►
From YouTube: Argo Contributors Office Hours 30th Sep 2021
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
Recording
okay
good
morning,
everyone
welcome
back
to
the
argo
contributors
meeting,
so
I'll
go
ahead
and
share
my
screen
with
the
document.
A
All
right
great,
so
we
have
a
pretty
packed
agenda
today.
So
let's
just
get
right
into
it.
Are
you
you
wanted
to
disagree?
Yeah.
I
think
let's
wrap
this.
B
Up
quickly
today,
as
we
have
so
many
items,
but
the
fundraising
discussion,
the
last
week
is
leo
and
ishita.
C
Yeah
sure
so
last
week
the
volume
was
not
much
for
discussions.
There
was
one
q
a
and
two
general
questions
which
were
asked,
so
I
could
answer
two
of
them.
One
of
them
was
regarding
rollout,
so
one
of
that
pending
it
was
not
as
a
q
a
it
was
just
a
question
which
was
added.
So
one
of
that
was
left
and
added
the
one
which
I've
added
in
the
chat.
I've
added
that
to
2.2
milestone
yeah,
that's
it
from
my
side
and
charging
was
fine.
D
Yeah
I
wanted
to
to
ask
jesse
if
this
is
maybe
probably
similar
to
the
problem
that
we
were
facing,
because
this
is
maybe
you
can
send
the
link.
E
Okay,
I
saw
this
same
channel
and
sorry
I
saw
the
same
message
in
the
slack.
I
didn't
notice
it
in
the
discussions.
E
So
something
didn't
add
up
here
and
they
mentioned
that
the
rollout
or
the
application
was
in
a
suspended
state,
but
nothing
in
this
rollout.
They
don't
have
anything
that
could
cause
it
to
become
in
a
suspended
state.
So
I
don't
sure
if
it's
incomplete
it
does
kind
of
look
incomplete
because
I
have
labels
and
nothing
is
on
the
labels.
But.
E
I,
like,
I,
don't
believe
anything.
Rollout
controller
is
doing
wrong
right
now
at
least
there's
nothing
that
could
cause
it
to
be
suspended
and
that's
what
I
asked
in
the
slack.
D
Okay,
I
was
thinking
more
about
the
aob
case
that
we
were
trying
to
fix
if
this
was
maybe
something
related
to
that
or
not,
but
yeah
there's
not
much.
No.
E
They're
using
basic
canary,
if
I'm
not
mistaken,
yeah
canary
without
traffic
management.
I
remember
so
so.
This
rollout
is
essentially
kind
of
like
a
rolling
update
and
we
don't
even
touch
or
care
about
any
service
stuff
like
they
don't
even
specify
a
service
for
us
to
modify.
So
the
only
thing
we're
doing
is
bringing
and
bringing
up
pods,
and
so
the
only
possible
bug
that
could
be
here
is.
E
If
we
we
brought
down
a
pod
before
it
could
be,
so
they
only
have
one
replica,
and
so,
if
we
deleted
that
and
then
brought
the
other
one,
that
could
be
what
they're
seeing.
E
But
I
I
I'm
pretty
sure
that's
that's
working
if
they
have
max
search,
25
and
we'll
search
up
and
then
and
then
max
and
available
zero.
I
think
we'll
also
make
sure
there's
always
one.
E
Anyways,
I
responded
in
slack
to
this
so
and
I
think
this
one
needs
more
information
I'll
paste.
My
response
in
this
thing
as
well.
C
D
C
Just
one
more
thing:
there
is
one
more
issue:
I've
started
in
the
chat
I
just
wanted
to
know
like
would
that
be
a
good
candidate
for
2.2
like
I
went
through
this
one,
it's
regarding
helm,
so
the
user
has
just
added
all
the
issues
which
he
he
has
seen
in
the
past
in
this
issue.
So
do
we
think
this
could
be
done
for
2.0.
F
C
That's
what
the
user
was
mentioning
in
this
I
couldn't
like.
I
didn't
get
time
to
reproduce
it,
but
yeah.
I
thought
of
taking
a
look
later
on
on
this
one,
so
just
wanted
to
know
like
I
could
invest
some
time
on
this
one
to
speak
and
discuss
the
next
week.
If
we
could
add
this
for
the
next
release,.
F
I
definitely
know
that
we
have
a
lot
of
complaints
recently
about
this
particular
problem
and
I
guess
that
issue
is
like
a
summary,
so
I
tend
to
believe
it's
true,
so
yeah
I
I
I
agree.
I
would
add
a
label,
you
know
and,
and
even
you
know,
add
it
to
the
milestone
and
add
a
label
gdp
to
2.1,
because.
C
The
user
also
has
raised
a
pr
similar
on
a
similar
line,
so
I
I
was
trying
to
even
review
the
pr,
but
I
had
a
little
time
crunch
for
this
week,
but
yeah
I'll
try
to
go
through
the
pr2
for
this.
Thank
you.
F
Yeah
yeah,
I
I
I
feel
like
I
also
didn't
get
time
to
get
there,
but
there
was
a
recent
change
in
helm,
and
so
you
know
like
helm,
dot
seven
is
looks
like
very
different
from
helm.
Three
dots.
I
think
we
have
four
or
six.
I
forgot,
but
basically
the
minor
version
that
we
package
is
very
different
from
the
most
recent
minor
version
available,
and
can
I,
if
I'm
not
mistaking,
we
were
hoping
we
had
to
upgrade
to
the
latest
version
and
we
could
not
upgrade
because
of
these
differences
and
so
yeah.
E
This
particular
issue:
okay,
okay,
so
yeah,
I
think
definitely
we
I
think
we
each
sounds
like
something
we
should
fix.
When
caveat,
is
I
I'm
wondering
if
it's
possible
for
so
since
it
could
be
a
disruptive
change,
even
though
it's
the
right
thing
to
do?
Is
it
will
be
possible
for
people
to
say
downgrade
the
helm
version.
F
In
place,
I
think
it's
kind
of
you
know
to
take
advantage
of
mostly
some
changes
in
helm
to
actually
fully
support,
cube
version
and
ipa
versions.
We
need
to
upgrade
and
we
could
not
upgrade
because
of
some.
You
know
just
the
upgrade
is
not
that
smooth.
In
order
to
upgrade.
We
have
to
make
changes
in
argo
cd
code
so
that
it
works
with
newest
conversion
yeah.
E
E
Right,
yeah
yeah,
but
because
I
I
yeah
my
point
is
that
I
think
it
might
be
we
we
may
want
a
way
to
say
someone
upgrades,
I'm
like.
Oh,
no,
I'm
not
ready
for
this
difference
and
rendering
of
my
templates
that,
let
me
stick
with
tom
3.6.
E
Okay
yeah,
it
sounds
like
something
we
should
do,
but
provide
an
escape
patch
in
case
3.7
is,
has
its
own
issues.
A
All
right
great
was
there
anything
else
that
we
wanted
to
discuss
from
the
triage
and
discussion.
B
Okay,
great
okay,
can
we
fix
the
primary
and
secondary
for
the
week?
Oh.
A
Yep,
absolutely
so
ishida
and
leo
did
it
last
week.
So
are
we
going
to
move
on
to
andre
and
maybe
jonathan
or
looks
like
this
order
might
be
a
little
bit
wonky?
How
do
we
want
to
do
this.
A
Yeah,
I
think
you're
right,
okay
is
jonathan.
Are
you
here,
okay,
so
jonathan's?
Not
here,
let's
see
jesse,
have
you
done
it
recently?
I
feel
like
you,
you
might
have
done
it
recently.
I.
A
Yeah
alex,
do
you
have
bandwidth
to
do
it
or
or
have
you
done
it
recently?
I
didn't
hear
you.
I
should
be
actually.
A
Sounds
good,
do
you
want
a
secondary,
or
are
you
good
by
yourself.
F
I
mean,
I
think
it's
good
to
have
a
secondary.
You
know
if,
if,
if
someone
hasn't
been
this
dependent,
then
please
join
me.
A
Who
is
that
speaking?
Was
that
ishida
yeah?
Okay,
so
we'll
keep
you
on
a
secondary
okay,
great!
Thank
you
all
right,
so
moving
on
the
contributors
meeting
will
be
moved
to
8
15.
alex.
Do
you
want
to
talk
about
that
briefly?.
F
Yes,
so
I
guess
we
spoke
about
it
during
class
meeting
and
you
know
we
kind
of
synced
up
to
flying
looks
like
there
is
no
more
overlaps
and
next
starting
from
the
next
week.
I
think
meetings
are
going
to
be.
You
know:
8
15
to
9
yeah,
9
a.m
sounds.
A
Good
yeah,
I
updated
the
window
already
awesome.
I
think
that
the
calendar
last
time
I
checked
said
eight
to
nine,
so
it
changed
it
like
five
minutes
ago.
Oh,
okay,
great
all
right,
sounds
good,
awesome,
all
right!
So
jesse
you
had
a
topic
about
how
values
from
get
repo.
E
Yeah,
I
think,
oh
wait.
Sorry
I
linked
in
zoom
chat.
E
You
can
bring
up
the
the
issue,
there's
actually
multiple
issues,
I
think
all
duplicating
the
same
feature
enhancement-
and
this
is
also
our
most
popular
thumbs
up
issue,
but
to
catch
people
up
our
users
have
a
strong
desire
to
use
the
helm,
repo
feature
of
argo
cd,
meaning
like
they
just
point
it
to
a
helm,
chart
and
then,
but
the
value
files
that
they
want
to
use
for,
that
helm
chart
be
stored
in
the
git
repo,
and
today
we
don't
support
that.
E
We,
our
workaround,
is
that
we
tell
people
to
point
their
app
to
a
git
repo
which
has
a
helm
chart
in
it
and
that
helm
chart
is
an
umbrella
chart
for
and
then
points
to
the
the
real
chart.
It's
it's
not
like
a
great
work
around,
because
the
values
file
is
actually
you
can't
just
you
have
to
nest
all
of
the
the
fields
into
a
and
to
one
one
level
deeper
because
of
the
way
umbrella,
charts
work,
and
so
it's
a
bit
of
a
bad
experience.
E
So
this
this
feature
request
is
to
support
home
repos,
but
with
value
files
that
are
stored
in
git,
which
on
paper
sounds,
you
know
pretty
easy
or
like
a
good
idea.
The
challenge
is
that
this
is.
This
will
be
the
first
time
argo
cd
has
to
deal
with
two
repositories
for
an
application.
E
It
has
to
deal
with
the
helm
repository,
which
is
monitoring
the
tip
of
to
determine,
if,
like
oh,
did,
anything
change
with
respect
to
the
version
or
and
stuff
and
then
and
it
today
we
cached
that,
like
which
attached
a
tip
and
then
we
detected
if
something
changed
now,
if
we're
referencing
something
in
a
kit,
we
also
have
another
thing
that
can
change
so
that
that's
the
technical
reason
why
I
probably
never
did
anything
about
this,
but
it
is
a
strong
use
case
that
I
I
think
we
should
address
so
to
to
summarize.
E
I
think
we
should
do
it,
but
I
think
there's
two
considerations
that
need
to
be
made.
One
is
the
caching
problem,
caching,
both
the
like.
Basically,
if
anything
changes
the
head
of
the
get
values
file
repo
or
if
the
helm,
repo
changes,
then
we
need
to
invalidate
cache
and
then
second
there's
like
a,
I
think,
there's
a
r
back
consideration
as
well,
because
our
project
repo
allow
list.
E
E
F
Think
I
kind
of
I
feel
like
maybe
my
main
reason
why
we
are
not
accepting
that
pr
is
because
it
introduces
a
lot
of
like
a
possibility
of
of
a
bug
and
support
work
just
for
one
age
case
so
and
but
I
agree
with
you
that
it
has
so
many
evidence
that
it
is
required,
but
I
would
try
to
find
a
way
to
you
know
if
we
introduce
support
for
kind
of
two
repositories.
I
would
not
limit
it
to
him.
F
E
Right
right,
one
thing
to
consider
is
that
there's
actually
a
pr
outstanding
for
this
that
hasn't
gotten
the
attention
and
then
well.
I
think
I
I
agree
with
what
you're
saying
I
don't
know
if
the
the
contributor
will
have
the
to
be
able
to
extend
it
in
that
flexible
manner,
especially
if
I
I'm
going
to
ask
them
to
raise
a
requirement
and
say
hey.
We
also
need
our
back
consideration
here
and
also
get
cash
in
consideration.
F
So
biggest
problem
and
yeah
like
it's
so
as
long
as
as
soon
as
you
introduce
git
caching
and
there's
actually
two
level
of
caches
like
we
will
cache
revisions,
and
then
we
cache
content
like
we're
trying
to
not
to
you
know,
generate
content
twice
so
yeah
that
then
that
pair
will
suddenly
became
more
complex,
and
I
suspect
that
it
can
be
almost
as
complex
as
if
we
were
trying
to
just
support
multiple
kids.
E
So
one
one
question
like
do:
one
of
the
suggestions
and
the
pr
wisdom
merge
this
or
as
close
to
as
is
without
the
caching
and
then
the
in
the
description
the
person
said
they
would
follow
up
with
a
with
that
they
don't
they
don't
mention
this
caching.
They
mention
that
as
detection,
when
the
fix
the
get
rebuilt
changes
are
we
are
we
okay
with
the
split
commit
in
that
way,
or
do
we
want
and
same
goes
with
our
back?
E
F
F
One
is
a
couple:
users
complained:
okay,
not
couple
like
okay,
there
was
an
issue
created
and
I
think
couple
people
confirmed
that
they
saw
it
in
production.
So
it
is,
it's
probably
it
is.
There
is
a
possibility.
We
can
get
a
less
remote
for
too
long
or
I
think
it's
a
h
case.
F
Maybe
the
feeling
is
that
we
cache
it
appropriately,
but
maybe
we
resolve
so
basically,
if
users
use
feature
branches
and
they
rewrite
history,
then
sometimes
argo
cd
tries
to
check
out
kamit
which
no
longer
belongs
to
any
branch
yeah,
and
then
we
get
to
fix
it
first
and
then
we
just
spoke
about
health
issues,
so
yeah.
F
Basically,
I
feel
like
our
answer
is
yes,
maybe
it's
okay
to
merge
it,
but
we
just
should
I
don't
want
to
introduce
instability
into
already
unstable
portion
of
the
system.
So
it's
just
due
to
bandits
like
we
were
kind
of
busy
recently
with
the
kubecon,
and
I
I
was
hoping
to
jump
on
bug,
fixes
and
like
fix
bug,
fixes
first
and
then
try
to.
E
This
is
what
I
think
I
think,
I'm
okay
splitting
the
r
back
into
a
separate
one,
but
I
I
think
I
really
want
to
make
sure
that
the
caching
is
part
of
this
one
yeah.
I
agree
because
so
well,
our
back
actually
seems
like
we
could
that
that
actually
doesn't
seem
too
terrible
to
do
like
quickly
after
there's
a
follow-up,
but
I
I
really
want
to
make
sure
that
our
back
is.
I
might
check
that
the
caching
is
considered
as
part
of
this
one
but
okay.
E
E
F
E
E
Yeah
yeah,
that's
what
I
mainly
want
to
get
out
of
this
is
that
at
least
we
decide
like
yes,
we'll
do
it,
and
then
it
may
take
some
time
to
power
through
this
pr
because,
like
I,
like
you
said,
there's
probably
a
lot
of.
E
E
I
think,
at
the
very
least.
We
owe
it
to
the
contributor
to
to
agree
on
syntax,
and
I
I
looked
at
it.
It
looked
okay
with
me
and.
F
Like
just
another,
I
want
to
try
one
more
time.
How
do
you
think
if
we
consider
introducing
multiple
sources
and
kind
of
you
know
like
kill
two
birds
or
how
is
it
you
know?
Basically,
if
we
introduce
field
sources,
and
so
it
can
has
either
git
repo
or
helm
repo,
and
then
we
make
sure
so
it
kind
of
provides
two
opportunities
like
it
can
be,
not
two.
It
can
be
like
many
sources.
Basically
that's
already
it's
better
than
you
know
in
this
particular
and
it
it
won't
have
to
be
just
film.
F
E
Plus
that
feature,
I
think,
would
be
nice
to
have
a
lot
of
times.
People
want
a
helm
chart
and
then
they
want
like
a
name
space,
but
they
want
that
name
space
to
also
have
like
some
annotations
and
stuff
like
that,
and
then
so
they
need
to
get
option
namespace,
but
they
also
need
to
get
up
to
their
home
or
they
they
they
want
to
help
chart
it,
and
we
don't
that
experience
is
also
hello.
F
F
We
would
create
a
temp
directory
for
a
combination,
I'm
just
thinking
about,
like
let's
say
if
you
can
assume,
if
you
can
use
relative
pass
in
helm,
repo
and
kind
of
you
know
step
out
out
of
the
home
repo,
and
you
should
expect
that
all
the
sources
mentioned
in
this
particular
app
spec
available
through
the
through
some
stable
relative
place,
then
we'll
support
this
use
case
as
well
as
multiple
sources,
and
it
may
be
the
same
same
amount
of
work
almost
and
it's
no
longer
specific
to
him.
Okay,.
E
I
see
what
you're
saying,
okay,
so
the
the
change
we
would
have
to
ask
this
person
to
do
then
would
be.
G
F
I
mean
yeah,
we
can,
at
least
you
know
if
I
I
feel
like,
maybe
that
person
spends
so
much
time.
I
guess
my
maybe
because
that
feature
is
so
much
required,
so
the
person
is
willing
to
keep
keep
pushing
it.
What
if
we
at
least
offer
a
solution
and
we
pursue
it
in
two
or
three?
E
And
just
do
it
ourselves.
Are
you
saying
yes
yeah,
I
I
I
don't
mind
that
with
a
tree
there,
so
I
okay,
then,
if
that's
the
case,
what
I
can
respond
is
here
are
the
requirements
that
we
have
for
for
us
to,
or
just
this
feature
in,
like
the
our
back,
the
the
the
framework
for
multi-repo
support
and
see,
if,
if
they're
willing
to
kind
of
keep
going
with
this,
if
not,
then
we
can
commit
it
to
the
following
release,
but
not
maybe
2.2
yeah.
F
And
I
also
you
know,
I'm
kind
of
in
the
background
I'm
trying
to
find
that
request
to
support
multiple
sources.
I
cannot
even
find
it
right
now,
so
I
would
you
know
I
just
wanted
to
consider
it.
You
know
have
discussion
about
this
as
an
alternative
way
to
operate
yeah.
So
I
guess
I
will
just
comment
the.
F
F
E
All
right,
I
think
I
have
enough
what
I
I
need
for
from
this
discussion
to
do
it
and
it
sounds
like
we
are
most.
I
think,
we're
on
the
same
page
actually.
E
I'll
I'll
respond
in
the
book
and
then
I'm
gonna
duplicate,
like
all
of
the
issues
to
that
one,
the
one
with
the
most
thumbs
up,
there's
actually
like
four
issues.
All
asking
for
the
same
thing.
A
All
right
great,
so
in
summary,
we're
going
to
agree
to
go
forward
this
for
2.2
and
tell
the
contributor
that
we'd
like
it
to
meet
these
requirements
that
we've
discussed
and
if,
if
you
can't
do
that,
then
we'll
target
it
for
2.3
to
do
ourselves
all
right,
great
okay.
So
next
topic
is
from
regina
to
israel,
ocean's
hacktoberfest!
So
do
you
wanna
discuss
this
regina.
H
Yeah,
I
just
think
we
I
mean
we
did
it
last
year.
I
believe,
but
the
just
opening
up
argo
cd
is
like
a.
H
I
don't
exactly
know
how
you
do
it,
but
able
to
count
towards
contributions.
F
I'm
yeah
I've
had
a
conversation
like
at
intuit.
We
had
internal
requests
to
support
that
event
and
I
feel
like
I.
I
also
want
to
participate.
You
know
it
at
least
like
I
feel
like
for
us
to
purchase
participate.
We
just
need
to
go
through
outstanding
issues,
find
good
for
new
contributors
and
add
have
cryptographic
label
to
them.
H
F
F
H
H
Yeah,
it's
not
specific
to
any.
I
think
intel
is
part
of
it
too,
but
it's
not
specific
to
any
company
necessarily.
F
Yeah
so
but
at
least
there
is
push
from
a
couple
sides
to
yeah.
So
I
was
planning
to
you
know,
spend
some
time
today
and
tomorrow
and
have
at
least
a
couple
dozen
of
such
issues,
labeled.
F
Yeah-
and
I
know
that
the
reason
why
I
was
asked
to
do
it
is
because
I
guess
you
know
intuit
will
sponsor
some
time
for
developers
internally
and
that
they
will
encourage
them
to
look
for
these
labels
and
contribute.
So
we
might
suddenly
get
you
know
a
bunch
of
prs
open
from
both
experienced
contributors.
F
I
Yeah,
so
I
I
think
it
is
probably
worthwhile,
and
luckily
we
have
a
large
enough
community
that
we
can
and
we've
got
some
people
triaging.
We
can
probably
go
and
knock
them
out
and
me,
but
just
to
turn
it
on.
I
think
all
you
do
is
add
a
label
to
the
repo.
It's
not
an
issue
level
thing.
It's
just
a
repo
level
thing:
okay,.
F
A
All
right,
great,
so
yeah,
we'll
label
the
repo-
and
I
I
do
remember
last
year
we
added
a
bunch
of
act
over
fest
labels
to
the
to
the
good
first
issue.
So
I
think
that's
a
good
practice
to
continue.
Okay,
so
pasha
you
have
the
topic
for
resource
tracking
changes.
You
wanted
to
give
a
demo.
A
Yeah
hear
all
the
stuff
sure.
K
I
hope
some
of
you
are
already
familiar
with
this,
but
I
will
provide
a
short
intro
and
problems.
What
I
am
solving
and
motivation
of
this
feature.
So,
as
you
know,
each
resource
in
argo
city
is
synchronized
from
hit
repo
marked
with
label
which
named
up
kubernetes
io
instance,
and
we
meet
problem,
and
a
lot
of
people
start
complaining
about
this
issue
that
they
cannot
create
here,
actually
few
issues
that
we
are
trying
to
solve,
but
most
of
most
important
of
them
that
first,
that
you
can
cannot
create
application
name
more
than
53
characters.
K
Its
first
issue.
The
second
issue
that
already
the
popular
home
charts,
already
use
this
label
and
in
case,
if
you
use
some
charts
and
not
synchronize
it
with
argo,
argo,
cd,
identify
it
as
own
resources
and
show
it
in
ui,
and
it
provide
problems
here
and
third,
one
is
the
same
thing
but
kubernetes
operators
so
and
here
few
more
issues.
But
this
is
main
problems
so
solution
here.
K
It's
provide
ability
to
switch
from
label
to
annotation,
and
for
this
we
define
a
tracking
mode
which
tracking
method
which
you
can
define
on
settings
level
and
on
application
level
on
application
level,
it's
high
your
override
settings
level
and
so
on,
but
we
have
three
levels.
First,
one
we
have
three
modes
method.
First,
one
is
label,
it's
default
one,
so
it
will
be
working
the
same.
K
Actually
why
we
need
it.
We
need
it
for
for
migration,
we
cannot
just
say:
okay,
guys,
we
remove
everything
to
annotation
and
now
everything
out
of
sync
for
you.
So
we
provide
a
smooth
way
for
migration.
You
can
go
to
your
settings
and
define
annotation
either
annotation
plus
label
method,
or
you
can
stay
the
same
way.
It's
not
required
for
you
if
you
not
meet
problems
with
application
names
and
so
on,
and
so
this
is
so.
This
is
actually
my
idea
that
you
can
go
to
your
configuration
map.
K
And
from
this
point,
argo
cd
start
to
track
resources
in
different
way
in
additional
in
annotation.
We
start
to
record,
so
I
want
to
find
this
section
yeah
in
annotation.
You
can
see
that
it's
that
we
start
record,
not
just
application
name.
We
start
define
additional
metrics
here
for
provide
uniqueness,
so
here
an
additional
issue
when
the
tools
multiple
applications
looking
for
same
resource,
but
they
can
be
in
different
name
space
and
and
so
on,
and
in
such
case
it
will
provide
you
ability
to
solve
this
problem.
Such
such
names.
K
So
I
would
briefly
demonstrate
it
and
explain
what
is
already
done
and
what
is
my
next
steps.
So
we
just
need
to
check
because
we
have
limited
session.
K
K
K
And
after
this
I
need
to
hard
refresh
application,
because
we
do
internal
cache,
but
it
will
be
solved.
It's
one
of
the
points
that
I
need
to
change,
and
you
can
see
right
now
that
resources
state
to
be
out
of
sync
for
you,
because
it
starts
to
start
to
track
applications
in
another
way.
You
can
see
that
this
still
use
test
application,
name
and
namespace
and
another
matrix
and
before
we
had
instance
test
so
after
synchronization
once
I
done
it,
it
will
be,
it
will
behave
in
the
same
way,
but
just
in
more
reliable
measure.
K
What
what
should
be
done
here.
All
these
things
that
I
already
pushed
covered
with
unit
test
and
end
time
test
fully
tested.
But
here
a
few
things
that
I
want
to
add.
First
one
is
ability
to
override
these
settings
on
application
level.
This
is
first
one
second,
one
its
ability
to
define
this
setting,
not
just
from
crd
cons,
not
just
from
configmap.
I
want
to
provide
ability
to
do
it
from
cli
and
from
from
ui.
K
It's
a
good
point
that
maybe
it's
good
place
to
provide
some
simple
settings
page
in
argo
ui.
These
few
points
and
yeah
and
that's
it
and
it
will
be
final
stage
of
this
of
this
feature.
I
I
definitely
assure
that
it
will
be
done
till
2.2
milestone.
I
hope
actually
get
it
working
in
coming
quick
next
week,
because
most
of
work
already
already
done
here.
F
I
see
a
bunch
of
questions
from
jc.
I
can
answer
some
because
I
think
it's
kind
of
I
might
know
the
answer
better,
because
I
was
you
know
working
on
proposal.
I
will
keep
first
try
see
we
can
discuss
the
name.
I
guess.
But
so
one
question
is:
why
do
we
have
it
on
up
level
and
system
level,
and
that
was
for
migration?
F
Basically,
we
were
hoping
that
in
the
as
soon
as
feature
is
released,
we
can,
you
know
mention
in
upgrade
nodes
and
then
release
block
that
we
want
everyone
to
eventually
migrate
to
annotations,
and
so
they
will
have
an
option
to
start
enabling
enabling
its
own
application
level
one
by
one
and
then
maybe
in
next
release.
We
can
change
default
so
that
people
will
have
now
option
to
switch
to
label.
E
What
is
the
the
not
smooth
thing
about
control
controlling
this
at
the
system
level
and
saying.
K
E
What
I
don't
understand
it's
like,
why
would
someone
enable
this
feature
on
at
the
app
level
if
like?
Why?
Wouldn't
we
they
just
go
to
the
system,
say
labels
plus
annotations,
but
we
would
maybe
have
a
way
to
prevent
the
other
sync
condition
from
happening
or
like
because
we
know
these
these
fields
are
we
control
these
fields
and
we
can
decide
like
if
they
should
or
should
not
factor
into
the
other
sync
status.
F
We,
when
we
switch
to
label
plus
annotation
label,
is
not
really
good
enough
for
tracking,
like
okay,
it
was
not
a
question,
but
we
have
label
plus
annotation,
because
we
wanted
to
you
know
a
lot
of
users
use
label
to
be
able
to
run
something
like
cube.
Ctl
get
all
these
resources,
all
minus
cell
and
then
application
name,
and
we
just
want
to
keep
that
feature.
F
E
F
F
Yeah,
so
it's
like
imaginary
case,
if
you
have
a
lot
of
users
and
one
of
them
you
know
like
is
a
difficult
user
and
he's
like
no
don't
touch
me.
I
don't
want
to
be
migrated
right
now,
so
you
can
do
it
in
waves,
like
you,
you,
you
can
migrate.
Whole
argo
cd
to
that
new
way,
except
one
particular
user
who
prefer
to
stay
yeah,
and
I
feel
I
I
understand
that
it's
like
additional
complexity,
and
you
know
the
pull
requests
that
user.
Do
they
it's
it's.
F
You
know
we
just
had
a
kind
of
discussion
and
we
were
trying
to
see
okay.
How
do
we?
How
do
we
ask
people
to
migrate
to
that
new
feature?
And-
and
that's
why,
like
first
decision
that
we
made
is,
it
has
to
be
controlled
by
some
setting
and
then
and
then
we
try
to
imagine
okay
like?
Is
it
enough
or
not?
And
so
we
propose
to
have
application
level
setting
as
well,
because
it's
kind
of
not
too
difficult
to
it.
Yeah.
E
Is
it?
How
does
that?
How
does
someone
opt
in
on
the
app
level
to
say
I
want
to
be
tracked
this
way.
F
E
See
I
that
it
seems
that
app
level
tracking
isn't
something
that
actually
when
it
doesn't
seem
like
a
concern
for
for
say,
like.
B
E
And
if
I
originally
thought
it
was
an
issue
of
user
experience
like
if
we
were
suddenly
there's,
things
went
all
out
of
sync
just
because
the
operator
decided,
but
that
that
part,
I
think
we
can
address
in
the
implementation
to
to
accommodate
for,
like.
Oh,
we
don't
mar,
don't
show
something
as
out
of
sync.
If
it's,
the
only
reason
it's
out
of
sync
is
because
the
set
the
the
global
setting
you
know
change
like
I
feel
like
there's,
there's
mitigations.
We
can
do
there
to
make
the
transition
smoother.
E
It's
okay,
but
adding
it
to
the
specs
seemed
a
little
weird
because
I
don't
think
any
end
user
would
actually
want
to
to
perform
that,
and
it
also
might
even
be.
E
Like
undesirable
behavior
to
for
for
them
to
have
that
ability
for
some
reason.
F
It
did
like
we
kind
of
the
pr
is
not
merged
and
it
was
too
late
because
we
realized
you
know
that
that
setting
kind
of
it
introduced
one
more
possibility.
You
can
have
two
applications
managing
the
same
resource,
one
using
label,
one
annotation
so
yeah
and
it-
and
I
I
didn't
think
that,
can
we,
I
think
yeah-
maybe
put
get
it
like
this
pull
request
is
not
yet
merged,
but
I
agree.
I
I
think
we
just
need
to
find
kind
of
think
about
this.
F
E
At
least
I
know
from
for
you
know
into
it
the
developer's
gonna
be
like
they
don't
know
why
the
heck
they're,
yeah.
F
Like
internally,
we
would
backfill
that
value
on
behalf
of
users
like
they
would.
They
would
not
even
know
that,
and
you
know
when
you
mention
that
I
thought
that
yeah
maybe
you're
right.
Maybe
it
should
have
been
done
on
project
level.
It's
like
a
little
better
because
you
kind
of
most
likely
you
would
set
the
setting
for
the
whole
project.
F
E
If
you
are
going
to
do
this,
though
it
seems
like
it's
also
not
a
spec,
I
wouldn't
put
it
in
the
spec,
because
this
is
a
transition
period
and
then
we're
it
feels
like
we're
baking,
something
in
the
spec
that
is
actually
temporary.
F
It
is
actually
like
the
steel
setting
with
the
station
will
have
two
values:
eventually
annotation
or
annotation
plus
label.
There
was
a
concern
to
even
keep
label
because
people
might
rely
on
it,
but
we
know
that
label
is
not
reliable.
F
It's
not
reliable
because,
let's
say
we're
going
to
truncate
application,
name
right
and
so
sorry
truncate
label
value,
and
so
we
might
have
in
the
kubernetes
cluster
a
bunch
of
resources
labeled
with
the
same
label,
but
but
they
would
belong
to
two
different
applications.
You
know
if
application
name
is
extremely
long
and
first
right
right
here,
yeah
so
yeah.
Basically
the
setting
will
not
be
deleted.
F
F
L
E
I
mean
we
already
have
kind
of
this
framework
in
place
because
we
have
ignore
differences.
It's
currently
at
a
kind
specificity,
but
we
would
need
to
ignore
differences
at
annotations.you
know
app
dot,
instant
site
instance,
and
so
the
essentially
to
to
to
have
that
feature.
They
ignore
the
the
that
difference
like
we
already.
We
have
the
code
to.
K
E
So,
okay,
sorry
and
then
last
question
is:
is
this
curious
why
we
need
to
track
names
based
in
the
id.
F
So
that
was
I
mean,
so
I
think
it
was
not
explained
like
any
description,
but
it's
there
in
a
proposal.
We
want
to
make
sure
that
if
someone
copy
a
label
from
the
resource,
typically
the
typical
use
cases,
one
resource
is
kind
of
a
template
for
another
one
and
operators
often
copy
all
labels
and
all
labels
honestly
and
annotations
actually
whole
metadata
from
the
template
into
you
know
the
new
resource.
F
F
F
F
E
No-
and
I
I
I
think
I
agree,
I
agree
that
it's
it's
a
more
consistent.
I
mean
coding
wise
to
just
refer
to
this
as
the
argo
cd
resource
id
and
there's
that's
just
the
id.
The
tracking
id
is
the
resource
id
and
then
yeah.
That's
good
enough.
F
E
A
Yeah
jesse,
I
think
if
we
want
to
have
a
discussion
about
the
design
system,
I
think
that
the
best
place
for
that
would
be
the
sig
ui
meeting
so
next
wednesday,
if
you
guys
want
to
tune
in
to
discuss
about
that,
I
think
that's
a
good
place
to
do
that.
K
E
That
was
why
I,
I
suggested
it
because
this
will
kind
of
blow
up
into
just
a
free-for-all
of
opinions.
So,
and
I
actually
want
to
remove,
even
including
myself
from
from
the
distractions.
A
And
to
be
clear
too,
with
that
meaning
anybody
that's
interested
can
can
join
it's
not
yeah
exclusively
for
ui
contributors.
It's
if
you're
interested,
please
join
so
all
right
yeah!
So
who
would
you
want
to
discuss
support
for
our
role
at
sun.
G
Yeah
sure
sure
sure,
thanks
yeah,
so
so
hi
everyone,
my
name-
is
hui,
I'm
currently
working
at
salesforce
and
adopting
argo
roads
into
our
cicd
infrastructure
yeah.
So
for
this
for
this
feature,
I
just
want
to
provide
a
little
bit
of
context
in
this
work.
So
when
we
adopting
the
argo
rollouts
in
our
production
system,
one
of
the
major
concern
from
the
leadership
team
is
the
lack
of
high
availability
for
the
controller,
which
means
there
could
be
only
one
controller
running
in
the
cluster.
G
Then
we
find
that
there
was
actually
an
existing
pr
that
implemented
the
aha.
However,
this
pr
was
like
submitted
back
in
early
june
and
now
has
many
conflict
with
the
existing
status
of
the
code
base
so
yeah,
and
also
since
the
originals
seems
no
longer
working
on
the
pr.
So
we
decided
to
take
over
the
work
from
him
resolve
the
conflict
test
the
pr
and
made
a
new
one,
yeah.
Of
course,
the
roger
the
original
authorship
is
still
preserved
in
this
pr,
so
also
don't
lose
any
credit.
G
So
this
implementation
follows
the
standard
approach
in
most
kubernetes
controllers,
so
we
use
the
leader
leader
leader
election
package
from
client
goal
pack.
The
client
go
library
which
basically
determine
the
leadership
based
on
the
ownership
of
a
list
object
in
the
cluster.
G
That
means
at
any
time,
the
replica
that
holds
the
lease
will
become
the
primary
primary
controller
and
then
it
can
enter
the
controller's
reconciliation
loop.
In
addition,
the
primary
controller
need
to
refresh
its
leadership
in
some
configurable
time
interval,
say
15
seconds
by
default.
The
the
other
replay
card
cannot
start
the
reconciliation
loop.
They
keep
retrying
acquiring
the
lock
in
configurable
time
interval.
I
think,
by
default.
We
set
five
seconds.
G
Then,
if
the
master
replica
is
done
and
cannot
be
restarted,
that
means
the
master
cannot
renew
the
least
lock.
Then
one
of
the
replica
will
acquire
and
take
over
the
lock
and
the
starter
is
the
controller's
reconciliation,
loop
yeah.
So
in
this
implementation,
although
we
provide
a
command
line
option
to
disable
the
hp
feature,
this
feature
is
enabled
by
default
due
to
two
reasons-
and
I
explained
in
the
comments
in
this
pr
firstly
defaulting
to
defaulting.
To
is
backward
compatible.
G
This
is
because,
when
there
is
only
one
instance,
this
instance
will
be
elected
as
the
primary
node.
So
there's
no
impact
on
the
existing
deployment
manifest.
E
That
was
gonna
be
my
concern,
but
I
think
I
think
that
that's
yeah.
G
I
think
harry
asked
the
same
question
and
I
explained
in
the
pr
good,
okay,
yeah
and
yeah.
H
G
The
last
but
not
least,
is
there
is
an
ongoing
work
which
which
to
add
a
new
health
check
part
to
the
controller,
because
currently
the
argo
in
the
in
the
deployment
manifest
we're
kind
of
using
argo
rollout.
The
metric
part
for
liveness
check
right
with
the
multi.
If
multiple
replicas
running
in
the
cluster,
then
only
the
primary
instance
can
expose
the
the
metrics
say
to
promises,
so
we
need
to
disk
distinguish
between
the
health
checking
part
and
the
metric
exploding
part.
So
I
filed
a
pr
in
the
e
sorry.
E
Oh,
so,
are
you
wait?
Are
you
saying
that
if,
if
you
run
with
multiple
replicas
for
the
controller
and
all
three
are
exposing
metrics,
are
we
tripling
the
counts
collected.
G
Yes,
if
we
don't,
if
we
don't,
we
are
not
traveling.
Actually,
if
the,
if
the
request
comes
to
the
non-primary
non-primary
instance,
then
the
count
basically
zero
right.
It's
not
it's
inaccurate,
oh.
G
G
A
separate
pr-
and
I
haven't
started
working
on
that.
Currently
it's
just
in
progress,
okay,
but
I
I
created
the
issue.
E
Okay-
and
we
are
you're
okay
with
this
thing
and
to
that
sorry,
one
that
1.2.
G
B
For
the
day,
thank
you
for
presenting
it.
I
think
it's
great,
I
think,
going
forward.
I
think
we'd
like
to
have
any
features
coming
from
the
computers
outside.
I
think
we
would
like
to
present
it
here
so
that
we'll
have
a
better
discussions
again
like
what
we
have
today
so
yeah
thanks
yeah.
Thank
you.
E
E
We're
here
is:
are
there
any
roll
out
outstanding
stuff
for
the
1.1
release
that
need
to
go
in,
like
harry?
I
saw
andre
just
tried
to
fix
something
with
the
seo
virtual
servers.
It
sounded
like
a
bug
that
should
be
also
a
part
of
one
that
one.
B
Yes,
it's
a
regression,
I
think
yeah.
So
I
think
extra
alex
made
one
release
cut
too,
with
a
couple
of
bugs,
but
from
the
community.
I
don't
really
see
any
other
birds,
so
we're
hoping
like
some
more
bugs
if
now
yeah,
but
otherwise.
I
think
we
should
be
good,
maybe
and
we
are
not
putting
any
features
in
here
yeah
after
the
release
cut,
so
it
will
only
box
up
regulations.
B
By
the
way,
so
there
I
think
alex
henrik
mentioned
that
they
would
like
to
release
it
in
the
during
the
kubecon
week.
Martial
enrique
is
there
so
I
think
on
october
11th
or
so
so
mostly,
they
would
like
to
announce
ronald's
1.2
1.1
as
a
part
of
that
announcement.
E
Okay,
one
more!
This
I
see
is
config
mentioned
plug-in.
2.0
short
answer
is
there
hasn't
been
much
progress.
We
the
engineer
that
was
working
on
that
had
to
take
a
leave
and
then
they
kind
of
went
in
limbo.
For
since
then,.
F
Started
by,
I
think,
sharma
started
working
on
it
and
may
kind
of
contribute
like
she
took
over
that
pr
and
made
it.
You
know
progress
a
little
further
and
now
she
is
also
busy
doing
some
stuff
to
support
incarcerated
into
it.
So
it's
just
there
is
no.
We
don't
have
an
actively
working
person
on
it
right
now,.
L
Okay,
so
it
sounds
like
maybe
you
could
use
some
help,
then?
Yes,
all
right,
yeah,
maybe
I'll
see.
What
I
can
I
can
do
is
do
you
know
is
the
pr
that
may
put
up.
Is
that
I
guess
it's
not
complete?
I
mean
I
know
it's
got
some
conflicts,
but
is
there
still
more
work
to
be
done
beyond?
What's
in
the
pr,
and
I
I
know
that
yeah.
F
And
then
I
think
we
were
hoping
to
you
know
we
we
knew
that
it
will
take
several
years
to
even
you
know,
do
what
we're
planning
in
you
know
in
the
proposal
yeah,
so
okay,
so
this
is
not
is
not
implementing
everything.
I
think
it's
kind
of
it's
like
a
first
step.
L
Okay,
so
I
guess
who
should
I,
who
should
I
talk
to
if
I
want
to
see
if
I
can
help
out
here.
F
I
think
yes,
I
will
see
everything
up.
You
know
I
think
mei
is
the
good
person
she
can
and
maybe
we
can
ask
may
to
present.
You
know
the
progress
in
next
or
we
can
do
it
offline.
We
can
just
ask
here
and
see
you
know
where
we
are
right
now
and
yeah
and
then
maybe
try
to
wrap
up
the
pr
that
exist
and
then
try
to
split
work
yeah,
but
we
definitely
need
help
like
we
like.
F
L
Okay
yeah,
so
I
guess
I
guess
we
can
discuss
this
further
in
slightly
slack.
The
best
way
to
talk
about
this
yeah,
okay,
cool
sounds
good
thanks.
A
All
right
well,
thank
you
guys.
I
think
that's
all
the
topics
we
had
for
today
so
I'll
go
ahead
and
stop.