►
From YouTube: Kubernetes SIG Cluster Lifecycle 20180130
Description
Meeting Notes: https://docs.google.com/document/d/1deJYPIF4LmhGjDVaqrswErIrV7mtwJgovtLnPCDxP7U/edit#heading=h.fb897yipgfuc
A
Hello
today
is
January
30th,
and
this
is
the
sig
clustered
lifecycle,
regular
meeting
I'm,
just
gonna
burn
through
agenda
items
and
feel
free
to
add
to
the
agenda.
I've
propagated
some
things
from
last
week
to
talk
about,
and
you
know
if
there
are
other
items
that
folks
need
to
talk
about,
feel
free
to
add
them
to
as
well.
A
So
most
of
my
feedback
is
from
followed
from
last
week
for
those
who
are
interested
Lee,
who
I
don't
see
in
the
call
right
now
and
I
had
actually
gone
through
and
triage
the
110
list,
so
the
110
list
should
be
reasonable.
There
are
Help
Wanted
items
on
the
110
list.
If
folks
are
interested
in
engaging
I
would
probably
execute
there
first
if
you
want
to
get
on
the
the
hot
seat.
A
A
If
anyone
wants
to
help
with
that,
what
we
should
probably
do
is
have
a
listing
of
Bach
commands
that
you
can
use
to
help
triage
it.
The
my
only
request
is
that
you
assign
the
triage
label
a
priority
if
you're
gonna
commit
to
it
this
milestone
and
whether
or
not
you
think
it's
a
Help
Wanted
item,
so
I
kind
of
reserved
the
help
wanted
to
like
Robbie,
myself
and
Ilya,
but
if
or
if
you
just
want
to
like
report
something
and
walk
away.
A
A
All
right
there
is
a
question
I
had
with
regards
to
all
this
triage
work.
There
was
a.
There
is
one
item
that
was
listed
for
the
110,
milestone
that
I
didn't
understand
and
I.
Don't
know
if
Robbie's,
not
here,
I,
don't
know
if
other
folks
had
thoughts
on
it.
But
it's
it
was
a
separate
line
item
from
the
couplet.
A
B
We
discussed
a
thief.
I
recall
attempting
my
interpretation
of
it,
but
I
think
what
happens
right
now
is
that
our
debian
or
RPM
packages
will
see
a
different
version
of
that
drop
with
the
drop
in
system
D
drop
in
file,
depending
on
on
you
know
what
version
of
Cuba
they
share
so
that
we
that
father
had
been
changing
over
time
in
in
our
packages
on
the
I
think
they
did
like
that
to
be
more
static.
A
B
A
A
You
can
always
come
back
to
that
I
think
when
Robbie's
around
I
know
that
Rob
has
a
list,
but
I
don't
know
where
it
is
and
I've
never
seen
it
publicly
of
what
things
are
blocking
Google
from
adopting
Kubb
ATM
installs
by
default.
So
if
anyone
else
has
a
link
to
a
listing
there,
that
would
be
super
helpful
because
I.
B
B
A
Put
a
note
in
the
doc
to
follow
up
with
Rob
regarding
that
I'm,
not
certain
what
they
can
all
share,
but
ideally,
if,
as
we
start
to
get
more
folks
involved
and
spin
up
and
help
wanted,
we
can
always
prioritize.
I
honestly
would
like
to
add
egg
flow
it,
because
if
you
get
gke
involved,
you
know
from
the
inception
point
of
view
like
they.
A
They
will
filter
a
lot
of
bugs
through
the
system
you
know,
otherwise
we
are
set
up
as
a
separate
test
suite
which
isn't
actually
testing
all
the
bits
live
right
right,
so
that's
always
been
our
problem
and
I
would
ideally
like
to
make
it
first-class
blocker.
You
know,
and
that
way
other
people
can
also
depend
upon
that
functionality
as
well.
So
as
the
adoption
group
grows,
they
can
also.
You
know
know
that,
there's
enough
eyes
on
this,
that
there
it's
stable,
right,
yeah
I
was
somehow.
B
Under
impression
that
cheeky
folds
used
a
bunch
of
internal
things
that
were
quite
google-specific
like
the
way
they
run
other
things,
so
I
was
some
harm
to
such
impression,
but
I
don't
know,
maybe
wrong.
I
mean
there's
like
general
faith
for
give
admin
NGK
or
they
may
have
fundamentally
different
view
of
the
world.
It's.
C
D
Is
a
Rodrigo
from
from
Google
here
and
just
in
order
to
give
you
a
sense
of
where
we
are
in
terms
of
non
cluster
API
and
at
UK
and
qadian
I.
Don't
have
a
list
so
I
think
Renee
should
follow
up
with
Robbie
on
there
on
the
list
of
blocker
I,
you
know
items
for
cube,
ATM
we've
been
we've
been
working
actually
on
designed
in
or
try
to
how
to
get.
You
know
UK
to
adopt
another
poster
UK.
D
A
Okay,
yeah
I
think
at
least
getting
a
list
going
would
be
helpful
because
then
you
know,
if
you
want
to
solicit
the
community
I'm
sure
people
would
stand
up,
knowing
that
the
broader
implications
would
be
wider
adoption
and
more
eyeballs
and
better
test
coverage,
etc.
So
it's
kind
of,
like
everybody
kind
of
you,
know
lending
a
hand
to
build
the
Amish
farm
type
of
thing.
Right.
Yeah
makes
awesome,
yeah.
A
The
only
other
item
that
I
had
was
a
request,
maybe
for
someone
else
if
they
could
update
the
big
cluster
lifecycle,
notes
to
reflect
all
the
different
subgroups
and
meetings
that
are
actually
in
this
meeting.
So
when
people
want
to
find
out
about
sync
cluster
life
cycle,
the
main
touch
point
is
through
the
community
repo
and
the
community.
Repo
is
actually
pretty
out
of
date
with
some
of
this
stuff,
and
the
only
show
is
the
main
meeting
and
not
the
subgroups
or
anything
else
like
that.
A
E
B
B
A
Don't
know
what
this
it
probably
would
be
asked
us
to
talk
with
rod
to
again
when
he
gets
back,
because
I
know
he
was
interfacing
with
other
groups
and
I
know.
Lucas
will
be
out
for
a
while.
I
was
not
directly
engaged
in
that,
so
I
can't
speak
to
it,
but
I
do
know
that
there
I
do
know
that
coop
spray
was
looking
for
had
already
integrated
and
I'm,
not
entirely
certain
what
the
states
of
cops
was
other
than
Justin
would
come
here
periodically,
but
I
don't
know
what
their
integration
plans
were.
Yeah.
A
A
Lee,
did
you
have
a
chance
to
look
at
any
of
the
Etsy
D
stuff
at
all,
I
know
you.
We
talked
about
a
brief
and
Monday.
No
okay
right,
there's
not
really
much
to
chat
about.
If
you
have
a
chance
to
look
through
the
backlog
as
we
continue
to
groom
lit,
feel
free
to
chime
in
if
you're
interested
in
executing
an
x
and
I'll
try
to
assign
you
where
I
can.
A
Canadian
mga
I
did
not
yet
I
figure
out
late,
Oh
Rob
is
back
on
that,
but
I
was
I
was
going
to
put
a
separate
milestone
for
GA
items
and
go
through
the
GA
list
and
try
to
get
a
canonical
set
of
backlog
items
that
we
need
to
get
done
in
order
for
us
to
reach
to
GA
they're
part
of
it
is
on
110
right
now.
Already
other
things
are
punted
out,
but
I
have
not
done
that
bit
yet
I'm
still
working
through
the
entire
list,
just
to
get
it
curated
and.
F
F
B
B
B
A
Gave
up
on
trying
or
the
the
standard
policy
is
we
don't
want
to
be?
First,
that's
the
statement
that
we've
made
for
a
long
time
was
that
we
would
rather
have
other
people
pilot,
and
we
will
be
a
happy
second
or
third
group
to
go
along
with
that.
I
know
that
control
was
possibly
going
to
be
one
of
the
first
groups
to
completely
split
off,
but
their
primary
reason
behind
that
is.
We
don't
want
to
own
all
of
the
other
technical
debt
that
will
come
along
with
it
and
I.
A
Don't
think
anybody
does
really
but
I
think
somebody's
got
to
write
that
bullet,
but
that's
been
the
standard
operating
procedure
for
this
thing
as
well
as
several
other
SIG's.
They
don't
want
to
be
the
first
group
to
like
eat
all
of
the
infra
and
all
the
other
costs
that
are
gonna,
be
associated
with
splitting
out
the
repo
and
trying
to
release
the
bits
and
coordinate
everything
together.
B
E
I
was
just
gonna
say,
like
I,
think
spanning
repose
might
be
possibly
introduced,
more
complexity
than
just
moving
the
repo
in
general,
but
my
question
here
was
mostly
in
alignment
with
like
what
exactly
is
this
technical
debt
we're
concerned
about
like?
Is
it
just
infrastructure
cost
or
is
it
the
cost
of
actually
getting
me
the
code
up
and
running
again
scandal,
oh
I,
think.
A
It's
mostly
infrastructure
and
tooling
costs
to
make
sure
that
it's
all
that
it
in
place,
and
sometimes
you
don't
want
to
be
the
first
person
to
like,
find
all
the
bugs,
because
you
spend
a
lot
of
time
just
dealing
with
that.
That
was
I
think
the
primary
drivers
again.
This
is
probably
a
good
conversation
for
for
Robby
to
as
well
to
chime
in
on
that's
my
understanding
of
the
history,
Joe
I
know
you
were
in
those
conversations
too.
If
you
have
a
hot
take
on
that
at
all
No.
E
A
F
Well,
you
know
cops
actually
doesn't
testing
it's
a
different
repost,
my
dentists
and
stuff.
For
me,
the
big
thing
is
the
release
stuff
right,
like
you
know,
it's
starting
to
look
at
stuff
like
like
the
Anago
stuff
like
making
that
multi
repo
would
be
a
big
project,
then
you're,
not
a
Google's,
actually
difficult.
A
Luckily,
Jeff
Grafton
has
solved
the
big
part
of
the
all.
The
artifacts
belong
in
one
repo
thing,
so
the
Basel
builds
for
1/10
have
all
the
artifacts
there,
including
the
RPMs
and
the
Deb's,
and
all
of
the
container
bits.
So
you
can
actually
build
everything
for
kubernetes
within
the
main
repository.
That's
not
how
I
released
yet
because
there's
still
a
problem
with
cross
builds
and
basil.
So
all
of
this
jiggery-pokery
is
there
in
the
mainline,
but
it
is
not
all
enabled
because.
F
C
F
A
A
So
we're
kind
of
we're
kind
of
reinventing
the
the
same
logic
and
reasoning
we
originally
had
with
wanting
to
be
second
or
third
I,
just
feel
like
everybody's
gonna
say
that.
Well,
it's
a
legit
reason
like
you
know:
somebody
needs
to
eat
the
cost
and
it's
gonna.
It's
gonna,
be
a
non-trivial
cost
right,
Joe.
A
H
F
Well,
I
I
think
you
know
right
now,
I
believe
most
people
install
cubeb
and
by
installing
their
OS
Edge's,
which
also
include
the
cubelet,
which
are
part
of
the
rest
of
the
sort
of
released
bundle
and
system
and
I
think
people
I
think
if,
if
we
release
cube
admin
out
a
band
with
different
versions
and
width
times,
Q
I
think
that
would
really
confuse
people
also
I
think
right
now
the
expectation
is
is
that
these
things
are
released.
You
know
right
now,
I
think
most
people
see
CQ
advertised
as
part
of
the
release.
I'm.
F
A
B
A
B
B
B
B
So
they're
saying
that
you
have
to
continue
discussing
this,
they
had
everything
lined
up
before
they
wanted
to
use
tortoise
luck,
charms
which
is
I'd,
say
confusing,
because
their
charts
and
they're
only
two
characters
that
are
different,
charms
and
and
juju
charms
would
make
more
sense.
But
besides
that,
it
seems
like
people
didn't
want
to
create
two
branches:
slash
juju,
something
Reaper.
A
A
So
that
way
people
can
contribute
without
having
to
feel
like
they
need
to
buy
into
some
other
CLA
or
some
other
process,
and
that's
that's
the
typical
barrier
that
most
people
we've
identified,
and
that
doesn't
mean
that
it
applies
to
everybody,
but
I,
don't
think
it
belongs
in
the
main,
kubernetes,
repo
and
I.
Don't
think
it
belongs
in
the
main
communities
org!
A
Is
that
they've
thought
about
having
a
hierarchical
structure
which
includes
SIG's
having
an
organizational
ownership
so
that
way,
sinc
cluster
lifecycle
could
have
many
repos
under
its
umbrella
right,
where
juju
terms
could
be
one
of
them,
and
then
people
can
coordinate
with
that
and
stuff
and
then
the
policy
everyone
a
dictate
that
could
be
determined
by
the
sig
higher-level
policy
that
would
apply
to
the
main
kubernetes
or
it
might
be
much
more
strict
than
what
would
apply
it
to
signal
organizational
structure.
Okay,.
A
B
G
A
Yes,
we
are
piloting
that
stuff
just
to
unblock
some
other
folks
who
don't
really
care
where
the
code
lives,
but
I
know
that
some
people
will
have
very
strong
opinions
about
where
things
should
live
and
why-
and
we
want
to
make
sure,
we're
judicious
and
thoughtful
about
how
we
roll
this
out.
So
they're
piloting
some
ideas
and
giving
feedback
and
I'm
sure
there's
probably
be
an
iterative
feedback
loop
as
part
of
this,
but
again
I,
don't
take
my
word
as
gospel.
A
A
Well,
yeah,
there's
there's
a
bit
there's
several
aspects:
there's
contributions.
This
is
kind
of
treading
in
other
areas,
not
just
cig
question
lifecycle
related,
but
it
it
mark.
It
treads
into
marketing
aspects.
It
treads
into
what
is
considered
blessed
by
the
community
right
and
we
want
to
make
sure
that
we're
we're
judicious
and
mindful
so
I
do
you
have
any
other
comments,
Joe
I'm,
trying
to
I'm
trying
to
tap-dance
around
this,
as
judiciously
as
I
can
well.
F
I,
you
know
I,
so
Brendan
has
a
proposal
out
there.
I
think
the
the
the
general
thinking
is
that
you
know
three
levels
of
projects:
there's
projects
that
don't
have
to
be
toys,
econ
but
associated
projects,
or
something
like
that
where
it's
like
they
get
to
use
the
C
and
C
f/c
la
they
get
other
build
tooling.
So
they
get
a
lot
of
the
process
and
it
sort
of
smooths
the
way,
but
there's
no
implied
endorsement
or
approval
going
on
there.
F
So
this
is
pretty
much
like
as
long
as
you're,
not
sort
of
violating
the
code
of
conduct
go
for
it.
The
next
level
would
be
sig
sponsored
where
it's
up
to
the
sig
to
actually
decide
how
to
do
this
stuff
and
then
and
then
the
final
level
is
its
core
stuff.
Now
this
would
be
applied
to
new
repos
moving
forward.
F
The
sig
sponsored
stuff
would
be
like
sub,
so
I
think
a
good
example
would
be
say
like
there,
the
reschedule
or
the
D
scheduler
or
whatever
it's
called
this
week
out
of
six
scheduling
or
the
the
the
no
problem
detector
out
a
signal
right.
So
those
are
things
that
are
sort
of
super
closely
associated
with
the
project,
but
are
sort
of
built
on
a
different
release,
cycle
and
sort
of
you
know
over
time.
F
You
know,
maybe
those
things
get
more
official
and
may
migrate
back
into
the
into
the
core
kubernetes
organization,
but
there's
no
guarantee
around
that
for
explicit
timeline,
but
the
questions
around
sort
of
which
organization
that
shows
up
into
whether
each
think
it's
their
own
org
or
whether
or
whether
the
the
sub
repos
have
sort
of
a
specific
name.
That
includes
the
sig
name
in
the
name
of
the
repo
I.
Think.
Well,
you
know,
that's
that's
still,
I
think
up
in
the
air
a
little
bit.
F
The
the
argument
against
the
org
is
that
org
sprawl
is
just
more
stuff
to
deal
with
and
and
especially
as
we
start
looking
to
attach
BOTS
and
sort
of
get
the
CLA
about,
and
all
that
stuff
going
up.
They're
just
getting
that
tooling
done
ends
up
being
a
real
pain
in
the
butt.
So
that's
something
that
I
think
is
the
lack
of
tooling
ends
up
slowing
a
lot
of
those
stuff
down
I,
just
repeating
a
lot
of
the
stuff
that
Tim
said.
B
It
seems
like
you
know
if
you
were
to
put
something
and
you
may
get
out
slash
Cabrini's
across
the
lifecycle,
slash
my
cool
project,
the
project's
name,
sort
of
you
know,
under
the
shadow
of
his
boring
name
of
the
organization,
there's
a
bit
of
a
marketing
problem
there.
So
it
doesn't
play
nicely
that
I
can.
F
The
idea
is
that,
then
we
want
to
make
it
clear
that
the
endorsement
is
coming
from
the
cig,
not
from
sort
of
all
of
kubernetes
and
so
there's
definitely
different
levels
of
endorsement
and
community
involvement.
That
goes
along
with
that,
and
this
is
part
of
just
sort
of
really
empowering
SIG's
to
actually
sort
of
on
their
own
destiny
around
this
stuff.
B
B
B
A
B
A
We
could,
but
we
definitely
want
buy-in
from
the
Google
side
because
they
depended
upon
that.
The
most
right,
like
all
the
testing
jiggery,
like
all
of
testing,
depends
upon
that
stuff
for
standing
up
clusters.
So
this
is
kind
of
this
goes
back
to
the
original
conversation.
How
can
we
get
coop
ATM
to
be
a
default
testing
deployment
for
Google
so
that
we
get
this
virtuous
cycle?
B
Yeah,
just
you
know,
I
think
we
should
shouldn't
forget
those
things.
I
mean
I,
just
find
it
very
difficult
to
deal
with
bits
and,
let's
say
like
searching
for
issues
everything
references
cube
ad
in
one
way
or
another,
on
the
yeah.
We
have
a
repo
farmer
and
with
some
other
issues,
but
that
doesn't
actually
that
money,
because
people
probably
choose
that
well.
A
A
You
know
we
should
open
the
issues
here
and
any
any
legacy
stuff
he
kind
of
filtered
out
so
I
put
that
list
in
those
queries
into
the
main
into
the
sink
cluster
life
slack
and
I'm,
hoping
other
folks
can
help
to
triage
that
bit.
I've
been
focusing
my
efforts
on
PRS
and
the
actual
Canadian
repo
items
to
get
them
all
sorted
through,
so
the
PRS
are
good
and
at
least
they
were
good
as
it
last
week.
A
B
B
You
have
been
I've
been
trying
to
to
get
to
to
get
the
green
off
PRS
on
the
issues,
but
it's
just
sort
of
the
minute
difficult
than
half
and
end
up
on
just
a
random
PR,
rather
than
anything.
That
is
more
specific
to
this
sake,
and
it
seems
like
oh,
it's
actually
kind
of
related
yeah
but
I
mean
that's,
probably
just
part
of
the
process
for
me
to
get
used
to
to
looking
through
them
and
find
my
ways
around
anyway,
yeah
so
cool
all
right.
Well,
we
exhausted
those.
Two
topics
are.
A
If
someone
can
poke
me
on
slack
with
how
Lucas
interface
with
those
folks,
that
would
be
super
helpful
because
there
are
a
couple
of
PRS
that
are
important,
that
we
need
to
get
in
and
done
and
I
just
want
to
make
sure
we
I
can
have
a
line
of
communication
in
line
of
sight
for
execution.
So
that
way
they
know
I
can
help
them
out
and
their
peers
aren't
fit
routing
right.