►
From YouTube: 20200309 - 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
March
ninth
edition
of
the
cluster
API
provider,
AWS
office
hours,
a
sub-project
state,
cluster
lifecycle
and
cluster
API,
just
a
reminder
that
this
meeting
is
recorded
and
will
be
posted
up
to
YouTube
later
and
that
the
kubernetes
community
guidelines
are
in
effect.
So
in
general,
please
be
excellent.
A
A
All
right
to
start
with,
we
do
have
a
couple
of
announcements
that
both
the
Virgin
0.50
VAR
c
0
and
r
c
dhawan
releases
were
previously
cut.
Since
their
last
meeting
we've
been
using
those
for
testing
with
the
B
1
alpha
3
release
for
upstream
cluster
API
and
to
better
assess
folks
being
able
to
test
like
the
cluster
kernel,
workflows
and
overall
user
experience.
A
B
So
I
wasn't
sure
whether
we
wanted
to
like
dive
into
this
properly
now
or
not
happy
to
do
a
bit
of
explanation
about
it.
But
we've
got
this.
For
instance,
proposal
we've
so
put
together
over
the
last
few
weeks,
which
is
basically
it's
try
and
add
spines
and
support
not
only
to
the
azure
but
also
AWS
and
GCP
providers,
but
the
sort
of
way
that
we're
planning
to
do
that
is
by
adding
some
fields
into
each
of
the
cloud
providers.
B
Now
I
spoke
with
Vince
about
this
and
I've
sort
of
summarized
what
he
was
saying
there,
which
is
that
there
might
be
concerns
from
each
of
the
provider
maintains
about
adding
fields
and
that
maybe
it
would
be
better
to
just
try
and
go
for
adding
support
for
machine
pool.
So
I
wanted
to
sort
see
if
there
are
any
thoughts
from
your
side
on
like
whether
this
is
likely
to
be
a
non-starter
or
whether
this
is
something
that
might
be
feasible
to
just
you.
B
A
In
general,
I
do
like
the
idea
of
officially
supporting
machine
pools
but
I
think
that's
kind
of
potentially
a
tangential
concern.
I
think
there
are
particular
use
cases
where
you
might
want
to
retain
kind
of
the
Machine
semantics
and
still
support
spot
instances
versus
having
to
take
on
all
of
kind
of
the
machine
pool
semantics
and
the
things
that
go
along
with
that.
So
at
least
I
think
on
the
80
of
us
died
we'd
be
open
for
changes
that
we
need
there.
I
can't
necessarily
speak
for
some
of
the
other
providers.
Yeah.
B
Ok,
that's
that's
good
to
know.
Vince
was
suggesting
that
the
GCP
will
be
fine
as
well.
I
just
need
to
then
get
an
AK
from
Azure,
but
yeah.
The
ideas
would
be
an
optional
field
on
the
machine
spec
that
then
could
be
used
either
by
the
machine
pool
implementation
or
the
machine
set
implementation
machine
implantation.
That
would
make
us
bones.
A
A
B
That's
that's
good
yeah,
okay,
in
which
case
yeah,
if
anyone
could
have
some
time
at
some
point
to
have
a
reef
room
that
make
out
some
comments
like
I
say
it's
quite
lengthy.
I
would
like
to
try
and
push
this
forward
because
we
want
to
try
and
get
it
done
in
the
next
cycle
in
the
beef.
Not
four
cycles
of
Kathy
yeah
would
appreciate
any
feedback
on.
C
A
A
A
It
had
some
questions
related
to
fixing
things
up
so
need
to
wait
for
that
to
be
addressed,
and
this
one
is
also
related
to
testing
I
believe
all
the
rest
already
have
milestones
attached
on
to
issues
tracking
issue
for
at
CDs
dollar,
or
this
is
related
to
an
upstream
issue
that
we're
seeing
I
think
we
can
easily
put
this
in
the
0.5
X
milestone.
Unless
anybody
disagrees.
C
A
So
it's
the
sum
that
I've
seen
at
least
when
I
dug
into
the
logs
I've,
never
seen
it
actually
causing
any
type
of
functional
issues,
at
least
with
CAPA,
so
I
think
there's
the
still
probably
worth
digging
into
a
little
bit,
but
I
don't
think
it's
as
critical
as
for
some
of
the
other
providers
that
are
seeing
it
more
frequently
right
now
and
other
ones
around
the
periodic
test.
Failure.
A
All
right
that
brings
us
into
what
we
have
left
for
0.5
dot,
0,
which,
although
it
says
passed
you
by
three
days,
is
actually
going
to
happen
this
week
we
have
the
add
instructions
for
existing
AWS
infrastructure.
I
think
this
one
was
just
waiting
on
additional
testing.
I
tried
to
do
some
testing
on
it
last
week,
but
was
never
able
to
finish
it.
A
D
A
All
right,
some
ami
is
missing.
The
AWS
CLI
tools
I
know
set
that
I
found
the
point
where
we
had
added
those
in
which
a.m.
I
didn't
yet
have
them.
So
this
is
just
a
matter
of
me:
go
on
head
and
doing
this
and
I'll
do
this
as
part
of
the
release
when
we
cut
it
ad
Docs
for
consuming
existing
eight
of
us
infrastructure,
that's
related
to
the
PR
that
we
already
discussed
value
weight
conversion
web
book,
behavior,
based
on
liggett's
review
of
cluster
api
I
plan
on
doing
this
later
today
and
into
tomorrow.
D
D
Is
that
if
you
submit
an
alpha
to
thingamajiggy-
and
it
has
something
that
would
cause
it
to
fail,
alpha
3,
open,
API
validation,
so
the
example
that
we
tested-
or
that
said,
I've
tested
was
you
have
a
machine
and
in
the
machine
stack
there
is
a
cluster
name
filled
in
alpha
3?
It
doesn't
exist
in
alpha
2.
D
The
expectation
is
that
there'd
be
a
label
in
the
alpha
2
machine
that
would
get
converted
to
the
Alpha
3
spec
field
and
then
ideally
go
through
open,
API
validation
to
make
sure
that
it's
present
and
because
of
the
way
that
the
API
server
does
its
order
of
operations.
It
appears
that
open
API
validation
happens
before
converting
to
the
storage
version,
which
basically
means
that
it
passes
alpha
to
validation
because
everything's
fine
it
gets
converted
to
the
alpha.
D
3
version
goes
against
the
validating
webhook,
which
doesn't
have
a
check
for
this
field,
because
it's
an
open,
API,
validation
and
then
because
it's
not
doing
additional
open,
API
validation,
it
just
persisted,
and
that's
neat.
So
technically
you
have
an
object
that
had
you
started
with
alpha
3,
it
wouldn't
have
failed,
open,
API
validation
wouldn't
be
a
CD.
The
reason
that
Vince
and
I
don't
think
it's
a
release.
Blocker
is
that
you
can
go
in
and
edit
the
thing
after
you've
created
it
and
fill
in
the
missing
fields
and
I.
D
Think
this
is
the
sort
of
thing
that'll
probably
be
on
a
case-by-case
basis.
Where,
if
we
find
problems,
we
can
fix
them
worst
case.
We
could
duplicate
our
validation,
so
we
have
open
API
and
webhook,
but
that's
kind
of
a
key,
so
mainly
the
fact
that
it
doesn't
break
your
informers
and
it
doesn't
break
lists
and
that
sort
of
thing
and
we're
going
from
an
alpha
version
to
another
alpha
version.
We
thought
it
was
okay
to
defer
any
further
investigation
or
resolution
until
a
dot
X
for
upstream
for
per
copy
yeah.
A
A
A
D
D
A
Yep
so
I'll
I'll
leave
it
in
the
milestone
for
now,
but
I'll
consider
it
potentially
some
that
we
can
bump
if
needed.
If
we
don't
get
to
it,
do
we
have
anybody
that
could
potentially
take
a
look
at
this
I.
A
Were
named
1156,
oh
wow,
okay.
This
is
the
work
from
the
Telus
folks
to
try
to
get
to
allow
to
define
no
key
pair
to
use
on
the
machine
by
updating
the
current
value
of
key
pair
that
we
have
a
pointer
and
allowing
specifically
setting
the
empty
string
to
say
explicitly.
Do
not
assign
a
key
pair
to
this
instance.