►
From YouTube: Weave & Chick-fil-A: Managing Fleets of Kubernetes Clusters... - Sean Drucker & Alexis Richardson
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
Weave & Chick-fil-A: Managing Fleets of Kubernetes Clusters w/ GitOPs - Sean Drucker, Chick-fil-A, Inc. & Alexis Richardson, Weaveworks
GitOps provides an operating model for managing multiple Kubernetes clusters at scale. In this session, Chick-fil-A and Weave will share examples from our production experience running many K8s clusters. Weave operates a global SaaS product in the cloud using Kubernetes. Chick-fil-A is running Kubernetes clusters at the Edge in restaurants. Both companies have solved the challenges of managing many clusters by combining Kubernetes' natural affinity for declarative infrastructure and its well-defined objects with Git to develop a management technique called “GitOps". D50In this session, you will learn what GitOps is and how it can help you D50effectively manage and monitor your Kubernetes clusters. We will share about the projects we have built to solve common management problems like CI/CD, K8s object generation, configuration management/drift, and observability of the stack.
To learn more: https://sched.co/GrXv
A
Hi
there
welcome
everybody,
I'm
very
excited
to
be
here
with
my
friend
Sean
Drucker,
and
what
we
submitted
our
talk.
We
had
this
kind
of
slightly
boring
sounding
title,
so
we're
gonna
have
quick
introduction.
This
is
gonna,
be
about
get-ups
who's
heard
of
get
ups.
Put
up
your
hand
if
you've
heard
of
get
ups.
Thank
you.
That's
mostly
everybody
and
you
may
have
not
have
seen
the
get
ups
sockpuppets,
who
will
tell
you
all
sorts
of
terrible
things
in
future?
A
Youtube
videos
come
and
visit
them
at
our
booth
and
see
fun
demos
of
Canaries,
but
today
we're
talking
about
get
ups
with
chickens
who
has
been
to
chick-fil-a
there
we
go
I
have
not
been
to
chick-fil-a.
I
would
like
to
introduce
you
to
Shawn
Drucker,
whose
distinguished
engineer
at
chick-fil-a
today
is
all
about
them.
They've
done
this
amazing
story
of
managing
a
huge
fleet
of
kubernetes
clusters
using
the
techniques
of
get
ops
give
ups
is
something
that
we
it's
named,
based
on.
A
Devops
practices
that
we've
works
did
running
our
own
kubernetes
clusters
and
started
writing
about,
and
we've
been
really
excited
to
have
all
these
people
meeting
us
who
we
would
never
have
met
before
who've
adopted
these
ideas
and
made
them
their
own
and
Shawn
is
going
to
show
you
a
new
open-source
project
at
the
end
of
today
that
implements
get
ops,
which
I
think
just
shows
the
power
of
the
idea
as
a
set
of
practices
so
I'm
gonna
hand
over
to
Shawn.
To
tell
you
a
little
bit
about
chickens
all.
B
Right,
Thank,
You,
Alexis,
yeah,
so
Shawn
Drucker,
a
distinguished
engineer
with
chick-fil-a
I
work
with
brian
chambers,
our
chief
technologists
and
I'm
honored
represent
all
the
engineers
that
we
have
at
chick-fil-a.
We
have
both
hardware
and
software
engineers
and
we
do
custom
solutions
there,
which
is
which
is
kind
of
unique
for
a
company
like
us,
so
Y
IOT.
So
yes,
this
is
a
good
apps
presentation,
but
our
story
starts
with
Internet
of
Things.
B
As
you
look
at
that
picture,
that
line
of
cars
actually
stretches
all
the
way
around
to
the
other
side
of
this
of
the
store.
So
we
have
a.
We
have
a
capacity
Challenge,
so
we
we've
got
a
challenge
to
try
and
serve
these
people
as
best
we
can.
So
what
little
some
fun
facts
about
what
happens
at
a
peak
hour
at
chick-fil-a
I
like
the
waffle
fries,
so
every
14
seconds,
I,
don't
know
if
I
could
eat
one
every
14
seconds,
but
but.
A
A
B
So
this
this
information
works
its
way
through
our
IOT
platforms,
so
we
basically
have
basically
have
2,000
restaurants
and
we
have
hundreds
of
thousands
of
internet
things.
What
is
the
thing?
A
thing,
a
thing,
so
a
thing
is
basically
anything
that
can
tell
you
about
itself.
So,
for
example,
we
have
we
have
chicken
fryers.
We
have
french
fries,
fryers
waffle,
fryers
grills
refrigerators,
even
a
tray,
even
a
holding
tray,
so
a
holding
tray
as
it
works
its
way
through
our
restaurant.
We
treat
that
as
a
thing,
but.
C
A
B
Well,
so
we're
just
starting
this
rollout
now
so
we're
just
starting
to
have
more
and
more
of
these
things
show
up
in
our
restaurant
and
we
need
to
be
able
to
control
them
locally
at
our
restaurant.
So
what
we've
done
here
is
we've
created
these
little
data
centers.
These
are
three
small
network
communities,
they're
very
limited,
a
gigabytes
of
RAM
each
and
we're
running
a
kubernetes
cluster.
So
we
have
a
how
much
Bitcoin
can
I
mine?
B
That's
the
thing
is
we
need
to
make
sure
these
are
co-opted
by
our
no,
but
so
yes,
so
we've
got
these
these
these
three
devices
they
fit
into
our
wiring
cabinet,
and
so
we
have
two
thousand
two
thousand
of
these.
We
will
have
two
thousand
of
these
we're
working
to
roll
that
out
kubernetes
clusters,
so
one
per
restaurant,
we
call
it
our
our
edge
compute
and
all
these
things
in
the
restaurant
actually
communicate
to
this
central
point
to
store
all
this
data.
B
B
B
Sure
sure
so,
our
main
service
that's
running
there
is
an
MQTT
broker
which
takes
all
the
messages
from
our
iot
devices
and
and
routes
them
around.
That's
our
main
thing
and
we
also
have
software
devices
so
the
the
devices
that
really
run
on
our
edge
compute.
There
can
also
be
a
software
device.
I
could
be
a
piece
of
software,
that's
acting
as
a
thing
to
do
some
sort
of
intelligence
to
to
listen
to
equipment,
to
tell
the
equipment
what
to
do
and.
B
This
this
stays
on
all
the
time
and
we
have
a
team
that
that
that
goes
out
to
in
the
tunes
to
all
this
stuff.
So
we
have
that
we
have
mm.
We
will
have
2,000
out
there,
but
so
wait.
You
know
what
just
happened
here.
We've
got
what
just
happened.
Yeah
we
have
mm.
Well,
we
will
have
2,000
kubernetes
clusters
out
there.
You
know
so.
B
All
right
exactly
it's
explained
so,
yes,
so
we've
got,
we
will.
We
will
have
2000
criminais
these
clusters
out
there
and
how
do
we,
you
know
with
our
new
problems
there?
How
do
we
do
the
communication
with
these
clusters?
You
know:
how
do
we
tell
these
clusters
what
to
do
the
consistency
of
those
clusters
what's
running
on
each
one
and
and
how
we
deploy
to
those?
B
So
we
created
these
the
seven
new
problems
that
we
had
and
what
we,
what
we
decided
would
would
work
on
this
is
get-ups
is,
is
basically
having
a
repository
in
the
cloud
as
well
as
a
ball
at
your
belt
server
for
our
secrets
and
just
to
have
our
our
edge.
Our
edge
can
poke
compute,
pull
down
the
pull
down
the
spec
files
and.
C
B
Big,
it
lives
in
the
clouds
so
up
here
in
the
cloud
we
have.
This
deployment,
which
we
call
fleet
and
fleet,
is
a
program
that
runs
in
the
cloud
that
actually
per
restaurant.
We
have
one
we
have
one
git
repository
per
restaurant
and
this
this
deployment
mechanism
will
go
through
and
start
to
deploy
the
spec
files
to
the
given
git
git
repo
in
the
cloud
and
then
on
our
edge.
We
have
this
product
called
vessel,
which
is
a
very
simple
thing.
All
it
does
is
locally
on
the
edge
it
clones.
A
B
Right,
so
it's
all
pull,
so
we
can't
at
least
that's
the
official
line.
Is
we
can't
get
into
our
kubernetes
clusters
at
the
restaurant
there's?
Obviously
an
emergency
way?
We
can
do
it,
but
for
the
most
part,
these
things
are
operating
on
their
own
pulling
down.
We
can't
get,
we
don't
get
into
them
from
the
cloud
that
would
be
a
security
issue
and
as
well.
The
cluster
needs
to
operate
in
a
disconnected
states.
So
if
we
lose
an
internet
connection,
this
cluster
goes
down.
Someone
needs
to
be
able
to
turn
it
back
on.
B
A
B
A
So
we're
just
gonna
take
a
step
back
now
and
say:
a
few
things
about
get
ups
is
a
set
of
patterns
so
that
anyone
who's
here
can
take
away
the
ideas
and
come
up
with
your
own
applications
that
follow
the
same
pattern,
even
if
you
don't
have
2,000
restaurants.
So
the
first
thing
is
this
is
part
of
DevOps.
A
This
is
about
developers
being
put
in
control
of
operations,
but
in
the
area
of
cloud
native,
principally
because
we're
in
a
world
of
automation
and
description,
we
have
descriptions
of
systems
and
we
use
those
descriptions
to
power,
automation
in
the
form
of
kubernetes
orchestrators
and
the
like,
and
hopefully
I
will
show
you
what
that
means
in
the
next
few
slides.
So
this
is
general
stuff.
Okay,
what
is
git
ops?
A
It's
an
operating
model,
it's
a
way
to
think
about
how
I'm
gonna,
run
and
use
and
manage
and
monitor
my
stack
that
apps,
the
services
and
the
clusters.
It's
also
a
way
to
deliver
software,
because
you've
got
software
and
gear
artifacts
that
are
things
like
containers
and
tools
for
pushing
those
into
the
system
which
is
a
way
to
deploy
and
update
your
cluster.
A
It
builds
on
one
or
two
decades
of
best
practices
from
SR
ease
at
places
like
Google
I,
believe
Google
do
a
lot
of
get
up
style
internally
and
some
ideas
from
computer
science
mark
Burgess.
You
may
have
heard
of
kind
of
came
up
with
all
of
these
ideas
in
the
1990s
believe
it
on
our
sort
of
model,
driven
practices
for
thinking
about
a
system
and
there's
a
set
of
principles.
A
way
to
speed
up
your
team.
A
About
a
year
ago,
Kelsey
stood
up
at
the
Austin
cube
con
and
described
it
in
my
opinion,
very,
very
simply,
as
like.
The
right
way
to
think
about
it
is
the
whole
world
of
your
computers
is
explained,
described
as
a
single
repo
or
set
of
repos
and
not
as
an
installation,
not
as
an
installation.
So
you,
you
move
your
operator
head
away
from
the
running
systems
onto
the
descriptions
and
think
about
just
changing
dos.
A
B
A
Ssh
just
hacking
into
your
system,
essing
with
it.
So
what
does
that
mean?
Basically,
here's
a
here's,
a
simplified
picture
of
how
you
might
think
about
your
stack.
You've
got
some
open-source
kubernetes.
You
might
have
had
some
things
added
to
its
and
custom
extensions,
maybe
a
patch
if
you'd
had
a
security
issue
recently,
perhaps
of
Cuba
Nettie's,
maybe
and
then
some
applications
in
config,
and
you
want
to
get
that
into
your
clusters.
A
That's
step
one,
and
then
you
want
to
do
the
one
other
thing
which
is
get
alerted
if
what's
in
the
clusters,
changes
from
what's
described
in
the
repos,
so
in
all
the
repos,
you
have
description
of
the
infrastructure,
the
machines,
the
clusters,
the
apps,
the
services,
maybe
other
things,
maybe
Canaries,
maybe
iste
ió,
maybe
graph
honor,
all
those
things
that
you
might
be
using
and
you
want
to
be
told
as
an
operator
if
what's
actually
there
has
changed
away
from
what
it
should
be
and
then
you
take
action.
How
could
it
possibly
change?
A
Don't
know
if
these
things
change,
we,
we
we
have
a
SAS
based
on
Cuban
areas
that
we've
worked
and
it's
constantly
drifting,
where
we
get
alerts
all
the
time.
So,
let's
break
that
into
four
principles
of
get
ops.
If
you're
doing
these
things,
you
are
doing
get
ups
one,
you
have
a
complete
description
or
asbestos
close.
You
can
get
to
complete
the
system.
That
is
a
declaration,
a
set
of
specs,
a
facts
about
your
system.
A
A
Second,
really
important
thing-
and
this
is
where
the
word
get
comes
in
everything
that
describes
the
system
is
versioned
using
a
suitable
tool
which
essentially
is
non
non,
reputable
I
guess,
which
means,
if
you
make
a
change,
you
get
a
log
of
that
change.
You
see
who
made
the
change
when
they
made
the
change,
why
they
made
the
change
who
did
what?
When?
And
why
gives
you
a
sense
of
why
the
system
is
in
the
current
intended
State
and
you
have
a
sequence,
so
you
can
go
backwards
and
forwards.
A
Changes
to
the
desired
state
must
be
approved,
usually
by
a
human
being,
but
could
be
automated
and
then
applied
to
the
system.
Autonomic
lis.
So
who
knows
what
this
world
also
normally
means
nobody,
because
it's
a
stupid
way,
to--what
autonomic
lis,
it
means
automatically
and
on
its
own,
so
the
system
on
its
own
gets
information
about
the
state
it
should
be
and
it
pulls
it
into
itself
and
then
it
goes
now.
A
A
A
When
you
have
something
in
a
version
control
system,
it
means
that
you
have
a
single
source
of
truth
about
your
desired
state.
This
is
very
important.
Many
things
that
go
wrong
in
its
tributed
systems
happen
because
we
don't
have
a
single
source
of
truth.
If
you
think
about
eight
2000
restaurants,
it's
very
important
to
have
a
single
source
of
truth
about
what
state
they
should
all
be
in.
A
A
Key
point
about
this
is
security,
so
when
you
talk
to
somebody
who's
worried
about
insecure
access
to
production,
well,
Gartner
released
the
report
recently
saying
one
of
the
main
threat
vectors
to
production
systems.
Now
is
the
CI
CD
pipeline.
If
somebody
has
a
lot
of
jenkins
and
somebody
gets
in
there,
then
maybe
they
can
hack
into
the
production
systems.
That's
something
that
you
see
a
lot.
How
can
we
prevent
this?
One
way
to
prevent
it
is
to
have
a
production
system.
You
cannot
ever
access
directly.
You
can
only
make
changes
by
updating
the
desired
state.
A
That's
in
the
repos.
This
is
very
important,
so
you've
got
this
concept
of
security,
boundaries,
separation
of
concerns
and
firewalls
and
finally,
really
important
observability,
always
checking
that
the
system
is
in
the
correct
state,
which
is
that
in
the
description,
the
correct
state
is
not
the
running
state,
it's
the
desired
state.
A
A
Is
github
actions,
updating
the
docker
registry
and
then
the
other
option
is
that
somebody,
an
operator,
might
change
in
config
there's
an
agent
which
in
this
case
is
we
flux
in
the
diagram
which
lives
inside
kubernetes
and
watches
the
docker
registry
and
the
config
repo
for
changes
and
based
on
information,
find
great
information.
Like
you
know,
labels
and
policies.
It
will
then
decide
what
to
do.
For
example,
it
might
put
it
into
the
staging
repo
yeah
faster
on
the
left
or
the
production
cluster
on
the
right.
So
this
means
you
can
do
things
like
promotions.
A
B
Yeah
I
think
it's
a
great
slide
to
so
for
for
chick-fil-a.
When
we
apply
to
the
kubernetes
cluster,
we
apply
all
our
spec
files
and
if
they
all
applied,
then
we
think
we're
good
to
go
for
the
most
part
and
they're
all
running,
but
we
don't
have
the
the
tooling.
That
really
can
can
tell
you
exactly
what's
going
on
with
that,
just
yet,
but
but
it's
it's
working
out
well
for
us,
but
this
the
sort
of
tool
gives
you
much
more
insight.
Yeah.
A
So
when
does
this
matter
the
most
one
when
you're
developers
would
rather
make
changes
in
get
than
learning
all
about
the
kubernetes
CLI
to
when
you
don't
know,
axle
out
direct
access
of
the
cluster
that
want
to
go
through
some
guarded
outpost
in
this
case,
making
changes
in
the
desired
state
and
three
over
to
chick-fil-a
when
you're
hitting
scale.
This
is
a
really
good
way
of
managing
big
numbers
of
clusters.
In
many
places
owned
by
many
people.
A
B
Let's
see
so,
what
we
have
here
is
chick-fil-a
open
source.
We
are
actually
starting
an
open
source
effort
at
chick-fil-a
to
publish
things
open
source
and
we
will
publish
these
tools.
I
actually
published
our
presentation
as
open
source
right
now.
So
this
is.
This
is
the
view
of
our
system
from
a
top-level
layer.
We
have
the
cloud
here
which
has
our
fleet,
which
is
our
Orchestrator
here,
which
which
pokes
spec
files
into
our
git
repository
or
pokes
secrets
into
our
vault
repository
and
then
on
the
right.
We
have
an
adds.
B
So
if
so,
taking
a
look
at
it
from
a
this
is
just
a
sample.
Get
lab
account.
We'd
have
a
we
have
a
repository
per
cluster
and
for
this
coop
can't
demonstrate
I.
Just
have
this
simple
pot
info
deployment.
What
you
typically
see
in
there,
something
we
actually
generate
things
using
reasoning,
helm.
We
actually
generate
templates
using
helm,
so
we
use
the
helm
engine
as
well,
but
we
won't
use
them
about
the
the
deployment
part
of
helm.
Would
we
do
that
ourselves?
So
we've
got
that
and
then
we've
got
Hoshi
car
bolt
and
it
has.
B
She
carved
Walt
he's
just
using
their
simple
secret
engine.
We
have
a
spot
per
cluster
as
well
to
store
our
secrets
so
inside
pod
info,
even
though
pod
info
doesn't
have
a
secret
I
just
put
one
here
anyway,
and
so
we're
just
storing
our
spec
file
in
involved
to
be
pulled
down,
so
how
this
works
is
really
a
quite
simple
process.
So
it's
basically
cloning,
a
repository
which
I'll
do.
B
So
I
just
clone
this
repository,
so
this
is
happening
on
our
edge.
Our
edge
will
actually
clone
the
repository.
Keep
that
repository
up
to
date
and
now,
if
we
now,
if
you
look
in
here,
we
have
got
our
that
repository
gallery.
Spec
files
pull
down
on
the
edge,
and
now
we
I
need
to
pull
in
our
need
to
pull
in
our
secrets.
So
we
call
a
vault
to
pull
on
the
secret.
B
So
that
pulls
that
down-
and
so
now
we've
got
a
state,
the
system
that
we
want
it
to
be
in
so
now
we
just
use
the
the
magic
of
kubernetes
and
and
apply
these
and
let
kubernetes
bring
everything
up
to
the
desired
state.
So
we'll
give
that
a
try
here
and
see
what
happens
so
we
use
this
big
long,
cube,
control,
apply,
cube,
control,
apply,
statement,
it's
it's,
but.
B
B
A
B
B
B
B
Is
so
vessel
can
actually
update
itself,
which
is
a
little
dangerous
if
you
think
about
it,
but
it's
it's
the
way
that
it's
the
way
that
we
work,
so
it
actually
works
actually
works
quite
well
for
doing
that,
and
then
you
can
just
see
it
just
polling
here.
It's
just
pulling
every
five
seconds
so
just
to
do
a
quick
change
here.
So
if
you
take
a
look
at,
if
you
take
a
look
at
what's
running
now,
we
still
have
that
we
still
have
this
pod
running
here.
B
A
B
Daesil
picked
it
up,
so
it
pulled
it
down
it.
It
actually
applied
it,
and
you
can
see
right
here
that
the
that
that
auto-scaling
changed
so
eventually
that
would
that
would
show
up
so
and
then
the
other
thing
that
vessel
does
quickly.
Is
it
just
kind
of
monitors
what
your
cue
control
apply,
does
it'll,
take
a
look
at
the
cluster
and
pulls
back
up
the
the
things
that
are
running
and
just
give
us
some
basic
visibility
into.
What's
going
on
in
each
cluster
cool.
A
B
Sure
so
so
for
home,
oh
sure,
repeat
the
question:
you
asked
how
we
used
helm,
how
we
use
helm
since
we're
not
using
helm
to
actually
deploy
what
we
do.
So
there
are
a
lot
of
helm,
charts
out
there,
a
lot
of
great
way
to
create
the
spec
files,
and
so
what
we
do
is
I,
don't
know
if
we
have
it
up
here.
But
what
we
do
it's
not
there,
but
we
do.
B
Is
we
use
the
helm,
template,
there's
helm
template,
so
you
can
do
helm,
template,
specify
an
output
directory
and
then
that
will
generate
your
templates,
actually
create
your
spec
file,
so
you're
never
really
using
how
to
actually
deploy
to
your
cluster
using
how
them
to
create
those
spec
files
and
you
those
in
your
get
repo
and
then
they
get
applied.
So
that's
that's.
How
we're
doing
it
good
question.
A
B
Right
so
this
chart
right
here:
that's
what
this
thing
is:
that's
what
this
fleet
is
right
here.
So
this
is
what
will
control,
how
we
do
a
deployment?
Let's
say
we
only
want
to.
We
won't
only
want
to.
It
knows
all
about
our
restaurants,
the
characters
sister,
restaurants,
where
they
are
all
that
that
full
set
of
data
about
the
restaurant
and
then
we
can
tell
that
to
push
into
whatever
get
repo.
We
want
to
to
start
rolling
things
out.
Take
a
look
at
the
responses.
B
Right
so
the
engineering
team
that
does
that
I'm
not
quite
actually
familiar
exactly
how
they
do
it,
but
they
do
get
shipped
pre-configured
images.
They
turn
them
on
and
it's
actually
configured
to
that
cluster
itself.
We
consider
an
IOT
device,
so
it
actually
goes
through
our
IOT
registration
with
our
cloud
to
authenticate
itself,
and
then
the
operator
has
to
authenticate
that
this.
This
cluster
is
now
active
in
their
restaurant
and
then
it
can
start
doing
what
you
talked
to
talked
about.
So
then
we
have
one
of
the
things
that
will
help
us
with
open
source.
B
Is
we
can
post
our
most
of
our
code,
now
open
source
that
we
don't
even
have
to
have
access
to
a?
We?
Don't
really
have
to
have
one
token
stored
there
to
get
to
our
vault,
and
then
we
can
do
everything
else.
We
don't
have
to
worry
about
downloading
images.
Things
of
that
nature,
so
that'll
help
us
as
well.
Gentleman.
A
Flux
is
a
bit
more
complex,
it
does
a
lot
more
things
for
you,
I
think
vessel
is
showing
you,
for
example,
the
way
that
it's
inducing
convergence
is
by
a
repeated
application
of
the
same
set
of
calls
where,
as
flux,
has
a
complete
model
of
the
system
and
actually
tries
to
do
fine-grained
things
based
on
labels.
So
you
can
do
things
like
promotion
policy
and
stuff
like
that
they're
very
similar,
though
in
the
spirit
I
mean
similar.
B
B
Great
question:
so
we
have
looked
at
hashey
car
vault
to
do
dynamic
secrets
for
certain
things
because
for
our
I
Oh
t
side
we're
using
OAuth
and
have
JWT
s
and
things
of
that
nature,
so
looking
for
siap
Revolt
can
exchange
a
JWT
for
some
other
secret.
That's
stored
and
use
those
engines,
but
currently
we're
only
using
the
static
secrets.
So
the
secrets
that
we
deploy
out
actually
gets
deployed
out
through
fleet
and
that
they're
static
sitting
there.
But
definitely
the
dynamic
secret
part
of
vault
is
is
an
important
feature.
Question.
B
Tell
us
what
our
open
source
program
sure
sure.
So
we
just
started
it.
And
if
you
look
here,
you
know
at
chick-fil-a
we,
these
are
the
three
starting
ones
and
ones
in
presentation,
ones,
the
security
thing
and
our
SSM
bootstrapping
for
bare
metal.
So
that's
actually
the
the
bootstrapping
that
we're
using
for
those
those
clusters.
B
But
we
are.
We
are
right
now
looking
at
things
that
we
do,
that
that
are
core
low-level
things
that
don't
relate
to
our
business.
So
if
there
are
things
that
are
usable
by
other
companies
that
that's
fine,
that's
great
and
have
a
community
for
that,
so
vessel
I,
don't
think
MacDonnell's
gonna
be
able
to
use
vessel
to
get
any.
You.
C
B
Advantage
you
gave
for
us
so
so
we're
looking
within
the
software
engineering
that
we
do.
We
are
building
a
software
engineering
organization
at
chick-fil-a
and
trying
to
pull
out
the
things
that
that
engineers
do.
That
may
be
useful.
You
know
to
the
open
source
community
as
a
whole
gentleman
in
the
brown
jacket.
B
So
this
chart
is
missing.
This
chart
is
missing
the
the
line
back
up
to
Fleet,
so
vessel
is
actually
applying
things
and
actually
monitors
doesn't
rudimentary
monitoring
via
cube
control
and
just
takes
a
look
to
make
sure
that
that
what-what
was
applied
is
actually
running
and
reports
that
back
up
to
Fleet
and
then
fleet
runs
in
the
cloud
and
then
we
can.
We
can
take
a
look
there
to
see
which
restaurant
is
not
behaving,
but
now.
A
Also
have
a
two
little
cube,
diff,
which
you
might
want
to
check
out,
which
is
checks
for
differences
between
the
description
and
the
kubernetes.
Okay,
I
think
we're
out
of
time
quickly.
We've
got
some
good
ops
stickers
up
here
for
people
who,
like
stickers,
please
thank
Sean,
Drucker,
yep
and
Thank
You
Alexis.