
bit
more
ephemeral
now.
So
why
use
a
service
mesh?
Well,
it's
to
overcome.
Some
of
the
challenge
that
we
just
talked
about
is
to
maybe
avoid
building
in
infrastructure
concerns
directly
into
your
code.
A
Maybe
it's
to
hand
that,
off
to
a
different
layer,
a
different
independent
layer
that
the
service
mesh
layer,
let
it
deal
with
with
some
of
these
concerns
the
the
rate
limiting
of
a
of
calls,
the
you
know
the
retrying,
the
the
you
know
matter.
I
may
be
injecting
some
chaos
doing
some
introducing
some
latency
and
some
jitter
I'm
having
fine
grain
granular
control
over
your
the
traffic
within
your
cluster
having
it
be
more
secure.
A
A
I
got
to
watch
some
of
what
I'm
saying
here
is
a
I
get
you
know
it's
true
it
it
often
deployed
on
top
of
a
container
Orchestrator
I,
don't
have
to
be
again,
I
think
the
more
services
you
have
the
more
prevalent
some
of
these
challenges
are
going
to
be
the
more
benefit
you're
gonna
find
from
a
service
mesh,
as
I
mentioned,
the
container
Orchestrator.
Some
of
you
may
be
asking
this
question:
hey
Becky,
you
know
I
thought,
that's
why
I
had
a
container
orchestrator
and
I.
A
Could
it
was
going
to
deal
with
keeping
track
of
you
know,
rescheduling
things
and
figuring
out
and
tracking,
where
they're
going
and
service
discovery-
and
you
know
and
hey
like
my
container
Orchestrator-
does
application
some
amount
of
application,
health
checking
and-
and
you
know
some
of
these
things
you're
talking
about.
So,
what's
the
deal
gone
around
talking
about
container
orchestrators
for
Timon,
postulated
that,
if
a
container
orchestrator,
if
a
project,
if
a
if
it
didn't,
have
cluster
management
and
scheduling
the
two
things
in
yellow,
you
couldn't
call
it
a
container
Orchestrator.
A
A
couple
of
years
ago
we
were
trying
to
figure
out
what
a
whether
or
not
to
call
it
container
orchestrators
or
something
else
and
which
one's
qualified
and
in
a
lot
of
respects,
that's
been
kind
of
my
definition.
Container
works
traders
have
had
their
focus
on
infrastructure
concerns.
Necessarily
so
we've
got
to
start
there
and
deal
with
distributed
system
concerns
at
those
lower
level
layers.
You're
gonna
hear
me
bastardize
the
OSI
model
a
little
bit
today,
but
I
think
it's
well
I
think
it's
well
well
intended,
and
maybe
it
will
help
be
helpful.
A
But
if,
in
my
mind,
container
orchestrators
are
most
prominently
focused
on
layer,
four
on
down
doesn't
mean
that
they
don't
do
layer,
seven
things
or
doesn't
mean
that
many
of
them
over
time
aren't
providing
for
application
level
concerns
the
code.
That's
you
know
the
workloads
that
are
running
inside
your
containers,
but
they
are,
but
they
certainly
don't
address
all
of
what
your
application
needs.
A
Another
common
question,
another
piece
of
infrastructure
that
you're,
probably
or
you
might
be
running,
is
an
api
gateway,
common
question
to
say
the
things
that
the
api
gateway
is
providing
rate
limiting,
for
example,
I
here
you
know,
service
master
is
providing
that
as
well.
So
does
this
mean
I
need
to
get
rid
of
my
API
gateway,
or
these
complimentary
I
show
you
them
both
well,
the
beautiful
answer
there
is
no,
it
depends.
A
Let
me
characterize
the
API
gateway
as
being
probably
primarily
concerned
with
north-south
traffic,
whereas
the
service
messages
is
a
bit
more
centered
around
east-west
traffic,
so
in
in
truck-
and
you
know,
clustered
traffic.
Certainly
the
surface
mesh
ingress
is,
and
egress
is
traffic
out
of
the
cluster
as
well,
but
you'll
often
find
a
deployment
model
where
you've
got
an
API
gateway
out
in
front
of
your
cluster
that
coming
through
ingress
into
into
something
like
kubernetes
and
then
into
the
service
mesh
itself.
A
We're
gonna
get
some
illustrations
going
on
here
in
a
minute
and
hopefully
make
this
a
bit
more
real
okay.
So
if
we
continue
to
need
a
container
Orchestrator
continue
to
probably
need
api
gateways.
We
continue
to
need
actually
a
litany
of
other
things
within
our
cloud
native
environments,
our
cloud
native
deployments.
A
A
The
importance
of
the
network
couldn't
be
more
prominent,
maybe
outside
of
a
distributed
system
like
super
critical
to
how
it
is
that
your
services
are
responding
either
young
and
and
getting
inside
just
sort
of
immediate
insight
without
changing
code
into
logs
and
metrics
from
those
those
network
requests
also
facilitating
distributed
tracing
for
those
requests.
We're
generally
talking
a
lot
of
HTTP
here
or
h2,
and
and
and
some
other
protocols
are
making
their
way
into
service
meshes.
A
Clearly,
what
you
need
is
a
service
mesh
I
briefly,
because
I
want
to
get
into
the
labs
I
think
we're
good
at
we've
got
by
the
way,
10
10
labs.
We
make
it
through
4
labs,
I'd,
be
awesome,
I
don't
want
to
jinx
it,
but
if
heck,
if
we
make
it
through
the
first
one
based
on
conference
Wi-Fi,
that
would
also
be
fantastic.
A
Fortunately,
I
didn't
drive
my
car
here.
You
cannot
go
out
and
key
it.
Moreover,
it
is
not
my
fault,
the
Wi-Fi,
so
just
you
know
this
definition,
you
know
makes
things.
You
know
very
obvious
what
the
heck
has
serviced
meshes.
It's
one
of
those
more
formal
definitions.
It
probably
doesn't
really
communicate
to
you
meaningfully
what
what
they
are.
A
Some
people
have
described
them
as
a
micro
services
platform,
partially,
maybe
others,
including
myself,
articulated
as
probably
a
services
first
network
services,
meaning
your
micro
services,
your
applications
being
concerned
with
them
and
their
needs
of
the
network
I'm
going
to
go
through
some
more
specific
examples.
We
call
out
a
few
point
being
is
orchestrators.
Perhaps
don't
get
you
all
of
what
you're
need?
You
know
neither
do
service
meshes.
A
If
they
do
the
service
mesh
in
general,
I
can
say
the
service
meshes
were
to
refer
to
them.
All.
My
hope
is
that
some
of
them
are
going
to
shift
left
a
little
bit,
maybe
integrating
into
the
CI
systems.
Maybe
you
know
to
the
extent
that
they
facilitate
Canaries
and
fine
grained
traffic
control.
Maybe
we
could
control
that
a
bit
earlier
in
the
pipeline,
someone
go
start
that
up
and
then
come
back
an
hour.
A
A
The
concept
there
is
the
service
mission
is
sort
of
the
next.
In
my
mind,
not
everyone
follows
the
same
cone
cloud
native
trail.
If
you
will
or
this
path,
but
a
lot
of
times,
things
start
with
containers
move
over
you
get
enough
of
them.
You're
going
to
contain
orchestration.
You
get
enough
services
running
in
there
turns
out.
You
need
another
layer
of
tooling
service
smash.
This
layer
I
think
helps
decouple
ops
and
dev
a
little
bit,
not
that
we
don't
want
these
two
groups
working
super
close
together
in
the
same
team.
A
It
decouples
to
the
extent
that
today,
I
guess.
Let
me
ask
you
this
there's
a
hopefully
a
thought-provoking
question
in
your
environments:
who's
responsible
for
defining
and
enforcing
a
rate
limit
or
in
implementing
a
circuit
breaker
who
determines
when
that
one
that
should
go
off
and
on
how?
How
often
you
should
retry
how
I
keep
it
using
a
lot
of
the
same
examples.
Service
measures
do
a
lot
more
than
than
what
I'm
referring
to
by
sometime.
You
know
certain
environments,
that's
the
developer,
may
be
in
certain
environments.
A
That's
the
operator
today,
more
often
than
not,
we
probably
find
that
people
are
addressing
these
concerns
in
their
application
code,
which
probably
you
know
going
back,
and
you
know
as
a
reminder
again
means
that
you'd
have
to
do
a
release
of
your
application
code.
You
know
roll
out
a
new
version
if
you
wanted
to
change
any
of
those
the
way
in
which
that
traffic
flows
and
if
you're
using
multiple
languages,
you
probably
aren't
necessarily
getting
uniform
observability
out
of
them
that
the
the
telemetry
that's
coming
out
may
not
be
the
same.
A
The
the
way
in
which
those
libraries
behave
and
the
capabilities
of
those
libraries
to
control
the
traffic
and
to
provide-
and
you
know,
provide
various
things.
It
probably
isn't
the
same.
So
a
service
mesh
provides
uniformity
here.
It
also,
and
to
the
point
of
like
helping
dev
and
ops
iterate
a
little
bit
more
quickly
and
maybe
diffusing
some
responsibilities,
I
like
actually
the.
If,
if
any
of
you
want
to
ask
those
questions,
I
hey
whose
responsibility
is
it
to
you
know,
taking
ups
define
the
retry
count?
If
any
of
you
work
well,
I'm,
not
I!
A
Think
you
know
it's
half
the
time.
It's
you
know,
I'm,
not
sure
that
responsibility,
maybe
is
diffused
or
rather
rather
sorry
can
confuse
and
maybe
maybe
bringing
in
an
additional
layer
helps
diffuse.
The
responsibility
helps
a
actually
here's,
a
separately
separately,
addressable
tool
to
control,
to
control
these
things,
to
provide
that
that
the
the
observability
the
encryption
at
the
traffic
control-
and
you
can
do
it
declaratively
and
yeah
mole-
and
that
can
be
versioned
and
maybe
doesn't
take
another
release
of
your
application,
and
so
hence
can
iterate
a
bit
independently.
Teams
can
address
it.
A
So,
okay
with
that,
maybe
literally
warm
up
your
fingers
get
the
Wi-Fi
revving,
let's
jump
it
to
the
labs.
The
not
only
was
the
other
URL
that
I
mentioned
an
excellent
resource
and
will
point
you
here,
but
much
of
today's
labs
will
be
on
layer,
5,
dot,
IO
you'll
find
it's
a
the
site
itself
is
a
bit
of
a
community
curated
set
of
information
about
service
meshes
you'll
find
some
other
resources.
You
may
or
may
not
find
interesting,
but
maybe,
as
a
first
step,
both
to
facilitate
some
in-class
questions
in
and
amongst
each
other.
A
Maybe
some
of
you
are
able
to
address
questions
from
one
another.
I'll
go
ahead
and
join
the
slack
will
we'll
be
joined
here.
The
slack
is
a
persistent
account,
a
persistent
channel
from
any
of
the
conference's
that
we've
given
talks
at
so
you
will
find
others
who've
been
who
are
virtual
former
classmates
of
yours
folks,
who
have
already
moved
the
tassel
over
who
probably
have
some
questions
so
may
be
a
good
place
to
engage
and
interact
with
that.
We
sort
of
preemptively
said
hey,
maybe
to
avoid
some
of
the
Wi-Fi
issues
we
have.
A
A
Ok,
first
lab
so
we're
here
at
cube
con.
Hopefully
many
of
you
are
sort
of
along
your
way
in
the
journey
where
it's
you're
familiar
with
docker.
Hopefully
many
of
you
actually
have
maybe
docker
for
desktop
installed,
to
the
extent
that
we
have
Wi-Fi
issues,
we're
gonna
resort
back
to
probably
that
tool.
We
shall
see.
First
lab:
if
you
go
to
layer
5
that
I/o,
you
will
find
a
link
to
the
github
that
lays
out
all
of
the
labs.
A
These
are
all
the
labs.
The
first
one
is
to
go
ahead
and
set
up
a
quick
kubernetes
cluster
today,
hopefully,
hopefully,
the
only
thing
you
need
is
your
browser
and
where
you
can,
if
you're
familiar
with
play
with
docker,
there's
also
a
play
with
kubernetes,
thanks
to
our
friend
Marcos
Nils,
a
code
docker
captain
of
mine
and
to
some
work
that
gireesh
is
done.
A
What
you'll
do
is
bring
up
a
couple
of
screens
of
pages,
so
so
in
one
bring
up
the
the
workshop
itself
and
the
overview
first,
one
setting
up
kubernetes
second
thing:
you're
going
to
want
to
do
is
go
out
and
go
to
pwk
play
with
kubernetes
dot
layer,
v,
dot
io.
You
will
hopefully
get
a
session
established
on
your
own
pw
k,
your
own
play
with
kubernetes
environment
in
there
we're
gonna
go
ahead
and
create
a
three
node
kubernetes
cluster
I
hope
to
do
this
relatively
quickly.
A
B
A
A
C
A
A
All
right
understanding
that
we
had
a
significant
number
of
hands
raised,
but
not
all
I
think
the
fallback
plan
here
is
and
please
don't
go
download
it
now,
if
you
don't
have
it
I
guess,
but
is
docker
for
desktop
super
convenient
way
of
getting
up
a
kubernetes
cluster,
a
single
node
kubernetes
cluster,
the
lab
will
work
on
a
single
node
or
on
three
nodes.
We're
gonna,
hopefully
walk
through
three
nodes
up
here,
but
single
node
will
work
as
well.
A
D
If
you
go
to
layer,
fountain,
io
and
scroll
down,
you
will
actually
find
the
current
event
like
an
oak,
yukon,
2018
or
yeah
cubic
or
north
america
2018,
and
there
he's
actually
not
showing
it
on
the
screen
now.
So
you
have
the
labs
links
from
lab
to
the
car,
then,
actually
the
get
up
material
and
the
lab
resources
is.
A
Okay,
so
what
while
that's
coming
up,
the
other
thing
to
the
other
resource
has
been
provided
today,
for
the
labs
is
we're
gonna
talk
about
sto
mixer
and
some
of
these
deals
points
of
extensibility
there's,
there's
an
adapter
in
there
from
SolarWinds
it
integrates
with
log
li
paper
trail
app
optics.
If
you
want
to
see
logs
or
metrics
coming
out
of
your
deployment,
there's
an
optional
lab
that
you
can
go
through
to
do
that.
On
the
on
layer.
V
io,
there
is
there's
a
lab
resources
link.
A
A
All
right,
we're
gonna,
we're
gonna,
do
the
switcheroo
to
docker
for
desktop
here,
those
that
have
got
play
with
kubernetes
up
in
your
browser.
The
instructions
on
on
github
we're
gonna,
what
we're
gonna
walk
through
this,
but
the
instructions
on
github
are
our
verbatim.
What
you're
going
to
want
to
do
for
the
play
with
kubernetes
environment
if
you're
running
docker
for
desktop
instead
almost
verbatim
instructions,
slight
tweaks,
we'll
walk
through
it.
Now,
though,.
B
A
D
D
So
you
can
actually
see
it
state,
like
you,
know,
a
few
moments
while
it
is
pulling
the
container
images
and
then
like
setting
up,
like
you
know,
a
bunch
of
the
things
that
are
needed
now.
If
you
actually
look
at
the
screen,
like
you
know,
so
here
is
the
cube.
A
cube,
Adium
join
command
that
was
was
was
part
of
the
output.
Now
this
is
very
crucial.
D
Like
you
know,
so
every
one
of
you
is
going
to
have
your
own
join
command,
so
you
cannot
like
share
between
others,
because
every
session
is
kind
of
like
you
know,
you
can
say
it's
sandbox,
so
you
know
so
whatever
is
in
your
sandbox.
Another
person
is
not
able
to,
you
know
will
probably
not
be
able
to
see
it.
So
this
cube,
ATM
join,
is
very
important.
D
So
again,
please
don't
join.
Please
don't
use
this
use
the
one
from
your
command
output.
Now
at
this
point
in
time,
what
we
can
probably
do
is
that
we
can
actually
check
the
node
status
on
this
one.
So
right
now
we
only
have
one
node
and
you
know
so,
and
you
can
see
that
like
and
it
says,
there's
one
node.
The
role
of
that
node
is
a
master,
but
it's
not
yet
ready.
The
readiness
here
actually
indicates
that
it's
ready
to
start
accepting.
You
know,
requests
for
for
the
loads
or
for
deploying
services
etc.
D
D
Sounds
good,
thank
you
cool,
so
the
first
node
the
first
command
was
actually
they
get
nodes,
so
this
actually
gives
the
status
of
the
nodes
in
a
similar
fashion.
Like
you
know,
let's
take
a
look
at
like
you
know
what
pods
were
actually
deployed
as
part
of
the
first
initialization
on
the
master,
and
you
can
see
it
has
actually
deployed
about
five
or
six
pods
and
you
can
see
like
you
know,
pretty
much.
Some
are
running.
Some
are
still
in
the
pending
state,
but
that's
normal
they'll
become
ready,
eventually,.
D
And
I
also
have
a
sample
output.
Like
you
know,
your
output
on
your
terminal
doesn't
have
to
exactly
match
the
one
here.
What
we
are
presented
here
is
just
a
sample,
and
now
the
next
crucial
thing
is.
We
have
to
actually
deploy
an
overlay
Network.
So
for
this
workshop,
like
you
know,
we
are
going
with
weave
because
it's
a
weave
net,
like
you
know,
because
it's
kind
of
an
easy
option
you
guys
are
free
to
actually
use
anything
outside
like
you
know,
but
for
this
workshop,
like
you
know,
we'll
just
stick
with
weave.
D
So
here
is
a
command
like
Ana
for
deploying
weave
in
like
a
single
command,
and
if
you
actually
look
through
it,
you'll
see
that,
like
you
know
as
part
of
this,
you
know
this
single
command.
Like
you
know,
we
have
the
some
interpolations
in
place
like
you
know,
which
do
get
the
kubernetes
version
and
accordingly
deploy
the
right
version
of
weave
appropriate
for
the
kubernetes,
so
so
I'm
just
going
to
be
pasting
it.
Okay,
if
you
have
questions
I
mean
I'm
here
yeah,
please
feel
free
to
interrupt
us.
D
D
Like
you
know,
it
also
deployed
like
in
our
service
account
cluster
role,
role,
binding
and
like
I'm,
essentially
the
daemon
set,
like
you
know,
which
runs
we've
net
so
now
at
this
moment,
if
you
actually
go
back
and
like
you
know,
do
I
get
nodes,
you
will
see
that
the
master
now
is
actually
ready.
That's
that's
kind
of
an
interesting
observation
at
this
point
in
time.
So,
essentially,
like
you
know,
it
is
at
a
point
where
it
can
probably
start
accepting
requests.
D
It
was
just
waiting
for
the
for
the
for
the
overlay
network
to
be
up
and
running
and
the
same
way.
If
you
actually
now
check
the
pod
status
across
all
namespaces,
you
will
see
all
the
parts
are
running.
So
the
interesting
thing
to
note
is
that
the
core
DNS
part
will
not
come
up.
Looking
until,
like
you
know,
the
overlay
networking
is
actually
set
up.
So
that's
a
tiny
observation
at
this
point
in
time.
D
So
the
next
thing
is
now
that,
like
you
know,
we
we
have
one
master
node,
that's
ready
and
you
know
ready
to
move
on.
That's
actually
started
like
an
adding
more
nodes
and,
let's
see
like
if
we
can
actually
add
more
nodes
to
the
cluster,
so
I've
just
added
another
node
I'm
gonna
be
doing
it
one
more
time,
so
I'm
adding
another
node.
So
I
have
two
extra
nodes
now,
but
they
are
not
yet
part
of
the
kubernetes
cluster
so
for
them
to
become
part
of
the
cluster.
D
So
I'm
just
going
to
go
back
to
the
master
node
scroll
up
in
my
output
and
find
that
join
command,
which
is
actually
then
the
command
like
you
know,
which
is
needed
for
adding
slaves
to
the
cluster.
So
I'm
copying
it
from
my
output.
The
same
way,
you
should
be
copying
it
from
your
cube,
adium,
init
output.
D
So
all
you
have
to
do
is
like
in
a
copy
this
and
then
go
back
to
each
of
the
slave
nodes
and
then
like
paste,
the
command
and,
like
you
know,
in
the
next
few
seconds,
you
will
see
that
the
slave
gets
connected
to
the
master
like
and
it
becomes
part
of
the
cluster.
So
the
same
way
I'm
going
to
do
this
for
the
other
slave
and
the
same
behavior
like
you
know
so
in
the
next
few
seconds,
like
an
a
should
actually
see
it,
become
part
of
the
cluster.
D
Now,
if
you
go
back
to
the
master
now
one
thing:
one
thing
to
notice
that
all
the
cube
CTL
commands
will
have
to
be
run
from
the
master,
because
we
don't
have
the
cause.
The
cube,
config
information
on
the
slaves
so
from
now
on,
like
other
than
the
join
command
you'll
not
have
to
run
any
other
commands
on
the
slaves,
so
go
back
to
them,
go
back
to
the
master.
And
now,
if
you
actually
do
cube
CTL
get
nodes,
you
will
actually
see
that
there
are
three
nodes
and
all
of
them
are
ready
to
accept.
D
You
know
requests.
So
that's!
That's
pretty
much
lab
1,
so
if
you're
here
you
pretty
much
have
completed
lab
1
and
at
the
bottom,
like
I
also
have
a
cheat
sheet
like
you
know,
because
you
know
it
just
sums
up,
like
you
know,
most
of
the
commands
which
were
actually
used
in
this
lab
for
you
to
get
up
to
speed
so
that
pretty
much
concludes
lab
1.
A
Very
good,
all
right,
hopefully
you're
with
us,
given
the
fact
that
I
think
I
think
most
everyone
is,
if
you
don't
just
again,
raise
your
hand
we'll
be
by
I
think
we're
having
more
the
more
times
we
do.
This,
the
more
success
looks
like
we're
having
a
high
success
rate.
This
is
good
lab.
1
you've
got
a
three
node
kubernetes
cluster
you're,
ready
now
to
deploy
Sto.
Let's
talk
a
bit
about
SEO,
specifically
and
its
capability,
isn't
it
as
a
service
mesh?
A
So
we
mention
these
things
before
observability
resiliency,
grant
traffic
control
security
policy
enforcement,
lots
of
things
promised
we
shall
see
if
they
are
delivered
spoiler
alert.
Much
at
you
know
they
really
are.
Actually.
Maybe
one
thing
to
highlight
that
maybe
isn't
always
isn't
always
declared,
and
that
is
that
distributed
tracing
is
that
much
of
this
is
provided
with
no
code
change
and
that's
of
significant
value
to
those
deploying
service
meshes
for
those
wanting
to
get
distributed
traces
out
of
their
mash.
A
You
you
do
need
to
ensure
that
your
application,
propagates
span
IDs,
does
propagate
your
HTTP
headers,
but
that
work
is
so
much
less
involved
and
instrumenting
your
code,
otherwise
observability
with
a
missing
graphic.
Thank
you.
Wi-Fi
is
the
number
one
reason
why
people
deploy
a
service
mesh
to
figure
out
what
the
heck's
going
on
and
I
think
it's
important
to
maybe
clarify
and
set
expectations
accordingly.
That
service
meshes
are
like
we
said,
a
services
first
network.
A
It's
focused
on
the
the
telemetry
that
you
get
out
of
here
is
focused
on
network
traffic
and
and
introspecting
that
deeply.
You
know,
but
isn't
necessarily
going
to
be
an
APM
if
you're
thinking
about
observability
in
that
sense
that
hey
the
workloads
that
are
running
inside
your
container,
it's
going
to
new
to
the
extent
that
you
use
it
for
distributed
tracing.
A
You
are
gonna,
get
some
insight
as
to
what's
going
on
inside
that
workload
inside
your
application,
but
much
of
the
logs
and
the
metrics
that
we
were
referring
to
in
general
are
a
bit
more
oriented
towards
the
network
traffic
itself.
How
heavily
inundated
your
services
are,
those
those
types
of
things,
things
that
you
know
outside
of
your
the
containers
themselves
very
interesting
to
most
people.
This
is
really
what
gets
people
hooked
on
service
mesh?
This
is
also.
Maybe
the
easiest
place
to
sort
of
ease
into
a
deployment
is,
with
the
observe,
the
uniform
observability.
A
That's
because
you're
not
necessarily
affecting
your
your
applications,
you're
not
telling
it
to
work,
behave
differently,
you're,
not
telling
the
infrastructure
to
do
something.
Different
you're,
just
saying,
hey,
there's
tons
of
data
flowing
through
here
send
that
up
to
me,
I
want
to
take
a
look
at
it
and
get
better
insight.
Many
people
do
deploy
a
service
mesh
for
these
other
capabilities.
Fine
grained
traffic
control,
but
a
native
integration
with
you
know,
annotation,
like
particularly
SEOs
case,
and
a
really
native
integration
with
kubernetes
and.
A
Seven
things
finds
that
there's
a
certain
user
signed
in
and
that
user
is
maybe
associated
with
a
different
account,
a
different
work
or
different,
or
some
other
moniker
that
the
mesh
can
redirect
that
traffic
based
on
either
reject
pattern
that
you've
defined
or
some
other
label.
It
inch
respects
all
that
traffic
and
and
can
send
it
to
different
different
ways,
maybe
send
it
over
to
your
to
your
dark
mode
UI,
even
though
you
didn't
actually
create
a
dark
mode
UI.
A
Even
though
maybe
it's
helping
you-
and
this
is
some
one-
it
sort
of
expand
your
mind
a
little
bit
with
some
of
the
capabilities
that
a
mesh
could
provide.
Maybe
it
switches
out
CSS
for
you
without
you
having
to
have
made
that
code
change.
Maybe
you
turns
out
in
the
future
we're
defining
dark
mode
for
UI,
as
inside
of
a
service
mesh
maybe
provides
some
resiliency,
so
we
talked
about
the
ability
to
set
timeouts
and
maybe
timeout
budgets,
maybe
maybe
deadlines.
A
If
we
get
a
chance
we'll
talk
about
those
health
checks
for
certain
you
know
and
circuit
breakers
pool.
Our
connections
also
lets
you
inject
a
bit
of
chaos.
If
you
want
some
latency,
some
jitter
those
things
pause
here,
let
it
soak
in
maybe
my
favorite
slide
I
just
like
to
hang
out
here
for
a
moment.
So
it's
it
calls
to
me.
Okay,.
B
A
Call
me
a
P
dn't
call
me
bigoted
and
wrong.
Maybe
or
you
know,
call
me
a
lot
of
other
nice
things,
but
I
would
submit
to
you
that
if
the
thing
that's
calling
itself
a
service
mesh
doesn't
have
a
control
plane,
I,
don't
know
that
it
necessarily
meets
the
definition
of
a
service
mesh,
and
so
let's,
let's
talk
about
that
a
little
bit.
Maybe
it's
do
in
this
case.
A
It's
architecture
too
high
level
components,
the
control
plane
and
a
data
plane
for
those
that
are
network
engineers
or
have
done
a
little
bit
of
our
network
savvy.
These
these
terms
may
make
some
sense
to
you,
even
as
you've
gone
to
digest
and
understand
something
like
kubernetes.
You
hear
the
term
control
plane
tossed
around
as
well.
Hopefully
these
two
things
make
a
bit
of
an
inherent
sense,
so,
to
the
extent
that
control
plans
are
out-of-band
that
the
traffic
that
flows
on
to
the
mesh
doesn't
litter
doesn't
doesn't
go
through
the
control
plane.
A
The
control
plane
is
where
you
define
policy
and
configuration
how
it
is
you
want
your
mesh
to
behave.
The
data
plane
is
where
you're
going
to
have
well.
It
depends
on
your
type
of
deployment
for
for
ISTE.
Oh,
it's
going
to
be
probably
a
lot
of
sidecar
proxies
service
proxies
many
of
them
deployed.
This
is
the
plane
that's
going
to
handle
the
in-band
traffic
management.
A
This
is
where
all
the
traffic
is
going
to
flow
through
network
traffic
is
gonna,
be
introspected
rerouted
sto
itself,
primarily
currently
composed
of
three
major
block
building
blocks,
pilot
citadel
and
mixer,
or
we
could
talk
about
what
each
of
these
provide
here
in
a
moment.
Maybe
more
interesting
right
now
is
the
notion
that
application
traffic-
it's
a
good
representation,
maybe
how
it
flows.
A
If
you
follow
the
yellow
line,
you'll
find
that
in
the
past
your-
and
this
is
a
kubernetes
example,
and
so
in
the
past
and
you've
had
in
this
case,
you've
got
an
application
called
foo.
An
application
called
bar
foo
when
it
needs
to
speak
to
bar,
would
probably
address
bar,
send
it
requests
and
they
would
directly
speak
to
one
another
as
you
go
to
deploy,
sto
and
the
sidecar
proxy
gets
injected
into
your
pods.
A
Foo
continues
to
address
bar
it.
It
sends
traffic
its
way,
but
unbeknownst
to
foo
is
the
fact
that
there's
a
transparent
data
plane
there
there's
a
transparent
surface
mesh
happening
there.
That
actually
is
the
one
that's
intercepting
the
traffic
figuring
out,
whether
or
not
that
communication
is
allowed
or
what
to
do
with
it
and
and
facilitates
for
to
get
on
over
to
the
sidecar
proxy
for
the
other
applications.
So
so
anymore.
Your
applications.
A
Don't
not
only
do
these
workloads
not
directly
talk
to
one
another,
but
but
neither
does
any
one
sort
of
external
to
the
cluster.
Rather
all
the
traffic
is
intercepted
and
hence
fine-grained
control
over
how
all
of
this
works
you're
beginning
to
see
a
little
bit
of
telemetry
reports
that
can
be
emitted
here
up
to
mixer
a
lot
of
the
capability
in
and
for
those.
How
many,
because
we're
at
envoy
con
yesterday
a
few
very
good
and.
A
You'll
note
that
the
default
proxy,
the
fault
battery
included
in
SEO
is
envoy
envoy
is
probably
not
even
my
mind
achieving
a
lot
of
its
objective
of
around
being
sort
of
the
the
universal
data
plane.
If
you
will
have
that
Universal
component
fantastic
project,
you've
heard
about
it
earlier
in
the
keynote,
you
probably
heard
a
crack
about
C++
somewhere
in
there
C++
11,
specifically
a
point
of
curiosity
for
me
folks
doing
C++.
A
Functionality
built
into
this
this
project
enjoying
who
is
that
actually
who's
done
reinvent
docker
con
and
now
cube
con
over
the
last
three
weeks.
Who's
who's
suffering
massive
jet
lag
with
me:
yeah
solidarity,
we're
outside
of
that
who
who
paid
attention.
Maybe
the
AWS
announced
a
service
map
mesh
I,
just
kind
of
highlighting
as
relevant
recent
news.
They
they're
using
envoy
there
as
well
lots
of
more
examples
of
envoy
usage.
Then
we
could
probably
call
out,
but
that
said
it
is
the
included
battery.
A
There
are
a
couple
of
other
projects
that
are
proven
that
that
battery
can
be
supplanted
by
a
different
proxy.
Some
of
you
might
have
heard
about
engine
mesh
in
the
past.
So
if
you
use
nginx,
that
was
a
project
that
helped
you
put
in
that
proxy.
Should
you
want
to
that
project?
Engine
mash
has
since
landed,
or
rather
rather
been
been
closed
down,
linker
DL,
so
at
one
point
and
had
shown
that
mccurdy
could
be
the
proxy
or
operate
as
the
proxy
within
an
sto
mesh.
Neither
of
those
two
efforts
are
necessarily
moving
forward.
A
Just
I
haven't
hasn't
been
a
lot
of
continued
I.
Think
one
and
in
part,
continue
to
ask
also
in
part
some
of
those
vendors
are
focused
on
providing
those
capabilities
elsewhere,
so
so
pilot
the
first
of
the
three
control
plan
component.
So
we
can
talk
about
pilots
a
bit
of
the
head
of
the
ship.
If
you
will
it's
the
component
that
interfaces
with
your
underlying
infrastructure
in
this
case
kubernetes
gets.
A
You
know
receiving
constant
updates
in
terms
of
new
service
discovery
in
terms
of
where
services
are
scheduled
or
rescheduled,
so
that
it
has
a
continual
understanding
of
the
state
of
the
cluster
and
where
workloads
are
so
that
it
can
then
transcribe
that
into
envoy
config
in
this
case
and
push
that
config
down
to
each
individual
envoy
proxy.
So
that's
a
large
part
of
its
its
responsibility,
so
it
manages
that
sidecar
configuration
mixer
is
a
point
of
extensibility
for
Sto.
It
is
the
it
does
it
a
few
different
things:
I'm
some
precondition.
A
Checking
and
it'll
handle
things
like
quota
management.
Maybe
the
rate
limits
have
have.
The
threshold
has
been
hit
for
a
given
API
that
you
have.
Maybe
you
got
an
unauthorized
user.
Mixer
is
the
place
where
you
can
in
run
any
number
of
adapters
to
help
you
better
interface
with
some
of
your
incumbent,
tooling,
and
maybe
that
tooling
is
monitoring.
Maybe
it's
other
types
of
doing.
A
Let's
take
a
little
bit
of
a
closer
look
at
mixer,
real,
quick
I.
Think
visuals
always
help
me
if
you
consider
that
there's
lots
of
traffic
flowing
through
the
cluster
and
these
sidecar
proxies
are
seeing
all
that
traffic
they're
sitting
there
and
they
can
just
generate
all
kinds
of
telemetry
they're
seen
all
kinds
of
traffic
I
may
be
a
simple-minded
example
that
works
well,
for
me
is
if
you've
run
engine
X
or
Apache,
and
you
you
go
into
the
logs
of
those
web
servers
and
then
you
see
you
know.
A
W
gets
you
see
people
in
traffic
going
against
your
your
web
server
in
the
very
same
way
you'll
you
can
see
similar
audit
logs
coming
out
of
requests
are
being
made
of
of
your
services.
You'll
also
see
a
bunch
of
metrics
being
generated
and
pumped
out
same
thing
for
the
traces
that
come
out,
so
those
are
all
being
sent
up
to
mixer
and
depending
upon
what
adapters
you
have
deployed,
there
are
many
more
or
a
number
more
adapters
than
what
I've
represented
here,
we're
going
to
go
through
a
few
in
our
labs.
A
That
will
help
you
integrate
with
your
tools.
Citadel,
is
you
know
more
or
less?
You
know,
as
a
certificate
authority
managed
by
SDO
helps
provide
verifiable
identity
to
each
of
the
services,
giving
them
a
unique
identity.
It
mints
certificates,
it's
able
to
then
I'm
hand
those
out
push
those
down
on
the
the
sidecar
proxies
to
the
on
voice,
so
that
you
can.
Then,
if
you
want
to
facilitate
mutual
TLS,
some
encryption
between
each
of
your
services,
it
does
that
rotation
and
handling
okay
back
to
our
labs
lab
lets,
get
a
Co
up
and
running.
A
There
is
a
tool
for
using
you
serious
CTL,
sto,
CTL
CTL,
that
part
just
just
to
be
sure.
We're
not
cuddling
anything
today,
just
to
be
sure
that
I
offend
anyone
who
calls
it
something
else:
okay,
okay,
my
job
is
done.
Okay,
all
right,
so,
let's
jump
into
lab
2
back
over
to
back
over
to
github
and
I'm
sure
face
to
you
know
we
have
the
labs
up
here.
Oh,
we
got
key
cloak
lab.
Okay,.
A
Yep
there's
an
optional
component
to
this
lab,
so
you
do
have.
It
is
a
bit
of
a
choose-your-own-adventure
adventure.
You
should
see
that
up
at
the
top
and
that
comes
in
you've
got
choice
of
which
adapters
you
want
to
deploy.
If
you
want
to
try
out
the
sto
adapter
for
the
optics
paper,
trail
or
log
li,
do
the
optional
component
of
the
lab.
If
you
don't
want
to
skip
skip
those,
then
just
continue
straight
through
the
lab
go
through
some
of
this,
so
we'll
go
out
and.
A
D
Interestingly,
like
I
think
we
we
lost
a
session
I
know
like,
and
some
people
were
actually
experiencing
the
same
problem.
You
probably
either
accidentally
closed
it
or
we
lost
the
session.
I'm.
Sorry
like
I,
give
it
up.
You
are
actually
free
to
actually
proceed
with
lab
tool.
I
can
go
on
looking
I'll.
Just
try
to
like
you
know,
do
my
best
to
quickly.
You
know,
pick
up
speed
and
get
there.
D
So
this
is
what
I
was
trying
to
explain
to
many
folks
here,
like
you
know,
so
the
establishment
between
your
browser
and
the
and
the
servers
are
actually
using
WebSockets.
So
if
your
Wi-Fi
has
problems,
you
know
latency
issues,
then
it's
probably
going
to
kill
your
session,
in
which
case
like
you
know,
you
might
have
to
start
over
yeah
just
like
this,
but
getting
up
to
speed
is
very
quick.
D
That's
the
main
reason
like
you
know
why
I
have
a
cheat
sheet
like
an
especially
at
the
end
of
slab
one
or
you
can
actually
just
follow
the
commands
pretty
much
here
and
like
an
if
you
guys
already
familiar
with
the
legend.
You
should
be
able
to
get
up
to
speed
very
quickly,
so
I'm
sitting
with
my
kubernetes
again
shouldn't
take
too
much
time
while
that's
coming
up
like
I'm,
actually
creating
two
more
instances
and
yep.
D
So
then
I'm
back
on
the
master.
So
for
the
people
you
know
who
are
actually
following
the
lab.
It's
it's
very
straightforward
like
enough.
So
here
we
have
choose
your
own
adventure,
like
you
know
like
described.
The
only
difference
here
is
that,
like
you
know,
we
have
a
specific
command
for
docker
for
desktop
or
like
docker
for
Mac.
In
the
sense,
like
you
know,
the
reason
is
the
first
command
will
account.
It
mostly
works
on
your
Mac's
or
Windows,
or
your
local.
There
is
a
very
high
chance.
D
One
thing
to
notice:
I've
seen
like
you
know:
people
run
into
this
issue
where
you
will
actually
see
some
errors
like
you
know,
which
are
that
hey
I
don't
recognize
this
or
that,
like
you
know,
for
example,
you'll
see
a
bunch
of
them.
If
you
see
that
that's
okay,
because
what
happens
is
when
the
cube
CTL
apply?
Does
you
know
this?
D
So
I
have
my
cluster
up
in
the
next
minute
or
two
and
then
I
have
to
just
get
my
my
CNI
or
the
overlay
Network,
okay,
cool,
so
I
think
I,
pretty
much
I'm
ready,
cube
city,
you
don't
get
nodes
and
yeah
I
just
have
to
give
it
a
few
more
seconds.
So
I
have
this.
If
you,
if
you
walk
through
the
tutorial
I
can
I
have
a
watch
commands
in
several
places.
D
Now
the
interesting
thing
which
watch
is
that,
like
you
know
it
actually
waits
and
keeps
like,
you
know,
reapplying
the
same
or
rerunning
the
same
command
like
you
know,
every
two
seconds
by
default.
Everything
is
configurable,
but
you
can
see
that
it
actually
is
will
be
on
that
screen.
So
now,
not
sure
if
you
guys
noticed
like
you
know,
some
of
the
nodes
are
not
ready,
and
now
they
are
ready.
So
I'm,
just
gonna,
kill,
kill,
watch
and
come
out
so
now
I'm
actually
going
to
be.
D
You
know
deploying
is
do
so,
and
this
is
pwk.
So
I'm
gonna
use
the
first
command
here.
It
just
calls
the
file
from
our
github
repo
and
it
actually
dumps
it
to
a
local
file,
is
T
or
RT
amo
and
if
you
guys
like
an
AR,
you
know
if
you
guys
want
to
take
a
look
at
the
file,
like
you
know,
feel
free
to
using
the
VI
editor
on
this
on
pwk,
even
on
your
local,
like
Ana
you're
free
to
use
your
own
editor
to
actually
open
the
file
and
look
at
it.
D
It
is
pretty
long
like
you
know,
because
it's
pretty
much
like
you
know
the
way
we
have
configured
it
is,
you
know
it
actually
installs
all
the
free
add-ons
like
you
know
that
come
with
this
deal
again
like
an
STI
deployments,
can
be
customized
but
again
for
this
workshop,
like
you
know,
since
the
focus
is
on
getting
your
hands
dirty
with
this,
do
like
you
know,
so
we
have
just
chosen
all
the
defaults
and
gone
with
it.
So
the
file
will
be
pretty
long.
D
You
know
nothing
to
be
worried
about
also
there
are
other
ways
to
actually
deploy
so
yeah
now,
like
you
know,
if
so,
the
newer
versions
of
his
tio
can
be
deployed
using
helm
but
again
looking
for
the
sake
of
simplicity
in
his
workshop.
But
you
know
we
have
just
you
know
we're
just
using
cube.
Ctl
apply
so
so
now,
I
actually
have
the
HTML
or
sitting
right
here
so
I'm
going
to
be
doing
a
cube,
CTL
apply
a
serial
and
you'll
see
it
I.
D
Can
it
actually
is
deploying
like
a
whole
lot
of
things,
so
this
is
kind
of
where
some
people
will
actually
see
issues.
If
you
see
issues
just
just
rerun
the
same
command
and
it
will
usually
be
gone,
it's
again
like
an
are
due
to
the
asynchronous
nature
of
like
another
way
the
things
are
deployed
now
this
actually
has
like,
like
I
said,
are
gonna,
deploy
quite
a
bit
of
confidence,
so
we'll
have
to
wait
for
them
to
come
up.
D
Now,
if
you,
if
any
of
you,
are
not
using
pwk
or
using
you
know,
an
older
version
of
coburn
Ares,
probably
worse
in
1.8
or
maybe
even
1.9,
you'll,
actually
see
an
error
like
this.
Like
you
know
which
says
you
know
some
competence,
you
know
cannot
be
deployed
because
you
know
this
particular
API
is
are
not
available.
That's
fine.
The
only
thing
you'll
be
missing
is
now
in
the
new
versions
of
history.
Oh
they
have
this
automatic
sidecar
injection.
D
So
once
a
particular
namespace
is
labeled,
any
pot
that's
deployed
in
that
namespace
will
actually
have
a
part.
I
mean
have
a
sidecar
proxy
which
is
injected.
Now,
that's
the
only
thing
you'll
be
missing
on
now.
How
do
we
go
about
that?
Is
that,
like
you
know
in
this
lab,
we
have
this
piece
like
nowhere,
I
mean
we
do
have
some
steps,
for
you
know
manually,
injecting
the
site
cars
like
I
mean
that
that's
also
there.
D
D
Now
they
take
a
few
more
minutes,
but
usually
once
we
actually
get
to
the
third
lab,
like
you
know
by
then
look
and
you'll
actually
have
everything
stabilized.
So
this
is
good,
like
a
knife,
if
you
see
a
similar
output
like
no,
that's
that's
fantastic.
So
now
one
thing
before
we
actually
move
on
is
actually
that
we
probably
mentioned
about
the
mixer
adapters
that
we
had
developed
one
for
app
optics
and
largely
and
paper
trail.
Well,
it's
a
same
adapter,
so
it's
called.
We
call
it
a
composite
adapter
because
it
does
multiple
things
at
once.
D
It
actually
ships
metric,
strap
optics
logs
log
ly,
not
make
cetera.
So
please
do
go
to
lab
to
know
the
optional
lab
too,
and
you
know
if
you're
interested
looking,
you
can
actually
get
your
app
optics
token
from
the
lab
resources
spreadsheet,
like
I
know,
which
Lee
already
mentioned
like
an
earlier
in
this
earlier
during
this
workshop,
but
for
log
li
you
have
to
actually
go
and
sign
up
now.
One
thing
to
note
is:
please
do
not
use
the
email.
That's
in
that
spreadsheet.
Please
use
your
personal
email,
because
what
happens
is
for
verification.
D
It'll
actually
send
a
link
to
you
guys.
So
if
you
use
this
email
like
an
IU
you'll,
probably
be
not
I
may
not
be
able
to
you
know,
move
forward.
So
please
use
your
personal
email
like
you
know,
for
the
lochley
sign
up
for
those
like
you
know
who
have
who
are
there
and
who
already
have
an
optics
token
and
a
log
Lee
token.
You
just
have
to
store
it
as
a
variable
in
your
shell,
where
we
will
be.
You
know
configuring
our
SolarWinds
adapter
now,
in
my
case,
like
I,
already
have
that
handy.
D
You
can
see
here
like
I've,
just
made
an
order
of
my
tokens,
like
you
know
the
optics
token
and
the
lovely
token
so
pretty
much
like
an
all
you
have
to
do
is,
like
you
know,
store
in
and
then
run
sorry
stored
in
a
variable
and
then
go
back
to
your
browser
and
then
just
paste
it
so
pretty
much
looking.
Oh
so
now
the
app
optics
token
and
the
ugly
token
are
stored
as
variables
in
the
shell.
Now
the
other
command.
D
Looking
here
pretty
much
all
it
does
is
it
gets
the
file
from
github
and
then
it
uses
sed
to
kind
of
do
an
inline
replace,
like
you
know,
off
the
placeholders
with
the
variables
like
you
know,
which
just
stored
in
the
in
the
shell
and
then
I'm
kind
of
writing
it
to
a
new
file
called
solo
and
start
yeah
mo.
So
if
I
go
back
here
and
then
run
this
so
now
a
file
is
created.
D
So
let's
take
a
look
at
the
file,
so
you
can
see
at
the
top
the
app
optics
access
token
and
the
logged
me
token
have
are
kind
of
injected
in
place.
So
now
this
file
is
pretty
much
like
Anna
configured
to
actually,
like
you
know,
be
deployed
with
this
deal,
so
you
can
actually
see
that
like
I'd,
actually
configure
it's
like
an
a
bunch
of
things
like
a
log
entry
handlers
and
then,
like
you,
know
the
metrics
etc.
Like
you
know
an
you
know,
so
all
those
are
you
know
this
is
whole.
D
Like
you
know,
an
operator
would
actually
be
configuring.
The
mixer
adapter
so
that
the
logs
from
from
the
proxies
will
actually
be
sent
to
log
Li
and
the
metrics
will
be
sent
to
up
optics
like
in
iron.
Again
those
are
Sullivan's
products.
Those
are
SAS
offerings,
so
hopefully
I'm
not
sure,
like
you
know,
I,
don't
think
we'll
get
to
lap
5.
But
if,
if
you
guys
are
doing
it
on
your
own,
like
you're
on
get
lap
5,
you
should
be
able
to
see
the
logs
and
metrics
like
an
appropriate
sounds
offerings.
D
So
now
we
have
sto
deployed
and
our
mixer
adapter
configured
yeah.
So
the
one
thing
like
you
know,
which
I
have
to
do
is
apply
the
changes
which
is
in
this
file
and
you
can
see
that
it
actually
created
a
handler
and
then
it
created
two
rules
and
then
it
also
created
a
log
entry.
So
it's
as
simple
as
that,
so
now,
sto
is
deployed
and
the
mixer
adapter
is
configured.
So
it's
as
simple
as
that.
D
So
this
is
again
like
an
all
command
like
you
know,
which
is
a
way
to
verify
if
sto
is
deployed
now,
even
before
doing
that,
if
you
guys
are
on
pwk
the
nice
thing
with
pwk
is
that
right
after
stos
deployment,
if
you
look
at
the,
if
you
look
at,
if
you
look
at
the
top
of
your
screen,
you
will
actually
see
a
whole
bunch
of
blue
ports
that
are
being
exposed
now.
This
is
the
way
pwk
kind
of,
like
you
know,
shows
the
ports
that
are
exposed.
D
So
it's
kind
of
a
nice
thing
to
see
so,
but
we
can
also
verify
that,
like
an
eye
using
the
all
command
like
an
ax
cube,
CTL
get
all
and
what
this
does
is.
It
actually
shows
all
the
components
that
were
actually
deployed
in
the
steel
system
namespace
and
you
can
see
a
whole
lot
of
pods
services
deployments
or
replica
sets
horizontal
bar,
auto
scalars
and
jobs
that
are
actually
deployed
along
with
with
a
co
deployment.
D
Again,
everything
is
configurable,
so
you
know
so
that's
kind
of
a
nice
thing
with
the
new
versions
of
history.
Oh
now,
the
next
part
here
is
that,
like
you
know,
talks
about
setting
up
his
TOC
tl.
One
good
thing
is
steel.
Cdl
is
not
mandatory,
because
one
great
job
which
the
co
team
did
is
they
have
actually,
you
know,
leveraged
the
CR
DS,
the
kubernetes
er
D
is
better.
So
now,
whatever
commands
pretty
much
like,
you
know
most
of
the
commands.
You
know
you
can
actually
run
using
this
to
your
CTL.
D
So
so
as
part
of
this
workshop,
like
again
the
deployment
of
sto
kind
of
app
deploy,
like
you
know,
all
these
components,
but
one
thing
to
notice
that,
like
you
know
you
I
mean
like
you
know,
you
cannot
just
quickly
access
them,
so
we
have
to
kind
of
you
know,
do
some
tricks
to
actually,
you
know,
make
them
accessible.
There
are
many
ways
to
exposing
services
that
are
deployed
on
kubernetes.
Now
in
this
workshop,
like
you
know,
since
you
know,
the
whole
purpose
is
to
keep
it
simple,
I've
actually
showcased
to
like.
D
If
you
look
at
looking
out
this
particular
section,
which
is
talking
about
configuring
add-ons
and
exposing
services,
I'm,
actually
providing
two
options,
one
is
option
one
using
node
port
so
where
the
cube
will
be
changing
the
coronaries
service
type
from
cluster
IP,
which
is
default
to
an
or
port,
so
that
you
know
it'll
be
just
exposed
on
port
on
the
nodes.
The
other
way
is
actually
a
bit
more
secure,
like
you
know,
which
is
actually
a
port
forwarding.
Now
one
thing
to
note
is
on
pwk.
D
The
North
port
is
the
easiest
way
to
actually
get
it
to
work,
but
if
you
guys
are
actually
using
dock
refer,
you
know
doctor
for
your
emissions
or
running
on
a
linux
cluster,
or
you
know,
using
a
some
other
hosted
service
like
you
know,
you
probably
might
be
able
to
use
the
port
forwarding
capability
again,
so
I'll
just
be
showcasing
again
graph
on
ax,
for
example,
I
can
also.
How
do
you
expose
graph
on
ax
on
pwk?
So
here's
a
command
like
you
know,
which
actually
does
that.
D
So
it's
nothing
but,
like
you
know,
say
cube
CTL
and
sto
system,
which
is
like
in
the
sto
system.
Namespace
I'm
gonna
show
how
to
edit
the
service
graph
on
ax
and
now
in
this
service,
like
you
know,
just
go
to
the
end
of
the
file
and
you
will
actually
see
the
type
now
the
type
as
I
mentioned.
Cluster
IP
is
the
default
service
type
I'm
going
to
change
that
to
Northport.
D
Now
one
thing
to
note
is
if
you
actually
watch
closely
on
the
screen
at
the
ad
at
the
top
of
the
page,
like
you
know
where
the
ports
are
exposed
right
after
I
save
this
file,
a
new
port
will
actually
become
available,
so
I've
changed
the
type
to
North
port
and
then
I'm
saving
it
yep,
so
I'm
saving
it
and
you
can
see
a
new
port
was
actually
added.
So
but
again,
look
at
how
do
you
find
that
again,
like
a
non
cube?
D
Ctl
actually
has
a
nice
way
of
doing
it,
like
you
know,
suggest
to
change
the
edit
to
get
service
graph
on
ax
it'll,
actually
tell
you
the
port
number,
so
the
port
number
is
3.
1
3
6
8,
so
at
the
top
just
find
that
port,
so
3
1,
3,
6
8
and
you
just
click
on
it
and
it'll
open
it
in
a
separate
tab
automatically,
and
in
the
next
few
seconds
you
will
actually
see
graph
on
a
load.
D
D
I'm
sure,
right
so
in
the
pwk
Enron
meant,
like
you
know,
that's
it
has
its
challenges,
because
the
way
it's
working
is
I'm,
not
sure
if
you
are
familiar
with
pwer
PWD,
every
node
you're
seeing
on
the
left
is
actually
a
docker
container,
so
it's
like
double
port
forwarding
would
not
work
like
it
also.
So
the
way
we
have
got
it
work
is
that
we
have
a
proxy
service.
That
kind
of
you
know
makes
this
thing
to
work.
D
So
we
just
reason
why,
like
and
I
have
to
use
this
north-north
port
at
least
I've
not
tried
the
North
post
or
either
for
the
port
forwarding
on
pwk
and
looking
at
it
to
successfully
work.
Maybe
if
you
have
like
an
upgrade
like
it's
good
to
know
exactly
I
mean
like
again
like
you
know,
since
this
is
a
workshop
and
this
session
is
only
valid
for
four
hours.
D
I
mean
because
it's
secure,
like
you
know,
and
you
can
actually
kill
the
port
forwarding
at
any
time
and
like
in
all
the
port
forwarding
is
gone
so
so
I'm
just
going
to
be
changing
this
to
an
old
port
and
let's,
let's
find
the
port,
which
is
3,
2,
7,
1,
8
and
you
can
see
a
port
is
now
come
up
and
instantaneously.
You
can
see
the
Jager.
You
are
loading
up,
so
the
graph
on
actually
came
up
like
you
know,
so
you
can
actually
now
meddle
with
it.
D
So
if
you
guys
are
mostly
with
us,
then
you
know
you
should
probably
have
sto
deployed
along
with
all
the
other
things
I
just
mentioned,
like
you
know,
on
your
PW
instance,
or
your
local
and,
like
I,
said
the
instructions
pretty
much
work.
The
same
on
your
local,
so
you
know
shouldn't
theoretically,
have
any
challenges.
A
You
thus
far.
Hopefully,
you've
got
a
three
note.
Clustered
you've
deployed
sto
your
beginning,
yet
familiar
with
it.
Your
you've
deployed
some
adapters.
You're
ready
to
deploy
like
I
said
that
this
is
a
modified
version
of
the
canonical
sample,
app
I
think
it
will
so
we're
not
going
to
be
able
to
so
I'd
hoped
for
lab.
For
some
of
you
have
worked
ahead
and
some
of
you,
some
of
you,
are,
if
you
complete
the
labs.
I
may
see
some
tweets
from
you
later.
There's
a
bit
of
an
easter
egg
in
there.
A
I'll
be
looking
for
those
with
the
time
that
I
have
left
I
think,
rather
than
describing
kind
of
how
the
rest
of
the
labs
flow
they're
pretty
well
described
on
github
those
labs
have
been
curated.
Quite
a
bit
rather
I
think
was
probably
maybe
more
beneficial
in
our
time.
Left
is
two
things,
one
to
point
out
a
resource
that
I
think
might
answer
a
number
of
questions
that
either
haven't
been
asked,
or
some
things
that
I
said
that
might
need
to
be
reinforced.
So
one
of
those
resources
is
well.
A
A
A
The
rest
of
the
time
that
maybe
one
other
thing
is,
if
you've
gone
to
layer,
five
I/o,
there's
a
bit
of
an
emerging
project
that
I'm
fightin
I'm
coming
from
Austin
Texas
today,
if
y'all
couldn't
get
my
myself
in
grow
up
here,
what
that
he's
much
funnier
in
my
head
when
I
do
that.
So
there's
another
project
that
they're
gonna
be
working
on.
It's
called
massery.
It's
intended
to
be
an
educational
playground
for
not
only
getting
a
bit
more.
A
It
may
be
in
an
easier
way
than
even
we're
doing
today,
maybe
getting
familiar
with
and
manipulating
a
service
mesh
right
now,
it's
sto
focused,
probably
look
to
add
compatibility
with
some
of
the
other
prominent
service
meshes
I'm.
So
keep
an
eye
on
that.
If
you
would
I
think
maybe
the
the
rest
of
the
time
is
the
best
spent
fielding
some
questions.
A
Yeah
good
question
so
I
sort
of
repeat
the
question
if,
as
you're,
if
you've
configured
mutual
TLS
between
your
services,
as
just
you
know
as
you
look
at
the
communication
patterns
here
and
you're,
seeing
the
sidecar
to
sidecar
communication
and
this
example
is
nginx,
but
but
is
that
communication
going
to
be
HTTP
or
HTTPS
episode
between
services
that
will
be
HTTPS?
It
will
be
a
secure.
Oh.
A
D
E
A
Yep,
so
within
East
you
episode,
the
question
is
within
SEO
is
the
sidecar?
It's
a
sidecar
communication
encrypted
by
default.
I
have
to
go.
Look
in
part,
I
have
to
go
look
at
the
QuickStart
and
whether
or
not
that's
enabled
by
default
I
believe
like
the
first
one
is
I
think
it's
called
out
on
SEO
dot
IO
like
hey.
If
you
don't
want
to
do
in
abel
encryption,
here's
the
here's
the
way
not
to
do
that.
So.
A
Yeah
yeah
good
great
questions,
so
the
question
is
the
question
is
a
bit.
You
know
there
there's
a
bit
of
overhead:
that's
introduced
as
you
go
to
inject
sidecars
adjacent
to
every
there's,
some
over
every
Apple.
You
know
and
that
overhead
comes
in
terms
of
like
CPU
and
memory
footprint
there
and
it
comes
in
terms
of
like
yeah
in
latency
and
Troy
section
of
the
network,
packets
and
they're
forwarding.
A
The
answer
is
there
is
overhead
the
if
you
go
to
in
this
case
envoy
and
envoy
proxy
dot,
IO
explicitly
that
project
states
that
they
don't
publish
performance
metrics,
because
it's
such
an
it
depends
on
your
infrastructure
and
on
your
like
so
many
things
that
that
said,
that's
super
unsatisfactory
to
me.
It
is
a
real
point
of
focus.
It's
something.
That's
got
to
be
done.
A
Right
is
in
part
why
C++
11
is
used
you'll,
find
that
other
prominent
service
proxies
in
the
space
use,
something
like
rust,
another
low-level
language
to
ensure
that
that's
done
super
efficiently.
You
might
have
also
seen
like
BPF
EBP
F
being
leveraged
in
here.
To
do
things
in
the
kernel
to
the
I.
Don't
have
a
there
are.
There
is
an
a
project
under
the
envoy
organization
on
github
that
is
and
I
think
it's
envoy
perf
that
helps
test
envoy.
A
Thank
you
very
much.
Oh
yes,
sir.
Maybe
last
question.
Yes,
they
are
so
just
you
know.
Maybe
the
last
question
I
said:
hey,
so
you
know:
are
these
resources
available?
Are
the
slides,
available
layer?
Five
dot
IO
is
probably
is
the
place,
is
the
hub
for
service
mesh
stuff,
and
so
thank
you
very
much
guys
I
hope
to
see
you
around
the
conference.