►
From YouTube: Kubernetes SIG Cluster Lifecycle 20170725
Description
Meeting Notes: https://docs.google.com/document/d/17J496IR2tXKw7k97fxwz2KUWOf9rpBD3pIEsmDiJQSw/edit#heading=h.uyqetzfp6u9a
Highlights:
- Deadline for feature issues is next week.
- Upgrading to etcd3.1.10 for the 1.8 release
- Working with sig-node and sig-apps on checkpointing & daemonset upgrade strategies to enable self hosting
- kubeadm adoption group formed with the initial meeting immediately following this meeting
- Thinking about more official structure around sig membership and committers
- kubeadm can now consume CI images
- kubeadm pgrade proposal out for review
- Folks are starting to work on a common cluster API; more details will be sent to the SIG mailing list
A
All
right
welcome
everyone
to
the
sequencer
lifecycle.
Meeting
on
the
25th
of
July
2017
go
ahead,
get
started.
We
had
a
slight
discussion
before
the
recording
began
about
features
for
1.8
and
the
future
freedoms
which
is
coming
up
in
about
a
week.
Jase.
Do
you
want
to
just
recap
that
briefly,
so
it's
caught
in
the
recording
yeah.
B
Absolutely
so
we're
the
feature.
Freeze
is
a
week
from
today,
August
first
and
essentially
that
deadline
means
that
we
need
to
have
basic
documentation
about
what
the
feature
is.
The
one-liner
that'll
be
used
for
marketing
and
it
sounds
like
there's
some
great
stuff
happening,
so
I
am
the
I'm.
They
release
lead
and
one
of
the
things
that
happened
last
release
was
that
there's
a
little
bit
of
rough
edges
around
the
communication
between
the
work
on
coop
Adam
and
the
release
and
Reece
block
exists.
B
A
A
C
C
A
D
Psa
is
the
first
PSA
is
there's
a
couple
PRS
to
update
it
CDs
to
Etsy
d3
110,
that's
not
controversial!
Really.
Those
PRS
are
both.
It's
like
when
you
are
to
update
to
client
the
other
PRS
updating
the
dependencies
internally
inside
of
what's
run
this
part
of
the
code
base
to
the
defaults
for
a
lot
of
the
coaster
up
stuff.
D
A
D
E
D
A
A
D
F
Well,
I,
don't
think
it's
controversial
that
whatever
install
method
that
folks
are
using,
whether
it
be
cops,
are
open.
Shifter,
gke,
there's
a
there's,
a
burden
with
the
the
CD
two
to
three
type
of
thing
and
I
and
I'm
not
sure,
based
on
the
way,
because
we
just
don't
have
any
sort
of
unified
view
of
this
stuff.
I'm,
not
sure.
F
If
there's
sort
of
an
easy
button
that
we
can
actually
create
for
folks
to
actually
make
that
problem,
work
I
think,
there's
lessons
to
be
shared
across
the
different
methods
in
terms
of
how
they
do
stuff-
and
maybe
you
know,
writing
the
guides
and
those
lessons
could
have
been
done
for
sure.
I
I'm
not
quite
sure
how
how
we
can
make
this
as
turnkey
as
we
want
to.
E
Well
completely,
for
example,
we
had
a,
we
had
a
turnkey
procedure
for
the
upgrade
from
a
city
street
sd3
in
the
one-five
and
remember
the
number
one
five
release
and
we
discovered
like
in
the
dying
days
of
the
release
that
it
nuked
every
AJ
cluster.
So
you
know
it
was
a
it
was
it
is.
The
decision
was
made
to
proceed
with
that
CD
three
I
think
certainly.
D
Less
behind
it
had
a
che
clusters
go.
There
was
a
process
that
we
had
evolved,
which
was
to
turn
down
your
members.
Do
the
upgrade
and
you
have
an
offline
upgrade
like
we
actually
had
this
documented
that
your
upgrade
will
be
offline
turned
on
your
members
to
one
offline
upgrade
then
turn
up
your
members
and
add
members
again,
but
this
is
history
now
I
think
for
going
forwards.
D
The
reason
why
we're
doing
a
3-1
10
upgrade
in
this
cycle
is
to
prevent
the
minor
version
bump,
so
we
want
to
keep
want
to
keep
in
lockstep
with
minor
version
additions
as
we
upgrade.
So
that
way
it's
totally
non-controversial
and
it
should
be
seamless
for
entities
clusters.
We
specifically
did
not
go
to
3.2,
because
there
is
little
even
the
slightest
amount
of
questioning.
E
D
E
You
see
what
I'm
saying
right
now
so
now,
if
you're
saying
we're
not
going
to
three
two
now,
that
means
we
don't
want
to
go
to
three
two
in
one
nine,
because
then
the
one
seven
two
one
nine
upgrade,
which
will
be
a
supported
upgrade,
would
be
the
same
upgrade
that
you
just
described
as
controversial.
It's.
A
F
E
G
A
D
The
second
PSA
is
I'm,
starting
down
the
checkpointing
road
for
self
hosting
and
I'm
deep
in
re-familiarize
a
re
re
re,
maybe
the
for
three.
If
familiarizing
myself
with
the
Kubla
code
and
I
need
to
talk
with
the
signal
folks,
hopefully
today
to
make
sure
I
insert
at
the
right
path,
because
it's
it's
a
wonder,
can
so
the,
but
it
doesn't
look
that
free
miss!
That's
words.
It
doesn't
look
that
difficult,
but
I'll
see
what
I
can
do
and
see.
If
we
can
get
something
in
place.
You
know
reasonable
amount
of
time.
A
D
D
A
B
H
A
H
Well,
there
were
some
discussions
whether
would
use
new
well.
First,
first
of
all,
Lehmann
sets
are
going
to
be
moved
from
the
extensions.
We
won't
be.
The
one
API
group
apps
we
1
beta
2,
as
well
as
all
other
apps
objects,
we're
going
to
be
in.
We
one
be
to
group
in
the
end
of
1/8.
That's
it
let's
go,
and
we
asked
for
this
list
at
first
then
delete
method
being
added
as
a
feature.
H
So
we
agreed
that
we're
going
to
use
a
new
upgrade
strategy
instead,
something
like
rolling
upgrade
at
first
when
delete
and
that's
a
net
new
feature,
probably
in
all
fighter
now
that
we
didn't
discuss
that
and
Aaron
is
going
to
assign
some
some
guy
from
chorus
to
work
on
it
and
I,
don't
know
if
them
will
I
mean
it
will
or
similar,
but
the
gaps
will
help
review.
At
least
they
taught
that
what
they
would
have
the
bandwidth
but
yeah
I
think
that's.
That's
mostly
yeah.
A
Yeah,
so
they
they
did
sort
of
commit
to
doing
reviews.
I
think
Eric
was
a
little
reluctant,
but
a
couple
other
people
on
the
call
said
they'd
be
happy
to
review
the
change,
so
I
think
we're
good
to
go
there
and
Aaron
did
say
that
he
had
someone
that
he
could
have
start
working
on
the
proposal
for
this,
so
they're
they're
gonna
start
diving
into
the
code
and
seeing
what
it
would
look
like
to
change
this
and
fire
off
proposal
all
shortly.
K
H
E
Indeed,
so
the
cube
ATM
working
group
is
kicking
off.
We
are
going
to
have
our
inaugural
meeting
immediately
following
this
meeting.
Sorry
for
the
short
notice,
I
don't
expect
I
want
to
be
able
to
attend,
but
hopefully
we
can
at
least
start
the
ball
rolling.
Very
we
sent
out
that
spreadsheet.
There
were
not
a
lot
of
times
which
worked
with
this.
One
was
basically
a
clear
winner.
I
think
there
was
only
one
person
that
had
a
negative
opinion
of
it
at
all,
so
this
was
a
the
best
slot
that
was
available
to
us.
E
There's
bright
white
over
that's
for
sig
note,
but
there
we
go
so
yeah.
So
after
this
we'll
try
to
use
the
same
hangout
I'm,
not
sure
that
will
actually
work,
but
if
we
can
the
same
zoom
we'll
try
to
use
the
same
zoom
if
we
can
and
if
not
I
will
send
around
a
different
zoom
in
slack
or
something.
But
hopefully
we
can
just
keep
it
open.
A
That's
it
yeah,
and
that
was
that
was
sort
of
previously
announced
as
something
that
we
were
starting
up
on
the
mailing
list.
So
hopefully,
even
if
people
missed
the
announcement
of
the
first
meeting
starting
today,
people
at
least
know
that
this
is
happening.
I
can
jump
in
next
week
that
that's
intended
to
be
weekly
or
bi-weekly
I.
E
F
F
That
hasn't
happened.
Yet
you
know
the
elections
haven't
happened
for
the
you
know,
we're
still
in
the
process
of
doing
the
elections
for
the
committee.
So
with
that
in
mind
after
the
for
the
steering
committee.
So
with
that
in
mind,
I
would
love
to
just
get
input
from
folks
here.
In
terms
of
of
you
know,
how
do
you
think,
like
a
good
way
for
us
to
sort
of
codify?
That
would
be?
How
do
we
I
mean?
This
is
not
an
urgent
discussion,
but
it's
it's
important
for
us
to
actually
get
there
eventually.
F
How
do
we
decide
who
has
standing
I?
Think
you
look
at
the
list
of
folks
who
have
you
know
committers
across
the
set
of
directories
that
that
are
their
owners
across
the
set
of
directories
that
matter
who's
a
member
of
the
sig,
but
not
necessarily
a
committer?
How
do
we
actually
go
about
doing?
What's
done
so
I
just
want
to
open
the
floor
and
get
some
feedback
in
terms
of
how
people
are
thinking
about
this
control.
G
One
question
speak
anything
against
modeling
as
close
as
possible
to
the
Apache
Software
Foundation,
okay.
So,
in
what
way,
what
part
of
that
deal
and
chill?
Well,
essentially,
all
the
different
projects
are
it's
up
to
them
to
decide
to
you
know
what
exactly
the
process
is,
but
you
have
to
have
a
new
your
charter.
You
have
to
clearly
say
you
know:
how
is
it
release
cut
and
how
you
know,
how
do
you
actually
become
a
committer
and
so
on?
To
put
like
all
of
that
and
then
I
appreciate
that
you
have
kind
of
boilerplate
whatever.
F
No,
that's
good,
I!
Think
yeah,
that's
that's
generally.
This
is
where
I
predict
and
assume-
and
you
know,
based
on
talking
with
folks
I
think
it's
gonna
go,
is
that
there
will
be
autonomy
at
the
sig
level
for
the
SIG's
to
actually
figure
out
how
they
govern
themselves.
There
will
be
some
requirements
around
what
you
just
have
to
have
something,
and
there
might
be
other
ways
that
you
know
other
other
tests
that
it
has
to
pass.
F
F
G
F
So
the
crossing
stuff
I
mean
that's
the
idea
that
there
is
you
know.
At
the
end
of
the
day,
the
buck
is
gonna,
stop
with
the
Governance
Committee.
The
the
Governance
Committee
doesn't
want
to
make
a
lot
of
decisions
directly,
so
they're
gonna
delegate
a
lot
of
that
to
think
big
architecture
or
the
API.
You
know
approval
process,
and
so
a
lot
of
the
Crossing
stuff
is
gonna,
get
discussed
there
in
terms
of
Technology
in
terms
of
policy
for
how
things
get
done,
that'll
be
delegated
to
other
places.
F
So
there's
a
good
there's
a
you
know,
like
conformance
what
does
it
mean
to
actually
be
a
conforming
cluster
that'll,
probably
be
driven
out
a
cig
testing.
What
does
a
kubernetes
release?
Look
like
that'll,
probably
be
driven
out
of
the
city
release
and
with
requirements
around
that
so
I
think
there's
a
sort
of
there's
a
bunch
of
places
where
we
have
cross-cutting
things
and
that
all
won't
get
concentrated
into
one
place
that
electrical
unit
across,
but
I,
think.
F
The
question
here
is
that
I'm,
assuming
that
there's
gonna,
be
some
ability
because
things
can
govern
themselves
and
the
question
is:
how
does
this
cig
wanted
you
that
what
makes
sense
for
this
cig?
We
have
this
sort
of
a
little
bit
of
a
split
personality
where
there's
a
lot
of
concentration
on
cube
admin,
but
obviously
the
scope
and
the
Charter
of
cluster
lifecycle
goes
beyond
that,
and
I
think
that
we
need
to
make
sure
that
we
do
service
to
that
larger
scope.
F
H
So
one
thing
I
guess:
we
have
to
figure
out
this
like
who's
the
well
now
we
have
the
own
it's
file
and
I.
Think
right
now
we
have
four
or
five
like
crew
vets
and
eight
to
nine
reviewers.
So
what's
the
closest
like
get
there
and
how
do
we
remove
folks
that
have
been
active,
I
mean
early
I
still
like
he
did
a
lot
of
effort
in
the
beginning
of
humans
life
cycle,
but
now
hasn't
been
active
for
quite
some
time.
F
Well,
I
think
one
change
that
I
would
like
to
see
is
right.
Now
the
owner
files
are
really
tilted
around
people
and
not
SIDS
and
I
actually
think
that
that
that
creates
a
confusing
activity.
If
I
want
to
make
changes
in
this
directory,
who
do
I
talk
to
right?
You
get
a
list
of
people
to
talk
to,
but
in
reality
it
should
be
hey
if
I'm
gonna
be
making
changes
in
this
part.
Here's
the
stake
that
I
want
to
coordinate
with,
because
there's
gonna
be
a
lot
of
stakeholders
there.
F
D
We
started
doing
this
with
six
scheduling
because
it
was
such
a
pain
because
those
are
fragmenting
across
the
codebase,
so
we
started
making
aliases
and
the
alias
files
exists
in
the
root
and
you
can
update
levels
and
create
whatever
aliases
you
want
right.
So
that
way
you
can
have
your.
You
know
your
stratifications,
that
you
desire
I,
think
that
is
a
reasonable
model
and
would
probably
work
well
for
this
case.
D
F
Yeah
and
and
I
think
that
also
I
mean
that
works
within
kubernetes,
/,
kubernetes,
there's
obviously
other
repos
that
actually
play
in
here,
also
and
hopefully,
more
and
more
repos
over
time
as
we
move
away
from
the
mono
repo
one
of
the
things
that's
on.
My
list
is
to
make
membership
of
teams
and
the
organization
and
repo
sort
of
permission
be
data
driven
off
of
files
that
are
checked
in
not
going
to
be
owners
explicitly
that'll,
be
something
similar
where
we
can
curate
that
in
a
sort
of
traceable,
auditable,
obvious
way
so
yeah
soon.
H
F
Yeah
I
got
a
right
that
tool,
I
haven't
sort
of
stubbed
out
right
now.
This
is
this
is
on
me
to
actually
do
that,
because
I
think
that
lets
us
move
in
that
direction
where,
where
we
actually
have
here's
a
file
that
says
who's,
you
know
you
should
have
permissions,
because
right
now,
github
tends
to
hide
that
stuff
and
it's
become
sort
of
a
confusing
mishmash.
F
If
who
does
what
when
and
where
I
mean,
maybe
that'll
be
part
of
the
community,
repo,
maybe
it'll
be
a
different
repo
but
yeah
exactly
right,
there's
gonna
be
one
place
that
says
or,
like
the
owners
aliases,
do
we
want
to
share
that
across
across
repos,
instead
of
having
to
be
just
a
kubernetes
kubernetes
thing?
Yes,.
F
C
C
Today,
in
terms
of
owners
and
approvers-
and
they
found
it
was
too
much
work
to
maintain
the
automation
for
that
it
was
not
valuable
work.
So
they
had
worked
on
on
reducing
the
the
number
of
the
stratification
of
permissions,
which
is,
if
we
decide
you
want
to
do
it,
we're
okay,
with
doing
something
like
that.
It's
nice,
because
kind
of
teams
are
rather
limited
in
their
permissions,
don't
like
the
full
set
of
things
that
you
get
from
my
kit
lab.
C
So
if
we
could
have
a
thing
where
it's
like,
you
know,
just
like
two
levels
like
you
could
read
the
repo
with
or
maybe
three
you
could
redo
PO,
you
can
write
the
repo
or
you
can
you
can
change
the
we
both
with
the
administrative
things,
I
think
that
would
be
nice
to
say
at
either
a
cig
or
a
cross
project
level.
So
we
could
have
you
know
I,
don't
think
we're
gonna
move
from
github
anytime
soon,
so
making
use
of
the
platform
would
be
nice
rather
than
not
least
of
my
personal
opinion,
ya.
F
C
F
All
right
so
I
don't
have
any
answers
here.
I
don't
want
to
take
up
too
much
time,
but
I
just
wanted
to
sort
of.
You
know
make
sure
that
folks
were
thinking
about
this
and
we
were
taking
some
time
to
think
about.
How
does
this
thing
want
to
govern
itself?
So
we
will
have
some
autonomy
in
terms
of
how
we
decide
things
so
yeah.
E
Guess
one
I
just
I
guess
until
like
our
goal
like
what
is
our
goal
like,
we
could
define
our
goal
as
like
adoption
of
cube
ATM,
and
maybe
we
expand
the
scope
of
cubic
atm
to
encompass
other
things,
but
I'm
not
sure
we
actually
have
a
have
a
concrete
goal
that
we
are
heading
towards.
Otherwise,
even
intermediate.
Let's
think
charter
Justin,
well,
yeah
I,
think
we've
drifted
from
the
sea
charter.
A
A
little
bit
right,
I,
wouldn't
know
I
guess
I
feel,
like
we've
been
focusing
on
a
subset
at
the
state
charter,
which
is
actually
I
went
about
later
in
the
meeting,
which
is
how
we
can
continue
to
drive
that,
but
also
expand.
What
we're
working
on
right,
I
think
it
was
somewhat
necessary
for
us
to
focus
the
beginning,
because
there
were
a
relatively
small
number
of
people
that
were
actively
working
and
if
everybody
was
doing
something
different,
we
wouldn't
have
made
any
progress.
A
F
Know
but
I
agree
with
Justin
I
think
it's
worthwhile
for
us
to
you
know
we
don't
have
to
do
it
right
now,
but
to
take
another
honest,
look
at
the
sink
charter
and
say:
is
this
where
we
want
to
go?
Does
this
represent?
You
know
what
our
goals
are
right
now
and
and
and
part
of
the
governance
of
the
sig
is
actually
over
time,
reexamining
making
sure
that
we're
staying
true
to
that
charter
or
modifying
the
charter
or
modifying
our
activities
to
make
sure
that
those
things
are
aligned
so
yeah.
That
seems
totally
reasonable.
Okay,.
F
A
Awesome
all
right
next
up,
Alexander
her
cat
on
github,
linked
to
APR,
so
Jeff
Grafton
has
been
adding
support
to
the
build
process
to
be
able
to
push
CI
images
as
we're
as
we're.
Building
each
CI
build
up
into
GC
our
buckets,
which
was
unblocking
the
PR
link
here
to
support
using
CI
images
in
cube
adventitia
tests,
which
is
something
that
was
on
our
list
for
1.8.
So
it's
awesome
that
there's
a
PR
out
Joe.
F
A
F
H
Well,
yeah,
it's
I'd
say
it's
definitely
possible
I
mean
it's.
We
just
have
the
image
repository
and
you
can
overwrite
it
to
whatever
you
want
and
well
I.
Don't
I
don't
know
like
if
we
have
the
pushing
tooling,
but
on
the
other
hand,
it's
not
that
hard
to
like
just
build
your
images,
let's
tooling,
for
that,
and
then
just
do
Daka
push
to
docker
hub
or
whatever,
and
then
in
cubed,
and
just
specify
that
I
want
to
use
the
docker
hub
use.
It
looks
as
instead
of
like
GTR
I
mean.
F
So
Lucas
I
would
love,
I
mean
I,
I,
think
it's
possible,
but
I
think
there's
a
certain
amount
of
like
I
want
to
be
able
to
build
all
the
release.
Artifacts,
you
know
with
j-b
des
in
the
name,
just
like
they're
built
for
the
real
release.
Right
and
and
any
any
Delta
there
I
think
is,
is,
is
gonna
introduce
issues
over
time.
So
that's
yeah.
This
is
there's
a
thing
of
degrees
there
for
sure
go
ahead.
Lucas!
Sorry!
Yes,.
G
H
H
Guess
that's
more
sig
releasing
like
how
do
I
make
this
release
I,
don't
think
actually,
there's
something
wrong
with
cubed
and
because
I
mean
if,
if
we
had
this
thing
that
you
could
push
to
job
Joe
Beda
like
docker,
hub
use
or
whatever
some
GC,
our
repo,
it's
it's
really
straightforward
hook
up
cubed
and
to
use
that
it's
just
this
release
thing.
That
is
not
great
I.
L
D
I've
literally
set
up
everything
by
myself
build
everything
from
scratch.
Do
my
own
thing
so
I
understand
all
the
bits
in
place,
but
the
problem
is
turnkey
automation
and
getting
buy-in
from
synchronous,
because
there
are
no
owners
right.
I
have
everything
there
for
from
an
RPM
perspective.
Right
and
I
would
even
like
to
turn
it
down
even
more
and
just
have
a
single
container
and
do
atomic
install
and
then
I'm
done
right.
D
Then
I
then
I
just
need
to
manage
a
single
container,
which
is
like
the
hypercube
ish,
with
the
baby
em
together
smashed
into
one
and
I'm
done,
but
the
there's
a
general
problem
of
ownership
of
the
release
process
of
the
least
release
artifacts.
So
I,
don't
know
how
we
solved
that
problem
and
I.
Don't
know
if
that's
within
our
charter
to
solve
that
when.
A
I
was
discussed
at
the
the
Leadership,
Summit
and
I
think
everyone
was
sort
of
leaning
towards
having
cig
release
own.
That
I
mean
that's
back
to
the.
What
our
Charter
is
I
think
that's
sort
of
well
outside
of
our
charter
to
try
to
solve
that
problem.
That
doesn't
mean
that
people
that
are
working
on
this
can't
also
work
on
that,
but
I
don't
know
that
we
want
to
do
it
out
of
the
auspices
of
just
a
cluster
lifecycle,
yeah.
F
For
sure
yeah
I,
just
I,
it's
a
it's
a
sort
of
Devon
test
and
sort
of,
hopefully
make
it
easier
to
sort
of
do
something.
Official
thing
as
as
we
Devon
test,
hopefully
we'll
avoid
some
of
the
fire
drills
that
we
hit
every
time
we
can
bring
it
all
together,
come
released
right
and
that's
I
think
the
the
the
ultimate
goal.
For
me
at
least.
H
I
When
the
CI
Working
Group,
we've
got
a
cost
cloud,
C
iPod
we've
been
working
on
that's
pulling
in
the
kubernetes
repos
in
building
the
artifacts,
we've
been
pushing
to
a
per
project,
docker
hub
or
registry,
and
deploying
via
our
cross
cloud
prisoner.
Well,
one
of
the
things
that's
kept
us
away
from
use
and
cube.
I
Admin
is
actually
being
able
to
pull
in
the
artifacts
her
commit
because
they
weren't
generated
during
this
sounds
like
it's
a
good
segue
to
maybe
bringing
in
cube
admin
so
that
we
can
use
our
own
artifacts
or
find
some
way
to
pull
that
into
our
pipeline.
We're
using
cube
admin
right
now.
The
reason
we've
stayed
away
from
it
is
being
able
to
specify
our
kubernetes
binaries
and
images
during
the
deploy
that
we
weren't
able
to
do
sounds
like
before
this
commit
okay,
so.
H
We
document
this
poorly
I
guess
so
Jeff
Rubin
is
kubernetes
version
specifies
like
all
the
tags
for
for,
like
it's
its
registry
and
cube
API
server
and
the
kubernetes
version
value,
and
you
can
specify
both
the
repo
like
name
and
tag,
the
version,
but
I
don't
know
it's
not
sufficient,
but
that's
possible
to
do
and
has
been
for
quite
some
time.
This.
A
Was
a
great
discussion
to
have
during
Justin's
next
meeting
about
Cuban
adoption,
so
I
think
one
of
the
things
to
discuss
there
is
is
what
is
preventing
everyone
from
using
cube
Atman
in
production
scenarios?
It
sounds
like
this
is
a
great
use
case
that
either
isn't
well
enough
documented
or
supported
is
just.
F
F
F
M
H
I
F
A
released
out
SH
out
of
kubernetes
last
kubernetes
that
just
builds
containers,
yeah
yeah.
You
should
look
at
stuff
in
in
in
github
coms,
I'll
put
it
in
a
slash,
kubernetes,
slash,
release,
cuz,
that's
how
the
real
releases
are
done.
That's
the
we
should
find
the
one
thing
that
we
all
line
up
on
here,
and
this
is
really
at
least
thing
obviously,
but
different,
really
assault
together.
A
repo
yeah.
F
Why
it's
like,
like
everything
else
in
kubernetes,
there's
like
12
ways
to
do
everything
right
so
this
is,
this
is
I.
Think,
hopefully
you
know
with
cig
release
getting
off
the
ground,
there's
an
opportunity
to
start
bringing
this
stuff
together.
That's
cool!
Unfortunately,
though,
is
pretty
googly.
That's
that's
the
one
thing
that
I
think
we
need
to.
We
need
to
work
through
and
it's
not.
You
know
it's
not
the
maliciousness.
It's
just
you
know.
A
D
D
I
know,
as
we
add
new
features,
yeah
getting
those
tests
and
automated
tests
in
place
before
a
release
are
having
some
level
of
I,
don't
know
what
we
want
to
call
it
during
the
end
of
the
cycle,
where
we
kind
of
update
our
test
coverage,
it's
probably
pretty
crucial
because
we
went
out
the
door
with
one
700,
with
some
pretty
massive
laws
and
then
I
see.
Lucas
is
smiling
because
he's
the
one
who
submitted
the
PRS
and
I
reduced
them.
So
I,
don't
know
if
your
comments
Lucas.
H
D
H
Yes,
I
think
most
of
them
when
pretty
minor,
but
there
was
actually
some
something.
That's
like
a
cherub
take.
That
was
a
regression
that
was
well,
but
but
otherwise
it
was
pretty
good.
I
think
most
like
low-hanging
fruit,
like
adding
the
the
notification
for
like
this,
is
happening
in
1/8,
with
a
24-hour
deadline
and
such
thing,
but
yeah
you're,
you
know
completed
right.
It's
there
were
some
quite
some
terrific's.
H
D
One
of
them
was
that
was
the
boy
I
think
Moyer
commented
on
to.
There
was
one
that
was
broken
by
default
and
it
was
a
regression.
There
was
also
I'm
trying
to
I'm
struggling
to
recall,
but
there
were
two
total
that
that
could
have
affected
people
that
were,
you
know,
had
used
existing
means
by
which
we
were
deploying
so
there's
just
a
gap
in
test
coverage.
Is
that
the
general
statement,
the.
L
H
More
like
integration,
or
so
we
had
unit
test
for
this,
but
the
problem
was
something
with
Cobra
initialize
in
its
solution
session
and
well
so
in
the
unit
test
they
were,
they
worked
well
but
like
for
real
sometime
like
cobra
hadn't
invoked
something
be
able
to
like
populate
these
kind
of
things.
So
it
was
well
kind
of
messy
but
yeah.
A
You
ought
to
say
one
other
other
plug
for
typing
in
the
notes
is
that
when
we
were
doing
planning
for
1.8,
everyone
agreed
that
testing
was
really
really
important
and
we
marked
it
as
a
p,
0
priority
and
so
sort
of
in
line
with
what
Tim
was
saying.
If
you're
adding
new
features,
please
make
sure
you
add
test
coverage
for
them,
and
ideally
we
also
be
backfilling
test
coverage
for
older
features
that
don't
have
it
Lucas.
Yes,.
H
It
would
take
quite
a
lot
of
time
for
me
at
least
like
reading
all
the
stuff
that's
needed,
but
if
Jacob
is
unable
to
I,
can
take
over
over
that
pot.
But
basically
we
need
this
like
on
every
PR
before
we
merge
as
like
statement
on
whether
this
breaks
qadian
or
not,
but
then
we
actually,
we
was
we
all
really
need
need
for
is,
like
you,
betta
media
tests
that
we
have
like
cuba
in
specific
eatery
tests.
H
That's
what
we're
lacking
and
that's
what
brought
what
caused
this
regression
that
Tim
talked
about
is
that
we
hadn't
like
we
didn't
test
yeah
we're
testing
all
the
time
like
the
normal
setup,
but
we,
for
example,
didn't
test
with
config,
and
this
broke
config
in
a
real
indication.
So
and
I
I've
asked
some
how
run
1
wil
will
work
on
this,
at
least
to
some
degree,
and
also
we
got
the
which
is
later
on
the
agenda
here.
Bootstrap
tokens
on
now
II
to
be
tested,
so
in
the
yeah
Eadie's
job
against
head.
H
We
have
some
feature,
bootstrap
tokens
EDS
running
and
they
are
all
green.
So
that's
that
was
one
of
the
requirements
for
getting
this
feature
to
beta,
so
I
think
we're
doing
better,
but
it
I'd
love
to
see
more
like
feature
cube
am
either.
Is
there
it's
not
that
hard
to
write,
but
it
definitely
takes
time
to
cover
a
lot
of
cases.
A
H
So,
just
a
heads
up
that
up
grace
proposal
is
converted
to
knock
down
without
styling,
but
I'll
fix
that
up
after
I've
addressed
all
the
other
like
comments
that
may
pop
up,
it
would
be
great
to
like
have
some
like
if
it
looks
good
I
mean
we've
discussed
this
for
for
some
weeks
now.
So
if
it
looks
good
to
everyone
just
hit
like
LG
TM,
when
when
you
have
fixed
needs
and
style
or
something
to
get
like
buying
from
our
own
sig,
so
it
can
be
merged
right
from
Joe
Tim
robots.
It
says
wrong.
A
H
K
F
A
I,
don't
leave
it
to
my
list
next
thing,
I
wanted
to
sort
of
briefly.
Mention
is
sort
of
along
a
line,
so
we
discussed
earlier
about
to
expand
the
scope
of
the
sig
outside
of
just
cube
admin.
So
cube
admin
is
sort
of
the
place
we
drew
the
line
initially
of
where
we
wanted
to
focus
because
it
abstract
it
way
all
of
the
underlying
infrastructure
for
sale
machines,
whether
you're
on
bare
metal
or
on
a
pile
or
or
somewhere
else
right,
and
so
we
could
say
once
you've
got
all
of
a
those
machines
up.
A
We
can
make
it
consistent,
install
experience
from
there
on
forward
and
drive
consistency
across
the
platforms.
I
think
it's
starting
to
become
a
little
more
clear
that
we
also
want
to
at
least
think
about
having
more
consistency
at
the
layer
below
that,
in
terms
of
it,
cluster
level,
API
and
so
I
know.
I've
had
some
conversations
with
a
couple
people
about
this
and
what
this
might
sort
of
look
like
Chris
Nova
had
sent
out
cubic
horn
with
sort
of
a
straw.
Man
I
think
it's
somewhat
based
on
the
Copts
API.
A
Also
I
think
there
is
general
consensus
from
what
I've
heard
that
people
want
this
to
be
a
sort
of
kubernetes
style.
Api
that
looks
like
the
rest
of
the
crew
needs.
Api
is
it
used
as
communities,
APEC
machinery
and
so
I
just
wanted
to
sort
of
say
that
we
are.
We,
as
in
a
couple
of
people,
myself
included,
are
gonna
start
looking
at
what
this
API
might
look
like.
A
I,
don't
think
we're
quite
at
the
point
of
spinning
off
yet
another
sort
of
sub
meeting
from
the
larger
group,
but
it's
something
that's
gonna,
hopefully
start
to
gain
steam
sometime
within
the
next
month
or
two
I
just
want
to
put
that
out
there.
If
people
are
interested,
please
ping
me
or
contact
us
or
C.
Let
us
know
how
you
wanna
get
involved.
A
N
F
That
note
I
think
there's
some
interesting
ideas
that
folks
are
starting
to
play
around
with
where
it's
writing
controllers
for
managing
structure,
also
versus
just
sort
of
one
shot,
typist
talking
so
I
think
the
Lucy
guys
have
done
some
really
interesting.
You
know
prototyping
and
work
here
also
and
I
know
that
that
Chris
has
been
syncing
up
with
them.
So
I
don't
know
if
that's
part
of
this
effort,
also
but
I
I
think
it's
super
exciting
stuff
ya.
A
H
I'm,
basically,
it's
driving
consistency.
If
we
had
three
layers,
one
is
like
provisioning
and
then
it's
bootstrapping
and
then
add-ons
justin
is
working
on
like
the
atom
manager,
which
is
layer
three.
We
have
cubit
and
fillet
it
too,
and
now
we're
gonna
look
into
something,
for
they
were
one.
Yes,
it's.
F
G
N
H
H
We
that
didn't
make
it
sig
release,
wasn't
created
like
kind
of
at
the
time,
not
nothing,
at
least
not
in
the
position
to
make
something
new
just
get
the
release
out
the
door
so
now
for
1/8.
Do
we
want
to
push
for
this?
To
get
these
out?
I
mean
I,
don't
know,
I,
don't
think
172
is
out
yet
on
cubed
em
like
it
was
released,
Friday
and
now
it's
depending
on
some
google
at
the
push
and
it's
well,
it's
documented
inside
of
google.
E
C
Think
you
would
need
to
put
it
in
an
ego
so
that
the
Googler
who
is
running
the
script
is
also
going
to
be
pushing
the
Deb's
and
rpms
and
we
could
try
and
voluntold
in
the
the
branch
manager,
since
that
person
is
a
Googler
and
should
be
familiarizing
themselves
with
an
ego
right
now.
So
we
can
try
putting
that
on
their
on
their
list
of
things
to
do,
and
you
can,
you
see,
see
him
more
issue.
C
C
H
It's
like,
as
part
of
this
effort,
it's
gonna
be
a
GCS
bucket,
instead
of
like
so
we
can
with
a
GCS
bucket.
We
can
have
different
permissions
like
we
can
have
the
stable
locked
down
to
just
Googlers
and
we
can
have
the
unstable,
lock
down
to
maybe
maintain
us
or
over,
like
to
somewhere.
That
thought
other
people
actually
are
able
to
push
CIB
or
whatever
yeah.
F
H
So
I
check,
I
looked
at
this
and
I
think
I
have
some
cycles
to
like
at
least
provide
guidance
and
I,
also
just
making
it,
but
it
depends
on
some
Googler
or
something
to
actually
make
GCS
buckets
for
it
like
take
responsibility
for
maybe
matey
making
sure
it's
working
like
whenever
any
release
and
those
kind
of
things
just
adds
up.
Yep.