►
From YouTube: 20220202 SIG Arch Conformance
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
Hello,
everybody,
I'm
marianne
clenans,
I'm
your
host
for
today's
gates,
conformance
meeting
on
february,
the
2nd
2022,
and
we
would
ask
you
to
please
abide
to
the
cncf
code
of
conduct
and
respect
and
be
your
best
self,
and
this
meeting
will
be
recorded
to
youtube.
So
you
can
be
yourself
doing
all
those
things
later,
so
I
will
share
my
screen.
A
And
that
does
not
show
what
I
wanted
to
show.
Let's
stop
the
share
and
share
the
screen.
Oh
yes,
it
did
show
what
I
wanted
to
show.
Yes,
okay,
so
looking
at
the
agenda,
so
john
sorry,
I
didn't
react
earlier
to
let
eleanor
know
about
the
move
as
well.
So
if
you
want
to
go
with
your
topic,.
B
Okay,
great
yeah,
so
one
of
the
things
that
we
do
assignably
right.
Obviously
we
run
the
ede
conformance
suite
by
default,
but
we
do
play
with
the
tags
a
little
bit
so
right
by
default
we
run
the
conformance
tests
that
are
non-disruptive.
B
You
can
run
the
full
conformance
suite
you
can
roll
run
just
like
maybe
one
quick
test
to
check
the
the
clusters
kind
of
up
and
running,
but
we've
got
some
requests
one
in
particular.
That
made
us
start
thinking
about
this
a
little
bit
more
really
too,
but
one
was
about
like
networking
and
saying
you
know,
there's
a
lot
of
networking
stuff.
We
would
like
to
test.
That's,
maybe
even
slated
to
eventually
become
conformant.
B
Can
we
test
that
just
but
the
problem
is
is
like
they
can't
just
grab
the
tag
sitting,
sick
networking,
because
it
just
grabs
way
too
much,
and
they
don't
want
just
the
conformance
test
that
overlapped
with
sick
networking,
because
they
would
kind
of
like
the
idea
of
well
what
about
the
things
that
are
maybe
going
to
be
conformant
in
an
ideal
world
right.
I
know
one
of
the
things
all
right.
B
The
primary
thing
that
this
this
group
covers
is
things
that
should
be
conformance
tested,
but
aren't
is
there
a
way
to
know
what
tags
that
covers?
If,
if
can
I
share
my
screen
real
quick,
I
know
you're
sharing
yours.
Can
we
trade
back
and
forth
real
quick?
Yes,
yes,
yes,
you
can.
Let
me.
B
Fantastic,
thank
you.
So
you
can
see
my
terminal
now.
B
Sonabooy
now
has
this
little
ede
command,
which
just
kind
of
bakes
in
the
dry
run.
So
we
have
the
conformance
test
list
so
that
you
can
more
quickly
identify
what
tests
you're
going
to
run,
but
then
also
a
way
to
change.
Rather
than
listing
the
test,
we
can
also
just
list
the
tags
and
the
number
of
times
those
tags
occur
so
like
if
you're
running,
just
the
conformance
suite
you
can
see
here
are
all
the
different
tags
you
might
hit,
which
by
itself
is
somewhat
interesting.
B
If
people
want
to
look
at
these
things
like
some
of
them,
there's
one
random
test
marked
fast,
but,
for
instance,
if
someone
is
doing
you
know,
sig
network
networking,
there's
all
sorts
of
other
test
patterns
that
may
or
may
not
be
relevant
to
almost
any,
but
alpha
clusters
right,
and
it's
just
it's
hard
to
know
how
these
tags
are
going
to
evolve
over
time.
B
And
I'm
wondering
if
anyone
has
thought
of
this
idea
of
kind
of
preemptively
going
and
marking
these
tech
tags
is
like
eventually
conformant
or
something
like
that
to
where
we
know
a
certain
test
or
a
certain
feature
is
eventually
going
to
evolve
into
that.
I
don't
know
which
one
of
these
features
are
right
now
only
marked
as
feature
because
they're
not
yet
tested
as
conformant
or
if
it
will
just
always
be
optional.
B
C
C
B
I
mean
are,
are
most
kept,
would
you
say
even
always
on
this
track
of
hopefully
eventually
becoming
ga
or
like
some
of
those
tests.
B
C
D
Yeah,
I
think
we
add
something
as
alpha
just
to
give
a
toggle
for
people
to
try
things
on
in
the
sense
that
you
know
we
want
the
default
to
be
something
else,
but
if
they
want
the
old
behavior,
we
would
give
them
a
switch
to
for
them
to
use
the
old
behavior,
and
we
would
do
that
as
an
alpha
and
then
we
rip
it
out.
So
I
remember
doing
it
at
least
for
like
some
time
out
situation.
D
Was
it
probe
timeout.
I
guess
that's.
E
Remembering
yeah,
it's
okay,
yeah,
that's
where
the
features
ga
intentionally,
but
it
was
like
it
was.
Never
it
turned
to
be
conformed.
It's
just
a
way
to
disable
like
return
to
previous
behavior
right
right.
D
And
the
other
related
problem
here
is,
you
know
there
are
things
like
rbac
which,
by
definition
which
everybody
uses,
but
we
can't
add
it
to
conformance
suites,
because
it's
an
optional
feature.
So
we
have
those
kinds
of
idiosyncrasies.
B
Too
yeah
that's
kind
of
the
core
one,
some
that
will
just
always
be
either
an
optional
toggle
or
a
one
of
many
choice.
Like
provider,
aws
provider
azure.
You
will
never
want
to
run
both
of
those
at
the
same
time.
It's
it's
always
one
from
the
list
right.
So
our
ingress.
D
Right
and
the
other
thing
here,
john,
is
these
tags
are
kind
of
like
debated
and
used
by
the
different
sigs.
So
some
sigs
do
better
with
the
tagging
stuff
and
other
things
are
not
so
we
basically
tell
them
like
they
will
come
to
us
and
say.
Is
it
what
was
the
example
like
windows
would
come
and
say:
hey
we
want
to.
D
D
So
yes,
we
currently
we
don't
have
a
place
to
a
way
to
indicate
that
the
specific
things
is
going
to
be
conformant
soon
or
something
like
that.
Right.
B
D
Right,
and
maybe
the
other
way
to
think
about,
it
is
like
maybe
run
one
bucket
of
data
tests,
but
then
that
could
get
mixed
up
with
it
needs
to
be
better
and
it
needs
to
be
like
what
we
end
up
doing
is
conformance.
If
you
turn
on
conformance,
we
switch
off
everything
else
right
like
so.
We
have
some
rules
like
that
right.
So
if
you
switch
on
performance,
you
should
not
have
any
tags
by
which
people
will
like
switch
things
off,
so
it
pretty
convoluted.
Basically,
that's
what
I'm
trying
to
say:
okay,.
E
Yeah
and
I
was
trying
to
clean
up
with
for
sicknote,
and
there
is
a
cap
proposal
I
posted
in
the
chat.
E
The
another
different
difficulty
in
signal
was
that
the
feature
may
indicate
special
environment
that
needed
to
run
this
feature,
even
though
it's
already
ga
or
it
may
indicate
that
there
is
a
feature
gate
which
is
completely
different
semantics.
It
may
be
on
a
path
to
conformance
or
like
node
conformance
in
our
case,
but
it's
still
marked
as
feature
so
this
cap
was
intended
to
clean
it
up
a
little
bit
for
signal,
but
also
it
spans
across
different
cheeks.
That's
why
I
put
it
on
sick
testing.
E
Okay,
I
presented
on
a
couple
times,
but
not
with
stick
testing,
so
it's
not
accepted
cap
yet
so
it
may
clean
up
things
a
little
bit
when
it's
accepted
and
we'll
start
doing
it.
B
Okay,
thank
you,
yeah.
That
would
be
really
useful
if
there
was
just
this
idea
of
that.
Consistency,
like
you
said
about,
does
feature
indicate
it's
a
flag.
Does
it
it's
different
type
of
gate
it's
triggered
by
this
or
that
if
we
had
a
couple
tracks
to
better
understand
that
that
it's
honestly
one
of
the
main
questions
we
get
from
sanibui
is
what
tags
are
there?
What
should
I
do?
B
How
can
I
run
things
and
it's
you
know
it's
hard
when
it's
distributed
and
everyone's
making
different
choices,
but
it's
also
pretty
low
priority
when
there's
so
many
other
things
to
do
so.
D
We
do
have
written
down
the
rules
that
we
currently
follow.
So
if
you
want
to
take
a
look
at
that
and
see
if
we
can
propose
something
to
the
community
to
make
it
better
make
things
better,
you
know
we'll
end
up
coming
like
if,
if
somebody
comes
up
with
like
an
idea
of
like
here
is
how
we
should
do
it
going
forward,
then
we
can
then
go
back
and
match.
You
know
the
different
use
case
scenarios
that
we
ended
up.
D
B
D
Is
also
you
know,
how
do
you
define
this
across
the
currently
maintained
stable
versions
as
well
right,
so
that
would
be
the
other
problem
that
we
face
at
that
point,.
B
Yeah,
that's
one
place
where
we
do
sort
of
appreciate
that
things
move
slowly,
because
we
can
kind
of
just
implement
we
filter
by
these
regexes.
But
honestly,
we
don't
check
the
version
and
just
choose
the
regex
afterwards.
We
just
hope
the
regex
will
be
consistent
across
a
smaller
set
of
tests
or
a
smaller
set
of
versions.
B
Okay!
Well,
thank
you
all
for
for
chiming
in
and
adding
a
little
bit
to
our
understanding
and
and
we'll
see
where
to
go
from
there.
But
I
appreciate
your
time.
C
B
Okay,
so
it's
not
looking
at
it's,
not
filtering
conformance.
C
Yeah
and
okay,
that's
an
informing
thing
for
us
that
we
should
probably
look
at
once
in
a
while
see
how
often
that,
like,
I
don't
know
where
the
notifications
go,
but
that's
an
important
job
and
then
another
one
is
there
is
for
cube
cuddle.
C
I
believe
a
list,
no
not
for
api
server,
there's
a
dash
dash,
help
command
or
exploring,
and
you
can
look
at
all
of
the
flags
because
that's
actually
what
we're
looking
for
is
the
features
inside
of
api
server
there,
and
it
tells
you
whether
those
individual
features
are
on
by
default.
C
And
so
there
might
be
this
overlap
of
check
for
this.
This,
the
job
that
jordan
set
up
and
then
for
all
of
the
flags
that
are
on
by
default
and
look
at
the
tests
that
pass
in
that
overlapping
set
and
that's.
B
D
Might
be
ga,
but
you
know
it's
not
non-bi-digital,
so
we
have
those
scenarios
as
well.
C
B
B
On
the
test
grid,
for
that
that
test
run
that
you're,
saying
ben,
has
and
and
I'll
double
check
those
future
flags
for
the
api
server
and
see
if,
if
that
could
help
us
out,
I
appreciate
it.
E
B
E
Can
review
prs
that
I
sent?
Was
it
still
helpful
or
maybe
leave
some
comments?
It'll
be
so
appreciated.
A
All
right,
thank
you
very
much.
Certainly,
you
became
host
when
I
fell
off
due
to
internet
access.
So
if
you
can
pass
that
back
on
I'll
share
my
screen
or
you
can
share
the
screen
with
the
confirmation
meeting
again,
you
go
to
participants
and
there
we
are
now
I'm
an
artist
again.
Thank
you
very
much.
Okay.
Let
me
share
the
screen
that
one
there.
Okay
next
topic
is
from
steven
right
api
service
test.
A
So
the
background
on
this
is
our
back
as
we
just
discussed
is
optional,
so
stephen
I
and
we
need.
We
would
like
to
use
that,
but
we
can't
so
when
you
can
lead
the
conversation.
F
Need
quite
a
bit
of
setup
and
the
way
that
I've
been
doing
it
is
using
a
image.
That's
been
used
for
some
of
the
other
historical
endpoints
around
api
service
and
jordan
picked
up
the
fact
that,
because
we
were
using
rbac
as
part
of
setting
up
the
test
that
started
the
whole
discussion
around
the
need
for
it.
Sorry,
it
shouldn't
be
used
because,
of
course,
it's
optional
and
that
led
to
the
those
endpoints
being
made
and
eligible
for
now,
maybe
for
a
profile
at
a
later
date.
F
But
the
problem
is
that
the
image
that
I'm
currently
using
to
set
it
up
seems
to
be
stuck
and
wanting
our
back
to
get
that
test
actually
up
and
running.
And
if
I
don't
do
that
part
it
just
won't
come
up
to
the
the
endpoints
might
come
up
stably
in
a
stable
format
for
me
to
test.
So
I
was
actually
hoping.
F
Clayton
was
going
to
be
around
to
get
a
bit
of
more
of
an
insight,
but
maybe
I
can
ping
him
unless
anyone's
got
any
extra
ideas,
because
I
seem
to
be
in
a
stuck
in
a
catch
22
off.
F
I
need
a
rbac
based
on
the
image
that
I'm
using
and
I
have
gone
through.
The
process
of
setting
up
a
new
image
is
or
getting
a
revision
to
an
image,
but
it's
just
going
to
delay
those
endpoints
getting
tested
and
also
there's
a
potential
issue
around
h.a
for
the
api
service.
Endpoint
there's
one
of
the
endpoints
things
they
have
a
bit
of
a
glitch
when
I
use
it,
and
I
that
comes
back
to
the
aha
part.
D
I
guess
at
this
point
we
can
drop
an
email
and
instead
of
trying
dm
ping
yeah,
then
we
have
some
something
tracking
that
that
all
of
us
can
watch
right.
Yeah,
and
this
is
the
api
service.
So,
let's
throw
it
out
to
the
apa
missionary,
mailing
list
and
yeah
know,
add
clayton
explicitly
clayton
or
jordan
and
jordan.
I
guess.
F
Yeah,
because
both
clayton
and
jordan
have
painted
on
the
original
issue
yeah,
it
was
just
trying
to
find
out.
What's
the
best
way,
do
we
need
a
separate
image
that
doesn't
need
the
rbac
components
so
that
the
endpoints
become
available,
but
also,
if
they,
if
there's
any
other
guides,
so
hopefully
someone's
got
a
potential
extra
documentation
that
may
help
me
understand,
what's
falling
over
or
not
falling
over,
so
yeah
we'll
look
at
getting
something
set
here.
A
All
right,
let's
roll
over
to
the
next
one
thanks
for
that,
then
yeah
we'll
send
out
that
email
and
get
some
some
feedback
from
api
machinery.
F
Okay
ryan:
can
you
just
open
up
the
link
for
that
controller,
go
and.
F
Okay,
so
if
we
scroll
right
to
the
bottom,
I've
got
just
that
last
section
here:
I've
after
setting
up
a
diamond
set,
I've
got
doing
a
list
of
all
the
revisions
based
on
our
label
selector
and
that's
doing
a
list
for
all
namespaces
for
controller
revisions
and
after
getting
that
list,
then
to
start
with
I'm
doing
a
get
to
just
look
at
reading
the
individual.
F
F
I've
got
some
some
initial
points
around
assumptions
that
I've
got
around
some
of
the
assumptions
for
controller
revisions,
how
they're,
potentially
working
and
a
few
things
around
the
hash
that
gets
created
and
we're
trying
to
guess
some
clear
guidance
around
how
the
create
and
some
of
the
updates
and
patch
process
works,
so
that
I
can
better
track
it
using
api
snoops,
because
at
the
moment
try
just
it's
not
a
well-used
sort
of
which
is.
B
F
A
popular
endpoint
that
most
people
sort
of
think
about
resource
and
I'm
just
having
some
issues
just
trying
to
find
some
extra
functionality
and
also
the
existing
integration
tests
are
using
a
fake
client
and
for
the
e2e
test.
I
believe
I
should
be
using
the
real
history
option
for
working
with
the
control
revisions
and
I'm
just
wondering
how
that
could
cause
some
potential
issues
and,
if
you
that
very
last
link
on
that
page,
there
yeah
that
one
there
on
for
that
line.
F
Sorry,
if
you
scroll
up
just
a
fraction
further
to
get
the
start
of
the
yeah,
sorry
to
get
the
name
of
the
function,
create
control,
revision
and
yet
after
it
does
a
little
bit
of
a
setup,
the
line
that
I've
highlighted
is
where
it's
actually
doing
another
rest,
another
call
back
out
to
say:
hey
now.
Let's
actually
call
do
the
call
to
actually
create
it,
and
this
is
where
I'm
like,
losing
the
threat
of
where
stuff's
happening
within
the
cl
the
control
plane.
F
It's
like,
I
thought
I
was
in
the
right
place
for
the
create
part,
but
it's
like
I've
sort
of
come
full
circle
when
I'm
losing,
who
is
now
actually
really
dealing
with
that
of
the
create
process,
because
at
the
start
of
the
file
at
the
very
top
round,
there
is
some
ref
there's
a
reference
about
a
new
controller
revision
resource.
Sorry,
if
we
just
brought
it
in
a
little
bit
constant
yeah
line
66
function
create
a
new
control
revision.
C
Can
I
summarize
real
quick
what
I
hear
you
saying,
because
some
of
this
is
maybe
not
straightforwardly
clear,
is
we're
trying
to
implement
a
test
that
does
what
the
controller
normally
does
and
interacting
with
the
api
server,
and
we
don't
consider
coverage
for
conformance
unless
the
e
to
e
test
suite?
Is
the
api
user
agent
communicating
with
the
api
server
so
we're
trying
to
implement
some
of
the
logic
inside
of
the
controller,
in
this
case
the
controller
history,
to
be
able
to
generate
the
correct
names
within
our
ede
test?
F
I've
got
initial
call
out
to
sig
apps
on
this
already
done,
so
I
think
I'll
probably
need
to
re-paint
them,
maybe
email
their
mailing
list,
probably.
F
I
posted
about
a
week
ago
this
this
document
that
we
just
had
like
just
before
getting
asking
for
feedback,
as
well
as
the
case
performance,
with
a
link
back
to
the
sig
apps
slack
channel,
and
he
said
he
was
going
to
have
a
look
at
it,
but
I
think
he's
just
snowed
under
with
stuff
yeah.
F
D
Yeah,
let's
do
the
same
thing,
see
gaps
mailing
list,
let's
hit
them
up
yeah
and
today
tomorrow
it's
going
to
be
difficult
to
get
anything
out
of
people
because
of
the
feature
caps
deadline.
D
A
E
Oh,
I
I
just
have
a
small
question
and
I
want
to
pick
your
brain
on
what
would
be
the
best
approach
here.
So
there
is
a
pod
overhead
feature
that
we
gain
this
release
and
this
powder
head
feature
if
you're
not
familiar
with
that,
you
can
specify
random
class
and
runtime
class
can
specify
overhead,
and
this
overhead
will
be
applied
to
every
port
that
uses
this
runtime
class
and
then
the
result
of
it
will
be
that
c
groups
will
be
allocated
with
more
memory.
E
Almost
more
cpu
depends,
like
just
overhead
will
be
added
to
the
c
group
sizes.
So
since
we
gain
this
feature,
I
was
curious.
What
can
we
do
to
like?
E
Do
we
want
to
do
conformance
of
that,
and
the
complication
here
is:
how
do
I
test
conformance
of
the
feature,
so
I
can
tell
that
the
port
will
be
like
overhead
will
be
added
to
support,
but
I
don't
really
have
a
way
to
validate
that
c
groups
will
be
allocated
with
a
proper
additional
size,
nor
I
can
like
I
mean
I
don't
even
know
like
how
it
will
work
on
windows,
and
what
do
I
check
on
windows?
E
F
I
had
a
look
at
the
test
plane
layout
that
link
able
to
be
brought
up
because
at
the
end
of
your
test
plan,
you're
looking
at
doing
a
lot
of
work
with
tubelet
and
on
the
conformance
test
requirements.
F
That
makes
the
endpoint
that
it
doesn't
make
the
conformance
test
requirements.
C
D
D
Yeah,
correct
yeah,
so
the
only
thing
that
we
can
check
is
you
know
what
is
accessible
through
the
api
server,
only
right
so
and
if
this
thing
that
we
are
switching
on,
if
it
doesn't
result
in
a
change
in
what
we
get
back
from
the
epa
server
upon
querying
later,
then
there
is
nothing
we
can
do
about
it
and
we
can't
basically
test
id
to
conformance
at
that
point.
D
All
right,
okay,
so
some
other
things
won't
even
have
c
groups
right
like
windows.
You
were
saying
so
I
think
this
can't
be
a
candidate
for
conformance.
E
Okay,
if
I
will
find
a
way
to
check
for
available
memories
that
are
left
on,
the
node
will
be
something
that
I
can
do
like.
Can
I
relay?
Can
I
just
take?
It
depends
on
that
that
available
memory
decreased
on
that
amount.
Why
shouldn't
I.
D
I
think
that
is,
I
mean
if
we
are
able
to.
We
should
be
able
to
check
the
memory
on
the
node
right
and
the
memory
changes
that
is
observable
through
apa
server
for
sure.
D
But
then
you
know,
are
we
going
to
do
this
on
every
node
or
you
know
we
get
into
that
thing?
E
Okay,
so
I
guess
simplest
will
be
just
to
validate
that
admission
worked
and
we
applied
the
other
hat
to
the
port,
but
then
we
wouldn't
check
the
functionality.
Oh
right,
okay,.
D
F
A
And
once
once
you
have
the
test
up
and
running
and
you're
careful,
some
checking
with
weather
api
still
pick
it
up
touching
the
endpoints
you
can
reach
out
to
to
myself
or
stephen.
I
have
here.
We
can
help
you
with
apis
or
you
can
just
download
and
run
it.
E
A
And
I
thank
you
very
much
and
you
get
20
minutes
of
your
day
back
and
thanks
for
all
that.
The
influence.