►
From YouTube: 2020-04-01 - Cluster API Office Hours
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
Wednesday
April,
1st
2020.
This
is
the
close
to
API
office
hours.
Meeting
cluster
API
is
a
sub-project
of
said
cluster
lifecycle.
This
meeting
is
being
recorded
and
will
be
posted
on
YouTube
to
the
SIP
cluster
lifecycle
playlist.
Once
it's
ready
and
please
use
our
meeting
etiquette,
which
is
to
use
the
raise
hand
feature
of
zoom.
A
If
you
do
have
something
that
you'd
like
to
say,
I
will
be
monitoring
and
will
call
on
you
and
also
we
follow
the
kubernetes
code
of
conduct,
which
is
generally
summarized
as
let's
all
be
nice
to
each
other.
Please
go
into
the
April
1st
section
in
the
document
and
add
your
name
to
the
attending
list,
and
at
this
point
what
we
like
to
do
is
say
welcome
to
any
new
attendees
and
if
anyone
is
interested
in
saying
hi
and
introducing
yourself,
I
will
pause
now
for
a
minute
to
see.
If
anybody
wants
to
do
that.
A
B
C
So
we
just
landed
a
readme
into
the
autoscaler
cluster
API
providers
that
I
got
the
PR
link
there.
I
guess
it
hasn't
quite
lurched
yet,
but
it
just
gives
some
basic
instructions
on
how
to
start
the
autoscaler
with
coaster
api
and
then
how
to
how
to
make
it
actuate
and
then
we've
also
added
a
cluster
API
label
to
that
repo
as
well.
So
if
you're
going
to
add
issues
or
if
you'd
like
to
put
a
PR
up,
you
can
see
the
I've
added.
C
The
link
here
is
to
filter
issues
based
on
cluster
API
and
then,
lastly,
our
next
steps
now
that
we've
got
this
landed,
is
kind
of
focusing
on
cleaning
up
the
code.
We've
got
there,
there's
a
few
kind
of
internal
improvements,
we'd
like
to
make
and
then
adding
more
unit
tests
and
end-to-end
tests.
So
that's
about
it
for
the
announcement
all.
D
So
I
have
moved
these
points
to
pros
all
from
Google
Doc
into
APR.
Now
there
weren't
any
comments
for
a
couple
of
weeks,
so
it
seemed
about
the
right
time.
So
I've
fixed
up
the
last
couple
of
comments
from
Jason
pop
that
in
the
P
R,
which
I've
linked.
If
you
haven't,
read
that
and
have
a
go
there.
If
you
have
then
please
review
in
case
there's
anything,
that's
changed,
I,
don't
think
there
is
and
I
added
everyone
who
I
could
recall
that
had
refute
the
document
to
the
reviewers
list.
A
B
Yeah
I
didn't
really
phrase
that
as
a
question,
sorry,
but
my
question
is
in
the
cluster
cattle
QuickStart
that
we
have
in
the
book.
I
know
that
the
versions
are
automatically
updated
as
we
release
new
versions
of
cluster
API
and
so,
for
example,
if
we
release
dot
like
let's
say
week,
release
0.3
dot
for
next
week.
That
means
the
book
QuickStart
will
gets
your
Doc's.
We
got
four
as
the
example.
That's
someone
should
download,
which
is
great,
but
what
I
was
wondering
is
how
are
we
do?
B
We
expect
that
might
be
a
problem
if
the
infra
providers
aren't
moving
as
fast
and
the
version
I
don't
expect
there
to
be
any
breaking
changes
in
the
knight
in
patch
versions,
but
what
if
there
was
something
that
was
compatible
with
the
existing
version
of
an
infant
provider
in
there?
That
means
we
are
allowing
our
users
to
kind
of
take
the
new
versions
before
we
even
tried
it
or
moved
our
tests
to
that
version
in
the
infer
providers
we
go
if
that
makes
sense,
so
I'm
just
wondering
if
that's,
why
is
it.
E
I'd
say
like
in
general,
copy
should
not
break
any
contracts,
and
if
it
does
so,
when
you
don't
publish
that
version,
there
is
some
precedent,
for
example,
controller
run
time
with
it.
This
the
breaking
change
went
in
like
without
noticing,
but
I
would
not
expect
like
0
3,
4
3
3
5
to
do
any
breaking
changes
that
would,
in
fact
that
mean
for
structure
provider
and
that,
if
that
happens,
there
is
a
bigger
issue
that
we
need
to
fix.
B
Right
would
it
make
sense
to
have
like
I'm
just
trying
to
think
of
waves,
or
we
could
have
I
know
this
shouldn't
happen,
but
what,
if
it
does
like?
How
can
we
find
out
if
it
does,
without
the
users
telling
us
I'd
love
to
have
ideally
a
test
that
tells
us,
instead
of
having
it
found
out
by
people
trying
trying
out
cluster
API
from
the
book
and
I'd.
F
E
Yeah
we're
actually
putting
the
end-to-end
story.
There
is
like
a
lot
of
issues
for
bridge
has
been
working
on
cluster
kado,
specifically
so
I
would
expect,
like,
probably
in
this
cycle
or
next,
to
have
an
enzyme
story
that
actually
is
PR
blocking
or
periodic
job
to
like
tells
us
if
something
is
broken,
on
faster
cuddle
in
it
or
upgrades
or
move,
etc.
B
A
All
right
sounds
like
we're
at
a
good
spot
there,
or
we
will
be
when
we
get.
A
G
G
This
test
will
be
implemented
in
an
in
cluster
API
and
then
in
details
should
be
just
a
figure
out,
but
then,
ideally,
each
provider
can
scale
to
the
same
test
in
a
nice
test
test
great.
So
we
have,
let
me
say,
a
consistent
signal
for
every
provider
and
this
relief
to
me
more
and
more
I
think
about
is
this
is
kind
of
starting
the
seed
of
what
can
be
a
cluster
API
confirmes
things
and
in
Safi
that
for
me,
this
group,
as
you
should
start
thinking
for
the
long-term.
A
A
H
Think
it's
acceptable
for
a
window
of
time,
but
I
think
as
we
mature
and
get
to
a
certain
point.
We
might
want
to
make
sure
we
have
a
clear
contract
with
our
client
libraries
to
make
sure
that
they're,
clean
and
concise
and
and
supported
for
a
given
window.
So
I
think
it's
fine
for
the
current
stage,
because
it's
very
nascent
but
the
if
you
saw
any
if
you
saw
the
fallout
from
letting
go,
for
example,
the
Twitter
rampages
that
have
occurred
because
of
the
breaking
changes
that
occurred
1.18.
H
I
E
That
would
complicate
the
release
process
a
lot
and
in
general
we
actually
went
the
other
direction,
which
was
to
remove
some
modules
from
this
repo.
Given
that
I
think
we
had
three
or
four
in
the
bus,
and
now
we
have
only
one
I
do
agree
the
like.
Maybe
we
can
just
put
this
or
like
or
like
a
list
of
restriction
on
for
its
Jo
3x
and
then
reevaluate
in
the
next
cycle.
A
C
Thanks
so
by
way
of
a
little
context
in
the
last
few
weeks,
I've
been
involved
in
several
kind
of
conversations
and
debugging
sessions
with
new
users
who
come
to
the
project
and
are
you
know
seeing
the
doctor
provider
on
our
QuickStart
and
trying
to
work
through
the
process
of
getting?
You
know
that
provider
running
so
they
can
just
kind
of
play
with
coaster
API
and
in
the
course
of
those
discussions,
I
managed
to
come
up
with
some
instructions
for
a
fedora
based
install
of
this
and
another
user.
Don't
think
he's
here
today
things
Alexander
Hughes.
C
He
put
together
a
set
of
instructions,
very
thorough
instructions
for
how
to
get
this
running
on
a
bun
too,
and
so
he
and
I
started
to
talk.
He
wanted
to
contribute
and
I
think
it's
a
great
idea,
but
as
we
looked
at
the
docs
I'm
kind
of
wondering,
if
we
shouldn't
take
out
the
doctor
information
from
the
quick
start,
you
know
keep
the
quick
start
focused
on
the
actual
providers
and
then
in
our
developer,
section
create
a
single
standalone
doc.
C
That
just
speaks
to
the
doctor
provider,
and
you
know
part
of
the
reason
I'm
bringing
this
up
is
because
the
advice
that
comes
from
some
of
the
people
who've
been
on
this
project
a
little
longer
than
myself
seems
to
always
be
well
if
you're
trying
to
use
the
doctor
provider,
this
is
meant
only
for
development,
and
it's
meant
only
for
testing
and
so
I
think.
Maybe
if
we
break
it
apart
from
our
actual
QuickStart,
we
could
have
a
document.
C
So
I
wasn't
quite
sure
what
the
next
best
step
in
terms
of
the
community
here
is
what
you
know
whether
I
should
put
up
just
a
PR
with
what
I'm
proposing
or
should
we
have
I
don't
know
if
it
needs
an
enhancement,
doc
or
not,
but
I
wanted
to
discuss
it
here
to
see
if
anyone
had
ideas
or
maybe
if
they,
if
they
think
this
is
worthwhile,
you
know
Alexander
would
like
to
collaborate.
So
maybe
we
could
get
something.
A
A
So
we
have
an
overarching
one
to
revisit
the
overall
documentation
structure
that
we
have
filed
so
really
you're
here,
I,
don't
know
if
you've
got
any
updates.
If
you,
if
you
and
your
team
about
any
chances
to
try
and
think
about
the
documentation
structure,
but
I
think
it
does
tie
in
to
your
question.
Mike.
J
Yeah,
definitely
if
we
want
to
do
a
break
out
on
the
conversation,
I
mean
I've
been
going
through
the
QuickStart,
a
bunch
I
stood
out
of
here
for
a
couple
of
tweaks
for
the
other
stuff,
but
I
would
love
to
like
good
more
of
a
team
together
that
could
kind
of
look
at
the
docs
overall
pass
just
me.
So
if
you
want
to
comment
on
that
PR,
maybe
we
can
get
a
time
together
next
week
and
talk
about
it.
Yeah.
C
Definitely
I
guess
my
my
concern
at
the
moment.
Is
there
guy
no
I'm,
not
sure
that
I
can
get
involved
in
a
deep
documentation,
kind
of
refactor,
but
I
think
I
think
we
could
excise
the
darker
stuff
from
the
QuickStart
and
then
just
make
a
solid.
You
know
kind
of
docker
provider
single
page
dock
on
its
own.
Yes,
if
that
feeds
into
a
larger
effort
to
refactor
the
docks,
I'd
be
happy
to
help
them.
Oh
cool.
A
A
Mike
I
think
that
it
shouldn't
be
front
and
center
with
the
quick
start
necessarily,
but
I
mean
if
we
wanted
to
have
a
link
on
the
quick
start.
That
said,
hey
don't
have
access
to
a
cloud
provider.
We
have
this
non-production
docker
provider.
You
can
use
and
go
over
here
for
that
documentation
or
something
along
those
lines.
Yeah.
C
C
A
C
B
I
just
wanted
to
bring
this
issue
up
again.
It's
been
open
for
a
while
and
right
now
as
priority
long-term,
but
I
know
a
bunch
of
people
commented
on
the
issue
saying
they
also
run
into
this
and
I
run
into
it
pretty
frequently,
if
I
forget
to
export
my
own
github
token
and
I
know
that
people
that
I've
pointed
to
the
quick
start
to
try,
it
out
have
also
come
back
to
me
saying
they
run
into
it.
B
A
G
This
one
is
one
testing,
so
there
is
not
a
sequence
of
action.
The
problem
is
that
the
limit
of
the
free
tire
of
oh
sorry,
the
rate
limit
of
the
github
API,
is
very
low.
This
is
an
it
is
enough
if
you
do
custom
control
in
it
or
the
simple
basic
operation.
But
if
you
start
testing
operationally
in
Aarau,
you
creationary
means.
K
G
L
E
C
B
After
talking
with
a
few
people
about
extending
past,
essentially
expanding
the
cathy
test
framework
to
includes
doing
some
conformance
testing
and
so
I
just
wanted
to
bring
this
up
and
I
think
the
issue
has
a
lot
of
details
and
people
have
already
started
conferencing
on
it.
So
if
you're
interested
and
you
have
any
opinions,
please
read
and
comments
hoping
to
get.
The
discussion
started
on
the
issue
first
and
then
we
can
potentially
have
a
follow-up
meeting
or
sync
if
needed.
B
E
Yours
so
for
TCP
X,
the
first
few
releases
mean
kind
of
like
some
bugs
internal
testing
and
also
like
some
external
folks
testing,
thanks
again
New
Relic
for
fine,
so
many
bugs
and
fixing
them.
So
the
goal
of
this
planning
is
kind
of
like
improve
the
robustness
of
the
code,
but
also
adding
end-to-end,
which
is
something
that,
like
we
are
missing
today.
So
if
you
would
like
to
contribute
or
like
you
have
ideas
on
how
to
kind
of
like
make
ACP
better,
there
is
a
plan
here
and
the
perfect
few
people
signed
up
for
it.
E
The
first
step
is
to
really
write
unit
tests
in
integration
and
end-to-end
tests
in
the
kcp,
so
everything
under
the
control,
plane,
folder
and
then
the
next
test
is
the
Nexus
kind
of
like
some
refactoring
there's
some
details
here
that
we'll
have
to
open
different
issues
like
the
first
one
is
removing
annotations.
If
you
look
at
the
decoder
at
some
of
the
code,
wasn't
really
entrant
and
the
first
step
was
to
remove
one
of
the
annotations,
and
so
we
want
to
remove
also
the
other
annotations.
E
These
are
notations,
for
those
of
you
are
not
familiar
with
the
code
base
or
use
to
keep
track
of
State
and
where
we
want
to
be
at
is
to
keep
track
of
State
by
reading
the
state
at
the
source,
rather
than
like,
like
storing
the
state
on
jeez
and
I.
Think
any
your
working
on
the
second
point:
25
77,
or
at
least
at
the
first
part
of
it,
which
is
adding
a
water
to
the
remote
clusters.
F
It
was
Vince,
great
and
I
put
behind
him.
So
last
week
we
had
discussion
on
the
copy
and
the
document
we
have
here
is
the
one.
So
he
we
had
asked
if
somebody
can
represent
the
ship
virtual
meeting,
and
so,
if
you
open
the
virtual
meeting
link,
we
had
definitely
discussion.
We
can
go
down
further,
so
the
portion
which
I
am
referring
to
is
related
to
cap
BK.
F
You
can
see
here
tentative
acceleration
of
CA
C
PK,
so
we
are
even
not
clear
what
the
exact,
but
you
can't
be
K
is
what
here
it
is
and
if
you
open
the
clusters
control
plane.
That
is
what
yeah
go
ahead.
Here
is
the
floor,
which
we
had
so
according
to
what
we
think.
If
this
is
required,
but
maybe
may
not
be
I,
don't
know
exactly
how
Vince
you
are
using
the
control
plane,
so
we
have
in
it.
Then
we
delete
before
that.
Is
we
clean
up
and
then
start
and
then
how
the
cluster
plane.
F
F
F
So
in
the
process
we
land
up
in
copy
to
pass
on
the
information
to
metal
cube,
and
so
at
this
time,
what
we
have
done
is
you
can
look.
There
is
a
code
also
behind
you
should
go
to
the
next
one.
Sorry
go
into
that
particular
place.
Yeah.
There
is
a
code
right
example,
a
ship
code,
so
we
our
code,
is,
we
would
like,
if
somebody
can
review.
M
M
You
know
in
general,
it's
very
helpful
if
you
have
a
you
know
if
you're,
if
you
have
a
question
about
trade-offs,
if
you
can
provide
something,
you
know
either
for
others
to
easily
reproduce
or
if
you
you
know,
you've
been.
Actually
you
know,
instead
of
instead
of
sharing,
let's
say
some
input
on
its
own.
That
needs
to
be
reviewed
so.
M
F
We
how
to
substitute
values
within
the
template,
and
if
we
have
to
substitute,
should
we
do
it
prior
to
giving
our
after
giving
to
cluster
API,
or
should
it
be
that
we
take
whatever
template?
You
have
interested
API
and
then
use
it
and
pass
the
value
to
it.
So
it's
a
passing
of
values
to
the
metal
cube.
That's
the
problem
we
are
facing,
which
one
is
metal
should
we
do
before
or
after?
F
L
Right,
yes,
so
if
I'm
on
some
question
is
your
or
taking
the
default
templates
that
we're
shipping
in
the
releases
you're
trying
to
use
customized
for
airship?
But
the
question
is:
there's
some
complexity
around
that,
and
should
you
just
make
your
own
template
completely
from
scratch
or
continue
to
use
the
one
that
we
released?
My
if
that
is
the
question
and
I
would
say
just
do
your
own
template,
however,
that
works
for
uses
airship
their
API
or
the
contract
for
us
as
kloster.
L
A
joint
is
not
necessarily
the
default
template,
which
is
more
of
like
a
QuickStart
how
to
get
quickly
started
with
clustered
API.
We
fully
expect
people
to
use
their
own
their
own
definitions
of
whatever
qadian
control
plane
as
long
as
they're
respecting
the
API.
We
do
this
in
cluster
API
provide
every
sphere
already,
so
we
have
a
little
tool
to
programmatically,
generate
the
cube
radio
and
control
template
cubic
qadian
control,
plane
template
with
specific
things
which
very
important
to
the
sphere.
C
A
A
F
Okay,
at
this
time,
since
we
are
still
debating
I
I,
think
I'll
wait
for
one
more
week
next
week,
I
will
have
better
inputs,
because
this
was
just
trying
to
hammer
out
yesterday
and
we
didn't
have
any
support.
We
didn't
really
understood
fully
what
copied
us
from
provider
point
of
view
for
Burma
T.
So
at
least
in
the
team
there
was
nobody
who
was
able
to
answer,
and
so
what
I
couldn't
get
was
really
limited.
F
But
tomorrow's
meeting
that
I
will
try
to
encourage
AT&T
people
to
attend
directly
because
some
of
the
decisions
are
driven
by
them
and
from
my
point,
I
am
trying
to
see
how
we
get
the
to
coordinate,
because
you
people
don't
have
time
to
come
to
us.
Our
people
don't
have
time
to
come
to
you
and
I
am
caught
in
between
trying
to
get
how
do
I
resolve
this
issue
and
I
also
went
to
last
week's
comments
made
by
Vince.
Then
you
layer,
something
on
the.
F
What
was
that
related
to
packet
host
or
something,
but
that
one
which
was
pointed
there,
that
code
is
now
no
more
there
I
think
just
disappear,
so
that
was
related
to
bare
metal
over
there.
So
if
you
have
that
still
the
link
are
some
code
somewhere
else.
Point
to
me
on
that
from
the
last
week's
you've.
H
So
ticker
Bell
is
there
open
sourcing
and
they
are
writing
their
specs
and
Docs
and
I
believe
they're
going
to
have
they're
gonna
be
working
with
the
community
in
us.
In
particular,
the
kepi
team
I
had
a
conversation
with
them
about
creating
a
provider,
specific
implementation
that
will
be
for
two
parts,
most
likely
one
part
packet
and
another
part
taker
Bell.
So.
F
H
F
F
A
What
why
don't
we
so
I'm
gonna
turn
over
to
Mike
and
then
we'll
move
on
to
the
next
topic?
Prakash,
if
you
are
interested
in
like
a
more
thorough,
deep
dive
of
the
moving
pieces
and
components
of
cluster
API
I
would
recommend
watching
any
of
the
deep
dives
from
any
of
the
previous
cube
cons,
they're
on
YouTube
they're,
recorded
and
available.
As
we've
said,
the
cluster
API
is
about
providing
functionality
with
contracts
for
custom,
resource
definitions
for
clusters
and
control,
planes
and
whatnot.
The
templates
are
just
examples,
so
the
functionality
is
there.
A
If
you
can't
get
it
working
with
metal,
cube
or
Tinkerbell
or
whatever,
then
maybe
there's
some
bugs,
but
in
general,
cluster
API
is
about
provisioning
clusters,
creating
control
planes
and
worker
machines
and
worker
nodes.
So
that
is
the
playground.
Basically
and
y'all
can
you
know
manipulate
the
templates.
However,
you
need
to
is.
C
A
C
Make
a
real
brief
ID
at
the
risk
of
throwing
more
for
Kyle
here,
I
just
wanted,
I
think
Prakash.
It's
really
important
to
look
at
this
comment
that
nadir
added
to
the
dock,
to
the
notes
here.
I,
don't
remember
if
he
actually
said
it
that
way
or
not,
but
like
the
key
to
understanding
here
is
that
the
contract
for
using
cluster
API
are
these
CR
DS,
not
necessarily
the
templates
that
are
being
pulled
out
so
like
I,
might
recommend
looking
at
looking
at
those
interfaces
to
understand
how
you
might
use
it
with.
F
N
Well,
rather
unpleasant
experience
in
the
Prometheus
zoom
meeting
very
like
this
one,
we're
like
20,
enthusiastic
young
people
came
along
and
showed
us
some
videos,
and
apparently
this
is
happening
a
lot
so
yeah
I
don't
know
whether
I
mean
I.
It's
a
nine
digit
number.
You
know
you
can
sort
of
ward
all
zoom
meetings
so
that
the
password
improves
that
and
then
there's
a
bunch
of
other
things
you
can
do
to
I
mean
maybe
you're
on
top
of
it.
Maybe
you're
kicking
people
off
as
I
speak.
N
H
Have
privileges
and
we
can
boot
people
at
any
time,
so
it's
there's
a
balance.
This
has
been
talked
about
inside
of
it's
not
it's
not
particular
to
this
group.
It's
it's
just
general
with
zoom.
So
there's
the
it's
a
code
control
X
team,
which
is
basically
we
had
this
problem
a
while
ago
and
we
changed
all
our
zoom
meanings
and
how
we
set
up
our
zoom
accounts
to
try
and
alleviate
some
of
these
concerns.
H
But
we
have
three
people,
four
people
on
the
call
who
actually
can
have
host
privileged
access,
so
we
can
boot
people
at
will,
but
there
there's
there's
dis
balance
where
we
we
can't.
If
we
lock
it
down
too
tight,
then
we
don't.
We
have
this
weird
problem
where
we
can't
get
contributors
to
join
like
the
people
who
just
want
to
be
able
to
join
randomly
so
we've.
We
followed
the
guidance
of
contributes.
H
H
A
H
O
I
was
gonna
say
it
might
be
worth
it
to
re-enable
the
waiting
room
by
default.
We
used
to
have
it
and
we
turned
it
off
because
it
was
kind
of
a
pain
because
you
would
have
to
one
of
the
moderators
would
have
to
sit
there
and
then
Nick
people
into
the
meeting.
But
given
the
current
climate
and
just
being
an
an
issue
for
other
meetings,
that
might
be
a
decent
compromise
route
that
we
could
take.
P
O
Yeah
I
was
thinking
more
from
the
use
case
of
being
like
we'd.
Have
somebody
actually
watching
the
room
and
like
a
co-host
in
addition
to
the
actual
host,
that
could
be
more
proactive
for
looting
people,
and
it
would
make
it
more
difficult
for
them
to
attempt
to
rejoin
if
they
are
booted
and
it
wouldn't
add
the
onus
of
having
you
know,
a
password
that
contributors
would
need
to
pass
in
and
we'd
be
publishing
publicly
anyway.
A
C
A
C
H
A
D
Just
gonna
say:
I've
never
seen
this
waiting
room
thing
before
so
I
just
have
one
quick
question
bout
that
which
is
like:
how
does
does
that
mean
by
everyone
who
joins
the
call
has
to
be
like
allowed
in
how
do
we
spot
new
actual,
like
contributors
to
the
calls
versus
like
cuz
I?
There's
this
like
a
good
set
people
here
who
come
most
weeks
like
they're
over,
so
you
can
get
recognized
and
that's
fine
but
like
how
do
we
like
triage
new
people?
And
it's
just
a
case
of
like
let
everyone
in
and
if
they?
N
A
Okay,
that
makes
sense
yeah,
so
I'm,
gonna,
I'm,
gonna
time
box.
This
and
just
say
you
know,
Tim-
will
follow
up
with
Cantor
back
see
if
there's
anything
that
any
new
guidance
if
we
do
start
getting
trolls
and
we'll
deal
with
them,
but
for
right
now
we
fortunately
fingers
crossed
haven't
had
any
problems.
Q
A
A
L
Alright
yeah
the
other
thing
that
came
up
in
just
Ninian
the
upstream
Katri
meeting
was
people
found
it
very
strange.
The
idea
that
we
don't
assume
two-way
communication
between
whatever
is
running
cluster
api
and
the
downs
and
the
downstream
cluster,
and
I
think
I
can
understand
that
in
an
on-premise
environment
in
a
way
right
that
assumption
definitely
with
hold
for
certain
public
clouds,
but
for
on
premise.
The
situation
is
quite
often
a
bit
different.
L
O
A
Well,
please
add
comments
on
25:54
with
ideas,
questions
and
so
on.
All
right,
we've
got
about
five
minutes
left.
So
let's
go
through
the
new
issues
and
thank
you
that's
for
getting
rid
of
half
of
them.
So
first
one
here
is
from
the
dear
and
sure
get
nodes
cluster.
All
named
cluster
roll
and
cluster
role
bindings
exist
when
upgrading
to
118.
Oh,
this
is
per
KC
beam.
L
Yeah,
so
it's
a
dislike
breaking
changing,
cube
ADM
for
118.
We
need
to
implement
that
there's
a
I
think
someone's
had
a
further
question
is
like:
how
do
you
determine
which
release
of
cluster
API
is
compatible
with
which
way
I
should
look
keep
enacted
base?
I
think
we
do
have
an
existing
issue
that
we
point
need
to
be
visit
that
as
well.
You.
A
A
O
A
O
I
worry
about
it
out
of
repo
too,
if
it's
just
a
drive-by
contribution
as
well,
because
if
somebody
sees
it
in
the
helm,
trucks,
repo
and
they're,
like
oh
use,
this
expecting
it
to
be
supported,
it
wouldn't
really
be
supported
there
either
if
it's
just
a
drive-by
contribution.
So
that's
what
happens
with
Niger.
H
Yeah,
that's
what
happens
with
99%
of
em
charts
like
if
you
look
at.
If
you
look
at
the
helm,
trucks
repo
today
or
even
like
over
the
course
of
the
last
three
years,
it's
just
been
a
mass
mashing
of
people
wanting
to
contribute
to
get
their
applications
out
there,
but
they're
not
maintained
so
I,
there's
nothing
we
can
do
about
what
community
wants
to
do.
I
think
we
can
only
control.
This
particular
repository
I
would
say
that
unless
there's
a
lunar
ship,
maintainer,
ship,
testing,
etc,
I
think
that
I'd
be
a
minus
one.
E
I
put
a
comment
in
the
link
PR
like,
which
is
this
also
duplicates,
are
all
pretty
much
all
the
mo
files
that
we
have
today
and
that's
like
already.
The
past
have
been
a
pain
to
maintain
so
either.
This
is,
if
we
include
in
the
repo
has
to
be
like
a
after
release
kind
of
thing
like
so
we
it's
an
artifact
of
the
release
or
it
goes
outside,
which
is
also
my
preferred
outcome.
G
You
want
I
can
provider
that
is
really
similar
to
Cuddy
praticed,
instead
of
creating
simply
Dockery
create
thought
which,
which
contained
not,
and
it
slightly
different
or
by
commenting
on
the
issues
that
being
these
as
a
developer
tool.
We
should
not.
We
should
treat
these
as
the
company
so
to
not
include
it
into
the
default.
The
list
of
provider,
but
eventually
documented
in
the
guy
I,
think.