►
Description
#sig-cluster-lifecycle
#capn
#capi
A
All
right,
I
believe
this
is
recording
now,
oh
there
it
is
hey.
This
is
the
tuesday
morning
cluster
api
provider
nested
office
hours.
This
is
as
always,
going
to
be
recorded
and
posted
to
youtube
later
today,
so
don't
say
anything
you
wouldn't
want
actually
shared
with
the
whole
world
and
we're
going
to
get
started.
A
B
One
thing
is
kubicon
you,
I
think
market
tennessee
working
group
has
a
panel.
It's
it
used
to
be
a
kind
of
session
for
the
reserve
session
from
working
group,
but
this
year,
tasha
plans
to
do
a
kind
of
a
panel
kind
of
thing.
Instead
of
just
you
are
you
know,
describing
the
project
progress
so
yeah.
B
Besides
that
they
write
a
you
know,
a
document
they
plan
to
put
to
the
cncf
blog
post
about
marty
hennessy,
because
I'm
not
sure
if
you
know
that,
so
I
think
it
it's
major
cloud.
B
Cto
mentioned
the
whole
style
multi-tenancy,
it's
kind
of
there's
a
term
looks
very
confusing
yeah,
especially
because
his
assumption
is
that
there's
no
multi-tenancy
requirements
within
our
organization,
which
seems
not
to
be
true
even
because
so
we
would
like
to
as
a
working
group
without
you
writer,
you
know
post
to
explain
the
multi-tenancy
terms
and
why
we
need
it
and
so
taxonomic
taxonomy
about
the
market
tendency.
Basically
I
I
I
see
they
have
a
draft
post.
B
I
think
I
I'm
going
to
give
a
link
to
you
guys,
chris,
if
you
have
time
just
take
a
look
they
describe.
You
know
we
see
as
a
cluster
as
a
service
margin
tendency
which
seems
okay.
Well,
just
just
take
a
look
about
the
content
to
see.
If
you
see,
if
you
think
that
any
every
right,
he
needs
to
be
improved
or
is
okay,
just
say
as
it
is.
Let
me
let
me
give
a
link.
A
Yeah,
if
you
can
drop
that
into
the
the
agenda
doc
under
the
psas,
I
just
put
a
little
line
item
for
the
taxonomy
doc.
That
you're
talking
about
and
and
blog
posts
is
what
you
said.
That
is.
The
plan
is
to
send
you
to
blog
post.
A
Thanks
for
adding
that
that
sounds
fantastic,
so
just
just
asking
for
reviews
of
that
and
any
changes
to
any
of
the
verbage
and
how
and
potentially
how
we're
doing
multi-tenancy
yeah.
B
A
Double
check
all
right
sweet.
I
appreciate
that.
Thank
you
for
the
for
the
panel.
What's
actually
happening
with
that.
B
You
know
trillionaire
panel,
you
have
to
have
people
from
outside
of
the
working
group
at
least
two
or
three
or
even
female
preferred,
but
this
year
again
is
is
it
is
like
you
know,
multi-tenancy
progress
report,
but
doing
in
a
form
of
panel,
don't
show
slides,
say
everybody
do
what
just
give
a
high
level
it's
kind
of
it's
kind
of
this
described
a
decision
decision
tree
like
what,
if
you
want
this,
if
you
want
this
kind
of
isolation,
what
do
you
want
to
go
if
you
want
that?
What
do
you
want
to
go?
B
It's
like
a
decision
tree
describe
description,
so
they
want
to
change
the
form,
basically
making
it
more
interesting,
instead
of
just
just
showing
the
slides,
nothing
more,
let's
just
just
trying
to
use
a
different
form
to
describe
what
I've
been
doing
and
what
are
the
recommendation
and
there
are
a
bunch
of
qa
in
the
end,
but
I
doubt
you
know
in
the
online
version
the
qa
q
a
always
not
that
good,
because
the
interactive,
the
tools
is
not
very
good
for
interactive
presentation
anyway,
but
let's
see.
A
Okay,
cool
all
right,
if
there's
anything
that
comes
out
of
that
for
any
of
us
here,
just
let
us
know
sounds
like
it
sounds
like
it'll,
be
an
interesting
effort.
Yeah.
A
No,
that
makes
sense
cool
and
out
of
curiosity,
did
you
end
up?
Having
any
did
you
submit
anything
to
eu
for
virtual
cluster
anything
around
there
yeah.
A
A
Yeah,
it
would
be
good
to
it
would
be
good
to
collaborate
and
see
what
we'll
see
what
we
can
do
there,
especially
if
we
can
do
something
between
between
companies.
I
need
to
obviously
talk
on
our
side
and
make
sure
that
would
be
possible,
but
yeah.
B
My
my
current
is,
I
have
a
thought
is
maybe
v2
or
just
we
I
can.
I
we
can
write
a
blog
post
or
something
about
the
mighty
cluster
thing,
because
I
I
think
I
briefly
talked
to
you
about.
I
feel
our
model,
assuming
that
you,
you
only
have
one
tenant,
but
you
can
actually
multiple
super
cluster
right.
This
is
another
way
of
doing
the
federation.
B
A
Okay,
yeah
just
keep
us
in
the
loop
I'll,
add
an
I'll
add
a
light,
a
line
item
to
this
agenda
as
well
as
you're
working
on
that,
where
you
need
help
where
you
need
reviews.
Anything
like
that,
I
think
pretty
much
anybody
on
here
would
be
would
be
helpful
for
that
cool.
That
sounds
fantastic
all
right,
so
we
don't
have
ciao
yet
and
that's,
I
guess
all
right.
I
just
wanted
to
check
in
on
where
things
were
I'll
share.
A
A
Yeah
and
we've
got
one
of
the
controllers
done
and
we
got
two
two
I
think
maybe
in
progress,
but
I
don't
know
I
haven't
seen
anything
yet.
Let
me
just
jump
over
and
share
chrome,
real,
quick.
A
All
right.
Can
you
all
see
captain
somebody
can
just
give
me
a
thumbs
up
or
something
yeah.
Yes,
yeah
all
right,
cool,
sweet,
so
yeah,
so
we
do
have
a
milestone
coming
up.
A
We're
probably
gonna
have
to
push
this
as
much
as
something
since
since
we're
one
week
out
just
to
be
clear
on
where
we
are,
but
we
basically
have-
and
we
didn't
even
actually
have
everything
categorized
under
here,
but
well
here
I'll
go
back
to
milestones,
so
we
basically
have
our
1.0
1.0
that
we're
trying
to
aim
for
the
first
release
of
this
that
that
might
actually
have
a
working
version.
A
I
do
need
to
do
need
to
do
some
updates
in
here.
So
we
did
finish
the
we
got
the
the
pr
merged
travis
pr
merged
on
here,
so
I'm
gonna
actually
close
this
out
for
std.
A
Awesome
awesome
that
then
yeah
we
can
probably
put
something
together
for
for
the
controller
manager
one,
so
we
at
least
have
something
in
place.
I
don't
it's
not
a
full
working
cluster,
but
well
that
actually
would
be
a
full
working
cluster.
If
we
have
all
three
of
those
might
come.
A
Yeah
is,
is
that
actually
what
chaos
gonna
be
so
chow
assigned
himself
to
this?
So
I
didn't
know
if
he
was
picking
it
up.
A
All
right,
I'm
gonna,
unassign
child,
then
and
then
make
sure
that
that
that
we
get
this
one
done
instead,
instead
of
doing
the
controller
manager,
one
because
I
guess
unless
you're
saying
that's
the
second,
the
second
controller
that
he's
working
on.
B
A
A
It's
all
good,
no
one
was
actually
assigned
to
it
directly,
so
yeah.
C
Will
talk
to
chris
yeah
about
that?
Let
chris
and
coordinate
all
these
things.
A
Yeah,
let's,
let's
let's
talk,
if
not,
I
I
probably
have
a
little
bit
of
bandwidth
that
I
can
throw
at
this
to
get
something
up
really
quickly.
Since
I
wrote
the
doc,
I
could
probably
just
just
run
it
together
real
quick.
I
just
want
to
make
sure
that
nobody
else
has
already
started
something
and
then
I'm
going
to
stomp
on
somebody's
toes.
If
I
do.
B
A
No
worries
take
your
time.
It
there's
really
no
pressure.
Here
I
mean
if
we
have
to
end
up
pushing
it
pushing
this
date.
That's
fine!
I
don't
think
that
it
really
hurts
anybody.
It's
just
a
it's
just
to
try
and
put
pressure
on
us
to
get
something
out:
okay,
cool!
So
let's
look
at
the
other
pieces
that
are
in
that
milestone
that
might
block
us
from
being
able
to
do
anything.
So,
let's
see,
we've
got
one
done.
I
think
this
can
be
closed,
because
this
is
all
documented.
A
Now
I,
this
is
good,
we're
actually
going
and
cleaning
these
things
out
and
continuing
down
through
this
list
with
the
high
level
user
stories
that
was
all
documented
in
the
ncp
doc
and
apparently
didn't
have.
It
wasn't
tagged
properly
on
my
pull
request,
it
didn't
actually
close
yeah.
I
just
checked
house
report.
A
Awesome,
okay,
cool,
then
wayne
I
can
chat
or
anybody
else
on.
This
call
can
chat
actually,
since
kasich
and
dt
both
work
with
with
us,
and
we
can
figure
out
if
we
can
get
something
out
quickly.
Let's
see
so,
we've
got
create
nested
controller
types
and
control
planes,
so
those
those
are
identical
things
and
I'm
gonna
dupe
this.
I
think,
what's
the
first
one
in
here.
A
That
I
don't
know
why
we
have
two
of
these
things
all
right
and
then
going
back
to
milestones.
We've
got
so.
I
need
to
go
update
this
dock,
but
that's
not
going
to
stop
us
from
being
able
to
release
a
1.0.
That's
just
dock
updates
yeah.
We
need
to
create
the
controller
for
the
nested
contr
nested
cluster
as
well.
That
should
be
a
really
simple
controller
to
make
yes
and
then
we've
got
base
level
architecture
yeah.
So
these
all
just
didn't
get
close
with.
A
With
closing
that
pr
might
be
closer
than
I
thought.
All
right.
Chow
definitely
already
did
all
this,
so
this
should
have
definitely
been
closed.
A
Going
back
did
that
not
close,
yet
there
we
go
okay,
so
we
need
the
nested
api
server
which
chad's
definitely
working
on.
We
have
the
updates
on
the
dock,
which
is
not
super
important,
but
we
can
we
can.
We
can
add
that
in
here,
otherwise
we
can
just
punt
it
if
we
get
to
the
1.0
at
that
point,
yeah
and
then
control
plane,
control
plane
did
this
not
actually
get
so
that
one.
C
A
A
The
I
think,
that's
probably
why
oh.
A
A
C
A
A
Milestones,
okay,
api
server
controller
manager,
and
now
I
need
to
go
back
and
open
up
33..
A
A
So
now
we're
looking
at
the
milestone-
and
we
have
everything
prioritized
except
for
I'm
gonna
punt
this,
because
it's
not
that
important
to
move
in
this
milestone
and
I'll
put
it
into
that
1.x
release
all
right.
So
we
just
have
a
couple
of
pieces
that
all
seem
like
they
have
a
sineys,
I'm
going
to
throw
this
one
in
there
as
well
and
kosher
could
diy.
We
can
all
tell
we
can
all
chat
about
how
we
can
get
these
done.
So
we
can
move
this
forward.
A
Cool
all
right,
so
that
puts
us,
so
our
plan
is
basically
in
by
march
31st
we're
on
march,
so
we
got
like
seven
whole
days
or
eight
days
until
that
actually
including
some
weekends
yeah.
So
let's
just
keep
trying
to
work
towards
that
and
see.
If
we
can
see
if
we
can
get
something
in
place,
that
we
can
actually
demo
and
and
work
with,
and
then
we
can
start
scheduling
what
what
is
next
to
actually.
C
A
C
I
have
a
quick
question
for
faye
sure,
so
I'm
working
on
do
some
stress
tests
on
a
multi-tenancy
code.
Also,
I
try
to
use
a
class
loader
which
can
launch
thousands
of
pod
and
all
different
resources
on
the
multi-tenancy
virtual
clusters.
C
B
B
C
B
Yeah
yeah,
but
we
have
a
paper
just
accepted,
so
I
think
you
can
see
the
paper
if
you,
if
you
want
to
see
the
paper
everything's
described
in
the
paper,
how
we
did
the
experiment,
you
can
you
have
you
you,
you
should
be
able
to
get
get
an
idea.
How
we
do
the
experiment
is.
C
Okay,
so
the
the
standard
tool
we
we
try
to
use
is
a
called
a
class
loader,
which
is
a
but
the
cluster
doesn't
have
a
lot
of
matrix
already
built
in
so
quite
some
materials
that
get
from
the
scheduler.
But
we
don't
have
a
scheduler
on
the
virtual
cluster.
So
I
have
to
change
a
little
bit
the
code
so
that
the
class
loader
can
get
the
important
pod
creation
time
pod
kind
of
a
scheduled
time.
B
C
A
lot
of
a
host
here,
but
the
problem
is
that
we
have
enough
hosts
here,
but
we
we
really
can
launch
it
using
this
all
the
major
tools
that
get
all
this
detailed
part
status
for
each
each
stage
like
a
scheduler
scheduled
and
the
finished
computer.
B
C
Okay,
do
you
want?
Okay,
do
you
want
to
publish
this
piece
of
code
if
I
did
some
changes,
no
wow.
B
C
Just
want
to
see
the
the
pod,
for
example,
I
launched
a
10
000
pack.
I
don't
know
how
quick
the
product
can
be
scheduled
on
the
super
cluster
from
virtual
cluster
I
launched
ten
thousand
part
and
how
quick
this
pod
can
be
eventually
scheduled
on
the
super
cluster
and
how
quickly
it
can
be
deleted.
So
this
kind
of
dynamics
I
want
to
measure
a
little
bit.
C
With
without
using
kcs
yeah.
B
B
Okay,
you
just
check
the
timestamp.
That's
the
way,
that's
the
way
that
we
did
so
we
we
want
to
even
break
down
that
that
latency,
so
that
that's
what
we
did
in
the
paper,
but
for
you,
if
you
just
want
to
know
the
sinker
latency,
every
just
just
just
check
the
creation
time,
the
difference
between
the
part
creation
time
that
you
have
temp
stamp
in
the
object.
C
B
C
Yeah,
we
need
to
capture
all
the
delays,
because
the
two
time
step
is
different.
One
is
a
from
the
super
when
it's
from
virtual,
then
we
can
correlate
these
two
parts.
The
part
name.
B
A
C
So
which
conference
you're
going
to
publish
the
paper?
I
see
dcs
five
cd
says:
oh
actually,
okay,
this
year.
C
C
B
A
C
C
Also,
and
also
very
quickly,
you
said,
you
use
a
virtual
couplet
to
do
these
things
so
virtual
cooperate
you
installed
on
the
supercluster
or
you
installed
on
the
super
super
cluster.
C
I'm
doing
some
testing
putting
the
virtual
class
virtual
cooperate
in
the
virtual
cluster
so
doing
some.
Some
testing.
B
A
Cool
that
sounds
awesome.
Is
there
anything
else
that
anybody
wanted
to
bring
up
going
once
going
twice
all
right,
so
we
can
break
now
and
then
basically
just
meet
up
next
week
and
we
can
get
those
controllers
moving
forward,
sound
good
everybody,
yeah,
okay,
cool
all
right,
y'all,
later
bye,.