►
From YouTube: Product directions for Configure
Description
For more details and discussion: https://gitlab.com/gitlab-org/configure/general/-/issues/68
A
So
let
me
start
I
will
start
with
the
proof
of
concept
with
the
agent
because
I
really
like
it,
let's
call
it
broker
asmik
I'll,
rename
it
so
I've
been
playing
with
Argos
CD
during
the
weekend
and
I
think,
obviously
pretty
nice
piece
of
product,
obviously
its
tenth
of
the
size
of
gitlab,
so
it
sticky
focuses
on
deployments
to
a
cluster,
nothing
gasps,
which
is
pretty
limited
use
case.
On
the
other
hand,
it
has
a
few
really
nice
features.
First
of
all,
I
could
install
it
without
any
hassle.
Just
following
the
dogs,
everything
works
perfect.
A
The
other
thing
that
I,
noticed
and
I
would
like
to
see
in
our
agent
as
well
is
to
be
able
to
have,
let's
assume,
that
I
have
production
cluster
back
of
his
closet
or
staging
cluster
whatever,
and
actually
with
Olga
city.
You
install
Olga
city
only
in
one
of
your
clusters
and
you
teach
Argos
82
to
reach
out
to
the
other
clusters.
So
I
think
this
is
a
real
nice
feature,
because
personally,
if
I
administer
cluster
I,
don't
want
any
extra
resources
on
my
production
clusters.
A
Obviously
so
we'd
all
go
see
that
you
can
do
that
and
I
think
this
is
what
we
should
aim
for
with
the
agent
as
well
to
do
not
be
required
to,
instead
to
all
the
costs
are
just
how
it
in
one
of
your
clusters
already
inside
your
internet
and
it
can
reach
the
other
clusters
even
with
cost.
What
actually?
This
is
Margot
lab
current
command?
A
This
integration
might
not
be
such
a
huge
issue
for
self-managed
instances,
so
this
is
one
thing
that
I
think
is
really
nice,
Naga
City,
the
other,
is
it
focuses
really
on
deployment.
I
think
that
they
made
the
right
choice.
Supporting
many
ways
of
configurations
of
the
deployments
like
the
simple
cube
city
are
many
fads,
manifests
customized
harm
case
on
that
etc.
So
I'm
pretty
sure,
sooner
or
later,
we'll
have
to
do
that
as
well
and
as
a
result,
I
would
instead
use
Argos
edy
behind
the
hood
or
make
it
super
simple
to
use
Argo
city.
A
One
thing
that
we
should
not
use
I
mean
that
we
should
not
encourage
to
be
used,
but
rather
copy
or
make
something
much
better
is
the
UI
they
have.
They
have
the
UI
something
similar
to
what
we
call
deployment
boards.
Just
their
UI
is
much
better,
which
is
a
very
poor
thing
to
say,
but
that's
the
truth
so
they're,
basically,
you
can
see
the
who
set
up
with
this
is
your
deployment.
This
is
a
service
here.
A
You
have
these
pods
attached
what
port
and
you
can
see
everything
in
much
more
details,
so
I
would
envision
our
deployments
board
in
this
way.
This
was
a
topic
around
Argos
city
and
the
proof
of
concept
that
we
are
currently
running
and
deployment
boards
actually
lead
us
to
a
different
topic,
which
is
environments
and
star
form.
A
Currently,
if
you
have
to
run
manual
apply
step,
then
you
have
to
find
the
pipeline
that
has
that
given
apply,
step
and
click
on
it
to
to
apply
or
things
instead
of
this,
it
would
be
much
nicer
to
be
able
to
say
that
okay
I
have
my
deployment
board
phase.
My
staging
here's,
my
production
and
I
see
that
some
items
are
waiting
on
the
staging
or
on
the
production
environment
to
be
applied,
of
course,
which
are
for.
We
can't
really
say
propagates.
A
These
commits
from
staging
to
production,
because
those
are
usually
different
directories
and
you
want
to
merge
those
on
purpose,
but
even
applying
them
to
making
the
final
manual
step
would
be
really
really
beneficial,
I
think
so
to
have
a
UI
for
that
and
third
step.
A
third
topic
that
is
that
I
want
to
highlight
is
again
related
to
the
to
the
broker.
A
The
document
is
broker
that
we
are
currently
building,
which
is
I,
think
we
should
somehow
be
able
to
tie
it
into
the
I,
wouldn't
say,
get
lopsy
IMO,
but
but
to
define
everything
through
code
and
leverage
the
values.
The
code
that
did
lab
already
has,
thanks
to
you,
thanks
to
develop
CI
yeah,
more
follow
so
somehow
leverage
it,
but
I
think
we
don't
want
to
use
the
CI
a
more
actually.
A
Finally,
fourth
topic
is
what
we
spoke
about
in
case
of
the
manna-fest
project
in
during
design
sprint,
that
we
envisioned
kubernetes
deployments
running
using
a
manifest
project
and
actually
have
a
separate
application
project
and
you
deploy
from
the
manifest
project.
I
think
this
is
a
great
way
forward
and
something
that
I
have
seen
many
cases
with
enterprise
users
and
they've
seen
only
one
different
case.
A
The
weave
discussion
I
had
last
week
where
they
had
a
single
project
lecture.
They
are
doing
project
work,
so
everything
is
a
single
project
there
and
I
would
like
to
extend
this
idea
of
a
manifest
project
to
all
our
infrastructure
related
code
bases,
which
would
mean
that
if
I
start
perform
project,
then
that
is
already
an
infrastructure
project
and
whatever
kind
of
project
I
create.
We
should
always
think
about
it
that
this
is
my
business
code,
and
here
I
have
some
kind
of
infrastructure
related
code
with
that
and
make
it
simple
to
be
used
accordingly.
A
So
there
are
always
kind
of
this
parent
child
pipelines
that
are
actually
a
premium
feature
right
now.
So
if
you
leverage
those
that
means
that
we
are
already
creating
premium
features
by
default,
so
I
think
this
last
part
is
a
bit
still
messy,
but
together
we
can
figure
it
out.
So
I
have
two
four
topics.
These
will
be
discussed
in
this
described
in
this
issue
as
well
for
discussion.
A
First
of
all,
what
can
we
learn
from
Argos
CD,
because
I
think
we
should
learn
from
them
and
boro
good
ideas
from
them
and
perhaps
even
use
them
behind
the
hood?
Second,
having
deployments
boards
I
first
taken
over
from
Argos
CD
second
for
tower
form
as
well,
then
having
the
manifest
projects
for
every
kind
of
deployment
job,
and
this
would
actually
even
allow
us
to
say
ku-band.
This
integration
is
just
a
trapa
for
project
here
and
it's
transparent
and
you
are
free
to
hack
it
away.