►
Description
In our thirteenth Dagster Community Meeting, the Dagster team showed new 0.15.0 and Dagster Cloud updates.
👨🏫 Agenda 👩🏫
Introduction: 0:00
Data Engineering with Dagster Course: 1:45
Community Spotlight: 3:00
Software-defined assets update: 4:29
Ops and Graphs with Software-Defined Assets: 13:13
Orchestrating dbt and python using Dagster: 26:13
Dagit Improvements: 38:36
Dagster Cloud Update: 48:03
Q&A: 52:41
🗞 Related blog posts 🗞
Dagster 0.15.0: Cool for the Summer: https://dagster.io/blog/dagster-0-15-0-cool-for-the-summer
Orchestrating python and dbt with Dagster: https://dagster.io/blog/dagster-0-15-0-dbt-python
🌟 Socials 🌟
Checkout (and star!) our Github ➡️ https://github.com/dagster-io/dagster
Check out our Documentation ➡️ https://docs.dagster.io/
Join our Slack ➡️ http://dagster-slackin.herokuapp.com
Visit our Website ➡️ https://dagster.io/
Follow us on Twitter ➡️ https://twitter.com/dagsterio
A
So
one
quick
announcement
first
is
that
we
have
a
couple
roles
that
we're
currently
hiring
for.
One
of
them
is
a
developer
advocate
which
would
be
with
developer
relations
team,
and
you
know
about
creating
educational
content
around
dagster,
helping
people
get
onboarded,
and
so,
if
you're,
someone
who's
excited
about
using
dagster
and
it's
curious
to
and
then
interested
in
making
that
easier
for
other
people,
please
reach
out
to
me
on
slack,
and
there
is
also
a
sales
engineer
opening
as
well
so
reach
out
to
us.
A
If
those
seem
interesting
and
another
exciting
announcement
is
that
we
are
partnering
with
co-rise
to
do
a
data
engineering
with
dexter
course,
so
the
person
who
will
be
teaching
that
course
dennis
hume
is
actually
here
with
us,
he's
a
staff
data
and
here
at
duchy
and
he's
gonna,
say
a
few
words
about
the
class.
So
dennis.
Do
you
wanna
tell
us
more
about
this.
B
Yeah,
so
I'm
really
excited
to
do
my
first
co-rise
class.
I'm
very
excited
that
it's
on
dagster
I've
been
doing
dagster
for
about
two
years
now.
So
a
lot
of
the
course
is
just
kind
of
bringing
some
of
those
practical
use
cases
of
bagster
and
it's
not
entirely
just
me
stealing
dagster
documentation
and
putting
it
in
a
different
format.
B
So
a
lot
of
the
course
is
going
to
be
project
based,
and
it's
really
just
that
the
goal
is
a
really
helpful,
like
zero
to
one
of
kind
of
the
role
of
orchestration
in
a
data
platform
and
kind
of
how
you'd
go
about
building
that
into
extra.
So
the
course
is
in
august
and
yeah,
I'm
really
looking
forward
to
it.
A
Okay,
so
I
will
do
a
community
spotlight.
What
I
wanted
to
start
doing
at
community
meetings
is
start
showing
off
some
of
the
content
that
our
users
have
created
since
the
last
meeting.
So
actually
there's
a
couple
things
out
from
dennis
who
just
spoke.
He
recorded
a
video
from
his
big
data.
Madison
talk
on
modern
data
platform,
orchestration
with
daxter,
which
is
now
on
our
youtube
and
a
four
reasons:
duchy
engineers,
love
dexter
cloud
blog
post.
A
A
C
C
So,
first
of
all,
we
made
a
bunch
of
improvements
to
dexter
core
we've
got
job
level
retries
and
improved
partitions,
backfill
ui,
a
set
of
improvements
at
a
python
api
and
much
more
second
software-defined
assets
are
now
fully
stable
and
ready
for
prime
time
so
clare,
and
I
are
going
to
spend
our
time
today
on
this
second
bullet
point.
But
if
you've
got
any
questions
about
the
other
parts
of
the
release
feel
free
to
ask
them.
C
So
software
defined
assets
being
ready
for
primetime,
brings
up
three
big
questions
and
in
the
next
few
minutes,
I'm
going
to
try
to
give
high-level
answers
to
each
of
these
they're.
When
should
you
use
software-defined
assets
oops?
Why
should
you
use
software-defined
assets
and
how
should
you
use
them
before
we
jump
in?
I
want
to
make
a
few
things
clear
at
the
outset.
C
The
first
is
that
op
graph
and
job
are
not
going
away.
Software
defined
assets
aren't
a
replacement
for
dagstr's
core
community
computational
concepts.
Ops
are
actually
the
core
unit
of
computation
that
occurs
within
an
asset
instead
think
of
software
defined
assets
as
a
layer
on
top
that
links,
ops,
graphs
and
jobs
to
the
long-lived
objects
that
those
ops,
crafts
and
jobs
interact
with.
C
C
C
Here's
another
one,
every
hour,
add
a
partition
to
the
events
table,
so
software
defined
assets
aren't
only
for
situations,
you're,
dropping
and
recreating
an
asset.
This
situation
is
still
a
good
fit
we're
just
updating
a
part
of
an
asset,
but
we
know
what
asset
we're
gonna
update
ahead
of
time.
We
know
a
part
of
the
asset
we're
gonna
update
when
we
launch
for
run
so
still
a
great
fit.
C
Okay.
A
third
example
every
day
send
emails
to
a
set
of
users,
so
this
is
a
bad
fit
for
software-defined
assets,
because
there's
no
assets
here.
The
goal
of
our
job
isn't
to
update
long-lived
objects
within
our
data
platform
is
to
send
emails,
but
you
can
still
use
a
vanilla,
dag
search
job
for
this.
C
C
C
C
Another
benefit
using
software
defined
assets
is
that
allows
you
to
kick
off
and
monitor
runs
from
the
perspective
of
the
assets
that
those
runs
are
trying
to
update.
So
if
you
got
some
table
in
this
case
a
table
called
stories,
you
notice
it's
out
of
date.
You
don't
have
to
go
searching
around
for
the
job
that
generates
the
stories
table.
You
can
just
navigate
directly
to
the
page
that
corresponds
to
the
stories
table
and
then
click
the
button
to
re-materialize
or,
if
you're
wondering
why
this
table
is
out
of
date.
C
Last
of
all,
using
software-defined
assets
often
means
that
your
code
ends
up
more
concise.
That's
because
specifying
the
inputs
to
a
software-defined
asset
defines
the
upstream
asset
that
it
depends
on.
So
you
don't
need
to
use
the
graph
or
job
decorators
to
explicitly
wire
up
your
dependencies.
They're
implicit
in
the
asset
definitions
themselves,.
C
C
The
long
answer
is
that
you
should
check
out
the
upgrade
guide
in
our
docs.
It
has
side-by-side
code
comparisons
that
show
jobs
before
and
after
using
software-defined
assets
and
I'll
post
a
link
to
that
in
the
zoom
chat
after
I
finish
talking,
but
I
want
to
give
a
few
top
level
guidelines
to
to
think
about
as
you,
as
you
add,
software-defined
assets
into
your
code
base.
C
The
second
one
is,
you
don't
need
to
go
all
at
once
start
by
using
software-defined
assets
for
just
one
of
your
jobs
and
then
go
from
there,
and
the
third
is
that
you
don't
even
even
need
to
convert
all
of
your
ops
to
assets.
A
common
pattern
that
we
see
in
dijkstra
jobs
is,
there
will
be
an
op
that
produces
an
asset
and
it
will
depend
on
another
op
that
doesn't
produce
an
asset.
C
D
Okay,
so
just
to
give
a
little
bit
of
context
in
o
fourteen-o,
we
we
introduced
a
reusable
data
entity
and
that's
the
software
defined
asset,
and
what
I
mean
by
reusable
here
is
that
if
you
have
an
existing
data
entity,
you
can
use
it
as
an
upstream
data
source
to
define
new
and
similar
downstream
assets.
So
one
example
here
is
this
transactions
asset
that
contains
rows
of
transactions,
including
data
such
as
the
date
amount,
vendor,
etc.
D
And
then
we
have
a
downstream
asset
here
that
accepts
transactions
as
an
upstream
asset.
The
asset
is
called
total
spendings
by
date,
and
that
will
take
the
transactions
data
frame
and
aggregate
that
to
calculate
the
total
spending
for
every
single
given
date.
So
then,
here,
we've
defined
a
an
asset
dependency
relationship
here.
So
there
are
a
couple
limitations
that
we
realize
about
defining
assets
via
acid.
One
limitation
is
that
complex
computation
is
difficult
to
decompose
within
a
decorated
function.
D
So
in
o
15,
no,
we
introduce
graph
back
assets
and
what
this
means
is
that
there
is
now
a
utility
method,
assets
definition.from
graph-
that
takes
in
a
graph
and
creates
an
assets
definition
from
this,
and
this
is
valuable
for
a
couple
reasons.
Some
of
the
use
cases
that
you
might
look
for
to
to
use
a
assets
definition
dot
from
graph
or
graphic
assets,
is
to
one
decompose,
complex
computation
into
multiple
odds.
D
So
if
you
define
a
graphite
asset,
you
can
use
the
ops
within
it
in
other
assets
and
in
other
non-acid
jobs,
and
so
you
don't
need
to
choose
between
using
ops
and
graphs
or
assets.
You
can
use
option
graphs
and
build
assets
from
those
existing
components,
and
so
you
can
also
incrementally
migrate.
Existing
opts
and
graphs
to
assets
through
our
new
api,
so
I'll
jump
into
a
demo
about
reusing
operations
across
multiple
assets
and
reusing
acids.
Just
to
make
this
a
little
bit
more
clear.
D
Okay,
so
say
that
we
are
an
airline
and
we've
recently
made
optimizations
to
reduce
the
number
of
layovers
that
passengers
might
experience,
and
we
made
this
optimization
in
2022,
and
so
we
want
to
assess
whether
the
number
of
layovers
in
2022
was
truly
less
than
the
number
of
layovers
in
the
past.
Historically,
we
have
a
couple
existing
assets
here.
The
first
one
is
passenger
flights,
which
contains
data
such
as
the
starting
and
ending
airport,
the
number
of
layovers
total
amount
spent,
etc.
D
And
then
we
have
a
downstream
asset
here.
That
is
the
u.s
layover
percentages,
and
what
that
contains
is
a
data
frame
that
breaks
down
the
percentage
for
x
number
of
layovers,
for
example,
like
50
of
flights
had
zero
layovers
25
percent
had
one
layover
so
on
and
so
forth,
and
we
want
to
ideally
use
this
existing
asset
to
create
our
2022
layover
breakdown
asset,
because
it
looks
like
a
lot
of
the
components
that
we
already
need
already
exist
within
this
asset.
D
So
there's
this
u.s
flights
data
frame
and
if
we're
looking
to
calculate
the
2022
flights,
we
can
filter
on
this
data
frame
to
get
back
that
result.
So
we'd
like
to
ideally
output
this
as
an
asset,
so
we
can
reuse
it
and
then
secondly,
there
exists
a
little
bit
of
computation
here
that
will
calculate
the
layover
breakdown
percentages,
given
a
data
frame
in
this
case
u.s
flights.
So
let's
refactor
this
so
that
we
can
make
these
components
reusable.
D
D
And
if
we
want
to
add
this
graph
as
a
graph
back
acid
to
our
repository,
we
can
use
our
assets
definition,
dot
from
graph
method
and
input
the
u.s
assets
graph,
and
what
this
method
will
do
is
it
will
assume
that
every
single
output
of
your
graph
here
is
an
output
asset?
So
in
this
case
this
graph
will
output
two
assets,
u.s
flights
and
u.s
layover
percentages?
D
D
D
D
So
let's
go
ahead
and
add
this,
and
now
that
we
have
the
u.s
flights
asset
that
makes
this
ant
and
the
layover
percentage
breakdown
off.
That
makes
this
pretty
easy,
so
we
can
define
a
new
graph,
let's
call
it
layover
breakdown,
2022
and
that
accepts
u.s
flights
as
an
input
asset
first.
Well,
we
would
want
to
return
back
the
layover
percentage
breakdown
of
the
u.s
flights
filtered
for
2022.
D
D
D
D
Cool
so
we've
successfully
rematerialized
this,
so
we
have
all
the
data
that
we
need
to
make
the
comparison
between
2022
and
the
pest
daggett
also
offers
a
lot
of
really
valuable
views
in
order
to
understand
the
relationships
between
your
assets
and
the
ops
that
underlie
them.
So
there's
this
functionality
here
to
toggle
viewing
as
an
asset
graph.
D
So
if
I
turn
this
off,
then
I
can
see
all
the
graphs
that
that
define
the
computation
to
generate
my
assets,
and
in
this
case,
if
I
open
up
the
u.s
assets
graph
that
I've
defined
here,
then
I
can
expand
this
to
view
the
ops
that
are
necessary
to
generate
my
assets.
And
then,
if
I
explode
the
graphs,
then
I
can
also
see
the
full
list
of
ops.
D
If
I
click
on
an
op,
for
example,
layover
percentage
breakdown,
I
can
see
every
single
place
where
that
op
is
invoked,
and
so
that
makes
it
really
useful
so
that
if
I
want
to
make
a
change
to
an
auth,
that's
used
in
many
places,
I
can
see
the
the
impact
that
it
will
have
on
my
data
assets
cool.
That
is
it
from
me.
C
Yeah,
so
so
in
the
world
of
software
defined
assets,
they're,
basically
two
sides
of
the
same
coin.
So
when
you
say
I
want
to
materialize
an
asset
that
will
launch
a
run
and
the
run
will
execute
your
code
that
produces
the
contents
of
the
asset
and
then
log
and
asset
materialization
to
say
that
that
to
say
that
that
asset
was
materialized.
A
B
E
All
right,
hopefully,
you
can
see
my
screen,
because
if
you
can't
this
is
pre-recorded,
and
I
can't
fix
it-
my
name's
owen,
I'm
the
software
engineer
at
elemental
and
I
work
on
the
open
source
side
of
dagster.
This
is
going
to
be
a
demo
talking
about
the
capabilities
that
we've
added
to
the
dijkstra
dpt
integration,
specifically
around
software
defined
assets.
E
Since
o140,
I
gave
a
similar
presentation
way
back
when
a
few
months
ago,
and
I'm
sure
not
everyone
here-
has
seen
that
so
I'm
going
to
start
from
the
beginning,
talk
about
where
we're
at
then
and
talk
about
the
changes
from
then
until
the
o15o
release,
so
starting
from
the
very
beginning.
What
we're
looking
at
here
is
a
graph
of
ops,
so
these
ops
are
loosely
arranged
in
kind
of
the
pattern
of
a
typical
pipeline
at
the
top.
E
E
Once
we
have
that
data
in
the
database,
this
can
be
done
with
a
tool
like
python
or
a
tool
like
error
by
5,
train,
etc.
We
want
to
run
a
dbt
project,
so
dbt
has
a
bunch
of
models.
These
models
are
essentially
sql
transformations
that
are
going
to
create
tables
in
our
database
and
all
these
tables
are
related
to
each
other
in
some
way.
E
Like
I
mentioned,
there's
a
ton
of
tables
that
are
going
to
be
updated
using
dbt,
so
it'd
be
great.
If
we
could
actually
have
detailed
information
on
what's
going
on
in
there
and
that's
exactly
what
software
defined
assets
and
the
dijkstra
dpt
integration
allow
you
to
do
so
if
we
go
ahead
and
instead
of
looking
at
this
operational
graph,
we
look
at
the
graph
of
assets.
We
get
a
completely
different
view.
E
E
Well,
maybe
I
only
want
to
run
these
two
dbt
models
in
this
current
invocation
and
when
you
did
that
before
in
a
lot
of
cases,
that
would
actually
just
end
up
running
the
entire
dbt
project
right
because
under
the
hood,
there's
still
just
one
task
so
that
command
translated
to
run
that
task
and
that
task
would
update
everything,
and
this
wasn't
ideal
for
a
lot
of
reasons.
The
main
one
is,
you
know
when
dbt
projects
scale
like
that
can
be
a
really
expensive
operation.
E
E
E
So
doing
this
from
the
ui
is
useful,
but
another
thing
that
you
can
really
imagine
want
to
do
is
say
like
okay,
I
have
this
big
graph
of
assets.
Maybe
I
don't
want
to
run
this
all
at
once.
Right,
maybe
I
want
to
schedule
different
parts
of
this
to
run
at
different
frequencies
and
that's
become
a
ton
easier
in
o15o,
so
I'm
going
to
go
ahead
and
pop
into
the
code,
but
just
as
a
motivating
example
imagine
we
want
to
refresh
this
order
forecast
model
and
then
everything
upstream
of
it.
E
So
we
want
to
make
sure
that
we're
refreshing
our
model
as
frequently
as
possible,
and
that
includes
all
the
upstream
dependencies
just
making
sure
that
it's
not
running
on
the
same
old
data
over
and
over
again,
and
if
we
look
at
that,
that
only
requires
some
of
the
models
in
the
dbt
project
in
order
to
work.
There's
all
this
stuff
down
here.
That
is
completely
unnecessary
to
order
forecast
model.
E
E
But
the
important
thing
here
is
at
a
high
level
we're
just
loading
some
assets
from
a
dbt
project.
I've
done
some
customization
to
make
all
the
asset
keys,
nice
and
organized.
This
is
optional.
We've
also
loaded
some
assets
that
are
defined
in
python.
So,
for
example,
we
have
these
raw
data
assets
that
are
being
loaded
from
this
raw
data
module,
which
is
right
here.
So
I
can
just
look
in
there
and
we
have
some
regular
python
assets
that
are
generating
some
random
data,
not
too
important.
E
But
here
we
can
see
how
we
actually
define
that
job,
so
the
first
job
that
we
were
looking
at
for
most
of
this
demo
was
this
everything
job
and
its
selection,
which
is
star
meaning
every
asset
in
the
repository,
but
we
can
also
define
sub
selection
jobs,
so
people
that
are
familiar
with
dbt
are
are
also
going
to
recognize
this
sort
of
syntax.
In
this
case
we're
saying:
okay,
select
everything
upstream
of
order,
forecast
model,
and
that
includes
everything
in
the
dbt
project,
but
also
anything
above
that.
E
So
if
we
go
ahead
and
go
back
and
look
at
what
dexter
has
created
for
us
with
this
refresh
forecast
model
job
we
can
see.
Indeed,
it
is
a
set
of
assets
that
only
include
some
of
the
dbt
project.
There's
all
these
dbt
assets
that
are
not
included
in
this
job
and
won't
be
executed
when
we
run
this
job.
E
B
E
So
that
is
one
of
the
major
improvements
and
I'm
just
going
to
go
over
a
cool
use
case
that
the
underlying
abstractions
that
we
improved
kind
of
enabled.
So
if
I
go
back
and
look
at
this
broader
asset
graph,
you
can
see
that,
like
sometimes
there's
going
to
be
cases
where
you
have
all
this
dbt
transformation,
then
you
have
some
like
python-based
stuff.
E
That
needs
to
be
in
python,
but
at
the
end
of
the
at
the
end
of
the
day,
like
this
asset
down,
here
is
creating
a
predicted
orders
table
and
it's
in
the
same
database
as
all
this
dbt
stuff.
So
it's
natural
to
ask
like
what,
if
I
want
to
use
dbt
to
transform
this
table,
for
example,
maybe
we
want
to
union
this
with
some
other
data
create
like
a
full
table?
E
So
I'm
going
to
walk
you
through
how
you
can
add
a
new
model
that
depends
on
this
python
stuff
and
it's
pretty
simple.
So
if
I
go
into
the
dbt
like
sources.yml,
this
is
declaring
all
of
the
external
sources
for
the
data
that
is
being
transformed
by
dbt.
So
all
I'm
going
to
do
is
add
a
new
source
for
that
asset
at
the
bottom
of
our
graph.
This
forecasting
predicted
orders
table
and
I'm
just
going
to
create
a
new
dbt
model.
E
E
E
A
C
A
F
Great
I'm
a
product
designer
on
elementor
working
on
a
core
team.
In
this
session.
I'm
going
to
talk
about
two
of
the
main
improvements
we've
made
to
daggett
in
the
0.15
release.
I
think
most
of
you
are
familiar
with
deckett,
but
for
anyone
who
is
new
daggett
is
the
sort
of
companion
web
app
to
daxter.
It's
we
think
of
it
as
the
sort
of
single
pane
of
glass
for
your
data
platform.
F
F
So
let
me
jump
into
the
demo,
and
let's
do
this
one
so
first
place
I
want
to
start
is
on
the
assets
page
itself.
The
assets
page
is
the
place
to
manage
and
monitor
all
of
your
dexter
assets.
In
o.15
we
have
introduced
we've
added
new
columns
for
the
material
latest
materialization
and
latest
run,
so
you
can
now
see
for
software
defined
assets
all
of
that
information
without
having
to
click
into
every
single
asset.
F
You
can
also
now
multi-select
any
set
of
software-defined
assets
and
quickly
materialize
or
re-materialize
them
right
from
the
list
view.
You
can
also
wipe
materializations
as
well,
and
then
we
also
introduced
a
bunch
of
quick
actions
on
software
defined
assets.
So
if
I
find
an
asset
that
I'm
interested
in,
I
can
just
jump
right
into
a
graph
of
the
upstream
or
downstream
assets
from
it
clicking
into
a
software
defined
asset.
Let
me
grab
this
one.
It
takes
me
to
what
we
call
the
asset
detail
page
here.
F
F
F
I
can
also
quickly
re-materialize
any
assets
within
within
the
the
detail
page
and
then
now
in
in
o.15.
I
can
use
these
sort
of
quick
selections,
so
I
can
quickly
select
everything
upstream
or
everything
downstream
of
this
asset
and
re-materialize
everything
I
see
on
screen.
So
a
really
quick
way
to
you
know
re-materialize
a
large
set
of
assets
also
in
in
0.15.
We
introduced
this
new
concept
of
an
asset
group,
so
asset
groups
are
defined
in
python
and
allow
you
to
more
easily
work
with
a
subset
of
your
assets.
F
When
you
define
an
asset
group,
it
now
appears
in
the
left,
nav
and
you
can
sort
of
quickly
jump
into
a
mini
version
of
the
entire
asset
catalog.
So
here
I've
got
an
asset
group
called
default
and
you're
able
to
define
as
many
asset
groups
as
as
you
like.
I
only
have
one
in
this
section
here.
It
takes
us
to
a
a
view
of
the
this
asset
group's
graph,
and
then
I
also
have
the
same
list
view,
and
this
is
sort
of
scoped
down
to
just
the
set
of
assets
that
I've
defined
in
this
group.
F
As
the
groups
give
you
a
new
way
to
sort
of
collaborate
and
share
status
information
with
your
your
stakeholders,
so
you
can
just
share
this
link,
and
now
they,
your
stakeholder,
has
sort
of
like
slimmed
down
view
of
the
entire
asset
graph
to
just
the
set
of
assets
that
they
care
about.
You
can
also
quickly,
you
know,
re-materialize
one
individual
asset
or
re-materialize,
the
entire
group
of
assets,
so
it
just
makes
working
with
a
subset
of
them
a
little
bit
easier.
F
That's
it
for
updates
to
the
asset
ui
in
no.15.
We're
really
excited
for
you
to
try
it
out
and
appreciate
any
feedback
that
you
have
throughout
it,
we're
going
to
continue
to
iterate
and
invest
in
in
this
experience
in
future
releases
too.
Next
I
want
to
actually
I
want
to.
I
want
to
quickly
talk
about
the
the
left
nav
in
previous
releases.
You
we
didn't
really
have
any
way
of
organizing
this
left
left
nav.
F
We
know
that
many
of
you
have
data
platforms
with,
like
hundreds
or
thousands
of
jobs
across
many
many
repositories.
Now
in
odot
15,
we
automatically
group
all
of
your
jobs
and
asset
groups.
By
repository,
you
can
still
filter,
I
got
to
move
my
zoom
there
we
go.
You
can
still
filter
your
repositories
and
kind
of
slim
down
this
view,
but
it's
now
much
easier
to
kind
of
you
know
organize
things
in
the
left
nav
just
by
repository
okay.
F
The
last
thing
I
want
to
talk
about
is
sort
of
a
pivot
away
from
software-defined
assets,
and
that
is
with
I'm
going
to
talk
about
partitions
and
backfills.
So
for
those
who
are
not
familiar
with
with
partitions
a
prediction,
job
is
a
job
indexer
where
each
run
corresponds
to
a
partition.
Key
most
commonly
this.
F
F
So
here
you
can
quickly
see
the
total
number
number
of
partitions,
the
number
that
have
failed
and
the
number
that
I'm
missing,
and
then
you
have
this
visual
representation
of
that
same
information.
This
makes
it
really
easy
to
kind
of
look
across
the
entire
set
of
your
your
partitions
and
find
gaps
where
you
may
have
missing
partitions.
You
may
have
a
set
of
failures
or
a
bunch
of
intermittent
failures.
F
This
high
level
status
information
is
really
useful
for
catching.
You
know
a
big
set
of
missing
partitions
that
have
happened
in
the
past.
If
you
remember
the
old
design,
you
had
to
sort
of
click
the
back
button
a
whole
bunch
of
times
in
order
to
get
navigate
all
the
way
to
further
back
partitions.
Now
you
can
catch
those,
you
know
just
with
a
glance,
but
a
more
granular
view
is
important.
F
When
you're
trying
to
debug
issues
with
your
partition,
you
can
still
get
to
that
more
step-by-step
granular
view
by
either
clicking
the
show
per
step
status,
button
or
just
selecting
it
anywhere
within
this
within
this
status
bar
selecting,
I
can.
I
can
navigate
this
set
by
just
like
move
clicking
in
different
parts
of
this
graph,
and
it
shows
me
the
step-by-step
status
information
here.
F
F
So
here
we've
made
it
a
lot
easier
to
just
like
quickly
select
all
of
the
failed
or
all
of
the
missing
partitions
or
or
both,
and
I
can
also
just
click
and
drag
within
here
to
select
a
series
of
partitions
or
manually
enter
a
set
that
I
want
to
I'm
just
going
to
grab
one
you
can
also
for
fail
for
failed
runs.
You
can
always
re-execute
from
failure
or
select
a
subset
of
that
job
as
well.
Here
I'm
going
to
just
trigger
a
backfill
with
one
partition
and
navigate
to
the
new
backfills
page.
F
Here
I
can
get
a
high
level
status
of
all
the
backfills
that
I've
run.
You
can
see
in
the
past.
We
have
you
see,
like
the
number
of
runs.
You've
actually
kicked
off.
You
can
click
into
any
of
these
to
actually
view
the
the
details
of
that
run.
Here
we
can,
we
can
go
ahead
and
click
in
click,
click
into
one
and
actually
see
like
the
run
page
for
it.
F
This
will
update
with
the
the
progress
of
the
the
backfill
we
just
ran,
and
now
you
can
also
quickly
terminate
any
unfinished
runs
in
a
backfill.
So
if
you've
kicked
off
a
really
really
large
backfill,
you
can
actually
just
terminate
if
you've
kicked
one
off
that
you
don't
want
to
run,
you
can
terminate
it
right
here.
F
That's
it
for
all
of
the
dagget
updates
that
I
have
to
talk
about
here.
I'd
encourage
you
to
check
out
the
odat
15
blog
post,
where
we
go
more
in
depth
about
some
of
the
other
updates
that
we've
made,
and
I
also
just
encourage
you
to
continue
to
use
the
dagster
feedback
channel
in
the
community
slack.
G
G
So,
instead
of
having
to
go
through
sales,
our
sales
team
is
actually
quite
lovely.
So
that's
that
that's
fine,
but
you
can
actually
go
through
and
just
swipe
a
credit
card
and
start
using
this
in
a
totally
permissionless
fashion,
which
is
super
exciting.
I'm
gonna!
We
don't
have
that
much
time,
so
I'm
gonna
quick
blast
through
these
features,
the
you
know
just
as
a
reminder.
We
use
this
hybrid
architecture,
which
means
that
you
own
your
code
and
your
data,
and
we
don't
see
that,
and
we
only
see
your
metadata.
G
We
support
running
an
ecs,
kubernetes
docker
and
it's
a
plugable
system
for
other
compute
runtimes.
You
know
the
value
adds
here.
You
can
offload
your
operations,
meaning
that
we're
going
to
host
the
web
server
for
you,
the
daemon,
the
metadata
database
and
the
only
thing
that's
left
for
you
to
host
is
effectively
this
open
source
agent
and
a
stateless
cluster.
G
We
automatically
upgrade
the
web
app
for
you,
which
is
great
and
while
maintaining
backwards
compatibility
with
your
code.
Lastly,
we
have
some
enterprise
features
so,
instead
of
having
to
deploy
a
completely
new
instance
anytime,
you
need
a
new
instance.
You
can
just
simply
click
a
button
and
get
a
new
deployment
to
onboard
additional
teams.
We
have
built-in
auth,
github,
google
and
octa
and
others
and
role-based
access
control.
G
You
know,
is
backed
by
a
company
and
we
need
a
viable
business
model
to
support
the
project.
You
know
we're
still.
You
can
still
sign
up
for
early
access
if
you're
interested,
and
then
you
know
we're
also
going
to
be
gathering.
You
know
we're
going
to
make
august
9th
a
marketing
event
and
we'll
be
sending
out
information
about
that,
but
we
want
user
testimonials
both
from
cloud
and
open
source
users.
So
you
can
help
us
and
we're
going
to
have
a
slick,
launch
video,
and
you
know
we
can
make
you
data
famous.
G
Perhaps-
and
you
know,
if
you're
interested
in
this-
it's
a
very
lightweight
thing,
just
kind
of
recording
something
nice
about
the
project
on
your
phone.
You
can
dm
molly
and
she'll
also
be
reaching
out
directly
to
community
members.
If
you're
already
cloud,
customers
or
you're
just
interested
in
hearing
about
upcoming
developments,
we're
going
to
start
using
the
dagster
cloud
channel
and
slack
to
both
update
new
features
and
also
to
preview
upcoming
features
to
get
feedback.
A
And
while
we're
waiting
for
some
questions
to
come
through
I'll
just
I
just
have
like
a
very
small
amount
of
closing
announcements,
which
is
I'm
gonna
post
a
feedback
form
in
in
the
chat
just
to
check
in
to
see
like
was
it
information
presented
at?
Are
these
meetings
helpful?
Are
there
things
other
things
you'd
wish
to
hear
about
so
feel
free
to
fill
that
out?
A
Also
earlier
I
did
a
community
spotlight,
and
one
thing
I
forgot
to
mention
is:
did
we
miss
you
in
the
spotlight,
if
you
created
something
or
in
the
future,
create
something
with
stackster
related
like
a
poster
talk
or
host
an
event?
Let
us
know
and
we'll
add
it
to
our
community
page.
You
can
either
reach
out
to
me
on
slack
or
you
can
find
a
contact
email
directly
on
the
community
page
at
the
bottom,
so
yeah
any
questions
on
all
this
exciting
of
these
exciting
updates.