►
From YouTube: Argo Contributors Office Hours May 5th 2022
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
All
right,
good
morning,
everyone
welcome
back
to
the
contributor
experience
meeting,
I'm
your
host
for
today.
So
let's
get
started
as
usual,
we'll
start
with
the
triage
discussion
primary
and
secondary
for
this
week
so
alex.
I
think
you
were
primary
this
week
and
jayden
are
you
here.
B
I
don't
have
any
particular
update.
I
noticed
there
was
a
questions
and
I
didn't
even
have
to
answer
because
that
there
was
a
basically
it
was
not
a
question.
It
was
a
discussion
and
someone
else
from
the
community
just
answered
it.
So
there
was
just
three
of
them
and
yeah.
That's
my
only
observation.
A
Okay,
cool
thanks,
alex
yeah,
okay
yeah.
If
that
was
it,
then
I
think
we
can
just
pick
next
week's
primary
and
secondary.
So
do
we
have
any
do
we
have
any
volunteers?
First,.
A
Okay,
that's
all
right!
I
can
volunteer
as
primary
for
this
week.
Does
anybody
want
to
be
secondary,
hey
remington,
it's
keith
yeah,
I
could
be
secondary.
Okay
sounds
good
thanks.
Keith.
A
Great
that
was
easy.
Okay
looks
like
we
just
have
one
other
topic
for
today
alex.
Were
you
the
one
that
added
this?
Yes,
so.
B
That
was
me.
I
know
that
I
think
we've
been
done
with
all
the
blocking
fixes
and
I'm
I
want
to
collect
all
the
pull
requests
that
we
want
to
include
into
release
candidate.
I
think
we
have
a
lot
of
open
uis,
and
this
is
the
one
that
I
know
about.
This
is
to
improve
a
new
feature
that,
and
this
has
to
go
in
2.4
release
candidate
because
it
kind
of
it's
it
would
be
a
breaking
change.
We
better
merge
it
before
we
release
the
feature.
B
So
basically
it's
a
call
for
everyone
who
contributed
and
if,
if
you
at
least
outstanding
players,
maybe
here
or
in
the
dock
or
maybe
just
in
a
slug,
I
would
appreciate
it
and
I
think
we'll
prioritize
merging
them
and
we
can
pretty
much
nothing
stop
us
from
creating
release.
Candidates
like
today.
A
Awesome,
did
you
link
this
one
because
you
wanted.
B
A
A
Okay,
does
anybody
else
have
any
off
the
top
of
their
head
that
they
can
think
of.
C
I
I
just
wanted
to
to
bring
up
the
topic.
I
think
it's
kind
of
related
to
the
to
this
pr
that
michael
pushed,
I
wanted
to
discuss
a
little
bit
about
the
terminal
feature
that
is
going
to
be
in
the
next
release.
C
I
read
that
it's
going
to
be
disabled
by
default,
but
while
deploying
that,
in
into
it
clusters,
we
can
see
the
button
right
it,
it
shows
up
in
the
ui.
C
So
I
guess-
and
I
wanted
to
to
know
your
opinion-
if
it's
if
it
should
be
disabled
by
not
displaying
anything
in
the
ui
right
and
not
having
this
this
visible
anywhere
and
only
if
the
admin
goes
and
enable,
then
things
will.
D
Appear
yep!
That's
in
this
pr
leo.
We
got
to
get
this
merged
so
that
just
that
button
disappears
in
our
local
environment
or
internal
environment.
B
D
So
there's
a
flag
that
lives
in
the
argo
cdcm
config
map
that
disables
it
on
the
back
end
and
then
I
added
a
field
to
the
settings
server
which
I
think
wasn't
initially
intended
for
other
stuff,
but
now
it
just
ends
up
being
stuff.
The
ui
needs
to
know
about
so
the
ui
asks
the
settings
endpoint
is
terminal
enabled,
if
so,
show
the
button,
if
not
don't
show
it
yeah.
B
We
could
have
done
it
for
using
our
bug
as
well,
but
I
I
guess
you
know
this
feature
is
special
and
we
could
use
our
bug.
We
have
api,
I
think
it's
called.
Can
I
api
and
technically
we
can
hide
all
the
buttons.
E
B
Yeah
I
mean,
can
I
I
exist,
but
I
can't
I'm
not
suggesting
to
change
approach.
I
think
this,
you
know
exact
feature
kind
of
special
it
it
can
be.
I
can
see
how
someone
just
don't
want
it
at
all
ever
and
kind
of.
D
E
A
B
Your
oh
my
approach,
I
think
it's
yeah
it's
like
for
later
on.
Basically,
it's
a
big
change
that
we
will
have
to
work
on
in
the
next
release.
I
would
merge.
B
And,
and
maybe
and
just
someday
we
have
to,
I
was,
I
think
I
was
saying-
the
cleanest
solution
is
to
use.
Can
I
api,
but
it's
like
way
more
work
and
maybe
and
web
terminal
deserve
its
own
setting
to
be.
You
know
enable
disabled.
A
Okay,
cool
anybody
else
have
any
other
thoughts
on
this.
A
Right,
it
looks
like
we
have
one
more
item
on
the
agenda
as
long
as
alex
was
that
it
for
that
topic
or
yes,
yes,
that
was
just
my
basically.
B
I'm
looking
for
you
know
if
anyone
has
any
objections
to
create
a
release.
B
I
mean
not
objections.
Basically,
if
you
know
what,
if
you're
aware
of
some
changes
that
have
to
be
made
before
we
create
release
candidate,
please
let
me
know,
but
this
is
what
I
I'm
aware
of.
This
is
the
only
pull
request
that
I'm
watching
right
now.
Yes,
please
bring
me
links
to
other
peers.
A
Cool
thanks
alex
yeah,
so
it
looks
like
we
have
one
more
item
on
the
agenda,
so
it
looks
like
an
issue
about
failing
e2e
tests
due
to
resources
exhaustion
of
the
github
runners.
So
excuse
me,
if
I
pronounce
your
name
wrong,
but
rashab
did
you
want
to
talk
about
it?
Yeah.
F
Yeah,
so
I
added
this
to
highlight
that
sometimes
the
integration
test
failed
due
to
it
couldn't
start
and
reason
could
be
timed
out
or
if
I,
if
I,
if
I
look
into
sometimes
looks
like
that,
github
memory
has
been
full
and
it
cannot
execute.
So
I
was
thinking.
Can
we
refactor
our
end-to-end
test
like
break
that
into
component
wise,
like
application
set?
F
Would
it
would
be
in
another
package
and
would
be
another
package
and
when
the
and
the
profile
would
start
the
process
only
so
it
would
first
start
the
process
and
the
argo
siri
test
would
run.
Then
it
would
also
start
the
other
process
or
build
at
the
separate
time.
So
how
how
that
could
be
done?
I
I
don't
know,
but
this
what
something
came
into
my
mind.
So
should
we
look
into
this.
B
B
So
I
think
I
know
what
causes
the
memory
issue,
because
someone
else
run
into
the
same
problem.
I
think
it
was
either
chetan
or
someone
I
forgot
who,
but
basically
we
use
gorman
to
start
the
components
of
our
argo,
cd
and
gorman
run.
This
uses
the
approach
file
that
I'm
sharing
right
now
and
this
profile
by
default
just
run
a
bunch
of
go
around
commands,
so
it
does
it
in
parallel.
So
basically
it's
building
argo
cd
like
four
times
I
guess
so.
F
B
We
have
a
flag,
called
pin,
mods
and
it
can
be
set
to
true.
So
if
it's
set
to
true
a
rose
file
assumes
that
binary
was
built
previously
and
it's
sitting
in
the
gst
rcd
directory,
and
I
think
it's
doing
it
we
test.
So
basically,
I
would
just
try
to
set
this
binary
to
true
in
e3
tests
and
you
know,
send
the
pull
request,
but
not
sure
how
to
basically
in
this
case
it
will
not
be
building
anything.
It
will
just
start
existing
binary
and
yeah.
F
B
I'm
assuming
we
don't
use
it
because
I
know
that
it
was
introduced.
You
know
for
internal
development
needs.
I
think
who
introduced
it
because
he
wanted
to
use
golden
debugger,
but
it
is
also
it's
kind
of
a
performance
optimization.
C
F
If
you
see
the
error,
it's
like
execution
paused
so.
H
B
G
I
have
a
couple
of
background
information
to
that.
Probably
so
because
I
I
I
noticed
that
as
well
on
my
local
machine
on
a
vm
with
8
gig
ram
right,
so
it
basically
yeah
had
my
vm
end
up
in
a
in
a
lock.
So
that's
why
we
decided
to
build
the
you
know
the
make
controller
make
server
make
wrapper
server
before
so
that
at
least
the
compilation
and
linking
will
be
done
sequentially.
G
B
G
G
G
That
will
that
will
pretty
much
eat
up
the
resources.
C
And
I'm
not
sure
if
line
424
is
going
to
be
really
evaluated
correctly
during
the
the
make
target
execution.
Maybe
we
should
put
this
inside
the
the
start
e
to
e
local,
like
we
already
have
a
lot
of
a
lot
of
variables
being
been
defined,
big
target,
good
yeah.
I.
C
B
B
A
Thank
you
all
right,
so
it
looks
like
ravi
added
a
few
pull
requests
so
ravi.
Would
you
like
me
to
share
that.
H
So
much
so
last
week,
alex
was
suggested
for
this
notification
engine
integration
with
ocular,
where
we
were
getting
this
object
and
missing
error.
You
asked
me
to
try
another
approach
where
we
could
introduce
that
the
fields
you
know
we
are
updating
the
runtime
object,
with
the
object
with
the
api
version
and
find
right.
H
Yeah,
so
this
is
what
I
was
doing
in
the
pr.
What
you
asked
me
is,
instead
of
doing
it
in
the
update
method,
you
asked
me
to
try
it
out
to
introduce
that
in
the
notification
controller,
wherever
it's
getting
triggered,
because
we
expected
that
here
the
object
kind
would
be
missing
when
it
comes
and
when
it
translates
down
structure
as
well
as
in
the
recorded,
go
in
the
send
notifications.
If
we
could
object
the
update
object
on
time,
it
would
resolve.
H
This
is
what
I've
tried,
and
you
know
I've
just
run
this
even
with
that.
I'm
seeing
this
issue
right.
Even
with
that,
you
know
I'm
running
into
the
object
frame,
missing
error
and
I
can
show
that
I've
just
put
a
log
message
in
the
controller.go
just
to
display
what
this
result.
B
Please
don't
yeah,
I
know
that
line
218
would
fail
if
you
don't
have
a
resource
kind.
Basically,
this
that
the
kind
must
be
inserted
before
line
2.
Oh
sorry,
2
210.
Basically,
the
unstructured
has
its
own
on
marshaller
and
this
guy
fails.
If
you
don't
have
a
google
can
let
me
let
me
just
give
you
one
second,
to
kind
of
confirm
it.
So.
H
The
the
you
know,
I've
I
put
it
in
the
logs
and
every
time
the
logs,
it
is
giving
a
version
and
kind
for
that
result
field.
I'm
not
seeing
an
issue
as
such.
Here
you
know
alex
so
here
it
is
doing
fine,
but
when
it
is
trying
to
you
know,
even
when
we
update
the
send
notification
object
runtime,
even
then
we
are
seeing
this
issue
so
so.
B
I
kind
of
I
you
know
when
it's
I
wanted
to
kind
of
double
check
myself.
I
think
I
just
did.
I
know
that
if
you're
trying
to
unmarshal
a
json
object
into
unstructured,
it
has
its
own
custom
and
muscler,
and
then
I
I'm
seeing
this
code
right
now,
so
it
tries
to
deserialize.
You
know
the
data
from
the
json
and
then
it's
specifically
looking
for
group
kind
and
it
dies
and
it
returns
an
error
if
it's
missing.
B
H
No,
it's
not.
That
reason
I
I
think
basically,
what
is
happening
is
this
is
the
first
time
when
the
object
is
being
called.
You
know
during
the
notification
engine
calls
the
the
the
result.
Object
is
fine,
but
you
know
as
we
as
it
processes
the
analysis
stuff
when
the
rollout
object
gets
updated.
That's
when
that
api
version
is
getting
deleted,
aversion
and
kind
are
getting
deleted
when
that
update
call
is
made.
H
So
what
I
was
thinking
is
wherever
we
make
this
update,
call
and
I
tested
by
enabling
this
object
kind.
You
know
setting
the
groupers
and
kind
in
the
update
next
to
the
update
calls,
and
I
don't
see
any
errors.
H
H
B
Then
then
you
know,
hopefully
we
will
not
lose
this
behavior
again
like
in
the
future
yeah.
That
would
I
mean
to
benchy.
I
think
that
would
work
as
long
as
you
don't
have
to
duplicate
the
same
thing
in
multiple
places.
F
C
Basically,
you're
saying
that
when
we,
when
you
invoke
update
status
in
this
api,
call,
the
return
update,
rollout
doesn't
have
the
gvk
correctly
populated
in
the
into.
H
H
I
created
an
issue
in
the
open
source
with
the
code
generate
and
congenital
issue.
I've
created
it.
I
haven't
seen
any
response
between
luckily
deals.
I
wanted
to
understand
why
they
are
deleting
it
if,
as
part
of
this
code
generation,
because
this
comes
basically
from
the
code
generation-
the
entire
client
setup
that
this
update
call
that
we
see
here.
This
is
all
generated
by
the
kubernetes
code
generation
tools.
H
We
don't
write
these
things,
so
it
is
in
the
upstream
from
which
this
error
is
kind
of
popping
up.
B
H
Essentially,
you
know
as
it
updates
this
object,
it
doesn't
populate
the
result.
Object
with
you,
know,
object
and
kind,
and
we
cannot
edit
anything
here.
So
so
that's
a
that's
a
problem.
H
No,
this
is
a
problem
in
the
the
client
generation
code,
not
in
our
club,
not
sure
yeah.
I
would
you
can
take
a
look
at
it.
B
As
well,
you
know.
B
That,
if,
if
you
created
the
issue
against
you
know
the
code
generator,
they
might
not
respond
because
they
might
think
it's
not
their
problem.
Basically,
that
you
know
the
code
generator
just
uses,
kubernetes
client
and
maybe
they're
expecting
that
kubernetes
client
itself
should
populate
group
kind.
So
yeah,
that's
just
a
suggestion.
So
this
is
an
input
generator
and
then
so
maybe
it
makes
sense
to
create
it
in
in
kubernetes
itself.
B
H
That
was
a
good
feedback
actually
after
that,
and
we
thought
that
maybe
put
it
into
a
comment
function.
So
thank
you
thanks
for
suggestion
and
leo,
if
you
you
know,
if
you
think,
if
you
want
to
look
into
it
and
see
if
there
is
anything
else
that
we
could
do
open
for
it
and
we
can
chat
about
it.
The
other
thing
I
wanted
to
check
is
on
the
post
promotion
analysis
for
canary
strategy.
H
So
today
you
know
there
are
a
couple
of
issues
identical
to
this
one
is
this
one,
and
there
is
another
issue
that
hari
created
on
running
analysis
independently
after
the
post.
You
know,
after
the
promotion
is
completed
right,
whether
it
could
be
january
or
blue,
green
or
experimentation.
We
could
run
this
so
so
so
there
are.
H
There
are
some
similarities
between
this
issue
and
this
ratio,
essentially
for
canary
once
we
start
the
analysis
and
the
post
promotion,
the
you
know
there
is
no
way
we
can
skip
the
reconciliation
for
that,
because
all
the
objects
have
been
promoted
and
at
the
end
of
the
canary
promotion,
the
the
current
pod
hash
is
equal
to
the
stable
replica
asset,
whereas
when
we
compare
this
logic
with
blue
green,
you
know
you
know
the
the
the
logic
there
is
different,
because
once
once
this
post
promotion
thing
happens,
you
know,
then
those
two
fields
are
not
yet
same
once
they
are
same.
H
We
can
stop
it,
but
in
this
case,
because
it
I
mean
it
is
inherently
doing
it.
H
The
the
thing
that
could
differ
from
this
issue
with
post
promotion
analysis
is,
if,
in
case,
the
in
the
canary
strategy,
the
post
formation
analysis
failed
after
say
five
minutes
of
the
enter
canary
getting
populated
then
do
we
want
to
revert
back
to
the
previous
one.
I
just
wanted
to
get
some
opinion
on
this.
H
So
for
post
function,
analysis,
let's
say
after
the
canary
got
updated.
If
we
continue
to
do
the
post
promotion
analysis
and
in
case
there
are
failures,
should
we
just
update
the
rollout
object
with
a
degraded
state,
or
should
we
revert
back
to
the
previous
best
known
state,
like
the
previous
replicas.
H
So
I
I
briefly
discussed
this
with
hari.
What
hari's
suggestion
is:
let's
just
mark
roll
out
status
as
degraded
and
send
the
notification
which,
which
happens
and
continue
the
analysis.
That
is
his
opinion
I
just
want
before
we
go
for
any
implementation.
I
just
wanted
to
get
if
everyone
is
aligned
on
it
or
you
know,
if
you
want
to
get
some
comments
on
this.
A
Great
was:
are
you
done
with
those
topics.
H
A
Okay,
great
thank
you.
It
looks
like
that
was
the
last
of
our
agenda.
Does
anybody
have
any
last-minute
topics.
E
I'll
just
mention
that
I
just
put
just
in
case
you
missed
it.
I
posted
a
link
to
that
kind
thing
for
some
upstream
bugs
around
it
that
you
might
want
to
look
into.
So
that's
all.
B
E
A
All
right:
well,
if
there
are
no
other
topics,
I
think
we
can
go
ahead
and
end
a
few
minutes
early.
So
thank
you
guys
for
attending.
See
you
next
week.
Thank.
I
You
one
other
thing
to
mention:
if
any
of
you
are
going
to
be
at
argo
contributor
summit
in
valencia
and
you
haven't
registered,
please
do
so.
I
think
the
link
is
in
the
meeting
notes
and
also,
if
you're
interested
in
presenting
a
topic.
Please
message
me
on
slack,
because
we
do
have
some
open
time
slots
on
there
so
I'll
make
sure
that
the
link
is
added
to
the
document.
If
it's
not
already
on
there,
it
should
be
there
from
from
last
week
yeah
it
is
so
I'll.
Add
it
again
there
you
go.