►
From YouTube: 20190730 sig cluster lifecycle
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
A
C
A
A
A
C
D
D
Sorry,
yes,
go
really,
sir,
for
releasing
binaries
specifically
for
cluster
API
projects
and
potentially
other
Singh
cluster
lifecycle
projects.
People
are
interested.
I
found
it
to
be
a
really
good
tool
for
auto-generating
and
release
notes
and
building
binaries
cross-platform.
You
can
take
a
look
at
that
link
and
you
can
see
it's
being
used
in
capti.
Right
now
integrates
really
well
with
github
actions.
D
We
can
automatically
print
out
a
a
change
log
and
get
all
your
binaries
uploaded
in
a
nice-looking
release,
and
all
you
have
to
do
is
tag
the
repo
with
an
annotated
tag
and
push
that
which
is
also
just
good
practice
for
code
modules.
Overall
I
had
a
really
good
experience
with
it
and
I
think
it
would
be
useful
to
use
across
the
board
we
and
by
we
I
mean
I,
rewrote
something
like
this
for
Kappa,
but
I
think
you
could
entirely
replace
it
with
Co.
D
A
Really
like
go
release
her
one
of
the
things
that
might
be
beneficial,
I.
Think
we
don't
actually
do
this
as
part
of
our.
You
can
have
templates
on
your
when
you
do
on
your
github.
Commit
messages
because
go
releaser
doesn't
really
respect
the
issued
template
stuff.
It
might
be
beneficial
to
have
a
standard
template
for
your
commit
messages,
because
that
would
then
publish
the
informational
bit
cleaner
for
go
release
er,
so
that
way,
you'd
have
a
go
release
or
page
that
has
like
a
unified,
feel
or
format
for
that
are
the
release
notes
so.
D
I
will
say
we
definitely
don't
really
have
that
right
now,
with
any
of
the
projects
we've
got
for
capo,
we
have
release
notes
that
we
generate
by
hand
by
looking
at
commit
messages.
At
least
what
go
release
will
do,
for
you
is
pull
in
all
the
commits
that
are
in
your
release,
and
then
it
will.
Let
you
go
in
and
edit
the
release
do
they
get
of
UI.
So
if
you
don't
like
the
way
go
release
your
looks,
you
can
absolutely
modify
it.
Yeah.
A
D
A
B
A
There's
no
there's
no
edicts
and,
in
fact,
you'll
find
everything.
That's
outside
of
the
main
KK
repo
is
a
is
a
wonderland,
a
tapestry
of
solutions.
I
think
what
I
might
ask,
though,
would
be
to
like
document
the
best
practices
potentially
implement
in
the
community
repo
underneath
cluster
sit
cluster
lifecycle.
Oh
that's.
C
F
Presumably,
we
can
contribute
fixes,
like
I,
have
my
own
version
of
this,
but
based
on
your
endorsement,
so
I'm
like
I
should
try
go
really
sorry,
but
there's
something
I'm
I
would
rather
like
it
feels
like
I
should
just
like,
send
a
PR
to
go
release
it
rather
than
like,
be
like.
Oh,
it
doesn't
work,
pay
so
I'm
gonna
like
throw
it
away
and
continue
working
on
my
own
right
right.
B
A
Well,
it
seems
a
little
weird.
Cuz
basil
is
just
basically
like
a
glorified
build
system,
so
you
know
you
can
shout
out
to
do
anything,
so
it
seems
like
a
be
like
a
final
shell
out
statement
for
a
build
process
like
if
I
were
to
call
make
release.
You
know
it'd,
be
like
the
final
step
of
these
process,
where
you
basically
shells
out
to
an
executable
so
and
I
think
documenting
this
all
would
probably
be
a
good
thing.
A
G
G
G
A
And
once
twice
their
dues,
the
next
topic
was
coop.
Con
North
America
I
wanted
to
get
a
read
out
real,
quick
with
regards
to
some
projects
that
wanted
to,
and
hopefully
you
can
edit
to
the
notes
who
are
the
people
that
are
interested
in
working
on
the
talks.
There's
usually
1%
project
I
think
the
general
take
was
from
last
meetings.
A
Is
that
unless
you
actually
have
a
release,
yet
we're
not
gonna
try
to
get
a
meeting
venue
for
that
until,
but
we
will
prognostic
me
lead
that
this
is
a
effort,
that's
ongoing
in
intro
sessions
and
maybe
even
give
you
some
limelight
to
discuss,
but
otherwise
for
the
developer
session.
It's
a
little
weird
because
you
want
to
talk
about
where
you're
at
and
how
people
can
get
involved
and
where
the
binaries
are
and
stuff
like
that.
A
A
Maybe
on
because
component
standard
and
add-ons
are
not
police,
yet
maybe
it
makes
sense
to
and
the
same
person's
doing
both.
Maybe
it's
a
makes
sense
to
and
unloved
them
together.
We
does
that
drive,
or
would
you
like
to
do
them
separately,
because
it's
a
little
weird
because
work
as
a
cig
I
have
to
like
have
this
long
conversation
with
Nancy
about
how
we're
structured
and
that
we
have
mini
talks
versus
some
other
sig.
That
only
have
a
couple
I
figured.
H
The
add
ons
thing
would
just
be
a
demo
for
as
a
subset
of
what
we
talked
about
during
the
COO,
video
and
conversations
and
then
so
like
we
don't
need
a
huge
section
or
session
for
add-ons
I
just
figured
oh,
this
is
a
feature
we
want
to
get
in
and
you
know
there's
anticipated
things
there.
So
unless
we
want
ahead,
have
a
contributor
session
for
that
I,
don't
think
we
need
something
the
the
component
config
thing
is.
H
F
H
I
think
something
that
would
take
a
marginal
step
forward
from
what
Lucas
did
in
China.
He
has
a
talk,
but
I
can
link,
but
it's
it's.
H
F
Because
the
the
other
thing
is,
if
it's
more
discussions,
then
we
could
have
one
room
which
might
make
Nancy
happier
with,
for
example,
just
to
pick
like
component
standard
add-ons
at
CD
manager
will
sort
of
like
gather
in
there
and
like
pick
your
sub
circle,
rather
than
saying
we
need
end
rooms,
we
can
always
ask
for
end
rooms
and
then
settle
for
n
over
two
yeah.
A
A
I
know
that
this
this
has
been
very
prolific
on
cluster
API
and
it
makes
sense
for
him
and
he's
been
asking
for
a
while
for
him
to
give
a
talk
on
cluster
API
I,
don't
know
if
they're,
usually
we
like
to
have
two
representatives,
where
possible,
from
different
companies
to
give
talks.
So
we
know
we
want
to
be
as
as
open
and
transparent
as
possible
to
make
sure
that
we're
doing
having
a
vendor
specific
venue
to
talk
about
these
things
should
I
solicit
the
cluster
of
geography.
E
H
So
enhancements
freezes
today
and
I
don't
have
the
kept
up
for
the
kuba
diem
integration
with
the
add-on,
installer
library,
I
did
publish
the
pr
two
weeks
ago,
just
showing
how
that
could
be
used,
but
yeah
it
hasn't
received
any
traffic
or
feedback.
I'm,
not
sure
if,
like
that
might
be
a
blocker
for
the
116
cycle
are.
I
Yes,
I
just
worried
that,
depending
on
what
we
decides
about
this
particular
area
of
sequester
psycho
I
worried
that
we
maybe
should
not
mention
mechanics
that
are
still
not
approved
in
the
public.
Maybe
we
can
present
some
options
during
the
Canadian
talk
with
regards
photos,
but
we
should
note
remote
the
mechanics
for
the
users.
A
We
talk
about
them
in
the
context
of
how
do
we
enable
them
sometimes,
but
most
of
the
time
when
we
go
do
deep,
dives,
we
don't
usually
prognosticate
album
features
unless
their
VIP
high-profile
stuff,
like
a
che,
is
a
good
example.
It's
been
asked
for
so
many
times
the
problem
I
think
with
and
users
can
kick
the
tires,
but
there's
no
guarantees.
I
I
think
we
can
do
a
similar
thing
for
add-ons.
We
do
have
to
make
sure
like
we
prognosticate
that
you
know
there
is
a.
Is
there
an
actual
release?
Yet,
though,.
H
No
there's
a
POC
that
just
shows
that
it
works
but
yeah
no,
no
release
and
no
existing
integration
in
Covidien.
Why.
A
I
Be
there
Fabricio
fabrizio
promise
that
he
wants
to
take
a
deeper
look
at
this,
so
I
can
discuss
it.
I
personally
think
that
maybe
we
can
mention
some
options
in
front
of
the
public,
but
I
don't
want
the
comedian
basically
session
to
turn
into
a
bike
where
you
know,
users
propose
different
solutions.
A
I'm,
just
going
to
I'm
officially
gonna
put
the
stance
on
that
I'm
going
to
talk
with
Antoine
bug
spray.
They've
they've
not
been
showing
it
to
give
readouts
here.
I,
don't
know
if
any
other
coups
great
folks
are
in
Nepal
I,
don't
see
Chad
either
so
I
want
to.
You
know
you
got
to
be
part
of
the
sig
and
work
with
them
to
Sagan.
You
want
me
ten
to
help.
Do
all
the
TPS
reports,
for
you.
A
Everything
else
seems
legit
I,
don't
think
sed.
Adm
actually
has
released
yet.
But
we'll
talk
about
that
during
the
intro
session.
I
do
plan
to
submit
an
intro
session.
I'll,
probably
be
the
the
giver
of
the
intro
session.
I'm
gonna
talk
with
Lucas
and
see
if
he's
gonna
be
able
to
attend
or
not.
If
he
wants
to
help
give
that,
maybe
it's
possible
that
Justin,
Lucas
and
myself
can
do
the
transit
of
the
intro
session.
I
So
we've
got
some
complaints
from
slack
for
people.
That's
we!
So
some
folks
are
subscribed
to
our
sequence,
life
cycle
main
mailing
lists.
We
actually
only
have
one
nowadays
and
basically
I'm
proposing
that
we
create
a
separate
one
it'll
back
in
the
day
we
used
to
have
sequence,
life
cycle,
failing
tests
or
something
like
that
and
I'm
proposing
that
we
create
the
new
one,
because
apparently
people
are
currently
subscribed
to
the
main
mailing
list
again
and
a
lot
of
spam,
because
our
qadian
tests
are
unfortunately
a
bit
flaky.
A
F
A
F
A
A
There's
a
million
ways
to
filter,
like
you
can
filter
from
whose
the
sender
and
everything
else
I
think
this
particular
one
like
it's
not
a
lot
of
spam,
maybe
once
a
week,
sometimes
it
gets
high,
moderate
modest
volume,
but
it's
not
a
terrible
amount
of
spam,
like
I
think
the
highest
count
day.
That
I
see
is
about
four
emails.
A
I
A
I
think
one
thing:
that's
good,
there's
there's
this
trade-off.
That
you'd
have
to
way
is
like
spam
versus
broad
visibility
right
and
if
you
have
a
separate
list
and
it's
outside
of
the
normal
purview
as
if
people
rotate
in
and
out,
they
will
invariably
the
communication
would
be
lost
and
they
won't
necessarily
know
that
this
being
spam
to
the
second
list,
or
that
they
should
subscribe
to
that
list
right.
A
A
lot
of
people
don't
even
know
that
they
should
subscribe
to
list
today
right
as
new
contributors,
they
view
slack
as
the
the
de-facto
way
people
communicate.
So
what
I'm
worried
about
there
is
information
loss,
I
would
say
if
it
becomes
a
real
issue
like
if
you
got
like
20
30
people
complaining
about
it,
I
think
we
can
entertain
the
idea,
but
I
think
for
the
time
being,
people
can
always
set
up
a
filter,
I
just
filtered
away.
As
just
a
mentioned.
Okay
sounds
good.
A
Once
twice
three
times
all
right,
I
didn't
put
it
in
the
meeting
notes,
but
we
should
do
the
sub-project
updates,
so
I'm
gonna
copy
stuff
from
below
and
hopefully
put
it
up
here
about
little
mirrors
typing.
Oh.
A
B
C
H
B
A
K
Heat
update
we
walked
this
week
to
get
the
feature
for
116
approval,
kept
and
and
announcement
issue
there
is
already
appear
see
for
customized
with
could
mean
before
someone
want
to
take
a
look
and
also
we
had
bug
in
a
coup
in
Phi
discovery
for
cubed
min
join.
That
was
physics
attend
and
charity
Kadeem
in
the
last
release
in
the
in
115
and
114,
and
we
are
a
mess
with
you.
G
A
G
We
are
we're
using
Hugo
with
the
the
doxy
theme.
It
is
basically
an
elephant
very
similar
to
how
kind
and
other
sub
projects
have
done
their
deployment
for
so
project
documentation.
Right
now
it
is
a
the
documentation,
will
just
reflect
the
version
at
head,
but
we
do
have
the
capability
of
swapping
between
versions,
I.
A
E
E
A
H
Yeah,
so
all
I
had
for
the
add-on
section
here
was
just
to
mentioned
proof
of
concept
from
the
Covidien
bit
feel
free
to
take
a
look,
it's
very
small,
but
it
does
add
a
new.
It
adds
an
API
tech
for
the
convictions
up
now
and
then
with
the
component.
Sorry
Justin
did
you
have
anything
I
missed
last
week's
meeting
because
I'm
just
back
from
vacation,
yeah.
H
That
criticism
so
next
moving
on
a
component
config
and
component
standard,
this
one
is
actually
quite
interesting.
Just
a
note
of
kept
about
basically
making
it
possible
to
know
whether
or
not
a
field
was
actually
provided
as
an
intent
from
whatever
system
you
know
created
the
config,
and
this
gets
into
a
bunch
of
really
ugly
API
machinery
and
go
UX
stuff.
H
This
is
just
a
really
interesting,
an
interesting
discussion,
so
Justin's
proposal
is
to
use
pointer
fields
for
everything
and
that
that
means
you
know
it's
a
trade-off
in
usability
and
it
might
be
a
little
bit
of
a
gnarly,
fine
and
replace,
but
it
certainly
certainly
is
a
way
to
do
things
that
will
work
and,
it
seems,
like
we've,
got
some
good
kind
of
moving
consensus
on
this.
At
the
moment
we
came
up
with
some
other
weird
ideas
that
also
don't
don't
really
feel
great.
Yeah.
H
A
I'm
definitely
interested
and
I'll
I'll
take
a
look
through
the
kit,
hopefully
this
week,
so
TLDR.
If
you
want
me
to
look
at
something
just
as
a
general
rule
of
thumb,
the
this
year
amount
of
spam
I
get
from
github
is
uncomprehensible,
I
think
as
a
general
pattern.
If
you
want
other
people
to
look
at
it,
just
assign
yourself
or
assign
another
person
to
take
a
look
at
it.
A
A
Next
up,
as
we
avoided
a
while
ago,
to
create
the
image
builder,
we
have
started
to
collect
the
image,
be
lighting
utilities
into
a
the
first
process.
That
motion,
I
kind
of
agreed
upon
was
first
collect
them
all
into
single
repository
and
then
to
start
to
try
to
produce
the
state
space.
So
right
now
we're
currently
in
collection
mode.
We
have
the
config
ATM
builder,
that
Moshe
had
created,
and
we
also
have
the
the.
A
M
C
M
This
priority
io
priority
to
be
serviced,
and
that's
that
adding
some
flags
so
I
think
one
of
the
things
that
we're
going
to
try
to
do
in
in
the
very
near
term
is
settle
on
on
a
autoconfig,
we'll
probably
be
looking
to
the
component
config
for
some
some
ideas
and
and
so
forth,
but
so
so
so
that
the
flags
don't
need
to
proliferate.
As
people
are
asking
for,
there
are
more
knobs
to
just.
A
M
Am
talking
yeah,
it's
it's
I
guess
there
are
there's
the
IO
IO
priority.
Sorry
I'm,
I'm
troubled,
exactly
off
the
top
of
my
head,
there's
r22
things
that
I
can
recall.
I
owe
priority,
which
is
which
can
be
set
with
the
I.
Oh
nice
command
and
think
that
in
the
dots,
that's
all
that
that's
shown,
but
I
think
there
is
I.
Think
there
is
also
something
something
else.
But
I
can't
I
can't
think
about
that.
M
A
I
think
my
only
comment
would
be
like
avoid
see.
Group
based,
IO
limitation
select
the
plague.
They
have
been
fraught
with
peril
since
epoch
and
there's
been
a
number
of
pick,
well-written
papers
and
all
the
other
accoutrements
that
are
associated
with
it.
You
can
get
yourself
into
a
forty
or
the
hurt.
L
M
Sort
of
trying
to
try
to
really
define
the
the
problem
better
last
week
and
didn't
didn't
wrap
that
didn't
wrap
that
up
I
can
move
America's
living
around
the
call
to
this
he
and
I
yeah.
Okay.
Here
we
talked
about
a
little
bit
it
the
yeah.
What
most
you
wanna
do
you
want
to
say
something
or
you
know
no.
M
Yeah,
so
the
the
issue
at
least
said
that
that
we
saw
it
was
actually
an
issue
that
we
felt
with
ku
ADM,
but
then
we
thought
okay.
This
is
this
will
also
apply
to
to
something
like
that
see.
The
ADM
is,
you
know
the
cluster
API
is
today:
we've
decided
that
we're
not
supporting
in-place
upgrades,
so
we
have,
we
have
replaced
upgrades
right.
M
A
See
where
you
go
you
this
is
like
the
mantra
mutable
versus
mutable,
in
my
pad,
where
the
who
owns?
What
like
the
distinguish,
the
distinguishing
feature
that
I
applied
to
this
is
that
cluster
api
is
strictly
immutable
by
definition.
If
you
want
to,
you,
can
create
an
operator
you'd
call
it
sed
an
operator
or
you
could
call
it
certain
Canadian
operator
or
something
like
that.
Those
operators
could
do
a
mutable
a
do
a
mutable
infrastructure
right
in
the
operator
stompin.
A
M
C
A
A
So
yeah
I'd
be
interested
in
reading
the
doc
or
happy
to
have
a
separate
conversation
that
they
have
everyone
here,
if
it's
kind
of
still
not
all
written
down
yet.
But
the
distinction
has
been
pretty
clear,
at
least
in
my
mind
and
talking
about
it,
especially
in
the
cluster
API
group,
as
we've
had
many
discussions
about
what
it
means
to
have
grid
but
I
think
having
that
written
down
somewhere.
M
Yeah,
that's
exactly
what
Y
is
what
what
I
was
trying
to
sort
of
distill
the
the
doc
into
and
that
yeseo
sort
of
defining
like
these
yeah
sort
of
the
boundary,
the
boundary
question
identifying
the
the
existing
tensions,
but
I
guess
I
I
still,
don't
don't
quite
understand
all
right,
all
right,
I
guess
you
know
one
of
the
tensions
that
I
see
is
that
for
something
like
like
cuvee
diem
right,
kuba
diem
is
what
we
you
know.
One
of
this
use
cases
is
one
of
the
one
of
the
goals
is
trying
to
do
it.
M
Trying
to
achieve
is
is
to
make
it
easy
for
you
know,
for
someone
interactively
to
you
know
to
to
invoke
this
tli
right.
So
somebody
somebody
has
a
bunch
of
machines.
Maybe
they're
manually
running
cuvee
diem
upgrades
on
them,
but
the
other
use
case
has
now
become
okay.
This
is
being
you
know,
maybe
orchestrated
by
by
some
like
by
an
operator
right-
and
you
know,
I
know,
for
example,
that
there
are
you
know,
least
fervor
for
like
the
AWS,
the
provided
a
right
for
Kappa.
M
There
is
the
assumption
that
it's
it's
not
going
to
be.
You
know
running
command,
you
know
invoking
anything
on
the
hosts.
You
know,
apart
from.
What's
in
the
river,
you
know
the
initial
bootstrap
script,
so
it
makes
me
wonder
like
what,
where
is
this
mechanism
for
even
invoking
something
like
cuvee
diem,
upgrade
and
and,
like
I
said?
M
That's
one
of
the
tensions,
the
other
tension,
I
think
is
you
know,
with
the
control
plane,
join
and
and
cuvee
diem
sort
of
handling
some
some
of
its
own
orchestration,
something
you
know,
things
that
could
be
handled
by
by
a
you
know
a
cluster
API
control
plane.
But
then
again
you
know
if
groovy
diem
also
needs
to
support
the
use
case
where
that
control
planes-
maybe
not
not
around
at
all,
so
it
yeah
those
those
sort
of
things
and.
D
M
Iii
will
we're
again
we
weren't,
you
know
we
weren't
sure
really
where
to
where
to
put
this.
A
Well,
I
mean
I
would
recommend
writing
up
a
doc
with
your
thoughts
and
the
boundary
lines
of
what
issues
you're
seeing
and
then
maybe
send
it
out
to
the
list.
So
we
can
have
a
more
informed
discussion
as
a
group,
maybe
during
the
next
two
cluster
in
our
cycle
meeting
I
do
agree
that
there
are
sets
of
problems
and
they're
currently
in
several
for
several
folks.
They
are
overlapping
concerns,
but
I
think
we
can
disambiguate
and
tease
them
apart
into
a
set
of
domain-specific
problems
for
each
tool
and
I.
I
F
F
No
because
the
version
skew
policy
has
been
designed
very
carefully
to
enable
this
upgrade
is
sort
of
the
answer.
There
is
a
bug
which
I
filed
a
number
of
years
ago
now
around
in
a
multi
master,
so
of
an
HJ
configuration
you
wish
every
one
and
should
be
more
careful
to
upgrade
API
server
and
then
controller
managers
and
all
these
things
and
I
don't
know
of
anyone
today.
That
is
actually
doing
that.
So
there
may
will
be
someone,
but
I,
don't
know
if,
like
cops
doesn't,
is
it
wrong
I?
F
A
Do
it
there's
a
bunch
of
magic
that
can
occur
in
like
because
the
system
is
so
resilient
to
failure
there?
There
are
a
bunch
of
ways
where
it's
it's
secretly
failing
on
the
second
and
third
node,
and
only
one
is
really
the
updated
one
and
eventually,
as
you
update
the
others,
everything
comes
back
online.
So
I.
This
scenario
has
existed
for
a
long
time,
especially
in
the
control
plane
notes
like
safer
here's.
The
concrete
example
that
you
upgrade
a
control,
plane,
node
and
for
some
reason,
as
part
of
the
upgrade
process
it
does.
A
The
conversion
for
API
objects
to
the
latest
incantation
of
those
API
objects.
The
old
control
plane,
nodes
that
existed
that
are
still
online
start,
maybe
perform
a
list
operation,
but
they
can't
actually
do
the
conversion
because
they
have
no
idea
what
version
it
is.
So
as
a
result,
they
implode
right
or
they
go
to
a
crash
or
they
fail
to
actual
calls,
but
because
everything
else
in
the
system
does
retry
as
it
back
off
the
this
API
server
becomes
unhealthy.
The
other
one
becomes
unhealthy.
A
Then
you
only
have
one
that's
online,
then
you
replace
the
other
ones.
So
it's
a
known
condition:
it's
not
the
best
condition,
but
it's
also
not
the
worst
condition
like
back
in
the
day
when
everything
was
blowing
up
at
regular
times
that
that
was
much
worse
than
this
is
I.
Think
because
we
are
a
fault-tolerant,
maybe
to
a
fault,
dad
jokes,
that
we
we
can
recover
from
a
lot
of
these
things.
I
F
I
think
what
you're
saying
is
solved
by
that
you,
you
should
always
have
take
the
Masters
first,
because
you
are
not
allowed.
It
is
not.
We
have
disallowed
and
I
pray
the
Masters
first
and
downgrade
that
nodes
before
a
downgrade.
If
we're,
because
you're
not
allowed
to
have
a
cubelet
that
is
ahead
of
the
api
server,
but.
B
So
like,
if
you
look
at
what
we
were
doing
and
say
the
close
to
API
upgrade
tool,
we
create
a
machine.
That's
got
a
newer
version
of
the
cube.
What
running?
That's
got
the
newer
version
you
want
to
roll
to
so
technically
for
a
pretty
period
of
time.
You
do
have
a
newer
couplet
with
an
older
set
of
API
servers,
while
you're
doing
that
upgrade
and
cops.
A
So
again,
we've
been
breaking
this
kind
of
rule
for
a
long
time.
It's
not
been
catastrophic
because
of
the
window
of
time
of
the
upgrade.
If
we
want
to
be
purely
seamless
like
you
know,
then
there
are
there's
work
to
be
done,
but
I
think
because
the
priorities
of
other
things,
it's
kind
of
this-
has
been
existed
for
a
long
time
right,
but
we've
been
able
to
get.
F
By
with
it
for
years,
I
think
also
API
server
has
tightened
up
the
guidelines
on
that
you're
not
supposed
to
do
both
introduced
and
start
using
a
new
field
in
the
core
control
plane.
In
one
version,
you
have
to
split
across
two
versions
where
you
should
split
across
two
versions,
which
helps
a
ton
right.
Anything
still
to
sneak
through
I
was.