►
From YouTube: 20190506 - Cluster API Provider AWS Office Hours
Description
No description was provided for this meeting.
If this is YOUR meeting, an easy way to fix this is to add a description to your video, wherever mtngs.io found it (probably YouTube).
A
Hello
and
welcome
to
the
May
6th
edition
of
the
cluster
API
provider
AWS
office
hours,
a
sub-project
us
a
cluster
lifecycle.
If
you
have
any
items
that
you
want
to
cover,
please
go
ahead
and
add
them
to
the
agenda
now
and
also
go
ahead
and
add
yourself
to
the
attending
list.
Too.
I
am
dropping
link
to
the
agenda
doc
in
the
chat
for
anybody
who's
attending
in
real
time.
A
To
start
with,
we
were
able
to
cut
the
0.21
release
this
past
Friday
Thank
You
Chuck.
So
we
have
that
out
there
now.
That's
mostly
just
bug
fix
release,
but
it
does
address
issues
that
people
are
definitely
hitting
with
the
0.20
release
and
then
next,
because
we
do
have
some
changes
coming
in
that
are
going
to
impact
tagging
of
the
resources
we
are
starting
to
track.
Our
0da
3.0
release
and
so
far
pull
request.
Number
of
706
is
the
only
item
that
I
have
for
tracking
there.
A
A
A
B
B
That
was
a
lot
of
email
back
and
forth
with
the
CNC
F.
So
I'm
glad
that's
done,
and
then
I've
also
got
the
next
item
here.
Oh
I
guess
maybe
no
nevermind
yeah
the
next
item
here,
an
image
promoter,
work
I
will
be
working
with
Harish.
Who
is
trying
to
get
more
involved
to
the
upstream
stuff
to
be
doing
some
of
the
image
promoter
work
for
both
I'm
hoping
he
can
also
like
kind
of
get
that
going
for
copy
or
closer
API,
but
it'd
be
good
to
have
in
both
repositories
as
well.
A.
C
Question
for
you
about
the
end
tests,
yeah,
okay,
so
what's
the
scope
of
those
that
like
is
intended
to
be
covered?
So
the
sorry
7:06
fixes
an
issue
with
like
load
balancer
integrations
from
the
cluster
API
in
or
out
of
scope
of
those
end
end
tests
should
I.
Add
MP
are
part
of
that
PR
that
validates
it.
You.
B
Know
so
right
now,
it's
not
I,
think
it's
it's
a
very
minimal
and
to
end
test.
It's
mostly
just
create
the
cluster
stand
down.
The
cluster
I
think
word
will
probably
have
to
have
a
plan
going
forward
about
what
is
or
in
not
in
scope
for
those
and
there's
a
whole
bunch
of
issues
that
kind
of
cover
what
or
were
thinking
like
what
people
have
been
thinking
over
the
past
few
months
has
been
to
it's.
A
Yeah,
so
one
of
the
things
that
we
definitely
want
to
do
for
the
intend
tests
over
time
is,
we
want
to
extend
them
to
do
full
conformance
tests
so
that
we
know
that
we're
actually
speaking
up
a
conforming
cluster,
but
that
still
would
not
validate
the
lower
balancer
test.
So
I
think
that
is
definitely
something
that
we
should
think
about.
Adding
it's
like
a
separate
kind
of
eating
test
in
there
and
I'm
not
sure
about
the
framework
they
were
using,
but
it
should
it
should
facilitate
adding
kind
of
a
second
parallel
test.
A
D
D
We
used
to
create
one
whoa
like
right
now
a
mess
we
actually
do
create
one
session
for
each
machine
that
comes
up
or
like.
If,
even
if
we
run
a
load,
balancer
would
create
a
new
session
and
give
the
Bliss
SDK
doesn't
actually
like
say
that
is
this.
This
shouldn't
be
done,
but
it
says
when
you
can,
you
should
catch
the
session
and
going
in
the
session
code
actually
like
shows
it
like.
It
doesn't
like
a
lot
of
like
back-end
calls.
D
For
example,
when
you
ask
you
what
role
and
to
like
kind
of
get
credentials,
that's
for
I
am
roles,
but
when
you
also
you
have
static
credentials.
Any
of
this
will
actually
block
you
when
you,
the
same
source.
Ip
makes
multiple
cause
what
the
same
or
like
multiple
sessions
with
the
same
a
SS
access
key.
D
So
for
that,
like
only
something
that
like
caches
these
sessions
and
we'll
need
to
make
sure
like
to
understand
how
like
that
works
for
like
screen,
so
they
can
expire
or,
for
example
like
if,
like
we
want
to
have
at
some
point
the
a
SS
provider
to
handle
multiple
credentials.
I,
don't
know
if
that's
go
for
now,
but
for
now
like
we
can
probably
just
cache
on
the
endpoint
or
region
name
and
but
if
anybody
knows
more
about
it,
please.
E
We
did
in
bolero
is
we
have
a
single
pod,
that's
running
that
will
operate
on
a
single
set
of
credentials,
and
so
it
creates
one
session
and
uses
that
for
the
lifespan
of
the
pod
and
as
far
as
I
know,
the
go
SDK
for
AWS
will
handle
refreshing
any
credentials
that
are
temporary,
like
the
STS
tokens.
So
I
would
say
that
until
and
if
we
need
to
handle
multiple
accounts
or
multiple
sets
of
credentials,
then
we
can
create
a
single
session
per
region,
probably
or
whatever.
E
F
G
Yeah,
so
I
don't
think
Dominic
is
on
the
call.
I
was
just
kind
of
going
over
that
with
PR
about
some
of
the
note
draining
and
I
was
just
curious,
is
if
the
plan
for
that
is
something
that
would
move
up
into
Cappy
at
some
point
and
if
you
know
how
kind
of
in
the
timeline
that
transitions
like
that
would
work.
Yes,.
A
G
Got
it
that
answers
that
I
think
I
also
have
been
next
item
so
I
believe,
maybe
six
weeks
ago
we
on
this
call
brought
up
that
we
were
interested
in
core
OS
support.
We
actually
brought
that
up
in
a
cluster
API
meeting
and
kind
of
after
some
back-and-forth
decided
to
kind
of
start
that
work
within
the
provider
and
then
later
bring
that
up
and
I
just
wanted
to
say
on
this
call
that
we're
gonna
start
diving
into
that.
D
F
So
untrue,
no
and
or
set
they'd
open
a
bug.
The
method
was
not
doing
what
was
not
behaving
as
expected
under
certain
scenarios
and
I
have
a
PR
out
to
fix
that
I
think
I,
I
think
Jason
and
Chuck
have
looked
at
a
tiny
knife
away.
Also
think
wins.
You
have
low,
so
looked
at
it,
but
what
I
wanted
to
discuss
is
like
long
term
with
like
the
node
lifecycle
and
work
stream
like
where
do
we
see
this
going
and
like
what
is
the
timeline
like?
Will
this
bill?
F
H
D
Was
gonna
say
like
I,
wouldn't
move
it
upstream
yet,
given
the
work
that
we've
been
doing
on
the
datatypes,
but
I
would
like
definitely
add
the
use
cases
for
it,
so
that
when
we
actually
go
into
an
implementation
detail
which
I
believe
this
should
probably
be
controlling,
lifecycle
management
or
we're.
Actually
this
could
be
a
reference
implementation.
F
D
I
think
somebody
else
yeah
go
ahead
in
the
in
the
new
in
the
new
proposed
like
if
things
work
out
like
as
we're
thinking,
there
would
be
like
the
control
in
humanity,
its
own
machine
so
like
they
wouldn't
like
you
already
know
it
doesn't.
Machine
would
either
be
like
you
really
know
like
if
you
have
one
machine
and
that
should
be
the
first
control
plane
and
if
you
join
the
others,
but
for
worker
nodes
you
it's
always
gonna,
be
a
joint
yeah.
D
F
No
Oh,
if
nobody
else
has
I
have
a
real
request.
I
think
I
know
I've
posted
about
this
in
the
channel
and
most
folks
may
know
I
don't
have
an
AWS
account
anymore
for
me
to
test.
So
if
somebody
can
set
me
up
with
a
temporary
AWS
account
that
I
can
share
I
promise
to
use
it
responsibly.
So
that
would
be
awesome,
so
just
wanted
to
get
it
out
there.
I
did
email.
I
did
follow
Chuck's
suggestion
of
emailing
a
horse
from
CN
CF.