►
From YouTube: SIG Cloud Provider 2022-06-22
Description
Agenda: https://docs.google.com/document/d/1OZE-ub-v6B8y-GuaWejL-vU_f9jsjBbrim4LtTfxssw/
Last call for enhancements freeze for 1.25
LKG - not an enhancement, but walter needs to write an email for sig release
Cloud provider flag https://github.com/kubernetes/kubernetes/pull/110066
Node Addresses flapping https://github.com/kubernetes/kubernetes/issues/109793
[elmiko] curious about 1.24 bump for cloud-provider-gcp
[bridgetkromhout] need approvals for cherry-picked bugfix backports of https://github.com/kubernetes/kubernetes/pull/107631
A
A
Welcome
everybody
to
Sig
cloud
provider,
it
is
June
22nd
today
go
ahead
and
add
yourself
to
the
attending
list
and
we'll
go
ahead
and
get
started
actually
to
kick
it
off.
I
will
give
everyone
a
minute
to
update
sub-projects
and
then
I
will,
as
I
go
through
the
modules
stop
with
stop
at
the
ones
that
have
something
to
to
update.
A
A
Thanks
all
right
so
for
provider
AWS
we
have
another
round
of
patch
releases
for
CCM
images,
I,
don't
think,
there's
any
features
necessarily
to
call
out.
We
did
have
a
tagging
controller,
which
is
useful
for
cost
allocation
and
AWS,
and
that
was
I
think
the
primary
reason
for
the
the
new
patch
releases,
but
other
than
that
I,
don't
think
anything
to
call
out
azure.
A
A
A
A
D
A
A
Yeah,
so
there
is
I
have
that
on
the
agenda.
There's
a
there's
a
issue
and
pull
requests
that
we
need
to
get
merged,
I
think
as
long
as
it
gets
merged,
I'd
like
it
to
get
get
merged
as
soon
as
possible,
because
that's
helpful
for
me,
but
I
think
as
long
as
it
gets
merged
in
this
release
cycle
we're
good
but
yeah.
It's.
D
A
Well,
I
mean
there
is
I
like
there
isn't
opportunity
for
someone
to
pick
up
a
more
substantial
change
here
so
but
I
like
I,
don't
know
that
anybody
has
the
bandwidth
to
do
it.
But
if
so,
I
can
kind
of
explain
what
somebody
could
look
at
in
terms
of
a
more
substantial
change
like.
Basically,
the
problem
is
that
cubelet
and
CCM,
so
this
fix
that
we
have
which
we
can
get
into
in
a
few
minutes.
A
But
the
root
of
the
problem
is
that
they
both
think
that
they
should
be
updating
the
node
IPS.
So
I
think
that
there
could
be
a
fix
that
prevents
them
from
both
trying
to
update
the
the
node
IPS
at
the
same
time.
So
a
better
separation
between
like
a
better
handoff
I
guess
but
yeah.
That
would
be
a
much
more
substantial
change
and
would
absolutely
require
a
cap
so
based
on
how
much
effort
that
would
be
and
how
much
time
we
have
I
don't
think
it's
possible.
B
A
C
Yeah,
like,
like
total
plus
one
to
what
Bridget
is
saying
like
this
problem,
seems
like
we
need
an
explicit
way
to
solve
it
and
it
seems
like
that's
not
going
to
happen
without
a
cap
or
something
because
we,
like
the
discussion,
keeps
going
back
and
forth
in
the
various
PR's
and
everything,
and
it's
like
having
some
implicit
way
to
learn.
This
just
seems
like
it's
not
going
to
cut
it.
You
know.
D
I
don't
know
if
we
can
solve
this
with
something
as
as
simple
as
a
as
a
well-defined
set
of
labels
of
who's
responsible,
but
yeah
I,
plus
one
with
what
Bridget
says.
I
think
we
don't
have
and
what
Nick's
saying
we
don't
have
enough
we're
not
going
to
get
a
a
cap
done
by
Thursday,
but
that
just
means
that
we
need
to
start
the
cap
a
lot
earlier,
which
means
we
probably
should
be
starting
to
discuss
it
now
for
126.
D
A
I
agree
yeah,
so
any
other
caps
that
that
anybody
knows
of
that
need
an
update
like
I.
Don't
think
we're
super
worried
about
like
the,
for
example,
the
credential
exec.
What
was
the
let's
see,
I?
Guess
the
the
cubelet
image
credential
provider
cap
win
beta
last
release
or
two
releases
ago?
Are
we?
Are
we
Walter?
Maybe
you
know
the
answer
to
this?
Are
we
concerned
with
taking
that
GA
at
all,
or
are
we
content
with
it
in
beta.
E
D
A
B
A
B
I
think
that's
1.25
and
Beyond
I.
Don't
think
that
hit
us
for
1.24,
because
yeah
the
the
other
thing
to
consider
is
maybe
take
a
look
at
all
of
our
cups
for
1.25
and
make
sure
that
our
test
plan
looks
good
and
our
docs
look
good,
because
stuff
was
getting
bounced
out
of
the
one.24
release
because
it
wasn't
really
set
in
terms
of
having
all
that
stuff
squared
away.
So,
even
though
you
think
something
is
in
like
go
make
sure
it's
in
oh.
B
B
When
you
think
oh
yeah,
our
feature
is
in
there
it's
it's
slated
to
go
into
the
release,
it'll
be
fine,
make
sure
it
has
test
Planet,
docs
or
it
may
be
bounced
out
of
the
release
for
not
having
taxable
and
tax.
Basically,
kubernetes
is
growing
up
and
getting
very
serious
about
some
of
this
stuff.
So
if
you
haven't
been
paying
attention
to
cigarettes
like
that's
the
tldr,
not
that
I'm
in
secret,
loose
I,
just
kind
of
get
it
by
osmosis.
A
D
But
I
do
need
I
I
am
I,
am
I,
am
more
than
I
want
to
admit
behind
on
writing
that
email
getting
it
approved
by
this
group
in
preparation
for
sending
it
to
Sig
release,
but
the
good
news,
such
as
it
is,
is
there's
not
nothing
kept
related
on
that
one.
We
we
do,
however,
need
to
spend
some
time
talking
to
Sig
release,
but
I
need
to
actually
as
sort
of
related
to
what.
A
Okay,
I'll
just
put
a
note
that
it's
something
that
we
need
to
do
I
mean
I.
Imagine
I,
don't
know
that.
There's
like
I,
don't
know
what
the
the
actual
deadlines
are,
but
I
imagine
we
want
to
make
significant
progress
on
it
in
the
125
cycle,
so
I'll
just
throw
it
in
there
cool
all
right,
so
yeah.
So
the
next
one
is
this.
Is
me
cloud
provider
flag?
A
So
basically
there
was
an
issue
where
somebody
was
complaining
because
the
cloud
provider
flag
was
marked
as
deprecated,
but
cloud
provider
equals
external,
is
a
legitimate
value
for
it
and
is
necessary
and
and
I
think
that
the
notice
to
deprecate
it
even
said
something
about
it
being
deleted
in
124.,
and
so
this
person
was
saying
like
there's
no
way
to
get
the
cluster
working
without
using
this
deprecated
flag.
So
they
were
trying
to
do
it
without
using
the
flag
anyway.
A
So
I
started
out
with
just
removing
the
deprecation
notice,
but
I
had
a
comment
from
someone
in
I
think
signode
that
the
API
is
no
longer
like
the
the
the
way
you
know
the
cloud
provider
flag
with
multiple
options
is
no
longer
really
necessary.
So
maybe
we
should
consider
cubicle
configuration
option.
A
That's
a
Boolean,
so
I
initially
said
I
am
not
sure
if
it's
worth
making
that
change,
because
it'll
force
people
to
change
their
tubular
Flags,
but
I
ended
up
going
ahead
and
like
just
updating
the
the
pr
to
include
that
change.
So
anyway,
I
haven't
I,
don't
think
I've
gotten
much
traction
on
it
since
that
yeah.
A
So
maybe,
if
people
from
say
cloud
provider
could
review
it
and
if
we
agree
on
it,
you
know
either
go
in
with
the
cube
of
config
option
or
just
leaving
the
flag
as
is
and
sort
of
changing,
or
you
know,
undeprecating
it
yeah.
Let's
decide
that
and
then
I'll
ping
Sig
node
for
review
again
yeah
go
ahead.
Walter.
D
So
I
I
sort
of
feel
like
deprecating.
It
is
the
right
thing
to
do.
What
would
you
think
about
doing
something
like
making
it
so
that
if
the
cloud
provider
flag
I
mean
this
I
think
there's
multiple
stages?
But
if
we
may
begin
supporting
the
cloud
provider
not
being
explicitly
there
that
it
automatically
switched
to
the
external
behavior
in
preparation
for
removing
it,
because
eventually
we
shouldn't
have
a
cloud
provider
flag
in
the
KCM,
the
cubelet
or
the
cube
API
server,
yeah
right.
A
D
A
Right
there
is
a
use
case
for
a
flag
in
cubelet
which
is
sort
of
the
reason
for
this
external
provider
field
configuration
option,
and
that
is
the
decision
as
to
whether
or
not
to
apply
the
taints
for
for
the
external
Cloud
providers.
So
in
some
cases
people
don't
want
those,
they
don't
want
Cuba
to
get
the
the
tanks
that
prevent
you
know.
A
So
the
idea
is
that
cubelet
is
supposed
to
apply
those
taints
on
node
startup
and
then
the
CCM
is
supposed
to
remove
them
once
like
the
node
addresses
are
initialized
and
all
that.
But
not
everybody
wants
that
to
happen.
So
you
know
I
guess
you
need
to.
You
need
some
kind
of
a
Boolean
flag
and
cubelet.
That
says
this
is
managed
by
an
external
cloud
provider
and
I
do
want
to
apply
those
tanks
or
you
know
this
is
some
other
kind
of
kubernetes
cluster
and
I.
Don't
want
them.
D
A
A
A
D
A
A
Yeah,
obviously
we
can't
so
that
was
one
thing
that
I
was
sort
of
wondering
about,
because
you
know
you
have
to
think
about
like
the
behavior
of
okay.
If
you
know
everybody's
gonna
have
cloud
provider
equals
external
set
in
their
Flags
and
then
this
new
configuration
option
comes
around,
so
we
have
to
think
about
like
okay,
so
new
configuration
option
is
unset
and
cloud
provider
is
set
to
external.
What
is
the
behavior?
A
A
B
A
B
A
A
A
A
A
But
yeah
I
think
as
of
right
now,
like
I,
think
the
well
myself
Iko
I,
don't
know
that
Andrew
ever
looked
at
it.
We're
happy
with
it
and
I.
Think
we're
just
waiting
on
signode
to
to
take
a
look
and
I
think
they
did
take
an
initial
look,
but
we're
just
waiting
for
a
for
a
final
approval.
So
yeah
like
with
this
absolutely
has
to
get
merged
before
125
I
know
we
have
some
time,
but
the
sooner
the
better.
C
Yeah
this
was
just
a
quick
one.
You
know
one
of
our
Engineers
was
asking
about.
He.
Mike
fedossen
had
put
up
a
PR
to
pump
the
the
dependencies
there
in
the
cloud
provider
gcp
to
1.24
and
I.
Guess
there
was
a
little
bit
of
back
and
forth
about
getting
it.
Working
and
everything
I
didn't
know
Walter.
Do
you
I'm
just
curious?
Do
you
know
if
there's
any
kind
of
progress
on.
D
In
addition
to
upgrading
like
making
sure
that
you
know
certain
feature
date,
settings
are
ported
over,
so
it's
not
always
a
straight
and
so
I
CC
doesn't
seem
to
come
to
these
meetings
anymore
or
just
Joe,
but
Cece
and
Joe
actually
wrote
a
fairly
detailed
document
on
the
best
way
of
doing
this,
and
so
I
think
Jacob
Jacob
is
going
through
that
it's
a
little
bit
more
work,
but
it's
more
likely
to
actually
get
all
the
tests
passing.
So
that's
that's
what
he's
doing
and
thank
you
very
much
for
that
work.
C
C
E
I
would
check
you
if
we
can
open
source
the
document
and
if
there
will
be
no
blockers,
I
will
put
this
as
empty
document
actual
plan
to
change
the
process
for
bumping,
because
it
happens
that
ever
released.
We
have
this
problem
with
pumping
the
diversion,
so
this
is
something
that
I
want
to
change
in
the
future,
but
this
is
more
in
the
scope
of
next
quarter.
Rich
Work,
something
like
this.
B
Awesome
yeah,
so
this
is
just
some
cherry
picks
that
they
were
a
little
delayed
because
the
original
PR
I
believe
it
was
before
he
went
out
on
paternity
leave.
Andrew
had
asked
for
a
little
more
detail
on
the
detail
that
added
the
Cherry
picks
then
languished
and
wondering
if
somebody
maybe
we're
waiting
for
him
to
get
back
from
dirty
or
maybe
somebody
else
can
take
a
look
at
the
back
ports
of
those
trade
picks.