►
From YouTube: Kubernetes SIG Federation 20170620
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
Hi
everyone-
this
is
June
20th.
This
is
being
recorded,
sig
Federation
meeting.
My
study
was
slight
delay,
not
too
sure
why
we
just
started
appearing
I
think
delay
on
our
end,
but
they
were
at
the
point
of
discussing
one,
a
pirate
even
before
we
do
that
on
the
video.
On
our
end,
we
have
a
new
person
in
room
jesse.
Key
generally
does
recently
he's
always
with.
B
The
youtube
project
so
I'm
a
a
new
transitioning
manager
for
Federation
and
multi
cluster,
along
with
storage
as
well
and
Michael
Rubens
working
with
him
on
kind
of
Shane
over
here
on
the
Google
side,
I've
been
at
Google
for
about
nine
and
a
half
years
on
G,
suite
and
enterprise,
and
before
that's
number
of
years
elsewhere,
I
think
about
16
years
in
industry,
whole
stack,
backends
front
end,
you
name
it
so
learning
a
lot.
Looking.
B
A
So
in
terms
of
agenda,
we
want
to
start
first
of
all
so
I'm
going
to
suggest
whatever
outstanding
one
seven
items
and
does
not
with
us
today.
What
he's
trying
to
do
right
now
is
take
down.
What
is
a
regression
has
appeared
as
a
regression
since
Friday
night
in
the
CCI,
a
constant
integration
test,
we
kind
of
puzzling,
what's
going
on
so
far
over
the
last
four
or
five
failures
they
have
not
been
Federation
related
or,
however,
it's
still
on
our
on
our
plate
to
prove
that
this
is
not
the
case.
A
A
C
I
think
I,
probably
just
surprised
everybody
with
a
bunch
of
detail.
If
anybody
has
any
thoughts
about
what
I
suggested
or
talked
to
it
last
week,
I'd
be
happy
to
hear
them
either
now
or
email
or
ever
anybody
else
wants
to
provide
feedback.
Do
you
think
we
have
way
too
much
work
to
do
and
the
hard
part
isn't
coming
up
with
stuff
to
do
it's
figuring
out
what
is
actually
most
important
to
tackle
this
cycle
and
I?
Think
that'll
have
to
be
a
collective
effort
to
try
to
nail
down
what
the
priorities
are.
C
A
D
This
is
where
people
that
memory
I
think
are
mistaken,
that
the
alphas
are
a
bit
gothic,
so
just
want
to
circle
back
on
the
on
the
the
spillover,
a
couple
of
items
you
know
from
one
point:
seven:
the
jobs
and
the
HP
80
I
just
wanted
to
find
out
where
the
artifact
we
were
almost
set
to
go
that
still
ending
to
be
merged
store.
We
haven't
started
emerging
yet
code.
C
D
C
Freeze
happens,
replica
sets
and
then
we
can
start
iterating
on
HP,
a
okay
as
far
as
jobs
goes.
I
think
that
it's,
my
understanding
is
that's
pretty
close.
I
would
like
to
make
the
requirement
that
we're
going
to
merge
jobs,
but
a
PR
author
will
be
responsible
for
moving
jobs
to
sync
controller.
Like
that.
Yes,
such
a
requirement,
we
can
merge
it
before
that,
but
it
has
to
be
in
like
every.
D
D
D
So
I
said
the
few
comments,
not
a
bunch
of
comment
from
you
and
a
fun.
He
said
he
replied
to
that.
You're
stateful
said
you
know
the
filtration
creative
staple
set
as
well.
So
I
guess
that's
going
to
be
so.
I
heat
cannot
come
back
and
present
it
next
week.
That's
I
think
that's
what
we
planned
I
did.
C
E
C
Sets
to
China
in
they're
not
entirely
sure
that
some
of
the
some
of
the
changes
related
to
global
quorum
are
going
to
be
in
keeping
with
the
design
philosophy
of
stateful
sets
I
think
there's
certain
sort
of
distributed
systems-
fundamentals
the
staple
sets
tries
to
achieve
and
I
don't
see.
It's
not
clear
to
me
that
the
proposal
has
written
today
will
actually
maintain
that
same
sort
of
capability
characteristic,
whatever
so
yeah.
D
C
Like
to
see
I'm
hoping,
we
can
communicate
on
that
on
review
in
advance.
I,
don't
I
think
there's
enough
holes
in
the
design,
but
I
don't
really
want
to
see
it
reviewed,
as
is
I
want
to
at
least
like
I,
want
at
least
somebody
whether
it
be
me
or
Clayton
or
somebody
else
above
it,
you
know,
have
a
reasonable
expectation
that
there's
no
major
flaws
in
the
design
and
then
have
people
go
through
and,
like
you
know,
if
there's
some
corner
cases,
we've
nest
well.
A
C
Yeah
I
mean
we
talked
touching
that
briefly
last
week,
it
escapes
me
as
to
whether
there
was
any
consensus
that
that
proposal
would
cover
sort
of
the
aggregate
status
view
where
there
was
just
access.
The
underlying
resources,
like
just
exposing
the
API,
basically
I'm,
going
to
refresh
my
memory,
did
I
forget
that
we
were
going
to
do
something
about
that
or
would
abused
talk
about
it
and
there
was
no
action
item.
Other.
D
C
I
went
to
the
document
before
the
meeting
last
week
and
just
noticed
that
it
seemed
to
be
focused
on
just
allowing
access
at
the
API
level.
Oh
maybe
the
thing
was
that,
like
it
seemed
to
me
that
there
really
was
a
need
for,
like
a
user
experience,
discussion
find
the
user
and
I
want
to
see
the
status
of
like
my
replica
set
and
see
like
where
the
replicas
are
distributed
across
which
clusters,
I,
probably
don't
want
to,
have
to
make
queue.
C
D
C
Allow
is
one
thing,
but
actually
exposes
the
users
another
I
guess.
My
concern
was
that
it
was
narrowly
folks.
He
did
be
narrowly
focused
on
enabling
access
to
the
underlying
clusters
without
actually
providing
user
experience
and
I'm
kind
of
like
Paul,
that
how
is
it
actually
useful
if
the
user
can't
really
use
it.
D
C
A
D
D
E
If
we
think
we
can-
and
it
seems
reasonable
to
use
this
time
to
do
that,
but
if
we
don't
think
we
can,
it
doesn't
seem
worthwhile
to
do
a
bunch
of
prioritization
and
then
we
can
come
back
and
redo
the
prioritization
when
we
could
either
use
this
meaning
for
something
else
or
just
have
the
time
back
to
do.
Work.
E
C
Be
able
to
do
the
new
stuff
in
the
now
family,
not
something
that
makes
sense.
These
are
something
about.
We
need
everybody
to
do
or
like,
and
we
just
have
people
go
in
and
like
remove
the
items
that
are
actually
completed.
We
have
a
completed
section
or
do
we
just
delete
them?
No,
no
probably
makes
sense
to
have
a
completed
section,
but
like
I,
just
I'm,
just
not
sure
we
need
everybody
involved.
If
something's
done
move
it.
If
it's
not
done,
leave
it
there
and
we
can
take
into
consideration
ordering
priority
I'm
inclined
to
agree.
G
G
Altar
is
designed
for
or
the
refrigerated
recipe
using
external
EMS
provider.
We
have
commutation
Matt
go
ahead
that
we
had
proposed.
In
addition,
an
additional
mechanism
has
to
contain
alternative,
and
we
spoke
about
it
in
this
meeting
two
weeks
ago
and
I
send
something
out
over
the
mailing
list,
a
while
back
so
I
don't
know
if
people
and
chanced
it
so
the
number
of
people
have
reviewed
it,
but
maybe
Christian.
You
want
warning
on
that.
A
H
J
A
C
I
think
it's
hard
to
just
sort
of
go
like
I'm
going
to
end
and
say
priority
is
this
without
considering
some
of
the
other
things.
To
my
mind,
staple
sets
is
certainly
useful
for
certain
percentage
of
users,
but,
like
I,
think
of
cube
without
staple
sets.
It's
useful
for
a
lot
of
things
and,
if
cube,
didn't,
have
a
lot
of
the
products
ization
nice,
these
around
it
staple
sets,
would
not
be
priority.
D
System,
Baba,
prospectively,
believe
II
spend
lot
of
cycling
back.
You
know,
and
this
is
kind
of
boring
from
our
perspective.
It
is
going
to
hide
because
the
customer
requirement
as
well
but
yeah.
So
so
that's
that's
what
even
we
went
to
a
lot
of
receptacles
to
buy,
go
through
put
together
a
design
document
in
all
this
thing,
I
guess.
C
The
question
would
be
like
at
least
from
the
right
outside
I'm,
seeing
like
a
need
for
a
lot
of
sort
of
foundational
stuff
so
that
we
can
prioritize
and
deliver
something
to
customers
that
we
can.
Actually
you
know,
support,
upgrade,
etc,
but
I
guess
my
question
would
be,
and
maybe
this
is
a
question
more
for
Clinton,
but
like
it's
just
you
as
well,
we
not
have
these
same
products.
Ization
concerns.
How
are
you
supporting
in
maintaining
like
Federation
in
production
today
for
your
customers
without
some
of
the
things
that
I've
been
talking
about?
Videos.
C
D
The
other
thing
is
also
from
our
perspective.
This
is
a
very
high
kind
of
a
priority
to
quality,
but,
as
I
see
other
items
with
the
list
that
we
need
to
kind
of
aim,
how
does
we
gonna
stack
up
against
those
items?
So
I
mean
I
I,
don't
want
to
tempt
them
bad
perspective,
because
the
time
of
time
and
effort
we
have
put
in
so
for
keeping
that
in
mind,
I
would
consider
off
our
standpoint.
E
D
A
Okay,
hi
I
weigh
in
on
the
next
one
version.
Scale
I
think
that's
I
go
to
ordinary
was
saying
about
Foundation,
and
we
just
kind
of
started
touching
on
this
in
the
last
cycle.
Is
we've
mostly
assume
that
people
deployed
in
Federation
a
long
time,
whatever
the
latest
version
is
of
kubernetes
put
in
a
world
where
you
have
cluster
admin
which
are
loosely
connected
to
to
to
I?
Guess
individual
operators
of
communities
dispense
being
more
of
an
issue,
so
I
think
it's
pretty
high
priority
for
anyone
who
wants
to
privatize
a
version
of
Federation.
C
A
You
know
I've
been
pretty
immersion
school
as
a
major
issue
and
through
you
sort
of
tried
to
try
to
get
while
I,
don't
to
explain
kind
of
what
their
perspective
is
with
respect
to
staple
debts
and
how
they
do
upgrades
I,
don't
know.
Maybe
Quinton
will
provide
a
better
or
more
complete
view
on
this
and
what
what
it
is
for
the
customers.
L
C
I
would
say
this
next
one
by
a
medium,
but
I,
don't
think
it's
very
hard,
so
I
would
expect
this
to
get
in
1a,
but
this
basically
means
that
I,
if
I
have
a
deployment
with
three
replicas
I
will
get
a
deployment
three
replicas
in
every
cluster
that
is
targeted
rather
than
trying
to
spread
it,
which
there's
I
think
that's
a
use
case.
We
have
for
some
customers,
so
it
should
be
pretty
trivial
to
implement
film
just
like
a
flag,
basically
on
the
Federated
object.
M
M
C
M
N
N
A
Simple
mantra
like
Korea,
so
I
think
it
would
underscore,
though
that
is
it's
very
feature.
Request
kind
of
speaks
to
the
fact
that
there's
some
somewhat
of
a
hole
and
in
the
features
provided
by
some
of
these
workload.
Types
in
that
you
know.
We
have
cluster
selectors,
that
you
can
deploy
over
selected
clusters
and
you
can
tell
the
user
go
directly
to
the
underlying
API
server,
which
kind
of
defeats
the
purpose
of
Federation.
And
then
this
one
again,
a
third
one
which
has
kind
of
fallen
through
the
cracks.
C
C
Many
things,
but
is
believing
I
mean
I
think
this
is
definitely.
This
is
going
to
be
important
for
some
read
our
customers
not
entirely
I'm
not
entirely
going
to
push
it
like
it's
a
high
priority
upstream,
though,
because
I
think
it's
something
we
could
probably
do
downstream
for
the
most
part-
or
you
know,
do
a
little
bit
of
work
upstream
or
down
if
we
have
to
and
given
that
there's
a
lot
of
other
high
priority
items,
I
don't
really
walk,
but
this
is
high
priority
either.
So
I
would
say
like
a
medium.
L
C
Right
I
thought
that
what
we
had,
maybe
when
I
when
I
met
you
guys
in
May
we're
talking
about
having
some
preliminary
work
where
you
would
be
watching
namespace
by
namespace,
or
at
least
you
were
going
to
think
about
doing
that,
so
that
didn't
actually
no,
no
good.
Okay,
that
should
definitely
be
on
the
list.
I
think
that
CC
I
mean
I
would
call
it
maybe
I,
don't
know
partition
Federation
or
something
like
that,
because
your
your
federating
selectively,
rather
than
just
like
all
resources.
C
Yeah,
please
put
that
in
on
the
list.
I
think
I
think
that's
important
I,
don't
think
that
I
think
that
there's
certainly
some
use
cases
where
you
just
want
to
be
able
to
cut
array
everything
and
then
there's
other
use
cases
where
you
want
to
selectively
do
that
and
not
give
full
access
to
the
underlying
cluster
I
think
it's
primarily
like
a
security
thing
and
being
able
to
partition
access
to
member
clusters
between
Federation
control
and
non
Federation
controls.
O
O
O
C
O
O
C
Maybe
we
need
another
line:
item
which
is
just
like
documented
pay
back
or
documented
back
I
mean
I,
know
it's
available
for
queue,
but,
like
I,
have
seen
some
people
asking
about
off
this
week,
saying
like
how
do
I
give
somebody
access
to
my
Federation?
Do
I
just
have
to
avoid
shift
them
like
you
can
take,
which
they
thought
was
kind
of
suboptimal,
I.
C
C
C
P
C
P
P
O
P
O
C
O
Be
fair,
though,
like
I
say
a
year
ago,
that
might
have
been
the
case
today.
Most
people
who,
through
strap,
who
are
doing
anything
other
than
just
kicking
the
tires,
will
then
just
set
up
with
a
correct,
slash,
delegated
off
solution.
It
was
probably
fairly
few
people
who
aren't
using
delegated
off
this
way.
O
P
A
I,
just
like
from
user
discussions,
we've
had
there's
kind
of
two
dimensions
to
it.
There's
the
we
Feder
eight
are
back
in
that.
Do
we
install
the
same,
are
back
rules
allowing
clustered
with
they're
controlled
by
the
Federation
control
plane,
or
do
we
support
a
federated
are
back
only
in
the
Federation
roller
and
there
has
been
faster.
So
we
want
individual
rules
on
some
requests
so
just
widely
copying
them
with
the
thing
is
already
available
without
without.
M
And
actually
it
depends
on
what
we
want
to
do
by
this
point.
We
could
use
the
same
controller
to
the
words
that
cluster
selected,
so
they
bind
for
utilities
configure
on
the
in
the
line
clusters,
but
you
can
add
a
custom
silikal
so
that
it's
on
the
fourth
edition-
and
we
continue
just
physics
in
controller
to
do
so.
C
M
A
Okay,
low
accelleration
on
a
tree,
I
guess
had
good
stewards.
Somebody
should
be
looking
at
this.
There
are
other
groups,
they'll
be
moving
out.
Of
course,
there's
no
there's
no
reason
the
Federation
would
be
in
the
core,
the
only
the
only
reason
right
now
think
Ria's
or
goal,
and
a
lack
of
kind
of
effort
and
time
for
people
to
move
it
out
and
yeah
gigantically.
A
L
C
Let's
say
low
at
least
four
one
eight.
That
would
be
my
suggestion.
I
think
that
the
the
other
thing
that's
related
to
this
is
how
we
decide
to
decide
a
federation
preferences
if
we
decide
to
just
continue
using
queue
types
in
a
federation
control
plane.
That
sort
of
is
a
very
different
approach.
Then
we're
going
to
actually
create
federation
specific
type
so
that
we
can
have
federation
specific
fields
without
having
to
mess
with
cube
the
cube
native
types.
C
M
Think
they're
active
they're
a
few
advantages
we
get
by
moving
available
like
we
can
add
another
one
on
it
like
now,
I
think
this
requirement
is
going
to
be
contributing,
for
you
have
to
be
part
of
the
on
forty
month
and
that's
why
we
cannot
add
existing
sig
members
as
so
it
goes
to
our
own
s5.
But
if
you
are
other
people,
he
can
pull
flexibility
of.
M
Adding
that
and
since
I
know,
sick
CLI
is
planning
to
add
cube
credit
out,
but
the
API
machinery
team
is
also
working
on
adding
or
removing
the
API
types
or
desertion.
Clayton
Kimberly
would
prophesy
no
more,
but
I
think
they
are
already
they've
already
moved
a
lot
of
this
amount.
So
the
dough
already
people
who've
done
yeah.
O
We
definitely
do
not
underestimate
the
scope
of
this.
It
is
easier
than
it
was
I
think
because
of
the
depth
of
the
Federation
control
playing
overlapping
code.
You'll,
like
service
kind
of
luck,
definitely
took
the
hit
on
some
of
the
controller
stuff,
and
some
of
the
API
server
stuff
with
Federation
also
touches
a
lot
more
core
code
and
reuses
a
lot
more.
That
would
be
the
biggest
worry
active
engine
like
the
namespace
deletion
controller
code,
for
example,
yeah.
C
Going
to
say
this
is
low,
but
perma
product
is
Asian
perspective
being
able
to
actually
have
you
know,
maybe
a
web
UI
or
something
interact
with
the
configuration
of
the
controller
manager
without
having
to
parse
like
the
template
and
try
to
figure
out
which
parameters
are
being
set.
I
think
this
would
be
important
but
I'm
going
to
say
lo
for
now
we
can
get
resources
on
it
and
it
becomes
a
priority
for
Red.
Hat
is
part
of
characterization
and
we
might
about
that.
M
C
J
C
So
I
mean
especially
the
high
priority
items,
I'm
hoping
we
can
start
working
on
them
like
this
different
max
honestly
I
mean
obviously
Quinn
has
to
chime
in.
But,
to
my
mind,
like
the
three
four
high
priority
items
here,
like
I,
think
those
are
pretty
critical
and
I.
Don't
think
we
should
waste
any
time
in
getting
started
and
sort
of
you
know
design
discussion,
preliminary
implementation.
The
goal
is
to
get
these
done
well
before
the
code
freeze
as
they
impact
a
lot
of
the
other
work.
C
You
know,
and
it's
going
to
guarantee
something
to
higher
priority
and
large.
It's
going
to
take
a
lot
more
time
than
we
actually
think
so.
Optimism
doesn't
really
have
a
place
here
as
much
as
I
like
to
assume
it
does
I
guess
the
action
item.
Everybody
has
something
in
the
backlog.
If
you're
the
owner
or
something
in
the
backlog
we
reviewer,
maybe
Lincoln,
make
sure
can
we
just?
Can
we
make
a
section
like
down
lower,
that's
like
done
or
something
or
do
we
just
want
to
delete
it?
C
And
that's
right
because
it's
already
and
it's
still
in
queue
17q
to
town
okay.
So
if
you
have
something
that's
in
that
list
and
is
done,
please
delete
it
and
maintain
things
going
forward.
If
you
finish
something
with
over
I,
think
that
should
be
kind
of
like
by
grooming.
This
list
of
things
should
probably
be
like.
The
first
thing
we
do
are
one
of
the
first
things
we
do
each
meeting
just
so
that
we're
clear
on
where
we're
at
I
think
you
yeah,
okay,
yeah,.
D
L
C
Yes,
I'm
hoping
we
actually
have
time
to
iterate
on
that
I'm,
hoping
that
you
Clayton
you
have
a
chance
to
take
a
look
at
it
as
well
and
maybe
even
come
to
the
Sun
review.
I
can
sort
of
understand
what
global
core
means
with
your
star,
like
how
global
correlates
to
local
core
and
whether
we
can
relax
some
of
the
constraints
that
were
put
on
staple
sets
when
I
when
they
let
its
federated
and
the
way
that's
being
proposed
anyway.
Yeah.
So
we'll.