►
From YouTube: SIG Cloud Provider 2021-09-29
Description
Discuss Webhook support in CCM KEP.
Kubecon planning
Mixed protocol in service support for cloud providers.
A
Welcome
to
the
september
29
2021
sig
cloud
provider
meeting,
this
meeting
is
being
held
by
sick
cloud
provider
kubernetes
and
the
cncf,
and
we
adhere
to
all
of
the
standard
practices
of
the
above.
A
lot
of
that
is
going
to
boil
down
to
please
be
polite,
can
consider
it
inclusive
of
all
of
your
fellow
contributors.
A
I
guess
that
one
yep
all
right,
starting
with
the
sub
project
updates
or
the
up
project
updates.
Apparently
I
don't
think
I
see
anyone
from
alibaba
or
baidu
here.
Amazon
do
we
have
any
updates.
A
All
right,
I
did
notice
that
we're
changing
the
owners
slightly
for
the
aws
system.
B
Oh
yeah,
that
was
one
of
the
owner's
files
that
I
missed.
I
think
it
was
in.
I
guess
the
case
I
o
repo,
so
there
was
some
inactive
contributors,
so
we
removed
those-
and
I
put
myself
and
matthew
wong,
and
I
think
I
need
a
lgtm
from
you
walter
on
that
one
you've
gotten
it
already.
A
Oh
perfect,
no
worries
anything
else.
A
Well,
azure.
C
Just
I
dropped
a
link
into
the
one
thing
that
I
came
to
chat
about,
which
was
this
is
actually
a
good
idea.
A
good
example
of
this
is
already
fixed
out
of
tree
and
reporting
the
fix
back
to
entry
yeah,
just
cherry
picking
in
the
fix
need
to
get
that
approved
emerged.
I
think
we
were
looking
for
andrew,
but
I
suppose
anyone
could
take
a
look
and
see
if
it
works
for
them
to
bring
that
to
people's
attention.
Yeah.
A
A
Slack
is
you,
can
you
can
either
sign
it
to
me
in
the
bug
itself?
Chef
taco
or
you
can
ping
me
on
slack
either
one
works.
Thank
you.
Yup
no
worries,
gcp.
D
I'm
trying
to
think
what
to
add
here
we
did
bump
to
what
is
it
121
of
open
source
kubernetes,
and
there
are
some
networking
changes
going
in.
E
A
No
worries:
do
we
have
anyone
for
openstack,
I
think
so.
F
Vsphere
yeah.
Sorry,
I'm
not
prepared
to
address
that
myself,
because
I
thought
andrew
would
be
here
and
not
going
to
tell
you
more
than
I
know
so.
I'm
going
to
have
to
not
provide
an
update
on
that.
A
No
worries
so
extraction
migration.
I
think
we
have
and
we're
starting
to
build
consensus
around
the
lkg
did
either
joe
or
kermit
have
anything
they
wanted
to
add.
D
The
design
is
available,
I
think,
there's
a
link
further
down
so
yeah
anybody
that
doesn't
know
what
alaska
that's.
No
good
testing
is
or
does
know,
but
wanted
to
provide
us
input
or
thinks
they
might
want
to
consume.
It
recommend
reading
the
doc
add
comments
there.
I
think
that's
the
stage
we're
at
just
looking
for
feedback.
A
Yeah,
the
other
thing
I
will
make
a
general
comment
on.
I
should
probably
sorry
I'm
talking
and
not
typing,
I'm
not
good
at
doing
both
at
once.
We
have
one
kept
for
this
coming
release
and
very
little
time
left
we're
hoping
to
start
adding
web
hook,
support
to
the
ccm
that's
what
the
predominantly
the
cap
is
about.
A
Persistent
volume
claims
support
it,
so
the
kep
is,
is
approved
for
alpha
and
there's
a
fair
amount
of
detail
on
the
work
that
is
available.
So
if
anyone
is
interested
in
doing
in
getting
started,
please
reach
out
to
me,
I
am
happy
to
help.
I'm
also
happy
to
I'm
also
hoping
to
do
some
of
that
work,
but
there's
quite
a
bit
of
work
to
do.
A
I
think
we
can
break
it
into
chunks
and
I
think
it
is
a
great
project
getting
started
if
anyone
is
interested,
and
obviously
that
is
one
of
the
things
that
we
need
to
do
to
finish
up
the
extraction,
migration
work
and
thank
you,
michael
for
the
typing,
also
just
fyi.
It's
taco
with
a
k.
A
Thanks
sorry
for
that
agenda
steve,
I
believe
you're
up.
F
Okay,
so
we're
planning
a
maintainer
track
session
at
kubecon
coming
up
in
two
weeks,
walter
nicholas
and
I
are
planning
on
being
there
live,
but
we
have
a
good
sized
portion
of
that
session
set
aside
for
what
we're
calling
cloud
provider
lightning
talks,
so
these
would
be
updates
from
the
individual
specific
cloud
providers,
with
the
expectation
that
this
would
be
a
three-minute
summary
of
where
you've
been
in
the
past
in
months.
F
What
state
you're
at
now
and
what
the
road
map
looks
like
going
forward,
particularly
with
regard
to
migration
out
of
tree.
F
What
we'd
like
to
do
is
get
that
lightning
talk
submitted
in
the
form
of
an
mp4
video
high
def
16x9
would
be
great
if
possible,
and
then
I
will,
assuming
I
get
them
at
least
a
few
days
before
I
should
be
able
to
edit
them
together,
so
that
we
have
minimal
disruption
just
hit
play
one
time
and
let
them
roll
rather
than
digging
up
individual
files
from
the
stage
when
we're
up
there
live.
I
think
it
will
be
better
if
we
get
these
in
advance.
F
The
other
thing
I'd
like
to
get
associated
with
that.
If
in
your
video,
you
show
slides,
give
me
the
slides,
because
the
conference
expects
that
a
pdf
of
the
slide
deck
gets
published
and
I'll
cut
and
paste
whatever
slides
you
used
into
the
main
deck
for
the
session,
so
that
an
attendee
who
downloads
it
later
we'll
have.
You
know
the
appropriate
experience
of
being
able
to
look
up
anything
they
saw
for
reference
later.
G
A
A
The
broadband
at
the
conference
is
terrible,
and
you
know
we
don't
having
all
of
this
pre-built
and
ready
to
play
is
going
to
make
your
cloud
provider
look
better
because
you're
not
going
to
be
sitting
there
hanging
on
a
broadband
and
making
it
look
like
your
cloud
provider
is
slow,
so
strongly
recommend
getting
steve
those
those
pieces
and
and
not
trying
to
do
anything
with
your
actual
cloud
provider
live
at
the
conference
where
possible.
It
would
also
be
nice.
A
F
Yeah,
the
other
thing
is
now
that
walter
brings
it
up.
I
think,
even
if
you
are
intending
and
gonna
be
one
of
the
moderators
like
me
for
the
vmware
cloud
provider,
that's
going
to
be
recorded
because
with
covid
going
on,
you
never
know
what
might
happen.
H
Hi,
my
name
is
peter:
I'm
maintainer
of
the
k,
ops
project,
so
we
have
a
lot
of
end-to-end
testing
in
place
and
we're
a
lot
of.
H
H
The
api
server
still
relies
on
the
cloud
config
for
the
pv
admission
plug-in,
and
I
believe
this
directly
relates
to
the
cap
that
was
mentioned
earlier,
and
so
what
happens?
Is
the
entry
code
doesn't
recognize
the
new
field
and
it
fails
the
admission
entirely,
and
so
I'm
wondering
how
we
should
handle
this.
H
H
Have
separate
cloud
config
files
for
entry
versus
out
of
tree?
I
also
suggested
you
know,
depending
on
the
timeline
again
whether
it
makes
sense
to
keep
the
schemas
relatively
aligned
for
entry
until
we
can
drop
the
entry
code.
But
I'm
curious
what
other
people
think
about
this.
B
Yeah
no
worries
well,
like
I
suggested
here,
you
know
in
the
in
the
cops
test.
What
we
could
do
for
now
is
just
have
a
separate
cloud.
Config
for
you
know,
have
one
for
api
server
and
other
entry
components
and
then
then
use
the
old
schema
and
then
for
the
external
ccm
we
can.
B
We
can
have
us,
you
know
its
own,
its
own
cloud
config
and
then,
as
soon
as
we
have
the
the
admission
controller
broken
out
and
running
as
a
web
hook
as
part
of
the
ccm,
then
we
don't
need
to
pass
cloud
config
to
the
api
server
anymore
yeah,
and
I
agree
that
there's
the
other
option,
which
is
to
keep
the
entry
cloud
config
in
sync,
but
I
think
we
that
sort
of
goes
against
the
current
policy
of
not
merging
features
entry,
and
you
know
these
will
be.
B
You
know,
changes
to
the
cloud
config
would
probably
be
considered
features.
So
as
long
as
there
is
a
workaround,
I
have
a
feeling
that
walter
and
andrew
and
others
will
prefer
it
prefer
that
we
use
that.
But
I
don't
know
what
do
you
think
walter?
Do
you
agree
with
that
assessment.
A
B
So
let
me
basically
in
the
test,
there's
and
and
peter
please
correct
me
if
I'm
wrong,
there's
one
cloud
config
file
right
and
it's
shared
by
all
of
the
components,
and
we
have
a
new
feature
in
the
external
ccm
which
uses
a
new
field
in
the
cloud
config,
and
so
that
cloud
config
is
also
passed
to
the
api
server
which,
because
of
the
persistent
volume
admission
controller,
is,
is
also
reading
the
cloud
config.
B
You
know
it
shouldn't
be,
but
it
is
because
of
the
the
pv
library
or
whatever
it
is
yeah
yeah.
So
you
know
the
workaround
is
really
to
to
use.
So
thank.
A
You
to
really
the
kcm,
then
am
I
missing
something
because
it's
I'm
sorry
they're,
just
I'm
trying
to
put
all
the
pieces
together
in
my
head.
I
would
imagine
and
correct
me
if
I'm
wrong
the
cube
api
server
is
ignoring
the
cloud
config
file,
so
it's
probably
only
the
case
cm
that
is
paying
any
attention
to
the
cloud
config
file
and
it's
doing
it
specifically
for
the
persistent
volume
controller.
H
G
A
Yeah,
I
think
then
I
have
to
agree
with
your
assessment.
We
can.
We
can
double
check
with
andrew
he's,
usually
the
most
hard
line
on
this
and
you've
got
to
get
both
of
us
to
agree,
and
I
don't
think
I
andrew
well,
so
I
think
you're
right
if
there
is
a
workaround
yeah
okay,
I
mean
this
is
obviously
also
going
to
block
the
the
credential
provider
featuregate,
so
yeah.
I
think
we
just,
I
think,
you're
absolutely
right
on
your
initial
assessment
peter.
A
I
mean,
arguably
and
again,
I'm
gonna
throw
throw
nick
under
the
bus.
Sorry,
nick
not
your
day,
you
at
one
point
nick,
I
thought
were
considering
the
idea
of
actually
fixing
this
with
a
con
with
your
own
controller
rather
than
a
web
hook.
B
I
I
my
memory
is
that
that
was
proposed
by
somebody
else
at
aws
and
I
don't
I
don't
know
fully
the
reasoning.
So
my
memory.
A
Of
the
reasoning
was
that
if
we
make
it
a
web
hook,
we
always
have
the
standard
web
hook
problem
that
if
the
web
hook
isn't
functioning
that
you
know
you
start
having
problems
where
either
the
mutation
isn't
applied
or
where
you
end
up
failing
the
right,
and
so
whoever
I
thought
was
that
amazon
was
preferring
a
controller
route
because
it
ends
up.
I
mean
it's,
it's
async,
so
you
have
some
async
problems,
but
you
you,
don't
you
don't
end
up?
You
know
either
failing
the
right
or
failing
to
do
the
mutation.
B
Yeah
I
mean
the
the
web
hook.
Issues
are
obviously
annoying.
I've
had
plenty
of
those.
You
know
they
cause
on-call
pain
operations,
pain
for
us,
but
if
there's
like
a
race
condition
which
I
thought
there
was
and
that's
the
whole
reason
that
I
mean
the
controller
approach
like
I,
you
know-
I
don't
know
the
details
of
this,
but
my
understanding
is
the
controller
approach
was
initially
the
proposed
replacement
and
then
it
was
rejected
because
of
some
race
condition.
A
Yeah
I
I
know
that
on
both
side,
ali
and
michelle
al,
both
googlers
on
this
but
are
working
principally
on
the
storage
side-
were
not
happy
with
it
as
as
the
gcp
solution,
and
so
they
were
looking
for.
A
web
hook
got
it,
but
it
is,
I
mean
at
some
level
we'll
we
will.
A
You
know
for
us
for
that
to
happen.
We
push
through
the
cap.
That
support
should
be
available
to
everyone,
but
doesn't
mean
that
everyone
has
to
go
with
that
solution.
But
if
you
prefer,
then
I
think
that
just
increases
the
import
of
getting
this
cap
into
alpha.
B
Yeah,
well
I
mean
maybe
I
mean
I'm
assuming
that
aws
does
not
differ
from
gcp.
So
if,
if
we
have
the
same
race
condition,
then
we
would
need
to
do
the
same
thing,
but
maybe
maybe
I
can
try
to
figure
out
who
was
was
going
for
the
the
controller
approach
and
see
if
they
know
more
than
I
do.
A
Sounds
good
all
right!
Thank
you,
nick
all,
right,
if
we're
done
there,
bridget.
C
Yeah
I
added
another
issue
because
I
thought
with
this
community
here.
This
would
be
a
good
crowd
to
talk
to
about
there's.
There
was
some
discussion
both
in
the
issue
and
at
the
club
provider.
Azure
meeting
andrew
brought
up
this
idea
of
what,
if
we
move
this
to
beta
and
1.23,
because
you
know
we
haven't
hit
the
exception
deadline,
yet
the
thing
about
moving
it
to
beta
is,
of
course,
as
we
all
know,
beta
features
around
by
default
and
then
there's
some
back
and
forth
there
about.
Well
what
what
do
we
want?
C
The
experience
for
users
to
be,
but
at
the
same
time
this
has
been.
You
know,
january
2020.
This
has
been
going
on
for
a
while.
So
are
we
generally
in
the
community
in
favor
of
moving
this
to
beta
or
yeah?
Basically,
I
was
just
seeing
what
people
thought
I'm
fine
with
filing
the
paperwork.
I
just
don't
want
to
cause
unpleasant
surprises
for
other
clouds.
C
It
looks
like
tim
hawkin
is
not
against
it
and
I'm
not
sure
if
our
aws
friends
have
weighed
in
yeah,
I
feel
like
I'm,
I'm
the
caller
first,
you
know
first
time
caller
I'll,
take
my
answer
offline,
I'm
not
asking
for
an
answer
right.
The
second
I'm
just
kind
of
asking
people
to
look
at
and
consider
this
maybe
put
their
two
cents
two
cents
in
on
the
issue.
A
There's
something
for
me:
yeah
go
ahead.
C
A
A
So
I
think
from
the
cloud
provider's
perspective,
I
would
like
to
get
each
of
the
major
cloud
providers
to
say
they
don't
have
a
problem
with
it.
If
you
could,
if
you've
gotten,
bowie
and
tim
to
say
yes,
then
gcp
is
obviously
okay
with
it.
A
If
you
know
from
there,
I'm
assuming
that,
since
azure
is
pushing
for
this-
that
your
azure
is
okay
with
it
from
there,
it
would
be
nice
to
if
we
could
get
sort
of
someone
from
the
ibm
red
hat
camp
to
say:
okay,
get
amazon
to
say,
okay
and
get
vmware
to
say,
okay,
and
I
think
if
that
group
is
willing
to
say
no
problem,
then
cl
the
sig
cloud
provider
has
no
objection.
A
Now,
having
said
that,
you
know
I
would
suggest
nick
I'm
guessing
you're,
not
the
best
person
from
amazon.
Is
this
more
a
key
short
thing.
B
Yeah,
this
is
definitely
kishore's
preview.
A
So
yeah
sorry,
I'm
just
going
to
try
and
find
you
the
right
person
to
chat
with
bridget
stephen
who's.
The
best
person
on
the
vmware
side.
A
Okay
yeah,
so
if
you
can
get
andrew
to
sign
up
for
vmware
and
el
mico,
do
you
know
who
the
right
person
started
for
the
red
hat?
Ibm
side
would
be
yeah
that'd,
be
richard
tyce,
okay,
so
sorry
to
give
you
homework
bridgette.
But
those
are
your
three
names.
If
you
can
get
those
three
people
to
sign
off,
then
I
think
sig
cloud
provider
has
signed
off
for
you.
C
Awesome
I
appreciate
it
and
this
that
homework
is
exactly
what
I
came
to
this
meeting
to
ask
about.
In
this
context,.
A
Then
I
think
we're
going
to
go
to
the
major
bugs
is
yeah
peter's
still
here.
So
one
thing
I
will
mention
peter
step
one.
I
apparently
click
press
the
wrong
link.
I
gotta
fix
the
bugs
club.
A
Should
probably
fix
the
bug
triage
to
cover
cloud
provider,
but
until
I
do
you're,
probably
better
off
filing
against
kk
than
the
cloud
provider
repo.
A
All
right,
then,
I
am
going
to
go
ahead
and
signaled
this
one.
A
Not
sure
it's
entirely
cubelet
note
problem
detectors
usually
run
as
its
own
service
and
then
there's
the
underlying
vm
image,
but
all
of
that
still
tends
to
be
owned
by
node,
okay,
cool.
C
Is
this
I
feel
like
I
gotta,
look
and
see
if
this
is
the
one
I
thought
it
was.
Is
this
the
one
that
I
was
asking
about
yeah?
This
is
the
one
that
we
mentioned
that
we
talked
about
earlier,
which
is
to
say
there
is
a
pull
request
to
fix
it,
that
we
need
to
that.
I
attached
you
to
okay.
If
you
click
on
the
pull
request
to
fix
it.
A
So,
in
fact,
what
I'm
going
to
suggest
then
is
they've
already
got
that
we've.
If
we've
already
got
a
fix
and
that's
who's
fixing
it,
then
I'm
just
going
to
assign
the
bug
to
them.
Wonderful.
C
And
yeah
I
assigned
you
on
that
that
pull
request,
because
you
said
that
you
needed
to
oh.
G
A
I
will
that
that
will
probably
happen
tonight.
Wonderful,
thank
you
all
right
and
joe
you're
already
assigned
on
this
one.
So
I'm
assuming
that
we're
the
right
things
are
happening
and
we're
just
waiting
for
bowie.
A
Awesome,
I
think
that
is
everything
we
have
so,
if
everyone's
good
with
it,
I
can
give
everyone
20
minutes
back.
C
A
We
are
not
having
well,
I
will
put
it
slightly
differently.
We
might
have
this
meeting,
but
if
we're
having
this
meeting,
it's
going
to
be
outdoors
with
the
people
who
are
attending
cubecon.
C
A
That
is
a
very
good
reminder.
Thank
you,
but
on
that
topic
I
will
actually
suggest
that
for
those
of
us
who
are
actually
attending,
I
would
actually
like
to
gather
in
person
it's
kind
of
hard
to
remember
what
cubecom
was
like
in
the
pre-covet
days.
But
this
is
where
we
build
our
community
and
building
trust
and
learning
to
work
with
each
other
is
integral
to
making
all
of
this
work.
So
I
I
would
very
much
like
to
get
everyone
together.