►
From YouTube: Argo Contributors Office Hours 21rd Oct 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).
B
C
C
D
Yeah
yeah:
well,
maybe
here
we
can
start,
and
I
because
that
I
wanted
to
introduce,
who
who
yeah?
I
let
you
introduce
yourself.
Look
at
this,
sir.
B
Hi
very
good
morning,
all
I
am
srinivas
working
at
harness.
B
As
a
distinguished
engineer,
I
have
been
working
with
harness
since
it's
founded
and
basically
like
I
designed
architected
our
harness
continuous
deployment
product
and
right
now,
I'm
actually
researching
on
the
our
go
and
we
were
actively
looking
into
how
how
best
we
can
integrate
with
our
go
with
ordnance
to
to
get
the
to
contribute
to
the
community.
D
F
F
G
D
D
I
guess
not.
Okay,
terry
is
not
in
the
meeting,
but
he
is
a
long
time
user
and
contributor
to
argo
workflows
and
he
became
an
approver
and
pasha
is
a
less
long
time
contributor
to
cd
but
super
active
contributor
and
he
became
a
viewer
and
that's
super
happy
to
have
you
right.
I
think
we
should
start
with
agenda
I'll
just
share
my
screen,
and
so
I
did
the
first
one.
You
know
like
five
minutes
ago,
so
I
guess
yeah.
D
We
just
don't
want
to
forget
to
get
feedback
for
you
know
triaging
and
discussions
if
if
there
is
any
feedback-
and
we
need
to
find
who
is
going
to
do
it
next
next
week
right,
I
guess
okay
jan,
could
not
join
us
today
and
pasha
yeah.
I
am
here.
I
C
I
D
I
know
that
we
actually
we
have
a
pull
request
open
for
a
quite
long
time
to
support
what
you
just
described
bulk.
Basically,
the
pr
introduced
bulk
operations.
It
maybe
already
has
bulk
deletions.
If
you
share
the
link
with
me,
I
will
just
you
know
at
least
link
it
yeah.
It's
very
sure
I
will
send
you
okay,
so
you
know
we
need
a
volunteer
for
next
next
two
weeks.
So
I'm
thinking
to
do
that.
Anyone
wants
to
you
know.
D
Actually,
no,
actually,
I
think
we
agreed
to
always
have
like
one
primary
who
has
more
experience
and
I
believe
daniel,
I
think,
contributes
to
events,
and
maybe
he
has
less
experience
and
sure
is
daniel
in
the
meeting.
So
yes,
we,
I
think
we
definitely
need
second
person
who
is
in
the
meeting
and
experienced
argo.
F
I
can
do
it,
I
think
I
yeah.
D
D
D
One
is
suggested
by
chandler
like
there
is
a
feature
and
I'm
not
sure
if
he
is
here.
Let
me.
D
What
is
the
well
discussed
in
the
ticket
looks
like
if
I,
if
I
understand
it
right,
there
is
a
okay.
I
I'm
assuming
chandler.
You
know
created
that
issue
that
introduces
support.
For
you
know:
health
checks
for
additional
customer
resource
and
pull
request
and
okay,
the
pull
request
was
created
like
on
september,
so
it's
been
a
month
and
we
have
oh,
there
was
a
review.
D
Okay.
I
guess
the
goal
was
to
get
attention
to
the
pull
request
and
you
know
I
think
he
the
goal
is
achieved.
We
need
to
finish
reviewing
and
assuming
it
has
already
a
long
threat
of
discussions
most
likely
that
care
is
in
a
in
a
good
shape
and
yeah.
So
I
guess
it.
You
know
now
it's
time
for
someone
who
has
approved
button
to
to
review
and
finally
approve
it
has
to
be.
Usually,
I
guess
either
me
or
jc.
You.
H
Can
you
can
give
it
to
me
yeah,
that's
the
I
just
it's
only
a
health
check,
it's
so
it's
only
it's
low
risk
in
terms
of
it
won't
affect.
H
D
D
Let's
move
on
so
next.
Actually
it
was
me
so
I
I
just
wanted
to
go
through
one
more
pull
request
that
introduces
changes
in
in
the
roadmap
and
yeah.
I
guess
the
goal
like
we
did
a
couple
of
weeks
ago,
actually
a
couple
weeks
before
coupon,
so
the
idea
is
to
add
roadmap
items
that
you
know
feature
some
enhancements
that
we
already
kind
of
like
and
we
talked
about,
but
I
didn't
want
to
just
do
it
by
myself
and
you
know
don't
ask
anyone.
D
D
Okay,
so
here's
my
list,
one
is
very
simple:
okay
and,
and
one
more
kind
of
disclaimer
is
that
I
pick
them
based
on
popularity.
A
lot
of
people
just
keep
asking
for
for
the
future,
and
so
what
I
did
I
just
sold
by
number
of
thumbs
ups,
and
then
I
check
you
know
the
most
straightforward
ones
and
put
them
into
roadmap.
So
here
is
the
first
one.
So
the
feature
is
about
adding
ability
to
ssh
into
into
a
pod
from
argo
cdui.
D
I
know
that
feature
is
straightforward
to
implement.
It
requires
a
little
bit
discussion,
mostly
about
rbac,
like
we
don't
want
to
just
introduce
this
feature
and
let
everyone
with
access
to
ssh
into
a
port,
but
in
terms
of
implementation
we
definitely
know
how
to
implement
it
and
it
has
like
24
from
comes
up.
So
this
is
a
popular
feature
and
I
guess
you
know
any
thoughts,
any
objections
to
to
have
it
in
the
roadmap.
H
There's
so
I
I
I
implemented
a
shot
actually
a
long
time
ago
to
argo
v1,
if
you
remember,
there's
one
problem
with
shelling
into
pods.
If
is
that,
if
you
don't
actually
terminate
the
shell,
you
actually
end
up
with
lots
of
orphaned
shells
into
the
inside
the
pod
left
running
inside
the
pod.
H
So
I
forget
what
we
had
to
do
to
mitigate
that,
but
it's
it's
not
totally
straightforward.
That's
just
keep
so
exactly
like
there's
some
cleanup
that
actually
has
to
happen.
I'm
gonna.
D
Have
some
kind
of
liveness
you
know
some
some
like
like?
Maybe
you
are
supposed
to
keep
sending
requests
to
kind
of
prolong
the
life
of
this
session.
H
A
D
D
I
think
we
still
want
to
leverage
all
the
features
we
have
as
much
as
we
can,
and
this
is
like
it
seems
to
be
cheap,
like
we
don't
need
to
introduce
any
fields
in
the
cache
like
we
don't
even
have
to
change
controller,
it's
just
a
instant
usefulness
for
for
users
with
a
relatively
small
effort,
all
right.
So
the
next
feature
is,
I
just
want.
Okay
next
feature:
it's
like
jc.
D
He
opened
that
feature
request
recently,
it's
not
the
most
popular
one,
but
it
resolves
the
most
popular
request
to
support
combining
help,
chart
and
values
file
in
a
git
repository,
and
I
feel
like
it's
kind
of,
we
talked
a
little
about
it,
and
maybe
I
want
to
talk
a
little
bit
about
it
again.
So,
basically,
the
request
that
we,
the
most
popular
request
we
have-
is
to
be
able
to
combine
off-the-shelf
health
chart
and
value
files
in
the
separate
git
repository.
D
We
have
open
pull
request
and
this,
like,
I
think,
it's
like
the
most
popular
pr
and
most
popular
feature
request
in
argo
cd
ever,
but
the
currently
implemented,
pull
request
has
some
caveats.
It
just
won't
work,
I
mean
it
will
work,
but
it
has
problem
with
caching.
Basically,
argo
cd
won't
notice
changes
in
the
value
file,
plus
the
amount
of
work
to
make
it.
D
You
know.
100
buck-free
is
the
same
as
kind
of
support,
support,
multiple
sources
for
the
whole
application,
and
if
we
support
multiple
sources
for
an
application,
then
you
know
it's
kind
of.
It
covers
the
helm,
plus
video
files
use
case,
and
so
we
talked
about
it
with
jc,
and
I
think
we
spoke
about
it
in
one
of
the
previous
meetings.
When
we
were
discussing
the
different
plus
value
files,
they
are
yeah
and
I
felt
like
it's
important
to
edit
to
a
roadmap.
This
will
require
some
designing
up
front.
H
So
on
the
on
that
topic,
so
the
the
the
issue
where
we
support
helm
values
in
in
a
git
repo.
H
D
H
That
that
was,
I
think,
with
the
reason
we
were
so
hesitant
about
supporting
this
and
why
we
closed
this
original
issue
and
then
now
we're
reopening
it,
because
we
do
want
to
do
that
for
help
helm
values
and
so
in
the
ui
we're
going
to
have
to
present
yeah
you're
in
sync
with
you
know,
helm
repo,
but
out
of
sync.
D
Yeah
with
think
so
I
think-
and
maybe
you
know
maybe
I
was
wrong
about
a
lot
of
changes
in
in
deep
server,
because
when
you
described
it
this
way,
I
realized
that
we
can
just
keep
people
server
like
untouched
and
simply
make
changes
in
controller
so
that
it
would
request
it
would
send
many
requests
to
reaper
server
to
get
manifests
for
you
know,
for
for
the
for
resource
and
kind
of
now
it
looks
less
dangerous
to
me
it's
most.
Maybe
a
lot
of
ui
changes
a
lot
of
some
some
controller
changes,
but
maybe.
J
D
And
people
kind
of
they
example
here
why
it's
useful.
It
was
like
someone
wanted
to
compose
an
application
out
of
multiple
helm
chats
because
if
you
have
an
application
and
sometimes
some
charts
split
into
parts,
because
I
don't
like
for
crds,
for
example,
convenience
season.
Yes,
yes,
go
separately,
and
it's
kind
of
now
we're
forcing
people
to
have
two
applications.
Even
though.
J
J
C
J
J
That's
very
that's
very
helm,
specific
solution,
but
then
in
in
general
I
think
yeah
like
like
I,
I
don't
have
a
strong
opinion
either
way,
but
just
that
I
think
so
far.
The
general
idea
that
an
application
is
effectively
tied
to
a
single
kit
url.
I
think
we
move
away
from
that
with
this,
which
is
why
this
probably
needs
a
little
cautious.
H
Yeah
great,
I
I
that's
why
we
didn't
want
to
do
this
in
the
first
place.
I
the
only
reason
I
reopened.
This
was
because,
from
a
back
end
technical
standpoint,
when
we
solve
for
helm,
values
in
the
re
get
repo
we
solve
for
this
essentially
and
then
so
we
could
choose
to
just
do
the
helm
one.
But
after
we
do
the
helmet
it's
it's,
it
will
be
probably
much
more
trivial
to
implement
this
one.
H
So,
but
I
agree
with
the
semantics:
it's
like
it's
a
it
could
be
considered
like
a
philosophic
philosophical
decision
and
it
is
a
technical
decision
for
this.
D
H
D
B
How
can
we
design
like?
Should
we
say
that
only
the
source
can
be
the
git,
or
only
the
path
is
different
since
the
problem,
it
is
like,
if
you,
if
you
have
different
paths
and
different
repositories,
then
the
whole
idea
of
the
githubs
like
which
commit
will
trigger
the
what
deployment
right,
yeah.
H
H
B
Maybe
I'm
like,
I
can
take
one
minute
to
think
how
to
solve
it
like
see,
for
example,
can
we
introduce
overlay
concept
kind
of
like
if
they
want
to
have,
they
can
have
two
parts
in
the
same
source
and
then
we
can
apply
the
order
of
the
paths
they
do.
It
is
that,
does
it
make
sense,
or
maybe.
D
We
can,
I
see,
advantage
of
multiple
sub.
You
know
right
now.
We
kind
of
have
support
for
multiple
passes,
but
it
only
works
if
you
use
plane
manifests,
but
a
user
might
want
to
combine
two
customized
based
applications
in
the
same
repo
under
two
different
buses.
This
is
not
supported
right
now,
like
we
and
yeah.
So
you
what
you're
saying
it
can
make
it
possible.
D
H
D
It
looks
like
it
would
be
three
different
parameters.
You
know
for
school,
which,
for
each
surprise,
yeah.
J
And
I
think,
besides
the
philosophical
question,
there
is
no
good.
There
would
be
no
good
way
to
explain
to
people
why
they
would
create
three
applications
versus
one
application
with
three
git
repositories,
so
which
means,
if
we
only
solve
this
for
the
helm
use
case.
We
have
a
very
strong
reason
for
that
that,
in
case
of
hem,
the
chart
can
come
from
one
source.
The
values
file
is
another
yaml
which
can
come
from
another
source.
So
there
is
a
strong
reason
to
bring
this
into
support
for
a
very
specialized
use
case.
J
D
Yeah,
I
think
it's
awesome
feedback.
I
will
definitely
you
know
I
will
I
maybe
I
will
keep
it
on
the
roadmap
and
maybe
it
will
have
less
priority
and
then
introduce
separate
item
just
for
the
film
and
yeah.
D
Because
feature:
okay,
yeah,
I
think
we
don't
want
to
like
design
it
right
now,
so
we
can
move
to
the
next
one.
So
I
added
the
next
feature.
It's
kind
of
it
was
implied
that
we
will
work
on
it.
So
if
you
know
that
we
are
working
on
config
management
tools,
enhancement,
there
is
a
like
pull
request
that
hopefully
will
be
merged
soon.
That
introduces
backend
part
of
it
and
so
basically
idea
that
it
will
be
easier
to
integrate
more
config
management
tools.
That
requires
things
like
python.
D
You
know
something
that
you
cannot
just
deliver
as
a
binary,
so
that
work
is
more
or
less
done
already,
and
the
next
logical
step
is
to
build
your
cli
integration,
and
you
know
so.
Finally,
user
will
have
first
class
like
experience
with
the
third
party
config
management
tool,
and
we
recently
got
you
know
additional
kind
of
reason
to
work
on
it.
So,
for
example,
helm
had
some
country
had
some
breaking
changes
and
maintainers
push
back
to.
D
Integration
for
helm
and
attempt
to
you
know
just
drop
him
related
code
from
argo
cd
and
and
defer
it
to
your
plugin.
So
if
we're
able
to
get
the
same
experience,
then
it's
great
for
project
because
then
we
have
independent.
You
know
like
people,
won't
have
to
wait
for
next
target
cd
release
to
get
new
home
version
so
yeah
and
to
me
this
is
like
least
questionable.
Basically,
we
already
yeah.
We
are
going
to.
J
E
D
H
Yeah,
I
I
think
the
the
goal
of
the
proposal
was
that
we
should.
H
It
should
ideally
be
possible
to
implement
our
current
native
support
for
things
as
a
plug-in,
because
if
that
is
possible,
then
we've
I
feel
like
we've
done
it
right
like
generically
enough
that
it's,
then
users
can
get
the
same
experience
with
native
that
they
have
with
native
ones,
with
the
plugins.
D
All
right
yeah,
so
that
sounds
straightforward.
So
the
next
item
is
I
I
got
embedded
about
it
during
kubecon,
we
spoke
with
several
argo
users
who
so
the
problem
is.
There
are
custom
resources
that
reduce
more
like
some
communities.
Customer
resources,
controllers
of
these
resources
produce
more
resources
but
for
some
reason,
do
not
create
ownership
relationship
between
those
and
so
as
a
result,
argo
cd
cannot
correctly
visualize
these
resources
and
so
and
people
just
think.
Okay,
let
us
programmatically!
D
Let
us
provide
this
configuration
to
point
argo
cd
to
a
label
that
has
a
pointer
to
a
parent
and
apparently
it's
a
pretty
frequent
request.
I
found
several
one
of
them
and
I
just
I
closed
some
of
these
duplicates
and
found
you
know
kept
one
which
has
most
number
of
thumbs.
Ups
and
and
plus
it
has
a
kind
of
implementation,
suggestion
already
and
yeah.
So
it's
a
smaller
feature,
not
so
much
work.
In
my
opinion
and-
and
you
know.
H
I
have
a
question:
is
this
to
show
parent-child
relationships
between
application
and
say,
like
a
managed
resource
that
would
otherwise
not
appear
in
the
ui
or
is
this?
Are
we
talking
about
also
showing
parent-child
relationships
for
managed
resourcing
producing
other
resources,
but
they're
not
carrying
an
ownership
reference.
D
Second-
and
I
didn't
fetch
the
first
actually
yeah
so
in
yeah.
This
is
like,
if
basically
rcg
managed,
for
example,
istio
operator
and
easter
operator-
produce
some
other
resources.
Okay,
and
there
is
logical
connection,
but
no
like
no
link,
no
explicit
link
in
metadata,
but
the
link
can
be
somewhere
else.
It
can
be
like
in
a
label,
and
so
the
suggestion
is
to
point
our
cd
to
that
label
and.
D
The
owner
references,
yes
yeah
and
I
feel
like
in
most
cases,
maybe
this
operator
crs
kind
of
wants
to
be
careful.
So
if
you
have
the
ownership
reference,
then,
if
you
delete
this
to
operator,
then
everything
it
produced
will
be
deleted
and
to
prevent
it.
I
guess
others
of
this
cr
do
not
create
ownership
reference,
but
the
reference
still
exists
and
it's
usually
in
a
label
or
in
annotation.
C
H
H
Yeah,
I'm
sorry
pvs
are
created
by
pvcs
and
there's
no
owner
because
they
don't
get
deleted
when
pvcs
are
deleted,
but
we
we
don't
want
to
show
the
other
thing
as
like
a
dangling
object.
H
So
so
we
have
these
special
treatments
for
native
kinds
and
then
I
guess
it
makes
sense
if
we
could
allow
users
to
configure
this.
I
get
at
a
system
level
right,
alex.
G
D
D
H
So
this
is
a
cosmetic
relationship
and
it
won't
affect
any
kind
of
cascade
deletion
right.
D
H
Yeah
this
this
would
be.
I
think
this
would
be
useful
for
for
for
users.
D
Right
so-
and
there
is
one
more
that
is
one
more
question:
oh.
H
We
actually,
with
the
upcoming
tracking
id
changes.
We
actually
have
indirect
back
reference
to
the
parent
object
if,
if
they
carry
oh
sorry,
it's
a
little
bit
different.
So.
H
Yeah
yeah,
so
that
I
mean
I
don't
think
it
can
solve
the
problem
completely,
because
not
every
operator
is
going
to
carry
over
all
the
the
parent
labels,
but
we
might
be
able
to
have
a
cheap
way
of
automatically
detecting
this
by
the
new
tracking
id
that
we
have.
H
So
if,
if
we
see
if
we
encounter
an
object,
that
has
a
tracking
id
that
doesn't
make
sense
because
it's
for
another
resource,
we
know
that
it
was
carried
over
by
the
true
parent
and
then
we
can
establish
a
child
relationship
like
maybe
automatically
even
okay.
I
mean,
let
me
make
sure
I
write
this
down
because
at
least
in
the
I'll
write
down
in
the
notes
here
and
then
I'll
update
the
issue
with
some
of
these
ideas.
D
D
So,
as
you
might
know,
we
already
have
a
way
to
customize
diffic
and
ignore
some
fields
from
the
resource,
so
that
arguably
don't
report
difference
in
those
fields
as
a
difference
but,
and
then
users
asking
for
kind
of
next
step
they
asking
for
ability
to
do
not
change
these
fields
during
thinking
and
kind
of
just
keep
it
untouched
completely
and
that's
what
that
feature
is
about
so
yeah
it
I
mean
we
can
debate
about
like
danny
who
you
might
know
him
he's
a.
I
mean
like
he
used
to
work
undergo
workloads.
D
He
proposed
that
feature
and
it
kind
of
solves
the
real
use
case
in
terms
of
workflows,
but
actually
the
raw
knock
strike
rollouts.
Yes,
I'm
pretty
sure
there
are
more
more
such
cases
where
you
just
don't
want
to
touch
the
field
completely,
because
you
know
that
maybe
some
controller
manages
the
field
that
it's
not
supposed
to
yeah.
So
there
are
edge
cases
like
knowledge
cases
for
the
future
very
popular
which
I
like
it
is
popular
yeah.
I
feel
like
I'm
coming.
D
E
D
H
I
think
server
side
still
honors
like
the
last
requester
that
changed
to
there
might
be
two
people
claiming
they
managed
the
resources,
but
I
think
I'm
guessing
the
last
one
wins.
I
don't
think
it
changes
the
behavior
of
how
edits
are
made,
but
it
probably
does
better
bookkeeping.
D
Yeah
so
yeah
I've,
I
think,
yeah,
we,
you
know
we
don't
have
to
like
just
discuss
how
to
implement
it's.
Just
you
know
I
just
wanted
to
make
sure
we
have
good
roadmap,
and
you
know-
and
I
didn't
want
to
put
features
into
the
list-
that
we
will
never
do
and
I
think
multiple
application
sources
is
kind
of
on
the
edge.
So
yeah.
H
I
I
would
say
we
should
commit
to
the
values
one,
but
not
the
size,
of
course
one.
So
if
you
wanna
kind
of
rearrange
it
right.
D
D
We
have
to
explain
what
the
time
line
view
is:
okay,
yeah,
but
you
can
kind
of
short
summary
that
we
hear
from
like
we've
seen
that
a
lot
when
we
support
end
users,
sometimes
argue
ui,
do
not
show
any
hint
about
why
things
goes
wrong
and
the
reason
is
it's
because
you
know
some.
The
answer
is
in
kubernetes
events,
but
they
are
not
tied
to
any
kubernetes
resource
in
the
application.
So
we
were
thinking
that
it
would
be
useful
to
simply
show
namespace
events
actually,
no.
D
Like
it
has
to
be
a
namespace
event
related
to
one
of
the
resources
managed
by
the
application,
but
users
don't
have
to
look
for
for
resource
and
click
on
it.
It
should
be
just
flat
list
of
all
such
events
kind
of
in
the
application
details,
page
yeah,
and
I
hope
that
you
know
the
error
event
would
be
like
super
visible
to
end
user,
because
it's.
H
Yeah,
so
just
I
guess
one
other
way
to
describe
it
is
like
something
bad
happened.
You
go
to
the
name
space
and
maybe
someone
goes
gmt'll
get
events
for
the
entire
namespace,
so
they
get
all
the
events
for
everything.
Actually,
that
doesn't
show
you
everything.
I
think
it
only
shows
you
for
popular.
H
No,
I
think
you
chose
it
for
everything,
but
then
you
take
those,
and
then
you
present
it
in
the
ui
in
the
in
a
timeline
that
you
can
then
click
on
individual
event,
things
and
you
can
also
visually
see
where
events
are
clustered
around,
so
that
that
you
know
hey,
there's
a
lot
of
activity
at
this
point
in
time,
something
bad
and
warnings
would
be
highlighted
in
red.
H
We
were
also
thinking
that
node
events
are
also
interesting,
because
if
a
node
became
unavailable
or
got
drained-
or
you
know
a
lot
of
there's
a
lot
of
issues
that
happen
around
the
node,
so
we
would
include
both
the
namespaced
events
as
well
as
node
events.
H
H
This
name
space
of
this
application,
then
yeah
and
we
should
filter
out
on
the
only
the
ones
relevant
to
the
application,
because,
just
because
you
have
something
running
in
cube
system
like
one
thing
running
a
cube
system,
maybe
you
don't
doesn't
mean
you
also
get
to
see
all
the
other
apps
in
cube
system.
H
F
H
Oh,
that's
yeah!
If
there's,
if
there's
precedence
for
this,
then
yeah
we
should
get
inspiration
from
that.
Should.
D
K
Hey
alex
yeah,
yeah,
so
yeah.
So
recently
we
added
the
feature
to
argo
road
controller
yeah.
It
looks
fine,
but
the
issue
is.
I
find
that
when
we
enable
to
replica
of
the
controller,
then
to
replica
will
explode
some
matrix
and
basically
one
top
type
of
metric,
which
is
retrieved
from
the
external
sad
say
they
would.
The
controller
will
do
a
list,
the
rollout
and
then
posted
the
the
matrix.
K
So
both
the
primary
and
the
backup
can
have
consistent
data,
but
for
the
reconciliation
statistics
right,
the
backup
one
cannot
expose
the
correct
data
saying
if
the
backup
is
not
working,
so
the
count
basically
will
be
zero.
And
then,
if
we
show
the
data
in
the
rafana
dashboard,
then
you
will
have
like
10
reconciliation,
loop
and
then
the
next
one.
Next
time
you
you,
you
query
to
the
back
one,
then
it
will
return
the
data
zero
right.
So
that
makes
data
inconsistent.
H
K
K
Another
solution,
another,
the
second
solution
I
put
here
is
add
some
label
to
the
reconciliation
metric,
like
the
instance
id
and,
and
the
second
label
called,
is
primary,
so
just
to
distinguish
this
metric
between
primary
and
the
secondary
one
important.
D
In
order
to
enable
leader
election,
did
we
already
give
our
back
permission
to
controller
to
manage?
I
know
like
talk
to
kubernetes
api.
K
Yes,
I
think
so
yeah,
okay,
the
controller
itself
knows
if
I
am
primary
or
not
yeah.
H
Okay,
so
would
couldn't
we
just
say:
if
the
controller's,
not
the
primary,
it
still
has
a
metric
endpoint
so
that
it
you
know
it
could
respond
to
a
health
check
if
it,
if
necessary,
but
when
it
emits
how
that
metric
endpoint
is
just
a
blank
screen.
H
Then
then,
it's
essentially
the
equivalent
behavior
that
we
have
today,
where
there's
only
one
worker,
that's
doing
all
the
work
in
yeah
there's
and
the
second
ones
won't
contribute
to
skewed
metrics.
I
think
it
would.
D
H
Assuming
it
it
doesn't,
it
does
the
right
thing.
Would
that
be
the
right
of
approach
to
this?
K
You
mean
return,
zero
data
right.
H
K
H
F
F
Actually
because
I
don't
see
the
issue,
but
the
load
increases
as
the
number
of
ports
limit.
These
events
and
we
already
seeing
at
least.
H
Yeah,
that
might
be
a
telegraph
problem,
because
it's
up
to
the
scraper
to
send
the
events
away
front
but
yeah
this
would
this
will
at
least
I
think
that
should
be
your
concern,
because
if
there's
one
less
thing
sending
stuff
to
wavefront
and
it
should
not
increase
the
load.
F
Yes,
yes,
I
think
that
is
my
other
comment
also
if
there
is
a
way
to
skip
it
actually,
yeah,
if
it
is
not
able
to
if
it
still
wants
to
emit,
but
there
is,
should
be
a
way
to
skip
those.
K
D
Thank
you
all
right
so
and
we
already
five,
you
know
run
out
of
time
and
have
to
wrap
up.
D
H
A
A
One
master
yeah
well
that
was
mainly
to
to
to
present
the
the
new
metrics
and
how.