►
From YouTube: SIG Cluster Lifecycle - Cluster Addons 20210511
A
B
It
could
be
yeah
hey,
so
I
guess
I'm
on
the
new
one.
So
I
my
name
is
dave
gordon.
I
joined
red
hat
about
nine
weeks
ago,
so
I'm
still
pretty
new
to
this
role,
but
I
am
the
olm
development
manager.
So
it's
nice
to
be
here
and
just
I
look
forward
to
working
with
all
of
you
super
welcome.
C
I
think
I'm
in
the
new
face,
the
other
one
welcome
dave.
My
name
is
camilo
macedo,
I'm
senior
software
engineer.
I
work
now
in
pe
team
with
the
project
portfolio.
You
know
predictor
framework
work
as
well,
so
I
we
presented
a
cheese
item
to
subject
in
the
operator
framework
meeting
and
today
the
people
told
you
that
should
be
nice
to
share
here
and
to
that
is
because
of
this
I
put
it
inside
in
the
agenda.
C
A
Oh,
let
me
see
or.
C
Yeah,
so
the
idea
is
just
to
share
cheese
documentation.
C
We
provide
a
few
documentations
in
kobe
builder,
repository
kubernetes,
documentation
and
sgk
as
well
about
it
how
the
plugins,
how
the
architecture
of
the
projects
work
and
how
others,
from
inside
of
the
red,
hatch
or
outside
of
your
head,
that
you'd
like
to
create
plugins,
create
solutions
such
as
we
have
to
scout
for
the
helm
or
acebo
projects,
or
now
the
java
one.
How
the
people
can
create
this
kind
of
solution
in
a
way
follow
the
same
standards,
the
the
same
design.
A
C
He
just
to
bring
a
little
bit
of
context
is
because
the
shgk
is
kobe
builder
are
now
integrated
projects,
so
they
discovered
the
same
things
more
or
less.
Of
course,
sdk
is
called
things
beyond
that,
but
they
are
integrated.
They
work
in
the
same
way
and
they
work
with
the
same
plugin
system.
So
by
adopting
city
standards
by
collaborating
refugees,
we
are
also
improve
and
to
take
advantage
of
the
community.
C
A
lot
of
people
came.
If
everybody
follow
the
thing
we
can
make
you
one
solution
from
one
to
working
another
tool.
Things
like
that.
I
hope
that
it
makes
sense
so
to
clarify
that
I
put
you
here
the
links.
Why
use
the
copy
builder
style?
That
is
the
style
that
is
used
by
sdk.
C
C
How
it
works
the
plugin,
how
shgk
works
for
the
language
basic
plugins
and
what
is
the
plugins
vision,
so
one
of
the
ideas
that
we
have
for
the
future
would
be
that
the
key
would
be
able
to
discover
the
plugins.
So
now,
if
you
want
to
have
it
the
java
plugin,
for
example,
we
need
to
put
youtube
solution
inside
your
sdg,
but
the
idea
in
the
future.
C
The
vision
that
we
have-
we
don't
know
if
it
is
your
happiness
or
not,
but
it
is
has
been
discussed,
is
one
one
thing
that
is
desirable
would
be
the
the
to
be
able
to
look
okay.
I
have
a
cheeseburger.
C
This
is
outside
the
plugin,
but
I
know
how
this
works
things
I
mean
able
to
use
that,
so
we
could
allow
external,
the
sdk
the
tools
work
with
external
solutions,
which
would
be
very
nice
so,
but
for
all
this
happens,
if
he,
if
the
teams,
if
he
who
is
creating
this
kind
of
solution,
follow
the
same
standards,
do
the
things
in
the
same
way
that
you
make
easier
if
you're
weaking
provided
solution,
probably
you'll
make
it
easier
make
the
things
integrated.
C
Otherwise,
it's
very
hard
because
we
can,
if
you,
for
example,
I
create
a
project
that
is
called
fuji
python.
You
is
called
a
python
predictor,
for
example,
and
you
don't
follow
these
standards.
C
We
sgk
has
a
plugin,
a
solution
that
you
do
everything
that
is
required
for
ansible
for
helm
and
for
google
now,
so
we
don't
need
to
create
a
plugin,
for
example,
apple
game
for
help
engine
plugin
for
other
language
to
achieve
the
same
goal.
This
is
easy
to
keep
him
maintained
since
he's
easy
to
improve
and
do
all
this
stuff.
C
So
because
of
this,
I
think
he
would
be
nice
shared
his
documents
and
also
his
ideas
as
well.
C
Also,
if
anyone
anyone
has
doubts,
I
mean
the
quarter
has
slack
and
also
in
the
kubernetes
ones.
My
nickname
is
c
macedo
m-a-c-I-d-o
fell,
free
to
ping
me
feel
free
to
reach
me
out.
I
can
help
you
understand
how
the
things
work,
so
I
can
help
if
you
have
doubts
how
you
could
develop
this
solution
as
well.
Anything
that
I
can
help
just
please
don't
hesitate.
E
Thank
you
kamilah,
like
I'll
I'll
jump
in.
I
think
this
is
really
great.
I
think
you
know
the
the
effort
to
do
java
is.
E
It
will
will
certainly
help
a
a
big
tranche
of
people
who
want
who
prefer
that
language
for
personal
preference
reasons-
and
I
think
you
know
it-
opens
the
gates
to
other
things
as
well,
and
I
think
I
think
this
is
really
exciting
to
see,
and
I
I
personally
am
not
sure
whether
it's
worth
the
effort
to
have
coup
builder
discover
plug-ins
versus
just
having
like
a
very
thinner
like
a
very
thin
shim,
but
that's
that
is
based
on
my
experience
of
like
the
initial
work
that
I
some
of
the
initial
work
I
did
to
to
do
like
the
declarative
plugin,
but
that
is
that
is
I'm
happy
to
you
know
the.
C
Idea
would
be
like,
for
example,
we
have
one
solution
that
it's
nice,
but
it
for
us,
for
kobe,
builder
or
for
sdk,
is
not
invaluable
to
keep
maintained,
but
would
you
like
to
provide
that
facility
for
the
users,
so
I
don't
know,
for
example,
integrated
we
see
any
other
thing
or
or
the
declarative,
for
example,
or
it's
called
fuji,
something
that
is
customized.
C
That
is
some
solution.
That
is
very
specific,
so
we
don't
want
to
keep
mounting,
but
we
would
like
to
provide
you
allow
the
tool
works
with
that,
so
people
could
create
different
solutions
like,
for
example,
downstream.
Downswing
is
a
great
example.
We
don't
want
to
put
something
that
is
very
specific
for
openshift
in
the
upstream
solution,
but
would
you
like
to
provide
solutions
that
you
help?
Who
would
you
like
to
do
things
that
are
specific
for
open
shift?
C
So,
like
a
I,
would
you
like
to
ease,
call
folds
something
specific
image
for
open
ships,
so
they
certified
criterias?
What
is
required
for
the
certified
criterias,
which
is
not
required
for
the
community
operators,
is
not
required
for
kubernetes,
for
example.
Then
we
could
use,
it
is
the
stereo
plugin
that
should
be
supported
only
downstream
and
do
not
ship
in
the
upstream
solution.
E
Yeah,
I
I
it's
a
it's
a
it's
a
it's,
a
noble
effort
and
I
think
it
definitely
has
value
it's
just.
I
just
worry
whether
it's
hard,
but
it
sounds
like
you
have
a
good
grasp
on
it.
So
so
that's
that's
good.
C
If
you,
if
you
will
check
the
the
issues
in
the
shk
repository,
for
example,
a
lot
of
people
ask
a
lot
of
things
really
different
things
like
I
can.
I
can
try
a
few
we
should
try
to
after
share
some
issues,
some
requests
that
are
very
specific.
I
would
like
to
integrate
it
with
you
cheese
framework,
for
example.
I
would
like
to
use
the
cheese
too.
I
just
like
to
use
the
cheese
vendor
so
are
very
specific
and
unique
desires.
C
C
The
people
have
a
lot
of
effort
doing
something,
but
today
we
need
to
have
a
lot
of
effort
again
to
make
the
java
solution
be
aligned,
as
we
fest
educate
so
make
that
work
together
and
see
how
we
can
keep
dutch
maintaining
easily
how
we
can
make
the
java
solution
work
with
the
generic
plugins.
That
are
the
things
that
are
provided
by
shgk
on
top,
for
example.
C
If
when
I
start
the
project
I
have,
it
is
knowledge
how
the
things
works.
I
could
take
the
decision.
You
know
I.
I
don't
want
to
follow
these
standards
because
if,
for
my
target
audience,
for
example,
it's
not
too
good
jesus,
you
take
the
decision
knowing
about
today.
Okay,
are
you
a
force
be
hard
to
maintain?
C
Because
for
me,
it's
important,
you
make
people
more
happy
in
this
way.
This
is
one
one
kind
of
the
decision,
but
he,
if
you
only
do
different,
for
example,
only
because
you
don't
know
that
you
have
has
a
standard
you
lose
the
advantage
should
take.
You
know
you
lose
the
opportunity
to
take
these
advantages,
because
after
you
did
the
solution,
you
need
to
spend
more
effort.
C
It's
it's
harder.
I
I,
in
my
humble
opinion
when
we
start
to
think
the
solution.
If
you
redesign
it's
working
this
way,
it's
one
effort
after
the
things
are
done
to
align
it
to
work
in
this
way,
it's
another
way
to
think,
and
maybe
maybe
you'll
not
be
valuable
anymore.
Therefore,
changing
we
can
lose
some
opportunities
here.
Take
advantage
of
the
community,
because
the
community
also
can
helps
like
java
java
developers
can
see
a
problem
ng.
C
That
is
following
the
standard
engineering
contributory
for
the
standards
with
the
the
leaving
kobe
builder,
that
use
it
by
sdk
providing
the
solution.
So
instead
you
often
solve
something
that
is
specific
for
java.
Then
you
solve
the
problem
that
she
you
help
everybody.
That
is
the
kind
of
advantage.
If
everybody
follow
up
these
standards.
C
C
F
Somewhat
philosophical,
I
guess
just
kind
of
wondering
somebody
brought
it
up
to
me
and
I
thought
it's
actually
worth
discussing
here.
F
F
I
mean
had
folks
seen
any
established
patterns
in
this
area.
Any
general
thoughts
on
this
just
thought.
That's
it's
kind
of
an
interesting
topic
to.
F
Discuss
I
have
like
very
little
thought
on
that,
but
yeah
I'll
bring
it
here
are.
F
Yeah
it
will
both
things
really
yeah,
because
it
becomes
our
operator's
responsibility
to
fix
everything,
and
if
you
know
if
previously,
software
was
managed
more
directly
by
a
user.
If
it's
an
established
project,
some
users
are
used
to
managing
in
a
certain
way.
They
have
learned
ways
to
troubleshoot
it
now
operator
comes
in.
It
has
to
take
the
responsibility
of
troubleshooting
everything
it
can
possibly
troubleshoot
or
otherwise
users
are
lost.
I
mean
that's,
that's
one.
One
line
of
thought.
C
For
you,
okay,
kubernetes
has
one
standard,
that's
one
convention.
Actually
that
is
the
status
conditionals.
That
is
for
your
map,
like,
for
example,
you
define
that
you
require
a
deployment,
your
service
and,
however,
you
know
something
doesn't
work.
Well,
so
you
you're
checking
this
yard.
They
start
as
conditional.
I
can.
I
can
try
check
the
link
here
and
share
with
you
ng
for
turbo
shoot
in
the
cluster.
Usually
I
I
use
the
logs,
so
your
whole
developer.
The
solution
came,
decides
the
level
logs
as
well.
C
So
you
can
configure
like
you.
Can
let
the
adaming
you
can
you
can
put
in
the
environment
variables
a
configuration
for
your
level
logs,
for
example,
like
a
debug
or,
however,
for
ansible
operators?
Scaffolding,
if
he
operator
suggests
his
configuration
will
be
in
place
because
this
I
escaped
about
the
language
will
be
dedicated.
C
Something
that
is
is
the
the
for
is
the
way
that
it
is
is
confused
by
the
foo.
So
you
can
put
like.
I
want
the
bug
logs
because,
usually
you
don't
you
want
a
lot
of
loads
because
you'll
be
hard,
but
if
you
have
a
problem,
you
probably
you
want
to
increase
the
level
of
the
log.
So
as
a
developer,
you
can
provide
this
feature
for
the
admins
in
your.
E
Yeah
I'll
also
mention
we
had
some
one
of
the
reasons
it
was
that
getting
off
adam
operators
into
chaos
was
delayed
was
because
we
had
some
feedback
from
people
who
were
used
to
the
old
way
of
doing
things
and
wanted
to
like
get
the
manifests
and
manage
them
manually
and
not
even
use
an
operator
or
see
what
those
things
were,
and
one
of
the
things
we're
working
on,
which
will
require
support
from
cluster
add-ons
is
the
ability
to
have
the
operator
run
like
in.
I
think
you
know.
E
We've
talked
about
this
before,
like
client-side
or
one-shot
mode,
where
it
outputs
a
manifest.
So
I
think
I
think
the
argument
of
supporting
that
or
for
supporting.
That
is
that,
if
you
we
can
still
package
things
as
operators,
even
if
users
prefer
to
run
things,
you
know
the
old
way
and
diagnose
them
manually.
E
I
think
we
should.
We
will
hopefully
get
to
a
point
where
the
operator
is
is
fairly
good
at
doing
the
simple
stuff.
I
don't
think
it
will
ever
be
able
to.
You
know
address
every
case
and
I
think
one
of
the
things
we
haven't
yet
done
is
like
the
pause
button
right,
where
you're
able
to
yeah
stop
an
operator
doing
things,
so
you
can
go
and
mess
around
and
see
if
you
can
get
it
working
again
and
that
sort
of
thing.
E
Presumably
be
that
sort
of
like
that
to
me
is
the
middle
ground
right,
that's
so,
on
the
one
extreme
you
have,
I
reject
all
notion
of
operators.
I
want
to
manifest
and
I
want
to
do
it
all
myself
and
we
can
support
that
with
an
operator.
E
On
the
other
hand,
we
have
the
operators
are
great
and
they
will,
like
you
know,
do
the
busy
work
for
me,
and
so
I
want
to
use
operators,
and
then
I
think
we
can.
We
can
also
support
the
the
middle
ground,
which
is
like
all
right.
The
operator
broke
in
a
way
that
I
wasn't
expecting
like,
but
I
was
I'm
using
the
operator.
Stop
the
operator.
Let
me
go
in
and
do
my
manual
remediation
as
I
would
do
and
then
once
either
an
update
of
the
operator
comes
along.
F
Okay,
I
see
yeah
and
is
there
something
there,
but
one
thing
that
that
that's
that's
quite
interesting,
justin,
the
the
the
other
thing
thing
I
was
wondering
about:
is
there
some
some
way
of
user
discovering
what
are
all
the
resources
that
an
operator
has
thrown
into
their
cluster
that
they
may
need
to
make
changes
to,
while
in
the
manual
override
mode
you
know
or
or
or
otherwise,
just
or
whether?
What
all
that
belongs
to
this
this
operator?
And
how
do
I
reason
about
it?
How
do
you
understand
it.
E
Yeah,
I
mean
very
much
so
like
there's,
there's
two
two
ways,
one
of
which
is
you
know
we
can
run
the
operator
in
client-side
one-shot
mode.
Presumably,
we
could
figure
out
a
way
to
run
it
in
client-side,
one-shot
mode,
without
even
touching
like
just
sort
of
get
the
cr
from
the
cluster
and
tell
me
what
should
be
there.
The
other
one
is
using
owner
riffs.
So
some
toshi,
who
was
oh
yeah,
it's
on
the
cool.
There
did
a
wrote,
a
tool
that
can
show
you.
E
I
think
one
level
of
owner
refs
the
the
objects
owned
by
by
a
cr.
It
would
be
nice
to
show
the
whole
tree,
but
you
know
we're
starting
at
the
beginning,
because
the
way
kubernetes
works,
it
is
not
the
fastest
tool
because
it
has
to
read
all
the
all
the
kinds.
But
it's
you
know
good
enough
for
diagnostic
purposes.
F
Yeah,
I
guess
I
guess.
I
wonder
whether
there
may
be
a
reason
to
debate
some
kind
of
some
kind
of
a
format
of
describing
like
the
kind
of
relationships
that
an
operator
defines
from
a
from
a
custom
resource
to
to
to
a
set
of.
C
Other
resources
you
have
the
watch
features
that
you
watch
the
resources
in
the
cluster
right.
The
goal
of
the
operator
is
you
ensure
a
stage
on
the
cluster
right,
so
you
have
one
feature
that
you
call
watts
what
it
is
does
observe
the
cluster.
So
you
can
define
what
you
observe
using
kubernetes.
You
work
with
labels.
You
use
the
labels,
you
make
annotations
in
the
resource,
so
you
can
have
fun
and
to
play
around
with
that
as
well.
F
Okay,
well,
I
guess
I
guess
I'm
just
sort
of
thinking
a
bit
more
broadly
right,
I
mean
yeah
and
that
this
this
is.
This
is
just
something
that's
been
brought
up
to
me
by
someone
else,
and
I
do
I
do
think
you
know.
I
didn't
necessarily
relate
this
to
some
of
the
discussions
that
we
previously
had
about
the
client
side,
the
one-shot
mode,
but
it
does
actually
relate
it
now
that
that
that
you
mentioned
justin,
but
in
a
way
yeah.
F
Just
sort
of
you
know
like
if,
if
I'm
authoring
an
operator
right
now,
I
have
to
sort
of
spell
out
everything
that
it
does
in
the
readme.
F
F
You
know
just.
I
just
wonder
what
that
that
that
could
be
part
of
some
kind
of
operator
definition
api,
because
there
are
things
that
exist
in
openshift
like
around
olm
and
such
things,
but
I
personally
with
my
recent
experience,
found
it
practically
for
very
limited
use
to
to
me.
As
an
author,
it
was
something
that
seems
to
be
useful
to
openshift,
but
I
I
couldn't
find
much
use
of
that.
F
G
G
To
see
where
you
felt
that
fell
down,
I
know
there's
a
lot
of
other
craft
around
it
that
we're
trying
to
sort
of
remove
in
newer
stuff,
and
I
think
the
direction
that
probably
operators
in
general
can
go
on
is
probably
more
of
an
associated
or
an
associative
relationship.
Rather
than
like
something
that
embeds,
let's
say
a
deployment
and
all
of
the
metadata
from
your
operator
on
a
cluster.
G
F
I'm
definitely
talking
to
something
different
than
csv
csv
lets
me
essentially
deploy
an
operator-shaped
thing
into
a
cluster,
and
you
know
tell
the
olm
operator
how
to
deploy
it
right.
What
I'm!
What
I'm
thinking
about
is
what
would
be
descriptive
to
the
user,
something
that
would
describe
the
function
of
the
operator
to
the
user
without
users.
Reading
the
code.
G
Yeah,
so
I
think
we're
trying
to
do
something
similar
in
that
situation,
where
you
basically
associate
arbitrary
traits
and
also
cluster
traits,
with
a
bundle
which
is
basically
like
all
the
stuff
that
goes
into
an
operator
and
then
that
stuff
could
be
tracked
at
runtime
on
a
cluster,
and
I
think,
we've
actually
pitched
something
like
this
in
add-ons
before.
F
Yeah
anyway,
I
just
thought
I
just
thought
it's,
it
seems
seems
sort
of
worth
discussing,
but
yeah
the
I
guess.
The
key
topic
is
how
do
how
to
address
the
transparency
of
what
does
this
operator
do?
F
What
should
I
expect
and
what
are
all
the
things
that
I
should
be
looking
at
when
I'm
using
it,
and
I
just
how
do
I
discover
those
things
yeah,
but
you
know
something
that
would
also
be
user
friendly
would
require
somebody
to
to
understand
a
complex
api
anyway,
that's
that's
just
an
idea,
but
like
yeah,
I
guess
broadly
we're
kind
of
saying
that
yeah,
the
the
client-side,
one-shot
mod
mode
and
the
the
sort
of
manual
override
modes
would
would
be
potentially
quite
helpful
for
some
of
the
challenges
that
that
users,
who
used
to
managing
things
more
directly,
have
but
possibly
even
other
users,
the
yeah
and
the
the
other
point
that
was
mentioned
is,
like
you
know,
reflecting
things
in
status
sounds
like
the.
F
E
Yeah,
I
think
I
mentioned
logs
and
events
as
well,
like
I
think,
conditions
I
think
are
great
for
structured
data
that
can
give
you
a
hint
as
to
where
to
look
and
then
yeah
like.
If
you
need
more
information
like
ideally
most
things
will
be
in
conditions,
but
you
know,
like
I
think
the
we
can
only
expose
in
a
structured
way
the
things
we
know
about
so
hopefully,
so
there
will
always
be
other
things
that
happen.
Yeah,
that's
always
going
to
be
in
the
logs.
C
But
I
I
totally
would
recommend
to
you,
like
you
work
with
levels
of
logs,
because
a
lot
of
the
ad
means
the
user
is
complaining.
This
operator
has
so
much
logs,
it's
confused
for
me
and
they
they
want
to
see
only
so
many
logs.
A
C
Humble
piano
is
the
best.
C
Andy
about
the
other
stuff
that
you
say
that
you
spoke
about,
she
managed
the
resource.
C
C
F
Well,
I
guess
I
guess
I'm
not
looking
for
a
specific
solution
here.
I
just
thought
I
would
to
to
bring
up
the
the
topic
in
general.
You
know,
I
think
I
think
we
will
we'll
be
coming
back
to
this
yeah.
E
I
think
it's
a,
I
think,
it's
a
very
fair
topic
to
bring
up.
I
think
I
think,
though,
if
we
can
get
to
a
point
where
we
we
shouldn't.
Therefore
we
shouldn't
reject
the
idea
of
add-on
operators
for
this.
What
we.
E
Can
we
make
add-on
operators
work
for
people
that
want
x
right,
and
we
in
the
chaos
project
think
that
we
can
get
an
operator
square
for
people
that
don't
want
to
use
add-on
operators?
In
other
words
they
want
to
package
it.
They
want
to
use
it
as
like
a
packaging
mechanism,
but
not
a
installation
mechanism.
They
want
someone
like
generate
the
manifest
and
effectively
apply
them,
and
there's
gonna
be
some
sugar
around
that
in
chaos.
E
But
you
know
like
that
is
a
pretty
extreme
point
of
view,
but
it's
if
we
can
support
that.
I
feel
like
we
are.
You
know,
that's
that's
my
that's.
My
sort
of
goal
is
to
make
it
so
that
when
people
come
with
these,
you
know
different
use
cases
that
we're
able
to
meet
them
with
the
same
tool
set
rather
than
having
to
create
a
new
tool
set
yeah.
F
Yeah
well,
thanks
alex
very
interesting
discussion,
I'll
think
more
about
this.
Of
course,.
C
F
I
guess
for
us
some
of
this
and
then
this
was
of
course
a
in
in
context
of
potentially
using
an
operator
to
install
celium
for
us
that
is
possibly
more
critical
than
certain
others,
certain
other
add-ons,
because,
as
a
is
a
component
that
provides
networking
functionality,
connectivity
to
the
entire
cluster
and
possibly,
of
course,
quite
possibly,
security
features
as
well.
If,
if
something
breaks
due
to
an
additional
layer
of
automation,.
F
It
may
have
rather
grave
consequences
for
functionality
of
the
rest
of
the
cluster.
So
you
know
all
sorts
of
things
do
need
to
be
considered.
F
It's
not
you
know
it's
not
an
add-on
that
provides
some
optional
or
some
of
the
auxiliary
functionality.
Whatever
you
want
to
call
it
it's
it's
kind
of
critical
to
the
operation
of
the
entire
cluster.
So
that's
that's
kind
of
one
of
the
the
the
reasons
various
angles
are
being
looked
at.
E
In
unrelated
notes,
for
you
elia
thank
you
for
enabling
wireguard
I'm
excited
to
try
that
out
have
another.
B
F
F
Thank
you.
Yeah
there
was
a
there
was
a.
There
was
an
echo
episode
with
liz
and
martinez.
Who
is
the
engineer
who
worked
on
berger
implementation,
which
which
should
be
on
youtube?
Echo
is
the
ebpf
psyllium
office
hours.