►
From YouTube: SIG Cluster Lifecycle - Cluster Addons 20210831
A
Right
welcome
everyone
to
the
cluster
add-ons
meeting
today
is
the
31st
of
august.
A
B
Thank
you.
Yes,
I
I
hope
this
isn't
going
to
be
too
controversial
because
we'll
see
the
kubota
declarative
pattern.
Library
currently
does
not
use
any
versioning.
So
all
the
versions
are
sort
of
pinned
to.
I
guess
it's
a
date
and
a
shot,
and
I
think
there
was
a
pr
from
ato
ato
88
who
proposed
bumping
the
kubernetes
version.
That's
like.
Oh,
we
actually
don't
have
versions
in
our
library
so
that
pr
is
linked.
B
I
think
it's
actually
currently
closed
right
now,
sort
of
in
anticipation
of
a
decision,
but
looking
at
like
controller
runtime
it
it
has
versioning.
So
I
just
want
to
call
people's
attention
to
this
library.
B
If
anyone
thinks
we
should
not
do
versioning
or
if
anyone
has
a
proposal
for
a
different
versioning
proposal,
then
please,
let
me
know,
but
otherwise
I
was
going
to
try
to
mirror
the
controller
runtime
versioning
numbers
actually,
if
we
can,
but
at
least
the
major
and
minor
I
don't
know
if
we'll
keep
up
with
patches,
but
it
seems
like
that's
the
that
seems
like
the
easiest
way
to
to
to
keep
things
aligned
and
you
can't
sort
of
use
one
without
the
other.
B
So
we
can't
use
computer
pattern
without
controller
runtime,
so
aligning
to
the
controller.
Runtime
version
seems
like
the
easy
way
to
go.
C
B
I
I
don't
think
there
are
any
huge
implications.
It
just
means
we
need
to
be
a
little
bit
more
disciplined
about
not
making
breaking
changes
without
bumping
the
the
minor
technically.
I
guess,
while
we're
still
on
major
zero,
and
I
think
it
actually
will
de-escalate
some
of
the
debates
about
whether
or
not
we
you
know
should
bump
versions,
because
we'll
sort
of
have
a
clearer
version
strategy
and
we
could
probably
have
more
than
one
version
open
at
once.
B
So,
for
example,
if
there
was
a
security
vulnerability
in
a
particular
version
of
one
of
either
our
library
or
a
dependency,
we
would
be
able
to
bump
it
and
backboard
those
bumps
without
requiring
people
to
go
to
the
latest
version
which
is
sort
of
what
they
have
to
do
today,
but
yeah
a
little
bit
more
branch
management.
I
guess,
but
not
not
as
a
library,
it's
not
too
burdensome.
In
my
opinion,.
A
B
I
do
hope
it
would
be
uncontroversial.
It
should
just
be
easier
to
consume
at
a
small
cost
to
the
library
maintainers,
which
is
actually
has
some
benefits
as
well.
So
I
think
it's
hopefully
seen
as
a
universal
good
thing
to
start
using
versions,
but
I
welcome
anyone
to
tell
me
why
it's
not
worth
the
not
worth
it
or
something.
B
I'm
adding
an
issue
now
and
I
will
link
to
the
existing
one
and
the
existing
pr
and
I'll
put
it
in
the
agenda.
Notes.
Gender
notes.
D
Yeah
yeah,
I
apologize,
it
seems
like
this
has
been
dragging
out
forever.
We
keep
getting
like
preempted
with
other
work
here,
but
we
managed
to
get
a
skeleton
up
for
how
we
think
everything
should
be
organized
with
some
sort
of
like
no
controllers
for
bundles
and
provisioner
classes.
So
just
for
some
some
background.
It's
like
the
the
super
essential
extensible
top
level
types
for
just
sourcing
content
on
the
cluster
and
then
managing
it
and
applying
it
over
time
and
so
like.
D
We
can
create
these
for,
like
any
number
of
things,
and
if
we
decide
on
like
the
format
for
add-ons
like
manifest
bundles,
we
can
sort
of
use
these
on-cluster
representations
with
a
provisioner
for
that
as
well.
So,
hopefully,.
D
That
now
that
we've
sort
of
exploded,
all
the
issues-
we're
tracking
internally
on
here
at
red
hat,
for
like
what
we
wanted
to
do
for
the
project
onto
the
issues
in
github,
people
can
feel
free
to
like
pick
up
and
contribute
when
they
want
to,
and
also,
I
think,
it's
a
little
bit
more
clear
about
like
how
you
would
add
a
new
sort
of
provisioner
to
this
repository.
If
you
wanted
to
include
like
one
natively,
so
the
issues
link
is
there
on
the
on
the
agenda.
D
But
that's
that's
about
it.
A
D
Yes,
so
we've
kind
of
like
broken
things
down
most
of
these
refer
to
like
these
plain
kids
models,
which
are
essentially
just
a
a
directory
with
a
bunch
of
manifests
in
them
and
some
metadata.
The
docs
will
be
published
to
this
repo
soon,
but
we
have
a
large
like
automate
and
build
testing
workflows,
which
is
just
like
some
generic.
D
Some
generic
stuff
for
like
getting
provisioner
running
on
on
the
cube
cluster
using
a
workflow
github
workflow,
that's
like
pretty
important.
We
have
perhaps
some
features
for
like
overriding
bundle
content.
So
this
is
like
maybe
changing
the
types
so
that
you
can
put
whatever
you
want
in
to
a
bundle
at
runtime
directly,
maybe
like
reference,
a
volume
or
inline
manifests
for
a
particular
type
of
bundle,
which
is
also
important,
but
other
than
that.
These
are
all
laterally,
lateral.
A
Great
yeah,
I
also
just
saw
at
the
top
of
the
agenda.
There
was
a
demo
from
rock
pack.
Did
this
already?
Did
you
already
do
a
demo
of
what
can
do
already?
D
Yeah
so
tim
who's
also
on
this
call,
did
like
some
prospective
stuff
and
just
prototyped
a
bunch
of
stuff
out
that
works,
but
was
not
necessarily
like
conducive
for
getting
like
upstream,
more
folks
contributing
to
it.
What
I
mean
by
that
is,
like
the
organization,
maybe
wasn't
where
we
wanted
it
to
be,
but
we
kept
having
things
preempt
us
here
and
tim
was
busy.
So
I
didn't
want
to
step
on
his
toes
by
giving
a
presentation
of
his
work
without
him.
B
Yeah,
it's
more
just
intended
as
a
a
gathering
place
for
future
topics
for
people
to
add
to
if
they,
if
they
want
to
that,
was
the
intent
of
that
section.
I
put
I
previously
put
the
the
versioning
discussion
in
that
backlog.
E
Yeah
hi,
my
name
is
james
and
I
guess
I'm
kind
of
wondering
at
one
point.
E
There
was
a
pr
for
cube
adm
to
like
pull
engineer
component
configs,
and
my
understanding
was
that
that
would
make
it
easier
for,
like
you
know,
bootstrap
time
add-ons
to
be
set
up
on
the
cluster,
and
you
know
maybe
I've
missed
it.
I've
been
trying
to
attend
these
meetings.
You
know
on
the
regular
boat
was
that
is
there
an
intent
to
move
that
forward?
Is
there
something
else
coming
that's
going
to
replace
that
just
kind
of
wondering
like
what
the
plan
is
for
bootstrap
time
add-on
stuff.
B
I
I
I
can
give
a
bit
of
background.
I
mean
I
think
I
contribute
more
to
k-ops,
so
I'm
trying
to
get
it
into
chaops
as
the
sort
of
first
proof
of
concept
of
where
that
goes
and
sort
of
tackling
some
of
the
trickier
issues
there.
First,
I
think
I
I
think
kubedium
lets,
you
add
an
additional
manifest.
B
I
don't
know
if
that's
true,
but
some
of
the
operators
that
are
purely
additive,
you
should
be
able
to
effectively
just
just
add,
with
an
additional,
manifest
and
sort
of
get
some
of
the
benefits
there.
So
that's,
hopefully
a
step
forwards.
I
think
there
are
two
cases
where
that's
tricky,
one
of
which
is
where
the
operator
is
currently
built
into
cube,
adm,
which
I
think
is
core
dns,
for
example.
B
So
if
you
wanted
to
somehow
swap
that
out,
that
might
be
a
little
trickier
and
the
other
one
of
which
is
where
either
the
configure
like
configurations
has
to
go
one
way
or
the
other
either
in
back
into
the
cluster
configuration
or
from
the
cluster
into
the
operator,
and
that's
a
trickier
problem
that
again,
we
don't
have
a
yet
have
a
great
answer
for
it.
B
How
can
easily
get
the
the
kube
dns
the
coop
dns
ip
address,
even
in
the
case
that
it's
sort
of
running
in
like
a
pre-mana
pre-generation
mode,
but
there's
also
the
option
to
that.
The
operator
will
get
those
get
some
information
by
discovering
it
from
the
cluster
after
having
been
installed.
B
So
I
think
that's
where
we
are.
I
think
most
of
the
progress
is
currently
happening
and
it's
not
great.
Progress
is
currently
happening
in
in
chaops
right
now,
but
I
hope
that
once
we
get
one
of
them
working
that
we
can
also
get
the
others
working
on
the
same
sort
of
patterns.
E
B
Exactly
I
mean
all
it's
going
to
do.
Is
it's
going
to
do
a
couple
apply.
I
actually
don't
know
whether
kuber
em
has
the
ability
to
do
that.
Apply
for
you.
It's
not
like
you
know.
It's
not.
The
hardest
thing
to
do
is
a
separate
line
in
a
in
a
bad
in
a
script
anyway,
so
I
don't
know
whether
they've
built
it
in,
but
that's
sort
of
the
you
know
for
anything,
that's
purely
additive
that
that
works
just
fine.
So
that
could
be
a
good.