►
From YouTube: Kubernetes SIG Service Catalog 2018-06-04
A
A
A
Repository
that
is
basically
on
creating
a
sort
of
a
user.
However,
you
want
to
call
it
our
back
rule
for
Service
Catalog,
and
this
will
load.
We
could
see
it
basically,
what
we
have
now
is.
We
have
a
bunch
of
roles
we
have,
but
they
are
basically
all
for
the
sort
of
administration
and
running
of
Service
Catalog
and
it's
various
pieces,
and
this
sort
of
assumes
that
everybody
wonderful.
A
A
B
A
A
B
Hear
if
I
raise
my
hand,
okey-dokey
yeah
I,
like
I,
am
really
not
sure
what
the
best
thing
is
to
do
here.
I
think
the
the
motivations
for
having
the
imperative
flag
for
this
like
make
a
lot
of
sense.
I
I,
regret
that
the
semantics
of
graceful,
elation
and
finalization
seem
to
be
kind
of
problematic
for
implementing
the
things
that
we
need
to
with
them.
B
That,
in
the
future,
we
can
have
something
that
works
well
without
having
to
have
an
imperative
flag.
I,
don't
think
that
we
should
block
it
better
until
the
semantics
and
Kuebler
just
right,
because
that's
a
big
boulder
though
I'm
in
favor
I'm
in
favor
of
continued
analysis
on
this
and
I.
Think
that
Morgan
does
this?
Does
this
understanding
match
up?
Is
it
like
fundamentally
the
issue
around
the
motivation
for
this?
Is
that,
like
you
delete
one
thing
basically
can
update
you?
Can't
you
get
it
back
out
of
the
stable?
Really,
yes,.
A
That's
part
of
it.
Well,
that's
this
particular
piece.
Yes,
that
is
that
I've
written
up
a
couple
of
extra
pieces
in
a
repository
and
I've
been
asking.
You
know
various
API
experts
on
it
and
there's
one
thing
that
might
preserve
the
existing
semantics
but
would
still
require
some
kind
of
extra
field
like
this.
A
Is
there
and
I
had
this
discussion
with
somebody,
but
I,
don't
remember
if
there
was
details
and
then
when
I
tried
to
look
up
details,
I
couldn't
understand
how
it
would
use
this
to
actually
implement
it.
So
is
there
a
way
to
intercept
something
like
the
delete,
call
and
then
change
it
into
the
update
call
so.
B
Not
I,
don't
say:
there's
nobody
to
it.
Here's
the
piece
of
code
that
does
it
I
mean
that's,
that's
exactly
what
the
API
server
does
now,
or
resources
that
have
graceful
deletion.
Yeah
I
server
both
get
a
delete
request,
either
as
graceful
deletion
for
this
right
I'm
not
into
an
update.
That's
basically
what
you're
asking
about
right,
yeah.
A
B
A
B
Implements
those
behaviors
and
like
the
problem
with
like
going
and
hacking
through,
that
is
that
then
we
would
potentially
lose
all
these
other
like
very
critical
behaviors
and
write
exactly
I
guess.
What
did
you
have
in
mind?
Morgan
like
you'd,
want
to
take
a
deletion
and
set
the
imperative
field
instead
of,
like
the
deletion,
timestamp
right.
A
Do
something
like
that
and
then
the
end,
the
only
thing
the
only
wrong
with
that
is
I
kind
of
get
econ
again
into
the
situation
where
okay,
well,
fine,
the
controller,
eventually
files,
the
real
delete.
Is
there
a
way
to
bypass
the
original
delete?
I?
You
know
it
gets
confusing,
really
quick,
so
I
was
just
so
that
was
proposed
as
as
somewhat
of
a
solution.
I
just
I'm,
not
sure
how
the
actual
implementation
would
go
so.
B
Like
in
a
vacuum,
there's
probably
things
we
could
do,
I
mean
one
one
option
might
be
to
have
a
special
sub
resource
or
something
like
that
that
had
those
types
of
semantics
I
think
we
could
definitely
do
that
all
right.
The
other
thing
to
consider
about
this
is
that,
like,
basically,
if
anything,
we
add
in
the
flavor
or
in
the
form
of
custom,
some
resources
and
like
API
type
like
are
our
complications
for
ourselves.
B
B
I
think
we'll
always
have
to
implement
the
right.
It's
right
and
if
you
just
delete
it
like
I,
don't
think
there's
a
way
not
to
support
that,
but
if
there's
another
way
that
it
provides
a
better
UX
I
think
it
should
be
on
the
table
to
do
that
thing
and
have
that
be
the
way
to
have
it
work
best.
So
one
question
that
I
have
for
Scott
is
got.
Have
you
tried
to
do
any
custom,
some
resources
for
your
needs,
yet
in
your
work,
they're.
E
B
F
B
F
B
A
Okey-Dokey
cool
we'll
have
to
find
a
time
to
talk
about
this
and
find
people
interested
moving
on
I
made
the
I
actually
I
did
the
tag
and
release
of
0.1
21,
but
the
chart
jenkins
website
is
giving
me
a
502
and
thick
owns
the
website
and
he
is
away,
and
so
we
don't
have
a
1:21
chart,
and
I
just
wanted
to
inform
everybody
of
this,
because
I
felt
it
was
something
you
don't
know
about.
It.
I
asked
her
the
people
watching
the
video
any
commentary.
A
A
G
Ok,
so
what
I'm
going
to
show
today
is
re-implementation
pot
presets
using
CR
G
and
a
mutating
Web
book
I
linked
the
repo
in
the
agenda,
but
I'm,
assuming
that
people
have
not
it
so
I
will
quickly
run
this
deploy
script
and
then
explain
it.
While
it's
running
in
the
background
and
try
to
explain
things
best
I
can
the
CRD
was
actually
built
with
coop
builder.
So
that's
part
of
the
reason
why
it's
in
the
standalone
repo,
my
next
action
is
to
bring
it
into
the
Service
Catalog
repo,
but
because
of
the
dependencies.
G
G
G
G
So
the
first
pod
looks
like
this.
So
it's
using
this
mutate
true,
so
it
will
match
the
first
pod
preset
and
the
second
one
is
deployment
as
two
containers
and
is
using
this
mutate
all
label.
So
whenever
I
described
the
first
pod,
you
can
see
that
the
environment
variable
6,
3,
7
eyes,
can
be
there
and
then
see
the
other
one.
G
One
of
the
things
that
was
done
in
the
reconciler
loop
of
the
CRD
was
to
implement
bumping
the
pod
when
a
preset
that
matches
is
updated.
So
if
you
look
here
whenever
the
mutating
web
hook
makes
modification
to
the
spec,
it
adds
an
annotation
with
the
pod
preset
and
the
resource
version.
And
then,
whenever
the
reconciler
loop
and
the
CRT
detects
that
a
matching
preset
has
or
a
mutation
on
the
speck
has
occurred,
it
adds
this
bounce
annotation.
It
will
actually
bounce
the
deployment.