►
Description
Nominations for co-chair
Discuss PR about konn-client leaks in ANP
Request help on webhooks in CCM KEP
Overview of AP Expression changes.
A
Welcome
to
the
december
2nd
2021
cloud
provider
extraction
working
group
meeting.
As
always,
these
meetings
are
part
of
the
cncf
kubernetes
project
and
we
are
beholden
to
their
general
rules,
which
include,
please
be
considerate,
polite
and
inclusive
of
all
your
fellow
members.
So,
as
mentioned,
please
be
inclusive,
polite
and
considerate
of
your
fellow
members.
A
Let's
go
ahead
and
fill
out.
I
will,
I
promise,
send
out
an
email
today
for
the
general
cloud
provider
meeting
next
week,
petitioning
for
everyone
to
attend
and
vote
on,
adding
a
new
chair.
Having
said
that,
I
don't
see
much
in
the
agenda.
Did
anyone
have
an
agenda
item
they
wanted
to
go
over
today.
B
C
B
The
pr
there
from
iago-
I
think
I,
since
this
was
merged
a
colleague
of
mine
and
I
were
trying
to
get
things
merged
over
to
kk,
to
use
this,
and
we
think
that
we're
running
into
sort
of
grpc
dependency
incompatibility
issues
that
would
cause
problems
there
and
we,
we
think
we're
at
the
point
where
we
need
like
a
more
complicated
branching
structure
in
connectivity
or
the
api
server
network
proxy
repo.
B
B
So
I
think
what
I'm
hoping
for
from
this
meeting
is:
who
are
the
folks
that
care
about
kk
getting
that
mitigation
and
how
far
back
might
they
care
about
it
being
reported?
Because
those
are
some
of
my
open
questions
for
for
what
I
want
to
propose.
B
A
B
What
I
want
to
mitigate
sort
of
a
memory
leak
problem
is
a
newer
version
of
connectivity,
client,
the
newer
version,
which
has
a
mitigation
from
that
pr
happens,
to
bring
in
a
grpc
dependency
transitively
at
sort
of
too
new
of
a
version
than
kk
is
expecting.
I
took
some
notes,
I
think
kk
master
uses,
grpc
1.40
connectivity,
0025
or
26
has
grpc
1.41,
so
we've
sort
of-
and
I
got
some
advice-
I
talked
about
this
with
jordan
leggett.
A
A
B
Interesting,
well,
the
that
might
that
might
allow
us
to
get
the
newer
client
into
oss.
You
know
master,
but
in
terms
of
back
porting,
if
you
go
further
back
than
1.23,
golang
is
116
at
those
earlier
versions.
So
we
may
also
run
into
an
issue
with
golang
versions.
A
B
That
sounds
great
to
me.
I
can
work
with
you
to
do
that.
Initially,
open
question,
though
stands
of.
Are
there
other
parties,
other
companies,
that
care
about
getting
this
mitigation
backported
in
kk
and
how
far
back.
A
A
So
it
would
certainly
be
interesting,
but
but
I
think
you
know
I
I
they
almost
certainly
care-
and
I
think
just
getting
this
dealt
with
in
the
short
term
in
the
connectivity
client
go.
Mod
file
is
probably
the
fastest
way
of
moving
forward.
B
A
Sure
and
feel
free
to
ping
me
after
the
meeting
I
can.
I
can
send
you
a
link
yep
thanks,
yep
cool.
Does
anyone
else
have
a
question
or
does
anyone
else
have
any
agenda
items
they
would
like
to
go
over.
A
All
right
I
have
I,
I
have
a
a
request
for
help.
A
We
have
one
outstanding
extraction
feature
that
no
one
is
working
on.
I
had
thought
that
we
had
a
volunteer
to
do
it
in
the
last
release,
but
they
never
actually
began
the
work
and
that
is
extending
the
cloud
controller
manager
framework
to
support
web
hooks.
So
we
have
a
cap
that
details
what
the
requirements
are,
but
someone
actually
needs
to
go
in
and
add
that
support
into
the
framework.
So
I
am
generally
just
going
to
ask
for
help
see
if
anyone
might
be
interested
in
doing
that.
Work.
C
Well
there
so
last
week
or
last
meeting,
I
said
that
I
was
going
to
go
and
ask
my
team.
If
anyone
was
interested
in
helping-
and
I
did
get
a
little
bit
of
interest,
so
I
will
follow
up
with
that
again
and
see
if
they're
still
interested.
It's
taken
me
a
little
while
because
we
just
had
a
manager
switch.
So
there's
a
little
bit
of
chaos.
But
let
me
follow
up
with
the
parties
who
expressed
some
interest
and
see
if
they
still
are,
and
I
will
let
you
know.
A
All
right
sounds
good.
You,
you
have
my
full
sympathy
on
the
manager
switch.
We
just
had
that
over
at
our
end
as
well.
So
I
know
exactly
how
that
goes.
C
Yeah
it's,
it
can
be
a
lot
of
fun
cool
I'll,
follow
up
with
you,
offline
and
yeah.
A
Joseph
is
on
the
the
former
one
doing
follow-up
on
their
side
and
sink
offline.
A
Awesome
all
right
did
anyone
else
have
any
items
they
wanted
to
discuss.
A
So
I
actually
have,
since,
since
we
have
joe
on
the
call-
and
I
I
I
saw
that
joe
had
done
some
interesting
work
in
the
api
server
recently,
I
actually
have
a
a
odd
outstanding
question
for
joe.
A
So
I
know-
and
I
think
the
answer
is
no,
but
I
want
to
confirm
because
I
I
did
think
it
was
some
really
interesting
stuff,
but
I
noticed
that
the
cube
api
server
and
the
api
machinery
team
have
added
some
very
interesting
api
expression
support
in
the
api
server.
A
I
mean
I
realized
only
alpha
right
now,
but
the
ability
to
basically
set
up
you
know
dynamic
bits
of
of
code
around
things,
and
I
was
wondering
if
there
was
any
way
we
might
be
able
to
actually
call
out
to
cloud
provider
specific
code
via
that
or
if
we're
still
having
the
the
same
sort
of
node
cloud
providers
are
no
doubt
goes
no
dynamic.
Library.
Support
is
still
going
to
prevent
us
from
being
able
to
do
anything
class
specific
in
that
api
expression
work
that
your
team
is
doing.
D
I
can
describe
it
first,
so
there
is
a
relatively
simple
but
still
quite
powerful,
expressed
language
called
cell
ceo.
It
sounds
like
common
expression,
language.
D
It
has
the
property
of
being
safe
to
execute
in
something
like
the
api
server.
It's
got
a
really
constrained
syntax,
it's
very
sandboxable.
It's
designed
ground
up
from
this,
like
security,
focused
people
thought
about
this
and
it
has
a
couple
other
properties.
One
is
that
you
can
type
check
it.
So,
unlike
a
lot
of
scripting
languages
that
you
could
throw
into
a
string
on
something
we
can
type
that
to
so
what
we've
done
in
1.23
is
we've
added
it
into
crds
for
the
purposes
of
validation.
D
You
could
say:
well,
I
you
can
accept
this
field
or
this
field,
but
not
both
or
you
know
you
could
all
you
know
you
could
do
simpler
ones
too,
because
open
api
already
has
some
really
basic
validation
rules,
but
you
can
do
things
like
cross
field
validation.
You
can
do
a
lot
of
conditional
checks
stuff
that
people
were
having
to
do
in
web.
Hooks
now
is
much
more
feasible
to
do
just
right
in
the
crd
itself
and
everything
self-contained.
D
You
don't
have
a
second
binary
running.
It's
really
nice
for
a
lot
of
properties.
I
don't
see
any
reason
why
you
couldn't
use
that
mechanism
in
other
places
we're
intending
to
use
it
in
other
places
already
so
we're
using
it
now
for
crd
validation,
but
crd
version
conversion
is
something
we'd
like
to
use
for
admission
control.
Is
something
we'd
like
to
use
it,
for
I
mean
being
able
to
check
if
you
know
a
a
label
with
a
particular
key
has
a
you
know,
set
of
expected
values.
D
Matched
is
a
really
simple
thing
to
do,
but
people
have
to
do
with
the
web
hook
today.
So
doing
that
with
a
simple
expression
seems
like
way:
cleaner
way,
more
efficient
and
way
easier
to
operate.
So
there's
a
bunch
of
things
you
can
use
it,
for
I
don't
see
any
I
mean
given
that
we're
already
using
it
in
the
api
server.
I
don't
see
a
lot
of
risk
to
use
in
other
places.
D
The
only
constraint
is
you
really
need
to
have
a
way
where
somebody
can
declare
through
the
api
what
they
want
done
so
like,
if,
if
you
have
some
kind
of
object
that
controls
part
of
your
cloud
provider,
some
api
object
or
api
resource
and
there's
some
field
that
could
be
set
to
something
that's
specific
to
your
cloud,
and
you
want
that
to
be
a
cell
expression,
and
that
would
be
useful.
I
don't
see
any
problem
with
that.
A
B
A
A
D
D
That
you
know,
people
would
naively
think
that
you
would
need
an
admission
controller
and
an
informer
and
a
bunch
of
other
stuff
for
that
you
might
actually
be
able
to
do
pretty
simply
but
yeah
there
are
equipments
like
if
you
can't
express
what
you're
trying
to
say
and
sell,
then
cell's
not
useful
for
you.