►
From YouTube: Kubernetes SIG Cluster Lifecycle Cluster Addons 20201124
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
All
right
welcome
everyone
to
the
cluster
add-ons
meeting
today
is
tuesday
the
24th
of
november.
A
We
had
a
bit
of
a
discussion
earlier.
We
weren't
quite
sure
if
the
meeting
was
going
to
happen
because
of
us
holidays.
B
We're
all
here,
though,
so
we
figured
we'd
record
this
since
we're
all
chatting
and
thanks
everyone
for
joining.
We
had
an
awesome
con
last
week
from
many
perspectives,
and
so
hopefully
some
of
you
were
able
to
join
in
with
those
you
can
almost
call
them
festivities
joe
joined
today
and
just
ask
the
question:
what
is
an
add-on
like
in
the
context
of
this
group,
and
so
we
started
chatting
about
that
a
little
bit
when
we
set
up
this
group.
B
Our
first
mission
was
really
just
to
provide
some
uniformity
and
reuse
in
the
space
of
installation
and
cluster
life
cycle
management,
which
is
one
of
the
reasons
why
we're
part
of
sig
cluster
lifecycle,
and
so
the
add-ons
that
were
primarily
focused
on
were
things
like
coup
proxy
core
dns
and
cni
implementations.
B
There's
a
huge
array
of
things
that
need
to
be
in
place
when
a
cluster
comes
up.
Other
categories
are
network
policies
that
you
might
want
in
place
from
the
very
beginning
of
the
lifecycle
of
the
cluster,
you
might
install
that
as
a
package
or
something
like
an
admission
controller.
B
B
When
you
upgrade
the
cluster,
you
probably
want
to
think
about
updating
some
of
these
add-ons
either
before
or
after,
but
then
joe
kind
of
started
talking
about
this
project
that
he's
working
on
called
cube,
add-ons
with
dt
iq
and
cube
add-ons
does
more
than
than
it's
a
more
generic
definition
with
just
helm,
charts
and
kudo
operators
being
available
for
install
so
justin
mentioned
that
we
could.
B
We've
really
been
focusing
on
keeping
the
space
as
uniform
as
possible,
because
there's
so
much
sprawl
in
people
who
are
or
in
projects
that
people
are
trying
to
use
to
do
things
on
top
of
kubernetes.
B
So
we
have
a
lot
of
libraries
in
place
or
that
we've
been
using
to
accomplish
this
group's
goals
and
we
would
love
for
people
to
use
those,
even
if
they're,
outside
of
the
scope
of
what
we
started
the
group
to
talk
about,
and
of
course
you
know
this
is
a
community.
So
we're
always
open
to
expanding
the
mission
statement.
You
know
in
an
intentional
manner,
but
yeah.
B
That's
kind
of
what
we
went
over
we
talked
a
little
bit
about,
could
build
a
declarative
pattern
as
well
and
what
that's
used
for
and,
of
course
we
have
the
docs
if
you're
tuning
in
at
a
later
date.
B
So
there
was
an
agenda
item
a
quite
interesting
one
for
talking
about
coup
builder
declarative
patterns
use
in
cluster
api
provider,
nested,
which
is
a
virtual
cluster
implementation
that
chris
hein
and
others
from
the
cappy
project
are
working
on,
but
chris
was
does
not
seem
to
be
able
to
join
today
it
is
thanksgiving
week,
so
we
might
just
push
that
to
next
meeting,
but
daniel
did
bring
up
that.
B
Maybe
we
should
talk
about
evan
evan's
cap
regarding
manifest
bundles,
and
I
do
think
that
this
is
an
important
item
for
for
us
to
address.
Justin.
You
mentioned
that
you
would
try
to
follow
up
with
evan
to
see
what
we
could
do
to
get
it
merged,
since
the
feedback
seems
to
be
pretty
complete.
There.
C
Yeah,
I
will
I
we
had
some
final
items.
There
was
a
little
bit
of
back
and
forth
on
discussion,
but
we
weren't
sure
whether
anything,
whether
everyone
wanted
to
put
anything
else
into
the
cap
itself.
It
does
seem
that
the
cap
is
in
a
perfectly.
C
B
Yeah,
I'm
feeling
pretty
good
about
the
contents
of
this
cap
and
I
and
I
do
think
that
it's
something
that
needs
to
be
addressed
in
the
community.
I
don't
have
any
contention
against
it,
so
I'll
I'll
say
so.
I
guess
that'll
just
make
it
easier
to
merge.
B
Well,
well,
those
were
our
primary
agenda
items
hopefully
we'll
be
able
to
get
to
this
cluster
api
provider
nested
topic-
and
I
guess
at
our
next
meeting,
which
would
be
the
8th
of
december
or
so
yes,
yeah
or
maybe
chris
can
come
by
in
the
slack
channel
again,
and
we
can
talk
more
on
it,
but
yeah
that
work
is
really
cool
and
we
should
you
know
if
you're
interested
tuning
in
later
or
if
you're
here
now
and
want
to
take
a
look
at
what
they
would
be
trying
to
use.
B
B
You
know
beforehand
in
a
declarative
manner,
but
in
this
case,
where
we're
actually
talking
about
scheduling
like
nested
api
servers
to
cluster
managing
crypto,
making
sure
that
things
are
hooked
up
together,
that
they
live
in
a
name
space
and
that
there's
an
easy
attraction
for
somebody
to
control
the
declarative.
Runtime
of
this.
That
you
can
upgrade
the
control
plane
like
there's
a
whole
slew
of
things
that
that
need
to
be
dealt
with
there
and
it
requires
packaging.
Some
manifests
into
the
operator.
B
It's
a
perfect
use
case
for
exactly
that
kind
of
thing
go
and
take
a
look
at
that
at
the
work
they're
doing
there.
Yeah.
C
I
just
say
like
I
don't
I,
I
didn't
really
understand
the
detail
of
the
question,
but
it
certainly
sounds
well
within
the
the
scope
of
it.
It's
certainly
it's
not
a
it's,
not
technically
a
cluster
add-on,
but
it's
as
you
say,
like
it's
exactly
the
same
thing
just
that
it's
actually
doing
like,
I
guess
a
different
cluster,
but.
B
They
would
be
using
the
cluster
custom
resource
from
their
project,
but
then
the
operator
would
be
feeding
that
to
actually
template
the
add-on.
You
know
and
then
manage
the
state
of
it,
and
maybe
there
are
some
of
the
shared
fields
you
know
from
that
that
we
might
want
to
have
inside
of
the
cluster
object.
C
Exactly
yes
like
it
would
be
wonderful
to
er
to
see
how
far
we
can
to
what
extent
we
can
reuse
the
the
patterns
from
add-ons
for
what
is
a
related,
but
slightly
different
use
case,
and
I
think
it's.
The
same
applies
to
joe
your
point
earlier
like
to
what
extent
sure
it
might
not
be
exactly
the
same
thing.
But
can
you
actually
sort
of
stretch
the
definition
a
little
bit
like
and
and
have
it
work.
B
B
Libraries
to
do
the
same
thing,
because
all
all
oh
yeah
yeah,
so
then
the
then
kudo
would
need
to
provide
a
way
to
generate
that
code
and
then
you
would
declare
it
from
yaml
and
say
hey.
I
want
to
use
this
library.
B
B
Yeah,
it
was
to
your
to
your
point
justin
about
like
not
really
knowing
what
the
question
is
that
chris
was
bringing
up
and
just
to
provide
a
little
bit
of
context.
Chris
did
dme,
he
was
wondering
like
hey.
Do
you
think
the
cluster
adams
group
is
healthy?
Do
you
think
that
what
we're
producing
like
has
longevity,
do
you
think
these
patterns
are
good?
B
Is
this
something
that
we
can
adopt
and
rely
on
continue
working
with,
and
I
I
felt
very
good
about
it
yeah
I
said
yeah
we
should,
you
know
totally
be
able
to
have
a
healthy
maintenance
relationship
if
you
decide
to
use
coup
builder,
declarative
pattern
and
the
kind
of
synergy
with
the
rest
of
the
ecosystem,
that's
being
built
inside
of
cluster
api.
C
I'll
feel
happier
once
we
get
it
integrated
into
cops
on
kubaydm
but
and
cluster
api,
but
I
I
think
we're
we're
making
good
progress
at
the
moment.
So
I
think
that's
it's
good.
B
Yeah,
we
could
probably
talk
with
rusty
and
fabricio
again
I
wasn't
able
to
ever
get
an
invite
to
their
kubernetes
next,
like
refactoring
session.
I
don't
know
if
that
ended
up
happening,
but
maybe
I'll
go
check
in
with
them
again
and
see
what
the
results
were
of
them.
Thinking
about
and
using
more
modern
patterns
to
kind
of
rebuild
parts
of.
C
Oh,
I
I
just
remembered
I
I
sent
a
sorry.
C
I
was
like
I
sent
a
pr
that
is
sort
of
odd,
because
it's
different
one
of
the
things
that
came
up
a
bunch
was
being
able
to
take
a
a
manifest
and
refactor
it
into
like
our
back
and
non
our
back
and
make
changes
to
it
like
that's,
based
on
what
sometimes
she
did
in
google
summer
of
code
and
the
idea
that
we
want
to
do
more
of
these
and
sort
of,
I
started
to
think
about
how
we're
gonna
actually
maintain
those
manifests
going
forwards.
C
We've
certainly
had
a
bit
of
a
challenge
with
doing
it
in
in
cops
where
we
can
check
in
the
mutated
version,
but
then
we're
never
quite
sure
why
things
changed,
and
so
I
sent
a
first
work
in
progress.
Well,
just
mvpr,
just
to
sort
of
start,
creating
a
pipeline
that
can
mutate
yaml
to
so
that
we
can
take
an
upstream
one
and
run
it
through
a
series
of
transforms
like
adding
labels
or
removing
labels,
or
things
like
that
and
produce
the
manifest
that
we
actually
splitting
it
out.
C
All
of
those
things
are
sort
of.
In
my
mind,
I
think
the
only
thing
it
does
right
now
is
removing
a
label,
I
believe,
and
but
I
think
that's
maybe
an
interesting
one
to
think
about
like
do.
We
want
to
proceed
in
that
direction.
To
try
to
create
the
manifests
from
upstream
manifests
until
we
can
eventually
get
people
building
cluster
add-ons
by
default,
which
will
happen
one
day,
but
until
then
we're
gonna
have
to
translate
from
helm,
charts
or
yaml
files
or
whatever
it
is
into
a
more
amenable
format.
B
I
think
I
just
linked
it.
It's
poll
89
right,
your
mvp
for
yaml
transformation.
Helper!
That's
the
one!
Thank
you!
B
B
There
we
go
so
yeah.
I
should
be
able
to
see
this
patch
that
I
have
here
just
justin's,
mentioning
that
we
need
to
manipulate
ammo
bundles
from
various
sources
right.
So
you
were
talking
about
helm,
charts
and
customization
directories
and
things
like
that.
C
Yeah
I
mean
it's
whatever
we
eventually
get
from
the
upstream
add-ons
projects,
whatever
their
preferred
means
is
so
the
hope
is
that
we
can
get
some
form
of
raw
yaml
out
from
whatever
pipeline
they
use
and
then
we're
going
to
change
it
in
some
way.
C
And
it
it's
a
little
verbose
because
it
reuses
a
library
in
customize
called
k-yaml,
which
has
the
very
nice
feature
that
preserves
comments.
So
if
you
do
have
comments,
it
won't
well,
then
maybe
you'll.
C
Exactly
right
it
does,
it
does
have
the
downside
that,
like,
for
example,
removing
a
label
is
a
little
bit
more
verbose
than
you
might
imagine,
because
so,
if
you
have
a
look
at
field
underscore
clearer
field,
clear
func
there
it
is,
you
can
see.
This
is
the
sort
of
you
sort
of
have
to
loop
over
the
a
tree
of
the
yaml
nodes.
You
have
to
basically
work
in
the
yaml
nodes
and.
B
C
You
do,
but
the
there
are.
There
are
layers
that
are
then
built
on
this.
That
are
much
more
palatable.
So
if
you
look
at
like
remove
dot,
go,
for
example,
you'll
see
that
we
end
up
in
a
much
more
like.
There
are
some
nasty
things
at
the
bottom,
which
hopefully
you
don't
have
to
look
at
too
much,
and
then
there
are
some
nicer
things
at
the
top
field.
B
Paths
yeah:
we
have
a
transform
object
from
somewhere
that
then
parses
the
field
paths,
and
it
says
that
in
these
fields,
which
are
okay
like
dot
notation
for
so
here
you're
even
doing
nested
things,
then
you
you
use
a
field
clearer
and
then
there's
a
predicate
function
that
you
put
in
that
says
like.
Oh,
if
it
matches
something,
is
that
how
that
works?
Exactly
right?
It
looks
at
the
keys
of
the
labels
and
yeah.
B
So
if
you
put
that
in
yeah,
the
predicate
is
looking
at
the
key-
and
this
is
the
remove
label-
filter,
okay,
cool
yeah,
that's
cool
anyway.
This
kind
of
stuff
is
exactly
why
we
have
this
project
so
that
this
code
doesn't
go
die
in
a
hole
somewhere
after
you
used
it.
For
one
thing
exactly
so,
thanks
for
that,
yeah
kmo
is,
is
a
pretty
useful
library,
but
definitely
a
very
different
strategy
to
working
with
kubernetes
resources.
B
Sweet
yeah
we
have
well,
I
guess
not
some.
Some.
Quite
a
few
people
join
this
meeting
anyway.
You're.
Would
you
be
looking
for
a
review
from
somebody
in
particular
on
this
justin?
Sorry?
I
didn't
know
about
this
patch.
C
Yeah,
I
mean,
I
think,
let's
start
off
with
feedback
on
the
on
the
idea,
whether
it
belongs
here,
whether
there's
some
other
tool
we
could
use.
I
do
think
it's
impo.
I
do
think
it'll
be
possible
to
embed
this
into
customize,
and
so
we
can
create
a
customization,
customization
yaml
to
describe
this.
C
It
may
also
be
possible
to
embed
it
into
the
kpt
tool,
which
would
also
work
the
same
way,
but
I'm
sort
of
starting
with
the
simplest
thing,
which
is
just
the
a
cli
pipeline,
but
we'll
have
a
bash,
so
we
would
use
this
using
a
sort
of
bash
series
of
pipes.
I
think
we
can
get
that
into
customize,
but
if
there's
some
other
way
that
we
should
be
looking
at
instead
like
that
would
be
super
helpful.
And
then
we
before
we
like
dig
into
the
the
code
details
right.
B
I
do
wonder
if
something
like
remove
go
could
be
feasibly
done
with
just
like
a
jq
program.
C
This
one
isn't,
but
I
I
have
I
have
more
ones
where
I've
done
things
like
remove
volumes,
so
you
can
certainly
ex
I'm
sure
we
can
express
it
as
jq
or
yq
or
whatever
it
might
be.
I,
the
the
real
hope,
is
that
we
build
things
that
are
actually
much
more
aware
of
of
kubernetes
kubernetes
in
general,
yeah.
B
My
favorite
jq
derivative
is
this:
by
by
itchy
ny
go
jq,
it's
a
pure
go
implementation
that
has
the
same
command
line
interface
as
jq,
but
it
also
supports
parsing
the
ammo
and
outputting
to
yeml.
B
That
is
a
good
feature
yeah
and
it's
a
single
statically
compiled
binary
instead
of
like
trying
to
find
an
installation
of
jq,
which
seems
to
sometimes
be
out
of
date
on
distress
and
cool
sweet.
Well,
let's
get
some
feedback
on
this
and
we'll.
Let's
bring
this
up
again
if
it's
still
open
in
two
weeks,
because
we
want
to
make
some
progress
on
building
a
little
bit
of
ecosystem
around
packaging
for
the
operators.
C
B
Yeah
and
certainly
there's
an
entirely
like
there's,
there's
a
whole
kind
of
use
case
or
persona
or
business
driver
or
whatever
you
want
to
kind
of
phrase
it
as
which
is
that,
like
these,
these
tools
can
be
really
hard
to
build.
They
come
with
overhead
and
complexity,
but
if
you
are
a
vendor,
you
know
or
a
project,
that's
trying
to
package
your
tooling
for
somebody
who
you
know
maybe
is
low
expertise
in
just
getting
started
with
kubernetes.
B
Then
the
operator
can
help
you,
you
know,
perform
upgrades
and
things
like
that
in
a
way
where
you
can
do
the
engineering
effort
and
it
reduces
the
burden
on
whoever's
kind
of
consuming
your
tool,
and
that's
just
going
to
be
an
ever-growing
ever-growing
market.
B
Great
thanks
for
submitting
that
patch
does
anybody
else
have
well?
Does
anyone
else
want
to
comment
on
that
in
particular,
if
not
suggest
your
own
thing,
otherwise,
you
could
probably
end
the.
B
B
We
some
time
what
is
this,
I'm
just
gonna
link
something
related
that
I
built.
Oh,
I
should
probably
put
that
in
the
meeting
agenda
here
we
go.
B
You
know
cncf
project
that
we
can
use,
or
it's
an
example
of
of
how
these
things
can
compose
well
together
for
users,
so
whatever
declarative
add-on,
deck
definitions,
we
have
whatever
packages
and
images
that
we're
shipping.
B
If
we
don't
use
the
application
things
that
are
built
directly
into
cluster
api,
you
can
still
integrate
with
them
around
the
artifacts
that
cluster
api
produces
like
the
config
secrets
and
things.
So,
if
that's
a
curiosity,
please
feel
free
to
fork
the
demo
and
try
it
out
yourself
and
sweet
well
that
links
in
the
agenda
document
as
well.
It's
on
repo
called
stealthy
box,
slash,
cappy,
hyphen
flux,
hyphen
demo.
B
I
think
we're
at
the
end
of
our
agenda.
So
thanks
so
much
for
joining
for
the
thanksgiving
week,
and
I
look
forward
to
seeing
all
of
you
as
well
as
hearing
more
from
chris,
the
next
time
that
we
chat
about
the
nested
provider,
work
that
they're
doing
for
virtual
clusters
and
how
that
can
relate
with
builder
declarative
pattern.
Let's
also
follow
up
on
evan's
cap
everybody
stay
safe
out
there
and
see
you
later.