►
From YouTube: Kubernetes - AWS Provider - Meeting 20220902
Description
Recording of the AWS Provider subproject meeting held on 20220902
Agenda - https://docs.google.com/document/d/1-i0xQidlXnFEP9fXHWkBxqySkXwJnrGJP9OGyP2_P14/
A
Good
morning
everybody
welcome
to
the
september
2nd
2022
meeting
of
provider
aws
meeting
a
sub
project
of
sick
cloud
provider
of
kubernetes.
Please
be
mindful
of
the
cncf
code
of
conduct
and
be
respectful
of
the
comments
you
make
and
be
inclusive.
So,
let's
go
ahead
with
the
meeting.
B
I
can
mention
that
the
there
was
an
issue
I
mentioned
a
couple
of
weeks
ago,
around
run
c
and
contingency,
which
has
actually
been
patched.
So
if
you
were
on
113,
you
can
now
go
to
114
and
all
as
well,
so
just
to
close
the
loop
on
that
which
I
mentioned
before,
and
so
we
did
some.
We
did.
Some
chaos
releases
that
basically
included
that
release
include
the
update.
C
Yeah,
so
I
guess
I'll
give
a
ccm
update.
We
have
a
bunch
of
cherry,
picks
and
releases.
C
Last
couple
days
there
was
a
kishore
you're
familiar
with
the
target
group
picks
for
123..
That's.
C
123
release
and
then
bash
over
the
other
ones.
There's
pretty
the
scs
rate
limit
an
issue
with.
I
guess
is
it
scs
assume
roles
where
yeah
it's
with
the
cs,
assume
role?
Yeah.
You
want
to
give
a
couple
a
summary
of
that
one.
D
Yeah
sure
so
this
was
related
to
the
when
we
try
to
talk
to
when
we
try
to
assume
a
role
in
in
any
account
and
that
role
doesn't
exist.
So
we
want
to
rate
limit
that,
like
how
frequently
other
controllers
are
trying
to
assume
that
rules.
So
the
fix
was
to
slow
down,
basically
cache
the
result
for
a
second
or
two
and
then
rate
limited.
C
And
then
there
was
a
fix
in
the
tagging
controller.
I
don't
know
what
the
issue
was
there
and
then
there
was
also
a
fix
with,
so
it
was
related
to
fargate
and
resource-based
naming
so
that
one,
I
guess
part
of
it-
was
totally
broken
with
resource-based
naming
or
something.
Oh.
So
anyway,.
A
I
have
an
issue
that
I
it
came
from
the
cloud
provider
ada
sick
cloud
provider
meeting.
I
I
just
want
to
bring
mention
here
as
well.
Let
me
bring
it
up.
C
A
A
So,
what's
happening
with
this
is
like
in
certain
cases
where
a
user
has
more
zones
like
the
local
zones
or
the
wavelength
zone.
The
auto
discovery
portion
of
the
subnet
resolver
is
it
breaks
because
it
finds
subnets
in
different
availability
zones
and,
as
of
now
like
other
local
and
web
web
zone,
is
not
even
supported
for
nlp
or
clv.
A
So
what
I
suggested
was
like:
we
restrict
the
auto
discovery
to
zone
type
of
ability
zone
only
and
when
nlp
supports
it
later,
then
we
can
revisit
this,
but
for
now
for
the
cloud
wider
aws,
we
will
restrict
to
ac
only
so
this
will
fix
the
auto
discovery
issue
that
was
reported.
I
also
talked
to
the
reporter
from
red
hat
and
he
agreed
to
do
a
pr,
maybe
this
week
or
the
next
week,
to
fix
this
issue.
B
A
We
do
for
public
and
private
yes,
but
these
are
the
new
zones
right.
These
were
not
present
before
it
just
got
added
recently.
It
looks
like
so
or
fairly
recently
and
if
the
customers
tag
their
subnets
in
those
zones,
then
auto
discovery
will
go,
find
them
as
well.
So
that's
what
we're
trying
to
prevent
here.
A
A
A
A
C
A
A
C
Okay
yeah,
I
guess
it
makes
sense,
it
does
seem
kind
of
gross,
but
it
seems
like
this
is
overstuck
with.
A
And
even
like
for
alb
like
it
doesn't
support
cross
cross
zone
types
right
now
like
it
has
to
be
in
the
same
zone,
so
we
still
need
further
enhancement
in
this
area.
So
even
in
future
we
might
want
to
say
okay
only
this
zone,
if
it
supports,
I
don't
see
like
immediate
support,
but
if
it
supports
then
for
future
we
will
have
to
say
okay
restrict
to
this
particular
zone.
C
Yeah
is
there
a
chance
we'll
want
something
more
configurable
where
we
choose
what.
A
So
this
this
is
only
for
clv
and
nlb
right.
This
is
a
cloud
provider
aws,
so
here
like
we
just
have
to
go
with
clv
or
nlp
in
the
load.
Balancer
controller
will
be
found,
fine
tuning
it
like
specifically
specify
what
zone
type
that
they
want
to
use,
because
we
want
to
support
alb
as
well,
so
there
will
be
able
to
provide
better
configurable
options
in
this
case.