►
From YouTube: Kubernetes SIG Cluster Lifecycle 20171114
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).
B
Basically,
I
wanted
to
just
go
over
and
bring
up
some
of
the
technical
differences
on
Cuba,
ADM
and
boot
cube,
since
they
are
sort
of
similar
projects
and
they
can
both
benefit
off
of
each
other
and
I.
Don't
think,
there's
a
lot
of
conversation
yet
on
how
they
could
perhaps
eventually
get
integrated
or
have
code
shared
between
the
two
sorry,
my
phone
screen
and
so
boo
cube
is
a
self
hosting
tool
that
is
in
the
incubator
and.
B
It's
not
intended
to
be
a
post
management
tool,
and
that's
where
qbm
has
its
main
differences.
I
think
is
that
there's
an
intent
there
to
have
it
be
a
management
tool
afterwards,
where
blue
cube
was
just
a
deployment
tool
to
get
kubernetes
up
and
running.
I
was
wondering
what
your
eyes
is
thoughts
about
that
were.
It
was
just
a
conversation
starter
on
how
we
can
improve
both
projects
and
perhaps
get
a
little
bit
of
parity
between
the
two.
C
So
I
had
a
quick
look
at
the
other
Google
Doc
I.
Think
it's
a
great
idea
trying
to
unify
all
sort
of
converged.
The
two
projects
and
yeah
I
think
the
biggest
differentiator
right
now
between
Cuba
diem
and
bewp,
I,
wouldn't
say
it's
too
much
of
a
conceptual
woman.
Actually,
you
said
that
Cuba
DM
is
more
of
a
management
tool,
but
I
would
say.
B
The
use
case
that
we
have
is
that
we
look,
we
take
boot,
cuba
and
we
generate
manifests
for
the
tectonic
product
at
core
OS
and
and
so
it's
a
bootstrapping
tool
where
we
can
customize
the
kubernetes
deployment
sort
of
ad
hoc
or
more
flexibly
than
cubm.
Does
it
currently,
and
so
that's
sort
of
the
use
case
where
were
coming
from.
B
D
Don't
think
there's
any
reason
we
can't
make
coup
Beatty
and
be
more
extensible
and
already
is
in
some
respects
to
do
its,
what
its
configuration
for
deployment
and
and
sometimes
I,
even
think
this.
This
is
rolls
into
anything
beyond
the
core,
for
lack
of
better
words
could
eventually
be
just
an
add-on
manager
right
of
some
kind
and
I
think
that
punting
that
to
that
idea,
actually
makes
everything
to
leave
it
a
little
bit
cleaner
because
then
you're
just
deploying
an
add-on
manager
which
could
be
yours
or
ours
or
anybody's
sprint.
B
D
Think
it
could
be
just
a
configuration
of
what
add-on
manager
gets
deployed
like
it's
just
specified
as
part
of
the
deployment
configuration
I
saw
that
I
know
where
Lucas
went.
There
is
he's
magic
he
switched
around
to
my
screen,
so
it's
kind
of
like
tile
one
popped
up
disappeared,
talked
up
at
a
different
time.
I
know
he
had
his
hand
raised
before.
D
A
Yeah,
so,
first
off
thanks
for
creating
the
duck,
I've
been
aiming
to
bring
this
up
for
too
long,
but
haven't
got
around
it.
So
I
really
appreciated
that
you
for
that
in
it
initiative,
yeah,
so
I
think
Cuba,
or
at
least
in
my
understanding.
We
all
want
Q
by
them
and
boot
cubes
to
converge
eventually
and
I,
mean
boot,
cube
kind
of
predated,
cubed
M
in
the
self
hosting
stuff
and
all
right
and
could
be
seen
as
like
an
experimental
playground
way.
You
like
made
a
lot
of
self
hosting
experiments.
A
A
E
B
A
A
Probably
not
a
staging
area,
but
adding
the
option
to
configure
the
pots
as
you
want
them
might
be
reasonable
right
for
the
use
case,
then
you
could
just
point
it
to
your
staging
area
yeah.
So
that
is
reasonable
to
me
and
it's
well
definitely
I
think
we
could
improve
and
the
phase
command
is
under
alpha.
Just
for
this.
This
reason
right
and
yeah
what
else?
What
I'm
gonna
say?
The.
B
D
Would
be
cool,
I
talked
with
you
Jim
no
more
than
14
minutes
ago,
and
we
are
finalizing
getting
the
initial
pod
ship
pointing
in
place
and
talked
about
the
path
for
testing
for
words.
So
we
should
have
that.
My
goal
is
to
get
it
merged
by
the
end
of
the
week,
but
I'm,
always
at
the
whims
of
the
node
team,
whether
or
not
I
like
it
or
not
so
I
have
to
play
by
their
rules.
D
D
F
Real
quick
with
respect
to
the
secrets,
checkpoint,
II
I
mean
I'm
sure
we
could
have
asked
something
along
the
lines
of
using
a
host
volume
amount
into
that's
here,
some
other
path
and
then
having
some
sort
of
sidecar.
That,
essentially,
does
that
that's
sinking
explicitly,
it's
not
ideal.
It'll
only
work
great
when
you
have
a
set
of
trusted
clusters,
because
those
things
will
be
sort
of
left
around.
You
know
for
the
plaintext
on
disk,
but
it's
it
definitely
is
is
not
definite.
I
would
say.
D
But
the
secrets
checkpointing
is
a
little
weird
and
very
contentious
right.
So
I
don't
know
if
we're
gonna
get
resolution
on
that,
because
it
it
kind
of
dries.
It's
like
one
of
those
weird
chicken
egg
problems
as
long
as
we
can
bootstrap
the
control
plane
and
then
rely
on
secrets
from
that
point,
but
it
seems
to
me,
like
you
guys,
are
using
checkpoint
secrets
for
other
conditions.
D
Besides,
just
the
control
plane
restart
right,
because
as
long
as
the
EPI
server
is
brought
back
up,
it
can
access
the
secrets
from
from
NCD
right
and
so
long
as
that
condition
exists.
Then
we're
then,
then
the
cycle
regenerates
everything,
but
if
you
have
conditions
where-
and
this
is
not
the
case
for
cube
ADM,
if
you
have
conditions
where
you
want
to
restart
something
that
relies
on
secrets
and
has
about,
has
amount
of
that
then
you're
in
this
weird
case,
right,
I.
C
Are
we
gonna
run
into
that
when,
like
the
self-hosted,
api
server
relies
on
secrets
for
TLS
credentials
so
because,
like
yeah,
we're
gonna
run
into
this,
then
right
yeah,
because
the
self
hosted
a
guy's
safe
and
won't
be
able
to
load
the
TLS
sites
for
at
CD
and
that,
yes,
it
was
gonna
need
some
kind
of
check,
pointed
secret
I.
Imagine
going
forward.
C
Yeah
and
then
another
another
difference
between
cube,
LEM
and
boo
cube
that
I'm
thinking
about
now
is
that,
from
my
understanding,
boot
cue
has
always
been
very
opinionated
about
a
lot
of
the
choices
it
makes.
So
they
just
you
know
you
deploy
flannel
as
the
cni
provider,
whereas
cube
idiom,
there's
a
bit
more
sort
of
vendor
agnostic,
so
yeah
that
one
that
might
be
quite
interesting
to
look
at
sort
of
the
differences
of
deployments
there,
because
you
know
we
don't
assume
that
we
have
a
function
in
DNS
server
and
that
affects
self-hosted
at
CD.
C
A
Regarding
the
letting
that
cuban,
with
more
a
management
tool
than
just
been
up
once
and
like
just
for
bootstrapping,
I
kind
of
agree
that
well
it
we
have
the
upgrade
functionality
now
right,
but
it's
more
for
user
convenience
and
and
like
more
for
this
flow
that
you
can
do.
I
have
any
bare
metal
host,
just
install
the
package
and
you
cube
at
a
minute
and
when
you
then
have
to
upgrade
eventually
or
sooner
rather
than
later.
A
You
just
issue
this
command
and
it
will
be
listing
all
so
we're
kind
of
staging
the
bootstrap
token,
the
bootstraps
token
like
UX
there
it
might
get
in
as
a
cube,
CTL
command
or
a
cube,
CTL
extension
command
or
whatever
in
the
future,
which
is
or
whatever
we
do.
When
we
reconsider
service
accounts
and
secrets,
lead
to
and
think
more
about
that.
A
A
The
the
main,
the
main
focus
of
cubed
M,
is
to
bootstrap
a
cluster
and
do
that's
really
well
with
the
minimum
effort
for
the
user
needed,
but
still
have
the
scope
set
where
a
small
well
have
this
scope
really
focused
so
cubed
M
only
does
two
operations,
it
talks
to
the
common
API
once
available
and
it
writes
files
to
disk.
These
are
the
two
only
ones
compared
to
like
cops
or
whatever.
That
sees
the
whole
cluster
all
nodes
in
the
cluster
knows.
A
A
Soon
as
possible,
absolutely
and
we're
kind
of
working
on
that
as
quickly
as
we
can,
as
Tim
said,
like
one
of
the
dependencies
are
well
checkpointing
in
some
form.
Robert
I
think
pointed
out
here
that
we
might
so
now
that
we
have
the
project
pointing
you
might
be
able
like
in
core
right.
You
might
be
able
to
remove
that
logic
and
only
do
secrets
for
the
time
being
until
we
might
eventually
do
checkpoint
secrets
or
whatever,
but
last
time
we
some
months
ago.
A
D
Who's
very
contentious,
some
some
issues
are
pretty
contentious
and
that
that
is
one
of
them
in
community
contentious
issues
unless
they
have
a
well.
You
know,
defined
set
of
here
are
the
known
issues,
and
here
are
how
we
are
addressing
them.
We'll
always
go
around
in
this
weird
circle,
and
some
of
those
issues
have
literally
existed
in
this
project
since
epoch
and
have
not
had
a
resolution.
So.
A
Yeah,
so
so
this
is
one
we
know
we're
gonna
run
into
sooner
or
later,
as
Jamie
said
as
well.
Cuz
well,
for
now
we're
we're
thinking
of
using
secrets
and,
at
the
same
time,
working
with
cigars
to
improve
the
authorization
mechanisms
in
companies
instead
of
like
punting
on
secrets
and
saying
it's
too
insecure
form
for
cluster
credentials,
but
yeah.
B
G
Question
which
is
like
we
don't
want
a
monoculture
right,
and
what
is
the?
What
do
you
want
to
happen
with
like?
Why
do
we
not
want
to
have
boot
cube
in
cube,
ATM
and
like
it
feels
like
we
decided
at
the
very
beginning
of
this
whole
exercise
that
we
were
not
going
to
have
a
monoculture,
and
we
were
going
to
encourage
to
us
like
weak
you
to
exist.
I.
B
Guess
the
goal
is
to
share
what
we've
learned
share,
what
Covidien
has
learned
and
see
if
there,
if
we
can
merge
either
make
boot
cube
into
a
library
potentially
make
it
something
that
cubm
brings
in
as
a
library
dependency
or
just
pull
in
code
into
idiom?
There's
not
a
firm
goal
for
blue
cube
in
general
to
change
it
as
it
is
today.
It's
just
bringing
up
the
discussion
on.
Where
do
we
want
each
of
the
project
to
be
going
and
how
they
could
potentially
benefit
from
each
other.
G
Is
to
like
show
that
blue
cube
is
better
right
like
implement
the
machines,
API
and
then
boot
cube
will
win
right
if
blue
cube
is
better
right
like,
and
it
feel
like
we're
all
very
good
about
like
borrowing
ideas
from
each
other
and
I
feel
like
it's.
It's
a
marketplace
of
ideas
and
more
than
the
code
is
like
the
idea
of
like
self-hosted,
which
you
know
came
from
Ben
Cuban.
Is
you
know
your
collective
idea?
I
guess
you
know
it
definitely
came
from
you
guys,
I,
don't
know
the
exactly.
G
From
and
that's
you
I
would
love,
I
would
love
to
see
a
machines
acre
implementation,
that's
based
on
what
I'm
not
sure
they're
to
plot
I
use,
but
like
whether
a
machines,
API
implementation
based
on
the
boot
cube
stack
or
is
that,
like
it
sounds
like
buki,
was
just
for
the
first?
The
first
node
is
that
right
and
then
so
what
that
apply
to
the
machines?
If
you
know
what
I
mean
I.
B
G
B
H
D
Moment
in
time,
but
there's
also
Redux
in
the
open
source,
which
is
with
enough
eyeballs,
all
bugs
are
shallow,
and
if
we
put
things
in
a
common
place
and
we
all
exercise
it,
it's
a
benefit
to
core
OS.
This
is
better
the
community
to
have
common
set
of
reuse
code,
because
in
that
case
you
are
at
the
point
of
having
many
people
leverage
it
versus
a
select
few
well.
G
I
mean
I've
been
being
pushed
into
arguing
a
case,
I
don't
necessary
with,
but
I
feel
like
we,
we
decided
to
have
many
installers,
right
and
and
I
personally
think
you
know
my
installers.
The
best
mission
went
right,
but
you
know
like
and
we
should
have
one,
and
so
it's
my
opinion
and
but
we
decided
not
yet
and
I
feel
like
the
idea
of
saying
everyone
should
adopt.
One
bit
of
code
is
not
what
we
decided
well.
D
G
D
A
Cube
Adam
is
special
Justin
in
in
that
regard,
that
it
is
not
a
full
installer
right.
It
is
a
tiny
bit
of
code
that
just
initializes
your
cluster
enough
to
do
the
MVP
thing
unless
pretty
much
every
everything
else
in
the
community,
except
for
boots
cube.
So
here
we
have
boot,
cube
and
cube,
am
alongside
each
other,
having
pretty
much
the
same
scope
and
trying
to
achieve
the
same
tasks
regarding
the
machines.
Api,
that's
tectonic,
that's
that's.
Maintaining
a
full
stack
and
tectonic
might
be
able
to
implement
the
machines.
A
Api
should
do,
but
Cuba
diem
and
boot
cube
are
very
limited
in
scope
and
small
enough.
Then,
where
we
want
convergence
to
have
less
codes,
maintain
for
both
parts,
and
we
then
also,
if
you're,
this
installer.
That
does
the
the
grant
full
thing
in
in
your
way,
you
will
have
a
lot
of
and
like
they're,
less
literally
70
of
these
full
installers
enter
and
that
provision
machines
do
do
that
thing
in
their
way
install
add-ons
whatever
I
am
the.
I
G
Adm
incubating
will
do
this
stuff
magically
for
you,
because
what
we've
learned
is
people
don't
want
to
use
anything
like
we
don't
use
anything
we
prescribe,
and
so
we
can't
say
if
it's
okay
to
say,
use
cube
ABM,
but
we
have
so
also
documents
it
and
we
have
to
say
if
someone
else
wants
to
maintain
something
different
from
Cuba
diem.
We
should
encourage
that.
We
should
welcome
that.
We
shouldn't
say:
oh
but
you're
doing
it
wrong.
Oh,
but
you're
gonna
submit
to
Moscow,
who
should
say
that's
great.
D
A
That
is
like
we
do
use
cubed
M
as
an
experimentation
ground
right
for
identifying
what
is
the
commenting
needed
for
N
bananas
cluster?
Then
we
have
a
lot
of
document,
documentation,
efforts
and
also
design
Docs
for
qubit
him.
Here's
exactly
exactly
all
the
things
that
cube
am
is
doing.
If
you
don't
want
to
use
cube
item,
you
still
have
this
documentation
that
you
might
want
to
do
this.
You
might
want
to
do
this
and
this
for
getting
API
aggregation,
for
example,
work.
If.
A
G
F
F
Think
part
of
the
problem
here
is
that
they,
because
there's
part
of
the
power
of
kubernetes,
is
that
there
are
so
many
ways
to
set
it
up
and-
and
there
is
no
single
best
way
to
do
it,
for
all
all
scenarios
like
something
that
people
somebody's
going
to
use
in
production
compared
to
turn
on
video
some
compared
to
a
stack
of
raspberry
pies
is
going
to
be.
You
know,
that's
a
tall
order
to
have
one
tool.
That's
actually
gonna
handle
all
of
those
things.
F
I
agree
that,
like
you
know,
cube
admin,
is
we
have
documentation
that
you
know
talks
about
here's
how
things
stitch
together,
regardless
of
the
tool
I,
think
it's
just
a
matter
of
time
and
funding
in
terms
of
what's
going
to
have
this
biggest
impact
on
the
most
part
instead
of
users
and
so
I.
Think
most
people,
at
least
in
this
group
right
now,
are
like
for
most
end
users.
The
biggest
impact
thing
we
can
do
is
actually
get.
You
know
something
that
is
disagree
with
also
having
documentation
that
covers
how
to
do
this
stuff.
F
F
G
I
mean
I,
think
I
think
my
feedback
on
that
is
it's
not
it's
not
my
battle
right,
like
I
my
view
is,
we
should
have
one
installer
tool
and
we
should
make
that
work
and
I
have
been
I've
lost.
That
battle
and
I
feel
like
the
people
that
want
many
installers
should
be
the
ones
documenting
it,
and
it
feels
like
you.
Baby
em
is
part
of
that
effort.
G
G
A
lot
of
these
decisions
written
down,
maybe
in
code
comments
and
things
like
that,
like
the
we
need
to
extract
that
in
a
way
the
people
that
are
not
gonna
use,
cubic
or
not
gonna
use
cube
ADM
that
don't
want
to
use
a
pre-baked
tool
that
they
still
feel
that
kubernetes
is
installable
and
that
they
don't
feel.
Oh,
you
know
that
I
have
to
use.
I
have
to
use
this
complicated
script
or
this
complicated
script
written
and
go
to
install
it
like
that
is.
F
F
G
F
Things
are
wired
together
is
still
pretty
opinionated
so
that
that
is
still
a
high
opinion,
type
of
way
of
doing
things
and
putting
stuff
together,
I
I'm,
just
I'm,
not
sure
you
didn't
know
like
yes,
I
would
love
to
see
he's
such
a
diamond
exist
I'm,
not
sure
how
we
make
that
happen.
When
everybody
is,
you
know
focused
on
honestly,
on
other
things
and
and
from
from
the
point
of
view
of
of
you
know,
companies
like
like
patio.
F
Our
focus
is
is
that
we
want
to
actually
have
something
that
is
some
Morton
supported
upstream
and
is
going
to
actually
work
across
the
large
large
large
set
of
environments
that
were
supporting
people
with
and
and
honestly
that
doesn't
relate
if
we
had
a
document
that
you're
describing.
That
would
be
awesome.
I
love
that
too,
but
but
in
terms
of
priorities,
that's
not
where
our
priorities
are
right.
Now,
yeah.
G
F
G
Feel,
like
I
feel
like
it
should
be
a
prerequisite
for
trying
to
make
you
a
liam
special
is
to
not
poison
the
well,
which
is,
in
other
words
like.
Why
did
cuba
keep
up
with
special
right?
Why
did
why
did
cuba
fail
and
why
to
keep
up
like
do
its
thing?
It's
because
it
didn't
have
documentation
outside
of
Cuba
and
then
the
the
sort
of
failure
pattern
was
people
would
would
evolve
the
the
requirements
for
installing
communities.
G
D
To
interrupt
everybody
here,
I
want
to
table
this
discussion
because
at
some
point
we
need
to.
We
have
other
agenda
items
that
we
need
to
reach
today,
and
this
is
kind
of
a
philosophical
argument
that
folks
can
debate
in
another
venue
which
I'm
totally
clueless.
If
there
are
action
items
that
folks
would
want
to
address,
I'd
be
happy
to
entertain
those
action
items.
So
if
people
want
to
sign
up
for
it-
and
we
can
add
them
to
the
agenda
list,
but
we
have
other
things
that
we
need
to
address
and
talk
about
to
there.
F
One
last
word:
I
want
to
say
there
just
an
eighth
poisoning,
the
wells
pretty
strong
language,
so
you
know
just
you
know,
that's
a
different
discussion
and
I
personally
went
through
and
I
put
it
in
the
bootstrap
tokens.
I
documented
that
in
a
way
that
was
not
super
cube,
a
cube
admin
specific
so
like
I,
think
on
a
feature
by
feature
basis.
We've
been
trying
to
make
sure
that
there's
actually
good
document.
What
we
don't
have
is
something
that
stitches
it
together
as
a
guide.
That's
that's
for
end
users.
F
Those
features
are
documented
outside
I,
don't.
G
Mean
poisoning
the
way
I
just
mean,
like
basically
seeding
the
seeds
of
its
own,
like
avoiding
the
same
mistakes
that
we
make
prior
previously
with
Cuba
right
and
and
that
it
goes
beyond
just
the
cube
up
project
Cuba,
but
to
the
whole
commands
project
and
I
like
very
much
what
you
did
with
bootstrapped
opens
where
you
know
where
it
doesn't
feel
like
a
Cuban
name,
specific
feature
and
we're
gonna
use
that
in
cops
and
other
things,
I
think
so.
I
think
that
I
think
that's
great.
Alright,.
A
Anyway,
so
we
did
the
CIA
pinning
last
last
release
as
Joe
and
Matt
discussed,
and
we
got
us
here
in
in
this
group,
and
we
said
back
then
that
we
would
convert
this
like.
We
would
require
the
CIA
to
be
be
past.
The
cubed
M
join
in
order
to
have
a
like
fully
secure
flow
where
this
workers
impersonate
masters
thing
can't
happen.
So
the
question
is
now:
should
we
make
cubed
M
join
just
talk
and
without
the
CIA
pinning.
A
F
You
know
this
is
a
better
plan
for
a
while,
well
I
think
it
definitely
improves
the
bootstrapping
security
stance.
I
think
we
have
a
very
clear
error
in
that
PR,
where,
if
hey,
if
you
hit
this
error,
just
add
this
flag
and
you're
good
to
go
I,
think
it
just
it's
it's
it
makes
people
stop
and
stop
and
think
about
it.
I
am
ok
with
this
going
in,
we
were
not
in
GA
yet
so
so
changing
the
changing.
D
Will
abstained
from
security
conversations
because
every
every
time
I
tread
into
it
I
start
to
uncover
more
things
because
I'm,
one
of
those
people
that
likes
to
dig
and
like
what
this
mean
and
then
you're
like.
Oh,
my
gosh,
that's
not
what
I
thought
it
meant
so
I
will
I
will
punt
and
leave
that
for
from
where
and
Joe
to
talk
about
the
other
aspects.
Other
things
I
kept
early
champion,
I'm.
F
A
H
A
F
A
You
you
get
the
cube,
Adam
join
command
on.
You
bet
him
in
it,
but
if
you
Claire
your
console
or
something
you
can
definitely
fetch
the
token
again
and
things,
but
it's
not
as
straightforward
like
it's,
not
as
user
friendly
as
it
could
be
its
cubed
and
token
lists
so
yeah
and
yeah.
We
also
talked
a
bit
about
component
config.
A
We
touched
on
it
as
well,
at
least
earlier
in
the
meeting
a
quick
status
update
on
that
is
that
cube
the
cube?
Let
API
is
gonna,
soon
be
graduated
to
be
des,
hopefully
in
time
for
1/9,
it's
gonna
be
pretty
tight,
but
my
totem
was
was
optimistic
when
I
talked
him
earlier
today
we
already
graduated,
or
so
the
first
thing
with
component
config,
is
to
move
out.
A
Currently
all
component
configuration
is
in
the
same
API
group,
so
the
first
thing
is
to
move
it
out
component
dot,
config
dot,
IO,
and
then
these
these
components,
config
types
will
be
in
in
different
directories
in
the
core
and
they
will
serve
as
the
the
API
to
talk
to
the
component
right
cube
proxy
was
moved
out
to
its
own
API
group.
Cubelets
is
next
up
cube,
scheduled
is
soon
they're.
A
Probably
most
of
them
are
gonna,
be
all
fine
1/9,
but
hopefully
they
will
graduate
to
beta
in
110,
so
people
actually
can
start
using
them
and
consume
configuration
from
a
file.
That's
versioned
and
has
all
the
API
machinery
things
all
those
features
in
it
automatically.
Instead
of
having
this
very
painful
way
of
upgrading
and
things
and
changing
behavior
with
flags.
D
That's
all
going
to
be
well
and
good,
except
for
the
one
pen
engine
will
be
the
API
server.
Everything
else,
I
think
I,
don't
see
us
getting
around
that
and
I
don't
even
want
to
tackle
it,
but
the
the
controller
manager
and
scheduler
absolutely
that
will
get
a
a
nice
upgrade
kick
and
just
put
it
into
a
file.
D
G
D
Part
of
it,
another
part
is
like
if
you
have
everything
in
a
configuration
file,
that's
stored
inside
identity
and
your
self
hosting
the
API
server.
How
do
you
get
you'd
have
to
check
point
that
config
map
on
to
the
file
to
is
on
to
the
file
system,
to
for
restoration,
so
that
that's
a
separate
thing
that
would
be
required
more
if
you
enable
that
feature,
config
maps
aren't
aren't
as
contentious
as
other
things,
but
it
still
gets
there.
Yeah.
A
And
to
clarify
to
us
in
a
meeting
you
can
still
like
you
can
deliver
a
file.
However,
you
want
it's
just
a
normal
file,
but
the
most
common
method
that
also
cubed
and
will
use
for
most
components
is
delivering
these
files.
We
are
config
mapped
as
Tim
said
so,
and
that
also
makes
it
easier
to
control
like
authorization
and
then
different,
like
changing
the
the
configuration
in
the
cluster
will
get
far
more
easier,
for
example,
for
a
queue
proxy
with
just
updating
the
config
map
and
very
real
rolling.
The
the
daemon
sets
maintenance.
G
Pretty
being
dumb
but
API
server
can
access
at
CD
and
read
it
directly
or
is
this
about
like
if
you
change,
like
your
HDD
configuration,
for
example,
the.
D
A
C
D
C
D
So
though
it
depends
upon
the
component
is
not
all
components
have
all
the
jiggery-pokery
to
do
dynamic
configuration
the
only
component
that
I'm
aware
of
that
actually
has.
This
is
where
it
will
auto,
detect
and
reload
and
I.
Don't
know
if
it
works
in
all
cases,
but
I
do
know
that
the
other
components
you
have
to
shoot
them
in
order
from
the
restart,
and
that
was
done
by
on
purpose
because
of
how
state
is
stored,
and
if
you
were
to
magically
change
the
configuration
parameter.
All
that
state
could
potentially
be
invalid.
D
A
D
A
Definitely
some
some
work
to
be
done
there,
but
yeah,
it's
cool
to
see
progress
on
the
cubelets
and
key
proxy
as
the
first
ones
and
then
the
scheduler,
then
the
controller
and
your
manager
and
lastly,
the
API
sir-
and
this
is
again
one
of
the
areas.
So
people
ask:
why
isn't
Cube
at
mga
and
it's
because
we
still
have
things
to
occupy.
Yes,
we
still
have
features
to
graduate.
We
still
are
waiting
for
component
config
to
be
a
thing,
because
when
cubed
and
goes
ta
component
config
will
also
go.
Ga
will
be
GA.
D
I'm,
a
huge
fan
of
like
G
aid
features
are
boring
as
all
get-out
or
g8
components,
because
that
means
you've
actually
done
the
due
diligence
along
the
way.
So
I
think
by
the
time
we
actually
get
to
GA
I'm,
totally
cool
with
letting
it
be
a
LAN
pool,
and
you
know
by
the
time
we
get
there,
it
should
be
non-controversial
exactly.
A
Is
it's
not
something
we
want
to
rush
like,
for
example,
right
now
in
the
config
config
file,
we
have
string
string,
Mac
maps
for
the
command
line
arguments
it
works
for
like
99%
of
the
users,
but
it's
unversed
and
and
no
schema,
nothing,
no
structure,
no
validation.
So
we
don't
want
to
graduate
that
and
hence
we're
pushing
on
components,
come
figure
instead
and
working
with
the
community
to
do
the
right
thing
and
then
someday
we'll
get
to
GA,
and
it
will
be
not
controversial
at
all
so
yeah
Robert.
E
Try
to
so
I'm
just
a
quick
PSA
that
we
have
a
cute
concession
now
it's
on
Wednesday
afternoon
for
a
sig
update,
please
let
Lucas
where
I
know.
If
you
want
to
help
chair
that
I
put
some
notes
in
about
failing
tests
in
test
grade.
If
you
have
a
chance,
please
look
at
test
grade
and
help
debug
and
fix
those
tests.
I
think
those
were
that
doesn't
mean
two
things.
We
also
skipped
Justin's
mission
statement
dark,
although
we
only
have
five
minutes,
so
we
have
to
do
the
next
week.
Well,.
G
A
You
and
yeah
a
note
on
the
failing
tests
so
pretty
much
well,
so
the
fails
of
the
cubm
upgrade
tests
are
expected
because,
because
of
the
way
upgrade
tests
are
written,
this
is
another
area
we're
working
with
sig
testing
on.
Currently
there
there's
a
lot
of
GC
assumptions
in
doctoral
tests
and
we're
working
getting
those
well
not
to
be
not
to
be
there
and
expressed
in
a
generic
way,
but
these
are
working
then,
but
the
one
that
is
actually
failing
for
real
is
the
Calico
test.
A
C
We
have
like
three
minutes
left,
so
yeah
I
just
wanted
to
sort
of
want
to
know
like
the
general
status
of
like
p1
items
for
the
for
the
one
nine
priorities,
because
ICO
freezes
next
week
and
I'm,
not
sure
you
know
whether
we
miss
anything,
whether
it
be
sort
of
have
folks
jump
on
stuff.
If
we
really
want
to
end
before
code,
freeze
so
yeah,
but
I'm
not
sure
like
whether
this
meeting
is
like
the
or
next
week
is
the
right.
C
You
know
time
to
have
that
discussion
right
where
we
are
in
terms
of
one-line
deliverables.
Maybe
we
can
know
that
to
tomorrow's
discussion.
Yeah,
okay,.