►
From YouTube: 20220825 SIG Architecture Community Meeting
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
Hi,
everyone
today
is
august
25th
thursday,
and
we
are
here
for
the
cigar
architecture
meeting
today.
We
are
very
light
on
agenda
today
and
rihanna
can.
Can
you
get
us
started
with
the
e2e
tests
that
need
some
help.
B
Thank
you
very
much.
Yes,
we
have
two
e3
tests.
The
links
are
in
the
agenda.
You
can
open
them.
They
are
basically
tests
that
we
have
prepared
to
get
in
just
as
a
release
cut.
So
code
four
happened,
so
these
are
ready
for
review.
I
think,
hopefully,
stephen,
I
think
both
our
api
machinery
endpoints,
is
fairly
simple
test
if
we
can
get
some.
C
Yeah
the
website
machinery
that
we've
got
both
here.
C
A
Okay,
okay,
this
seems
very
straightforward.
Let's
get
this
on,
say,
apa
missionary.
It's
already
marked.
Let
me
just
triage
it
quickly.
B
A
B
Okay,
I
think
I
thought
after
this
meeting
I
would
take
it
to
do.
They
mean
I'll
go
to
a
machinery
meeting.
If
I
don't
get
review
and.
A
C
Yeah,
so
for
the
resource
quota
status
to
start
with,
if
we
just
go
to
the
code.
C
C
So
the
first
thing
I
do
is
do
the
patching,
and
once
I've
done
the
patch
of
the
status
then
we're
doing
a
check
on
if
the
value
that
I
set
there
of
750
is
correct,
as
well
as
the
annotation
that
was
set
as
well,
that
I
picked
up
from
another
test
and
then
I
go
through
doing
the
update.
A
Oh
hang
on
so
this.
E
C
Do
it
all
the
status
endpoints,
because
we
need
to
associate
both
hitting
the
endpoint
with
the
actual
e2e
test,
name
we're
actually
having
to
simulate
what
the
controller
would
actually
be
doing
so
then,
api
snoop
can
both
see
what
the
name
of
the
test
actually
hit
the
actual
status
endpoint.
So
if
you
go
back
through
all
the
service
ones,
pretty
much.
Every
status
endpoint
has
to
be
tested
this
way,
so
that
we
can
associate
the
endpoint
being
hit
by
the
actual
e2e
test
name.
A
How
is
the
default,
our
back
rules
for
resource
quotas?
Will
it
let
like
sono
boy
or
the
youtube
test
patch
things
up
this
way.
C
This
is
similar
to
another,
our
back
sorry
resource
quota
test
that
got
in
just
before
code.
Freeze
where
I'd
gone
through
and
clayton
had
checked
off
it's
it's
basically
the
same
pattern
that
I've
used.
C
It's
just
the
fact
we're
hitting
status
end
points
in
comparison
to
heading
just
the
normal
in
resource
endpoints,
okay,.
E
E
Right
I
was
saying
if
yeah,
the
quota
controller
would
see
that
the
spec
was
applied
and
then
later
patched
the
status
itself,
and
so,
if
you
were
setting
a
status,
that's
different
than
what
the
control
quota
controller
would
be
doing,
which
is
what
this
appears
to
do
you
would,
you
would
get
different
results.
Is
all
I'm
trying
to
impart
like.
If
this
is
saying,
I
desire
a
quota
with
500
millicourse
of
cpu
and
half
a
gig
ram,
and
then
the
status
block
is
later
patched.
E
C
Yeah,
I
I
I'm
picking
up
the
the
process
that
I've
used
for
the
other
status
endpoints,
I'm
using
the
same
technique,
and
I
really
haven't
had
anyone
a
lot,
because
I'm
not
too
sure
how
the
status
endpoints
are.
They
just
basically
are
read
like
more
of
a.
This
is
what
the
current
underlying
status
is,
and
it's
not
like.
E
In
the
case
of
quota,
the
status
is
basically
what
is
enforced,
and
so,
basically,
when
you
say
I
desire
a
cpu
and
a
gig
of
ram,
you
haven't
gotten
that
granted
you
to
yet
until
the
quota
controller
has
reflected
those
exact
values
on
status,
and
so
in
this
particular
case
like
I
would
expect
spec
and
status
to
always
be
equal
and
the
only
cases
where
they
might
not
be
equal,
if
I
recall,
was
like,
if
you
reduce
quota
relative
to
what
you've
already
allocated
but
either
way.
E
What
strikes
me
as
awkward
on
this
test
is
that
the
status
block
doesn't
match
the
spec
block,
and
I
I'm
yeah,
that's
probably
the
best
thing
I
could
say.
A
Yeah
and
the
other
thing
that
I
can
say
is
like
you
made
some
changes
here
and
that'll
be
stomped
by
the
controller
you
know
so
you
might,
it
might
start
becoming
flaky
depending
on
you
know
when
it
runs
and
when
the
controller
updates
things
right.
C
Yeah,
the
the
initial
tests
that
I've
done
haven't
done
any
flakes
per
se,
but
of
course,
I'm
only
just
doing
some
initial
very
small
test
runs
on
a
pretty
well,
it's
a
coin.
Cluster.
That's
got
multiple
controllers.
I've
tried
to
make
it
as
production
to
like
having
the
external
load
balancer
and
stuff.
So
I'm
just
I
I'd
need
some
extra
guidance
or
I.
A
So
so,
basically,
what
derek
was
telling
you
before
was
like
do
the
same
thing:
that
the
resource
quota
controller
would
do.
Okay,
you
know
exactly
simulate.
You
know
what
you
put
in
on
the
bucket,
where
you
need
to
set
it
and
then
what
the
resource
controller
would
would
set
the
values
to
like.
If,
if
it
maps,
then
it
really
doesn't
matter
whether
the
resource
code,
controller
stomped
on
it
or
your
changes
actually
passed
the
cause,
the
updated
values
right.
D
A
E
A
Okay,
okay,
so
since
derek
spoke
up
after
signing
direct
for
this.
A
Okay
sounds
good,
any
did
you
add
anything
else
while
we
were
talking
ryan.
B
Oh,
it
might
be
anything
close
by
video
yeah
internet
as
we
went
through
it
this
week
we
looked
at
these
and
said
we
should
investigate.
Maybe
these
are
ineligible
because
it's
optional,
but
we
haven't
looked
at
it
at
all.
I
thought,
while
we
have
experts
in
the
room,
if
there's
any
opinions
about
whether
these
are
optional
or
not
else,
we
can
just
go
dig
through
the
documentation
as
we
try
and
test
them.
D
A
Well,
the
question
is
the
I
don't
know,
I
don't
know
how
to
phrase
it.
I
think
I
guess
the
best
way
to
phrase
it
would
be.
Our
back
is
not
a
pre-required,
not
not
a
prereq
for
the
e2
conformance
test.
F
F
So
no,
this
is
not
related
to
our
back
directly.
So
I'll
ignore
the
fact
that
I
think
arvac
should
be
part
of
performance,
because,
whatever.
D
F
B
F
Yeah
those
apis
are
just
ask
they
they're
exposing
the
ability
to
ask
the
api
server.
Hey
can
set
actor,
do
something
how
the
api
server
evaluates
and
responds
to.
That
question
is
up
to
the
configuration
it's
traditionally
our
back,
but
it
can
be
web
hook
or
node
authorizer.
All
sorts
of
things
can
respond
to
that
request,
but
that
api
itself,
I
think
it's
pretty
critical.
The.
D
F
F
E
F
That
one,
I
think
it's
just
a
scoping
mechanism,
so
it's
a
way
of
because
it's
a
separate
resource,
so
you
could
assign
authorization
to
use
that
endpoint
within
a
namespace
and
thus
limit
the
set
of
questions.
You
can
ask
the
api
server
to
things
within
that
increase,
but
it's
not
dependent
on
like.
Is
it
a
service
account
calling
it
or
like
mo,
calling
it
or
they're
calling
it?
It's
it's
just
a
scoping
mechanism
of
that
broad
api,
the
middle
one,
which
is,
I
have
some
arbitrary
actor
and
some
arbitrary
action.
Can
that
actor?
A
You
know
if
without
the
cubelet
won't
work
is
fair,
but
the
people
could
be
running
this
in
an
environment
where
they
have
their
own
cubelet
or
something
like
like
that
right
right.
A
A
F
I
I
think
I
might
be
missing
some
context.
Why
are
we
writing
tests
for
this
stuff?
Is
it
just
for
conformance.
A
Yeah,
it's
not
being
tested
right
now
in
performance,
so
what
we
end
up
doing
is
we
look
at
the
list
of
apis
that
are
not
being
tested
and
we
put
it
in
a
bucket
to
say:
hey,
we
don't
need
to
test
these
things
or
we
need
to
test
these
things
and
then,
when
it
goes
to
the
second
bucket,
then
the
rest
of
the
process
starts.
F
I
I
I
think
it
would
be
fair
if
you
wanted
to
skip
the
first
and
third
one
I
might
ask
like
david
eats,
if
what
he,
what
his
feelings
would
be
on
that.
Second,
one
tested.
A
Okay,
so
how
about
an
issue
and
a
email
to
sig
auth
and
we
can
debate
there.
A
A
D
No,
I
think
I
was
trying
to
inject
something
at
that
moment.
Okay,
I'm
back
now
I
just
dropped
a
link
to
the
list
of
remaining
end
points,
and
so
this
is
the
list
we'll
continue
to
go
back
and
ask
for
remission
until
they're
done
and
then
we'll
at
that
point,
we'll
probably
revisit
which
of
the
ones
that
we've
decided
are
not
available
eligible
performances.
We
want
to
bring
back
and
that's
where
we
might
start
discussing
things
like
our
back
and
and
other
plans.
D
B
Well,
thanks
for
that
feedback
I'll
make
the
issue
and
then
I'll
bring
it
to
the
sig
off
meeting
and
have
some
discussion
here
there,
because
there's
also
a
list
of
endpoints
that
is
not
eligible
for
various
reasons.
Like
storage,
like
vendor-specific
things
that
cannot,
you
cannot
expect
people
to
test
for
it
because
it's
very
specific
to
a
vendor
or
it's
not
or.
B
Oh,
it's
not
testable,
or
it
needs
giblet
to
work.
There's
various
rules
that
disallow
it
for
conformance
so
I'll,
bring
the
rules
and
I'll
bring
the
the
end
points
and
I'll
run
it
past
the
group,
because
we
have
spent
time
on
other
end
points
and
then
after
much
effort
it
turns
out
it's
not
eligible,
so
we
rather
avoid
as
we
really
write,
we
want
to
rush
the
last
ones
in
to
get
to
100
conformance
testing.
We
are
sitting
at
95
at
the
moment,
so
almost
there.
A
Good:
okay,
a
few
more
people
have
trickled
in
hi
ben
mayank.
Did
you
have
anything
to
talk
about
today?.
A
Okay,
good
okay,
so
if
there
isn't
anything
else,
let's
wrap
it
up
for
today
and
see
you
all
next
time.