►
From YouTube: Kubernetes - AWS Provider - Meeting 20191213
Description
Recording of the AWS Provider subproject meeting held on 20191213
A
Hello,
everybody:
this
is
AWS
bi-weekly
meeting
I,
have
your
host
moderator
just
in
Santa
Barbara
I
work
at
Google,
a
reminder.
This
mini
you
speeding,
report
recorded
them
will
be
put
on
the
internet,
and
so
therefore,
please
be
mindful
our
code
of
conduct
I
have
pasted
a
link
to
our
agenda
in
the
notes,
but
sorry
in
the
chat,
but
please,
if
you
have
things,
add
them
on
there.
We
have
a
couple
of
things
on
the
agenda:
thanks
Nick,
thanks,
Peter
and
I.
B
B
So
if
we
can
I
think
the
goal
for
this
would
just
to
be
identify
a
handful
of
critical
issues
that
we'd
like
to
see
somebody
work
on
for
the
1:18
release
cycle.
So
if
we
can,
you
know
if
we
can
identify
owners,
that's
fine,
but
we
don't
have
to.
We
can
just
identify
issues
and
then
over
the
next
week
or
so
try
to
identify
owners
for
those
issues.
So
with
that,
let's
get
started
and.
A
B
B
B
A
Reasonable
it'll,
say
yeah,
it
looks
like
it
has
pretty.
It
looks
like
they're
proposing
schema
changes
as
well,
so
I,
wouldn't
I
would
classify
this
while
it
may
be.
A
limitation
I
would
classify
this
as
a
feature
more
than
a
bug.
We
can't
unless
I'm
we
wanted
to
actually
address
it.
It's
not
it's
not
gonna,
be
a
small
thing.
It
would
probably
be
kept.
For
example,
yeah.
B
B
A
B
B
But
yeah
I
mean
like
what
I'm
saying
is
that
if
somebody
is
picking
this
up,
I'd
like
to
see
so
andrew
has
identified
that
it's
more
than
just
AWS,
but
this
this
would
be
something
that
if
someone
was
interested
in
picking
this
up,
you
know
I
think
yeah
writing
a
kept,
for
it
would
be
a
good
starting
point.
Yes,.
A
A
Excuse
me.
We
do
add
tags
I
think
so.
Some
of
this
could
also
be
addressed
if
someone
talked
to
the
AWS
console
team
and
suggested
they
put
the
tag
more
prominently
in
there
in
the
display.
So
again,
so
we
now
hagit
with
things
like
I
think
the
namespace
in
the
name
of
the
service
and
the
in
ec2,
for
example,
the
the
name
tag
is
displayed
very
prominently,
I
assume
not
the
case
in
as
last
time
I
looked,
it
was
not
the
case
in
the
yelly
right.
B
B
B
Guess
I
would
guess
we
would
still
allow
for
the
the
randomly
generated
change.
I,
don't
you
know
like
to
stay
backwards
compatible.
So
if
you
didn't
want
to
write
to
pick
a
name,
you
could
still
randomly
generate
one,
but
the
other
thing
is:
what
about
other
cloud
consoles?
Do
they
you
know,
do
they
have
similar.
C
A
A
A
B
A
A
B
A
A
That
was
the
it
was
that
that
second
comment,
that
I
think
was
the
slightly
problematic
one
like
today
we
use
in
our
internal
reconciler.
We
use
the
port
as
the
primary
key
is
my
recollection,
and
now
we
would
have
to
use
port
plus
proto
four
times
per
toe
as
the
primary
key,
which
I
mean
it's.
Not
it's
not
it's
not
crazy,
but
it's
it's
certainly
a
slightly
bigger
change
than
this
would
be
and
would
probably
touch
more
of
the
code.
It
should
be.
B
B
A
I
mean
I,
don't
think
I,
sorry
I,
don't
even
think
it
needs
a
cap
in
that
it
is
certainly
missing
functionality.
It
looks
like
looks
like
MWA,
so
I
were
sort
of
looking
for
it,
I
just
okay.
That
person
commented
at
the
end,
NW
triple-a
s
I'm
les
I
guess
so.
Maybe
they
maybe
we
can
like
work
with
them
to
make
it
work.
It's
it's
basically,
just
a
PR
that
that
does
that
I
guess
yeah.
A
B
A
A
A
A
B
And
it's
just
so
I
guess
it's
could
potentially
be
an
issue
with
the
the
yum
package
for
cubelet
queue:
medium
QC.
Yes,.
A
B
B
B
C
B
A
D
A
My
good
job
dogs,
trying.
A
Name
override
is
like
a
way
to
specify
node
name,
there's
only
a
lot
of
confusion
about
host
name
and
node
name.
We
don't
currently
allow
I
think
this
is
related
to
the
issue.
Wish
I'm
surprised
hasn't
been
more
uploaded
around
like
allowing
arbitrary
node
names,
but
we
don't
currently
allow
them
and
so
I
think.
B
B
A
A
And
the
reason
we
do,
that
is
because
we
in
some
places
we
only
get
the
node
name
on
some
of
our
calls
and
so,
for
example,
attaching
to
a
when
we
attach
a
volume.
We
need
to
know
what
ec2
instance
ID
to
attach,
and
so
we
have
to
map
from
node
name
too
easy
to
instanceid
and
the
way
we
do
that
today
is
we
query.
A
A
Assist
there
we
are
so
they
get
asks.
Why
do
you
want
to
do
this.
C
B
B
A
A
Yeah,
exactly
like
that's,
we've
had
we've
had
the
higher
priority
or
the
higher
priority
reorganization
of
the
code
rather
than
doing
that,
but
but
on
the
flip
side,
now
that
we
are
I,
think
everyone
is
now
using
the
approach
where
we
only
have
one
copy
of
the
code,
so
we
could
at
the
same
time
refactor
it.
Yes
I
guess
it's
just
a
matter
of
humans.
A
D
A
C
C
Don't
know
if
Justin
you
wanna,
okay,
sorry
sure
that
the
PR
it.
A
D
A
A
C
It
cool
this
is
a
cherry
pick
of
a
commit
Justin
made.
It
landed
in
117
and
I
was
saying.
If
we
get
back
according
to
116,
it's
a
to
fix
in
and
end
test.
I
wasn't
sure
the
cherry
pick
policy
is,
and
someone
was
deferring
to
Justin
to
see
if
it's
suitable
for
a
terrific,
so,
okay,
something
you
can
do.
They're
voting.
A
C
A
Okay,
cool.
Thank
you.
Thanks
for
bringing
it
up
all
right,
I'm
gonna
stop
sharing
the
screen,
because
this
is
just
taking
over
everything.
Any
kind
of
tell
what's
going
on
anymore.
I
stopped
this
other
side,
all
right
back
to
normal
all
right,
Nick
back
to
you,
I!
Guess,
thanks
for
that,
Peter.
B
D
B
B
B
A
B
B
B
B
B
B
A
There
is
logic
robot
once
set
up,
one
won't
pick
subnets
and
therefore
is
so.
We
try
to
set
it
up
in
all
the
zones
or
attach
things
in
all
the
zones,
but
we'd
like
the
subnets,
where
we
are
running
which
are
relevant
to
the
cluster
based
on
the
tag
and
other
heuristics,
and
then,
if
we
exclude
private
subjects,
because
private
subnets
can't
be
don't
work
with
a
public
load
balancer,
and
so
that's
why,
in
this
case,
subnet
one
though
or
sorry
the
one
that
has
access
is
the
one
that's
attached.
It
sounds
like
somehow.
B
A
B
A
B
A
B
A
B
A
B
A
A
B
B
A
This
might
also
be
one
of
the
things
that
we
can
fix
and
like
the
ingress
feed
to
efforts
which,
like
been
ongoing
for
almost
as
long
as
this
issue,
I
guess,
I,
don't
know
what
this
is:
someone
exposing
a
service
directly
or
whether
they're
exposing,
for
example,
nginx
and
a
low
bouncer
inside
their
cluster
I.
From
the
sounds
of
things,
it
sounds
like
they're,
exposing
a
service
directly,
in
which
case,
perhaps
this
is
our
thing
that
could
be
covered
by
ingress
v2,
which
is
in
keeping
with
what
I
Marshall
said.
A
B
I
would
knowing
very
little
about
this
issue.
I
would
be
inclined
to
say
that
it
sounds
like
I
mean
it
sounds
like
a
thing
that
somebody
might
want,
which
is
to
you
know.
You
can
configure
a
load
balancer
in
AWS
to
listen
on
443
and
then
forward
to
port
80
right
and
that's
what
he's
asking
for
right.
B
A
It
does
it's
sort
of
the
nice
thing
about
ingress,
so
the
bathroom
ingress
ingress
has
traditionally
been
there
seven.
So
it's
different,
but
nice
thing
my
ingress
is
that
it
is
a
different
object.
It's
a
different
interpretation
so,
like
we'd,
say
the
internal
service
we're
declaring
that
the
service
we're
declaring
is
internal
and
then,
in
addition,
we
have.
We
expose
that
service
externally
and
there's
some
fancy
stuff
that
happens.
In
this
case
we
expose.
There
were
four
four
three:
we
do:
CSL
decryption
in
ELB
or
yeah.
B
A
B
So
I
guess:
okay!
So
if,
if
ingress
works
as
a
future
for
this,
that
could
be
an
option,
another
option
would
just
be
to
sort
of
have
a
have.
It
have
somebody
take
a
stab
at
new
annotations
that
better
map
to
what
people
want
to
do,
or
maybe
give
a
little
bit
more
power
to
to
the
person
defining
the
service
and
allow
them
to
do
things
like
this,
where
you
have
a
different
listener
on
the
the
load
balancer
versus
the
service
or
the
the.
A
B
A
Alright,
thank
you
for
joining
that.
No
do
we
have
a
meeting
in
two
weeks.
I
guess
this.
Next
is
the
final
question
which
we
will
evaluate.
It's
a
is
the
13th
of
Friday,
the
13th
that
makes
the
27th,
which
is
I,
think
we
should
go
ahead
and
cancel
that
I.
Don't
think
I
think
the
attendance
would
be
super,
sparse.