►
From YouTube: 20180820 sig cluster lifecycle aws cluster api
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
folks,
this
is
the
sig
cluster
lifecycles
AWS
cluster
API
implementation,
working
group
or
sub
project,
it's
more
of
a
sub
project,
and
we
only
have
one
agenda
item
today,
which
is
Jason,
has
a
doc
which
is
a
distillation
of
requirements
around
an
MVP
that
kept
you
plans
I'm
working
on,
and
we
can
kind
of
talk
about
how
this
overlaps
with
other
requirements
that
other
folks
might
have
and
who
might
be
able
to
contribute
to
what
going
forwards
on
an
MVP
so
Jason.
If
you
want
to
share
your
screen.
C
Yes,
all
right-
and
this
is
the
same
document-
that's
linked
in
the
agenda
notes
as
well
so
going
through.
We
had
a
few
different
documents
created
from
the
kickoff
meeting.
One
of
them
was
a
basic
requirements,
talk
another
one
was
the
lessons
learned,
doc
and
I
tried
to
kind
of
distill,
both
of
those
to
come
up
with
distilled
kind
of
MVP
that
we
can
target
for
the
initial
implementation
going
through.
C
There
were
some
things
that
weren't
exactly
defined
in
the
in
the
requirements
list,
so
the
filling
in
gaps
from
the
feature
list
right
here
is
just
me
trying
to
codify
some
of
some
of
those
gaps.
So
one
of
them
is
we
don't
mention.
I
am
rules,
or
instance,
profiles
at
all,
so
for
the
MVP
figuring
it
was
the
path
of
least
resistance
and
and
most
amenable
for
end.
Users
would
be
to
require
a
predefined
set
of
iam
roles,
and
instance,
profiles
to
be
used.
C
That
way,
we're
not
requiring
end-users
to
give
credentials
to
the
controller's
here
to
actually
create
those
those
things,
because
we
know
that's
a
an
issue
for
some
organizations
and
then
we
also
don't
mention
security
groups.
So
I,
you
know
just
putting
down
security
groups
can
be
created
on
demand
or
or
we
can
use
predefined
ones
for
the
MVP.
C
The
other
thing,
that's
not
really
that
we
don't
really
go
into
detail
with
the
requirement
stock
is
the
actual
bootstrapping
steps
and
for
hefty
Oh
heavy
owes
use.
We
would
like
to
require
pre-baked
images
that
already
contained
the
binaries
requires
input,
and
we
would
also
prefer
to
go
with
a
cloud
in
it,
driven
approach
for
kind
of
driving,
cube,
ATM
and
then
also
exposing
the
cube
ATM
config
within
the
provider
config
for
providing
customization.
C
A
D
A
Okay,
one
of
the
things
that
we
had
talked
about
with
regards
to
pre
baked
images
is
we
have
external
tooling,
which
we
could
probably
donate
to
the
community
somewhere
with
regards
to
baking
images
for
Koob
idiom,
just
because
it
sits
on
hefty
or
assets
right
now,
but
we
can.
We
can
absolutely
give
to
community
to
make
it
more
community
focused
for
some
of
these
things,
so
that
way
they
can.
The
image
baking
can
be
generic
for
everyone
in
the
community
can
contribute
to
that
and
make
it
fresh
for
us
in.
E
E
F
A
F
So
what
I
mean
to
say
is
work
on
me
or
the
worker
pre-baked
image
versus
the
master,
pre-baked
image,
I.
Think
I'm
understanding
you
right
for
a
guess.
We
already
provide
the
eks
worker,
no
Tommy,
we
publish
it
so
I
think
it
would
make
sense
to
actually
publish
the
book
a
node
army
as
well
and
I.
Guess
that
makes
sense
from
the
control
plane
standpoint.
A
No
now
this
is
this
is
kind
of
specific
to
an
implementation.
Detail
that
we
kept
do
want
to
work
on.
This
doesn't
include
like
flavorings
as
I
know.
Chris
was
working
on
to
determine
like
nomenclature
for
us
to
talk
about
like
if
folks
wanted
to
pick
eks.
We
absolutely
want
to
have
the
capabilities
in
this
implementation
of
cluster
API,
for
people
to
spin
up
ETS
clusters
yeah.
A
But
what
we
have
do
not
the
Royal
we
so
I'm
gonna
have
to
like
preface
my
wheeze
I,
don't
know
I'll
just
start
saying:
F
do
but
have
dia
what
is
probably
gonna.
Work
on
first
is
just
providing
that
base
cube,
ATM
control,
plane
layer
for
folks
to
just
bootstrap
and
get
rolling,
and
if
folks
are
willing
to
execute
for
the
ek
flavor
that
we
can
provide
some
of
that
abstraction
abstraction
in
the
provider
continue
makes.
A
Right,
that's
a
broader
question.
The
purpose
of
this
one
is
to
make
it
easy
to
support
one
to
end
clusters
so
through
a
centrally
managed
minute
control
plane.
Ideally,
this
would
be
a
standard,
but
it
remains
to
be
seen.
Standards
always
like
you
always
set
forth
an
initiative.
I,
don't
I,
don't
want
to
like
go
into
this
thinking
that
this
will
become
the
shiny
city
on
the
hill.
I
want
to
do
step,
one
first.
Okay,
sense.
F
C
C
Anything
that
would
relate
to
potential
impedance
mismatches
with
the
chlorin
current
cluster
api
definitions,
and
for
that
I
put
the
e
KS
based
deployments
and
the
autos
Gatling
groups,
because
those
don't
necessarily
mesh
in
exactly
with
the
machine
actuator
interface
there.
There
might
be
some
extra
work
to
kind
of
shoehorn,
those
in
to
fit
properly
with
cluster
API.
So
that's
generally,
why
we
avoided
those
two
topics:
no
I'm.
A
Gonna
I'm
gonna
interrupt
you
again
because
it's
my
job,
David
or
others
who
have
already
implemented
stuff
like
this.
What
are
your
thoughts
here
because
I
know
that's
a
big
topic
in
the
larger
cluster
API
conversation.
G
G
A
E
E
Users
can
create
machines
or
controllers
can
create
machines
if
we
wanted
to
implement
a
SGS
or
he
has
controllers.
Those
controllers
would
have
to
create
the
machines
for
the
user,
and
so
we
have
sort
of
a
an
inconsistent
owner
of
the
machine
object.
Is
it
a
user
or
is
it
an
ASG
controller
I
think
we
need
to
resolve
that
I.
Don't
think
that
I
have
an
answer
for
it
in
this
meeting,
so
I
think
it's
a
fair
place
to
say
we're
deferring
this
decision,
but
I
think
we
have
to
solve
it.
E
A
C
Moving
on
the
next
deferred
feature
is
separate:
control
of
public
and
private
network
selection
between
the
control
plane
and
the
workers.
The
feature
document
actually
lists.
You
know
public
private,
for
control,
plane
and
worker
nodes
separately
to
kind
of
strip
things
down
for
the
MVP
I
kind
of
booted
being
able
select
those
independently.
G
So,
on
our
end,
we
we
would
like
to
be
able
to
select
which
news
are
on
a
public
network
and
versus
the
private
network,
particularly
if
you
have
a
server.
That's
part
of
a
hybrid
cloud
right
now,
where
some
some
workload
is,
is
communicating
via
VPN
to
your
internal
network
versus
other
workload.
It
just
needs
to
be
public.
C
G
E
F
C
I
could
skip
down
here
to
the
kind
of
diagrams
that
I
created,
basically
the
first
ones
of
kind
of
like
a
fully
public
network,
single
ad
deployment
of
both
the
control
plane
and
workers,
and
the
second
one
is
basically
a
private
networking
single
AC,
where
both
the
control
plane
and
the
workers
are
in
a
private
subnet
Expo.
The
API
services
exposed
through
a
public
ELB
and
any
other
access
to
the
hosts
would
be
through
a
bastion,
a
bastion
instance.
That's
kind
of
where
public
and
private
is
coming
from.
Does
that
clarify
it
at
all?
Yeah.
A
This
is
what
we
have
Specht
out
as
in
we
as
an
app
deal,
so
I
have
to
preface
it
that
we
will
execute
on,
but
for
an
MVP
within
like
a
reasonable
time
frame
that
week
and
that
we
will
try
to
work
on.
But
if
other
folks
are
willing
to
throw
down
and
have
resources
to
execute
on
these
pieces.
We'd
absolutely
welcome
them
and
would
want
to
have
them
part
of
the
development
process.
D
C
C
G
C
The
next
one,
the
next
few,
are
around
accessibility,
the
first
one
purely
private
API
server
access
I
wanted
to
try
to
avoid
that.
Just
for
simplicity
of
the
MVP
I
figure,
the
most
common
generic
use
cases
for
the
people
who
would
be
willing
to
try
out
the
cluster
API
generally
would
not
have
issue
with
a
publicly
accessible
API
server,
and
then
we
could
look
at.
You
know
how
we
can
support.
You
know
purely
private
access
down
the
line
and
similarly
the
VPC
private
link.
You
know
very
related
to
that
same
topic.
F
C
A
Haven't
made
a
question
for
folks
what
see
our
eyes
other
than
redhead,
which
I
already
know
the
answer?
What
see
our
eyes
are
folks
using
by
default
currently
for
their
cluster
stand-ups,
because
in
112
this
is
going
to
change
not
seeing
I
CRI
David,
so
like
container
D
versus
Kryon
versus
duck
or
something.
F
A
A
A
A
F
F
F
C
H
C
Yes,
the
idea
would
be
it's
law.
The
only
requirements
for
this
MVP
that
that
I'm
proposing
would
be
that
cube.
Atm,
the
cubelet
and
cube
cuttle
are
installed
on
the
on
the
system
and
and
some
of
the
requirements
around
that
it's
Dre
pretty
minimal.
That's.
H
Fine
I
I
deal
with
a
bunch
of
customers
that
get
hung
up
on
the
concept
of
whose
neck
can
they
choke?
Who
can
they
plain
blame
at?
Therefore,
there's
a
bunch
of
enterprise
support
issues
that
come
up
even
though
they're
adopting
kubernetes
so
just
trying
to
keep
my
own
the
forward-looking
thanks,
sorry
for
the
interruption,
no.
G
C
G
A
C
B
A
A
We
could
have
one
Tommy
to
rule
them
all
if
you
wanted
to.
Is
it's
not
going
to
take
up
that
much
extra
space
to
do
to
put
all
the
containers
in
one
image
versus
another?
Ideally,
we
wouldn't
need
containers
or
access
to
the
outside
world
just
to
do
the
bootstrapping
portion.
That's
up
to
other
people,
I.
C
A
Are
there
any?
Are
there
I
want
to
make
sure
this?
Is
the
community
understands
that
this
is
not?
This
is
a
decision
that
we
have
do
have
said
that
we
are
going
to
execute
on
and
it's
not
by
no
means
the
only
way
or
the
only
path,
but
this
is
the
new
year.
These
are
the
pieces
that
we
plan
to
execute
on.
We
will
resource
this
on
our
side.
A
Are
there
areas
like
mentioned
Cesar
mentioned
wanting
to
add
user
data?
Are
there
other
areas
that
folks
would
like
to
help
in
or
contribute
on
or
be
a
part
of,
or
just
you
know,
I
want
to
make
sure
that
other
folks
are
involved
and
understands
that
this
is
not
us.
Trying
to
you
know
push
the
community
I
want
the
community
said.
This
is
what
we
will
resource
and
the
community
can
say
other
things
too
as
well.
If
you
want
other
things
to
work
on
it,
we'll
try
to
do
our
little
best
to
accommodate.
F
A
E
E
A
From
our
side,
we
absolutely
want
to
fast
follow
what
he
casts
after
we
get
the
first
MVP
done,
but
if
folks
want
to
work
on
it
now,
we
we
happily
try
to
help
with
the
abstraction
pieces
and
provider
configuration
to
make
it
possible,
but
we
we're.
We
have
decided
that,
like
we
only
have
so
many
resources
to
be
able
to
do
this.
G
E
C
A
Do
have
questions
about
like
group-wide
utility
tools,
around
AWS
and
sort
of
best
practices
that
I
was
hoping
to
leverage
I
know
that
Chris
was
going
to
work
on
that
document
with
regards
to
like
utility
wrappers
in
an
ideal
world,
it
would
be
great
to
have
like
a
canonical
set
of
utility
wrappers
to
help
support
the
ideas
idempotency
for
doing
certain
operations,
because
I
know
that
that's
a
big
kicker
for
a
lot
of
folks,
she's
gonna
be
out
for
a
while,
but
I
don't
know
if
other
folks
had
a
thoughts
or
opinions.
There.
A
D
You
I'm
I'm
still
a
bit
unsure
about
what
what
some
of
the
things
that
class
a
coyote-like?
Will
it
create
read
pcs
or
will
it
not
I
mean
those
kind
of
questions
decide
what
kind
of
utility
functions
we
need.
C
Yeah
exactly
so,
you
know
in
for
the
MVP
we're
looking
at
by
default,
creating
a
V
PC
if,
if
one
is
not
specified
already
the
subnet,
the
you
know,
there's
a
subnet
below
that.
You
know
for
the
easy
that
we're
targeting
the
security
groups
for
the
control
plane
and
the
worker
nodes,
the
load
balancer
that
we're
going
to
use.
You
know
pretty
much
all
of
the
infrastructure
you
know
south
of
you
know.
The
I
am
role
and
instance,
profiles
that
we're
going
to
require
be
passed
in
and
pre-created.
A
Yeah,
this
is
the
kind
of
detailed
information
that
I
think
helps
because
there's
a
large
user
pool
of
information
that
we
want
to
leverage
it
put
it
in
putting
in
the
community
repo
that
we
don't
want
to
like.
We
don't
want
to
repeat
the
same
mistakes
yet
again
in
another
implementation,
so
I'll
try
to
reach
out
I,
don't
know
where
that
document
lives.
I
know
Chris
planned
on
working
on
there
does
anyone
else
know
where
that
might
live
are.
G
A
If
you
want
me
to
get
something
moving
along,
but
I
think
we
should
probably
get
someone
on
the
review
for
that
one
get
the
dock
into
sort
of
spec
form
in
the
repo
with
a
reference
and
then
also
a
bunch
of
clauses
that
outline
like
you
know
we're
gonna
potentially
address
this
in
the
next
milestone.
So
I
think
it's
on
it's
on
our
responsibility.
Right
now
to
get
the
backlog
back
in
place
and
get
reviews
started
to
get
rolling
and
I.
Think
just
open
issues
in
PR
is
I,
think
I.
A
Ideally,
when
we
discussed
it
internally,
we
would
like
to
go
along
with
the
112
release
cycle
just
to
get
something
out
there.
So
at
the
right,
an
ideal
world
alpha
hits
at
the
same
time
as
112.
We
don't
want
to
do
anything
before
alpha,
so
that
will
always
be
our
Gator,
but
I
know
that
Google
is
trying
really
hard
to
to
make
alpha
hit
and
112
or
right
right
after
112
a
possible.