►
From YouTube: How Spinnaker is Shaping Delivery Excellence at SAP - Serge Poueme & Mohamed Bakr, SAP
Description
How Spinnaker is Shaping Delivery Excellence at SAP - Serge Poueme & Mohamed Bakr, SAP
Serge Poueme, a Site Reliability Engineer, along with his colleague, Jessie Metivier, a User Assistance Developer, share how they’re leveraging Spinnaker at SAP. Join them as they uncover why they chose Spinnaker as their CI/CD hero, the architecture they’ve designed, and of course, how it’s allowed their company to push projects quickly to the cloud! In their talk, they will also showcase how they integrate Spinnaker with their current DevOps tools and build starterkits for internal teams — zeroing in on a team that now confidently ships code with Spinnaker!
Looking for something specific?
0:00 Introduction
0:48 Journey towards CI/CD
1:18 Why Spinnaker?
3:21 Several iterations
4:20 Company numbers
4:46 Spinnaker performance
4:57 Struggles before Spinnaker
6:30 How Spinnaker helps
8:34 Example deployment of agent operator
10:20 Spinnaker for application deployment
11:40 Deployment pipeline
14:41 How SRE supports these projects
14:58 Architectural overview (Jenkins, Spinnaker, Kubernetes)
15:58 Spinnaker landscape
18:09 Implementing Spinnaker
20:23 Self-service API
21:20 CD Starter-kit
22:52 Thank you/Contact
23:03 Q&A
SUBSCRIBE CDF CHANNEL ►► https://tinyurl.com/y4e4434n
ABOUT US
The Continuous Delivery Foundation seeks to improve the world’s capacity to deliver software with security and speed. Learn more at https://cd.foundation/
A
Hi
everyone
and
welcome
to
cdcon.
I
hope
you
are
enjoying
the
main
stage
presentations.
My
name
is
serge
prem,
I'm
a
senior
devops
engineer
within
the
site,
liability
engineering
team
at
sap
and
today,
I'm
here
to
talk
to
you
a
bit
about
our
spin
acres,
shipping
delivery
excellence
at
sap
with
my
colleague,
mohammed.
A
So
bef
before
we
dive
into
the
details
on
how
we
use
spinnaker
in
at
sap,
we
just
want
to
briefly
talk
about
our
journey
towards
cacd,
which
essentially
started
a
couple
of
years
ago
when
our
team
was
created,
and
we
were
actually
tasked
to
build
a
standout
cicd
pipeline
blueprint
for
all
the
products
that
would
be
deployed
by
the
site,
reliability,
engineering
team,
so
we
went
through
a
couple
of
solution.
Assessments
did
several
proof
of
concepts
and
finally
selected
spin
acre
as
a
continuous
delivery
platform.
A
Spinaker
is
secure,
it
offers
authentication
and
authorization
and
teams
can
adjust
it
to
meet
the
security
needs
and
spinnaker
is
also
extensible.
So
it's
very
easy
to
create
new
custom
stages,
write
plugins
in
order
to
to
achieve
new
types
of
capabilities
on
spinnaker
and
of
course,
integration
was
one
of
the
main
factors.
A
A
A
lot
of
automation.
Looking
at
adoption
challenges.
Also,
we
will
talk
about
that
a
bit
later
and
we
have
a
couple
of
great
success
stories
to
tell
so.
We
currently
serve
20,
active
customers,
so
to
say
internal
customers.
Most
of
them
are
development
teams,
and
this
is
approximately
160
developers,
206
pipelines
that
are
deploying
to
multiple
cloud
environments.
So
we're
talking
here
about
sap
cloud
platform,
aws,
microsoft,
azure
or
google
cloud
platform,
and
now
I'll.
B
Be
very
extra
careful
that
we
don't
break
anybody's
else
applications
and
we're
also
dependent
on
other
teams
for
successful
builds.
By
that
I
mean,
whenever
one
team
is
trying
to
deploy
their
application
into
production,
their
application
has
to
go
up
to
be
successful
for
our
application
to
go
to
production
as
well.
B
So
when,
if
one
team,
their
application
fails
for
some
reason,
we
are
stuck
in
our
deployment
to
production
and
that
causes
us
to
have
a
lot
of
time
wasted
by
seeking
external
help
or
other
teams
to
fix
the
issue
or
fix
the
issues
in
their
application.
So
we
there
was
a
wait
time
and
it
was
a
bit
exhausted,
exhausting
to
to
to
fix
these
issues
that
we
don't
usually
have
control
of,
and
so
by
that
we
were
very
not
and
we
needed
to
change.
B
This
is
when
we
start
using
spinnaker,
and
it
helped
us
a
lot
in
our
automation
for
the
infrastructure
deployment.
The
first
part
we
used.
We
used
this.
We
have
a
service
that
we
use
called
garner.
It's
an
open
source
sap
tool.
What
it
does
it's,
actually
an
orchestration
tool
to
deploy
a
kubernetes
cluster
across
multiple
cloud
and
providers,
and
we
were
able
like
search
team,
is
able
to
configure
that
with
the
spinnaker.
B
We
can
spin
up
our
environment
from
scratch,
with
just
a
few
clicks
to
deploy
the
infrastructure
infrastructure
components
to
the
application
as
well,
and
all
that
is
done
through
the
pipeline
with
spinnaker,
and
here
is
an
example
of
how
we're
deploying
one
agent
operator,
just
like
I
wanted
to
point
out
here.
An
example
like
is
the
yama
file
is
pretty
simple:
we
just
grab
the
reference
file
from
git
and
then
it
gets
deployed,
and
then
we
get
another
reference
file
for
the
one
asian
custom
resources.
B
B
B
B
How
big
is
the
cluster
and
what
memory
is
you're
going
to
use
and
all
that
we
can
do
that
easily
with
with
spinnaker
here
and
here,
as
you
can
see,
this
is
an
example
of
the
deployment
that
we
have,
how
we
do
get
the
config.
Then
we
with
success.
Then
we
post
the
vault
and
we
start
our
cluster
with
salinas,
which
is
the
service
for
a
search
team
that
we
use.
B
B
One
is
the
dev
one
is
stage,
and
one
is
production
for
the
dev
cluster.
We
created
two
namespaces
one
is
a
playground
and
that
one
was
requested
by
the
developers
because
they
needed
a
safe
space
where
they
can
test
and
their
in
their
applications
in
the
cluster
and
see
how
it
acts
with
different
components
there,
and
this
is
plus
the
the
development
namespace
that
we
have
in
the
same
cluster,
and
so,
whenever
developer
pushes
their
code
in
their
own
environment
in
their
own
branch.
Sorry,
it
gets.
B
B
B
B
After
passes
certain
threshold
for
each
for
each
step
until
it
reaches
100
percent,
if
all
is
clear
and
it
passes
with
no
big
variation
or
no
problems.
100
traffic
goes
to
the
new
to
the
new
ad
version
and
then
the
pod
of
the
older
version
it's
deleted.
And
then
we
send
a
push
deployment
event
to
donate,
trace
for
monitoring.
A
A
So
what
we
did
is
first
making
sure
that
our
spinnaker
landscape
is
central
to
our
service.
So
spinaker
instance
can
actually
target
all
the
different
kubernetes
clusters
of
all
the
various
customers
that
we're
working
with,
and
it
can
also
talk
to
a
vault
service
in
order
to
retrieve
secrets
that
I
use
during
deployment.
A
And,
of
course
we
have
some
monitoring
in
place.
We
are
able
to
to
to
monitor
the
the
matrix
from
spin
acre
on
primitives.
We
also
have
a
dynatrace
that
gives
us
a
lot
of
insights
on
how
spinnaker
is
doing,
and
we
have
a
log
management
system
that
allows
us
to
to
to
read
the
the
output
logs
from
the
microservices
in
real
time
and
troubleshooting
issues.
A
So
we
support
actually
dozens
of
projects
using
this
architecture
and
one
of
the
main
things
that
we
did
in
order
to
to
allow
the
deployment
of
the
spinnaker
landscape.
Let's
say
to
make
the
deployment
of
the
spinnaker
landscape
easier
is
to
create
an
automation
that
relies
on
jenkins
and
essentially
gets
a
spinnaker
configuration
from
a
git
repository
and
is
able
to
deploy
that
spinnaker
configuration
using
some
specific
steps
or
stages.
So
this
gives
the
capability
to
the
teams
to
perform
some
pull
requests
for
changes
to
spinnaker.
A
Those
pull
requests
can
be
reviewed
and
approved
when,
when
they're
approved,
there
is
a
merge
process
that
happens,
and
then
jenkins
receives
a
web
hook
that
triggers
a
pipeline
and
perform
all
the
required
options.
So
the
same
pipeline
can
be
used
to
restore
spinnaker
in
case
of
a
disaster
recovery,
but
it
can
also
be
used
for
rolling
updates,
and
this
has
essentially
allowed
us
to
to
support
a
central,
spinnaker,
landscape
or
service
very
easily.
A
A
Certain
aspects
like
the
service
accounts
that
are
used
some
post
deployment
patches,
for
example,
patching
the
spinnaker
deck
ui
service
to
make
sure
that
it
is
exposed
to
the
end
users
and
we've
been
improving.
This
landscape
configuration
over
time
to
make
sure
that
it
supports
all
the
the
aspects
of
how
we
want
to
define
and
control
the
spinach
landscape,
and
in
addition
to
this,
we
also
tackle
adoption
challenges.
A
So
we
faced
several
adoption
challenges
during
all
those
projects
and
there
are
four
specific
things
that
we
focus
on
to
make
sure
that
we're
able
to
tackle
them.
So
the
first
one
is
our
self-service
api,
so
that
api
is
a
way
for
customers
to
register
their
kubernetes
clusters,
docker
registries
or
elm
providers
to
the
central
spinnaker
configuration
without
knowing
the
details
of
elliot,
and
that
has
been
a
very
important
part
of
the
onboarding
process
to
the
spinnaker
landscape.
A
The
second
one
is
our
documentation,
so
we
write
and
share
in-depth
documentation
on
our
wiki
page.
It's
very
useful
because
it
tells
teams
how
to
use
the
service
and
also
what
are
the
constraints
and
limitations,
and
we
make
sure
that
the
documentation
is
always
updated
as
we're
going
through
changes
and
updates
on
the
spinnaker
landscape.
A
So
startup
is
a
way
for
us
to
make
the
the
development
of
pipelines
easier
and
we
maintain
both
continuous
integration,
starter
kits
and
continuous
delivery,
starter
kits
in
the
case
of
spinnaker,
all
the
spinnaker
pipelines
actually
baked
into
our
continuous
delivery
data
kits
so
teams
are
able
to
reuse
those
kits
those
pipeline
templates
and
then
quickly
build
spinnaker
deployment
pipelines
with
very
minimally
fault
and
the
the
fourth
thing
we're
doing
to
make
sure
that
we
can
increase.
Adoption
of
spinnaker
is
coaching,
so
we
provide
coaching
to
all
our
customers.
A
A
A
So
this
has
helped
improving
our
onboarding
experience,
but
also
helped
us
getting
up
the
service,
because
our
team
members
can
actually
focus
on
other
tasks
like
security,
while
onboarding
and
configuration
changes
are
handled
by
the
api,
and
this
is
also
a
very
critical
part
for
for
the
success
of
the
service,
and
here
I
just
want
to
briefly
show
you
an
example
of
a
continuously
restarted
kit.
So
it's
essentially
a
git
repository
that
sims
can
fork
and
it
comes
with
what
we
call
spin
templates
and
spin
config.
A
So
these
are
just
spinnaker
configuration
files
for
pipelines
and
spinnaker
templates
for
pipelines,
and
we
have
built
also
a
wrapper
on
top
of
our
er
and
spin
cli,
which
are
the
the
common
command
line.
Interface
is
used
to
interact
with
the
spin
acre
api
when
it
comes
to
manipulating
the
pipelines
as
code
and
teams
can
make
changes
to
the
git
repository.
A
They
can
make
some
commits
again
pull
requests,
reviews
these
changes
are
approved
and
then,
as
soon
as
they
are
approved,
there
is
a
trigger
that
starts
the
jenkins
pipeline
and
you
get
the
jenkins
pipeline.
We
call
area
or
spin
cli
to
create
the
spinnaker
application
for
the
team,
publish
the
templates
that
the
team
is
using
and
plan
and
create
all
the
pipelines.
A
So
this
has
allowed
the
teams
to
reduce
their
learning
curve
and
also
the
time
to
to
to
to
to
create
actually
the
time
required,
sorry
to
create
spinnaker
pipelines
in
production
and-
and
this
ends
our
presentation.
So
thank
you
for
listening
to
us.
I
hope
you've
enjoyed
it
and
we
are
now
opening
the
flower
for
q.
A
thanks
so
hope.
You've
enjoyed
the
session.
We
are
opening
the
floor
for
q
and
a
so
feel
free
to
to
ask
any
questions
you
may
have.
B
A
Okay,
so
faye
is
asking
how
much
change
do
the
teams
need
to
make
for
the
starter
kits
for
their
app,
so
very
little
changes.
So
the
name
of
the
the
name
of
the
pipeline
that
has
to
be
rendered
is
the
first
change
and
then
the
name
of
the
application
that
that
is
where
the
pipeline
is
being
deployed
and,
in
addition,
we're
asking
the
teams
also
to
specify
the
name
of
their
elm
repository
and
also
the
name
of
the
kubernetes
cluster.
So
these
are
the
typical
parameters
that
I
used.
A
So
sergio
is
asking
about
an
easy
way
to
install
spinnaker
so
yeah,
so
we
we
used
eliad
at
the
beginning.
It
wasn't
actually
easy,
but
we
did
we
built
that
jenkins
wrapper
on
top
of
it,
but
today
I
would
suggest
you
try
to
live
the
operator.
It's
actually
really
interesting
and
and
it's
it's
it's
easy
in
that
process.
But
at
the
same
time,
if
you
can
there
is
a
new
cli
called
klit
that
will
be
released
by
the
spinnaker
team.
A
A
B
A
So
there
was
also
a
question
about
the
starter
kit,
so
I
just
listed
an
answer
to
the
chat
for
for
transmit,
unfortunately,
not
yet,
but
we're
looking
into
it.
It's
something
that
eventually
we
will
do
so
pavel
is
asking
if
we
played
around
with
argo
cd
or
jenkins
x.
Yet
yes,
actually,
there
was
an
evaluation
paper
that
we
worked
on
for
ago
cd
and
before
selecting
spinnaker,
we
used
jenkins
x.
A
There
were
teams
asking
for
jenkins
x,
it's
actually
picking
up.
There
are
some
internal
projects
where
jenkins
x
is
actually
used,
and
argo
cd
was
also
used
for
for
several
projects.
At
the
same
time
that
we
were
building
a
spinnaker
so
that
there
is
no
limitation,
usually
in
the
environment,
actually
at
least
in
our
landscape.
A
Regarding
those
two
tools,
but
teams
are
just
trying
as
much
as
possible
to
get
a
solution
that
allows
them
to
to
start
up
very
fast.
So
we
there
is
not
too
much
traction
actually
from
now
on
those
two,
because
there
is
a
lot
of
focus
being
put
on
on
spinaker
and
jenkins,
and
also
a
lot
of
libraries
being
built
around
those
two
tools.
B
Okay,
so
I
see
another
question
about
the
canary
and
yes
it.
There
was
a
kit
provided
by
search
team
for
the
canary
and
we
did
have
to
do
some
configuration
regarding
sto.
But
this
configuration
were
done
in
the
deployments
for
the
kubernetes,
but
and
we
had
to
create
some
configuration
in
the
canary
pipeline
kinetic
configuration
in
spinnaker
itself.
B
A
Yeah,
so
we
we
also
built
a
custom
spinaker
stage
to
be
able
to
handle
vault,
and
we
started
working
on
the
volt
plug-in
that
in
the
early
days
of
the
spinnaker
plugins,
so
we
started
building
it.
It's
not
final
yet,
but
we
we're
looking
into
into
completing
it
probably
this
year
and
probably
releasing
it
also
as
an
open
source.
Plugin.
A
What
we
see
more
and
more
is
actually
teams
building
some
some
scripts
to
sometimes
when
secrets
are
dynamically
generated
to
generate
those
secrets
and
call
the
vault
api
to
inject
them
and
then
spinnaker
can
retrieve
them
in
in
in
the
deployment
stage.
One
thing
to
know
is
that
our
vault
is
internal,
so
it
is
not
exposed
to
the
internet
and
there
is
a
dedicated
link
between
between
the
spinnaker
landscape
and
bolt
hope
that
answers
the
question.