►
Description
GitOps for Keptn - The Keptn GitOps Operator
The Keptn CLI is very powerful for the configuration and imperative management of Keptn. But when integrating Keptn into the CI infrastructure, it is useful to be able to check in the configuration and also triggering deployments in a GitOps fashion (simply doing a git push to the configuration repository). During the adoption of Keptn at Dynatrace, we wanted to make the usage of Keptn transparent to the developers and therefore, we created an operator, which deals with a declarative configuration checked into git to control Keptn.
In this user group, Thomas and Dieter will give you an overview of the keptn-gitops-operator and typical use-cases.
Source code: https://github.com/keptn-sandbox/keptn-gitops-operator
Timecodes:
00:00 Intro
02:20 How the operator works
09:16 Demo
15:59 Q&A
26:40 Closing remarks
Learn more: https://keptn.sh
Get started with tutorials: https://tutorials.keptn.sh
Join us in Slack: https://slack.keptn.sh
Star us on Github: https://github.com/keptn/keptn
Follow us on Twitter: https://twitter.com/keptnProject
Join our Keptn community newsletter: https://keptn.sh/community/newsletter/
A
Today
we
have
a
very
exciting
topic:
it's
the
github's
operator
for
captain
and
although
I've
already
seen
some
small
parts
of
it,
I'm
really
excited
to
see
what
it
can
do
and
actually
how
you
can
use
it
and
how
it
makes
your
life
easier
by
using
the
g-dops
operator
for
captain.
In
addition
to
the
captain
control
plane
that
you
already
have
today,
I'm
joined
by
two
great
engineers,
thomas
hi,
thomas.
A
And
I
will
hand
over
to
you
just
in
a
second,
but
I
would
also
like
to
announce
that
we
do
have
a
meeting
document.
I
will
put
the
link
to
the
meeting
document
in
the
chat,
so
if
you're
joining
here
live
in
this
captain
user
group
meeting,
please
take
a
look
at
this
document.
Just
fill
in
your
name,
so
we
can
basically
can
keep
track
of
attendance
and
later
on,
you
will
also
find
the
link
to
the
recording
in
this
document.
Everything
will
be
recorded.
A
We
will
also
put
it
on
on
on
youtube
on
our
channel,
and
so,
if
you
want
to
watch
it
later
on,
that's
totally
fine
as
well
before
we
get
started.
One
very
small
remark
from
my
side,
as
we
are
now
having
a
very
stable
release,
cadence
of
captain,
I
really
encourage
everyone
to
always
use
the
latest
version
of
captain.
It's
it's
really
the
latest
and
the
greatest.
A
I
can
really
say
this
for
captain,
so
the
new
version
will
come
out
on
thursday
or
friday,
so
make
sure
to
take
a
look
at
our
website
and
it's
available
and
make
sure
to
upgrade
your
captain
installations
when
they
are
ready.
So
with
this,
let's
see
the
latest
and
greatest
from
the
community
and
from
the
time
thomas,
I
think.
First,
I
will
be
handing
over
to
thomas
to
take
a
little
bit.
A
B
B
In
our
opinion,
we
we
found
out
that
captain
itself
works
very
imperatively,
so
when
you
have
to
when
you
want
to
create
the
service,
you
have
to
type
in
captain
create
service.
When
you
want
to
send
an
artifact,
you
have
to
type
captains
and
artifact,
and
if
you
want
to
add
additional
additional
resources,
you
have
a
captain
add
resource
command.
B
B
So
the
next
slide,
please
so
first
things:
first,
what
is
an
operator?
An
operator
can
be
used
to
encapsulate
operational
knowledge
and
application
code.
So
an
operator
is
a
piece
of
software
often
encapsulated
in
a
container,
and
it
has
some
logic
in
there
to
know
how
to
transition
for
to
transition
from
one
state.
So
I
mean
you
need
to
speech
the
transit.
The
desired
state
to
another
state
and
to
another
state,
so
the
state
we
have
in
the
real
world
the
operator
can
maintain
the
state
of
the
managed
objects.
B
In
our
case,
the
captain
configuration
so
it
recreates
objects
when
they
get
deleted.
So
in
our
case,
when
we
create
the
captain
service
and
we
delete
it
on
the
control
plane,
it
gets
recreated
afterwards,
yes,
and
it
might
be
use.
It
is
a
very
cool
thing
to
to
that.
It
also
deletes
the
payment
objects
when
the
primary
source
is
deleted.
B
Yes
and
what's
in
an
operator,
so
we
have
kind
of
an
operational
knowledge
encapsulated
in
a
controller.
We
have
a
description
of
a
desired
state
in
the
kubernetes
world.
It's
often
the
custom
resource
and
we
have
control
loops,
which
checks
the
desired
state
which
matches
the
real-world
state.
So
next
slide,
please
so!
Yes,
I
talked
about
the
keytops
operator.
B
So
how
can
a
keyptops
operator
look
like
you
see
on
the
in
the
middle
of
the
of
the
screen,
an
operator
you
have
a
definition
of
what
the
operator
should
manage
in
a
custom
resource
definition
in
kubernetes.
In
our
case,
this
might
be
a
captain
service
in
the
captain
project
which
I'll
co
I
will
come
to
later.
B
B
B
In
our
code
repository,
we
have
a
captain
configuration
which
consists
of
an
hailing
chart
of
services
of
of
test
configurations
and
so
on,
and
we
have
besides
the
application
code,
and
this
gets
transformed
in
a
ci
step
to
the
captain
configuration
we
all
have
in
the
in
the
in
the
git
upstream
repository
and
in
the
control
plane
in
the
captain.
Control
plane
leaves
the
captain
github's
operator
in
the
middle
of
the
square,
and
this
captain
keytops
operator
has
a
knowledge
about
captain
project.
B
Should
captain
manage
and
based
on
this
information,
the
captain
github's
operator
talks
with
the
captain
api
and
is
able
to
create
and
delete
services
at
the
moment.
So
if
you
specify
a
new
service
in
the
in
the
configuration
repository,
this
gets
replicated
to
captain
and
in
on
the
in
the
second.
The
second
thing
the
captain
operator
is
able
to
do
is
if
something
in
the
in
the
branch
the
captain
operator,
watches
changes.
B
Then
a
new
deployment
gets
triggered,
and
this
is
all
this
is
based
at.
First,
on
the
on
the
committee
of
the
of
the
watched
branch
and
in
the
second
line
on
the
on
a
hash
of
the
directory
of
the
service
in
the
crypt,
in
the
captain
configuration
repository
and
in
our
deployment,
we
have
a
separate
execution
and
control
playing
cluster
and
afterwards
the
captain
helm
service,
watches
the
deployment
events
and
the
deployment
events
gets
triggered
when
something
changes
and
deploys
this
to
applications.
B
C
C
So
I'm
going
to
watch
it
because
then
we
can
actually
see
it
when
it
changes
and
in
this
window
I
have
the
checked
out
version
of
the
captain
repository
and
I'm
now
in
the
death
branch
and
within
the
death
branch,
which
is
my
my
stage,
is
a
helm
chart
and
I'm
going
to
change
the
helm.
Chart
I'm
going
to
change
the
values
under
the
armor
and
scale
up
the
application
to
two
replicas.
A
C
A
23.
cool,
so
what
if
I
understand
it
correctly?
What
really
happens
here
is
the
the
git
operator
watches
the
the
repository
on
github.
So
that
might
be
a
I'm
not
sure
if
this
is
a
public
or
private
repository
and
at
latest
every
30
seconds,
it
will
recognize
the
change
and
we'll
then
send
immediately
the
events
to
captain,
because
it
was
quite
fast
when
we
saw
the
event
in
the
bridge
and
also
when
we
saw
the
the
new
instance
of
the
potato
head
or
the
potato
server
application.
A
B
A
C
B
C
A
B
Yes,
we
have
a.
We
have
a
yes,
a
repository
in
the
captain's
handbooks
and
there's
kind
of
dark
hands
kind
of
installation
instructions
there.
So
what
you
need
to
install
this
is
to
install
the
crds
which
are
in
there,
so
you
have
to
check
out
the
the
repository
at
any
at
any
way
in
the
current
in
the
current
state,
as
this
is
in
the
very
early
stages.
B
As
I
mentioned
here
in
the
slides
and
aft,
and
in
this
repository
there
is
a
kind
of
a
kind
of
a
make
file
and
running
a
make,
install
command
in
the
from
the
repository
installs,
the
custom
resource
definitions,
I
think,
and
afterwards
there's
also
a
cubesat
run
command.
This
can
this
will
also
be
packed
in
the
helm
chart
chart
afterwards
and
using
this
you
can
run
the
captain
operator
and
afterwards
there
are.
B
There
are
two
things
you
have
to
do,
so
you
need
a
captain,
slash
config
file
in
the
in
your
git
repository,
but
this
is
also
mentioned
in
the
in
the
installation
system.
Instructions
and
you
have
to
create
a
custom
resource
for
the
captain
project.
You
want
to
watch
and
afterwards,
so
the
captain
github's
operator
is
looking
for
the
secrets
and
so
on,
for
the
for
the
git
repository
in
the
cable
namespace
and
afterwards
connects
automatically
to
the
captain
to
the
key
to
the
git
repository
and
does
the
rest.
B
But
yes,
as
I
said
before,
it
is
in
a
really
really
really
early
stage.
So
it's
a
0.1
version
at
the
moment
and
we
are
about
to
create
the
helm
chart
for
that
to
make
the
installation
experience
a
bit
a
bit
more
convenient
yes,
but
we
we
wanted
to
present
this
to
you
in
a
very
early
stage,
so
you
could
also
give
yes
you
could
you
could
also
propose
your
ideas
to
other
make
contributions
or
whatever
so
we
are.
B
A
A
B
So
also
as
as
I
stated
on
on
the
slide
here,
so
we
are
also
thinking
about
creating
projects
via
crd
at
the
moment,
but
that's
not
implemented
at
the
moment,
but
that's
not
implemented,
but
we
also
did
not
want
to
manage
every
project
which
is
in
captain
via
the
operator
because
it
might
be
intended
that
someone
wants
to
manage
it,
manage
his
project
impurely,
I'm
so
excited
to
do
this
via
the
custom
resource
definition.
At
the
moment.
A
Got
it
I
think
that
answers
the
questions
thanks
thomas
then
there
is
another
question
coming
in:
does
it
only
work
with
a
github
repository
or
others
like
gitlab
or
people
can
also
support
it,
and
how
does
it
relate
to
the
internal
git
repository?
That
is
also
managed
by
captain
kind
of
a
technical
question,
but
maybe
you
can
also
comment
on
this.
B
Okay,
so
it
so,
we
only
tasted
it
with
github,
but
it
should
work
with
every
key
to
propository.
As
far
as
I
know,
because
so
it
will
work
with
every
git
repository
which
can
be
used
with
captain.
That's
it
this
way,
because
it
only
uses
the
the
secret
which
is
written
by
captain
and
takes
the
git
credentials
from
there
and
watches
watches
exactly
this
git
repository.
I
think
this
was
the
first
question.
A
B
A
A
Cool
yeah,
I
got
it
and,
as
you
also
mentioned
it's
in
an
early
stage,
so
everyone
that
wants
to
to
contribute-
I
guess,
can
also
jump
in
here-
write
an
issue
or
raise
a
pull
request.
I've
also
linked
the
repository
in
the
the
meeting
document
and
I
will
paste
the
link
to
the
meeting
document
also
once
again
in
the
chat,
because
we
know
that
a
couple
of
folks
have
joined
a
little
bit
later.
A
B
D
Maybe
I
can
quickly
jump
in
here.
I
I
kept
quiet
for
a
while,
but
you're
right
we
are
with
captain
all
eight
one.
We've
introduced
a
new
concept
of
managing
secrets
with
captain
right
now.
I
think
the
secrets
are
just
stored
as
kubernetes
secrets,
but
the
idea
going
forward
is
that
you
can
also
define
and
replace
and
put
in
your
own
secrets.
D
Correct
the
idea
is
that
we,
you
know
that
a
a
kind
of
a
secrets
I
mean
the
secrets
will
be
available
for
captain
and
the
captain
participating
services,
okay
and
so
whether
it
whatever
you're
writing.
They
should
then
be
able
to
access
secrets,
not
all
the
secrets
there's
also,
if
you
look
at
the
captain
secrets
implementation
right
now,
when
you
create
a
new
secret,
you
also
specify
a
so-called
scope,
so
the
scope
in
the
future
will
then
define
who
has
access
to
those.
D
D
B
A
Yeah,
I
think
it
also
kind
of
it
is
a
very
nice
bridging
the
gap
between
event
driven
or
data
driven
and
in
github
streaming,
because
for
the
bootstrapping
you
you
can
now
use
a
lot
of
if
it's.
If
you
need
already
a
couple
of
files
for
creating
the
projects
and
the
sli
slos,
you
just
put
them
in
the
gift
repository,
but
for
all
the
integration
with
other
tools.
We
are
still
using
the
the
event-based
approach
and
I
think
those
two
really
can
can
work
great
together.
B
A
A
So
no
questions
yet
then
I
would
have
like
the
the
final
or
maybe
obvious
question.
So
what
is
the
kind
of
next
step
for
for
this
operator
if
we
think
of
this,
of
a
0.0.2
release,
or
maybe
a
0.1
release?
What
what
are
the
next
features
that
we
we
can
see,
or
we
could
expect
from
from
the
captain
operator.
B
So,
at
least
from
all
from
use
cases
we
have,
we
have
ourselves,
we
need
some
kind
of
installation
mechanism
for
that
operator,
so
we
need
a.
We
need,
a
hailing
chart
or
another
mechanism
which
which
can
be
used
to
install
it,
because
to
be
honest,
currently
it's
not
really
the
nicest
way
to
the
nicest
experience
for
for
us
to
install
this.
B
So
if
anyone
would
like
to
raise
a
pull
request,
I
think
I
I
think
I
will
I
will
most
probably
accept
it,
and
if
someone
has
ideas
what
we
could
improve
and
what
would
be
useful
for
this
operator,
I
think
every
I
think
there
might
be
very
cool
use
cases
for
the
operator
which
we
would
be
which
we
would
like
to
implement.
But
as
I
said
before,
it's
also
it's
also
depending
on
what
the
community
needs
from
the
operator,
and
it
should
help
anyone
and
not
only
us.
B
So
I
think
it
would,
it
would
really
help
for
the
maturity
of
the
operator
if
there
were
more
tests,
because,
to
be
honest,
currently
the
the
opera
was
written
in
a
few
hours,
so
they
are
not
done
many
tastes
in
there,
but
yes,
we
are,
we
would
we
will
have
to
write
them
at
any
way.
So
if
any
someone
likes
to
write
tests
he's
very
welcome
and,
as
I
said,
I
will
accept
every
pull
request
for
that.
A
I
think
it's
really
cool
that
we
now
have
some
kind
of
a
really
greenfield
project
where
you
can
start
with
cloud
native
technologies,
write
an
operator
join
thomas
and
defend
writing
and
operators.
So
if
anyone
is
up
to
for
this
challenge,
please
reach
out
to
thomas
and
tita,
also
in
the
captain's
lag
and
joined
them
in
making
this
a
really
great
project,
and
I
think
it
already
is
a
great
project.
A
I'm
really
excited
to
see
the
next
releases
of
this
me
too
with
so,
there
were
no
more
questions
coming
in,
so
I
think
we
can
basically
close
it
a
little
bit
earlier,
thanks
titus
thanks
thomas
for
presenting
for
explaining
for
answering
all
these
questions.
We
will
have
our
next
captain
user
group
in
about
one
month.
Let
me
just
find
the
date
real,
quick
for
you.
A
It
will
be
on
the
18th.
May
18th
we'll
run
our
next
user
group,
and
there
will
be
another
captain
development
community
meeting
on
thursday.
There
will
be
a
new
release
coming
out
this
week,
so
make
sure
also
to
watch
this
new
updates
in
the
the
captain's
lag
for
example,
or
on
twitter,
and
thanks
again
and
really
looking
forward
to
see.
What's
coming.