►
From YouTube: Argo Contributors Office Hours Apr 21st 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,
I
hope
everybody
can
see
my
screen.
Yes,
I'm
going
to
be
your
host
today
and
starting
off
with
the
triage
and
discussion.
A
As
you
can
see
here,
dan
and
john
was
they
were
for
the
assigned
for
triage
last
week.
So
john
and
then
are
you.
B
Online
yeah
I'm
here
I
I
didn't
actually
get
a
chance
to
do
much,
so
I
should
probably
continue
to
be
secondary
for
this
week.
C
Thanks
john
and
now
I'll
volunteer
for
primary
okay,
thanks
jan
I
was
well.
I
think
I
was
on
turn
last
week,
but
I
had
videos
so
I'll.
Take
this
this
week,
yeah.
A
I
remember
you
said
you
were
getting
some
ptos
right
cool
all
right,
so
I
guess
we
sorted
out
the
triage,
so
we
have
for
now
two
topics.
If
you
have
any
topic
you
want
to
be
discussing
today,
please
add
to
the
to
the
list
here
in
this
document,
so
starting
off
with
the
first
topic
of
the
day.
This
is
from
michael
weird,
repo
server
errors
for
helm
in
get
repo
with
dependencies.
Michael,
you
want
to
elaborate
a
little
bit
about
this.
D
Yeah
I've
got
to
drop
off
in
a
couple
minutes,
but
it
was
just
a
really
strange
bug
where
someone
was
using
like
a
customized.
Actually,
I
think
a
helm
umbrella
chart
and
they
had
several
dependencies
as
like
external
helm,
repos,
they
were
getting
an
error
from
the
helm.
Repo
ad
commands,
which
was
like
hey
this
repo's,
already
been
added
and
manifest
generation
was
failing
because
of
it.
So
no
idea
why
helm
repo
had
would
throw
that
error.
D
What's
even
weirder
is
that
when
they
went
and
added
those
repositories
directly
like
created
repo
secrets
for
them
individually,
the
errors
stopped
so
like
something
got
cached
in
a
way
that
let
the
home
repo
ad
command
succeed
anyway,
really
weird
air.
They
have
a
work
around
now,
which
is
nice,
but
I'd
love
if
other
folks,
familiar
with
manifest
generation,
would
take
a
look.
A
D
D
E
I
think
I've,
maybe
I'm
hoping
that
bug
is
fixed.
I
need
to
build
one
more
time,
but
I
know
that
there
was
a
issue,
so
we
we
were
supposed
to
supply
some
name
when
we
adding
repository
and
then
the
first
implementation
was
to
just
use
a
hostname
as
a
name
and
name
doesn't
really
matter.
Name
has
to
be
something
just
because
home
wants
that
name
and
then
and
hostname
was
a
bad
decision,
because
you
can
have
two
repositories
on
the
same
hostname
and
the
error.
E
Then
there
happens
and
in
the
most
recent
version,
we've,
okay
and
basically,
what
else
important
is
we
use
hostname
as
a
name
only
if,
unless
users
specify
some
some
meaningful
name
for
for
the
repository
and
what
the
suspect
happens
here
is
that
maybe
they
added
each
repo
individually
and
they
provided
some
meaningful
name,
basically
like
some,
some
alias
for
each
repo
and
that
fixed
the
problem.
But
and
in
the
most
recent
version
of
argo
cd
we
no
longer
use
hostname.
D
So
they
said
that
they're
running
233-
I
might
be
misremembering
that,
but
so
it
made
me
think
that
the
original
fix
hadn't
solved
it
that
what's
the
source
of
truth
for
the
user
specified
repo
name
is
that
the.
E
E
This
is
that,
basically,
that
name
is
to
be
able
to
kind
of
register
a
repo
as
with
a
visit,
rs
yeah,
and
then
that's
one
use
case
and
second
use
case.
Sometimes
like
user
can
reference
a
repo
without
any
areas,
but
and
that
works,
but
we
still
have
to
specify
name
for
you
know
for
repo
ad
command
and
that's
why
we
had
to
choose
something
like,
like
course,
name.
D
E
D
A
All
right
can
we
move
on
yep
cool
thanks
thanks,
michael
thanks
alex.
So
next
topic
is
something
I
wanted
to
to
bring
up
something
we
were
discussing
yesterday
into
it.
Basically,
looking
at
the
release
and
just
bringing
the
the
topic
to
discuss
the
release
process
related
to
argo,
cd
and
argo
rollouts.
A
So,
for
example,
if
we
look
at
the
current
plan
for
the
milestone
for
2.4
the
due
it's
due
to
may
1st
right
and
we
we
have
four
issues
closed
at
this
point,
so
I
guess
one
of
the
one
of
the
things
that
we
discussed
is
the
possibility
to
just
respect
what
the
what
the
milestone
plan
original
plan
was,
and
at
at
this
date,
whatever
is
in
master,
we
create
a
a
release,
candidate
branch
and
release
from
that
that
would
cause
us
to
have
more
constant
releases
instead
of
having
sparse
releases
like
what
we
have
today-
and
I
just
wanted
to
bring
up
this
topic
for
discussion
to
see
what
what
you
go.
C
I
think
we
discussed
this
in
the
past,
probably
yeah.
I
I
think
we
came
to
the
conclusion
that
that
what
you
just
described
is
the
way
we
we
want
to
go
forward.
Okay,
so
like
yeah,
I
guess
you
said
that
I
mean
may
1.
It
may
also
be
may
3rd
or
may
5th,
but
just
like
around
the
date,
we
we
should
push
out
the
release
to.
E
C
Yeah
I
mean
I
mean
a
short
delay
is,
is
not
something
that
that's
mission
critical
right,
so
the
the
due
date
for
the
milestone
should
basically
act
as
like
yeah
as
a
predictor
right,
so
that
people
know
okay
around
may
1
a
new
version
will
be
released.
I
mean
that's
also
for
for
like
the
vendors
right
who
have
products
based
on
our
cd,
like
red
hat
or
like
code
fresh,
so
that
they
can
adapt
their
release
cadence
accordingly,.
C
And
also,
I
guess
it's
pretty
important
for
the
community
to
to
have
like
at
least
a
class
r
estimation
of
when
to
expect
a
new
release,
because
that's
that's
also
what
I'm
always
hearing
from
people
like
a
pr
got
merged
to
master,
right
and
and
the
first
question
people
are
asking.
So
when
will
it
be
released?
When
will
we
have
a
release.
B
C
That
fixed
merged
so
and
yeah,
so
that
my
my
opinion
is
that
we
should
stay
with
with
that
release,
cadence
that
we
say:
okay
may
one
plus
minus
a
couple
days.
Maybe
if
it's,
if
there
is
a
real
hard
bug
that
we
have
to
fix
right,
then
maybe
we
can
postpone
by
two
weeks
or
something
like
that
right.
A
Yeah,
basically,
I
was
looking
at
this
list
yesterday
and
I
I
got
maybe
the
first
time
I
I
see
this
milestone,
so
it's
we're
getting
close
to
the
day
and
we
are
eighteen,
eighteen
percent
complete,
I'm
not
sure.
If
the
pre
in
the
previous
review
previous
releases,
we
were
at
the
similar
pace,
but
if
that's
okay-
and
we
all
agree
with
that-
we
can
just
cut
master
by
around
this
date
and
release
whatever
is
in
there
and
maybe
move
those.
A
E
A
So
it's
it's
a
kind
of
normal
procedure
to
just
take
those
those
those
tasks
and
move
to
the
next
one,
yeah.
Okay,
all
right
and
the
same
applies
for
rollout.
I
think
rollout
was
a
little
bit
more
delayed
last
time.
E
All
right
cool
yeah
rollout.
Actually
I
don't
think
we've
committed
to
have
release
for
loud
every
three
months.
Just
because
it
is
you
know
it's
smaller,
it
doesn't
have
ui
like
it
has
ui,
but
it's
just
different
type
of
project.
Like
you,
a
new
deployment
strategy
doesn't
get
invented
every
three
months,
so
it
it
has
less
frequent
releases.
E
E
Doesn't
really
follow
two
months
it
basically
release
happens
on
demand
like
as
soon
as
critical
number
of
changes
accumulated.
A
A
All
right,
so
I
guess
that
wraps
up
the
topic
I
wanted
to
discuss
thanks
thanks
everybody.
Next
one.
A
F
So
go
for
a
question,
hi
everyone.
So
last
time
I
worked
on
like
after
conversation.
I
understand
that
we
have
problem
with
performance
testing
and
sometimes
we
meet
problems
here.
So
I
started
to
work.
I
don't
have
good
presentation,
but
actually
I
will
just
try
to
describe
and
feel
free
to
ask
questions.
So
could
I
share?
F
Yes
go
ahead,
so
so
what
I
did?
I
actually
cr
thanks
for
coders
for
bring
to
me
cluster.
I
can
share
this
cluster
with
everyone
who
wants
to
have
some
access.
I
think
we
need
to
say
the
site
level
permissions.
It
can
be
maintainers,
but
it's
actually
cluster
for
community
that
we
can
use
for
everything
that
we
want.
And
now
I
am
running
here,
performance
testing,
and
for
that
I
have
created
three
components.
F
F
F
I
provision
here
100
applications
and
also
I
have
here
10
10
destination
clusters
around
10.
I
don't
remember
exact
number
and
what
I
am
doing.
I
have
application
which
named
simulator
and
this
simulator
just
doing
a
random
activity
for
call
reconciliation
inside
cluster,
and
so
each
few
minutes
I
am
like
delete
service
change,
port
number
change,
replicas,
number
and
so
on
and
in
such
case
controllers
start
to
trigger
changes
and
to
do
reconciliation,
and
also
I
have
another
component
that
restart
cd
every
morning.
Why
I'm
doing
that
for
applying
new
version
of
argosy
different
master?
F
You
can
see
reconciliation
time,
time
of
reconciliation
and
in
case,
if
it
will
be
grow
up,
we
will
see
immediately
and
we
can
react
on
that.
Somehow
we
can
check
when
changes
happens
and
and
so
on.
Actually,
my
future
plans
to
connect
I'll
just
manage
around
that,
but
it's
something
that
I'm
not
complete.
Yet
I
think
it
will
be
done
next
week.
E
Yeah,
it's
awesome,
so
it
looks
like
you
know
the
real
production
or
qcd
that
continuously
doing
something.
F
Yes,
actually,
yes,
we
can
improve
it,
but
for
now
I'm
just
changing.
I
just
remove
some
components
from
application
like
service,
I
s
originally.
I
did
replicas
count
change,
but
I
stopped
to
do
this
because
its
provision,
real
application
and
cluster
just
stopped
to
work,
because
if
I
will
run
100,
helm
gas
guest
books
inside
cluster,
it
will
just
consume
a
lot
of
resources.
E
Four
seconds
from
one
to
four
seconds:
actually
there
is
something
I
cannot
explain.
I
don't
know
that
lower
graph
qphs
api
activity,
it's
supposed
to
be
zero
all
the
time.
So
something
is
this
is
happening.
This
panel
here
is
to
catch
a
problem.
Basically,
if
you
see
this
constant
api
calls,
that
means
I
mean
it's
not
a
problem.
It's
just
argo
cd.
E
Okay,
I
was
like,
but
that's
the
I
mean
it's,
not
a
blocker.
I
guess
it's
it's
still.
It
would
catch
the
performance.
E
F
B
E
E
I'm
just
going
to
suggest
something
to
increase
visibility
because,
and
then
next
question-
I
guess
it's
to
leo
and
michael
like
because
we
have
a
kind
of
open
source
demos
domain.
It's
managed
in
basically
intuit
manages
this
domain
name.
How
do
you
think?
Is
it
possible
to
point
let's
say
performance,
dashboard,
ocd
url
to
sorry
I
mean
this
auto
generated
name
is
like
impossible
to
remember,
and
so
is
it
possible
to
point
performance.therefore
cd
to
that
url
in
the
hosted
zone?
E
That's
intuitive
images
made
I'm
talking
about
basic,
so
we
have
this
url
already
this
hostname
available
and
I'm
I
just
put
it
in
chat,
it's
easier
to
to
type
than
talk
and
then
what
I'm
suggesting
is
to
introduce
one
more
cd
performance.
A
E
A
So
basically,
what
you're
asking
is
for
us
to
register
an
additional
sub
domain
or.
E
And
I
I'm
not
for
explaining
this,
I
think
for
feedback
as
well.
I
I
feel,
like
I
mean
I'm
suggesting
it
just
so
that
we
care
about
that
instance,
because
if
we
don't
you
know
it
might
be
that
we're
just
going
to
forget
that
url
and
no
one
ever
going
to
open
it
again
if
we
have
cd
performance,
apps
kind
of
hostname,
and
maybe
it's
mentioned
in
the
readme
file-
and
we
explained
that
yes,
this
is
we
use
it
for
testing.
A
F
C
On
that
note,
I
was
wondering
so
whether
we
should
maybe
ask
the
cncf
for
like.
So
what,
from
what
I
understood
is
that
the
the
argo
approach
io
domain
is
still
managed
by
intuit.
C
And
maybe
so
the
the
idea
is
maybe
so.
My
idea
is
maybe
to
ask
the
cncf
to
take
this
over,
because
I
mean
we
donated
the
project
to
the
cncf
right.
As
far
as
I
understood,
and
maybe
they
can
also
sponsor
the
resources
that,
from
from
what
is
behind
there
right,
so
that
that
it's
not
going
on
on
on
into
its
bill
on
the
I
I
guess,
the
gcp
bill
and
the
the
aws
instance
for
the
performance
testing.
E
I
I
know
that,
since
you
have
already,
they
already
gave
us
hardware
I
mean,
but
basically
they
give
they
can.
We
can
get
servers,
but
the
problem
is,
it's
like
we
would
have
to
maintain
kubernetes
ourselves,
but
that's
sorry,
it's
it's
kind
of
an
option
too.
E
E
Okay,
no,
I
take
it
back.
Precious
cluster
won't
like
we
want
to
test
performance.
It's
not
okay
to
test
performance
on
one
k3s,
tiny,
cluster,
yeah,
okay,.
F
E
What
I
was
describing
like
for
demo
instance,
kind
of
like
a
real
kubernetes,
is
not
like
a
requirement.
I
would
just
maybe
have
a
single
machine
and
then
a
single
kts
cluster
on
that
machine
that
uses
all
the
resources
and
it
would
be
as
good
as
current
jc.
You
know
kubernetes
cluster
from
from
google,
but
for
performance
testing
it
wouldn't
work.
I
agree
it's
like
for
performance
it
just
it
won't
test
anything.
E
G
G
C
I
mean
it's
just
like
you
know
if
just
an
imagination
like
like
a
company,
pulls
out
right.
Let's
so
just
let's
assume
code
fresh,
it's
going
to
pull
out
someday
and
and
then
this
will
all
be
gone
right,
probably
or
if
like
intuit,
pulls
out
one
day.
So
then
argo
approach,
io,
would
be
like
you
know
what
what
happened
with
that
like
with
the
domain
right
so.
A
I
guess
we
are
were
from
from
my
understanding
what
to
what
alex
explained.
I
I
I
kind
of
understand
it
that
that
we
have
two
different
two
different
scenarios.
One
is
cd
apps
argo
proj
that
can
be
deployed
in
a
in
a
tiny
cluster,
so
we
can
use
some
cluster
provided
by
cncf
and
easily
deploy
a
tiny
cluster
there
and
have
the
demo
cluster
available.
A
E
Was
trying
to
say
that
sorry,
but
yeah
cncf
gave
us
just
vms
right
now.
Basically,
we
can
create
a
vm
without
asking
them.
I
think
jan
has
access
to
the
project
and
ian
and
jc
it's
like
since
manager
success.
We
just
need
to
tell
them,
please
add
one
more
person,
but
we
already
can
create
those
vms,
and
that's
why
you
know
we
could
set
up
the
demo
cluster
today
without
even
asking
cncf
because
they
already
approved
it.
E
But
you
know-
and
then
I
was
thinking
of
to
to
for
performance
testing.
I
think
we
need
managed
kubernetes
cluster,
because
we
don't
want
to
manage
cluster
ourselves,
and
for
that
we
will
have
to
ask
cncf.
So
we
can
still
ask
them.
We
can
tell
them
you
know,
can
you
sponsor?
Can
you
pay?
Can
you
take
over
the
bill
for
gcp
accounts
yeah,
and
in
this
case
it
would
work
like
we
will
not
spend
any
it
basically
for
us,
it's
no
different,
just
a
different
organization
and
different
manage
access.
A
Okay
and
and
from
your
demo
pasha,
it
seems
that
this
performance
cluster
is
configured
in
aws
under
code.
I
would
assume.
F
F
A
It's
under
me,
I
I
have
the
responsibility
to
fix
it.
E
Cool
yeah,
that's
the
kind
of
inconvenience
I
hope,
I'm
not
revealing
anything
in
general,
but
basically
into
it.
It
applies
the
same
rules
to
all.
Google
account.
Google
cloud
accounts
and
that's
why
I
know
that
it's
it
was
hard
to
get
access
and
now,
like
leo,
will
have
to
go
for
the
same
exercise
to
also
get
access
to
that
account,
and
I
know
that
intuit,
it's
just
technically,
I
guess
impossible-
to
share
access
with
anyone
outside
of
it
to
do
it.
So
that's
yeah,
even
cluster
itself.
G
E
A
One
one
last
question
regarding
the
performance
tool
framework:
pasha:
do
we
have
any
piece
of
documentation
somewhere?
If
let's
say
I
want
to
give
it
a
try?
No.
F
F
F
A
Think
that
wraps
up
the
topics
for
the
day
anybody
else
have
any
last
minute
topic.
You
want
to
discuss.