►
Description
This is a discovery and scoping discussion about how we can improve the project import process with sample data that can benefit both the quality performance tests and customer-facing demos with rich data. The quality team has rake tasks that use the Faker package to generate large quantities of sample data and the Customer Success team uses a custom built Demo Designer UI application and database that then uses the GitLab API to import the created records into a GitLab instance.
The mutual outcome of this discussion is the creation of a "demo-data-as-code" approach using YAML or JSON that is sequenced similar to how CURL calls to the API would be. This allows different ways to create the demo data with different tools but a common output that can then be parsed by GitLab's project import and future group import abilities.
Learn more in the issue.
https://gitlab.com/gitlab-com/www-gitlab-com/-/issues/7397.
Participants:
Mek Stittri (Director of Quality Engineering)
Jeff Martin (Sr Demo Systems Engineer - Customer Success)
Kyle Wiebers (Quality Engineering Manager, Engineering Productivity)
A
Great
thank
you
for
that.
So
just
to
recap
that
the
folks
who
were
watching
the
recording
this
is
a
conversation
between
us,
quality
and
and
Jeff
was
leading
the
demo
systems
from
home
sales,
and
we
are
figuring
out
how
we
can
work
together
to
create
a
mechanic
so
that
we
can
spin
up
demo
projects
and
also
tested
it
easily
in
his
lab
as
a
feature
and
I
think
we
were
discussing
how
to
structure
this
so
I'll
pass
it
on
to
Jeff
to
run
through
what
he
was
a
pointing
out
just
now.
Thank
you
sure.
B
A
B
We're
we're
coming
out
from
the
approach
of
this
is
right
now
the
QA
team,
in
the
quarter,
you
for
quality,
your
QA
quality.
Okay,
the
quality
team
has
some
demo
data
generators
that
are
very
comparable.
What
you'd
see
with
the
unit
testing,
where
it's
doing
fake
your
data
and
things
of
that
nature
and
generating
a
lot
of
you
know
random
data
to
see
you
know.
B
You
know
everything
kind
of
in
a
current
state,
so
you
could
see
what
a
full
get
live.
Implementation
looks
like
without
actually
having
to
implement
it
yourself
and
get
in
quickly.
One
of
the
constraints
that
we're
faced
with
is
that
works
well,
for
someone
is
the
first
time
ever
seen,
get
lab
and
they
just
want
to
tinker
around
in
a
sandbox
for
our
customer
success
team.
B
The
marketing
is
put
out
as
well
as
just
some
of
the
story
that
we're
trying
to
tell
and
the
main
challenge
that
we're
faced
with
is
when
you
use
faker
data,
you
don't
have
control
of
the
story
being
told
and
so
we're
discussing
what
are
the
options
of
creating
what
that
story
looks
like
our
Acme
company
and
with
the
products
the
Acme
company
has
and
what
that
looks
like
for
you
know
four
of
the
samples
that
we're
going
with
and
so
we're
playing
with
ideas.
This
is
all
work
in
progress.
B
It's
all
in
pencil
right
now,
but
that's
just
kind
of
the
recap
of
where
we're
at
so
far
so
picking
back
off
where
we
left
off
mech.
So
if
we
go
and
look
at
demo
data,
how
does
it
imported
us?
It
in
port
I
think
one
of
things
we
should
look
at
is:
can
we
build
off
with
that
get
lab
CI
file
from
the
infrastructure
as
code
type
of
design
model,
so
that
no
matter
how
the
data
gets
generated
right?
It
could
be
that
a
unit
test
level
all
this
gets
generated
right
now.
B
A
Right
that
could
bring
that
in
that
really
speaks
to
to
me.
I
think
I.
Think
I
mentioned
in
the
issue
as
well,
like
I,
was
envisioning
data
as
code,
where
the
the
language
of
describing
the
demo
data
and
I
also
want
to
say
it
is
really
below
into
our
test
data
right
now.
Our
test
data
is
very
fragmented
yeah.
We
don't
have
a
UI
that
can
like
hey.
A
This
is
like
test
user
one
and
all
that
and
start
to
like
menu
plate
on
a
design
aspect,
not
not
implementing
so
coming
into
the
top
level
view
it
gets
great,
so
yeah
I
think
that
really
speaks
to
us.
I
think
I
think
I
a
a
yeah
mo
structure,
ish
that
describes
the
structure
of
like
groups,
projects,
users
and
how
the
CI
is
configured
I,
think
that's
a
great
start
and
then
what
is
that
it's
kind
of
like?
Would
that
be
a
working
solution?
You
think.
B
I
think
so
you
know
right
now,
I
feel
more
comfortable
on
the
JSON
side
of
things
that
I
do
in
the
yeah
mo
just
because
we're
working
with
an
API.
So
my
head
stays
there,
but
I
understand
that
the
mo
is
basically
a
lot
of
what
they
get
loud.
Be
a
competency
is,
in
terms
of
a
community
I'm
curious
to
connect
the
dots
on
this
to
some
of
the
code.
B
So
when
we
look
at
this
all
we're
doing,
is
we
have
on
the
demo
designer
a
database,
my
sequel
database,
that
has
the
values
in
it
and
then
we're
using
reference
objects
of
what
it
ties
to
right.
So
we
actually
go
in
and
we're
creating
these
different
objects.
This
is
where
I
think
the
dependency
design
challenge
is
going
to
cause
problems
for
us.
B
Did
I
kill
it
out
there?
It
goes
when
I
look
at
a
playbook
and
I'm
trying
to
create
an
issue
you're
choosing
a
lot
of
dependencies
right,
so
you're
choosing
milestones
that
already
exist,
you're,
choosing
ethics
you're,
choosing
labels
that
exist,
one
of
the
things
that
I
think
we'll
run
into
with
any
sort
of
yamo
or
JSON
or
inserted
text
is
the
parent
object
being
created,
and
this
is
why
I
mentioned
terraform,
because
terraform
it
has
a
parsing
engine
to
understand
what
the
dependent
these
are
and
I
worry
about
that
from
a
demo.
B
Data
perspective
is
if
we
need
a
dependency
management
system
of
some
kind,
because
you
can't,
you
know,
create
an
issue
with
something
and
then
you
know
no
type
of
label
and
ylabel
doesn't
exist.
We
can
do
it
auto
detection,
but
then
you
end
up
inconsistencies
right.
You
might
end
up
with
instead
of
the
six
labels
we
have
here,
you
end
up
with
25
random
labels.
A
faker
came
up
with
right,
yep,
that's
the
main
concern.
I
have
right.
A
That
I
I
think
we
just
have
to
pick
one
data
model
and
see
the
implementation
can
solve
it,
and
if
that,
if
that
one
data
model
it
be
so
bald
that
I
think
we
can
solve
probably
a
majority
of
the
other
models.
But
that's
a
great
that's
a
great
call
out
this
there's
a
PHP
engine
that
you
have
right
now
is
it?
Is
it
just
like
a
creation
wizard
for
additions.
B
Window
models,
correct
yeah,
and
so
it's
it's
just
providing
a
UI
to
build
a
data
model
that
then
gets
parse
through.
It
runs
API
calls
right,
and
so
this
is
the
Creed
project
issues
here
it's
going,
and
this
is
right
here.
This
is
all
date,
calculations
right-
and
this
is
looking
at
all
those
dependencies.
B
If
it
has
a
milestone,
go
get
the
milestone,
get
the
idea,
the
milestone
from
the
gate,
lab
database
grab
that
milestone,
ID
and
then
use
the
you
know
the
SDK
to
then
populate
it
in
with
those
ID
numbers
or
with
those
relative
calculated
dates
right
now.
What
we
can
do
with
this
is
instead
of
opting
an
API.
This
could
then
I
could
do
a
you
know.
Do
you
want
to
import
this
into
gate
lab
using
the
API?
B
Or
do
you
want
to
export
this
as
a
JSON
or
yamo
file
and
then
parse
this
out
and
then
I
can
use
the
engine
I
have
because
I
already
know
what
my
dependencies
are
here
to
sequence
it
so
it'll
always
ensure
that
from
top
to
bottom,
all
your
dependencies
are
up
top
and
all
your
you
know,
your
other
records
are
down
below
right.
So
it
knows
it
already
exists.
A
Good,
ok,
one
option.I
idea,
I,
wasn't
planning
dive
into
the
technical
details,
were
happy
to
I.
Think
that
we
could
less
the
complexity
here
if
we
were
to
help
enhance
the
export/import
mechanism
to
make,
for
example,
the
dates
like
the
create
dates,
all
that
could
essentially
be
dynamic
and
what
you're
dealing
with
on
this
side
is
essentially
relative
numbers.
For
example,
if
you
were
to
create
a
project
right
on
you
soil,
instead
of
due
date,
yeah
instead
of
setting
a
static
date,
it
could
be
like
say
due
date,
two
weeks
after
the
project
was
created.
A
B
And
I
think
that's
what
we
do
here.
That's
what
this
whole
system
is
about
right
as
we
as
January
1st
2020
as
our
today
date.
When
are
you
doing
import
whatever
you
plug
in
it
does
a
sub
day
or
an
ad
date
based
on
that
right
got
it
yep
yeah,
so
so
I
think
what
we've
we've
crossed,
that
bridge
already
to
some
degree
right.
A
Yeah,
so
if
we
were
to
make
this
in,
if
we
want
to
leverage
the
rake
tasks
and
the
import
export,
the
employee
export
should
be
able
to
handle
this,
and
is
that
the
concern,
because
I
come
from
reading
that
the
blocker
so
far
to
use
to
pick
the
things
that
are
preventing
us
from
using
import/export
is
the
making
sure
the
dynamic
nature
of
things
are
created
correctly.
That's
right!
Now,
it's
a
very
high
touch
from.
B
From
that,
it's
what
it's
one
of
several
challenges.
Yes,
so
the
dates
are
part
of
it.
The
thing
is
that
the
dates
are
going
to
appear
in
different
places,
and
so
the
question
would
be
is
when
you
import
it
in
you'd
almost
need
some
sort
of
mechanism
to
read
through
ok,
this
projects
had
a
start
date
of
July
1st
of
2019,
and
it
has
timestamps
that
are
all
throughout
it.
What
you
want
to
understand,
then,
is
you
want
to
have
some
future
casting
right?
B
So
is
your
date,
the
date
it
was
exported
is
that
can
become
today's
value.
Or
is
it
one
where
you
want
to
say
I,
let
things
populate
through
here,
but
I
need
to
go
or
do
a
reset
in
time,
and
that's
the
challenge
that
you're
going
to
run
into
is
how
do
you
choose
today's
value?
Do
you
let
the
user
pick
right
and
then
how
do
you
dynamically
shift?
Those
dates
and
how
do
you
shut
them
across
all
records
with
the
right
dependency
controls,
I.
Think.
A
A
I
think
we're
just
gonna
ingest
with
what
what
the
demo
data
structure
is,
but
in
when
you
create
them,
create
the
in
the
get
lab
feature
version
you
can
either
create
it
Ben
or
you
can
create
it
and
specify
a
date
where
that's
the
global
date
that
everything
else
is
gonna
pick
it
back
from
and
thinking
from
the
technical
aspects
that
we
need
to
augment
the
import
export
mechanism.
Add
the
ability
to
to
set
those
dates
from
the
start.
In
those
controllers,
that's
right
now
is
theirs.
A
It
weighs
2018
I'm,
just
gonna
create
2018,
and
then
it
has
to
be
like
if
elsewhere,
okay,
oh
this
says
this
is
under
the
export
demo
project,
our
test
project,
we're
gonna,
create
it
on
this
date
and
and
then
there's
some
logic
to
like.
If
not
if
it's
not
certified
use
today's
date,
it
was
specified
use
that
day.
I
think
that
that's
like
the
main
crux
at
the
time,
yeah.
B
And
that
that's
what
I
try
to
solve
this
iteration,
it's
working
for
what
I'm
doing
today
right,
and
so
you
can
go
through
the
exercises
that
you
need
to
to
to
think
through
the
engineering
challenge
on
the
on
the
rail
side
of
it.
But
it's
the
relative
date
does
make
it
easier
to
do
an
ad
and
sub.
It's
just
these
for
the
sprawl
of
all
the
dates
that
you're
going
to
run
into
some
challenges
with
and
catching
all
them.
B
It's
the
big
thing,
the
other
piece
that
we're
also
working
on
that
is
beyond
the
scope
of
our
first
iteration
of
a
project.
Template
is
we're
looking
at
ongoing
project
demo
data,
where
we
actually
have
scheduled
events
that
happen
so
a
background.
Job
is
looking
for
that
relative
date
and
says
in
six
days
from
now.
It's
going
to
do
it
commit
at
you
know,
1
p.m.
3
p.m.
and
5
p.m.
with
face
pipeline
jobs,
so
we're
also
doing
scheduled
future
dating
as
well
for
an
ongoing
perpetual
demo
right.
B
A
Ci
schedule
is
something
of
a
weak
spot
in
and
I
would
even
say
the
test
data
coverage,
because
right
now
it's
just
create
under
fly.
Something
that's
like
in
the
background
running
is
something
you
want
to
solve,
even
in
our
staging
environments,
yeah
for
context,
so
I
think
I
think
I'm
totally
I
think
we
realign
on
that
yeah
that
takes
care
of
the
employee
X
for
my
cancer
mandates.
What
else
is
the
other
thing
that
we
need
to
touch
on
in
terms
of
the
missing
Export
Import
pieces?
B
There's
quite
a
bit
of
it,
there's
a
piece
of
technical
discovery:
I,
don't
know
what
I
don't
know
at
the
moment,
but
you
know,
and
it
look
like
what's
in
place
right
now
that
I've
tested
you
know
the
issue
is
issue
boards
and
labels.
They
they're
all
working,
just
fine
right
when
I
get
into
some
of
the
branches
in
the
commits
I
might
run
into
API
weaknesses
of
what
I'm
working
on
that
I.
Don't
know
right
and
I
might
find.
B
You
know
how
do
you
dynamically
modify
certain
things
now,
I
think
this
is
where
I'm
going
to
look
at
some
of
how
the
performance
tools
that
you,
your
team,
has
built.
You
know
how
they
do
it
you
know
is,
is
can
I
adopt
that?
Do
we
need
something
differing
Ally
I
haven't
gone
through
that
discovery
process,
yet
to
know.
B
A
B
You
know
obviously
the
roadmaps
and
such
or
generate
a
base
from
the
date
values
and
the
epics
and
the
issues,
but
that's
one
of
the
big
things
we
have
to
look
at
is
how,
let's
just
say:
let's
not
underestimate
how
much
the
group
might
involve
some
more
complexity
than
we
think
it
is
today.
Okay,
yeah.
B
B
B
You
know
it's
import
from
template
area
and
basically
that
button
there
is
going
to
read
those
text
files
and
then
we
can
have
a
repository
with
multiple
files
in
it
and
get
love
as
a
product
will
simply
read
from
those,
and
we
can
use
source
control
to
maintain
those
templates
bingo.
Yes,
that
is
that
that
lineup,
clearly
yeah.
B
A
I
think
it's
a
great
addition.
The
way
we
are
the
way
we
treat
dependencies
right,
I
think
it
lab
the
planet.
She
will
treat
those
output,
those
text
file,
demo
test
data,
definitions,
yeah
as
a
dependency,
and
you
can
see
if
we-
if
we
do
this
right
few
iterations
and
now
it
might
even
have
to
be
like
hey.
This
is
like
a
release
1.0
that
gets
packaged
with
get
lab
13
of
five
and
yeah.
Now
you
have
this
list
of
demo
projects
you
can
just
like
and
then
delete
it,
and
then
it's
gonna
be
ongoing.
A
B
A
A
So
it
might
be
in
a
shape
of
a
working
group
when
we
get
more
traction
of
it
right
now,
I
I'm
still
in
the
I,
think
it's
still
in
the
requirements
gathering
face
and
planning
out
what
we're
gonna
be
building
and
I
think
we
are
like
80
percent
there
in
terms
of
requirements
ish.
So,
let's,
let's
recap:
we
have
one
minute
left
so
Jeff
get.
If
you
can
send
us
that
one
project,
you
think
it's
great
candidate
to
cover
like
the
first
iteration
of
what
you
want
to
see.
A
B
So,
adding
on
to
that
is,
you
know,
give
you
a
group,
so
you
can
see
what
a
group
looks
like
when
it's
important
you
can
work
on
the
requirements
around
how
groups
get
important
I
can
give
you
the
sample
that
link
house
right
now.
I,
don't
think
that
should
be
the
gold
standard,
but
it's
something
that
least
get
a
copy.
You
do
a
proof
of
concept
with
as
a
baseline
right,
there's,
definitely
a
story,
that's
still
being
drafted
around
it
and
so
we'll
end
up
with
the.
What
is
that
right
approach
in
the
coming
weeks?
B
The
other
aspect
is
from
my
side:
I
will
work
on
exporting
in
Tamil
and
then
give
you
samples
in
the
mo
and
saying
here's
the
mo
and
then
here's
the
ticket
lab
project
is.
Can
you
get
this
to
match
an
import
and
match
that
test?
I'll
take
a
first
stab
at
the
yeah
mo
structure,
and
then
we
can
do
some
iterations
upon
naming
conventions
or
organization
whatever.
B
B
Again
sure,
yeah
and
I'm
starting
a
sprint
next
week,
it
might
be
you
know,
later
next
week,
I've
had
a
couple
delays
for
other
unrelated
projects,
but
my
goal
is
to
end
up
with
the
proof-of-concept
with
this
by
the
end
of
June
or
the
first
week
of
July,
and
so
the
hope
is
it
from
a
progress
milestone
at
least
we'll
have
all
those
pieces.
I
do
not
know
the
mo
output.
A
B
So
what
what
I'll
end
up
happening
is
is
twofold,
so
there's
a
common
toward
the
bottom,
it's
it's
all
the
same
issues
that
sometimes
the
timelines
moved
around.
So
the
adding
of
project
and
group
level
API
to
the
demo
designer
the
section,
the
API
resources
for
demo
designer
in
that
doc.
I'm
gonna
copy
it
into
that
issue,
but
it
shows
you
which
API
resources
I'm
going
to
be
building
for
the
June
2020
iteration,
okay,.
B
The
get
live,
API
and
so
I
actually
go
to
the
get
lab.
Api
resources,
documentation,
so
everything's
built
upon
that
and
all
I'm
doing
is
copying
that
data
model
onto
the
PHP
side
for
the
dev
designer
and
allowing
that
to
create
all
the
resources
and
then
populate
them
in
a
dynamic
generated
way
right.
So
that's
all
it's
happening.
B
So
all
the
things
that
I'm
putting
in
as
functionality
are
there
under
the
sub
bullets
in
the
dock,
that'll
move
into
the
issue
in
the
coming
days
and
then
the
ad
generated
sample
data
using
the
marketing
persona
is
the
second
half
of
that
dock.
In
terms
of
what
the
story
we're
telling
who
the
fake
company
the
products
you
know
what
the
products
are
doing,
who
the
the
the
people
are
that
are
performing
the
actions
commits,
etc.
B
A
Yeah,
it
has
me
thinking
whether
if
I
know
we
were
discussing
break
time
and
I
want
to
dig
in
to
do
it
further,
maybe
with
the
help
of
the
development
folks,
even
if
the
import-export
is
limited
in
functionality.
Maybe
this
this.
This
provisioning
mate
has
and
should
just
be
like
a
task
inside
get
labid
calls
all
the
external
API
s
if
the
API
is
rebuilding
have
already
those
added
functionality.
A
B
You
gotta
tell
all
the
code
I'm
building,
that's
what
that
does
so
feel
free
to
use
that
as
a
model
for
looking
through
that,
because
an
eventualities,
a
future
iteration
could
be
good
a
project
import,
and
it
gives
you
a
summary
of
all
of
the
things
in
this.
Like
here's
all
the
groups
that
are
there,
there's
all
that
and
has
checkboxes
and
you
choose
what
to
import
exactly.
B
A
I
don't
want
demo
day,
I
want
quality
test
data.
It's
like
I
want
test
data
from
quality,
and
this
is
another
checkbox,
any
yes,
yeah,
okay,
so
I
think
I
think
we
have
action
here.
So
let's
this
run
point
on
on
that
and
now
I
will
also
give
this.
According
to
Tim,
Tims
Alan
and
the
the
team
is
currently
working
on
the
import.