►
From YouTube: Spinnaker as a Platform: Driving Cloud Adoption in Healthcare - Joel Vasallo, TAG - The Aspen Group
Description
For more Continuous Delivery Foundation content, check out our blog: https://cd.foundation/blog/
Spinnaker as a Platform: Driving Cloud Adoption in Healthcare - Joel Vasallo, TAG - The Aspen Group
Adopting the cloud is at the forefront of many companies goals in the coming years. To truly recognize the power of the cloud most companies need to undergo a digital transformation; avoiding forklifting monoliths and adopting DevOps practices. While exciting, this also poses many risks, challenges, and key decisions; such as tooling, rearchitecting, and ultimately modernization of applications. We all want to leverage containers and Kubernetes but there should be a paved road to get there. The goal of platform engineering teams should be to build paved roads to drive adoption and empower engineers to build, deploy, and own their applications.
In this talk, we will share some strategies from TAG (The Aspen Group) on how we did this using Spinnaker including:
Basic overview of the technical challenges we encountered when we started our cloud journey
Brief overview of Spinnaker and why we chose it at TAG
How Spinnaker was used to drive adoption of the cloud and ultimately transformation to Kubernetes as the core
Considerations and solutions to meet and address compliance requirements using Spinnaker
Next phase of where the platform is going using Spinnaker as an engine
A
Awesome:
cool,
hey
everybody,
Welcome
to
the
Spinnaker,
Summit
and
pleasure
to
be
here.
My
name
is
Joel
vasalo
I'm,
going
to
be
talking
today
about
using
Spinnaker
as
a
platform
in
the
specifically
industry
that
I
work
in
today,
Healthcare
there's
my
Twitter
and
my
details.
There
I
don't
worry,
it'll,
be
at
the
end
too.
A
So
just
a
brief
note
of
what
Aspen
group
is
Aspen
group
is
essentially
made
up
of
a
bunch
of
smaller
organizations.
Most
importantly,
Aspen
Dental,
along
with
all
the
various
parties
that
we
have
in
Clear
Choice
well
now
for
Urgent
Care
and
even
things
like
vets,
so
it's
kind
of
trying
to
become
a
One-Stop
shop
if
you
will
for
healthcare,
so
it's
actually
pretty
cool
space
to
work
and
a
lot
of
great
engineering
efforts
going
on
there.
A
So
back
in
2019,
we
were
tasked
with
the
digital
transformation.
You
know
we
heard
it
about
it.
I
don't
know
about
you
at
all,
but
I
keep
seeing
it
on
LinkedIn
the
digital
transformation.
The
cloud
is
your
digital
transformation.
It's
your
highway
to
success
right,
but
let's
be
real
when
you're
starting
off.
Sometimes
it
looks
like
a
dirt
road
you're
like
okay,
so
much
for
that
highway
right.
Where
do
we
go?
How
do
we
start
this
off?
A
It's
it's
a
lot
to
learn
so
2019
a
synopsis
approvals,
lengthy
Change,
Control
boards,
hey
I,
gotta,
go
talk
to
somebody
well,
this
team
needs
to
approve
it,
a
lot
of
changes
and
a
lot
of
architectural
decisions
that
need
to
be
made
rollouts
those
approvals
only
added
time
to
your
to
relying
to
rolling
out
your
software.
It
was
time
consuming
and
oftentimes.
A
Those
changes
were
done
off
hours,
hands
up
whoever's
done
like
a
Sunday
night
deployment
like
three
in
the
morning,
because
that's
like
the
only
time,
yeah,
yeah
verification,
manual,
qioa
testing,
it's
great
but
sure
boy
doesn't
get
old
and
monitoring.
You
all
have
pig
Duty
right
like
yeah,
it's
working,
cool.
A
Our
application
architecture,
you
know
we
had
the
dreams
of
you
know.
You
hear
Microsoft
talk
about
dot
net,
core
containers,
the
cloud
I'm
like
yeah,
but
we're
still
stuck
in
a
monolith.
We
have
this
giant
Windows
app
with
you
know
it
seems
like
it
has
like
900
apis.
It
just
keeps
growing.
So
how
do
we
kind
of
like
go
from
there
to
what
our
dreams
were?
We
just
didn't
have
the
technical
expertise
at
the
time,
nor
like
the
vision
and
Direction
on
how
to
get
there.
A
A
And
we
were
one
of
them,
but
I
argue
that
a
lot
of
folks
just
took
their
Data
Center
and
tried
to
cram
it
into
the
cloud,
and
we
all
know
how
that
goes
right,
whether
it's
cost
effective
right.
The
cloud
was
never
about
being
cost
effective.
It
was
about
giving
you
capacity
on
demand
such
that
it
could
produce
the
potential
for
cost
Effectiveness
right.
A
If
I
only
have
three
users
visiting
my
website.
Do
I
need
50
servers.
No,
let's
scale
that
nice
and
down.
If
I'm,
in
an
environment
where
I
have
millions
of
users
watching
or
streaming
or
doing
something,
maybe
we
should
scale
up,
but
that
that
Ebbs
and
flows,
so
maybe
we
scale
back
down
when
that
peak
of
traffic
exists
or
goes
away,
always
remember
that
users
will
always
remember
your
bad
experience.
A
You
know,
and
potentially,
if
it's
your
first
time
using
a
service,
you
may
not
come
back
right,
especially
if
you
visit
think
of
a
restaurant.
First
time
you
go,
the
food
was
horrible.
Do
you
ever
say
like
hey?
Let's
go,
let's
go
back,
because
this
is
gonna
be
great
right.
Most
of
the
time,
you're
tepid
you're,
very
limited
to
going
back
same
thing
applies
for
the
services
that
we
write.
Ultimately,
the
websites
that
we
support.
A
So
with
all
those
things
in
mind
in
2021,
we
said
that
enough
is
enough.
We're
gonna
do
devops
we're
going
to
form
lines
that
we're
gonna,
build
it
we're
going
to
grow
it.
If
you
see
my
past
talks,
I
always
talk
about
like
the
steps
to
building
a
landscape,
and
you
know
picking
your
environment.
We
did
that.
We
have
our
source
control.
We
have
our
CI.
We
have
our
CD
tools
are
great.
Look
at
it,
terraform
I'm
sure.
If
someone's
talking
about
terraform
today
kill
FCI
I've
seen
block
I
saw
a
tweet
yesterday
about
that.
A
A
Why
am
I
going
to
give
our
most
senior
senior.net
developer
interface
to
basically
just
write
terraform
all
day
that
just
really
doesn't
build
a
good
practice
right,
like
it
also
kind
of
leaves
like
a
bad
taste,
they're
kind
of
like
battling
your
code
then
battling
terraform,
yaml
and
then
going
to
HCR
HCL
right.
So,
let's
kind
of
walk
through
what
what's
kind
of
going
on
to
kind
of
visualize
it
we
had
an
issue
of
environment
overload,
so
we'll
start
off
non-prod
prod,
most
likely
how
we
started
spoiler.
A
We
have
a
lot
of
developers,
so
they're
all
trying
to
get
their
changes
in
and
they're
always
stepping
over
one
another
and,
more
importantly,
we
have
a
QA
team
out
there
saying
what
am
I
testing
exactly
this
non-product
environment.
You
know
they're
trying
to
figure
out
whose
merge
request
made
it
in
and
how
they
got
there
and
when
should
I
cut,
you
know
task
and
ultimately
it
would
lead
to
times
where
the
deployments
are
either
slow
or
we
had
to
lock
down
environments
to
have
QA
to
have
a
solid
testing
environment.
A
So,
okay,
I've,
said
what
we
did.
We
had
terraform
right.
Another
environment
qat,
so
now
our
q18's
happy
I
think
it's
a
little
testing
environment.
They
could
lock
it
down
whatever
they
need,
developers
can
keep
developing
and
we
had
a
production
environment,
and
that
was
great
until
it
wasn't
where
a
developer
pushed
the
product
issue
and
then,
ultimately,
it
was
a
doctor's
problem
and
you
know
limited
understanding
of
how
to
debug
this
multiple
doctors
had
that
problem,
so
we're
just
pushing
fixes
and
pushing
fixes.
A
So
then
we
said
you
know
what
let's
create
another
environment,
this
time,
we'll
call
it
stage.
Yes,
it's
called
stage,
and
it's
in
production.
Don't
worry,
that's
another
story
for
another
day,
so
we
have
this
environment.
Now
we
have
one
or
two
a
select,
a
handful
of
doctors
that
we
test
and
experience
with
before
we
roll
it
out.
You
know-
or
you
could
argue,
that
this
should
be
a
cult
canary
and
that's
great
until
you
start,
you
know
their
apis
in
production,
so
other
people
start
to
consume
those
apis.
A
One
service
is
not
a
problem.
Don't
worry!
It's
not
just
two
services,
but
let's
just
assume
a
lot
of
services
are
hitting
it
and
now,
though,
internally
you're
also
causing
a
problem,
other
consumers
and
other
Dev
teams,
so
spoiler
alert,
step
UPS,
let's
create
another
environment,
and
now
we
have
this
dative
environment,
no
more
environments.
This
is
it.
We
promise
that
this
is
this.
Is
it
no
more
so
the
devops
team
kind
of
felt
like
this,
so
we're
just
laying
the
road
in
front
of
us
as
developers
are
adding
new
features.
A
A
So
at
this
point
we
had
to
recognize
all
the
hard
work.
We
did
look
the
tools
that
we
had
and
everything
that
it
worked.
It
totally
worked.
We
were
building
infrastructure,
people
were
able
to
deploy,
but
you
always
have
to
ask
what
else
can
we
do
to
get
better
right?
So
we
had
GCE,
VMS
and
windows.
We
had
Cloud,
run
Cloud
functions,
containers
and
serverless.
A
We
had.
We
even
had
support
for
cos
container
optimized
images
in
Google
Cloud.
We
still
were
in
a
position
where
developers
are
figuring
out
the
implementation
details
to
scale
your
app
go
talk
to
devops.
They
controlled
your
infrastructure.
You
wanted
to
open
another
Port
devops
is
the
one
in
charge.
Iam
our
security
teams,
along
with
devops,
had
to
figure
out
what
policies
to
add
and
a
lot
of
times
those
were
handled
behind
human
requests.
Hey
my
app
needs
this.
A
Our
technology
and
our
deployments,
especially
with
tools
like
Spinnaker
nowadays,
are
deploying
hundreds
of
times
a
day.
So
there's
really
no
time
for
a
human
delay.
Debugging
Legacy,
Windows
debugging
was
painful
if
I
have
to
install
like
perfmon
or
like
the
power
tools
on
Windows.
One
more
time
I'm
going
to
scream
inconsistency.
A
We
didn't
have
that
developer
experience
that
we
talked
about
earlier,
so
Dev
teams
were
essentially
making
their
own
Pipelines
and
importantly,
since
it
lacked
integration
teams,
like
our
data
engineering
team,
ended
up
building
their
own
experience
on
how
to
database
migrations.
So,
instead
of
having
a
one
seamless
experience
for
our
developers,
they
had
to
to
play
their
app
migrate.
Their
database
redeploy
their
app.
It
was
essentially
hopping
around
different
teams
and,
more
importantly,
different
pipelines.
It
was
confusing.
A
We
took
a
strong
investment
in
building
a
delivery
engineering
team,
so
this
delivery
engineering
team
is
connecting
those
dots
of
those
tools
connecting
those
tools
they
don't
necessarily
piece.
Well,
together
we
have
terraform,
we
have
gitlab,
we
have
Spinnaker,
we
have
bug
snag.
How
do
we
build
a
nice
unified
story?
So
a
lot
of
the
tools
are
built
together
in
code.
A
How
do
we
standardize
the
data
that
flows
between
these
systems?
How
do
we
create
a
reliable
metadata
that
goes
throughout
from
a
commit
all
the
way
to
your
deployment
and,
ultimately,
like
I
said
earlier?
How
do
you
work
alongside
devops
transforming
That
Rock
Solid
awesome
infrastructure
to
a
developer,
facing
tool
to
build
a
good
developer
experience?
A
At
the
end
of
the
day,
both
teams
are
charged
with
building
compelling
solutions
that
are
simple,
minimize
contact,
switching
and
provide
accurate
and
fast
feedback.
We
always
ask
ourselves:
will
this
Empower
our
Engineers
to
take
action,
fix
the
problem
or
fix
the
quality
of
life
of
their
release
process.
A
There's
a
lot
of
things
that
go
into
building
deliverable
apps,
first
off,
not
to
be
corny,
but
it
does
require
teamwork
right
I
mentioned
earlier,
our
app
was.net
too.
That's
like
when
I
graduated
college,
like
that's
Jesus,
like
it's
old,
so
and
it's
also
running
Legacy
Windows
environment.
So
how
do
we
get
there?
It
was
a
lot
of
migration
and
a
lot
of
pain
and
effort,
and
also
the
data
basis.
Migrations
had
to
kind
of
go
alongside
with
that,
so
teamwork,
it's
good.
A
It's
good,
be
nice
to
your
co-workers,
culture
of
empowerment,
I
always
debated
using
this
image
because
I'm
tired
of
it
but
developers
developers
developers.
What
are
you
doing
to
shift
left?
It's
easy
to
say
you
shift
left,
but
what
does
that
mean
in
your
organization?
What
are
you
giving
your
developers
to
not
only
build
the
right
tools
but
also
help
them?
We
don't
just
give
our
developers
yaml
I
mean
you
can,
but
that's
how
you
cause
outages
right,
because
we
all
make
mistakes.
A
We
want
to
provide
them
guard
rails,
not
Gates,
and
that's
where
you
transform
your
existing
teams.
Like
your
networking
team,
your
devops
team,
your
security
team,
as
subject
matter
experts,
hey
I,
have
to
make
this
app
available
on
the
internet
for
the
mobile
endpoints
help
me
get
there.
Networking
could
sit
down
talk
about
the
Ingress
controllers
and
what
you
need
to
do.
Security
can
verify
your
app.
Do.
Some
pen
tests
make
sure
you're
deploying
stuff
to
secure
stuff?
A
How
do
you
connect
those
dots
and
B
points
of
reference
for
your
developers
to
find
a
simple
goal?
For
us?
It
was
easy
click
of
a
button.
We
want
to
click
a
button
and
Magic
should
happen.
Obviously,
there's
a
lot
of
work
that
goes
behind
that
click
but,
as
we
know,
make
it
simple
and
ultimately
centralize
and
build
around
a
common
delivery
platform
and,
in
our
case
spoiler
alert.
It
was
Spinnaker.
A
So
I
always
like
to
ask
this
for
frame
of
reference.
How
many
folks
In
This
Crowd
are
new
to
Spinnaker,
okay
cool,
a
lot
of
people
know
it
so
that's
good,
but
regardless,
whether
you're
watching
on
the
stream.
If
you're
new
welcome
and
we're
happy
to
have
you
here,
it's
a
great
community
and
thank
you
for
taking
time
of
your
day
and
watching
this
talk
so
a
little
bit
about
Spinnaker,
let's
create
a
Netflix
and
it
was
Battle
tested
by
many.
A
It's
an
open,
active
and
growing
Community.
As
you
can
see
here
today,
we
have
a
lot
of
great
companies
using
it
and
growing
the
platform
from
where
I
once
was,
and
it's
it's
growing.
But
more
importantly,
there's
still
like
it's
been.
The
beauty
of
Spinnaker
is
that
it
supports
not
only
from
a
community
perspective,
but
it
also
has
like
partner
engagement.
So
if
you
need
that
Professional
Services
love
and
attention,
we
have
folks
like
Armory
and
optimax
who
are
graciously
here
today.
A
It
supports
multi-cloud,
so
it
supports
Google,
Cloud,
kubernetes,
AWS
and
other
Opera
and
other
providers.
It
enables
you
to
create
automated
release
processes,
leveraging
CI
CD
event
triggers
in
your
pipelines.
There's
a
ton
of
them
I'm
not
going
to
go
into
that
today
and
ultimately,
now
that
it's
backed
by
The
Continuous
delivery,
Foundation,
we
kind
of
have
a
little
bit
more
backing
in
terms
of
a
bigger
community.
So
it's
great
to
see
the
project
go
from.
A
You
know,
essentially
an
open
source
to
read
by
Netflix
to
you
know
where
it
is
today
very
impressive
for
us.
It
made
sense.
We
looked
at
our
Cloud
landscape,
we
were
ec2,
Lam
does
some
GCE
well
majority
GCE
and
we
wanted
to
go
towards
kubernetes
right
and
it
allowed
us
to
do
that.
This
is
just
kind
of
a
sample,
but
in
this
environment
we
have
our
ec2
instances
on
the
bottom
and
right
on
top
that
same
deployment
going
to
our
kubernetes
cluster.
A
We
found
that
the
pipelines
allowed
you
to
create
a
visual,
simple
step
for
each
item
in
that
pipeline,
that
performed
complex
operations
But.
Ultimately,
the
pipelines
were
just
Json
pipeline
templates
and
stages.
So
what
that
meant
is
that,
after
a
while
of
kind
of
going
in
the
GUI
and
clicking
and
dragging-
and
you
learned
everything
and
you
had
your
private,
your
ideas,
you
could
take
those
and
basically
build
templates
templates
out
of
your
pipelines.
A
lot
of
companies
do
this.
A
A
You
know,
Spinnaker
always
is
talked
about
as
a
software
delivery
tool
which
at
its
core
it
is
but
I
think
the
biggest
thing
is
that
it
allows
you
to
lay
out
your
business
process,
especially
for
people
who
may
not
be
Tech
experts,
or
rather
learning
this
new
world
right.
They
may
not
understand
how
you're
triggering
their
tools,
but
if
you
could
show
them
in
one
view
where
they
sit
in
that
flow,
it
gives
them
confidence
that
you're
not
working
around
them
or
you're
screwing
the
details
right.
A
That's
that's
at
least
how
I
always
think
about
it,
but
even
with
sin
that
you
gotta
remember,
there's
a
lot
of
it's
not
just
deploy
my
container
there's
a
lot
of
strategies
behind
it,
whether
you're
going
into
blue
green
you're,
doing
Highlander,
where
you
just
don't
care
you
just
want
to
roll
forward.
You
want
to
do
a
weight-based
rollout.
Do
you
want
to
do
a
canary
based
rollout,
there's
a
lot
of
things
you
have
to
consider
and
does
it
support
all
the
things
that
we
want.
A
Many
companies
may
not
need
this,
and
or
it
may
change
on
environment,
but
can
I
do
an
approval,
simple
things
like
this
give
people
context,
especially
when
you're
looking
at
compliance,
which
is
something
in
the
healthcare
industry.
We
have
to
do
tracking
that
someone
authorizes
change
and
keeping
a
record
of
that
that
record
that
that
that
approval
propagates
through
the
pipelines.
So
you
know
who
clicked
the
OK
button.
A
What
about
integrating
quality
and
testing
you
know,
I,
said
earlier:
our
testing
and
quality
was
manual
at
the
time,
but
now
they're
investigating
in
test
Suites,
there's
a
whole
slew
of
things
that
you
can
do
in
that
world,
whether
you're
doing
something
as
simple
as
Postman
or
using
an
off-the-shelf
tool
like
Locust.
But
there
are
some
simple
things
you
can
do
to
do
some
performance
testing
as
well
as
API
testing
or,
ultimately
writing
your
own
framework,
which
is
a
big
Endeavor,
and
it
should
always
be
taken
with
some
thought.
A
Is
there
integrating
database
migrations
right,
I've
seen
many
organizations
do
things
like
liquid-based
migrations
before
they
deploy
I
talked
about
it
earlier,
our
database
team
built
their
own
pipeline.
What
about
bringing
them
into
the
fold?
It
may
not
always
run,
but
it's
there
and
it
may
not
even
be
needed
in
every
pipeline.
Does
a
UI
pipeline
need
a
database
migration,
no
Change,
Control
right,
I
talked
about
it.
A
You
know,
I'm
not
discrediting
it,
but
I
think
it
has
to
be
evolved
into
the
world
that
we're
deploying
fast
so
out
of
all
these
operations
at
the
end
of
it.
What's
the
harm
in
sending
all
that
data
somewhere
in
our
case,
it's
a
simple
jira
ticket,
it's
an
API
for
it.
If
you
want
to
be
nice,
write
an
API
and
decorate
it
and
put
some
nice
little
colors
and
links
out
to
various
providers.
A
It's
always
you
know,
because
again
that
flow
of
metadata
and
then
ultimately
like
what
about
your
SRE
teams
right,
the
dashboards
and
metrics
are
standardized.
They
talked
about
that
earlier,
try
and
keep
things
as
homogeneous
across
their
applications
as
possible.
But
if
you
truly
are
admitting
kind
of
the
same
metrics
word.net6
we're
running
in
kubernetes
most
of
the
time,
ninety
percent
of
your
metrics
are
probably
the
same.
What's
the
harm
in
just
giving
them
a
basic
dashboard,
hey
whether
you
use
it
or
not.
Here
you
go.
Here's
a
free,
grafana,
dashboard.
A
A
Well,
has
anyone
tried
so
in
addition
to
that,
it
was
also
confusing
right.
Does
anyone
remember
like
the
first
time,
like
you
know,
I'm
sure
everyone
in
the
crowd's,
probably
you
know,
done
their
tutorials
and
stuff,
but
remember
that
first
time
you
made
a
kubernetes
ammo
and
you're
like
looking
at
it
you're
like
what
the
hell
is
going
on
here,
all
right.
A
Think
of
that
moment,
in
a
in
in
a
world
where
you
went
from
building
infrastructure
manually
or
creating
an
intera
form
to
translate
in
that
right,
how
do
I
even
make
a
DNS
record
for
a
kubernetes
service
right
spoiler,
we
took
a
holistic
View
at
kubernetes
as
to
what
it
could
automate
for
us
and
for
us
we
saw
DNS.
It
could
be
automated
with
something
like
external
DNS,
those
certificates
that
we
were
chasing
either
GoDaddy
or
some
random
cert
provider
that
we
had
and
had
to
go
email
and
go
buy.
A
Maybe
we
can
start
manager
run
our
own
certificate
store,
Secrets
management,
because
we
all
know
that
kubernetes
does
not
encrypt
your
secrets
by
default.
It
encodes
so
make
sure
you
encrypt
the
secrets.
Somehow,
how
do
you
load
balance
and
routes?
Well,
nginx
has
an
Ingress
controller.
You
can
use
this
CL
as
a
service
mesh
or
others.
That's
a
great
one
since
I
just
put
SEO,
because
that's
the
one
I'm
used
to
it
doesn't
matter.
There's
a
lot
of
great
offerings
out
there.
A
At
the
end
of
the
day
we
started
small
I,
always
like
to
say:
don't
let
perfect
be
there
to
be
good.
I
didn't
say
that
that's
Voltaire,
but
that's
a
good
phrase
to
keep
in
mind
constantly
improve.
Let's
take
the
case
of
that
app
earlier.
We
knew
where
we
wanted
to
go
microservices
we
hear
about
it.net
python,
kubernetes,
that's
a
dream,
but
sometimes
like
a
bad
nightmare.
The
monolith
.net
Windows.
How
do
we
tackle
this
right?
So
do
we
just
start
trying
to
wait
for
this
to
be
done,
or
do
we
figure
something
small?
A
A
You
know
I'm,
not
the
ideal
State,
because
at
the
end
of
the
day,
you
should
be
using
static
hosting
for
your
static
websites,
but
it
gave
us
a
simple
way
of
proving
out
that.
Can
we
create
a
Spinnaker
pipeline?
Can
we
develop
that
yaml
to
deploy,
because
if
we
keep
a
agnostic
approach
as
to
what
we're
deploying
where
we're
not
building.net
specific
things
into
the
pipeline,
we
can
then
service
any
language
that
we
have
in
the
company.
A
Going
back
on
that,
the
biggest
thing,
though,
here
honestly,
was
that
it
won
support
from
our
business
too,
because
a
lot
of
folks
get
scared
that
you're
moving
infrastructure
and
they
don't
see
they
saw
that
how
long
deployments
took
before.
But
then
business
users
were
able
to
go
into
meetings,
saying
hey.
How
long
will
it
take
to
add
the
logo
and
by
the
end
of
the
meeting
you
were
able
to
add
that
logo
people
see
the
value
and
that's
a
solid
win
for
the
business
or
the
folks
on
the
business
side.
A
So
at
the
time
we
kind
of
came
back
to
our
developers,
saying
hey.
How
are
you
doing
on
that?
Linux
migrationfrom.net
I
I
had
to
generate
this
GIF,
because
somehow
they
moved
the
monolith
once
I
think
they
saw
kubernetes
and
how
fast
the
web
team
was
going.
They
somehow
moved
the.net
app
in
like
three
months
like
I'm,
like
holy
crud,
like
what?
How
much
you
guys
are
sleeping
right.
You
folks
are
sleeping.
It
was.
It
was
an
impressive
migration,
but
we
essentially
enabled
the
highway
to
migrate.
Your
apps.
A
We
had
our
monolith
running
in
the
cluster,
but
then
we
also
built
a
path
for
them
to
migrate
out
their
microservices.
Hey,
slash,
patient
no
longer
goes
to
the
monolith,
build
an
Ingress
route
and
go
to
that
patient
microservice
leveraging
something
like
istio.
We
were
able
to
Canary
critical
components
of
our
application
in
our
payment
API,
which
ultimately
makes
how
much
how
we
get
paid.
So
we
want
to
make
sure
that
that
works,
and
then
that
could
extend
past
that
that
model
can
extend
past
to
all
your
various
sub-components
as
well.
A
But
again,
at
the
end
of
the
day,
where
we're
at
we're
not
just
going
to
give
developers,
yaml
I
like
this
quote
that
one
of
my
team
members
said
is
I
could
see
how
kubernetes
makes
onboarding
super
simple
because
he
saw
it
you
create
it.
It
creates
a
DNS
structure,
creates
a
certificate
but
then
again
type
will
suck
and
kubernetes
is
super
unforgiving.
You
know
nothing
stinks
more
than
having
to
debug
a
a
line,
38
yaml
and
then
you
find
out.
A
A
A
This
is
how
you
deploy
your
code
and,
more
importantly,
if
we
can
automate
that
we're
looking
at
building
a
growth
on
a
tool
that
we
open
source,
that
we've
been
a
big
open
source
contributor
in
formats,
but
essentially
provider
developer,
safe
defaults
with
customizable
templates
and
an
enforced
overrides
developers
can
tell
us
what
they
want
and
we
build
those
pipelines
and
that
infrastructure
to
their
spec
and
if
it
doesn't
allow
it,
we
can
then
have
a
conversation
like,
for
example.
Maybe
we
restrict
how
many
public
websites
you
can
have
right.
A
We
want
to
look
at
things
like
service
mesh
to
enable
a
canary
header
right.
Maybe
traffic
goes
to
prod
and
if
you
have
some
specific
header,
you
go
to
another
cluster
or
to
another
environment,
a
deployment
inside
the
cluster.
So
something
like
service
mesh
again
service
mesh
doesn't
have
to
be
SEO
just
use
what
you're
used
to
or
what
you
would
like
to
invest
your
time
in
leverage
and
grow.
A
Canaries
I
mentioned
it
earlier,
but
canaries
are
no
longer
an
environment
they're
a
feature
of
your
environments,
so
you
can
have
a
stage
environment,
a
development
environment
with
canaries
in
Flight.
You
know
think
of
the
world
where
a
developer
is
building
a
piece
of
software,
they're
iterating,
the
iterating,
the
iterating
and
then
when
they
merge
to
Mainline.
A
We
want
to
visualize
our
performance,
so
this
is
kind
of
some
live
traffic
from
our
deployment,
but
we
want
to
give
those
tools
to
our
developers,
so
they
can
see
whether
deployments
are
working
or
not.
This
is
a
tool
called
chiali,
but
it's
surface
through
tool,
metrics
in
Prometheus
and
open
Telemetry
and
then
ultimately
giving
them
the
grafana
dashboard
that
I
talked
about
earlier
again.
Shifting
left
doesn't
mean
you
just
put
the
onus
of
operating
the
tool
to
your
developers.
A
You
give
them
the
tools
to
make
those
calls
to
set
those
alerts
to
essentially
debug
their
application.
Give
you
as
much
context
when
they
truly
call
you
they
really
need.
You
hear
it
a
lot,
but
we
want
to
look
at
maybe
simplifying
a
lot
of
our
internal
developer
internal
developer
portal.
Like
today,
you
know
you
create
a
repo.
Then
you
got
to
go,
find
how
to
do
your
code
if
you're
good
you'll
need
that
Ci
or
what
CI
steps
do
I
need.
A
Maybe
it
provides
you
some
suggestions
and
ultimately
your
repo
is
created
for
you,
your
CI
CD
is
created
for
you.
More
importantly,
all
those
things
I've
talked
about
your
health
and
performance
are
linked
to
there,
whether
they're,
integrated
or
not.
At
least
you
get
helpful
links
where
you
can
just
have
one
experience
to
see
the
life
cycle
of
your
entire
application,
and
if
you
want
to
be
nice,
even
provide
things
like
Docs.
A
So
a
lot
of
talking,
let
me
just
put
it
all
together.
The
key
takeaways
first
off
is
understand
where
you
are,
how
you
got
there
and
where
you
want
to
go
so.
I
talked
a
lot
about
like
the
road
map.
As
to
you
know
the
manual
to
like
kind
of
the
devops
to
like
now,
the
future
of
the
delivery
engineering,
slash
devops
tools
at
the
end
of
the
day
should
be
unified
to
build
an
experience
right,
the
tool
they
shouldn't,
be
they
shouldn't
Define.
A
Your
experience
build
a
culture
of
education
and
empowerment
for
your
engineers.
Again,
I
really
strongly
feel
that
your
Tech
teams
become
subject
matter.
Experts
to
what
you're
talking
about
shifting
left,
we
say,
shift
left
developers
if
you
don't
just
give
them
the
tools
without
any
help,
provide
those
developers
and
Engineers
guardrails
not
Gates,
give
them
the
safe
default.
At
the
very
least
you
can
get
into
the
production
and
then
you
can
Cascade.
A
Maybe
you
can't
deploy
a
public
app,
but
at
least
you
can
get
into
it
as
an
internal
app,
and
then
we
can
have
that
discussion
as
to
how
to
expose
it,
make
your
pipelines,
bring
complex
processes
and
teams
together
again
showing
how
all
those
dots
had
different
teams
ownership
behind
it.
It
builds
a
conversational
flow
as
to
when
that
dot
turns
red.
They
can
go
talk
to
the
security
team.
It
doesn't
always
fall
on
your
devops
team,
build
and
Empower
those
teams
as
well
to
become
more
and
hopefully
out
of
this.
A
You
learn
that
Spinnaker
is
awesome,
so
shout
out
we
do
have
a
tech
blog
where
I
talked
a
little
bit
about
how
to
get
started
with
Spinnaker
operator
on
GCE
or
GK
and
IEP.
That's
a
pretty
big
article
again.
They
mentioned
earlier
just
to
talk
about
how
you
install
Spinnaker.
So
this
is
how
we
install
it.
So
if
you're
interested
I'll
check
that
out,
but
regardless
thank
you
so
much
for
having
me,
it
was
truly
great
to
be
back
after
a
long
time.
So
thank
you.
Everybody.