►
From YouTube: Kubernetes SIG Cluster Lifecycle 20180619
Description
Meeting Notes: https://docs.google.com/document/d/1deJYPIF4LmhGjDVaqrswErIrV7mtwJgovtLnPCDxP7U/edit#heading=h.pzy4m7bi9skj
Highlights:
- 1.11 release status
- Need to define a support matrix for kubeadm upgrades
- Test status
- Where to document alpha features
- Pending docs PR
Planning for 1.12 will start next week.
A
Hello
and
welcome
to
the
June
1920
18
edition
of
Cigna
cluster
lifecycle
meeting.
Hopefully
one
of
our
last
meetings
before
the
commemorates
1.11
release
and
as
such,
the
agenda
is
mostly
focused
around
what
we're
doing
for
that
release.
So
Tim
you
wanna,
kick
us
off
with
the
cube
admin
milestone
issue.
B
Okay,
I
just
have
a
couple
questions
with
regards
to
some
of
the
documentation,
things
that
are
in
flight.
The
only
one
that
I
haven't
seen
live
Amir
is
the
set
up
copy,
edits
and
I,
don't
necessarily
know
if
we
need
to
modify
these
they're
already
exists,
the
main
documentation,
but
this
is
a
separate
one.
Are
there
updates
to
the
stock
that
are
in
flight
in
the
PR,
so.
C
B
D
That
one
is
probably
gonna
take
some
time
too
so
the
implementation,
because
it's
probably
gonna,
take
some
time
to
update
as
there
being
a
lot
of
implementation.
Detail
changes
in
the
latest
in
the
registries.
But
if
we,
if
we
could
send
a
quick
patch
or
something
that
the
current
that
just
states
that
the
current
implementation,
because
dark
is
for
110
and
then
whenever
we
have
time
to
do
111
as
soon
as
possible,
of
course,
but
not
is
not
blocking
for
a
new
Reese.
So
then,
then
we
can
just
say
that
here
we.
C
B
D
D
B
B
C
C
B
C
D
B
D
B
D
B
That's
part
of
this
comments
here
and
I
wanted
to
there's
a
broader
discussion
about.
Should
we
even
be
documenting
alpha
things
at
all
in
the
main
Docs,
or
should
we
be
linking
out
configuration
stuff
to
Doc's
that
we
host
on
the
cobia
idiot
or
marked
on
pages
of
the
cuvette
DM
repo,
because
putting
major
out
putting
major
documentation
out
there
in
the
wild
for
alpha
grade?
D
So
it
the
the
way
we
handle
the
config
stuff
is
we
handle
it
on
a
bi-level,
but
since
companies
like
what
is
it
called
like,
the
policies
are
very
strict,
which
is
super.
We
and
we
still
intend
to
change
the
structure
we
have
Tabata
over.
Obviously
so,
but
we
still
treat
the
consumers
of
the
config
as
beta
that
we
always
have
an
upgrade
pads
and
as
long
as
as
long
as
you
have
something
you
can
go
from
version
to
version
and
always
be
like
successful
in
that,
so
that
I
I
do
think
this.
B
Then
we
should,
we
should
definitely
put
some
clauses
and
warnings
that
warning
you
know
you're
still.
We
with
that
exact
clause
that
you
stated
we
tried
to
treat
this
as
beta,
but
this
is
an
alpha
config
file
which
could
potentially
have
incompatible
changes
going
forwards.
So
as
long
as
you
put
that
clause
at
the
top
I'm
fine
with
making
that
modification
yeah.
B
Know
Chuck
it
and
ribbon
are
finishing
up
the
h8
documentation,
which
is
you
know,
yeah,
it's
a
lot
better,
but
the
upgrade
h
a
configuration
is,
which
is
part
of
chucks.
I.
Think
note
here:
yeah,
it's
related
to
this
note
is
not
currently
working
because
of
how
we
change
things.
Then
talk
about
that
Chuck
yeah.
F
So
I
guess
this
is
kind
of
a
conversation.
I
wanted
to
have
as
I
joined
and
got
ramped
up
on
coo
Vidya,
and
then
there
are
a
lot
of
different
methods
of
installing
kubernetes
that
we
provide
and
we
don't
necessarily
care
about
all
of
the
options.
I
think
it
can
help
us
going
forward
to
stay
focused
and
have
some
kind
of
a
support
matrix
of
the
configuration
options
we
do
care
about
and
want
to
support
for
various
reasons.
F
So,
for
instance,
right
now,
if
you
have
an
external
@cd,
it's
not
possible
to
run
Kubb
atmo
by
Covidien,
create
apply
it
just
won't
work.
You
can
do
a
lot.
You
can
do
like
manual
certification,
sort
of
starts
moving
around
to
get
it
to
work,
but
it's
not
something.
That's
supported
out
of
the
box,
all
right.
D
B
Is
a
broader
question?
We're,
like
we've,
never
actually
said
we're
even
alpha
great
support
for
AJ
and
we're
kind
of
trying
to
tread
there
by
the
time
we
get
to
GA.
So
like
a
lot
of
the
work
that
Chuck
and
Jason
have
done
in
simplifying
the
instructions
and
getting
at
CD
secured,
and
you
know
that
the
work
that
Lee
also
did
and
as
well
as
fixing
a
lot
of
the
upgrade
issues
that
currently
that
existed,
is
to
get
it
to
that
stage
right.
But
we
haven't.
B
We
haven't
officially
said
that
we
support
this,
so
I
love
folk,
scree,
ated
the
documentation,
but
they
were
community
driven.
They
weren't
necessarily
driven
by
the
sig
itself.
So
I
would
just
want
to
be
clear
that,
like
we
need
to
make
sure
we
have
formalization,
and
we
had
some
of
that
in
the
original
Docs,
but
I
think
it
belongs
in
the
Canadian
repo.
That
says,
like
here's,
the
current
feature
list
and
here's
what
the
grade
of
support
is
for
kerb,
ATM
I
would.
B
D
D
B
It's
in
the
client
bill,
repository
DM
repo,
absolutely
go
for
it
and
we
should
link
to
it
right,
but
not
in
the
mainline
Doc's
right.
That
was
one
of
the
things
where
we
don't
even
want
it.
We
don't
even
want
to
tell
users
about
alpha
grid,
feature
support
unless
they're
looking
for
it
right,
because
it's
more
confusing
to
the
to
the
newcomers
to
the
community
than
it
is
to
actually
illuminating
them
or
wanting
to
try
something
new
right.
D
Okay,
but
but
I
was
thinking
like
if
we
say
that
we
have
both
external
HDD
and
internal
at
CD
that
we
support
shouldn't.
We
say
that
that
we,
this
we
support
these
two
modes,
that
we
support
both
external
to
the
internet
city
together
with
static
bones,
for
example.
It
isn't
that,
and
everything
is
like
GA
of
that.
Is
it?
Isn't
it
something
we
should
mention
or.
F
B
This
is
specifically
for
the
H,
a
configuration
that
was
deployed
via
the
AJ
set
up
previously
right.
This
is
not
just
if
somebody
else
had
their
own
certs
and
certificate
environment
where
we
didn't
have
to
set
it
up.
You
know
it
was
done
completely
external,
then
it
should
not
be
a
problem,
but
I
think
this
was
the
partial
configuration
that
was
deployed.
Is
that
correct,
yeah.
D
C
D
Okay,
but
yeah
of
these
I
would
go
in
for
GA
just
like
quickly
as
you
asked,
I
would
say:
externality
internet
CD,
together
with
single
master
self-hosted,
is
the
the
matrix
we
want
to
support
then
will
we're
gonna
add
something
multi
masks
that
we
don't
haven't
agreed
on,
yet
as
alpha
and
and
then
I
don't
know.
If
we
ever
want
to
do
self
hosting
like
that,
much
well,
we'll
just
see
yeah
I've
kind
of
it's
a
it's
a
question
mark
atleast,
yeah.
B
A
A
A
B
D
D
So
that's
that's
my
TPS
and
the
cubed
cubed
I
mean
integration
stuff
fixes
the
document
version
file
thing
the
other
issue
that
you
had
in
my
system
there,
so
whether
we
want
to
keep
them
in
the
milestone
with
this
PR
fixes
or
whether
we
want
to
just
move
them
to
the
next
like.
So
so
we
just
send
a
clear
message
that
to
the
release
team
that
are
also
looking
at
this
milestone,
as
these
create
a
new
documentation
based
on
user
stories,
document
control
and
cubed
configuration
file.
D
B
C
B
What
they're
concerned
about
is
the
next
part
which,
which
is
the
test
there's
a
couple
of
tests
that
are
showing
up
as
red
and
I.
Think
Aisha
is
on
the
call
to
the
the
first
test
that
I
looked
at
that's
a
little
concerning
is
the
Committee
mgc
upgrade
to
stablemaster
no
comedian,
110
and
111
looks
fine.
It's
just
the
master.
One
has
been
red
with
a
bunch
of
other
tests
that
I,
don't
think,
should
be
red
besides
the
ones
that
are
known,
knowns,
yeah.
B
D
B
D
C
B
D
But
yeah
just
to
clarify
I,
think
that
the
upgrade
110
and
111
also
it
doesn't
test
latest,
beat
us
the
tests
latest
latest
on
the
race
patches.
So
so
it
does
I,
don't
think,
there's
a
difference
if
we
kind
of
beta
or
not
or
like
an
RC
or
not.
It's
it's
testing
from
latest
latest
on
the
branch
anyway.
So.
B
E
B
D
B
That's
the
first.
The
second,
the
second
test
are,
is
not
related
to
Q
medium,
but
they
wanted
to
talk
about
it
here
because
Roberts
here,
so
we
can,
we
can
play
forever
on
us.
Is
that
there's
a
bunch
of
flakes
going
on
with
the
GCE
appeared
to
us-
and
this
is
all
out
of
/
cluster-
that
I'm
worried.
Okay,.
A
B
A
A
B
E
She
was
here
just
a
second
ago
yeah,
so
we're
we're
not
sure
we're
trying
to
get
a
little
more
data
because
of
the
way
test
grid
only
shows
a
limited
window
of
time.
We
wanted
to
see
just
anecdotally
if
people
feel
like
this
has
been
something
that
historically
has
actually
been
flaky
in,
because
this
release
is
pretty
clean,
then
are
we
just
noticing
that
this
is
the
one
area
that,
where
we're
seeing
some
red.
A
A
It
does
look
like
there
are
test
runs,
or
we
see
like
quite
a
few
failures
and
I
wonder
if
there's
also
some
failure
correlation
going
on
where,
if
there's
one
test
that
fails
a
couple,
others
sort
of
come
along
into
the
failure
party
yeah,
because
a
lot
of
the
ones
where
you
see
one
fail,
that'd
be
kind
of
a
swath
like
if
I'm
not
looking
at
em
screen
anymore,
but
back
on
611.
There
were
quite
a
few
that
failed
together
on
6/10
that
were
quite
a
few
Liefeld
together
and.
E
A
It's
yeah
I
wouldn't
be
surprised
if,
if
it's,
you
know
like,
let's
kill
a
node
and
wait
for
a
new
one
to
show
up
if
if
there
could
be
flakes
and
the
infrastructure
that
that
takes
a
little
bit
longer
than
we'd
expect
or
the
overall
system
is
like
slightly
changes.
Parameters
of
how
fast
things
happen,
and
if
we
have
like
hard-coded
time
outs
than
those
things
are
going
to
start
to
flake
over
time.
B
B
Usually
what
it
happens
is
it
goes
to
the
cig,
lead
the
sigelei,
then
figures
out
what
the
problem
is
and
then
actually
creates
a
PR
to
fix
the
test
if
it
if
they
can
I
think
what
we're
currently
doing
is
like
it's
one
of
those
things
where
it's
like
passed
on
potato.
No
one
wants
to
own
this
hot
potato,
because
flakes
are
not
fun
to
debug.
They
require
a
fair
amount
of
time
and
effort
to
be
able
to
root,
cause
them
and
sometimes
they're
cascading.
So
hot
potato
Robert
yeah.
B
B
D
B
So
I
want
to
make
sure
that
when
we,
when
we
do
something,
I'm
not
hit
too
hypocritical
of
calling
other
things
out
along
the
way,
so
I
just
wanna
make
sure
that,
like
the
stated
goal
is
that
we
will
strive
to
try
to
not
to
put
mainline
Alfa
features
in
the
main
box
because
of
the
support
nature
of
it.
But
as
long
as
we
have
a
clause,
you
know
in
the
front
of
it:
I'm,
okay,
with
it
yeah.
D
B
You
can
link
to
developer
Docs
like
in
our
repo.
We
have
developer
Docs
right
and
we
can
actually
have
developer
Docs
for
those
things
and
it's
a
totally
legit
thing
to
do
and
I
think
having
a
separate
section
which
denotes
as
developer
or
prototype
or
alpha
or
like
warning
warning
danger.
It
is
required
yeah.
G
C
B
We're
not
all
there
yet
so
what
we're
gonna
do
now
for
as
a
stopgap
is
update,
Lucas's
PR,
to
have
a
clause
at
the
top,
but
we
absolutely
need
to
like
address
the
documentation
structure
for
112
and
let's
do
it
earlier
in
the
cycle.
They
talked
with
the
doc
folks
so
they're
on
board
about
like
what's
the
process,
policy
and
procedures
that
we
have
for
how
the
documentation
should
roll
and
we
started
making
a
good
chunk
of
that
restructuring
in
111.
But
let's
like
let's
finish
it
off
in
112,
does
that
seem
fair,
yeah.
C
D
B
D
B
D
Not
sure
I
I
could
like,
if
you're
fine,
with
having
one
commit
more,
that
fixes,
like
I,
think
it
was
two
unit
test,
failing
just
some
like
silly
stuff,
but
anyway,
I
think
what
there
were
two
if
you're,
okay
with
having
one
commit
more
I,
think
he,
as
the
I
mean
he
is
probably
sleeping
right
now
here
to
the
time,
sounds
I.
Don't.
B
B
A
A
B
D
B
D
So
folks
are
aware,
because
now
we
say
that
the
the
deef
it
should
fetch
stable,
112
stable,
112
txt,
which
doesn't
exist
yeah,
so
so
that
is,
that
is
expected
and
fine.
You
just
have
to
like
put
points.
It
combination,
combination,
112,
see,
that's
why
we
merge
this
really
as
late
as
we
can
before
the
RC,
which
is
we
should
have
no
delta.
With
with
the
real
release.
G
D
G
D
No
now
now
I,
remember
why
why
we
don't
do
this
yeah,
we
can't
no
kiss,
because
when
so,
when
we
do
this
because
cubed
and
CLI
110
should
after
111
is
created,
we
should
see
a
cube.
Nmcli
110
should
still
create
110
clusters
by
default,
otherwise,
otherwise
it
will
attempt
to
create
a
new,
a
cluster
Thank
You
Betim
itself.
This
and
fail.
C
We
kind
of
already
covered
the
documentation.
I,
don't
really
have
to
talk
about
this.
We
have
like
some
new
peers
and
if
everyone
can,
please
have
a
look
at
your
peers
to
see
like
what
updates
you
have
to
make
Lukas
for
your
PRA
I.
Think
Jennifer
has
started
doing
a
copy
edits
and
we
should
probably
let
her
finish.
First
and
yeah
I
think
everything
is
up
to
date
in
terms
of
the
pRSM.
We
are
good
to
go.
D
B
So
I
I
don't
exactly
know
what
the
logistics
are
for.
Updating
the
release
notes
are
I
talked
with
I
was
on
the
release,
call
before
this
one
and
we're
supposed
to
be
doing
PRS
to
that
separate
repository
with
the
actual
release,
notes
and
I
do
know
that
Nick
is
behind
on
communicating
with
a
bunch
of
other
folks
and
there's
a
bunch
of
missing
details,
so
I'm
gonna
touch
base
with
both
Nick
and
Josh
I.
B
D
So
the
thing
is
and
why
basically,
what
I
was
trying
to
say
with
in
the
release
release
channel
is
that
having
having
this
as
a
G
doc?
Yes,
it
doesn't
work
in
China,
but
it's
which
is
sad,
but
it's
way
easier
to
actually
move
stuff
around
on
a
like
continually
and
for
a
lot
of
more
people
than
having
to
actually
send
the
pull
requests.
Wait
for
wait,
a
couple
of
hours
for
it
to
be
merged,
and
then
you
notice-
oh,
it's!
It
now
has
a
rebase.
D
You
know
you
have
to
rebase
versus
just
moving
this
stuff
route
because
there's
a
lot
of
just
okay,
it's
hearing
unsorted
now
I
should
associate
the
sig
with
it
and
this
isn't
wrong
sig.
Let's
do
it
in
some
other
place
and
let's,
let's
like
just
upper
case,
all
the
sig
here,
because
now
we
have
capital
s
but
not
IG.
Such
things
are
easily
changed.
D
Someone
from
the
release
team
should
should
do
the
I
mean
github
seems
to
have
failed
all
the
automation
around
this
as
the
PR
numbers
are
in
and
contributor
cuz.
We
were
just
don't
have
like
links
in
all
places
so
before
anyone
starts
to
do
doing
anything.
This
it
seems
like
nobody
has
no
sing,
have
done
anything
to
this
yet
so.
D
D
E
D
C
C
D
B
D
B
Think
112
a
ga
should
be
one
of
the
listed
items.
Let's
do
a
tree
view
type
of
view
from
a
single
dock
so
like,
let's
keep
it
realistic,
I
think
one
I
think
GA
is
achievable
112,
given
the
resources
we
have
now
like.
We
have
way
more
people
in
manpower
than
we
had
before
yeah,
so
we
can,
we
can
probably
accomplish
it.
So
I
think
I
think
having
a
single
dock
is
totally
reasonable
and
I
can
start
the
dock
and
send
it
as
part
of
the
as
part
of
the
list.
Okay,.
D
G
D
Who
asked
me
the
other
day
whether
we
could
like
move
a
from
Cuba
Nathaniel,
because
that's
what
we
have
in
our
like?
You
know
the
apps
repo
or
what
is
up
channel
I,
don't
know
what
it's
called
so,
but
it
does
work
on
like
way
more
than
any
home,
but
we
just
got
that
one
place
in
Google's
infrastructure
and
pushing
all
the
stuff
there.
So
so
getting
that
CN
CF
own
dimension
would
be
super
as
well
as
like
doing
basil.
I'm
gonna
ask.
B
That's
like
a
broader
release,
problem
question
which
has
existed
for
a
long
time.
I,
don't
know
how
to
cope
enough.
Eyeballs
I
can
do
it
from
steering
and
like
because
it's
got
to
happen
from
the
top
down
right,
like
yeah.
The
system
is
currently
in
place
to
support.
You
know
continuation
of
the
cycle,
but
it's
got
to
be
like
an
edict
from
on
high
that
says,
like
Michel
as
a
project
move
the
whole
ship
off
of
Google
release
process
onto
community-driven,
CN
CF
process.
D
Yeah
I
mean
getting,
or
at
least
trying
to
execute
on
some
of
that
stuff
for
want
well
would
be
kinda,
nice,
but
obviously
we'll
we'll
see
what,
because
everyone
agrees
that
we
should
move
it
to
CSF
forever.
But
it's
like
just
a
matter
of
getting
the
the
person's
dedicated
to
doing
it
and
the
top
level
with
all
that
now.
B
D
D
B
D
D
D
D
D
D
D
Yeah
so
I
mean
I
for
Jenson
rpms.
We
should
definitely
update
today,
if
possible,
like
yeah,
get
the
pr
in
before
the
cut
tomorrow.
So
our
basil
artifacts
for
the
RC
at
least
have
the
right
version.
If
anyone
is
using
that
I
don't
know,
but
aren't
you
able
to
if
you
branch,
the
stuff
aren
t
able
aren't
we
able
to
merge
the
CRI
CTL
PR
increment
is
release
at
this
point
or
merged.