►
From YouTube: Argo Contributor Experience Office Hour 21th Jan 2020
Description
No description was provided for this meeting.
If this is YOUR meeting, an easy way to fix this is to add a description to your video, wherever mtngs.io found it (probably YouTube).
A
Okay,
hello,
everyone
welcome
to
one
more
argo
contributors,
experience
meeting,
so
I
see
I
think
we
have
some
new
faces
today,
at
least
not
unfamiliar
names.
So
if
you
are
joining
to
the
meeting
for
the
first
time,
can
you
please
introduce
yourself
and
then
we
can
jump
into
agenda.
B
C
Hi
I'm
wes.
I
recently
started
looking
at
a
project
related
argo
cd,
the
progressive
rollouts
controller
and
wanted
to
contribute
something
in
that
general
area.
So
I
figured
I'd
join.
C
Getting
into
the
controller
just
the
promotions
from
one
application
to
another,
so
like
let's
say
you
have
a
staging
environment
and
it's
set
to
sync
from
master.
The
master
branch
or
main
rather
main,
will
resolve
to
a
a
get
shaw
right
and
you
want
that.
Get
shaw
to
then
be
set
as
the
thing
to
sync
for
production.
C
So,
if
staging
passes-
and
you
have
that
you-
you
argo
cd,
the
status
of
the
application
object
in
kubernetes,
after
all,
the
phases
and
waves
and
everything
is
successful,
basically
take
whatever
sha
that
was
deployed
and
set
it
to
a
different
application
being
the
production
application,
possibly
even
a
different
cluster,
to
the
shaw
that
was
just
deployed.
C
A
Yeah
got
it
okay
and
I
think
we
have
one
more
new
identity.
It's
the
nickname
is
one
er7ca.
Can
you
please
introduce
yourself.
E
F
F
A
All
right,
let
me
start
sharing
my
screen
with
agenda
items,
so
we
have
a
lot.
Thank
you
so
big.
I
I
added
mine
yesterday
at
about
argo
cd
notifications,
and
it
was
only
a
gender
item.
A
Today
we
have
more
so
I
will
try
to
be
as
short
as
possible,
so
we
can
talk
about
everything
all
right.
So
let
me
spend
five
minutes
talking
about
our
ocd
notifications
and
I
didn't
plan
to
you
know,
give
any
demos
I
just
wanted
to
highlight
that
the
project
is
kind
of
evolving
and
there
is
a
lot
of
interesting
work
that
needs
to
be
done
and
a
lot
of.
A
Basically,
all
that
came
from
early
adopters
people
asking
for
more
features,
and
it's
looking
for
contributors
so-
and
I
will
try
to
kind
of
you
know,
explain
as
much
as
I
can
within
the
time
and
hopefully
someone
might
get
interested,
and
I
just
wanted
to
spend
like
one
minute
to
just
remind
what
argo
cd
notifications
project
is
and
then
I
wanted
to
basically
share
the
next
milestone
link
like
one
minute.
In
case
you
never
heard
about
argus
identifications.
A
It's
it's
a
set
car
that
you
can
install
into
argo
cd,
namespace
and
it
provides
integration
between
argo,
cd
and
different
notification.
Services
such
as
email
or
slack
or
telegram
and
list
is
kind
of
growing
and
pretty
much
this
overview
page,
which
you
see
on
the
screen,
is
kind
of
it
gives
you
the
shortest
possible
way
to
install
it,
and
maybe,
if
you
execute
this,
I
think
it
basically
have
like
five
steps
you
basically
going
to.
A
You
would
try
most
of
the
features
of
arc,
ocd
notifications
so,
and
these
steps
explains
that
you
need
to
install
an
additional
deployment
into
your
namespace
and
then
you
can
install
pre-configured
kind
of
set
of
settings
that
help
you
to
decide.
A
What
kind
of
notifications
you
might
want
to
get?
And
finally,
you
just
need
to
pick
an
argo
cd
application
that
you
want
to
get
notifications
about
and
you
just
need
to
apply
an
annotation.
Sorry,
it's
that
step,
and
this
is
kind
of
high
level
summary
of
argus
identifications
and
it
there
are
much
more
kind
of
information.
A
So
there
are
different
type
of
notification
types.
I
would
say
it's
here.
We
call
them
templates
so
basically
different
ways
to
generate
notification,
message
for
different
type
of
events,
and
we
have
a
way
to
configure
events
and
we
call
them
triggers.
A
It
has
much
better
description,
and
but
during
that
meeting
I
wanted
to
highlight
that
project
is
not
even
not
even
close
to
feature
complete
and
we
have
a
set
of
kind
of
buckets
of
features
that
we
can
choose
to
work
on,
and
I
just
wanted
to
list
that
you
give
you
that
set
of
buckets
so
first
of
all,
people
asking
to
add
more
integrations,
so
more
notification
services
in
particular,
there
is
already
a
request
to
integrate
with
microsoft,
teams
and
irc.
A
This
is,
and
I'm
pretty
sure
that
list
is
going
to
grow
because
maybe
tomorrow,
like
signal
is
getting
popular
and
people
might
want
to
integrate
with
signal.
So
this
is
one
and
second.
A
A
I
already
noticed
that
early
adopters
who
fargo
cd
come
up,
is
you
know
a
better
set
of
triggers
and
templates
that
they
enhance
them
and
I
feel
like
we
just
need
to
contribute
more
of
more
triggers
and
more
templates
into
so-called
so-called
catalog
of
triggers,
which
is
basically
a
yaml
file,
which
has
a
definition
and
of
triggers
and
templates
and
another
area
where
argus
identification
is
going
to
grow.
I'm
pretty
sure.
So.
Notifications
is
kind
of
misleading
name
very
soon.
A
After
the
first
release,
a
lot
of
users
suggested
that,
in
addition
to
just
notifications,
notification
services,
argos
identifications
can
send
messages
to
just
to
just
hooks
basically
a
generic
web
hook
that
can
do
whatever
you
can
imagine,
and
we
already
have
examples
of
how
you
can
start
a
jenkins
job.
For
example,
when
application
sync
completed
or
you
can
push
send
a
message
to
github
and
update,
commit
status,
and
I'm
pretty
sure
if
we
you
know,
if
we
talk
with
end
users,
you
will
see
even
more
examples
of
such
integrations
and
yeah.
A
A
Embed
approve
button
into
the
slack
message,
so
user
could
get
notification
about
successful
sync
and
then
let's
say
if
it
includes
rollout
using
user,
can
click
approve
button
right
in
slack
and
yeah?
There
is
a
ticket
for
that
and
already
it
has
several
thumbs
ups
and
I'm
pretty
sure
we
can
expand
it
to
all
type
of
integration
notification
services,
I'm
pretty
sure.
Maybe
telegram
has
the
same
type
of
functionality.
A
Maybe
it
has
a
button.
If
it
doesn't
have
a
button,
it
could
be
just
a
link
which
you
can
click
to
approve
the
same
process
yeah
and
there
is
we
have
a
milestone
and
that
milestone
has
set
of
tickets
and
all
those
I
mean
all
these
tickets
are
kind
of
ready
to
go.
There
is
no
need
to
start
discussion
about
how
to
like.
I
think
most
of
them
are,
you
know
pretty
straightforward
and
it's
a
great
place
to
start,
and
you
know,
after
that
we
can
start
thinking
about
more
open-ended
likes.
Like
chat
ups.
C
I
have
a
question:
does
argo
cd
already
today
emit
kubernetes
events?
Yes
does,
does
it
do
it
for
all
major
state
changes.
A
It
no
it
I
mean
it's
I've,
I
think
currently,
agustin
vacations
mostly
use
events
as
a
audit.
Look
so,
and
I
think,
for
most
important
changes.
D
Yeah
for
phase
changes
like
if
it
became
health,
the
health
status
change
or
the
sync
status
change.
If
you
consider
those
as
the
major
exchanges,
then
the
answer
is
yes.
C
Okay
yeah.
I
was
just
thinking
along
the
lines
of
something
maybe
similar
to
argo
cd
notifications
and
that,
if
there's
a
project
that
can
watch
for
specific
kubernetes
events.
B
C
Then
deliver
notifications
or
web
hooks
or
whatever
that's
just
a
little
bit
more
abstracted
and
not
and
could
be
used
for
many
other
use
cases
as
well.
I
like
abstractions,.
A
I
I
feel
like
yeah.
There
was
a
discussion
like.
Basically,
this
is
not
the
first
attempt
of
notification
service
for
in
general,
for
like
kubernetes
related
projects,
so
argos
identifications,
I
think
it's
kind
of
the
benefit
it
brings-
is
that
it
is.
A
A
little
bit
narrow,
but
basically
it's
focused
on
argo
cd
use
cases
and
that's
why
it's
just
easier
to
use
it
so,
and
I
think
I
was
it's:
documentation
is
kind
of
kind
of
trying
to
communicate
it.
So
the
one
of
the
biggest
differences
is
that
you
can
start
using
it
with
no
configuration
because
it
it
knows
kind
of
what
you
might
be
interested
in,
and
it
knows
what
kind
of
events
and
notifications
you
might
want
to
get
yeah.
So
this
is
right.
C
Does
argo
cd
notifications
use
kubernetes
events
in
the
background
it
it.
A
Does
not,
and
but
it
uses
it
just
it's
just
a
controller
that
keep
watching
argo
cd
applications
and
it
uses.
Basically,
it
has
list
of
so-called
triggers
and
it
tries
to
you
know,
execute
each
trigger
against
each
application.
If
and
if
any
trigger
returns.
True,
that
means
an
affiliate.
Notification
should
be
sent.
D
In
an
interesting
time,
I
think
maybe
we
should
we
have
documentation
of
the
design
rationale
of
why
we
chose
not
to
go
with
like
a
controller,
a
general
purpose
controller
and
and
went
with
a
library
approach
instead.
So
why
don't
we
move
to
the
next
topic
and
I'll
wes
I'll?
Put
the
link
to
the
notification
design
document
that
explains
the
different
choices
that
we
did?
D
There
was
attempts
before
where
we
did
try
to
monitor
any
type
of
resource,
including
events,
but
that
general
purpose
controller
ended
up
being
not
so
useful
because
of
the
the
complexity
of
having
to
describe
a
a
way
to
take
that
change
and
then
convert
it
to
and
the
notification
event
got.
C
G
Yes,
I
think
yeah
yeah
I'll
do
that.
Okay,.
A
Thanks
awesome,
yeah
and
I
okay
and
just
like
last
think
about
notifications.
I
feel
like
in
general,
we
chose
to
work
on.
You
know
this
kind
of
side,
projects
and
separate
people.
It
really
helps
to
move
quicker,
but
as
a
side
effect.
We
just
need
to
do
some
extra
effort
to
get
attention
from
both
users
and
contributors.
That's
why
I
think
I'm
going
to
write
a
couple
more
blogs
about
argus
identifications
and
mentioned
that
during
that
meeting.
A
Okay,
that's
all
I
had
about
this
topic.
I
think.
D
Okay,
I
posted
the
link.
Is
there
also
is
this
recording
when
we
discuss
this.
A
Yes,
recording
I
will
need
to
yeah
I
you
know
I
will
post
send
the
link.
Oh
wait,
a
second,
it's
right.
There,
oh
no,
never
mind.
A
D
I
I
might
be
able
to
find
it
while
you
continue.
Okay,.
A
Thank
you
right.
So
next
item
is
community
housekeeping
ux
ui
sessions.
I
guess
should
be
q
a
that
all
this.
F
Yeah,
I
think
most
of
the
jargon
from
here
is
by
me
a
bunch
of
different
things,
so
I
think
yeah
I
think
in
general.
I
wanted
to
check
if
we
could
organize
ourselves
a
little
better
with
respect
to
the
ui
ux
discussions
that
we
should
be
having,
because
I
think
we
all
agree
that
you
know
significant
guides
changes
go
in
which
sometimes
modify
the
experience.
F
It
is
fine
to
have
an
open-ended
meeting
like
this
one
that
we
have
there,
but
it's
just
that.
I
want
to
ensure
that
you
know
we
have
a
little
more
organization
so
that
we
can
have
the
right
folks
at
the
right
meetings
when
we
want
to
discuss
these
things.
So
my
my
proposal
in
general
is
if
we
can
mark
at
least
one
of
the
weeks
like
sorry,
so
so
so.
F
H
A
I
I
agree
that
maybe
it's
good
to
use
the
same
meeting
you
know
and
discuss
your
changes,
how
about,
if
we
just,
instead
of
doing
it
every
third
meeting,
what
if
we
just
just
like
a
label
in
the
same
in
the
document,
ux
your
ux
and
if
we
have
more
than
let's
say
four,
then
we
just
transform
next
meeting
into
into
your
ui
ux
discussion.
F
Yeah,
we
could
do
that,
so
I
think
the
main
risk
with
that
is
so
okay,
I
mean
that's
what
I
had
proposed
as
well
initially,
but
I
think
the
the
risk
with
that
is
that
if
for
two
or
three
meetings,
we
have
no
ui
ux
labels
there,
folks,
who
typically
would
have
blocked
this
time
of
the
day
for
this
meeting
would
have
something
else
on
the
calendar.
Okay,
and
what
happens
is
you
know
this
slot
is
completely
lost
for
those
folks,
because
you.
B
F
At
least
two
or
three
meetings
were
not
productive
enough,
so
I
could
flip
that
actually,
what
I
could
do
is
every
third
meeting.
At
least
we
effectively
have
a
ui
ux
focus
meeting,
and
if
there
is
no
agenda
on
it,
then
I
will
happily
go
and
repurpose
that
as
a
normal
meeting
typical
meeting
okay
and
vice
versa,
if
if
there
are
folks
who
want
uiux
feedback
on
any
of
these
other
days-
and
they
say
hey-
I
really
want
this
on
the
agenda.
F
Then
I
think
one
of
us
and
that
could
be
me-
I
could
actually
go
and
catch
hold
of
the
ux
folks
that
we
have
at
red
hat
and
ensure
that
they
do
attend
this
in
case
they
were
planning
not
to,
but
but
those
are
exceptions
right,
but
I
think
in
general,
if
you
come,
we
could
come
with
the
process
that
you
know.
Every
third
meeting
is
you
are
ux
focus,
then
I
think
serena
bridget,
both
those
they
can
actually
join
without
having
to
move
the
calendars
around.
That
would
look
good.
I.
I
E
Yeah,
I
think
it's
a
great
idea
as
well,
because,
yes,
I
do
it's
in
in
saving
time,
especially
for
for
the
ux
people
who,
when
we
come
on,
it's
not
necessarily
always
ux
related.
I
think
that
will
be
greatly
appreciated
for
from
our
time
side
for
sure
so
appreciate
it.
If
you
guys
do
that.
A
Okay,
so
the
next
topic,
I
believe
I
know
what
so
it's
a
question
about
pull,
request
right.
No.
A
I'm
sorry,
yes,
housekeeping.
Okay
assignment
of
roles
for
github
q,
a.
F
Yeah,
that
was
by
me
again
I'll
quickly,
go
through
what
I
really
intended
here.
So
I
think
we
are
getting
a
lot
of
questions
on
the
github.
I
think
question
and
answers
that
we
have
there
instead
of
github
issues,
and
I
think
thank
you
jesse
for
actually
starting
that,
while
it's
a
good
thing
that
people
are
moving
there,
I
think
in
general
we
need
to
ensure
that
we
respond
to
them
quick
enough
and
I
think,
as
maintainers
and
leads,
I
think
you
know,
jan
alex
and
jesse.
F
You
folks
already
engage
a
lot
on
github
issues
and
prs,
and
it
is
obvious
that
you
wouldn't
always
get
a
chance
to
go
and
answer
questions
on
github
q
a
so
I'm
proposing
that
on
a
rotational
basis.
Every
week,
if
somebody
could
volunteer
to
say
hey
like
it's,
it's
some
it's
something
like
a
triager
role,
but
more
of
a
question
answer
role
where
you
know
the
role
that
person
takes
up
for
that
week
is
to
go
to
github
question
and
answers
for
argo
cd,
for
example,
and
ensure
that
questions
are
answered
on
time.
F
If
you
don't
know
the
answer,
it's
fine
to
you
know
poke
through
the
question
a
little
more
and
you
know
catch
hold
of
somebody
like
alex
to
go
in
and
do
that.
But
at
least
this
way
we
ensure
that,
if
somebody's
asking
a
question,
the
turnaround
time
is
quick
enough
and
they
do
depend
on
this
mode
of
support
than
the
previous
mode,
where
it
was
probably
easier
to
go.
Github
issues
it
used
to
notify
a
bunch
of
people.
F
So
with
that
I
would
say
if,
at
the
end
end
of
this
meeting
or
now,
if
somebody
says
hey,
I
would
I'd
I'd
like
to
take
up
that
role.
That
would
be
good,
but
prior
to
that
alex
yan
jesse.
What
do
you
think
and
everyone
else.
A
I
yeah,
I
feel
like
it's
a
great
proposal,
because
it's
we
just
had
the
exact
same
discussion
pretty
much
last
week,
so
we
spoke
with
about
it
with
jc.
I
think.
Currently
we
pretty
much.
We
use
intuits
team
stand
up
to
do
that
and
it's
clearly
not
scalable
and
we
wanted
to
move
it
offline
and
basically
somehow
make
it
visible
to
everyone,
this
triage
process
and
what
you're
proposing
is
even
better.
I
feel
like.
If
we
have
you
know
official
list
of
people
who
volunteer
to
do
that,
and
we
have
rotation
yeah.
A
F
And,
like
I
said
like
stand
up,
is
pretty
again:
it's
pretty
internal.
You
might
not
have
everyone
in
there
who
ask
those
questions
like,
for
example,
in
github
discussions
right,
you
might
have
like
it.
It
could
be
a
simple
question
with
a
tedious
answer,
which
means
hey.
How
do
we
do
this?
We
all
know
the
answer,
but
probably
will
take
you
15
minutes
to
write
a
nice
long
answer
and
which
is
why
we,
I
think
it
should
be
rotated
any
other
thoughts
on
this.
F
A
Of
by
default,
his
yeah
jc
and
yeah.
F
I
mean
I'll
I'll
say
other
than
the
maintainers,
because
the
maintainers
would
anyway
get
pulled
into
a
lot
of
things
right.
So
I
think
this
would
be
a
non-maintainer
role
to
begin
with,
at
least
just
so
that
you
know
we
can
take
some
load
off
the
maintainers
so
that
you
can
focus
on
the
issues
and
the
pr's
in
general
that
you
do,
though,
that's
also
everyone's
job.
A
F
Could
you
could
you
click
on
the
link
github
discord
on
on
zoom?
I
just
thank
you.
I
I
F
Right
and
I
think,
while
while
trying
to
talk
to
him,
I
figured
I
I
started
thinking
with
he's
asking
a
general
question
about
helm
using
crds,
and
then
I
found
out
that
no
he's,
probably
talking
about
argo
cd
are
the
helm
distribution
for
the
argo
cd
release,
for
example.
So
I
think
these
are
very
general
questions.
Some
of
them
do
need
you
to
go
and
actually
look
at
past
issues
read
documentation.
F
Maybe
you
know
talk
to
somebody
on
the
channel
and
then
frame
an
answer
and
put
it
so
yeah.
That's
how
I
mean
this,
so
the
role
role
would
mean
effectively
to
first
on
talk
to
this
person
here
and
figure
out
what
exactly
the
person
wants
to
know.
Sometimes
the
questions
are
very
vague,
they're,
just
a
one
one
liners
or
two
liners.
F
The
idea
is
to
go
in
and
ask
and
figure
out,
get
some
more
details,
and
if
you
can
answer
answer
the
question
great,
if
you
can't
talk
to
somebody
who
knows
probably
a
portion
of
the
answer
and
then
you
can
frame
the
answer
yourself
so
as
jan
mentioned,
this
is
an
excellent
way
to
know
argo
cd
because
number
one
you
you
don't
have
to
make
prs.
That
may
potentially
take
a
lot
of
time
and
may
need
that
level
of
focus
and
attention
number
two.
F
D
Sorry,
I
had
a
thought
just
now:
we
currently
we
don't
have
a
slack
channel
for
the
the
members
of
the
argo
crutch
or
maintain
right
actually
do
we
have
one
for
yeah
members
are
and
or
maintainers
you
have.
D
Yeah
I'm
wondering
if
that
would
be
useful.
Somehow
in
let's
say
there
was
a,
I
don't
know
if
there
was
any
kind
of
questions
or
stuff
that
were
specific,
I
I
don't
want
to
keep
adding
slack
channels
because
yeah,
but
I
just
wanted
to
throw
the
idea
if
there
if
there
was
useful
for
other
reasons,
like
you
know,
talking
about
the
triage
like
rotation
and
reminders
and
stuff
like
that,
then
yeah,
maybe
maybe
it
would
make
sense.
I
To
have
a
private
channel
right,
that's
what
I
was
yeah,
because
I
I
can
understand
it's
it's
often
it
might
not
be
easy
to
to
ask
such
questions
in
public
or
so
maybe
private
generally
would
be.
I.
F
I
wouldn't
recommend
a
private
channel.
To
be
very
honest,
I
think
I
think
to
be
fair.
I
would
say
I
mean:
let's
go,
let's
go
through
this
scenario
like
and
and
and
to
be
fair
in
the
last
few
days
he
even
asked
a
bunch
of
dumb
questions,
and
I
think
we
all
saw
that
right.
So
I
think
in
general
I
would
either
go
and
ask
an
argo
dev
because
I
know
there's
a
small
number
of
people
in
there
who
would
be.
F
You
know
listening
to
my
knife
question
and
if
I'm
okay
with
that
I'll,
do
it
there
and
if
I'm,
not
okay
with
that
I'll,
probably
ping
alex
separately
and
say
hey,
you
know
what
I
have
a
very
nice
question.
I'm
super
embarrassed
to
ask
about
it
to
others
and
I'm
probably
gonna
take
care
of
it.
But
the
thing
with
the
private
channel
is,
you
may
not
have
everyone
there
who
might
be
able
to
answer
it
so.
D
Actually
yeah
for
this
specific
use
case,
I
say
it
wouldn't
actually
even
work
because
sharma
she's
on
our
team
right
but
she's,
actually
not
a
member
of
argo
project
she's
like
working
her
way
up
there
because
she
recently
joined,
and
so
actually
it
wouldn't
even
work
in
this
scenario.
So
I
think,
without
a
concrete
need
for
a
private
channel
that
I'm
thinking
about,
let's
not
create
one
yet
yeah.
I
Yeah
sure
no,
it
was
just.
It
was
just
an
idea,
so
maybe
yeah.
I
Can't
understand
if
if,
if
people
don't
want
to
ask
yeah,
I
always
say
there
are
no
stupid
questions
right,
they're
only.
I
J
F
I
F
A
A
A
Peers,
I
actually,
I
think
we
just
I
remember
it.
I
was
reviewing
it
and
then
I
think
we
were
rushing
to
create
one
of
eight
release
and
then
it
just
slipped
kind
of
we.
So
last
during
class
meeting,
we
talked
about
this
type
of
prs,
so
we
have
several
several
of
them,
so
good
prs
really
close
to
get
done
and
we
just
need
to
complete
it.
So,
and
this
is
one
of
those
which
have
to
be
completed
because
it's
it's
a
good
feature.
A
We
agreed
on
design-
and
I
think
I
just
I
don't
have
time
yet
to
to
to
look
at
it
and
as
far
as
I
remember
last
time
I
looked
at
it,
I
had
a
comment
and
I
wanted
to
propose
an
improvement
but
to
propose
it.
It
would
take
me,
you
know
like
30
minutes,
and
I
just
could
not
find
this
20
minutes.
F
Awesome
yeah.
I
was
asking
some
questions
for
the
last
two
days
that
in
general,
why
do
we
need
this?
Because
there
was
no
description,
no
github
issue,
I
kind
of
engaged
with
this
person
for
a
while,
and
there
were
some
changes
that
came
after
thank
you,
but
so
I
think
so.
I
think
I
think
I
I
wouldn't
want
you
alex
to
all
always
go
in
and
be
the
first
person
to
do
it
in
general
and
and
that's
to
say
for
the
maintenance
as
well.
We've
got
a
good
number
of
new
reviewers
in
the
group.
F
I
would
say
this
is
a
call
out
to
reviewers
that
you
know
let's
go
through
all
the
old
pr's
which
haven't
gone
in
for
a
while,
and
yes,
your
new
reviewers,
your
your
green
tick
mark
makes
sense.
So
I
would
say
you
should
go
and
do
it.
A
We
agreed
to,
and
basically
I
promised
to
just
start
the
conversation
and
I
wanted
to
copy
old,
all
pr's
that
need
discussion
and
at
least
talk
with
all
maintainers,
and
then
we
can
basically
decide
if
we
want
to
close
some
of
them,
because
some
of
them
are
kind
of
you
know.
Maybe
yes,
maybe
no,
but
at
the
same
time
I
know
that
at
least
couple
dozen
of
them
like.
Definitely
yes
and
then
we
can
make
a
decision
if
we
want
to.
A
You
know,
reach
out
to
original
contributor
first
and
even
if
contributors
no
longer
have
time
to
complete
it,
it
makes
sense
for
us
to
kind
of
take
over
and
complete.
For
example,
there
was
a
feature
that
a
lot
of
people
keep
asking
for
it's
a
and
it
was
created
more
than
one
year
ago.
I
doubt
the
same
person
you
know
can
keep
working
on
it
like
right
now.
If,
yes,
that
is
great,
if
not,
I
would
not
just
throw
away
all
his
work
instead.
I
would
just
you
know.
F
F
Yeah,
I
think
that
makes
sense
I
think
in
in
general.
Yes,
I
think
if
you
could
close
this,
this
would
be
good
and
also
in
general,
like
so,
I
think
from
the
maintainer's
perspective.
If
you
could
go
through
that
quick,
you
know
skim
through
the
prs
and
close
whatever
is
not
supposed
to
be
open.
F
Subsequently,
if
you
could,
if
our
reviewers
could
actually
go
in
and
take
a
look
at
the
other
pr,
so
that
the
first
level
of
review
is
done
and
any
obvious
problems
with
the
pr
should
be
caught
earlier.
F
A
F
Yeah
so
so
this
is
a
question
I
asked
a
while
ago
and
and
that's
fine,
we
didn't
answer
it
because
this
was
probably
a
meant
for
a
fee,
so
this
is
in
github
discussions
itself
yeah.
So
so
my
question
is:
in
general,
I
looked
at
the
oidc
config
that
we
have
today.
There.
B
F
F
What
that
helps
is
if
some
something
is
not
purely
oidc
compliant.
You
can
still.
If
you
can
specify
the
author
and
the
token
url,
you
can
go
ahead
with
the
auth
dance.
I
was
just
wondering
if
there
was
any
specific
reason
why
this
was
not
supported.
I
can
go
and
look
and
make
an
enhancement
to
see
if
it
works
out.
So.
D
Is
auth
url
and
token
url,
typically
in
the
well-known
configuration
broadcaster
and
then
I
see
okay
and
then
and
then
some
projects,
I
guess
some
oidc
providers,
don't
advertise
that
in
their
well-known
configuration,
is
that
the.
F
F
F
D
Okay,
I
think
I
remember
your
question.
I
didn't
understand
the
the
details
now.
I
think
I
understand
a
lot
better
thanks,
so
I
don't
have
any
yeah.
I
don't
think
I
don't
have
any
objection
to
introducing
those
two
fields
to
assist
with
the
the
login
exchange.
I
think
the
only
difficulty
you
might
run
into
by
adding
those
is
that
the
library
that
we
might
be
using
might
not
understand
how
to
you
know,
supply
that.
D
If
I
recall
correctly,
I
think
it
might
just
like
accept
the
well-known
configuration
url
and
then
it
handles
all
the
handshaking
from
there.
But
if
it's
possible
to
specify
this
separately
and
achieve
the
same
result,
I'm
yeah
I'm
all
for
it.
D
Yeah,
I
don't
think
there
would
be
any
security
implications
to
this,
because
it's
all
still
administrator
controlling
these.
These
things
so
yeah
yeah.
I
F
F
F
Could
version
across
different
versions
of
argo,
cd
or
maybe
just
different,
reduces
for
argos?
Let's
say
you
want
to
implement
a
breaking
change
in
that
and
you
can
advertise
it
using
the
general
crd
validation
schema
that
hey.
We
are
deprecating
this
we're
going
to
move
to
something
else
so
because,
especially
because
this
has
grown
pretty
big
and
they
wonder
if
it's
time
to
break
that
out
of
a
config
map
and
have
our
own
crd
altogether,.
A
C
So
is
this:
is
this
sorry,
I'm
just
asking
for
clarification?
Are
we
talking
about
like
the
argo
cd
pod
itself
and
and
how
it's
configured
when
it
boots
up,
or
are
we
talking
about
something
else?
It's
a
it's.
A
D
Yeah,
this
is
the
config
that
says
like
okay:
here's
how
to
configure
oidc
here's,
how
to
configure
local
users;
here's
how
to
actually
repos
are
also
configured
in
there
and
it's
kind
of
grown
over
time.
C
Yeah
would
would
there
be,
I
would
say,
the
only
thing
I
can
think
of
is
a
risk
of
getting
into
a
a
catch-22
situation
where,
in
order
to
load
the
crd
properly
or
do
something
you
need
to
have
the
crd
some
information
in
the
crd
or
something
like
that.
I
don't
know
if
that,
maybe
that's
not
applicable,
but
it
you
I
I
can
see
there
being
some
sort
of
catch
22.
D
It's
actually
a
valid
point,
there's
actually
today,
because
we
inline
structured
data
as
a
key
in
the
configmap
and
that
structured
data
better,
be
well-formed.
Otherwise
you
can't
load
it
it's
a
similar
problem,
and
there
are
s,
has
been
situations
where,
like
people,
administrators
who
make
a
mistake
and
that
data
would
somehow
blow
away
their
configuration.
A
I
I
kind
of
I
was
working
recently
on
argos
identifications
and
I
think
I
was
trying
to
kind
of
apply
low
links
from
argo
cd
and
I
chose
different
ways.
So
I
basically
in
argos
identifications,
there
is
a
tool
and
that
tool
help
you
to
manage
declarative
configuration.
It
just
produces
declarative
configuration
and
I
think
we're
trying
to
do
the
same.
A
C
C
You
make
a
cli
tool
that
knows
how
to
make
a
you
know,
your
structured
yaml
or
whatever
it
is
inside
of
a
config
map,
because
you're
absolutely
right
that
config
map,
really
the
data
value,
is
a
string
right,
and
so
it's
really
easy
to
mess
that
up
and
there's
not
a
lot
of
tools
that
will
validate
the
data
value
of
a
config
map
because
it
in
the
at
the
end
of
the
day.
It's
just
a
string
but
having
some
way
to
generate
it
might
be,
might
be
interesting.
C
I
use
jsonnet
today
and
I
can
so
I
can
actually
go
from
a
you
know
strongly
typed
if
you
will
object
and
then
use
a
json
standard
library
which
is
manifest
to
you,
know,
yaml
or
manifest
to
json,
and
it
will
take
that
object
and
convert
it
to
a
string
that
that
is
then
valid
for
a
config
map.
That's
what
I
do,
but
it's
the
same
idea
right
start
from
a
strongly
typed
start
from
a
strongly
typed
representation
and
then
convert
it
to
a
string.
C
F
Think
so
I
think
on
that
right,
so
I
think
the
bit
which
I
think
might
really
help
with
respect
to
a
strongly
typed
api.
Is
that
number
one
you
can
avoid?
I
mean
it's
it's
hard
to
make
a
mistake
with
that,
because
your
cli
is
effectively
cube.
Ctl,
it's
going
to
stop
you
from
doing
something
with
the
schema
things
with
the
schema
validation.
Things
is
not
valid
number
two.
You
can
actually
build
out
a
ui
which
can
be
totally
dynamic,
which
means
you
don't
have
to
build
a
form
with
hard-coded
uniform
fields.
F
You
can
potentially
load
the
crd
schema
and
have
that
drive
your
entire
form
so
which
means
if
tomorrow,
we
add
five
more
elements
into
the
config.
Your
form
can
be
totally
dynamic
and
it'll
just
load
things
that
will
show
up
so,
which
means
your.
We
can
potentially
have
a
ui
way
of
config
making
this
configuration,
and
that
could
be
a
one-time
effort,
basically
to
understand
the
schema,
validation
and
generate
form
fields.
F
And
yes,
I
think,
and
then
the
second
thing
like
I
said
it
will
like
the
schema
validation
itself,
is
going
to
ensure
that
your
content
is
of
the
right
types.
At
least,
of
course,
something
can
go
wrong
after
that
as
well,
but
that's
same
with
the
config
map,
anyway,
yeah.
D
One
thing
I
wanted
to
talk
about
is
that
the
see
the
config
is
not
actually
user
facing
right,
the
typically,
when
I
think
of
a
crdi.
It
is
something
that
I
expect
end
user
to
unders,
possibly
understand
like,
for
example,
the
argo
cd.
D
The
app
spec
is
something
they
can
modify
in
the
in
the
edit
panel,
but
for
this
configuration
the
only
person
who
would
ever
modify
that
is
an
administrator
of
the
argo
cd
instance
and
then
no
end
user
is
ever
going
to
understand
or
need
to
know
this
this
structure.
So
apparently
this
is
for
operators.
D
Yeah-
and
that
was
actually
one
of
the
rationale
why
notifications
went
with
a
config
map
approach,
because
we
actually,
I
was
actually
the
one
who
originally
went
to
alex
and
said,
hey.
D
I
think
we
need
a
crd
to
have
structured
representation
of
notification
configuration
because
configmap
is
just
too
unwieldy
and
then
we
we
quickly
came
to
the
realization
that
that
end
users
are
never
going
to
write
this
thing,
because
it's
always
a
notification
manager
or
operator,
and
then
at
that
point
we're
like
okay,
we're
willing
to
make
that
person's
life
a
little
harder
if
it
makes
the
end
user's
life
simpler,
like
they
don't
have
to
know
about
these
things.
D
To
be
created
like
once,
yeah
only
in
the
namespace
of
the
of
the
controller,
and
that
would
be
the
only
place
where
that
crd
exists
in
the
in
the
cluster.
Pretty
much
like.
F
But
if
the
same
cr
is
created
in
the
name
space
where
the
argo
city
workloads
exist,
it
would
recognize
it
kind
of
similar
to
let's
say
the
config
map
in
that
respect,
that
you
can
create
the
same
name,
config
map
with
the
name
and
space
just
that
now,
every
user
in
the
cluster
will
know
that
hey
there's
a
crd
called
argus
reconfig.
Let's
see
what
that
does.
That's
a
problem.
C
C
F
Yeah
I
mean
that
that
should
be
exactly
same
as
how
today
we
have
reconciliation
for
sorry,
so
we
have
something
that
watches
those
specific
config
maps
and
secrets.
So
if
somebody
changes
those,
it
basically
gets
notified,
picks
up
those
changes
and
sends
the
information
to
the
different
subscribers.
F
Of
a
secret
or
a
config
map,
so
we
already
have
that
it's
just
that,
instead
of
watching
a
secret
or
config
map,
we'll
watch
a
specific
type
and
then
like
I
said,
we
can
also
interact
with
the
operator
there
by
saying
hey,
looks
like
you've
put
in
here,
50
fields.
Out
of
that,
these
20
are
totally
garbage
fix
it
and
only
then
it'll
be
applied
else
will
not
apply
it,
so
I
was
kind
of
going
there
first.
Does
that
help
from
what
you.
C
Think
yeah
a
little
bit,
I
was
actually
wondering
if,
if
I
don't
know
how
again
I'm
the
first
time
here
so
breakout
sessions
for
this,
like
I
do
want
to
move
on,
it
looks
like
there's
four
other
bullet
points
here,
but
I'm
interested
in
this
and
what
kind
of
designs
and-
and
you
know,
problems
and
solutions-
and
everything
else
is-
is
there?
C
How
would
you
go
about
taking
this
kind
of
offline
from
this
particular
meeting
and
furthering
the
discussion.
A
Yeah,
I
think
continue
it.
Sorry,
honestly,
I
feel
like,
and
I
wanted
to
say
that
I
I
was
really
interested
in
the
same
topic
as
well,
and
I
know
that
sharma
she
works
on
kind
of
the
same
problem
right
now,
she's
trying
to
make
it
easier
to
configure
argo
cd
settings.
So
this,
I
think
two
of
us
you
know,
really
want
to
be
in
that
conversation.
C
C
F
I
think
in
general
I
would
say,
like
things
things
like
these
would
should
start
from
a
discussion.
It
should
go
through
a
feature
proposal,
proper
design
proposal,
and
only
then
you
know
it
can
even
be
considered
for
merge,
but
yeah,
I
think,
having
a
github
discussion
to
begin
with
would
be
great.
A
And
I
I
wanted
to
mention
that
the
reason
I
was
so
kind
of
interested
in
that
idea.
I
noticed
that
so
you
know
that
flux
2
recently
was
created
and,
I
think
flux
developers
they
kind
of
learned.
You
know
the
same
issues
from
flux
one
and
they
had
opportunity
to
re-implement
it
and
they
chose
this
way.
They
basically
use
crd
and
they
have
two
that
produce
3d
and
it's
working
really
well,
so
they
already
proved
it's
a
good
approach.
A
At
least
I
noticed
that
flux,
2
users
really
like
that
there
is
a
cli
that
you
can
use
and
it
produce
files.
So
it's
kind
of
they
declare
they've
declared
you
from
the
beginning
and
that's
very
attractive
yeah,
and
that
we
should
as
well
yeah.
We
should.
F
Yeah,
I
think
I
think
to
add
to
that,
even
in
the
argo
city
operator,
that
we
have
there,
for
example,
so
there
I've
seen
that
there
is
a
cm,
so
there's
an
argo
series
crd
there,
which
effectively
spins
up
an
argo
city
instance-
and
I
can
see
that
crd
is
gradually
bloating
to
actually
include
strongly
typed
config,
that's
present
in
these
config
maps,
which
is
where
I
figured
out
that
hey
if
we
are
putting
in
strongly
type
configs,
because
operators
seem
to
be
enjoying
this.
F
When
I
say
operators
I
mean
admins
seem
to
be
enjoying,
you
know
using
strongly
typed,
because
it's
a
lot
less
error
prone.
Why
not
actually
have
the
discussion
upstream,
if
you're
interested
overall
in
having
a
conversation
on
this
so
yeah?
There
are
a
couple
of
points.
You
know
good
candidates
there
to
check
on.
C
Oh
yeah,
you
could
actually
make
an
argo
cd
operator.
I
love
that
these
terms
are
ambiguous,
sometimes
but
yeah,
an
argo
cd
operator,
so
the
operator
is
what's
in
charge
of
managing
the
argo
cd
instance
itself.
Oh,
that
exists.
That's
the
oh!
That
already
exists
yeah!
Well,
then,
when
that
wouldn't
that
be
like
go
down
that
path
right,
because
that
path,
if
you
have
an
operator
that
you
need
to
tell
that
operator
how
to
how
to
configure
the
argo
cd
instance
already
right.
F
F
I
know
how
to
configure
argo
cd,
and
this
is
what
I'm
going
to
configure
in
that
crd
or
the
config
map,
so
there's
that
second
level
admin
who's
going
to
be
using
this,
not
the
first
level
admin
whose
job
is
to
just
you
know:
hey.
You
know
these
yamls,
look,
fine!
These
role,
bindings,
look,
okay,
it's
not
taking
something
major
and
I'm
just
going
to
say,
go
run
it,
but
how
are
go
cd
itself
functions?
F
Yeah,
I
think
we
discussed
this
on
slack
a
bit
and
I
just
wanted
to
finish
three
more
minutes.
Yeah.
I
think
I
I
do
see
in
the
ui.
There
are
like
two
buttons
like
sync
and
refresh,
and
then
in
some
places
we
have
something
called
app.
Resync
would
be
nice
if
somebody
could
give
an
overview
of
that.
Just
that
I
know
I
understand
it
correctly.
A
Okay,
so
I
can
start
from
refresh:
it's
refresh
refers
to
controller
reconciliation,
so
by
default
controller
you
can
sell
application
every
10
minutes,
and
sometimes
you
want
to
do
it
quicker
than
that,
and
this
is
refresh
so
refresh.
Is
it
just
if,
when
you
click
refresh
button,
it
simply
put
an
annotation
on
application
and
once
controller
noticed
its
annotation,
it
knows
that
reconciliation
needed
now.
It
reconciles
it
and
remove
annotation,
and
this
is
very
easy
way
to
kind
of.
A
So
basically
you
can
click,
you
can
trigger
refresh
using
cube
ctl,
you
can
just
use
cube
ctl
patch,
and
then
you
can
use
cubicity
of
watch
to
watch
for
application
and
just
you
basically
waiting
for
a
moment
when
annotation
disappears
from.
C
A
C
All
right,
do
you,
do
you
mind
if
I
I'm
I'm,
I'm
still
also
a
new
user
at
argo
cd.
I've
been
using
it
for
a
couple
months
at
work.
If
I
give
you
my
kind
of
layman's
explanation
of
what
I
think
those
two
things
mean,
and
then
you
can
correct
me
so
so
with
refresh
there's
a
there's
two-step
operation,
the
first
step
is,
is
you're
watching
a
particular
get
ref
right,
and
that
might
be
a
branch
or
it
might
be
a
shaw.
C
If
it's
a
branch
or
a
tag,
you
have
to
resolve
that
to
a
sha
and
then
go
and
get
the
manifest
that
exists
at
that
point
in
time
in
the
git
repo,
then
compare
that
to
what's
in
in
in
kubernetes
right
so
refresh
is
that
step
is,
is
identify
what
it
is
that
you
might
need
to
sync
right,
and
then
sync
is
the
actual
is,
is
you
know,
for
a
lack
of
a
better
term
cube
ctl
apply
whatever.
That
is
that
you
that
you
found
out
is
that
accurate?
Well.
D
Yes,
it's
yeah,
I
think
that's
like
that's
a
great
way
to
understand
it
from
end
user
perspective,
so.
A
I
mean,
as
part
of
think
you
need
to
get
manifest
somehow
if
the
manifest
might
be
crashed
already
in
this
case,
there
is
no
need
to
talk
to
github.
If
not,
then
yes
need,
you
would
have
to
go,
clone
repository,
generate,
manifests
and
then
keep
ctl
applied.
So
sync
is
pretty
much
cuba
ctl
applied.
A
D
C
A
refresh
right
like
in
order
to
sync
you
must
also
refresh,
but
because
they
are
they
can
be.
You
might
want
to
just
refresh
without
making
any
changes.
Just
tell
me
what
you
would
do
if
there
are
changes
right
and
that's
what
refresh
would
do
and
then
sync
is
your
opportunity
to
then
go.
Do
that
so
a
sync
requires
a
refresh,
but
a
refresh
can
be
done
without
syncing,
yes,
yeah.
A
The
hardest
level
means
you
or
highest
level,
means
that
you
need
to
execute
less
remote,
the
git
operation
and
get
the
latest
commit
of
of
a
target
and-
and
then
you
need
to
you,
know,
compare
it
with
the
live
state
and
kind
of
compare
manifest
one
by
one
and
basically,
when
you
click
refresh,
that
means
you
need
to
reconcile
application
with
this
highest
level
of
the
heaviest
possible
reconciliation
type
and
the
other
types.
For
example.
A
Let's
say
if
port
goes
up
and
and
down
in
this
case,
we
just
know
that
port
is
not
managed
directly
and
we
simply
kind
of
refresh
list
of
child
application
resources
and
store
it
for
for
ui.
So
this
is
kind
of
the
least
expensive
way
of
reconciliation.
H
D
C
Would
would
a
given
that
you've
asked
the
question
and
I
can
definitely
see
why
there
is
confusion?
Would
this
be
an
opportunity
to
to
potentially
change
what
these
terms
are
like
what
what
is
presented
to
the
end
user,
because
I
can
I
can.
I
can
see
how
sync
and
refresh
to
some
people
mean
the
same
thing
and
it
would
be
confusing
to
like.
I
don't
understand
what
the
difference
is,
but
something
like
maybe
poll
right,
especially
you
know
what
poll
means
in
in
get
terms
right
poll
means
to
go.
C
Get
the
upstream
changes
right
doesn't
mean
apply
them,
it
just
means
get
them
right
and
sync
could
potentially
mean
okay.
Now,
now
that
you've
got
the
upstream
changes,
go,
do
something
with
them,
but
I'm
wondering
if
there's
an
opportunity
here
to
maybe
change
what
these
terms
mean.
Considering
you
brought
it
up
as
a
question
yeah,
so
we
could
do
yeah.
D
The
person
yes
ask
the
question
how
they
drop
off.
To
be
honest,
I
think
if
we
were
to
change
again,
I
think
we
we
really
should
only
have
two
user-facing
concepts.
Sync
and
refresh
resync,
I
think,
is
showbic
asked
the
question,
it's
slack,
because
I
think
he's
going
into
the
the
code
right
now
and
we
should
never
expose
the
word
resync
to
our
end
users.
I
think
that
will
definitely
confuse
our
end
users.
D
So,
with
the
with
just
sync
and
refresh
I
I,
I
can't
see
a
good
reason
to
change
that
terminology
to
to
our
users.
I
think
that
would
cause
maybe
more
confusion
and
I
think,
a
lot
of
more
than
just
eyebrow
razor.
I
think
you'll
you'll
get
objections
to
that
rename
as
well
like
the
api
is
named.
That
way.
The
crd
fields
are
auto.
Sync,
you
know
all
those
things
are
all
already
kind
of,
in
my
opinion,
set
in
stone.
C
I
think
we
don't
have
a
what
do
you
call
a
dictionary,
a
dictionary
yeah,
a
term
an
index?
If
you
is
that
the
term
I
don't
know
no.
I
D
I
Not
it's
not
complete,
so
yeah
and
the
core
concept:
okay,.
H
A
It
actually
it
has
it,
but
maybe
it's
not
you
know
we.
No
one
knows
it
like.
Even
we
didn't
remember.
C
D
Yeah
alex
I'll
post
that
maybe
we
jump
to
wes's
topic,
because
showbig
asked
a
bunch
of
questions,
he's
also
left
so,
oh
william,
but
since
wes
first
time
I
thought
I
didn't
want
him
to
walk
away
with.
C
C
I
mean
I'm
happy
to
bring
this
offline
too.
I
just
wanted
to
bring
up
a
a
potential
feature,
I'm
working
on
a
poc
for
this,
but
I
think
that
this
is
simple
enough,
that
this
could
just
be
a
simple
feature
in
argo
cd-
and
I
briefly
explained
it
earlier,
but
I'll
do
it
again,
which
is
this
idea
that
you
have
multiple
applications
that
depend
on
each
other
right.
C
So
you
have
a
staging
version
of
your
app
and,
of
course,
app
is
overloaded
here
so
I'll,
say
microservice,
because
argosy
doesn't
use
that
term
anywhere.
So
I
have
a
microservice
and
I
deploy
that
to
staging,
and
so
that
is
one
that
is
an
application
for
argo
cd
right.
It
goes
to
my
staging
cluster
and
then
I
have
another
application.
It's
the
my
production
version
of
that
microservice
and
I
want
to
promote
the
resolved
shaw
of
the
you
know
of
my
kubernetes,
manifest
from
staging
to
production.
C
C
At
the
same
time,
I
only
want
staging
deployed
and
then
only
when
staging
is
healthy,
so
using
all
of
the
built-in
functionality
of
argo,
cd
right
waves
and
and
phases,
you
can
do
basically
anything
you
want
right
like
post
sync,
you
can
run
various
jobs
to
you,
know,
run
tests
or
you
know
whatever
you
want
right
and
at
the
end
of
the
day,
you'll
have
an
application.
Health
state
right.
That's
like
this.
Application
was
successfully
synced.
Everything
ran
and
everything
exited
with
exit
code,
zero
and
we're
good
to
go
at
that
point.
C
Take
whatever
resolved
git
sha
that
happened
and
push
it
to
a
different
application
within
argo
cd
that
so
that's
the
underlying
concept
of
at
least
what
I
want
to
do
and
and
what
I
started,
the
other,
the
other
guy
named
his
project,
argo,
cd,
progressive
rollout,
progressive
rollouts
or
something
like
that,
and
I
think
he
might
be
wanting
to
do
a
little
bit
more
than
what
I'm
describing,
because
I
really
took
some
time
to
boil
it
down
into
the
essence
of
what
this
feature
really
is,
and
I
don't
want
to
do
any
more
than
just
this
bare
essence.
D
I
have
a
couple
questions:
have
you
have
you
so
people
have
been
able
to
achieve
a
similar
use
cases
using
an
app
of
apps
pattern
in
conjunction
with
sync
wave
annotation
on
the
apps?
Are
you
familiar
with
that
that
technique.
C
C
D
Apps
pattern
and
I'll
caveat
this
with
the
the
fact
that
we're
trying
to
move
away
from
that
pattern,
but
the
apple
apps
pattern
is
basically
you
have
an
umbrella
app,
which
is
containing
nothing
but
child
applications
and
that's
something
you
you
commit
and
check
into
the
git
repo
and
so
and
then
your
staging
and
qa
and
prod
child
applications
are
have
a
wave,
a
sync
wave
annotation.
So
you
want
your
qa
to
go
first,
so
that
has
a
wave
of
zero
and
then
your
staging
had
to
go.
Oh.
C
D
The
umbrella
app
has
no
way,
because
it's
just
a
single
app
that
you
think,
but
as
the
the
process
of
syncing,
that
umbrella
app
will
kick
off
like
because
in
cargo
cd
just
sees
these
as
resources
applications
as
resources
at
the
end
of
the
day
and
so
waving.
D
Those
applications
are
no
different
than
waving
like
a
config
map
before
you
know
a
deployment,
it's
the
same
exact
process,
and
so
what
people
have
done
to
to
achieve
a
similar
use
case
is
that
they
have
a
they
use
the
app
of
apps
pattern,
but
they
apply
wave
annotations
to
the
child
apps
in
order
to
achieve
a
pipeline
effect.
Oh.
D
C
D
That
is
all
that
helps
replace
part
of
that
pattern,
in
that
it
lets
you
kind
of
move
away
from
it
makes
it
easier
to
manage
those
child
applications
one.
What
it
doesn't
solve
is
that
that
waved
approach
that
I
just
mentioned,
and
we
had
another
proposal
about
something:
I'm
calling
it
a
a
a
sync
job,
but
I
don't
know
if
that's
gonna
be
what,
but
the
idea
behind
the
sync
job
is
that
just
like
you
have
kubernetes
has
like
a
job
object.
D
That
just
is
is
a
one-time
execution
of
a
a
pod.
Basically
right
and
then
you
know,
argo
workflows
is
another:
a
concept
of
a
ephemeral
run
of
something
in
in
argo
city.
Sync,
job
is
an
idea
where
that
that
thing
is
kicks
off
and
then
sinks
a
bunch
of
applications
in
some
some
order,
and
maybe
things
can
happen
between
that
or
whatever,
but
and
maybe
it
it.
D
It
looks
at
labeling
to
understand
how
to
do
the
ordering,
but
it's
basically
provides
a
more
sophisticated
way
of
chaining
together
sinking
of
applications
through
a
specification
of
how
you
describe
how
you
want
that
to
happen,
so
it
would
be
more
powerful
than
the
waived
annotation
approach
and
that's
something
we
have
been
actually
talking
about
for
a
long
time.
I
just
haven't
had
the
time
to
or
a
person
a
resource
to
to
actually
investigate
that.
So
so.
C
So,
actually,
another
follow-up
question
regarding
that.
You
know
the
the
what
is
open,
open
question,
which
was:
when
does
argo
cd?
Consider
something
synced
right
so,
like
as
an
example,
a
job
can
be
successfully
deployed
and
started
right,
but
not
necessarily
ended
right.
That
job
doesn't
have
to
end
for
argo
cd
to
be
like
yeah,
I'm
good
on
my
end
right
whether
that
job
succeeds
or
not
is
not
anything.
I
need
to
be
worried
about
that,
potentially
would
be
a.
D
The
answer
is
pretty
easy
for
sync.
Basically,
if
you
run
it
should
be
the
same
answer
if
I
ran
keep
cl
diff
on
the
on
this
resource
and
that
keeps
you
diff
says
no,
there's
yeah,
there's
no
difference,
then
we're
in
sync
and
that
and
then
that
the
application
sync
status
is
in
the
aggregate
status
of
all
of
the
resources.
Are
they
all?
Is
there
any
difference
and
and
that
aggregates
up
to
the
application?
And
if
there's
no
different
any
of
my
application
resources,
then
I'm
in
sync
got
it
okay,
so
it
doesn't.
D
D
Of
an
object
so
so
yeah,
so
that's
why
we
have
two
statuses:
the
sync
side,
which
is
nothing
but
is
rna
differences.
Health
is
understanding
looking
at
a
single
resource
and
looking
at
status
and
understanding.
Okay,
do
I
have
three
out
of
three
replicas
and
all
are
all
those
three
replicas
available?
Then?
Yes,
my
deployment
is
considered
healthy,
and
so
we
have
built-in
health
checks
for
many
of
the
native
kubernetes
kinds,
which
makes
sense
to
have
a
health
status.
But
things
like
this
config
map
has
doesn't
make
sense
to
have
health
right.
F
C
Okay,
that
sounds
good
I'll,
post
a
I
guess,
a
discussion
about
this,
and-
and
maybe
we
can
follow
it
from
there
on
the
github
project.