►
From YouTube: Enabling GitOps with the GitLab Kubernetes Agent
Description
For this event, we have Nicholas Klick back for a third time!
Nick will uncover the GitLab Kubernetes Agent which is an active in-cluster component for solving any GitLab to Kubernetes integration task.
The Agent empowers operators to manage GitOps style Kubernetes deployments into private networks, behind a firewall or NAT, with the goal of building easier and more secure Kubernetes-based applications.
Come learn all about this new alpha feature directly from the GitLab team working on it!
This will be a Virtual Event! We'll update the event with the Youtube stream link well before we're slated to start. Thanks to our Sponsors StockX and Palo Alto Networks for the support to enable us to continue putting on these events!
B
Hey
everybody,
I'm
mario
lauria,
welcome
to
another
virtual
meetup
of
orchestra,
the
ann
arbor
group.
We
all
know
and
love
I
this
is
this:
is
we
have
over
50
people
that
rsvp'd
for
this
meetup,
I'm
so
excited?
We
have
nicholas
click
coming
for
a
third
time.
He
got
so
bored
during
cobid
that
he
did
all
this
testing
and
playing
and
he
gave
us
this
whole
presentation
he's
like
I
have
to
talk
about
this.
I
need
something
to
do
so.
B
B
But
after
that
meetup
happening
in
end
of
october
october,
21st
to
be
precise,
we
don't
want
to
interfere
with
halloween
after
that
we're
going
to
actually
kind
of
go
on
a
hiatus,
and
I
think
the
big
thing
for
us
is
to
wait
out
a
little
bit
until
we
can
have
everyone
back
together
in
person.
That's
our
goal,
that's
what
orchestra
really
is
about.
What's
what's
truly
brought
together
is
the
networking
I
have
everybody
to
thank
for
that.
All
of
you
you've
made
this
place.
Pretty
amazing.
B
You've
made
this
community
just
remarkable.
We've
come
very
far,
but
I
think
everyone
is
getting
a
little
bit
tired
of
evening.
Video
calls
and
things
like
that,
especially
when
you
just
got
done
with
a
full
day
of
video
calls
and
meetings,
so
ourselves
included,
and
so
I
think
discussion
and
communication
will
continue
in
slack,
and
you
know,
I've
had
meetings
with
all
sorts
of
of
you,
one-on-ones
and
virtual
lunches
and
I've.
I've
heard
from
tons
of
you
of
ideas,
and
things
like
that.
B
So
I
really
appreciate
those
keep
them
coming.
Obviously,
jeff-
and
you
know,
and
bob
and
george
are
still
very
active
in
the
kubernetes
community
and
we're
still
trying
to
see
each
other
here
and
there
we
can
in
ann
arbor,
but
I
think
that
the
community
will
obviously
live
on,
but
we're
going
to
take
little
hiatus
from
doing
these.
These
events
and
that's
that's
good
fall-
is
busy.
We
just
had
kubecon
eu
we've
got
many
other
events.
I
know
re
invents
coming
and
you
know
it's.
B
C
All
right,
well,
mario.
I
certainly
appreciate
that
introduction
and
I
certainly
appreciate
everyone
showing
up
to
this
to
this
talk
after
a
long
day
of
video
calls,
I'm
sure
so
yeah
tonight,
I'd
like
to
talk
briefly
about
the
kubernetes
agent.
This
is
a
new
project
that
we've
been
developing
at
gitlab.
C
My
name
is
nicholas
click,
I'm
an
engineering
manager
at
gitlab,
so
yeah.
First
of
all,
I
just
wanted
to
say
thanks.
Everyone
for
coming
out
do
a
real
brief.
Who
am
I
I
manage
the
configure
team
at
gitlab
and
the
configure
team
is
primarily
focused
on
working
on
terraform
and
infrastructure,
as
code
features
as
well
as
kubernetes
management.
C
I
wanted
to
to
market
some
of
our
some
of
our
terraform
stuff
before
I
get
into
the
kubernetes
stuff.
So
a
couple
of
the
things
we've
been
doing
with
terraform
on
my
team.
Is
we've
been
working
on
the
gitlab
managed
terraform
state?
So
if
you
happen
to
use
terraform-
and
you
also
want
to
use
gitlab,
you
can
actually
store
your
terraform
state
within
gitlab.
C
We
also
have
a
terraform
plan
widget,
which
will
allow
you
to
output.
Your
your
changes
into
the
mr
into
the
actual,
mr,
so
you
can
see
those
changes
there
as
well.
We
we
recently
took
over
the
terraform
provider
for
gitlab.
We
took
full
ownership
of
it
and
are
moving
it
forward
to
try
to
get
it
up
to
parity
with
the
gitlab
api.
C
One
has
been
managing
clusters,
so
basically
we
allow
you
to
to
basically
set
up
clusters
on
gke
and
eks
and
associate
those
clusters
with
your
gitlab
instance.
We
also
let
you
to
bring
your
own
clusters
and
and
associate
those
with
gitlab,
and
then
we
we
have
a
managed
apps
component.
Where
you
can,
you
can
install
a
kind
of
a
suite
of
apps
into
your
cluster
and
oops
yeah
I'll
just
show
a
couple
screens
of
that.
So
in
the
kubernetes
management
we
you
know.
Basically
you
can.
C
You
can
create
a
cluster,
it's
very
ui
based
what
we're
currently
doing,
and
you
know
after
you
set
it
all
up.
You
can
you
can
see
your
clusters,
here's
an
example
of
where
you
can
install
and
uninstall
those
managed
apps
into
your
cluster,
and
so
so
a
lot
of
what
was
built
previously
was
it
was
sort
of
very
ui
based.
It
was
kind
of
more
focused
on
on
helping
people
get
started
with
kubernetes
kind
of
their
first
entry
point
into
into
using
it,
but
there
there
were.
C
There
are
a
number
of
issues
with
this.
Essentially
so
right
now,
with
this
kubernetes
management
features
that
we
have.
It
requires
your
cluster
to
be
open
to
the
internet.
So
you
you,
you
need
to
allow
gitlab
to
be
able
to
talk
through
a
certain
ingress
at
some
point
directly
to
your
cluster
and
then
also
there's
a
requirement
for
to
basically
give
git
lab
cluster
admin
rights
so
kind
of
like
a
more
privileged
access
to
your
cluster
and
then
finally,
it
kind
of
had
exclusive
support
for
these
push-based
changes.
C
So
you
would
you'd
make
a
change
and
then
and
then
push
it
out.
It
didn't
really
allow
for
this
kind
of,
like
pull
base,
get
ops
which
is
becoming
more
popular.
So,
due
to
many
customer
many
many
many
customer
requests
and
and
just
our
own
analysis,
we
decided
to
create
a
new
prod
new
product
called
the
gitlab
kubernetes
agent,
and
essentially
our
design
goals
with
this
were
to
solve.
C
The
issues
that
are
described
above
so
essentially
allow
it
to
work
in
a
situation
where
you
don't
like
your
cluster
can
cannot
be
accessible
through
the
public
internet
like
it.
It
might
maybe
it's
utilizing
that
gateway
or
basically
like,
like
you,
don't
want
public.
You
know
requests
coming
in
from
from
your
gitlab
instance,
and
then
also
we've
had
a
lot
of
requests
for
for
something
pull-based
get
ops
related.
So
this
this
solves
that
as
well.
C
So
this
is
a
project,
it's
called
it's
all
encapsulated
under
the
get
lab
agent
project
and
the
the
gitlab
agent
is
really
it's
kind
of
several
things
combined
and
I
kind
of
want
to
talk
about
what
those
pieces
are.
So
for
one
there's,
there's
an
in
cluster
component,
which
actually
runs
inside
your
inside
the
kubernetes
cluster.
C
This
is
called
agent
k
and
then
there's
a
server
side
component
called
the
kubernetes
agent
server,
and
this
is
a
golang
process
which
runs
to
the
side
of
a
gitlab
instance,
and
then
we've
we've
updated
the
gitlab
rails
api
to
be
able
to
communicate
with
the
kubernetes
agent
server,
and
essentially
these
three
things
combined.
We've
we've
enabled
excuse
me
bi-directional
streaming
where
we
kind
of
like
leave
open
this.
C
This
kind
of
this
event
string
so
that
we
can
have
constant
communication
between
your
rails,
your
cass
and
your
agent
k,
as
well
as
other
components
in
the
gitlab
architecture,
and
then
the
the
actual
get
ops
mechanism.
We're
using
a
project
called
get
ops
engine
and
I'll
go
into
that
later.
What
that
is,
it's
essentially
a
a
core
component
of
argos
cd
or
it's
what
they're
they're
aiming
to
be
their
their
get
ops
engine
within
argo
cd.
C
So
this
is
the
high
level
architecture
for
for
the
agent
you
can
see
at
the
top
there's
a
kubernetes
cluster
and
within
that
lives.
The
agent
k
we're
using
grpc
bi-directional
streaming
to
kind
of
allow
that
agent
k
to
communicate
out
to
the
kubernetes
agent
server,
which
is
running
again
to
the
side
of
the
gitlab
rails
app
and
that
that
that
kubernetes
agent
server
allows
this
kind
of
the
brokerage
between
between
requests
from
the
the
get
lab
rails
app
and
your
cluster,
so
to
kind
of
dive
into
each
piece
of
it.
C
What
it
does
is
it
it
handles
the
authentication
when
agent
k
is
kind
of
initially
or
when
it's
initializing,
then,
whenever
it
makes
requests
it's
handling
the
authentication
to
the
rails.
App
excuse
me
it
all
it.
It
also
fetches
the
agent's
configuration
from
each
corresponding
git
repository,
because
a
single
kaz
instance
is
able
to
manage
many
many
different
get
repositories
and
we'll
kind
of
go
into
the
structure
of
how
you
set
up
these
kind
of
get
ops.
C
The
the
kind
of
configuration
that
you
need
in
order
to
enable
this
get
ops
mechanism
and
then
as
well.
It
matches
the
incoming
requests
from
gitlab
to
existing
connections.
So
they're
not
like
we're
not
having
to
constantly
open
new
connections,
and
then
you
know
it's
also
polling
these
manifest
repositories
and
we'll
get
into
what
the
manifest
repositories
are
in
a
little
bit.
So
it's
kind
of
pulling
these
manifest
repositories
looking
for
changes
which
will
then
execute
the
the
kind
of
get
ops
engine.
C
So
how
does
the
how's
the
agent?
How
does
the
agent
get
up
work?
So
essentially,
you
know
again.
This
is
this
is
kind
of
similar
to
what
I
showed
before
agent
k,
it's
gonna
fetch
configuration
and
and
the
manifests
from
kubernetes
agent
server.
The
kubernetes
agent
server
is
kind
of
constantly
polling
for
changes
to
those
githubs
repositories.
C
C
Now
the
this
is
a
little
so
so
I
mean
basically
yeah.
I
don't
know
if
you
can
see
my
oops
so
basically
what's
happening
here.
Is
your
your
pushing
code
changes?
Those
code
changes
will
need
to
update
your
manifest
file.
That
file
again
is
being
kind
of
like
it's
being
watched
over
whenever
it
changes.
Those
changes
are
pulled
in
by
the
kubernetes
agent
and
it's
those
are
applied,
so
we
are
leveraging
the
get
ops
engine
under
the
hood
again.
C
This
is
a
project
that's
owned
by
argos
cd
or
the
the
people
behind
argo,
cd
and
essentially
what
they're.
What
they're
planning
with
this
is
that
it
will
become
the
backbone
to
to
enable
get
ops
within
argo,
and
the
plan
originally
was
for
also
to
to
be
the
backbone
behind
flux,
and
that
was
their
original
agreement.
I
think
flux
has
since
changed
their
plans,
but
we
we,
we
also
have
have
an
option
to
kind
of
go
with
what
they've
changed
or
switch
to
as
well.
C
I
think
it's
called
the
the
get
ops
tool
kit
or
the
ops
toolkit,
something
like
that.
Yeah
and
like
I
get.
I
can
talk
about
that
in
a
little
bit
so
anyway,
we're
using
the
get
ops
engine.
Essentially
we
have
gotten
in
early
with
this,
and
and
the
reason
that
we
went
with
this
rather
than
straight
up.
Argo,
cd
or
flux
is
that
we
felt
like
this
more
closely
matched
the
the
kind
of
level
of
of
integration
that
we
were
looking
for.
C
Flux
and
argo
cd
are
much
are
kind
of
much
higher
on
the
on
the
kind
of
implementation
layer
like
like
they
have.
You
know
they
might
have
a
ui
or
there
might
be
more
feature
complete
and-
and
we
like,
we
were
looking
for
something
a
little
bit
more
at
a
lower
level
where
we
could
get
the
functionality
of
git
ops,
but
also
build
our
own
tooling.
On
top
of
it,
so
we've
actually
been
heavily
involved
in
working
with
the
get
ops
engine.
C
In
terms
of
you
know,
bug
fixes
or
contributing
to
the
road
map,
and
so
we're
very
excited
about
about
working.
You
know
like
working
with
them
and
working
with
the
community
to
make
that
make
that
a
really
valuable
project,
so
I've
kind
of
gone
over
the
the
architecture
of
what
we've
done
and
how
it
works,
and
I
kind
of
wanted
to
go
through
how
to
set
up
how
this
is
set
up.
C
C
C
So
currently,
the
only
way
to
install
the
kubernetes
agent
server
is,
with
the
get
lab
helm,
chart
we're
planning
to
get
the
the
the
agent
into
omnibus
and
to
get
lab
omnibus
in
this
milestone.
It
should
should
be
in
there
and
then
it
should
be
on.com
on
gitlab.com
either
this
milestone
or
the
next
milestone,
so
for
the
very
latest
by
the
end
of
next
month.
We're
expecting
for
this
to
be
available
on.com.
C
Just
simple
helm
install
and
the
the
one
thing
to
note
is
this
global
cast
enabled
true
needs
to
be
so
so,
like
you
can't
just
install
like
the
latest,
get
get
lab,
helm
chart
and
expect
it
to
be
turned
on.
You
need
to
make
sure
to
set
this
additional
attribute
and
then
the
the
next
step
is
that
you
need
to
set
up
a
configuration
repo,
and
so
basically
all
this
is
is
it's:
is
it's
a
repository
that
has
a
file
with
this
type
of
directory
structure?
C
C
Like
each
one
of
those
will
have
a
config,
yaml
and
and
that
configural
is
structured
like
this,
so
it's
get
ops
manifest
projects
and
then
this
this
id
is
actually
the
the
path
like
the
the
the
get
lab
url
or
the
the
gitlab
path
to
your
the
the
project
that
you
want,
the
the
agent
to
be
monitoring
and
I'll
go
into
that
example.
Here
in
a
little
bit.
C
C
It's
configure
examples
get
ops
and
within
that
project
I
have
a
manifest.yaml
file.
So
you
can
see
you
know
it's
two
different
projects.
It
is
possible,
so
I
I
want
to
be
clear
that
you
don't
have
to
use
two
different
projects.
It's
just
easier
to
explain
initially,
to
excuse
me
to
other
people
that
you
kind
of
have
one
project.
C
You
would
have
it
pointed
out
itself,
basically
so,
okay,
so
yeah
like
after
we
have
this.
This
configuration
repo
again,
which
is
this.
This
is
your
configuration
repo.
We
need
to
remember
this
agent,
one
okay,
so
so
I
named
it
agent
one,
and
so
what
I
need
to
do
is
now
at
present,
because
this
is
still
very
alpha
in
in
nature.
C
But
if
you
go
to
the
kubernetes
agent
documentation,
there's
there's
the
graphql
commands
to
run
in
order
to
create
your
agent
and
also
receive
your
your
token,
which
you're
going
to
need
in
the
next
step,
and
so
our
we're
we're
rapidly
working
on
a
ui
so
that
you
don't
have
to
kind
of
do
this.
But
I
wanted
to
explain
it
just
so.
C
It
was
clear
so
yeah
you,
you
run
rails
console
you,
you
run
these
commands
in
order
to
create
the
your
agent
record,
create
the
token
and
then
that's
going
to
that
token
you're
going
to
to
utilize
when
you
create
your
kubernetes
secret
in
this
step.
So
you
need
to
put
that
right
here
where
it
says
your
agent
token
you're,
going
to
insert
that
token
there
and
that's
going
to
create
your
your
kubernetes
secret.
C
Then
the
next
step
is
that
you
have
to
like
now
the
secrets
there.
You
have
to
kind
of
kick
start
your
agent
k,
and
so
this
basically
in
in
our
kubernetes
agent
docs.
We
have
an
example
of
a
resources
gml
but
like
you
could
use
kpt
or
you
could
use
whatever
you
wanted.
Obviously,
like
you
don't
need
to
use,
and
this
is
not
the
full
file.
I
want
to
be
totally
clear
about
that.
C
If
you
go
to
the
docs,
this
is
just
a
part
of
the
file,
but
I
just
wanted
to
highlight
what
you're
what
you're
doing
here
is
your
you
know
it's
going
to
pull
down
this
image
this
this
agent
k
image,
and
then
it
also
points
to
this
cass
address,
which
is
the
address
of
your
cast
instance.
And
so,
if
you
were
running
this
with
a
self-hosted
gitlab,
then
you
would
obviously
point
it
at
the
the
url
and
the
port
where
your
cas
is
running.
C
C
C
So
if
you
set
up
the
the
the
the
gitlab
agent
associate
that
with
with
an
existing
cluster,
we
want
to
be
able
to
detect
all
your
applications
that
run
on
there
and
pull
in
all
the
information
about
that
into
gitlab.
We
also
want
to
be
able
to
do
real-time
data
streaming
to
be
able
to
see
what's
happening
kind
of
in
real
time
on
on
the
on
the
cluster
as
well.
C
C
So
currently,
this
is
available
in
the
gitlab
helm
chart
and
we
expect
that
it
will
be.
It
will
definitely
make
it
on
to
us
this
month
and
then
it's
gonna
be
a
bigger
process
to
get
it
onto.com
and
right
now
in
omnibus
and
the
helm
chart
is
the
premium
feature.
So
it's
not
it's
not
a
core
yet,
but
we're
kind
of
we're
currently
negotiating
with
the
product
team.
C
B
Okay
awesome
thank
you
nick,
that
was
that
was
pretty
fast
as
well
yeah.
Any
questions,
please
type
them
in
the
the
slack
channel.
If
you
have
anything
after
the
fact
too,
we
can
always
forward
them
to
nick
nick.
Why
don't
you
enlighten
us
with
your
email
and
twitter
and
how
to
how
to
best
contact
you.
C
My
product
manager
would
love
to
chat
with
you.
He
does
interviews
all
the
time,
we're
always
trying
to
learn
what
what
people
like
what
they're
using
what
their
practices
are,
what
their
use
cases
are.
I
think
you
know
every
time
we
do
these
interviews.
It
really
helps
us
with
our
road
map
and
what
we're
planning
to
do
so.
C
B
C
C
So
if
let's
say
you're
generating
like,
if,
if
you
can
gener
generate
that
manifest
yammel
or
or
you
can
have
a
set
of
repositories
with
manifest
yaml
files,
you
can
you
can
sync
up
all
those
at
once
to
like
a
set
of
clusters,
and
we,
just
literally,
I
was
talking
with
the
guy
with
people
today
about
name
spacing.
So
what
we're
going
to
do
is
the
the
configurable
we're
going
to
allow
you
to
define
a
name
space
for
each
each
manifest
files,
so
you'll
have
separate
a
kind
of
segment.
C
Everything
does
that.
Does
that
help
answer
the
question
a
little
bit
yeah,
so
it
does
right
now
it
only
searches
for
a
manifest
yaml
file.
It
does
not
descend
into
directories
again,
so
I
want
to
explain
something
so
like
gilad.
We
have
this
this
concept
of
mdc,
which
is
a
minimum
viable
commit.
So
we
build
the
most
basic
thing
that
will
work
and
we
get
that
out
as
fast
as
possible.
So
focusing
on
kind
of
syncing.
B
C
Yeah
definitely,
it
definitely
seems
like
it's
moving
forward.
We
we've
been
contributing
a
bit
to
it
like
they
definitely
have
a
road
map.
I
am
not
the
most
up
to
speed
on
their
exact
road
map.
I
do
know
that
it
it
is,
it
definitely
is
moving
forward.
Let
me
just
take
a
look
at
some
things.
They're
trying
to
do
so.
C
Yeah
I
mean
they.
You
know
they've
been
focused,
it
seems
like
they
got
the
resource
cache
working
and
you
know
I
mean
like
like
like
what
we'd
like
eventually
is
the
the
manifest
generation
to
have
a
little
bit
more
kind
of
like
an
easier
way
to
do
that
as
well
and
and
separately
from
the
get
ops
engine.
I
want
to
talk
those
six
steps
that
I
talked
about
again.
This
is
the
kind
of
minimal
viable
way
of
us
doing
this.
C
B
C
Let's
see
see
yeah
I
mean
so
we
are,
we
are
hiring
now
I
I
will
say
that
whenever
kova
gets
over,
we'll
be
hiring
a
lot
more.
So
definitely
you
know
like.
Like
keep
that
in
mind.
We
it's
it's
great
culture
at
get
lab.
100
remote
we've
been
remote
forever.
You
know,
we've
got
people
all
over
the
world
about
1400
folks,
it's
a
really
great
place
to
work.
What
else
is
interesting?
C
B
C
Yeah,
so
that's
that
is,
we
haven't
built
that
yet
so
yeah
future
functionality
yeah.
So
I
mean
the
the
idea
would
be
that
we
would
be
able
to
inspect
your
cluster
and
kind
of
like
you
know,
annotate
and
figure
out
what's
there
and
then
pull
that
into
git
lab
and
provide
some
dashboards,
provide
some
monitoring
metrics
things
like
that,
but
have
it
be
very
automatic,
very,
very
simple
to
set
up
because
right
now,
the
problem
is,
is
that
these
managed
apps?
C
B
Great,
thank
you.
I
think
that
definitely
answered
his
question.
That
was
that
was
good,
great
presentation.
Thank
you
again,
nick.
If
there
was
a
whole
crowd
here
of
40
people,
they'd
all
be
clapping
very
loudly,
and
then
they
would
get
up
and
get
more
pizza
and
mountain
dew,
hopefully,
hopefully
sooner
than
we
think
that
will
be
a
realistic
possibility.
B
I
want
to
thank
everyone
that
joined
today.
Some
really
great
questions.
Any
any
further
questions
nick
will
answer
in
the
channel.
You
can
also
email
him
and
click
at
gitlab.com,
and
you
know
he's
he's
very
involved,
so
you
know
just
just
ask
for
him,
and
one
of
us
can
help
route
you
as
well.
Thanks
to
my
my
team,
bob
jeff
and
george,
I
could
not
do
this
without
them.
They've
helped
make
this
happen.
You
know
the
planning.
B
B
Things
like
that
lawrence
from
solo
is,
is
a
huge
get,
we're
really
excited
to
have
them,
and
that
is
october
21st
that
wednesday.
So
we
will
remind
everyone
about
that.
I
will
link
it
here
in
the
slack
channel
as
well.
I
I
really
do
miss
all
of
you.
If
anybody
has
any
pictures
from
previous
meetups
when
we
were
in
person,
please
feel
free
to
share
them.
We're
we're
at
least
myself
trying
to
collect
as
many
as
we
can.
I
posted
one
of
the
snowy
day.
B
I
think
it
was
in
february
of
this
year
that
we
had,
which
seems
like
so
long
ago
now.
So
thanks
for
for
everyone,
who's
who's
going
to
be
actually
have
some
new
people
join
the
channel
one
or
two
of
which
I've
I've
either
I'm
working
with
or
have
worked
with
before
so
great
to
see
everybody
so
with
that,
take
care
and
we'll
see
everybody
in
three
or
four
weeks,
thanks
so
much
for
logging
on.