►
From YouTube: Kubernetes SIG Federation 20170417
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).
B
B
B
C
Yeah
I
mean
my
bus
edition
was
just
to
clarify
that
you
can
see
there,
whatever
nine
things
only
already,
and
hopefully
we're
going
to
be
documenting
even
more
than
us,
so
just
breaking
it
down
into
a
couple
of
standard
sections
that
we
cover
each
week,
we'll
each
each
meeting.
My
peers,
who
I
can
do
that
the
other
suggestion
I
had
maybe
while
we're
on
the
topic
of
note,
is
just
a
particular
darkened-
is
getting
extremely
an
unmanageably
long.
C
D
C
I
think
I
mean
the
table
of
contents
is
definitely
got
unmanageable
because
of
the
length
of
it
at
the
moment,
but
I
spares
me
find
it
quite
useful
to
be
able
to
go
back
to
you
know
a
given
place,
I
guess
in
theory,
you
can
search
for
it,
but
cannot
pretty
consistent
updates
and
things
Oh
once
I
shrunk
it
down.
I
can
see
if
it's
a
problem
and
it
is,
we
can
remove
the
total
abbreviated
or
something
like
that.
E
F
There
isn't
much
to
talk
about
it.
The
CI
tests
are
green.
We
are
working
on
three
subnets.
We
have
turned
on
pace
of
mix
on
us.
One
then
I
cook
for
week
intentionally
do
not
start
reporting
on
the
PR
yet
as
an
unbeatable
page,
because
we
did
not
want
to
cause
unnecessary
anxiety
among
non-federation
developer,
because
we
are
still
not
very
confident
that
these
things
are
green
on
all
the
peers.
F
G
G
F
So
I
don't
know
exactly
what
you
have
in
mind,
but
I'm
going
to
my
cover
much
as
possible,
a
national
everything
that
I
missed
I.
Guess:
ok,
so
for
like
at
the
beginning
of
the
previous
quarter
to
one
we
started
out
with
the
goal
of
making
a
post
submits
blocking
because
we
thought
that's
all
to
the
network.
F
So
today,
if
you
go
to
submit
your
dashboard,
you
see
a
bunch
of
blocking
post,
submit
jogs,
posta,
Mita,
suites
and
Federation
test.
Suite
was
pretty
much
completely
broken
at
the
beginning
of
q1,
so
we
fixed
a
whole
bunch
of
infrastructure
issues
and
by
the
end
of
q1
we
were
in
a
pretty
good
shape,
actually
at
the
beginning
of
q2.
F
So
we
went
to
the
testing
fra
forks
and
ask
them
if
he
could
make
different'
job
the
CIA
job
for
submit
job
blocking,
and
they
said
they
did
not
want
to
add
any
more
blocking
jobs
to
see
I,
because
their
goal
was
to
remove
all
the
blocking
jobs
from
the
sea
I
blocking
list.
So
the
goal
is
still
to
remove
all
these
see
I
jobs
and
make
the
place
amid
shock
jocks
blocking
for
all
dps.
So
the
idea
is,
if
we
can,
let
appear
in
them.
F
That
means
we
should
pass
all
our
tests
and
things
should
work,
so
they
asked
us
to
go
back
and
make
a
place
of
my
jobs
blocking
and
we
stop.
This
was
about
two
weeks
ago
and
we
started
working
on
making.
Please
submit
jobs,
blocking
I
fixed,
a
bunch
of
infrastructure
issues
worked
with
Eric
Twitter,
the
FSA,
and
now
we
have
a
picnic
job
which
is
mostly
green
on
alternate.
Oh,
that's
where
we
are
romina
I,
don't
know
what
I
guess
so
I.
Let
you
add
the
details.
G
Suites
that
you
would,
then
you
know
create
free
or
sig,
and
it
would
lock
you
when
they
broke
at
this
point.
I
think
that
they're
abandoning
that
idea
or
pushing
away
from
that
idea,
and
so
the
strategy
that
we
had
had
Madhu
just
outlined
of
coming
up
with
a
federation
set
of
tests
that
would
lock
the
queue
when
they
stopped
working.
G
C
G
G
Let
me
try
to
tell
you
so
I
think
that
I'm
still
learning
about
this
too,
but
I
believe
that
a
the
CI
tests
are
ones
where
you
can
have
more
than
one,
but
basically,
when
you
have
more
than
one
PR
and
there,
when
you
do
a
PR
blocking
one,
it's
basically
this
one
PR.
If
this
PR
doesn't
work,
then
that
PR
is
not
submitted.
The
COI
test.
I
G
I
B
Declare
I
see
I
guys
are
the
ones
I
turn
post.much
they're
learning
one
head
to
making
sure
like
get
all
the
days
pass
on
head
and
provide
that
before
the
pr
even
was
it
so
obviously,
p
monsters,
salad
better.
Even
before
merging
a
PR
you
find
out
if
it's
making
tests
see
I
months
after
merging
with
on
them
and
make
sure
that
the
display
so
that
Isaac
testing
once
more
be.
A
G
E
G
Is
independently
tested,
but
even
if
we
have
the
pre
it
looks
like
what's
going
to
happen,
is
that
we
are
still
going
to
need
in
this
sig
somebody
like
a
build
cop
who's
going
to
notice
things
that
occur
after,
let's
say,
soak
testing,
or
you
know,
issues
that
have
come
up
that
are
more
subtle
that
possibly
break
the
test
in
head.
After
that,
I
was.
G
That
of
these
tests
are
running
in
and
I'm
trying
to
work
on
that,
because
I
think
that
is
a
not
cool
and
be
not
scalable.
You
know
a
it's,
not
cool
in
the
community
sense
it
shouldn't.
We
shouldn't
have
anything
that
relies
on
only
Googlers
being
able
to
maintain
it,
especially
for
open
source
I
mean
what's
going
to
this
world,
and
so
I
think
these
are
two
open
issues.
G
C
Would
love
to
have
him
do
that,
because
I
then
everything
you
described
has
always
been
the
case.
We've
always
had
a
subset
of
C
at
what
I
will
call
sea
access
which
are
automatically
run
tests
which
needs
plastic
or
PR
Deathmatch,
and
then
we've
always
had
another
conservative,
like
150
/,
mentioned
circus
disruptive
to
separate
its
run
outside
of
the
pre
segment.
G
F
C
C
F
C
B
Now,
if
it
won't
be
much
did
one
water
molecule,
then
we
will
need
a
human
to
look
at
it
like
a
fit
great.
We
won't
know
that
it's
broken
and
more
p.
Oz
will
keep
on
getting
worse
and
like
that
s
can
even
group
it
can
regress
even
vote.
So
that's
why
we
want
someone
to
locate
that
Israel
and
make
we
can
have
little
irreverent
break.
Someone
gets
an
email
and
a
second.
C
G
C
G
C
G
G
B
F
B
C
C
G
Think
Quentin,
please
don't
apologize.
This
is
this
stuff,
even
for
all
of
us
who
sit
right
next
to
each
other
and
are
talking
to
the
Google
people.
This
is
confusing
and
it
led
to
a
lot
of
back
and
forth.
So
I'm
really
happy
that
we're
getting
broad,
broad
agreement
on
what
the
heck
is
going
on
in
this
meeting.
Please
please
ask
everyone,
ask
questions
and
make
sure
they
understand
this,
and
especially,
you
know
what
we
don't
understand.
What's
going
on,
because
it's
as
possible
we're
missing
stuff.
C
G
C
Just
FYI
was
not
directly
related
to
just
say
that
we
have
a
proposal
into
the
CNC
x2
still
cross
the
MTF
project,
testing
environments
and
that
would
be
sort
of
hosted
by
the
sea
and
CF
and
willing
to
play
the
games
like
testing.
You
notice,
Prometheus
against
there's
different
versions
to
deny
dependent
on
bond.
So
this
is
some
other
left
with
them.
So
cool
and.
G
B
B
Is
something
I
clatter
discussing
with
Christian
as
well?
Whenever
a
test
fix
then
I,
like
two
main
reasons?
Why
fix
one
is
if
we
ran
out
of
good
as
you
are
there
something
wrong
in
the
test
infa
if
we
need
to
go
in
the
project
in
ulsan
come
on
for
that
we
need
Googlers,
but
the
other
test
messages
from
one
merge
the
code
and
because
of
that
does
the
regression
and
that
so
it
is
broken
and
for
the
second
one
we
can
non.
Google
is
because
lossless
inaccessible,
so
non
google
is,
can
still
go
on.
B
That
is
good,
see
it
broke,
they
can
look
at
the
logs
figure
out.
This
is
a
completely
are
in
just
about
it
and
even
for
quotations
they
can
look
at
the
logs
find
out
where
it
is
because
of
quota
and
if
they
can
bring
any
one
of
the
Googler
that
can
you
run
this
come
on
that
also
works,
but
it
needs
there
is
the
big
o
who
finds
out
what
is
the
problem,
and
it
was
the
solution.
I
think.
G
It
makes
sense
to
start
with
the
build
cups.
We
don't
start
this
with
the
community
and
have
you
know
the
build?
Comp?
Rotation
include
Googlers
and
non
Googlers
so
that
when
it
ends
up
being
a
quota
issue,
then
the
the
build
competition
on
google
just
kicks.
You
know
a
googler
and
then
we
can
look
into
it,
but
just
like
I
think
it
was
makyo
who
said
that
earlier
you
know
for
the
cases
that
aren't
you
know
they
can
address
them.
Yeah.
C
So
I've
got
a
proposal
to
kick
this
off,
which
is
that
we
have
one
preferably
non
Googler
as
the
primary
build
cop
there's
a
you
know,
approximately
once
a
week
exercise
that
you
kind
of
come
back
to
this
group
approximately
once
a
week.
Can
you
make
sure
that
everything
is
unblocked,
or
at
least
you
have?
You
know,
action
items
to
get
the
stuff
working
and
we
have
that
rotation
so
to
you
on
for
about
a
quarter.
C
The
reason
I
suggest
that
is
because
I
mean
then
that
person
is
a
bit
like
mother
has
been
the
kind
of
point
guard
for
q1,
and
maybe
somebody
else
wants
to
do
q2,
and
if
you
have
a
single
person
for
a
quarter,
then
at
least
you
get
some
kind
of
consistent
progress
and
they
can
fix
systemic
issues
rather
than
just
reactively
dealing
with
individual
pickle.
Our
items,
because
some
of
these
things
are
systemic
rather
than
10
fixes.
C
Rotation
we
have
to
work
on
the
middle.
The
night
is
if
the
circus
break,
and-
and
you
also
don't
need
to
get
some
understanding
for
you
if
you
get
lunch,
but
you
do
you
know
during
the
course
of
the
week
when
something
goes
wrong
or
I
deal
quicker
than
that.
You
do
need
to.
You
know,
do
something
about
it
and
he
does
needs
are
going
to
be
on
vacation
per
week.
You
need
to
get
someone
else's
for
your
job
for
the
week
away.
That's
that's
what
I
mean
sorry
I
wasn't
very
clear
about
this.
J
I
think
the
idea
is
especially
right
now
is
the
test.
Infra
is
still
not
necessarily
we
haven't
done
a
build
car
for
a
while.
So
you
don't
know
what
sorts
of
issues
are
going
to
come
up
having
one
person
do
it
for
a
long
time
means
that
that
person
can
deal
with
longer-term
systemic
issues
and
have
a
higher,
more
comprehensive
understanding
of
those.
So
then,
over
the
course
of
a
quarter
can
be
better
equipped
to
figure
out
how
to
fix
them.
We're
going
to
be
rotating
every
week.
You
have
just
do
the
log.
J
B
J
B
Away
like
once
a
test,
I've,
green
and
stable,
and
we
have
a
big
cut,
these
systemic
issues
would
be
a
lot
less
as
well
and
like
everyone
would
contribute,
and
even
the
bill
con
like
he'll
start.
Maybe
he
finds
out
some
issues
when
he
starts
and
then,
after
two
weeks
or
whatever
his
duration
is
he
passes
on
to
the
next
one.
B
J
A
J
D
Not
not
if
there's
two
responsibilities
here,
one
is
kind
of
looking
at
the
long-term
health
test,
wives
of
Federation
and
the
other
is
basically
fixing
day-to-day
like
acute
versus
chronic
like
problems
or
ever.
You
want
to
call
it
I,
don't
necessarily
think
those
need
to
be
the
same.
People
I
think
it
like
having
continuity
for
the
the
long-term
view
is
a
good
idea
and
then
having
as
many
people
as
possible
rotate
through
the
build
composition
like
if
you're
an
active
contributor,
you
probably
have
a
responsibility
to
see.
D
C
Now
that
makes
a
lot
of
sense.
In
theory,
I
think
it
practice
right
now.
You
know
what
actually
fails
is
a
test
so
that
you
know
this
is
Bailey
bake
up,
guys
going
to
see
broken
tests
and
then
he's
going
to
spend
some
time
diagnosing
and
then
he's
going
to
get
to.
The
fact
of
air
is
a
systemic
problem,
because
we,
you
know,
run
out
of
quota
or
we
don't
think
clusters
properly
or
whatever
best
stress
causes
these
tests.
Well,
then,
he
has
to
hand
over
all
that
space
to
somebody
else
who's.
C
Now
a
job
is
to
fix
those
things
and
I,
just
I.
Think
in
theory,
once
we
get
those
too
much
better
separated
when
a
test
fails,
it's
almost
exclusively
due
to
the
fact
that
that
some
PR
has
gone
on
the
trim,
they've
gone
and
that
caused
the
test
to
fail
some
systemic
there's,
some
problem
with
us
with
the
coup
benetti's
iteration
systems
is
the
cause
of
it.
Then
then,
what
you
say
makes
perfect
sense,
but
I
think
we're
quite
a
long
way
off
from
there
as
myself,
and
therefore
merging
those
two
roles
might
make
sense.
C
Systemic
problems
are
actually
do
tues
code
that
we
have,
for
example,
tests
that
are
not
properly
written
then,
who
you
know
who
has
to
go
and
fix
those
it's
difficult
to
tease
all
these
pieces
apart
today,
I
think
I
didn't
they
would
you
tease
the
part
in
the
future,
but
they
are
not
at
the
moment.
I
think.
E
That's
a
pessimistic
view,
I
think
what
we
worked
on.
First
quarter
of
Union
side,
the
quarter,
if
you
look
at
I,
mean
the
case
in
point:
a
tenet,
CI
CI
and
represents
probably
the
worst
case
scenario
and
running
all
these
tests.
In
parallel,
we
just
have
to
translate
that
infrastructure,
what
happens
in
SEO
into
the
blocking
tests
and
that's
what
Madhu's
been
working
on
and
we're
almost
there
so
the
build
cop.
B
I,
don't
want
that
the
booty
call
predation
guy
to
be
with
the
scapegoat
for
every
test
related
like
evenly
factoring
badly
written
tests
that
shouldn't
the
big
on
should
not
be
required
to
do
that.
It
should
be
the
author
who
wrote
that
and
because
I
should
chime
in
to
fix
all
of
those
things.
We
can
say
that
this
is
the
brick
oven
full
of
water.
He
is
fixing
everything
tessellated
consolation,
all.
B
E
I'll
just
say
it
is
a
skill,
at
least
knowing
where
to
look
in
that
pile
of
logs,
and
it's
somebody
that
everybody's
an
active
contributor
should
know
how
to
do
finding
a
failure.
At
least
we
initially
was
a
matter
of
like
half
an
hour's,
and
now
it's
probably
done
with
couple
seconds
when
you
know
where
and
how
to
search
logs.
F
All
so
I
think
we
should
take
a
page
out
of
Google
sorry
book
and
probably
put
this
in
the
big
rotation
responsibility
somewhere
that
the
bill
calls
were
there
for
a
week
as
a
big
copper
two
weeks
they
they
should
work
on,
like
I,
don't
think
they'll
be
spending
their
entire
time
looking
at
at
DCI
dashboard
or
the
pr
dashboard
and
fixing
things.
This
thing
shouldn't
happen
that
often
so
it's
probably
also
his
response
is
responsibility
to
leave
the
system
in
it.
D
That
kind
of
seems
like
a
tall
order
to
me
like.
Doesn't
me
wrong?
I
think
that
being
proactive,
you
know
is
great,
but
I,
don't
know
that
we
necessarily
have
everybody
on
Federation
team.
That's
like
yes,
I
have
all
the
free
time
in
the
world.
My
employers
are
housekeeping
to
do
now.
I'm
test
work,
no.
F
D
Guess,
like
I'm
all,
for
you
know
all
things,
testing,
I,
just
I
would
be
concerned
that
that
would
be
sort
of
a
disincentive
for
people
stepping
up
to
participate
if
we
make
it
more
work
than
it
absolutely
has
to
be
like
great.
If
people
were
to
step
up
and
do
that
doesn't
mean
aww
bless,
you
hear
the
requirement
might
be
a
bit
too
much.
Yeah
I
agree
with
you.
I'm.
C
Just
wondering
we're
kind
of
running
out
of
time.
Initially,
you
don't
have
a
separate
discussion
about
this
outside
of
this
particular
meeting,
or
she
only
do
that.
We're
gonna
start
off
with
a
put
two
rotations,
one
quarterly,
which
is
the
long-term
fix
guy
and
one
every
the
any
make
sure
the
tests.
They
agree
me
that
sort
of
approximately
the
at
least
the
first
milestone
if
we've
agreed
there
has
been
good
and
when
we
can
refine
it
further
with
separate
discussions.
B
D
It
basically
means
you
implement
a
type
adapter
for
like
a
type
like
fig
maps,
and
it
automatically
will
create
a
controller
and
run
integration
tests
and
there's
a
couple
pr's
up.
It
will
actually
enable
upgrade
and
e
2
e
tests
in
the
same
mechanism.
So
basically
you
implement
the
type
adapter
and
you
guess,
controller
and
you're
testing
pretty
much
done
for
free
what
it
means
very
basic.
But
it's
certainly
a
good
starting
point.
I
E
A
C
E
D
Not
type
specific
it'll
run
type
specific,
just
be
cheaper
to
maintain
what
it
makes
sense,
so
there
shouldn't
be
a
loss
in
coverage.
That's
certainly
my
goal,
I'm
not
trying
to
say.
Oh,
we
don't
need
to
run
this,
we're
going
to
run
the
life
cycle
of
every
single
type.
Any
we're
going
to
continue
to
do
that
description
says
or
we're
just
going
to
stop
validating
things
actually
have
nothing
to
do
with
actually
interacting
with
the
underlying
clusters,
which
is
a
couple
of
the
cases
involving
Felicia,
okay,
right.
C
There,
the
next
okay,
huge
benefit
of
this,
is
that
you
know
if
you
run
and
how
you
mentioned,
that
you're
running
reasons
for
proposing
to
run
the
integration
tests
in
parallel
with
and
dentists
one
one
other
possibility
would
be
to
run
them
before
the
intern
test
much
faster
and
if
they
well,
then
you
don't
need
to
remediate
is
because
you
need
to
have
a
better
already
and
that
could
claim
you
know
substantially
on
the
intent
testing
time
because
essentially,
wouldn't
need
to
run.
Nonetheless,
integration
test
passed.
D
Hi
would
be
supportive
of
that
I
think
the
general
reason
that
people
don't
do
that
is
because
maybe
I
don't
know
I,
don't
how
long
integration
tests
actually
take
the
unit
integration,
job
versus
e2e
I've
seen
cases
where
they
would
run
them
in
parallel.
Just
because
maybe
the
integration
tests
only
takes
five
minutes
to
run
or
the
job
takes
five
minutes
to
Rhonda.
That's
five
minutes.
You
could
be
attempting
to
run
into
each
other
yeah.
C
D
C
This
is
how
the
kinetics
testicle,
if
you
know
right
from
the
beginning,
is
that
we
had
unit
test
integration
tests
and
end-to-end
tests,
and
you
didn't
bother
running
the
integration
system.
This
unit
is
fast
and
he
didn't
run
father
enemy
team
tasting
this
integration
test
passed,
and
you
know
this.
He
makes
me
not
quickly.
You'll
see
how
it
gets
rejected
quickly.
No.
D
B
Want
to
make
one
big
point
regarding
demon
sites
as
discussing
it
with
Christian.
We
now
support
ruling
updates
of
demon
sets
and
covetous,
and
we
need
to
add
that
support
refrigeration
as
well,
and
if
you
do
that,
since
the
newly
adapted
does
not
support
extension
points,
yet
maybe
we
should
wait
for
updating
demon
sense
to
use
the
adapted
a
down.
As
learning
points
is
to.
D
B
D
D
H
D
B
D
Don't
think
that
there's
a
huge
amount
of
work
so
I
mean
you're
talking
about
replacing
existing
damage
that
controller
with
its
equivalents
and
getting
integration
you
each
house,
so
the
advantage
of
doing
this
now
is
you're
not
really
wasting
effort.
You're
going
to
build
on
that
when
you
want
to
add,
like
update,
rolling
updates
in
court
and
you've
already
done
the
work
to
enable
the
integrate
that
faster,
you're
going
to
have
to
obviously
extend
it
to
testing
as
well.
I
B
D
K
K
D
The
basically
the
next
step
is
actually
rewriting
the
adapter,
so
it's
testable
and
like
it's,
which
to
me
means
that
it's
modular
and
out
that
you
can
unit
testing
the
nice
elation
and
because
things
are
separable,
it
should
be
relatively
straight
towards,
at
extension,
points
to
add
things
like
scheduling
the
rolling
updates
of
court,
but
in
by
just
taking
the
existing
sync
adapter
and
trying
to
do
that
is
I.
Don't
think!
That's
really
know.
D
Right
right,
but
I
mean
the
the
goal
of
extending
that
is
not
actually
like
modifying
the
adapter
having
a
custom
version
of
it.
It's
actually
just
it's
configuring
it.
The
goal
here
is
having
like
a
generic
framework
for
synchronization
that
can
be
configured
with
either
types
or
functions
to
specialize.
Yes,
but
not
actually
overriding
and
implementing,
like
a
new
version
of
that
controller
that
make
sense
yeah.
B
So
in
so
the
next
item
on
the
agenda
was
clip.
Diversions,
cute
and
I
saw
a
lot
of
discussion
on
the
missing
list
and
mendacious
so
I
try
to
summarize
everything
that
was
being
discussed,
and
so
Quinton
replied
this
morning.
Israel
and
I
think
this
is
something
I
totally
agree
with
an
hour
to
make
sure
with
everyone.
Other
people
agree
as
well.
B
C
Think
that's
the
whole
point,
sorry
is
that
the
clusters
themselves
could
be
managed
by
anybody,
and
there
are
many
cases
where
they
explicitly
will
be
managed
by
different
parties
different.
Then
you
know
different
for
each
cluster
and
different
than
the
Federation
control
claim,
for
example,
as
a
multinational
organization
which
ends
in
a
regional
data
centers
run
by
the
head
of
regional
data
centers.
C
B
J
C
Kind
of
the
members
that
so
the
difference
is
clearly
in
a
cluster
management
situation.
You
have
a
mastering
have
nodes
and
there's
a
rule.
That
is,
the
master,
must
be
ahead,
I'll,
go
equal
to
the
british
of
the
nerds
and
that's
that's
reasonable
and
across
the
management
scenario,
because
they
typically
have
someone
managing
the
cluster
and
they
can
coordinate.
Amongst
you
know,
there's
more
than
one
person
I
can
coordinate,
but
the
core
multi
clusters
multi
cloud
provider.
Is
it
saturates?
It's
just
not
feasible.
You
know.
C
B
So
I
played
that,
instead
of
discussing
with
what
requirements
we
can
input,
we
decide
that
we
cannot
and
instead
discuss
on
what
happens
if
there
is
a
version.
Skillet
motives
of
you,
especially
the
second
point.
Yet
what
happens
if
a
particular
features
not
supported
in
the
interline
listed?
What
should
we
do
this
and,
as
I
was
like,
they
were
proposed
it's.
What
we
do
right
now
is
little
best
effort,
but
we
can
give
using
a
choice
that
maybe
we
should
generate
an
error
and
user.
Some
bits
are
replicas
set.
B
The
example
given
goes
for
scheduling,
impede
if
the
user
submits
a
replica
set
with
that
we
set
and
filtration
knows
that,
like
not
all
clusters
support
that
feat,
then
it
should
prob'ly
generate
an
error
at
that
time.
Oh,
it
should
do
a
best
effort
that
for
clusters
that
support
that
field,
it
creates
a
bigger
set
with
that
weak
side
and
for
those
that
doesn't
it's
like
today,
it
drops
them
and
if
they
charge.
J
A
fee
of
proposals
to
me
know
how
much
work
this
will
take
the
sound
white.
This
could
be
a
nightmare
of
matrix
cases
because
I
have
no
other
cluster
that
does
isomers
requested.
It
doesn't
support
something.
I
have
a
bugfix
original
cluster
that
had
to
plug
in
something
so
I
can't
use
it
and
be.
The
number
of
cases
here
seems
God,
multitudinous
and
like
this
could
just
be
chasing
chasing
things
all
over
holders,
but
probably
did
not
do
have.
B
G
I
I
really
worried
that
a
generic
mechanism
is
going
to
be
hard
to
to
maintain
in
reality
across
in
a
production,
environment
and
I
basis,
just
on
Google.
Basically,
what
we
do
is
we
have
like
a
master
that
we
then
say
supports
the
last
I
think
to
two
or
three
revs
and
versions
of
the
nodes,
and
this
seems
to
be
a
very
similar
situation
and
even
doing
just
that
with
those
constraints,
it's
created
a
lot
of
problems
when
things
fall
through
the
cracks
and
genetic
answers
don't
always
seem
to
to
fulfill
all
of
the
requirements.
G
B
C
For
something
worth
mentioning
here,
which
is
that
the
two
scenarios
are
quite
fundamental
difference
in
a
sense
that
there
isn't
actually
a
formal
API
between
the
nodes
in
the
master
and
a
lot
of
the
stuff
is
kind
of
well
I,
guess
to
lesser
extent.
There
is
that,
but
it's
much
less
strict,
API
aggressor
nation
then
occurs
at
the
cluster
API
level.
So.
C
The
cluster
Federation
knows
that
there
is
a
cluster
knows
precisely
which
version
the
cluster
is
running,
which
version
of
the
master
and
commends
just
behave
as
if
it
is
dealing
with
a
cluster
of
that
version,
and
we
already
have
backwards
compatibility.
I
believe
so
that
again,
this
is
fundamental
to
the
design
of
cuban
Eddie's.
E
A
good
in
the
in
the
queue
cold
case.
You
talk
about
a
client
that
submitting
just
kind
of
a
fire-and-forget
type
of
update
in
our
scenario
where
we
have
two
controllers:
running
potential,
different
versions
and
trying
to
sync
I.
Don't
know
what
comes
to
mind
if
I
get
dinner
boots
and
all
sorts
of
that
dear
yeah,.
J
And
some
sort
of
like
gigantic
testing
behemoth
where
like,
if
we're
going
to
say
this,
is
how
it
works,
and
then
we
don't
test
it
well,
then
you
can't
see
the
teller
works.
I
mean
now
we
have
to
have
tennis
against
like
what,
if
you
have
a
1.5
1.4
cluster
deficit.
So
here
what
16
1.3
cluster
with
crap
like.
B
J
D
We
testing
doesn't
have
to
be
I,
need
a
full
cluster.
We're
just
talking
about
the
API
here.
Not
is
not
actually
that
hard
to
validate
being
so
long,
but
good,
hi,
tiny.
Sorry.
I
sorry
too
kind
of
interrupt
this
though,
but
we're
really
at
a
time-
and
I
want
to
take
this
opportunity
to
suggest
that
we
start
needing
more
right,
more
awesome.
We
still
have
stuff
to
talk
about,
and
we
don't
have
any
time
to
talk
about
it.
D
E
I
would
say
yes
also
because,
for
example,
not
this
wednesday
next
wednesday
is
features.
Repo
free-
and
I
know
you-
people
have
something
they
intend
to
heaven,
17
that
are
not
in
the
features
recall
and
the
powers
that
be
in
17,
look
down
on
people
who
don't
pay
attention
to
their
features
people,
so
they.
D
C
I
I
G
D
Week
can
we,
you
know,
do
every
week,
but
can
we
also
meet
again
this
week?
We
find
time
to
try
to
get
through,
I
mean
I,
guess
my
thought
is
we
want
to
cover
we're
gonna,
get
it
the
sense
of
all
the
things
that
we
want
to
do,
and
now
we
have,
let
me
actually
have
to
make
sure
there's
going
to
be
features
proposed
for
any
of
the
things
we
require
it
for
wednesday,
which
doesn't
suggest
meeting
next
week
and
having
a
couple
days
to
scramble
and
get
feedback.
E
C
C
D
I'm,
a
beyond
annotations
I'd
like
to
at
least
have
people
talk
about
that.
I
know
that
I
haven't
done
a
good
job
in
a
document
really
laying
everything
out
cohesive
way.
So
maybe
I
can
do
some
work
on
that,
but
I
really
don't
think
we
continue
to
to
just
say:
oh
yeah,
we're
just
going
to
use
annotations
for
things
like
policy.
D
I
think
we
reach
need
to
consider
alternatives
or
potential
future
paths,
and
if
there
is
something
that
we
can
do
and
17,
ideally,
we
would
get
to
work
on
that,
rather
than
just
kicking
down
the
can
further
down
the
road.
Maybe
my
concerns
overblown
and
maybe
I'd
be
happy
to
be
told
that
but
I
think
ignoring
it.
That
isn't
a
good
strategy.
B
D
J
J
I
guess
if
there's
anything,
that's
really
pressing,
you
can
consent
emails
about
it
to
the
list,
and
hopefully
sync
up
about
they.
Never
over
email
that
are
very
important,
yep
I
guess
features
may
be
the
most
important
thing.
It
might
be
worth
doing.
Some
like
talking
about
that
a
little
bit
over
email,
there's
things
that
people
want
to
worry
about
there
so
that
we
can
get
started
and
make
sure
enough
time
to
get
things
ready
for
the
deadline.