►
From YouTube: Kubernetes SIG Service Catalog 20170919
Description
- F2F in Raleigh in November?
- One image vs multiple images
- 0.0.21 release tonight
- K8S names of services and plans
A
B
A
So
there's
going
to
be
an
open,
serviceworker,
API
working
group
face
to
face
meeting
in
Raleigh,
North,
Carolina,
November,
28th
and
29th.
That's
Tuesday
and
Wednesday
seems
like
a
good
opportunity,
since
we
have
a
heavy
overlap
with
that
group
with
folks
that
are
in
the
sig
Red
Hat
it
looks
like
is
gonna
host
it,
so
I
think
I
should
be
able
to
get
additional
space
for
another
day,
either
Monday
or
Thursday.
I
wonder
if
people
are
interested
in
doing
another
face
to
face
I
think
it
would
be
a
great
idea.
B
A
Hearing
that
Doug
is
gonna
make
sure
that
dinner
is
financed.
Somehow,
so
you
want
to
eat
on
money
that
comes
from
Doug
join
us
no
need
for
consensus.
Now.
We
should
probably
do
the
doodle
thing
and
send
a
message
to
the
mailing
list.
I
guess
I
can
take
that
as
an
action
item
ruffery
to
the
room,
Michael
QA.
B
Yeah
so
my
hands
up
so
there's
someone
in
there
I,
don't
know
if
that
was
you
Paul,
that
I
said
not
just
Monday
or
Thursday,
but
maybe
Monday
and
Thursday.
That
might
be
worthy
of
doing
a
doodle
poll
as
well
to
see
what
people
wanted
me
to
days
or
just
the
one.
B
B
A
A
E
B
Okay,
so
I'm
next
in
a
queue
so
I
I
actually
like
the
idea.
It
reminds
me
what
they
did
with
hypercube
and
I.
Think
that
was
probably
a
nice
way
to
go.
I
think
it
makes
it
a
little
bit
easier
for
programmers
of
this
stuff,
mainly
because
what
Paul
said
less
of
a
chance
of
things
getting
out
of
sync,
and
if
we
can
collapse
it
down
to
one
binary,
I
think
that'll
be
helpful
as
well.
B
Anybody
else
I'll
be
some
comments.
D
A
If
we
put
them
into
a
single
image,
and
especially
into
a
single
binary,
you
wouldn't
have
to
run
them
in
the
same
container
you
could.
The
idea
would
be
to
keep
basically
everything
about
our
current
chart
setup
the
same
except
use
the
same
image
and
either
different
binaries
in
the
same
image
or
an
all
in
one
binary
like
hypercube,
that
you
could
pass
an
argument
to
that
was
like
hyper
catalog,
API
server,
bla
hyper
catalog.
This
is
not
my
suggestion
for
a
name,
hyper
catalog
controller
manager,
wah.
B
A
In
ode
to
dot
or
ODOT
ODOT
20,
there
was
a
mist
feature
gate
and
a
controller
manager.
I
have
folks
that
are
blocked
on
this,
because
you
currently
can't
turn
it
on.
As
far
as
we
know
in
the
controller
manager
and
the
I
wonder
if
people
are
ok,
cutting
Godot
21
tonight
to
unblock
several
teams
in
Red
Hat.
A
A
Morgan,
would
you
open
up
the
issue
for
names
and
mosey
on
down
to
the
bottom
of
that
thing?
All
right
there
we
go
so
I
have
talked
a
little
bit
about
this
with
Doug
in
Vejle
yesterday,
so
we
came
to
and
I
also
spoke
with,
Clayton
Coleman
last
week,
who,
if
people
don't
know
who
he
is,
he's
one
of
the
owners
of
kubernetes
and
spent
quite
a
bit
of
time.
A
Over
the
last
week
talking
and
thinking
about
this
stuff,
we
have
a
a
proposal
that
this
is
basically
the
synthesis
of
ideas
from
myself,
Doug
Clayton,
VLA
and
here's.
How
it
goes
is
that
the
TLDR
and
I
will
explain
in
detail
in
a
minute,
but
first
the
TLDR
is
that
in
service,
instant
spec
users
will
use
the
human
readable
names
of
service
classes
and
plans
to
specify
which
service
and
plan
they're
on
the
controller
will
write
into
another
set
of
fields
called
service
class,
ref
and
service
plan.
A
A
All
right,
so
the
exhaustive
list
is
that,
and
there
will
be
some
repetition
here:
we're
gonna
use
the
open
service
broker
ID
for
kubernetes
names.
For
now
we're
going
to
add
an
external
name
field
to
the
service
class
spec
and
service
plan
spec
after
we
move
what
is
in
those
resources
in
morgan's
pour
requests
11:06
into
spec,
instead
of
just
being
exploded
in
the
in
the
resource,
we're
gonna,
add
external
name,
and
that
will
be
the
human
readable
name.
A
We're
gonna
keep
external
ID,
so
there's
going
to
be
external
name
and
external
ID
users
will
specify
the
human
readable
names
via
these
I
called
them.
Osp
service,
name
and
OSP
plan
name
fields
in
the
service
instance
spec
they're,
going
to
be
new
fields
added
to
the
service
instance,
spec
for
service
class
RAF
and
service
plan
RAF,
and
the
controller
is
going
to
maintain
those
fields.
A
There's
precedent
for
this
pattern
in
a
couple
different
ways
in
the
ways
that
persistent
volumes
and
persistent
volume
claims
work,
I'm
happy
to
explain
those
if
people
aren't
aware
of
them,
but
the
TLDR
is
that
there
are
several
different
ways
that
you
can
specify
criteria
and
a
persistent
volume
claim,
which
is
how
you
use
a
persistent
volume
in
Cooper
Nettie's
in
your
namespace,
and
the
long
and
short
of
it
is
no
matter.
What
way
you
pick.
The
persistent
volume
controller
writes
the
persistent
the
name
of
the
persistent
volume
into
another
field
in
the
spec
and.
A
This
is
basically
the
same
pattern
so
to
talk
through
some
examples
say
that
I
I
make
a
new
instance
of
a
my
sequel
service.
With
the
tin
plan,
I
would
put
in
OSP
service
name,
my
sequel,
I
would
put
in
OSP
plan
name
10
and
I
submit
that
I
push
that
to
the
API
server
the
controller
when
it
processes.
This
is
like
hey
the
service
class
raft
and
service
plan.
A
Raf
are
blank
I'm
going
to
look
these
I'm
going
to
look
at
these
services
and
plans
up
and
I'm
going
to
set
those
references
and
I'm
gonna
update
the
object
and
by
the
way
at
one
thing,
I
missed
going
through
this
list.
Is
external
name?
Is
a
selectable
field
on
service
class
and
service
plan,
meaning
that
I
can
do
a
list
request
with
a
field
selector
that
says,
give
me
the
services
that
match
the
service
classes.
With
this
external
name,
give
me
the
plans
with
this
external
name.
A
A
B
A
E
How
about
now
that
better
yeah
it
seems
like
it's
a
pretty
big
dock
and
you
know
it's
over
there
and
we've
got
a
year
of
meetings
in
it.
So
maybe
maybe
we
should
either
shift
all
the
current
contents
into
a
you
know
an
archival
dock
and
then
you
know
have
a
pointer
or
shove
it
all
in
yet
somehow
or
something.
But
you
know
it's
a
big
dock.
That's
all
I'm,
saying
yeah.
D
A
D
B
Okay,
next
I
guess
I
should
ask
any
other
questions.
Comments
on
this
topic,
all
right,
Morgan
next
up
table
of
contents.
Is
this
still
an
issue?
I,
don't
know
you
want
to
talk
about
it
or
you
just
want
to.
If
you
know
everything
I'm
just
wondering
given
the
previous
discussion,
is
this
still
a
desire
of
yours,
I.
E
Mean
you
know
it's
nice
to
have
contents
know
how
what's
in
there,
that's
all
well
going
to
go
ahead
and
talk
about
the
issue.
You
get
a
nice
pretty
table
of
contents
at
the
top
and
I.
Don't
know
if
you
need
to
have
the
recording
part
in
there,
but
I
think
going
forward.
It
could
help
to
look
at
the
I'm
gonna
go
back
to
the
patient
place.
You
can
go
back
to
the
place
really
but
I
guess
if
it's
a
quarterly
thing
that
maybe
it'll
only
be
an
order,
we
had
70
pages.