►
From YouTube: GitOps and Spinnaker - a DevOps Style
Description
Srinivas Kambhampati, Lead Customer Success at OpsMx, explains GitOps and how Spinnaker can drive the DevOps process around a GitOps model.
A
A
A
Time
as
well
as
our
presenters,
you
have
found
yourself
in
the
cd
foundation's
online
meetup.
We
have
these
monthly,
I'm
so
excited
to
have
you
here
today
I
am
tracy
reagan.
I
am
the
ceo
and
co-founder
of
deploy
hub
and
also
the
community
director
for
an
open
source
project
called
ortelius
that
is
being
incubated
at
the
cd
foundation.
A
We,
like,
I
said
we
do
these
monthly
last
month
with
cdcon.
I
hope
you
guys
all
attended
it.
It
was
pretty
awesome.
I
thought
it
was
a
really
amazing
conference.
Actually,
I
think
that
all
the
speakers
were
spot
on
and
everything
that
the
c
foundation
did
for
building
out
a
true,
diverse
community
around
the
discussion
of
continuous
delivery
was
they
did
a
great
job?
Maybe
next
year
we'll
be
doing
it
in
person,
but
I
do
hope
they
keep
it
a
hybrid
event,
so
everybody
can
be
participating
next
month
is
august.
A
We
are
going
to
take
august
off
again,
so
we
had
june
off,
but
we
are
going
to
take
august
off
it's
a
summer
month
and
everybody's
pretty
busy
and
we're
going
to
kick
off
september,
I'm
hoping
to
have
waleena
velasquez.
I
think
I'm
not
sure
how
to
say
her
last
name,
she
is
one
of
the
primary
contributors
of
the
jenkins
project.
I
think
she's
on
the
jenkins
governance
board
and
I'm
going
to
have
her
speak
on
how
to
get
involved
in
an
open
source
community.
A
The
cd
foundation
itself
is
an
open
source
community.
We
have
special
interest
groups
that
you
can
get
involved
in.
We
have
spinnaker
as
an
open
source
project
jenkins,
ortilius
screwdriver,
we're
looking
at
bringing
on
a
few
more
in
the
upcoming
year,
so
getting
involved
and
giving
back
to
the
community
that
you
are
in
is
kind
of
an
important
piece
of
this
community.
A
A
World
of
this
kind
of
operator
kind
of
format,
with
infrastructure,
as
code
and
learning
about
you
know,
operators
day
one
and
from
day
one
to
day
in
operators
to
spinning
up
infrastructure
environments
that
has
started
to
come
into
even
at
the
application
level.
So
we're
starting
to
have
more
of
a
discussion
around
get
ops
and
I'm
so
happy
today.
A
That
srini
was
willing
to
chat
with
us
about
get
ops
and
spinnaker
and
how
those
two
are
going
to
be
working
together
and
what
that
what
they
call
that
style
of
devops
will
look
like
shirini
I'm
going
to.
Let
you
introduce
yourself,
but
he
is
a
devops
architect,
and
I
know
there
are
other
others
of
you
who
have
that
title
and
you
know
how
hard
it
is
right
now
to
sort
of
keep
track
with
where
we're
going
with
devops.
A
So
this
is
a
great
opportunity
for
you
to
ask
questions
even
even
bring
up
different.
You
know
topics
around
get
ups
and
and
and
get
ops
with
devops,
because
this
is
something
that
the
option
x
team
are
really
really
focused
on
and
thinking
about.
So
please
don't
hesitate
to
ask
questions,
and
there
is
you
can
just
post
them
in
the
chat
and
I'll
keep
track
of
them.
You
can
also
post
them
in
the
in
the
q
a
that
works
as
well.
B
B
So
we
are
dedicated
to
making
the
process
of
continuous
delivery
easy
and
successful
for
everyone.
We
have
a
large
number
of
customers
of
every
size
from
the
most
sophisticated,
such
as
salesforce
cisco
to
organizations
who
are
now
highly
successful,
even
though
they
had
little
or
no
cd
experience
when
they
started
with
us.
B
So
let
me
dive
straight
into
what
is
githubs
so
just
want
to
start
with
a
quick
story
I
have
here
at
home.
I
have
taken
two,
our
laptops,
one
old
and
one
relatively
new
and
made
a
small
kubernetes
cluster
out
of
them.
This
old
laptop
is
kind
of
slow,
but
it's
okay.
I
made
it
the
master
node
and
the
other
newer
one
with
relatively
higher
memory
is
my
worker
node.
B
So
it's
one
master
one
node
kubernetes
system
and-
and
the
thing
is
that
this
laptop
with
higher
memory,
has
a
strange,
interesting
feature
that
it
keeps
crashing
randomly
it
crashes
once
I
think,
a
couple
of
days
or
something
with
this
normally
the
laptop
would
not
be
very
useful,
but
guess
what
in
the
kubernetes
model,
I'm
able
to
use
it
just
fine?
I
have
deployed
applications
and
I
do
development
on
that
in
the
kubernetes
cluster.
B
B
So
what
happens
is
whenever
the
the
laptop
crashes
and
I
simply
reboot
it-
the
control
plane,
the
kubernetes
control
plane,
has
a
single
loop
sort
of
thing
that
keeps
going
and
checking
if
my
target
environment
is
in
the
state
that
I
want
it
to
be
and
whenever
it
deviates
from
that,
it
simply
brings
it
back
up
so
in
in
a
very
simple
model.
This
is
what
git
ops
is.
B
It
has
version
control
built
in
pr
approval
process
built
in
as
a
single
source
of
truth
gives
you,
the
ability
to
roll
back
to
a
previous
unknown
state,
pinpoint
the
cause
of
failures
quickly,
deploy
new
services
and
so
on.
Everything
required
to
run
an
application,
including
the
infrastructure
configuration
code.
Everything
should
be
version
controlled,
and
this
is
really
not
a
new
concept,
as
tracy
pointed
out
infrastructure,
as
code
has
been
there
for
some
time
manufacturing
control
process.
B
B
So
there
is
an
inherent
feedback
loop.
So,
as
I
mentioned
in
in
kubernetes,
the
control
plane
components
keep
pulling
the
cube,
ctrl
cube,
ctrl
response
by
telling
the
states
of
the
part
and
so
on
so
forth,
and
they
initiate
actions
and
events
that
will
cause
the
cluster
state
to
come
back
to
where
it
wherever
it
was.
Even
if
you
as
much
as
reboot
the
node,
and
so
we
have
version
control,
we
have
automation.
B
I
guess
at
a
very
simple
level-
and
I
have
been
around
in
in
in
devops
before
devops
existed,
we
used
to
call
it
system
integration,
we
used
to
call
it
system,
build
or
release
engineering,
and
so
on.
There's
so
many
names
before
devops
came
in.
The
answer
is
actually
very
simple:
whatever
is
designed
should
be
developed,
whatever
is
developed
should
be
tested
and
whatever
is
tested
should
be
deployed
very,
very
simple.
B
While
the
answer
is
simple,
the
implementation
gets
rather
tricky,
and
this
is
trickier
when
you
have
multiple
parties
involved.
There
are
different
teams,
there's
a
development
team,
the
testing
team
and
there's
another
set
of
people
who
approve
it
different
environments,
different
criteria
for
each
of
the
environments.
B
Typically,
production
environments
will
have
a
totally
different
criteria
from
other
environments,
so
gitops
as
a
paradigm
in
a
way
is
the
enabler
enabler
of
the
shift
left
that
we
talk
about
where
we
are
ensuring
that
the
system
reflects
the
changes
and
stays
in
sync
with
the
desired
state.
At
a
very
simple
level,.
B
B
B
All
right,
thank
you,
so
having
a
single
source
of
truth
for
infrastructure
and
applications
also
ensures
that
all
the
parties
are
collaborating.
So
this
is,
it
definitely
promotes
collaboration
and
every
dialogue,
every
change.
Every
approval
is
automatically
captured
in
git,
and
this
supports
full
auditability
of
the
changes
and
in
customer
environments,
where
we,
such
as
banking,
finance
and
all
where
auditability
is
a
critical
requirement.
B
It's
ideally
suited
for
environments
that
are
constantly
changing
and
when
you
deploy
faster
and
more
often,
the
changes
are
not
just
in
the
code,
but
sometimes
they
are
also
in
the
configuration,
the
config
maps,
the
secrets
and
so
on
so
forth,
and
this
adds
to
the
overall
dynamic
to
the
change.
That's
happening.
B
So
it
is
possible
to
actually
take
the
configuration
of
spinnaker
and
kind
of
apply
it
to
a
running
spinner.
For
instance,
we
could
take
pipelines
and
apply
it
so
equivalent
of
of
simply
take
this
pipeline
and
make
it
happen
and
spinnaker
is
really
well
suited.
For
that,
and
the
best
part
is
that
spinnaker
also
plugs
into
the
infrastructure.
B
B
So,
as
you
can
see,
when
you
talk
about
githubs,
typically,
there
is
already
a
source
code.
Repository
developers
are
comfortable
using
it.
It's
a
it's
used
across
the
landscape
in
the
in
in
organizations
and
whenever
the
push
happens,
the
pipeline
triggers
and
automatically
triggers,
let's
say
a
jenkins
build
and
the
jenkins
build
might
also
result
in
some
automated
testing
and
then
finally,
it
goes
to
deployment
and
then
it
gets
deployed
on
on
the
cluster.
B
The
example
we
have
given
in
the
kubernetes
cluster
and
spinnaker
will
go
ahead
and
deploy
it
in
the
kubernetes
system,
and
this
typically,
this
entire
pipeline
is
orchestrated
by
spinnaker.
Spinnaker
has
excellent
pipeline
orchestration
system
in
place
and
it
plugs
into
all
these.
You
know
other
players
in
the
system.
A
B
Right
so,
let's
take
a
quick
look
as
to
how
this
model
can
be
further.
Let
me
dive
deeper
into
it,
how
we
are
doing
it
with
our
customers,
how
it's
done
at
ops,
we
store
the
spinnaker
configuration
in
a
git
triple,
as
I
said,
it
need
not
be
git,
it
could
be
s3
or
any
other
types
of
repositories,
and
in
spinnaker
there
are
multiple.
B
It
also
looks
at
secrets
and
injects
them
into
these
files,
so
that
halyard
can
use
those
secrets
and
configure
them
secrets
could
be
account,
passwords,
git
tokens
and
and
such
so.
This
is
because,
typically
we
don't
want
to
have
secret
stored
in
git.
It
may
have
access
to
many
other
people
who
we
basically
don't
want
to
give.
B
Let's
say
passwords
to
our
database
and
so
on
so
forth,
so
the
secrets
typically
might
come
from
another
source,
let's
say
a
vault,
back-end
or
ecr,
or
you
know
there
are
so
many
secret
registries,
secret
stores
from
cloud
providers
and
those
secrets
are
injected
when
we
put
that
in,
and
customers
typically
like
to
have
a
separate
spinner
car
for
production
than
for
rest
of
the
environments.
So
we
could
use
the
same
git
repo
and
have
different
branches
for
different
instances
of
spinnaker.
B
If
you
have
and
then
we
also
have
another
repo,
it
could
be
the
same
ripple
for
the
for
the
purpose
of
depiction.
I
have
shown
it
has
a
different
ripple
which
stores
the
pipelines
and
spinnaker.
As
I
said,
a
pipeline
is
nothing
but
a
json
file
very
convenient.
It's
a
text
file,
it
can
be
checked
in
it
can
be
pushed
to
the
git.
B
So
what
it
does
is
that
it
also
gives
us
an
opportunity
to
do
some
minor
editing
of
the
pipelines.
For
example,
we
could
ensure
that
the
pipelines
are
not
editable
in
the
ui.
We
could
change
the
account
names
or
something
for
different
environments,
and
so
on
so
forth,
so
some
editing
or
modification
to
the
pipelines
is
possible
as
the
pipelines
are
being
pulled
from
the
git
and
applied
into
the
into
the
system,
because
spinnaker
has
built-in
triggers
for
git
and,
of
course,
many
other
sources.
B
So
whenever
we
make
changes
to,
let's
say
a
pipeline
in
a
git
repo,
it
could
be
automatically
be
applied.
So
if
you
configure
the
spinnaker
configuration
changes,
the
configuration
changes
could
be
many.
Sometimes
we
have
to
increase
the
memory
or
the
cpu
limits
of
different
spinnaker
pods.
Maybe
we
want
to
add
additional
accounts.
B
So
this
gives
us
the
ability
to
make
sure
that
we
have
a
central
source
of
truth.
Where
everything
is
stored,
everybody
can
look
at
it.
Different
parties
can
collaborate,
and
yet
the
secrets
are
not
there,
people,
we
don't
have
to
express,
expose
the
secrets
and
the
secrets
may
actually
be
different
for
different
environments.
B
So
a
production
spinnaker,
for
example,
may
have
a
different
set
of
secrets
and
they
get
automatically
applied
with
the
same
configuration,
and
this
is
the
model
where
we
are
also
using
promoting
of
pipelines
and
what
is
pipeline
promotion
whenever
we
want
to
make
a
change
in
a
pipeline
that
is
deploying
to
production.
Let's
say:
obviously
we
don't
want
to
try
it
out
there.
B
B
A
B
Uncontrolled
changes
to
environments
on
somebody
going
and
making
some
product
change
to
a
production,
server
and
so
on,
needs
to
be
prevented
and
organizations
are
looking
at
how
to
do
that.
How
do
we
prevent
uncontrolled
unauthorized
changes
to
the
production,
environment
and
gitops
is
a
step
in
that
direction?
B
B
So
in
in
good
old
days,
there
were
very
few
people
who
knew
about.
Databases
knew
how
to
extract
data
using
sqls
and
so
on.
No
ui-
and
you
know
the
rest
of
the
organization
is
dependent
on
this.
So
we
have
somewhere
there,
maybe
not
at
such
an
early
stage,
but
the
fact
still
remains
that
we
don't
have
all
the
tools
needed
to
make
the
gitops
work.
B
The
you
know
single
source
of
truth,
but
we
still
need
a
lot
of
processes,
the
equivalent
of
the
kubernetes
control
plane,
if
I
may
say
so,
because
the
processes
are
different,
let's
say
for
production,
the
configuration
may
be
different
and
so
on
so
forth
and
as
I
said,
deployment
yaml
is
somewhere.
There's
a
template
and
the
template
is,
gets,
you
know,
configuration
substituted.
B
B
This
does
take
a
lot
of
tools
and
even
if
you
see
something
like
terraform,
that's
infrastructure
as
code,
it
does
have
loops.
It
does
have
some
logic
which
is
imperative
and
does
break
sometimes,
so
this
entire
githubs
model
still
needs
to
evolve,
and
then
we
have.
You
know
the
whole
issue
around
visibility,
because
now
we
have
a
single
source
of
truth
and
we
need
to
see
what
is
happening
in
the
system.
B
So
we
need
a
lot
of
tools
to
provide
the
visibility
as
to
what
is
happening
in
the
environment,
because
clearly
the
the
feedback
loop
in
many
cases,
including
spinnaker,
is
lacking
on
one
side.
Spinnaker
is
able
to
create
in
several
instances
it's
able
to
show
what
is
there
in
the
infrastructure,
the
actual
state,
but
it
still
does
not
have
a
control
system
or
a
loop
to
bring
it
back
to
the
desired
state.
So
certain
tooling
is
still
required,
and
this
entire
githubs
model
will
mature
as
time
progresses.
B
We
are
working
on
the
tooling
that
will
further
support,
git,
ops
and
also
this
whole
model
of
high
speed
deployment,
which
requires
that
we
have
some
sort
of
a
verification.
So
we
have
a
machine
learning
based
algorithms
that
do
the
verification
we
have
a
policy.
This
is
what
tracy
and
I
were
talking
before.
We
started
that
we
have
policy
integration
that
allows
pipelines
to
ensure
that
there's
a
policy
in
place.
So,
for
example,
what
we
talked
about
github
same
thing
when
you're
installing
we're
deploying
something
to
a
a
dev
environment.
A
Sweeney,
thank
you
so
much.
This
is
a
fascinating
conversation
really
I
you
know
when
I
first
started
talking
to
my
peers
on
this
topic
of
get
ops,
I
would
say
you
know
I
have
a
problem
with
seeing
how
git
ops
will
scale.
I
know
where
I
know
it
can.
In
terms
of
the
architecture
side,
you
can
have
as
many
operators
running
as
you
want
in
every
cluster
and
it
can
scale
find
that
direction.
A
So
I
think
that
I
think
it
looks
like
you
know.
Spinnaker
and
the
opsimex
team
is
really
thinking
about
how
to
solve
these
front-end
problems,
and
on
that
I
want
to
let
everybody
know,
I've
invited
you
to
talk.
So
if
you
have
a
question
there's
just
that,
you
know.
Please
just
don't
be
shy,
just
ask
the
question
and
that
way
we
get
it
recorded.
So
anybody
who's
watching
this
on
recorded,
which
is
the
majority
of
our
viewers.
A
They
can
hear
your
questions
so
please,
if
there's
a
question
about
git,
ops
or
any
aspect
of
what
sweeney
covered
just
go
ahead
and
put
it
out
there
and
to
get
people
thinking
shrini.
What
are
your
thoughts
on
how
this
is
going
to
look
in
a
in
a
microservice
environment
where
we
have
so
many
moving
parts.
B
Yeah
so,
as
you
correctly
pointed
out,
when
things
are
simple,
you
have
one
or
two
pipelines,
two
three
deployments:
it
looks
all
nice,
but
it's
when
you
have
a
micro
services
based
application
with,
let's
say,
a
hundred
different
micro
services.
So
this
is
what
we're
working
with
one
of
our
customers.
B
Things
can
get
very
tricky.
How
do
we
know
that
all
the
100
applications
I
mean
the
source
code
related
stuff,
the
configuration
related
stuff
and
everything
has
moved
nicely
between
dev
staging
qa
and
production
so
that
this
this
company
actually
had
teams
of
people
just
going
and
verifying
that
everything
is
right
before
it
goes
live
and
when,
once
we
have
go
move
towards
github
style,
the
configuration
control
comes
built-in.
B
So
we
can
be
sure
that
at
least
the
same
stuff
is
being
moved
across
environments,
and
what
it
lacks,
though,
is
is,
is
the
control
the
system,
the
feedback,
the
equivalent
of
the
control
plane,
so
to
speak
of
kubernetes,
where
we
can
actually
go
and
check.
If
somebody
modified-
let's
say
a
config
map
manually
in
production,
the
system
automatically
corrects
it
and
puts
puts
it
back
to
where
it
should
have
been
from
it.
So
that's
that
part
is
of
the
tooling
is
still
you
know
further
ahead.
The
road.
A
So
I
can
give
so
we're
looking
at
two
sides
of
this
puzzle
here
when
you
think
about
things
like
flux,
there's
a
there's
that
is
the
operator
itself
now
I
don't
know
if
flux
cd's,
starting
to
build
out
of
devops
kind
of
solution
around
it,
but
flux
is
primarily
that
operator
and
spinnaker
would
be
integrating
spinnaker
doesn't
care.
If
there's
what
operator
you're
using,
I
don't
believe
it's
just
making
sure
that
the
correct
yaml
file
is
being
pushed
into
the
repository.
A
So
you
know,
there's
there's
things
like
flux.
You
know,
rancher
has
an
operator,
I
can't
think
of
the
name
of
it.
At
this
moment,
terraform
has
an
operator.
It
lives
outside
of
a
cluster.
So
when
you
think
about
get
ops,
you
have
to
think
about
the
two
sides.
You
have
to
think
about
the
cd
pipeline
side
and
you
have
to
think
about
the
operator
side.
Flex
is
primarily
an
operator.
A
I
don't
know
if
there's
something
called
flux
cd,
that
is
like
argo
cd,
I'm
assuming
that
all
of
these
operators
are
going
to
put
some
kind
of
a
front
end
on
those
product,
their
tools.
So
they
have
better
insights
in
terms
of
what
it's
doing,
but
I
don't
just
I
don't
think
flux
is
pushing
a
full
cd
pipeline,
but
I
could.
B
A
B
A
Well,
that
means
we'll
be
stopping
about
five
minutes
early.
I
usually
try
to
keep
these
to
about
45
minutes,
because
it
makes
it
a
much
more
pleasant
experience
when
you're
watching
it
from
a
recorded
point
of
view,
so
sweeney.
Thank
you
so
very
much
for
your
time
today
and
for
your
insights
into
get
ups.
I
don't
think
this
will
be
the
last
session
we'll
have
on
get
ops.
I
think
this
is
the
one
of
the
you
know.
A
If
anybody
is
really
interested
in
learning
more
about
get
ops
or
getting
involved
in
the
conversation,
you
should
check
out
the
cncf
github
repository,
look
for
the
get
op
sig
and
all
that
information
will
be
there,
and
on
that
I
will
call
this
meeting
to
an
end
and
thank
you
so
much
for
everyone.
This
will
get
published
and
posted
up
to
the
cdf's
online
meetup
youtube
channel.