►
From YouTube: Kubernetes SIG API Machinery 20200923
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,
officially
I'm
recording
hello.
Everybody
today
is
september,
23rd
fall
is
with
us
on
the
northern
hemisphere.
Spring
is
in
the
south
hemisphere.
This
is
the
api
machinery
meeting.
A
We
have
a
number
of
topics
today
and
we
are
going
to
get
to
them,
so
we
can
make
a
good
use
of
the
time.
The
first
topic,
I
believe,
is
from
david
ashbal,
who
is
here
in
the
call-
and
I
will
give
it.
B
To
him
yeah
not
too
much
unless
people
have
questions,
I've
been
working
on
the
implementation
for
the
tracing
cap.
I
mostly
just
want
to
ping
this
before
it
gets
too
close
to
the
kept
deadline.
B
Yeah
thanks
for
doing
that.
Is
it
still
the
same?
The
last
state
I
thought
in
I
think
you
had
changed
it
to
just
be
about
like
request,
pass
and
not
include
changes
to
controllers
or
that's
correct,
yep
yeah,
which
I
think
was
okay,
but
I
forget
what
this
speak
was.
B
We
were
looking
into
the
egress
selector
stuff
and
making
sure
that
that
worked,
and
you
asked
for
some
more
time
to
look
at
it,
since
it
adds
a
new
egress
to
the
api
server.
I
think
you
also
might
have
wanted
david
to
take
a
look.
I
think
you
commented
something
like
that.
B
I
think
it
just
needed
more
time
last
release,
but
oh
right
right,
because
because
now
api
server
will
call
out
to
the
monitoring
the
collection
point
whatever
that
is
yeah,
that's
correct,
yeah!
Okay,
that's
that's
definitely
worth
the
I'm
glad
I
thought
of
that.
That's
definitely
worth
considering.
B
Okay,
I'm
pretty
busy.
You
may
need
to
aggressively
nag
me
on
slack
okay.
I
can
do
that
thanks
daniel
yeah.
It
might
also
be
good
just
to
send
a
note
to
the
api
machinery
mailing
list-
okay,
yeah,
just
just
that
you're.
The
thing
is.
B
A
Yep,
what
is
the
deadline
for
enhancements
for
120?
How
much
time
do
we
have.
B
A
Okay,
we'll
make
sure
you
can
ping
me
also,
if
nobody's
paying
attention-
and
I
will
try
to
find
the
people.
I
think
all
right,
thanks,
very
good.
Moving
on
jq
michael,
are
you
on
the
call.
C
Oh,
so
we
have
another
member
on
the
call.
Okay.
She
he
wants
to
present
the
question
or
I
can.
I
can
go.
D
Yeah,
hey
so
hi,
so
I
I
just
had
a
question
about
like
so
we
have
a
lot
of
clusters
that
have
like
loads
of
pods
and
some
of
the
cluster,
which
have
like
less
number
of
watts,
but
having
a
very
large
amount
of
data
in
their
prospect.
So
what
happens
is
like
if,
if,
if
we
have
a
controller
which
is
having
a
memory
limit,
that
is,
let's
say,
500,
mb
or
1gb,
which
is
kind
of
limited,
but
so
so
with
the
client
go
cache.
D
What
happens
is
like
we
cache
each
and
every
pod
object
in
the
memory,
so
we
run
into
these
memory
limits
so
right
now,
what
workaround
we
have
done
is
like.
We
have
written
a
custom
controller
in
which
we
are
stripping
down
the
port
object
and
keeping
everything,
keeping
only
the
required
details
in
our
cache.
Instead
of
storing
everything
and
we
are
using
the
pagination
to
basically
we,
we
are
basically
getting
the
pagination
output
and
shipping
down
the
pod
object
before
sending
it
to
like
the
page,
paginator
aggregator.
D
B
D
B
And
I
yeah
usually
says:
there's
a
metadata,
informative
chat.
B
We
can
I
look
at
the
the
issues
briefly
here
and
it
does
sound
like
something
that
some
controllers
would
want,
so
I'm
actually
gonna
tag
instally
to
see.
If
there's
it
just
makes
sense
in
like
the
controller,
runtime
or
cute
builder,
or
something
like
that.
E
Yeah,
okay,
would
that
be?
Would
it
be
a
wanted
api
feature
to
only
request
fields
that
you
want,
like
some
apis
offer
that
you
select,
which
fields
you
want
and
you
get
an
impartial
object.
B
So
that's
yeah,
yeah,
selecting
exact
like
requesting
specific
fields,
is
not
something
we
offer
and
it
would
kind
of
be
a
big
deal.
So
I
don't
see
it
happening
like
I
wouldn't
say
we're
like
against
it
or
anything
it's
just
it
hasn't.
We
we
haven't
had
a
the
the
horde
of
users
with
pitchforks
and
stuff
haven't
shown
up
demanding
that
feature.
Yet
I
I
mean
I
I
think,
there's
probably
I
I'm
I'm
sure
that
there
have
been.
You
have
issues
requesting
it.
B
There
just
hasn't
been
a
critical,
a
critical
mass
and
it
is
difficult.
So
it.
F
F
I
would
also
say
that
metadata
or
potentially
even
further
restricting
to
namespace
and
name
is
probably
a
bigger
bang
for
the
buck,
because
it
would
at
least
namespace
and
name
would
eliminate
any
potential
confidential
content,
while
still
letting
people
see.
What's
there,
we've
wanted
that
in
the
past
and
then
eventually
decided.
You
know
what
we
can
just
afford.
The
memory
to
put
all
in.
C
Thanks
daniel,
do
you
mind,
link
us
to
the
media
invitation
so
that
we
can
track.
B
Yeah,
all
all
all
tag
here
in
a
second
thank
you.
A
Very
good
looks
like
it's
something
that
we
can
follow
up
on
the
issue
itself.
Thank
you,
daniel
for
tagging,
sony
and
everybody
for
commenting.
Finally,
thank
you.
David
eats
for
joining
we've
been
moving
the
order
of
the
items.
I
think
we
can
go
to
the
main
item
that
we
had
today,
which
is
the
pga
features.
G
F
Okay,
there's
still
some
open
questions
and
discussion
on
the
doc,
but
it's
it's
moving
right
along,
so
we
were
trying
to
look
and
assess
what
features
we
still
have
in
flight
that
need
to
be
worked
in
either.
You
know
in
some
cases
you
know
you
decide
to
eliminate
an
alpha
feature
or
you
indicate
how
you're
going
to
keep
moving
forward.
F
So
if
we
look
at
the
pre
ga
features,
you
can
see
the
list
that
we
have.
I
don't
think
people
are
going
to
be
hugely
surprised
by
what's
here,
because
we
talk
about
it.
Pretty
frequently
server
side
apply
is
a
big
one.
It's
been
progressing
consistently
priority
and
fairness
has
also
been
progressing
consistently.
More
improvements
were
made
in
119,
but
we
aren't
going
to
rush
something
to
beta
until
we're
confident
that
it
actually
does
what
we
want.
F
The
allow
insecure,
back-end
proxy,
it's
code
that
we
own,
but
that
was
heavily
off,
dependent
that
one's
probably
going
to
move
forward
to
ga
api
response.
Compression
is
one
people
probably
never
even
heard
of
I'm
pretty
sure
it
was
broken
when
we
merged
it
and
it
never
moved
forward.
F
F
I
think
the
other
ones
on
the
list.
At
least
the
next
couple
of
daniel
might
recognize
more.
B
Yeah,
I
just
pinged
ciao
about
the
remaining
item.
Count
he
actually
thinks
it's
beta
and
that
there's
no
blockers
and
moving
it
to
ga,
so
that
probably
just
needs
a
little
bit
of
a
little
bit
of
work.
B
B
Okay,
okay,
yeah:
we
should.
We
should
figure
that
out,
remove
self
link.
I
know
wojciech
is
super
motivated
about
that.
So
I
expect
that
to
make
progress
back.
That's
pretty
sure
I
reviewed
a
change
that
actually
implements
it.
B
Yeah
yeah,
I'm
pretty
sure,
maybe
there
it
was
implemented
in
alpha,
but
then
when
he
turned
it
on
by
default,
like
a
bunch
of
tests
were
broken,
so
we
fixed
the
test,
I'm
pretty
sure
it
repeated
that
okay
selector
index.
I
don't
actually
that's
not
jogging
my
memory.
What
does
the
selector.
F
Index
give
me
it
was
the
faster
way
to
do
a
field
selector
and
label
selector.
It
was
like
a
pre-computed
hatch,
oh
yeah,
that
one.
B
Yeah,
I
think
it's
done
and
working
but
yeah,
I'm
I
if
it
just
went
beta.
I
think
we
can
definitely
wait.
Another
release
to.
F
Give
it
ga
I'd
like
to
some
storage
cash
problems
have
been
latent
and
I'd
like
to
give
it
a
little
bit
more
soak
time
before
saying
we're
done.
B
Yeah
yeah,
I
recall
making
a
few
rounds
on
that
one
and
I
think
it
was
in
a
pretty
good
state
when
it
emerged.
F
I
guess
morning,
hitters
is
me
and
jordan.
We
got
it
in
119..
F
I
need
to
go
double
check
the
promotion
criteria,
the
big
ones
were
the
admission
web
hooks,
but
those
got
done
in
19
and
I
know
he
has
some
work
where
he
wants
to
standardize
a
way
to
do
fields,
but
I'm
actually
pretty
happy
with
the
feedback
we're
getting
back
from
it.
Already.
I'd
probably
be
willing
to
take
this
one
to
ga
in
either
120
or
121..
B
Yeah
yeah
and
I
think
jordan's
motivated
so
and
then
at
the
end.
Here
are
two
things
that
I
ran
across
that
don't
technically
have
caps
one.
Is
this
controller
manager
component
too
big,
there's
a
long
issue
with
a
lot
of
people
asking
about
a
cap
and
like
I
think,
one
attempt
that
didn't
go
anywhere,
so
I
recommended
just
closing
that
wait
for
a
owner
to
show
up
and
be
motivated.
F
F
That's
the
the
same
one
I
remember
was
lexus
and
then
stefan
had
a
few
comments,
but
you
know
there
was
some
interest
in
like
it'd
be
cool.
If
we
had
this,
but
I
don't
remember
anyone
saying
this
is
important
enough
to
me.
I'm
going
to
work
on
it.
B
F
B
Sure
yeah,
I
can
double
check
with
with
walter.
If
he's
not
on
the
call,
oh,
you
know
I
think
he
may
have
had
some
training
or
something.
F
B
I'll
add
a
comment,
and
then
this
this
last
one
I
I
didn't
write
it
down
very
good.
I
said:
stop
breaking.
I
should
have
said:
stop
unintentionally,
breaking
client
go
consumers
because
sometimes
we
it
is
intentional
and
that's
good
right,
like
we
add
context,
hopefully
everyone,
even
those
on
the
receiving
end
of
that
believe
that
it
was
worth
it.
B
The
only
thing
I
can
think
of
is
like
a
like
a
a
free
submit
test
that,
like
lists
all
the
public
packages
and
methods
and
warns
us
when
they
change
that
still
wouldn't
get
100
of
the
way
there,
because
you
could
have
changes
that
don't
breakage
that
doesn't
show
up
from
the
public
interface.
But
anyway,
this
could
use
an
owner
if
somebody's
interested
in
you
know
helping
our
users,
not
hate
us.
F
F
And
I'm
not
sure
that
every
aspect
should
be
held
to
the
same
bar
right,
some
are,
are,
you
know
really
really
annoying?
Others
are
you
know,
sort
of
helper
packages
where
maybe
it
just
it's
better
to
make
this
change,
and
it's
not
like
changing
every
single
client.
This
is,
if
you
happen
to
be
using
this
feature
of
a
work,
something
changed.
I
don't
have.
F
B
Yeah,
my
point
isn't
to
make
it
completely
impossible.
My
point
is
to
make
it
visible
and
conscious
right,
like
I,
I
think,
if
we're
going
to
break
users
and
then
someone
in
a
high
level,
owner's
file
should
be
looking
at
that
and
being
like
yeah.
This
is
worth
it
yeah.
That
was
my
only
thought,
but
that
makes.
B
Rate,
if
nobody
signs
up
to
do
it,
it
probably
won't
happen.
A
Okay,
so
is
there
on
this
long
list?
I
think
there
is
a
there
is
a
part
of
this
which
is
update
for
the
people
in
the
meeting
and
future
generations
is.
Is
there
a
particular
ask
that
we
are
having.
F
I
wanted
to
make
sure
everyone
was
aware
of
what
we
had.
There
are
a
few
and.
B
Yeah
people
should
be
aware
of
what
these
features
are
and
what
we're
doing
also
some
of
these
could
use
some
attention,
but
I
think
we
should
make
a
issue
and
a
checklist
something
like
that
to
to
identify
owners.
I
wouldn't
want
to
limit
it
with
just
people
on
this
call,
although,
if
any
of
you
feel
attracted
to
one
of
these
items
by
all
means.
A
Feel
free
to
claim
it,
I
think,
as
part
of
what
is
going
on
in
all
the
other
six
and
in
general,
in
the
kubernetes
community.
We
we
want
to
have
a
tracking
mechanism
for
these
features
and
and
other
features
to
come
to
see.
You
know
the
the
graduation
status,
the
ownership
status
if
they
are
making
progress
or
not,
etc.
So
this
is
probably
the
first
step
for
that,
and
then
we
will
put
it
in
a
place
where
it
is
clearly
accessible
and
public
and
we
will
try
to
keep
it
up
to
date.
A
And,
of
course,
as
daniel
was
saying,
the
ones
that
don't
have
a
clear
owner
or
the
ones
that
may
require
some
help,
are
great
opportunities
for
people
in
this
column
outside
of
this
boat.
To
do
some
work.
A
What
else
did
we
have
anything
else?
So
no
tracing
the
question
about
play
and
go
the
pga
features,
and
if
that
is
everything
we
had
before
the
last
point
is
from
me:
there
is
a
requirement
to
change
the
zoom
settings
in
all
the
kubernetes
meetings,
so
I
will
be
working
on
that
property
on
friday.
I
I
give
you
a
fair
warning,
because
it's
a
long
mailing
list
and
you
will
be
receiving
some
updates.
A
A
A
We
may
want
to
use
zoom
in
the
weekly
background
soon.
Maybe
I
will
take
this
opportunity
to
change
that.
We've
been
using
google
mid
and
I
think
if
we
use
zoom,
we
can
record
them
and
put
them
in
a
in
a
regular
youtube
playlist.
So
it's
accessible
too.