►
From YouTube: 20180515 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
B
So
I
want
to
share
some
feedback
that
we
got
on
an
issue
that
was
pretty
good.
It
was
closed
out
a
while
ago,
I
think
early
in
the
year
and
a
bunch
of
people
had
found
a
bunch
of
problems
with
it.
I
think
this
is
an
issue
worth
reading
through
that
highlights,
have
Canadian
pre-flight
checks
and
the
warnings
and
the
suggestions,
maybe
isn't
the
best
way
to
approach
that
particular
issue
and
I
was
thinking.
B
The
the
main
issue
here
is
that
when
cuvee
diem
doesn't
find
her
eye
control,
it
will
say
you
should
go,
get
Craig
control,
but
that's
it
doesn't
really
that
doesn't
really
solve
the
problem,
because
a
lot
of
people
don't
have
the
cooling
runtime.
Are
they
going
like
toolchain
setup,
so
I
think
I
think
I
thought
that
was
an
interesting
way
that
we
got
feedback
for
that
particular
problem.
B
That
was
not
necessarily
expected
and
also
there's
versioning
issues,
but
I
think
the
bigger
piece
here
is
that
we
were
discussing
adding
cry
control
to
the
cuvette
TM
packages.
I
guess
I
guess
thinking
about
it,
there's
not
a
ton
of
discussion,
just
something
I
thought
that
would
be
good
to
highlight
and
like
bring
to
the
surface.
But
if
anyone
had
thoughts
on
that
or
one
that
yeah.
C
I've
been
looking
into
looking
into
that
since
last
week,
I
noticed
also
that
GC
is
going
to
ship
with
its
Christ
CTL.
The
next
question,
the
taint
the
same
time.
It
goes
ta
goes
Kia
in
111,
so
it
makes
a
lot
of
sense
for
us
to
also
ship
it
in
our
depth
and
rpms
but
forth,
and
we
can
start
utilizing
it
inside
of
Cuba
diem,
for
example,
for
the
free
pre-flight
checks.
And
then
we
don't
I
mean
it's
not
user
friendly
to
say.
Go!
Get
this
because
right,
I
said
like
setting
up
a
goal
on.
C
A
The
basic
gist
is
people
always
got
confused
by
using
doget,
so
the
the
pre-flight
check
basically
said
like
we
don't
have
this
command-line
utility
we're
gonna
start
to
bundle
as
a
very
specific
version
of
that
command
line,
rich
utility
that
we've
vetted
in
some
way
shape
or
form
with
the
Deb's
or
RPM
packages
as
part
of
the.
So
that
way,
the
pre-flight
check
should
always
pass
right.
A
So
this
lends
itself
to
other
things
which
might
be
worth
while
talking
about
like
there
are
enablement
in
the
broader
packaging
in
ecosystem
that
are
going
to
beta
in
this
next
cycle
and
I,
even
I,
even
highlighted
in
a
separate
PR
where
I
don't
think
Covidien
should
be
doing
some
things
like
loading
kernel
modules.
That
scares
the
heck
out
of
me,
but
we
should
absolutely
have
the
capability
of
having
the
packaging
loading
of
kernel
modules
if
a
person
wants
to
support
a
feature
set.
A
So,
for
example,
I
PVS
is
the
concrete
example
that
I'm
referring
to
where
it
would
be
confute.
It
might
be
another
confusing
case
where
it
said:
kernel
module
is
not
loaded.
You
know
to
load
your
kernel
module
for
this,
but
I'm
not
entirely
certain.
If
we
should
for
something
like
that,
we
should
do
in
a
case
by
case
basis
or
if
we
should
start
putting
more
stuff
into
packaging,
that
people
can
then
install
or
default
via
that
way.
I
think.
C
A
A
A
C
We
sell
for
list
images
last
week,
but
now
that
we
have
access
to
CRI
CTL
as
it's
going
G
and
we
package
it
everything,
it's
really
easy
to
just
involve
CRI
CTL,
pull
against
the
CRI
socket,
so
I
think
as
we
have
got
so
many
user
requests
like
this.
This
year's
Cuban
in
live
I
think
it
makes
sense.
Now
that
we
have
for
the
first
time
we
actually
have
one
solution.
We
can
go
forward.
C
C
Doesn't
work
with
doctor,
it
gives
that
the
cube
back
cubelets
starts
up
or
in
the
beginning
of
the
cuban
process,
cube
lattice
crash,
hoping
in
other
words,
not
running
it's
waiting
for
a
file
to
be
existent
and
the
cube
see
the
cubit
is
the
one
that
is
running
the
doctor,
CRI
demon
inside
of
itself.
So
in.
C
A
Stimulus,
that's
kind
of
thing,
though
likely
you're
gonna
confuse
people
in
the
very
beginning,
because
the
whole
purpose
of
this
is
UX
experience
for
people
who
are
doing
on-prem,
installation,
readapt,
installs
right
and
we
punted
on
doing
a
pull
specifically
because
we
didn't
have
the
apparatus
for
doing
all
of
the
CRI
runtimes.
But.
C
A
C
F
D
C
Yeah
I
think
I
think
that
shouldn't
be
an
issue
later,
because
we're
gonna
ever
gonna
use
a
ga
version,
not
only
pre-release
questions.
So,
okay,
we
have
a
lot
of
other
stuff
to
talk
about
so
I
think
we
should
proceed
and
or
yeah
we
can
also.
When
we
have
this,
we
can
also
in
the
beginning
of
the
cubed
M
in
it
flow
we
can
pre
pull.
So
we
can
sail
with
better
error
messages
than
before
and
say
it
seems
like
your
runtime.
Whatever
can't
pull
this
image
because
you
have
no
internet
connection
or
whatever
grant.
A
Perhaps
we
should
update
the
docs
as
part
of
this
work
to
say,
like
as
a
pre
step.
Try
validating
your
people.
You
pre
pulling
your
images
because
for
a
lot
of
folks,
the
initial
in
it
that
there's
I
don't
even
know
how
many
issues
I've
seen
people
somehow
having
issues
with
an
it
and
a
lot
of
them
get
conflated.
A
C
I'm
thinking
it
could
be
actually
in
executed
by
a
normal
pre-flight
check.
So
what
we
would
the
first
thing
that
would
happen
is
like
check
whatever
we
have
in
pre
for
checks
now
and
if
all
those
are
okay,
pre
pulled
images,
because
it's
gonna
happen
at
some
time,
so
better
and
preflight
checks,
interface
and
anything
else,
because
if
you
don't
explicitly
don't
want,
you
can
skip
skip
it
with
the
prefatory.
C
A
C
G
C
G
C
C
So
I
investigated
the
source
code
this
week
and
sent
like
five
pr's
or
something
to
make
you
bet
and
able
of
handling
to
external
rushes
of
the
API.
Most
of
those
things
should
now
be
merged.
It's
there's
only
more
that
I
think
Tim
endless
hasn't
having
the
list
to
review,
and
then
we
have
build
applicated
version.
1
alpha
2
tour
request
where
I've
just
taken
everything
we
have
in
version
1
alpha
1
copied
it
and
that
we
have
over
to
and
started
using
alpha,
2
internally.
C
A
Want
to
make
sure
we
have
test
coverage
yeah
on
this,
so
I
don't
know
whether
or
not
we
should
add
a
test
to
the
comedian
intent
test
suite
or
if
we
will
inherently
get
I
think
which
should
be
in
comedian
intent,
test
suite
honestly,
because
I
don't
think
the
main
and
and
test
suite,
even
though
the
upgrade
tests
will
indirectly
validate
piece
further.
They.
A
C
Sort
of
Coffee
map,
yeah
I'm
gonna
post
a
dark
to
the
signal
in
this
as
well
for
the
more
grand
plan
for
alpha-2
as
we're
this
late
in
the
cycle.
Not
all
the
expected
changes
to
the
API
will
will
be
made
due
to
the
lack
of
necessary
time,
but
where
we're
going
to
improve
it
incrementally
and
hopefully
in
the
next
next
version,
112
will
have
either
version
1
beta
1
that
has
another
schema,
but
still
backwards
compatible
or
version
1.
C
C
A
I'm
I
have
pretty
I,
talked
with
a
bunch
of
folks
and
they're,
pretty
uncomfortable
with
enabling
the
default
update
in
this
cycle
and
getting
Peters
PR
in
and
a
lot
of
people
to
tinker
with.
It
makes
absolute
sense
to
me
having
the
default
configuration
point
to
a
single
config
file.
People
are
very
uncomfortable
with
that
at
this
stage
of
the
game,
because
it's
had
like
very
little
cycles
underneath
it.
C
Yeah,
so
I,
don't
think
you
block
dynamic
configuration
will
even
graduate
to
beta
so
I'm
not
planning
to
do
that,
but
instead
we're
gonna.
The
flow
will
essentially
look
like
you
run
cube
at
a
minute
inside
of
a
cubed
n
API.
We
also
have
embedded
the
cubits
component
configuration
by
default.
The
cube
of
component
configuration
is
like
sets.
Things
like
or
use
authorization
for
endpoints
like
security
features,
don't
enable
the
read
on
report.
Things
like
that
that
are
cluster
specific,
also
like
set
the
right
dns
domain
or
DNS
endpoint,
which
is
basically
everyone
struggles
with.
C
That's,
let's
ever
try
to
configure
a
new
service
service
subnet,
so
such
things
will
be
in
the
cubed
configuration
and
then
this
is
marshalled
and
written
to
disk.
At
the
time
of
cubed
on
it,
it's
gonna
be
written
to
warm
lib,
cubelet,
config,
dot
llamo.
The
qubit
will
read
this
file
when
it
starts
to
execute
yeah.
A
This
is
not
I'm
not
comfortable
with
like
you're
gonna
change.
The
default
initialization
of
the
coop
I
mean
I,
get
that
I
know
what
you're
trying
to
do
and
I
agree
that
we
should
definitely
do
it.
I
just
think
that
we're
so
late
in
the
110
cycle
to
make
this
change
is
a
little
scary,
because
we
haven't
had
a
lot
of
time
to
qualify
and
make
sure
that
everything's
working.
A
So
you
know
it
just
scares
me
to
do
it
this
late
in
the
cycle.
Okay,
if
we
did
it
in
the
beginning
of
1-10,
I'd,
probably
be
feeling
much
more
comfortable,
but
we
were
t-minus
four
days
to
get
that
in
I.
Think
enabling
getting
Peters
PR
in
and
enabling
that
option
for
us
to
test
with
makes
a
ton
of
sense
but
defaulting.
It
scares
me
a
little
bit,
I
think.
C
It's
worth
to
do
it
later,
because
if
we,
if
we
even
try
to
go
ta
I,
so
the
last
quarter
is
always
the
worst.
So
because
that's
a
lot
of
conference
like
that
cube,
constant,
reinvent
and
everything.
So
if
we
can
at
all
go
to
this
CEO,
we
should
probably
do
it
in
112
after
analyzing
everything
we
need
to
do.
I
think
I'm,
pretty
optimistic
about.
C
C
C
A
G
G
G
You
know
I
work
very
strictly
9:00
to
5:00
I
work.
I
have
time
off
coming
up,
I,
just
I,
don't
see
us
getting
anything
more
major
than
very
small
incremental
improvements
done
in
the
next,
especially
like,
given
the
speed
of
PRS
in
cover
entities
like
the
turnaround
on
merging
a
PR
is
like
a
day
best-case
time
right.
A
This
here's,
my
question:
do
is
anyone
else
besides
us
using
the
loaded,
config
file
for
the
coolant
is
to
season
is
GK
using
it.
I
can
ask
if
other
people
are
using
it
and
have
cycles
under
their
belt
to
verify
and
and
you're
gonna
do
most
of
the
work.
Then
I'd
be
less
scared,
but
for
me
it
seems
like
pure
enablement
and
we
own
the
bag
right.
A
A
H
Bigger
concern
is
less
around
the
actual
config
file
usage
itself,
because
the
config
file
I've
used
in
use
myself
without
any
issue.
It
seems
to
work
as
as
its
intended.
The
worry
that
I
have
is
around.
We
have
this
disconnect
between
what's
managed
by
cube
ADM
itself
and
what's
managed
by
the
packaging.
So
we're
going
to
have
this
chicken-and-egg,
where
we
want
to
update
cube
ATM
to
use
a
config
file,
but
then,
in
order
to
pull
in
the
change
that
consumes
a
config
file
by
the
cubelet.
H
C
C
If
we
take
it
from
the
beginning,
so
we
cube
at
a
minute
it
generates.
Marshalls
config
file
right
at
the
disk.
Cublas
starts
up
use
uses
the
config
kubaton
uploads
the
configured
marshalled
config
map-
that's
lovely
in
its
sequence,
then
we're
on
join
time.
If
we
join
in
one
element,
if
we
join
a
110
cubelet,
it's
not
going
to
write
any
config
down
if
it
joins
a
111
queue,
but
it's
gonna
fetch
a
config
map
and
write,
write
it
down
to
disk
right.
C
C
We
now
write
write
down
the
new
config
file,
the
112
config
file
to
a
local
disk.
Then
we
have
in
our
upgrade
instructions
that
you
should
go
and
do
up
to
get
upgrade
afterwards
right
and
that
will
update
the
and
that
is
less
severe
now,
because
before
those
changed,
the
versions
or
the
drop
in
file
change
change.
Before
this
now
the
dropping
file
won't
change
any
more
with
the
config
file.
C
H
Just
worried
that
we're
messing
around
with
something
that
has
a
lot
of
different
kind
of
edge
cases
here,
you
know
what,
if
somebody
has
a
modified
systemd
drop-in
file
that
they're
relying
on
to
work
on
to
modify
the
behavior
right
now,
like
they're,
updating
the
CRI,
you
know
the
use
of
CRI.
Now
we
have
this
kind
of
weird
process
of
how
do
we
migrate?
A
A
There's
no
reason
we
can't
have
it
be
an
opt-in
right
like
we
have
an
optional
in
as
a
you
know,
you
can
opt
into
this
new
behavior,
but
you
know
we
haven't
had
enough
time
like
everything
else
to
to
make
sure
that
we
vetted
all
the
edges.
So
there
could
be
conditions
that
could
occur
and
then
will
default
at
1:12,
I
think
defaulting.
The
change
in
1:10
gives
me
it
makes
me
feel
like
everyone,
I
love
them
or
sorry,
one
alone,
my
bed
yeah
did
defaulting
this
in
111.
Is
it's
just
scary
to
me.
A
G
C
C
If
you
want
to
like
hear
hear
about
the
the
flow
of
this
but
I
think
we've
dealt
with
so
I
mean
it's
easier,
it's
an
easier
flow
than
what
we
have
right
now
anyway,
because
our
current
flow
is
it's
just
like
it's
worse
simply
so
so
we're
dealing
with
with
world
stuff
already
and
a
simplification
would
be
would
be
really
nice
and
I.
Think
it's
better
to
do
that.
Simplification.
Now,
if
that,
if
something
happens,
it's
better
than
if
something
happens
for
GA
is
the.
C
G
C
A
So
I
think
there
just
a
unease
I'm
getting
this
change
in
place
and
fully
abetted
I.
Think
if
you
spec
out
the
full
issue
and
all
the
edges
where
we
can
rally
on
it,
why
don't
we
do
that
first
to
alleviate
concerns
because
there's
too
many
open-ended
questions,
so
why
don't?
We
have
a
single
issue
we
can
rally
on
and
the
comedian
repo
that
identifies
exactly
the
flow
and
folks
can
poke
holes
on
it.
We
can
reconvene
by
tomorrow
to
talk
about
in
more
detail.
A
H
C
A
There
were
a
number
of
at
some
point.
We
need
to
update
the
release
notes
for
the
cycle,
but
there
are
a
number
of
fixes
that
both
Lee
and
Jason
worked
on
with
regards
to
upgrade
fixes
on.
They
were
very
thorny
and
in
the
110
and
111
cycles,
so
those
are
all
in
now,
I'm,
not
aware
of
any
other
ones
that
are
outstanding.
Besides
your
your
last
at
CD,
one
Jason
I,
don't
know
if
that
got
merged
or
not.
I
C
A
A
C
C
B
C
C
F
C
A
Okay,
I,
don't
have
anything
else,
I
think
for
folks
looking
to
engage
I
think
there
I've
shuffled
some
PRS
to
help
wanted
if
they
want
to
work
on
those
things.
There's
also
feel
free
to
work
on
the
documentation
stuff.
If
you're,
not
if
the,
if
the
issue
isn't
marked,
is
active,
that
means
anybody
can
can
have
at
it.
I
did
see,
live.
A
Amir
I
did
see
your
one
PR
with
regards
to
the
development
dock
stuff
and
usually
all
the
development
docks
I
think
the
other
person
said
that
they
usually
always
go
into
the
cuvette
DM
repo,
so
I
think
that's
the
only
suggestion
dev
docks
go
there.
Community
facing
docks
usually
go
in
the
other
in
the
site
website.
Repo.
C
J
J
C
A
Ivan
just
mentioned
that
a
second
ago
I'm
gonna,
we're
gonna
timeout
and
wait
until
next
week,
Monday
or
so
just
to
make
sure
anyone
who
there's
any
dissenters.
They
can
comment
on
that
thread.
But
then,
once
a
by
next
Monday
I'll
just
pull
the
trigger,
maybe
on
Sunday
I'll,
pull
the
trigger
and
then
just
transfer
the
repos
over.
C
Yeah
and
so
like,
if
you,
if
you
want
to
contribute
to
one
11
release,
you
can
also
just
click
the
one
11
milestone
and
go
through
the
issues
there.
If
it
seemed
if
something
seems
odd,
that
it
shouldn't
be
in
111
milestone,
please
shout
and
then
we
can
take
a
look
at
it,
I'm
trying
to
burn
down
the
list
into
something
more
manageable
from
like
it's
always
like
in
the
beginning
of
this
of
the
cycle.
C
We
are
nearly
everything
to
the
next
milestone
and
then
at
some
point
we
have
to
cut
and
say
this
is
not
for
this
one
and
bumped
in
the
next
one.
So
that
is.
That
is
one
thing.
If
you
see
something
odd,
please
they
say,
and
if
you
can
work
on
something
in
that
milestone,
that
isn't
active.
Please
do.
A
Just
for
a
point,
there's
28,
open
issues
and
I
believe,
like
nine
of
them,
are
Doc's
so
that
that
means
there's
17
or
19
open
issues,
most
of
them
and
a
bunch
of
them
are
active.
So
that
means
there's
already
PRS
for
them.
So
they'll
be
closed
out.
You
know
in
a
day
or
two.
So
if
you
subtract
those
out
of
the
next
year,
you're
there's
a
lot
less.
Actually
things
in
flight
currently
yeah.
A
If
what
may
make
sense,
maybe
a
couple
weeks
from
now
go
through
like
a
planning
session,
because
there's
a
lot
of
big
ticket
items
we
have
and
breaking
down
those
big
ticket
items
into
actionable
things.
We
actually
have
a
game
plan
for
a
bunch
of
these
big
items
like
couplet
dynamic
configuration,
the
the
self-hosting
Sentinel
process
stuff,
the
in
dealing
with
upgrades
in
a
cleaner
fashion,
the
H
a
configuration
stuff
in
dealing
with
that
in
a
cleaner
fashion,
using
phases
and
shifting
phases
around
those
are
huge
mammoth
tasks.
A
C
Yeah
also
after
code
freeze,
I'll
update
my
Cuban
interior
dock
because
I
wrote
it
in
December.
Now
it's
obvious
it's
time,
but
it's,
for
example,
I,
don't
think
or
we'll
see,
but
but
things
like
the
serving
certs
cubelet
stuff.
We
don't
need
that
for
tree
I
also
talking
to
people.
Actually,
it's
less
straightforward,
then
I
like
or
we
base.
It
basically
requires
a
list
of
machines.
So
in
order
to
have
secure
API
server
to
cubelet
communication,
you
need
to
in
beforehand
know
what
the
qubits
are
going
to
be
and
somehow
verify
them.
C
So
either
you
do
that
yourself
using
like
an
approved
flow
or
you
just
go
in
secure
like
we
do
I'm
I've
thought
about
this.
The
latest
week
and
I'm
gonna
post
an
issue
on
something
called
security
profiles.
We
could
add
in
future
future
Cuban,
inversions
I,
don't
even
think
that's
a
thing
for
we
have
to
do
for
GA,
but
it
totally
makes
sense
to
let
the
user
specify
how
secure
they
want
to
be
and
based
on
that
get
or
lose
weight.
Tonality
I
do.
A
Think
you
know,
as
we
brought
up
earlier
in
this
topic
in
in
in
the
beginning
of
1:10,
we
ate
it's
super
hard
in
the
configuration
changes
so
like
you
weren't
around
originally
Lucas,
but
there
was
a
number
of
very
painful
points
that
came
up
as
the
110
upgrade
process
in
the
configuration
file.
I
think
getting
that
to
beta
I
think
is
a
requirement
for
GA
no,
and
we
could
talk
about
that
in
a
later
time.