►
From YouTube: 20181106 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
Hello
today
is
Tuesday
November,
6
2018.
This
is
the
standard
clustered
life
cycle
call.
We
have
a
couple
of
agenda
items
on
the
list
and
other
things
to
discuss.
So
if
you,
if
you
want
to
discuss
something
in
this
call,
please
add
it
to
the
agenda.
The
applet
should
be
opened
up
for
anyone
who
is
on
the
mailing
list,
starting
off
31
and
31
I.
Don't
know
if
he's
on
yet.
B
Yeah,
this
is
the
pyramid.
She
is
my
guitar
ID
yeah
I
wanted
to
talk
about
this
after
I
got
the
comments
on
the
four
requests.
I
have
the
link
to
the
pull
request,
so
basically
that
the
change
adds
a
new
feature.
It's
a
new
add-on
that
you
can
deploy
on
clusters
to
help
with
DNS
latency.
It's
like
a
cache
for
kube
DNS.
The
change
I
have
is
adding
support
for
it
in
the
ete
scripts
and
GCE
class
recreation.
I
was
told
yesterday
that
we
should
we
should
support
it
for
all
the
other
tools.
B
A
So
then,
if
it's
alpha,
my
request
would
be
to
add
a
feature
gate
to
something
like
Kubb
ADM.
If
folks
want
to
tinker
with
it
has
been
to
test
it
as
part
of
their
deployers.
They
have
the
capability
to
do
so.
But
if
it's
defaulted
only
on
GCE,
that's
not
enough
for
the
community
to
understand
what
you
know,
how
to
enable
it
for
their
deployments
right
and
we
should
probably
also
have.
B
B
A
I
think,
eventually,
that
will
be
a
requirement
as
it
gets
to
GA.
It
is
not
is
definitely
not
a
requirement
for
Alfa,
but
it
would
be
nice
to
have
some
feature
gated
enablement
and
just
regularly
touch
base
with
the
sig
to
let
them
to.
Let
folks
know
who
you
know
there
are
people
on
this
call
right
now
who
deploy
across
any
number
of
environments
for
a
comedian
might
be
one
of
them,
but
you
know
Justin's
here
where
he
works
on
cops
and
towards
here.
A
B
A
C
Jenny
folks
and
I
do
hope
that
we
yeah
I
mean
I
was
just
gonna
say
like
I
think
this
would
be
great
to
get
into
maybe
it'd
be
good
test
form
for
the
bundle
as
well
right,
like
we
shouldn't
block
up
on
the
PR
on
the
bundle,
but
I
mean
like
this
is
like
the
bundle
is
I.
Think
our
real
answer
here,
like
we've,
we've
always
had
a
sort
of
like
we.
Certainly
it's
it's
polite
to
mention
breaking
changes
to
all
the
installation
tools.
C
We'd
only
have
a
mechanism
to
do
so
and
we've
had
like
I
I,
don't
know
if
we
have
previously
had
a
strict
policy
that
you
had
to
get
things
into
cube
ATM
before
you
could
merge
them
right
if
there
are
experimental
features
or
helpful
features,
I
think
I
personally
think
that,
like
the
cops
API,
we
don't
necessarily
want
you
to
feature
a
person
designing
the
cops
how
that
surfaced.
So,
like
I
am.I,
we
have
a
different
perspective,
I
guess
no.
A
C
D
So
Tim
are
you
proposing
that
we've
sort
of
blocked
introducing
new
features
that
we
are
trying
to
test
based
on
cube
ATM
is
that's
what
it
sounds
like
you're
saying
for
this
PR
is
we'd
like
to,
or
the
author
would
like
to
try
enabling
some
optional
new
functionality
to
see
if
it's
useful
and
I
feel
like
the
point
where
we
want
to
put
in
qadian,
is
when
we
flip
it
on
by
defaults.
Well,.
A
C
A
A
C
D
D
Because
we
don't,
presumably
we
also
want
to
open
a
tracking
issue
for
cops
in
cube
spray
and
each
of
the
other
sort
of
signal
sub
projects
right
like
I,
think
that's,
maybe
what
Justin's
point
was
initially
is
cube
idiom,
a
different
class
of
sub-project
than
the
other
sub
projects.
Right.
If
you
look
at
our
our
sig
page,
we
have
lots
of
sub
projects,
many
of
which
are
installers.
A
I
think
this
is
a
broader
like
question:
I,
don't
it's
it's
weird
in
that
comedian
is
part
of
the
main
KK
repo,
which
is
both
beneficial
and
not
at
the
same
time,
but
I
do
I,
do
think
management
of
how
we
disseminate
this
information
is
super
important
I
think
that's
actually
a
much
broader
question
that
I
think
needs
to
be
answered.
I,
don't
necessarily
want
to
block
her
PR,
just
make
sure
that
other
people
are
informed.
So
that
way
we
can
other
installers
can
can
solve
this
problem
going
forwards.
C
I
would
be
fine
with
a
an
issue
in
some
repo
somewhere
with
just
the
maintainer
z'
ping,
but
like
CC,
donut
I.
Think
that's
reasonable,
I
think
it's
you
know
in
the
past.
Have
you
got
to
go
up
in
an
issue
with
each
issue
with
each
project
but
I?
Think,
like
honestly,
that
was
unreasonable
at
me.
It's
just
too
much
work
and
it's
too
easy
to
miss
one
until
he
said
not
know.
If
someone
has
missed
one
right
words
so.
A
How
does
this
sound
is
a?
Is
a
straw,
man
open
an
issue
within
KK
and
see
see
some
of
the
main
maintainer
z'
for
a
couple
of
the
key
installers
antoine
is
on
here
myself
and
justin
and
robbie,
and
we
can
loop
in
other
folks,
potentially
the
cluster
API
implementations
that
might
use
something
else.
Cuz
I
don't
even
know,
and
we
should
probably
send
an
email
to
the
list.
A
I,
don't
know
how
we
want
to
do
this
in
the
future,
but
this
is
gonna
be
a
broader
problem
for
us
that
we
should
probably
start
to
think
about
because,
as
as
features
get
promoted
like,
there
usually
needs
to
be
a
knob
somewhere
and
the
installers
control
the
knobs
right
and
eventually
over
time.
If
we
have
the
grand
unified,
what's
called
component
configuration
and
bundles
it'll
be
all
magical
and
rainbows,
but
until
that
time
we
need
to
figure
out
a
way
to
to
to
loop
everyone
else
into
the
efforts
that
are
ongoing.
B
A
And
twenty
minute
you
were
the
first
and
agenda,
but
you
were
here.
You
had
a
question
about
a
survey.
Yes,.
E
A
G
Survey
was
already
executed
and
the
results
were
gonna
be
outsole
and
I
see
some
overlapping
between
this
survey
Antoine
and
the
one
that's
country
back
state.
So
if
you
joined
the
country
back
shadow
on
slack,
I
think
that
you
can
communicate
it
and
basically
gather
feedback.
If
this
is
a
possible
survey
to
publish
publicly
okay.
F
A
G
E
Thinking,
okay,
we
look
whatever
I,
think
something
like
quite
more
specific
to
life
cycle
and
say:
are
you
tickling?
Are
you
using
her
to
create
your
certificate
for
the
city,
yes
or
no?
What,
instead
of,
are
you
using
if
it's
a
custom
one?
What
kind
of
language
is
it
an
SI,
Belko,
etc
so
quite
understand
most
of
ecosystem
around
really,
specifically
lexical?
Are
you
using
Kiba
TM
and
if
no
I,
that's
a
truck
such.
C
Awesome,
yes,
I
think
my
microphone
is
very
laggy,
so
I
thought
just
by
I'm
talking
over
anyone,
but
yes,
I
sent
the
we
have
the
sed
ATM
kept,
which
is
been
out
for
a
couple
of
weeks
now.
I
think
we
sent
it
to
the
cig,
toaster
lifecycle,
mailing
list
and
sort
of
started,
a
lazy
consensus.
I
didn't
see
any
disagreement.
I
think
we
sent
it
to
the
sed
main
list
as
well:
I,
just
I,
guess:
I
served
lost
track
of
the
process.
I
guess
the
next
step
is
to
approve
it
and
create
a
sub
project.
A
D
A
A
E
C
So
I
might
like
me
Mike
again,
you
said
we're
starting
with
the
HCDA
DM,
which
is
a
command
line
tool
from
platform
9,
which
I
think
essentially
does
you
you
around
you're
responsible
for
copying
it
and
they're
rebuilding
automation
on
top
of
that,
but
there
will
always
have
to
be
some
form
of
and
we're
books,
even
with
the
automated
layer.
That
always
happen
least
some
form
of
authentication
I.
C
Think
Tim
was
suggesting
doing
something
involving
the
bootstrap
token,
if
I
recall
correctly,
but
honestly,
I,
don't
think
we
have
a
crystal
clear
design
yet
on
how
we
can
make
that
smoother.
I.
Think
if
you
have
any
good
ideas
like
I
think
you
mentioned
bold
integration
like
I
think
we
would
love
to
be
able
to
make
that
supported.
I.
C
For
you
know
you
we
decide
on
a
sufficient
token,
and
then
we
enable
you
to
grab
it
fur
on
the
master
when
you
join.
Let's
do
forgot
the
details
of
what
sufficient
is,
but
I.
E
C
D
C
C
C
Yeah-
and
it
also
has
like
all
the
host
decisions-
have
an
external
source
of
truth,
which
they
can
rely
on.
So,
yes
and
presume
a
have
their
own
automation
systems.
Yes,
we
should
I
should
clarify
that
to
be
open,
source,
installers
and
I
guess
one
of
the
ways
we
could
do,
that
is
to
look
at
the
ones
which
are
sub
projects
of
semester,
lifecycle
and
any
other
important
ones
that
come
up,
but.
E
E
C
D
But
I
think
what
I'm
saying
is.
If
we
just
say
you
know
boot
cube
cluster
API
cops,
I,
don't
know
if
how
active
qubit
qaw
s
still
is
cube,
deploys
and
maintenance
notes.
We
don't
care
about
that
same
thing
with
Cuba,
Cuba,
DM,
cube
spray
and
maybe
mini
cube
right,
like
those
are
kind
of
the
main
ones
that
we
we
feel
like
actually
have
momentum
going
forward.
I,
don't
know
about
cube
idiom
di
uni
cluster
humor,
Nate's
anywhere
department.
G
C
D
A
D
D
Mean
depends
on
the
target
time.
Target
audience
is
cluster
installers.
Then
it's
itself
being
GA
doesn't
really
make
much
sense
if
nobody's
actually
using
it
right,
and
if
the
target
audience
is
end-users
using
it
directly
in
the
manual
mode,
then
it
doesn't
matter
if
cluster
installers
use
again
so
much
I.
C
Was
just
think
I
just
don't
know
what
graduation
criteria
does
so
I'm
I
was
a
little
fuzzy
on
what
it
meant
and
sort
of
what
the
thing
was.
I
I
did
write
the
draft
I'm
like
and
says
consider
someone
is
used
by
the
majority
of
OSS
cluster
installations,
so
I
guess
I!
Guess
by
always
this
customization,
like
terrify
cluster
installations
made
by
OSS
tooling,
rather
than.
C
But
the
cluster,
the
majority
of
the
clusters
is
what
I
was
thinking.
So
yes,
but
yes,
we
you're,
not
gonna,
have
a
survey.
We're
not
gonna
know
if
there's
some
dark
pool
of
kubernetes
clusters
in
some
of
the
data
center
run
by
the
NSA
somewhere
right.
We
just
don't
know,
but
yes,
so
I
don't
I
ii
think
about.
If,
if
we
want
to
say,
is
used
by
cops
and
coop
spray
and
mini
cube
and
q
ATM
and
a
s
and
y,
then
that's
that
works
for
me
as
well.
D
D
D
A
The
original
notion
of
how
to
use
Kemp's
it's
a
weird
constraint
of
this
particular
problem
and
that
capsules
originally
done
designed
to
replace
the
feature
process
for
alpha
beta
G,
a
promotion.
So
that
way,
it
was
clear
to
all
folks
who
were
looking
at
the
caps
that
this
is
the
canonical
source
of
truth
for
the
features
and
that
we
think,
because
it
was
programmatic.
You
can
rip
that
out.
So
you
can
see
a
global
set
of
features
and
what
their
current
state
was.
G
A
So
it
sounds
like
the
biggest
conundrum
or
question
is
like
what?
What
does
the
process?
Look
like?
Not
necessarily
this
agreement
done
that
we
should
have
this
as
a
sub
project,
but
logistics
of
how
we
disseminate
information,
which
I
think
is
probably
fine
as
a
TBD
at
this
point
and
probably
needs
refinement
within
the
kept
itself.
The
kept
process
itself
is
that
a
fair
statement.
A
Yeah
I'm,
that
that
seems
almost
like
a
meta
topic
which
we
should
potentially
bring
up
with
arch
and
not
necessarily
related
to
this
particular
thing,
but
could
apply
to
several
folks
who
are
trying
to
do
stuff.
I
will
put
it
on
the
arch
list.
Are
there
any
other
questions
comments,
complaints
concerns,
otherwise
it
generally
has
a
plus-one.
A
A
H
H
The
reason
this
issue
came
up
most
recently
was
because
we
had
a
sort
of
a
large
push,
a
large
merge
that
we
wanted
to
make
back
upstream,
and
we
have
been
doing
a
lot
of
a
work
in
a
fork
of
the
karez
tree
so
that
we
could
get
around
things
like
code
freezes
and
you
know
just
our
own
development
process.
So
when
we
put
that
merge
back
up,
it
was
very
large
and
raised
some
flags
about
hey
you
don't
this
is
the
right
place
to
be,
and
we
agree
with
that.
A
So
I
hadn't
made
a
question.
They
don't
know
the
answer.
This
not
gonna,
be
nice
to
know
it'd
be
like
who
are
the
target
consumers
of
this?
That's
not
just
canonical
so
like.
Are
there
other
folks
on
the
sig
that
be
willing
to
sponsor
the
work
or
want
to
have
a
common
home
where
you'd
have
a
community?
That's
a
shared
community
within
this
group,
so
typically
there'd
be
like
a
number
of
shared
interests
like
SE
dat
and
has
almost
every
installer
would
like
to
be
able
to
use
that.
A
So
there's
there's
a
shared
group
of
folks
who
are
willing
to
help
maintain
this
and
work
on
these
things.
Typically
as
part
of
an
adoption
for
a
new
project
or
sub
project.
Inside
of
the
community
sings
board,
there
is
a
an
owner's
file
lists
of
who
are
the
core
maintainer
Xand
interested
parties
who
are
going
to
help
contribute,
maintain
and
keep
this
up
to
date
so
that
it's
not
just
a
something
that
would
folks
push
over
the
wall,
periodically
sure.
H
So
I
don't
have
a
good
answer
for
that
I'm
happy
to
look
through
the
sig,
List
and
figure
out
who
I
think
would
be
ideal.
Co
maintainer,
certainly
canonical
would
be
the
primary
one
that
would
be
interested
in
putting
juju
related
source
here.
So
I
can
take
that
to
find
you
know,
sort
of
that
combing
tainer,
spot
I.
D
D
H
So
we
started
this
and
we,
when
we
first
put
our
source
up,
we
likened
the
juju
related
bits
to
cops
kou,
Mary's
anywhere
stuff
like
that,
and
so
we
just
sort
of
followed
what
other
projects
like
ours
were
doing.
So
that
was
the
initial
motivation
to
get
us
landed
in
the
in
some
way
related
to
the
chimeras.
Orgs
I
think
it
makes
sense
when
people
used
you
to
find
the
charm
source
living
alongside
the
other
Kate's
projects.
Just
because
we
do,
we
do
touch
a
lot
of
them.
D
Anywhere
is
is
on
its
path
to
the
kubernetes
act.
Once
we
can
get
it
there.
The
all
the
new
stuff
we're
creating
is
under
kubernetes
things
and
that's
where
all
sort
of
the
sync
some
projects,
ideally
would
live,
and
there
were
in
some
discussion
dirty
right
folks,
since
they're
an
incubator
right
now
about
sort
of
where
they
should
end
up
so
I
think
it
between
those
two
I
would
certainly
vote
for
being
under
six
instead
of
under
committees.
D
G
Mean
usually
the
biggest
piranha
mystery
if
we
need
to
introduce
entrant
tests
related
to
germs
and
eventually
somewhere
somebody
pings
pings
us
about
some
test
failures.
We
need
to
find
the
maintainer
to
take
a
look
at
these
and
by
also
introducing
like
maintainer
to
a
six
copper86
repository.
G
A
A
typical
way
to
proceed
forward
here
that
other
sub
projects
have
done
is
create
almost
like
a
shell
repo
with
some
of
the
details
and
percolate
information
and
other
information
inside
of
the
owners
files
and
solicit
send
an
email
to
the
sig
requesting
information
about,
like
other
people,
are
interested
in
maintaining.
Please
look
at
this
separate
repo
with
this
Warner's
file
listed
over
here
and
then
folks
can
comment
on
it
and
then
propose
just
like
a
voting
similar
to
what
Justin
did
so
that
way
they
can
be
added
as
a
sub
project.
E
A
Yep
I
agree:
Justin
will
cause
because
he
got
the
mic
delay.
You're
gonna
say
something
I.
A
A
A
Everything
else
is
like
you
have
all
these
deployment
tools
deploy
all
over
the
planet,
and
sometimes
the
signal
is
not
the
best
right
and
it
makes
a
lot
more
sense
to
have
some
of
these
things
be
periodic
signals
than
it
does
to
have
it
be
PR
blocking
for
every
PR,
and
that
way
you
put
the
maintainer
z'
for
those
upper
other
projects
on
the
hook
for
managing
that
signal,
because
some
of
the
problem
with
some
of
this
other
stuff
is
you
get.
Flakes.
A
Providers
are
like
the
biggest
source
of
flakes,
who
knew
that
I
didn't
an
AP
eyes
were
a
thing,
but
I
don't
know
if
anyone
had
any
thoughts
about
that
one,
which
was
about
the
kind
that
the
second
bit
was
understanding
where
the
where
the
current
state
of
the
coupe's
pre-integration
was
I.
A
So
kind
that
the
purpose
of
integration
with
kind
is
to
take
out
some
of
our
peer
blocking
jobs.
We
have
so
many
PR
blocking
jobs,
it's
kind
of
bananas
now
so
every
time
any
user
submits
a
PR
or
any
developers
from
it's
a
PR.
It
goes
off
and
it
basically
deploys
clusters
for
all
of
these
different
providers.
And
then
it
also,
you
know,
runs
in
ten
tests
and
gets
back
the
signal.
But
the
problem
with
that
is
that
it's
very
flaky
it's
notoriously
flaky,
especially
if
it's
done
on
a
per
PR
basis.
A
G
G
E
Okay,
yeah
I
wasn't
responsible
for
the
PR,
but
is
there
any
way
to
do
it?
So
was
it
like
just
say
we
tried
this
image
and
we
report
the
status
results
and
starting
a
job
that
going
to
do
everything
on
the
CRA,
because,
right
now
we
we
we
have
already
up
our
CI
in
place
and
we're
going
to
use
it
also
on
the
emitter
and
back
aid
and
different
provider
and
I.
Think
it's
much
more
easy
to
say.
E
A
The
biggest
benefit,
the
only
the
major
benefit
was
knowing
that
you
could
always
deploy
the
latest
version
of
kubernetes
using
coop
spray.
That
was
kind
of
the
biggest
thing
that
we
were
looking
for.
If
it's
periodic,
then
we
don't
care
like
if
it's
periodic
and
reported
to
be
a
sort
of
federated
testing
into
test
group.
Just
so
that
folks
have
a
signal
at
the
end
of
the
release
cycle
to
say.
Oh,
this
deployment
tool
is
totally
failing.
A
It's
not
it's
not
via
PR
it'd,
be
periodic.
So
the
way
the
test
grid
periodic
jobs
work.
Is
they
basically
pulled
master
of
latest
kubernetes,
every
time
windowed
segment,
and
then
it
builds
and
deploys
that
be
a
standard
sort
of
it
says
on
deployment
on
GCE
right
now
and
you
can
have
Feder,
you
can
have
other
things
to
deploy
to
other
targets
that
cops
doesn't
for
a
tip
us
obviously,
and
then
it
runs
the
intent
test
suite
in
the
reports
the
results
back
now.
A
E
C
I
I
actually
would
be
in
favor
of
like
moving
to
kind
and
doing
less
of
the
cloud
tests
like
today.
We
have
something
we
rated
Westlaw
stole
quota
and
so,
like
I'm,
pretty
sure
afraid,
like
I'm
made
for
passing
any
PRS,
but
parent
we're
still
passing
so.
Apparently,
people
who
think
we
test
enough
but
I,
think
the
thing
which
I
would
find
useful
when
I'm
looking
at
and
like,
for
example,
an
AWS
PR,
is
to
be
able
to
ask
for
that
test
to
be
run
and,
like
we've
had
this
on
other
ones.
C
G
G
A
Think
it's
almost
critical
because
we've
we've
gotten
into
this
state
for
a
we've
been
in
the
state
for
a
long
time.
It's
almost
pathological,
where
you
know
people
do
slash,
retest,
slash
retest.
You
know,
like
all
the
time
over
and
over
and
over
again,
to
try
and
get
something.
I
was
even
gonna,
make
a
joke.
T-Shirt
for
goop
Conda,
slash,
retest,
slash,
rotisserie,
test
LG
TM
merged,
but
it's
it's
frustrating
for
developers
in
it's,
not
yeah.
I,
don't
think
it's
beneficial
for
anyone,
especially
from
a
cost
perspective.
G
G
A
G
A
My
goal,
talking
about
it
within
the
sig,
is
to
understand
what
the
time
horizons
will
be
and
whether
or
not
there
are
interested
parties
to
help
move
their
jobs
to
periodic
s--
to
help.
You
know,
get
people
out
of
the
blocking,
merge
queue
problem
and
to
try
and
accelerate
that
turn
that
it
takes
so
that
way
we're
you
know.