►
From YouTube: SIG Cluster Lifecycle 2020-02-11
A
A
If
you
want
to
upgrade
to
the
next
client
go
version
like
when
it's
released
you're
going
to
see
some
breakages,
so
basically
you
have
to
apply
changes
to
your
source
code
and
if
you
are
like
a
maintainer
over
the
Boyer
and
cube
ADM
cops
coaster
Bay.
So
basically,
this
thread
here
outlines
some
of
the
changes.
Also
Mike
Dennis
wrote
to
that
can
help.
You
apply
the
changes
to
your
code
base
and
finally,
we
have
also
not
a
PR
here.
That
is
also
adding
some
options
to
some
of
the
methods.
A
B
A
The
plus
and
minus
here
mean
that
basically,
this
there
incompatibilities
between
a
certain
version
of
plan
goal
and
a
question
of
kubernetes
that
lava-like,
this
change
in
particular,
is
different
because
it
is
going
to
break
your
companies
if
you
want
to
move
to
a
new
version
of
Clan
go
well
previously,
the
interface
remained
the
same,
but
there
are
some
other
breaking
changes
which
some
users
might
not
see.
In
this
case
we
have
this
interface
breakage.
A
A
Please
add
your
project
updates
I'm,
going
to
give
a
quick
one
for
cube,
ADM
of
ability
or
other
kept
about
Cuesta
status,
the
application,
so
when
queue
mediums
creates
the
coaster,
it
creates
a
cube,
ADM
config
config
map
in
the
cube
sees
keep
system
namespace
and
inside
the
config
map.
There
is
discuss
the
status
object
that
is
used
to
track
the
API
server
endpoints
in
an
HJ
deployment,
so
there
is
a
change
to
red
so
that
they
can
propose
that
we
should
use
annotations
to
track
these
API
endpoints,
and
this
is
all
this
also
effects.
A
Hcd
members,
we
have
a
kept
there
here
is
the
link,
and
also
we
have
a
work-in-progress
PR
that
marks
Costas
Thanos
has
deprecated
and
starts
implementing
the
usage
of
the
annotations
that
the
application
is
against.
This
is
a
beta
feature
for
bits
of
correct
me.
If
I'm
wrong,
I
think
we
are
going
to
remove
it
in
two
versions
from
committee.
It.
C
C
D
A
No,
this
is
not
directly
ready
to
see.
La
Costa
status
is
a
is
a
object
that
is
managed
by
cue
medium.
It's
an
implementation
detail.
Most
users
should
not
care
about
it,
but
we've
seen
people
track
the
elements
of
this
coastal
status
in
high-level
deployments.
But
what
I
mean
by
high
level
deployments
is
when
somebody
raps
cue,
medium
in
a
custom
infrastructure,
deep
wire,
for
instance,
and
they
might
want
to
track
the
question
status
endpoints.
A
C
So
we
are
changing
the
content
of
the
cubed
min
config
config
map,
okay,
so
this
and
by
deprecating
piece
of
information
which
is
stored
in
this
config
mode.
That
is
called
a
cluster
status.
Okay,
this
is
impact
only
the
tools
that
read
the
Cooperman
config
map
and
make
use
of
the
cluster
status
inside
config,
but.
B
I
think
you
so
I
think
the
cross-project
news
that,
for
instance,
group
is
that
at
the
last
cups
update
and
also
the
last
cost
Adams
update,
we
demo'd
an
integration
of
cops
with
cluster
add-ons,
so
sort
of
similar,
but
maybe
a
little
different
way
than
Lee
is
thinking
about
that
and
got
a
ton
of
great
feedback.
So
thank
you
to
everyone
for
that,
but
hopefully
we'll
get
those
two
projects
working
together
and
maybe
some
Covidien
and
restaurants
work
together
as
well.
B
E
Yes,
we've
been
trying
to
get
some
infrared
up
for
building
the
images
in
a
pipeline
by
a
CR
CD.
We
have
got
some
agreement
to
have
a
new
account
spun
up.
We
just
need
a
plus
one
from
I
think
Justin
or
autumn
a--they
for
creating
the
this
account.
So
if
you
look
at
a
conversation,
Timothy's
just
us
for
a
plus
one
at
the
bottom,
so
just
the
ping
there
so
I'm,
assuming
he's
asking
for
a
plus
one
yeah.
E
That's
fine
so
with
us,
then
we
will
be
able
to
at
least
start
spending
up
some
bold
scible
the
images
and
then
we
need
to
have
a
discussion
in
terms
of
where
cube
deployed,
config
admin
and
the
ansible
based
solutions
where
we
converge
on
I,
don't
know
Justin
what
your
day
like
tomorrow
is.
If
you
could
build
it,
join
the
image
folder,
not
face-to-face
office
hours
tomorrow
or
if
any
time,
maybe
we
can
set
up
another
discussion
and
kind
of
discuss
it.
B
B
E
E
So
this
a
bunch
so
there's
actually
three
different
sub
projects
of
image,
folder,
so
I
think
we
kinda
need
to
decide
on
the
direction
and
what
we're
gonna
support
and
what
we're
not
going
to
support
and
where
we're
going
to
put
it
so
I
think
there's
a
better
work
that
we
need
to
do
before.
We
do
that.
E
Okay,
because
one
of
the
problems
is,
is
that
there's
code
here,
but
the
output
is
not
available
anywhere
or
it's
available
in
different
places.
So
the
cat
free
project,
for
example,
will
produce
an
image
over
here
and
then
link
to
it
on
the
Kaveri
page
Masek
Kappa
a
will
do
the
same,
so
they'll
produce
an
image
from
their
workstations
and
upload
it
into
an
account
and
then
link
to
it
from
from
their
home
page.
E
So
this
this
project
is
to
just
create
the
project,
and
then
the
idea
is
that
we
use
Google
Cloud
bolt
to
spin
up
VMs
for
nested
virtualization
that
we
can
use
to
create
at
least
cap
cap
V
or
VMware
and
OpenStack
and
other
images
and
then
potentially
use
those
the
single
source
into
AWS
and
GCP
and
Azure
as
well
and
or
to
both
on
those
providers
using
credentials
that
are
injected
into
Google.
Cloud
bolt
Thank.
B
E
So
my
understanding
is
that
this
PR
is
actually
probably
not
gonna
work
and
we'll
probably
only
get
access
to
a
service
account
that
has
a
bucket,
and
then
we
iterate
on
that
to
add
more
permissions
to
it,
as
as
we
go
along
rather
than
getting
all
the
permissions
that
we
could
potentially
need
and
then
try
to
reduce
it
down.
So
we
want
to
start
with
the
minimal
and
then
add
permissions
one
by
one
and
whatever
else
is
required
in
order
to
make
it
functional.
Yeah.
A
E
E
A
H
All
right
so
yeah
we
released
one
seven
I
think
last
week,
but
it's
been
in
the
works
for
like
a
month
were
actually
up
to
one
seven
two
and
the
big
highlights
there
are
the
new
docker
driver
is,
is
an
expert
in
experimental
stage.
We
added
like
a
container
pause
so
that
you
can
keep
mini
cube
like
the
VM
running,
in
the
background
with
like
as
few
resources
taken
as
possible
and
we've
we've
both
shortened
startup
time
and
and
lower
like
CPU
resources.
In
terms
of
when
you
start
up.
H
Regards
to
overlapping
scopes,
we
have
talked
with
Ben
in
terms
of
like,
what's
going
forward
like
we're
open
to
collapse
to
collaboration,
and
it
seems
like
he
is
at
the
end
of
the
day,
we're
just
most
interested
in
like
what's
best
for
the
kubernetes
community
at
large,
the
the
the
docker
the
vm
free
deployments
has
been
on
our
roadmap
for
a
long
time.
So
it's
something
that
was
like.
We
were
always
interested
in
doing.
I
H
J
We've
just
been
working
towards
the
v1
alpha
3
release.
We
still
had
a
very
active
development
going
on
regarding
machine
pools,
machine
health
track
and
qadian
control
plane.
All
of
them
are
still
expected
to
land
in
v1,
alpha
3,
but
we
are
planning
on
cutting
a
pre-release
build
on
Friday
and
not
all
of
those
features
may
actually
make
it
in
there
or
be
fully
complete
in
that
pre-release
build.
But
you
know
we
wanted
to
make
that
available
for
especially
for
developers
of
providers
so
that
they
can
start
testing.
A
I
A
I
J
B
And
not
at
the
file
level,
and
it
wouldn't
be
in
the
case
in
for
working
group
payment,
because
I
presume
this
bucket
is
owned
by
its
infra
I
would
I
don't
know,
but
I
would
speculate
that
I
know
that
other
clouds
support
profile,
reporting
and
so
I
would
speculate.
That
GCS
also
supports
it,
and
I
do
have
a
pipeline
when
we
move
it
over.
We
will
be
able
to
activate
that,
but
I
don't
know
what
are
we
looking
at
here?
We're
looking
at
the
it
will
be
on
the
top
level
bucket.
There
will
be
options.
B
K
B
B
A
Yeah
nobody
complaints
about
this
except
the
the
Coast
API
jobs
and
we
already
modified
them.
So
if
anybody
can
place
in
the
future,
we
should
just
tell
them
to
do
the
same
by
there
is
potential
install
dependency,
the
complete
dependency,
if
needed,
but
that's
it
very
much
in
the
future
I
guess
we
are
going
to
have
a
consistent,
build
from
the
same
specs.
The
main
problem
here
was
that
we
are
using
multiple
specs
from
different
constants,
so
I
am
happy
that
Mike
helped
us
with
merging
that
Pierre.
I
You
know
sorry
one
quick
question
at
that:
do
you
know
if
we
decided
if
we're
moving
all
system,
D
files
out
or
just
the
like,
Devon
rpm
specs,
because
previously
they
were
sort
of
intertwined
and
I
believe
the
like
unit
files
at
least
are
useful
to
have
possibly
independently
other.
For
example,
container
D
has
a
system
the
unit
without
actual
packaging
in
medieval
I,
wonder
if
we
can
find
some
way
to
reconcile
that.
A
I
Right
so
the
actual
respect
for
the
RPM
makes
sense,
but
like
the
files
that
go
in
it,
like
the
system,
be
droppin
the
cubelets
system
D
right
now,
those
happen
to
be
stored
with
the
spec
but
and
I
think
it
doesn't
make
much
sense
to
need
to
go
to
K
release
to
find
like
a
cubelet
system.
D
file
like
just
the
service
unit
file
yeah.
A
I
I
A
I
I
A
L
Thank
ya,
nice,
nice
improvement
on
the
indian
test,
its
verifying
the
unit
and
join
now.
So
that's
thanks
to
my
memories,
been
contributing
in
a
few
ways,
and
I
would
for
one
of
the
things
that
our
time
is
working
on
is
certificate
renewal,
and
I
I
would
like
you
know
if
anybody
is
interested
in
that,
especially
if
you
have
you
know,
if
you
have
used,
let's
say:
renewal,
a
certificate,
renewable
coop
ADM.
L
Yet
what
the
you
know.
What
what
the
issue
is?
It
looks
like
Koo,
baby,
a
coupe
API
server
does
need
to
be
restarted,
to
be
able
to
communicate
with
that
CD
again,
so
yeah
I'm
I'm
digging
into
that
a
little
bit.
But
if
anybody
has
has
ideas
or
suggestions,
that
would
be
very
welcome
and
yeah.
That's
that's
it
you're.
L
I'm
not
sure
I
from
what
I
can
tell
there's
a
API
server
uses
the
sed
client
library,
which
appears
to
have
that
that
that
library
is
shared
by
the
sed
server
and
yet
the
client
and
as
far
as
I
know
it
does
reload
certificates
from
from
disk
when,
when
they
change
so
yeah,
it's
not
it's
not
clear
to
me
right
now,
where
you
know
where
the
problem
lies,
but
we
looking
into
that.
It's
not
it's
not
really
a
blocker
for
for
certificate
renewal.
That
is,
you
know
the
user.
L
L
A
C
L
Yeah
absolutely
I
I
think
that
the
the
latest,
the
latest
comments
are
at
the
various,
the
very
bottom
they're,
not
in
line
and
yeah
I'm
going
to
try
to
spend
a
little
time
to
figure
out.
You
know
what
what
is
what
is
happening,
I
think
it's
further
the
comments
that
I
left
the
latest
ones
anymore
or
at
the
very
bottom
and
yeah.
L
Maybe
maybe
there
is
maybe
there's
a
bug
and
if
I
can
find
it,
you
know
definitely
file
file,
an
issue
and
if,
if
I
can
get
results,
that
would
be
great
all
around
you
know
for
certificate
renewal
I,
as
far
as
I
know
that
this
kind
of
behavior
this
need
to
restart
curve.
Api
server
has
been
around
since
I.