►
From YouTube: SIG Cloud Provider 2021-10-27
Description
Plan to accelerate Leader Migration KEP
Review of KubeCon
Review cloud-provider-gcp/282
Release of AWS 2.3 with LoadBalancerController
Need to look at release versioning policy
A
Welcome
to
the
october
27th
2021
sig
cloud
provider
meeting.
As
always,
this
meeting
is
part
of
cncf
kubernetes
and
sig
cloud
provider,
and
we
have
some
codes
of
conduct.
Those
primarily
boil
down
to
please
be
respectful,
polite
and
inclusive
of
all
your
fellow
members,
contributors,
etc.
So
you
know,
let's,
let's
think
of
others,
and
with
that
I'm
going
to
go
ahead
and
start.
Let
me
share
the
agenda.
B
All
right,
so
thank
you
for
saying:
okay,
okay,
everybody
help
me
okay,
so
all
right,
so
the
so.
This
item
is
actually
a
follow-up
from
last
extraction
meeting
so
like
for
the
folks
who
don't
understand
what
dynamic
machine
before
so
leader.
Migration
is
a
mechanism
that
allows
sj
cluster
to
upgrade
from
one
set
of
leader
election
log,
two
controllers
to
another.
B
So
what
we
have
discussed
last
night
is
that
we
still
we
will
need
a
little
bit
more
than
what
we
had
last
time
to
make
it
a
j
feature.
So
in
summary,
we
are
going
to
change
the
graduation
criteria
to
like
the
two
following
items.
First,
one
is
to
test
that
end
to
end
in
oss
in
sa
control
plan
setting,
which
means
the
easy
case,
which
is
like
one
single
node
upgrading
from
kcm
to
case
and
plastician
is
not
going
to
be
enough.
B
So
my
implementation
plan
is
that
I'm
going
to
add
external
system
support
to
chaos
which
already,
which
the
current
state
I
believe
is
that
keops
have
support
for
all
providers
of
gcp.
So
I'm
going
to
just
add
the
only
remaining
one.
Then
I'm
going
to
contribute
the
support
for
freedom,
immigration
to
kelps.
B
So
the
second
one
is
like
we
are
going
to
test
that
with
at
least
two
cloud
providers,
but
we
don't
really
need
the
each
cloud
provider
to
provide
the
the
whole
test
suite
we
just
need.
We
just
need
the
signals,
like
somebody,
somebody
from
one
cloud
partner
say:
oh,
this
feature
works
for
hours
for
our
cloud
providers,
and
that
should
be
good
enough
and
we
need
two
cloud
providers
at
least
so.
This
is
my
proposed
change.
Any
ideas
about
that.
A
C
A
Awesome
and
then
I
think,
the
two
things
worth
iterating
on.
We,
we
didn't
finalize
the
details
and
I
think
we
agreed
not
to
finalize
the
details
in
a
meeting,
especially
since
the
approvals
need
to
happen
in
a
cap
anyway,
but
the
final
details
on
how
we
can
actually
put
in
automate
what
what
would
amount
to
automated
sufficient
automated
testing
of
the
aha
migration.
D
Yeah.
I
like
that.
I
like
the
chaops
idea
that
seems
like
a
like.
It
seems
a
lot
more
productive
than
trying
to
do
something
in
cuba.
So
and
you
know,
because
it
adds
long-term
value,
I
think
in
a
direction
we
want
to
go
anyways
plus,
I
think
it's
a
much
more
tractable
work.
I
mean
jared
you'll
you'll
have
to
make
a
judgment
call
on
that,
but
we
also
have
experts
that
could
probably
contribute
to
that.
So
you
wouldn't
necessarily
have
to
like
do
everything
yourself.
So
thank
you
for
finding
that
option.
D
E
F
D
A
A
C
A
A
I
think
it
went
pretty
well,
we
did
also
get
some
additional
slides,
so
I
think
ibm
gave
us
a
slide
right
at
the
last
moment
and
you
know
not
not
to
jab
you
too
much
there,
el
miko
and
we
also
got
a
a
a
a
fairly
late
edition
from
the
microsoft
thanking
actually
two
of
the
primary
contributors
we
have
from
microsoft,
china,
so
props
to
both
of
them.
That
was
definitely
nice
to
see
them
getting
called
out
for
their
good
work.
E
A
C
A
Yep
I
mean
it
was
interesting,
it
was
sort
of
interesting
I
mean
I,
I
think
all
of
the
all
of
the
snippets
for
all
the
cloud
providers
were
good.
It
is
an
interesting
thing.
You
give
everyone
a
one
sentence,
topic
of
what
you
want
them
to
talk
about,
and
the
different
ways
of
interpreting
that
one
sentence
is
sort
of
interesting.
A
C
I
mean
it
wasn't
necessarily
on
your
plate,
but
I
had
asked
your
opinion
last
week
and
you
you
know
you
seem
to
concur
with
the
general
thinking
there.
I
did
give
you
a
shout
out
in
the
comments,
but
I
figured
I'd
just
bring
it
up
again.
Maybe
if
you
or
or
someone
you
know
joe
or
someone
could
take
a
look
at
it,
I
don't
know
I
apparently
am
not
assigned.
Let
me
fix
that.
Let
me
just
I
should
have.
I
should
have
assigned
you.
I
just
kind
of
you
know.
A
A
C
Yeah
and
we
were
getting
some
like
weird
user
reports
of
there
being
like
like
permission
issues
or
something
like
this,
the
ccm
the
way
we're
deploying
the
ccm
it
looked
like
yeah
that,
like
you
know
it
in
some
configurations
on
gcp
like
you,
could
be
denied
access
to
the
alpha
sdk
endpoint,
or
something
like
that.
Based
on
your
permissions,
or
I
didn't
quite
understand
what
the
deep
level
of
that
was.
C
C
A
C
F
F
A
That
is
actually
pretty
important
and
also
means.
I
really
hope
that
gets
out
of
alpha
me
too.
All
right,
I
may
have
to
go
drop
a
note
to
mr
hawkin
as
he.
I
would
hope
he's
already
aware
of
this.
As
he's
both
big
networking
chair
and
the
dual
stack
lead.
But
let
me
let
me
reach
out
thank
you
for
the
the
heads
up,
bridget,
alibaba
and
baidu,
I
believe,
are
not
present.
Amazon
nick
keisha.
G
From
my
side,
so
we
released
the
version
2.3
of
the
load
balancer
controller,
which
is
a
minor
release
with
some
features
like
we
have
started
the
ipv6
support
with
application
load
balancers,
which
we
will
extend
to
network
load
balancers
as
well.
Eventually,
we
also
support
endpoint
slices
for
ip
targets,
some
load
balancer
attributes
for
nlb
some
subnet,
auto
discovery
enhancements
for
the
2.3
release.