►
From YouTube: Kubernetes Community Meeting 20160428
Description
We have PUBLIC and RECORDED weekly video meetings every Thursday at 10am US Pacific Time.
https://docs.google.com/document/d/1VQDIAB0OqiSjIHI8AWMvSdceWhnz56jNpZrLs6o7NJY
Demo of KPM; Overview of Protobuf Work; 1.3 Key Features Update; SIG-Openstack update
A
Let's
go
hello
good
morning
all
good
day
all
given
that
there
are
many
time
zones
today
is
Thursday,
April
twenty-eighth
and
we
are
meeting
to
have
a
recorded
community
meeting
with
the
Cooper
Nettie's
community.
So
the
agenda
today
is
having
a
demo
of
KPM
a
package
manager
and
deployment
tool
for
Cooper
Nettie's
Clayton
Coleman
is
going
to
be
talking
about
the
proto
buffer
work,
that's
going
on
and
then
I
have
TJ
Coulter
men
set
up
to
talk
about
the
1.3
key
features
update,
and
then
we
have
an
open
spot
for
any
cig
business.
A
Nothing
exciting
scheduled
up
there
yet,
but
I
need
you
in
a
special
interest
groups
when
I
give
updates.
That
would
be
great
and,
lastly,
I
need
a
volunteer
note
taker
for
the
day.
I
know
I
ask
this
every
week,
but
that's
because
we
need
one
every
week.
So
if
somebody
could
volunteer
and
take
notes,
that
would
be
fantastic
and
we
will
start
with
Antoine
LeGrande,
showing
us
k
p.m.
are
you
ready,
Antoine?
Yes,.
A
B
Okay,
so
I
would
do
a
demo
of
K
p.m.
so
it's
absolutely
studied
that
three
months
ago
to
deploy
our
prediction
application.
So
we
now
put
it
open
source
so
because
we
think
we
can,
it
can
help
the
community
to
do
the
time.
So
it's
a
package
and
deploy
manager
as
I
say
it's
too
deep
operate
deployments
in
prediction
and
also
there
is
a
registry
to
to
share
quickly
some
feature,
and
that
was
important
for
us,
so
to
be
declarative
and
even
button.
B
So
we
want
to
be
able
to
deploy
multiple
time
as
an
acacia,
breaking
anything
also
the
fact
that
its
server
side,
so
everything
has
an
API
to
beat
the
back
edge
to
deploy,
and
so
it's
very
useful
to
integrate
our
KPM
to
a
dashboard,
are
continuous
integration
tool,
also,
of
course,
to
manage
end
version
of
our
resources.
Our
application
setup
crystal
once.
B
B
First
to
install
our
KPM,
it's
a
python
tool,
so
you
can
just
a
pipe
installed
KPM
together.
That's
all
it
takes.
Then
you
have
a
come
online
with
commands
cool,
so
one
of
them,
for
example,
is
kept
em
list
by
default.
We
all
stop
on
a
public
registry,
so
you
can
leave
a
package
from
that.
But
as
I
say,
you
can
connect
two
frigates
out
of
one:
that's.
What
are
you
just
put?
A
trace
option
creates
registry
okay.
B
So,
let's
see,
what's
like
it
looks
like
it's
pretty
simple:
there
is
it's
just
a
directory
with
a
manifesto
channel
and
then
template
directory
where
you
put
your
resources,
so,
for
example,
the
config
not
depend
on
service,
etc.
So,
as
I
said,
as
a
directory
suggest
the
name,
you
can
put
complete
some
values
and
then
you
put
the
g14
to
manifest.
B
So,
let's
see
what
so
many
face,
look
like
kids
are
so
quite
simple,
just
submitted
at
our
run
your
package,
so
the
name
of
the
package
version
which
is
important,
the
differed
by
adverse
of
short
on
page.
So
if
I
ever
can
include
another
variable
which
is
quite
suit,
or
in
this
case
another
part
is
just
the
list
of
your
resource.
That
is,
that
aizen
tompa
directory.
But
you
can
pattern
like
you,
get
stopped
them
say
which
order
you
want
them
to
be
deployed,
and
also,
you
can
add
information
to
this
resource.
B
For
example,
we
are,
we
have
a
key
protected
are
just
say
to
the
kept
an
account.
You
can
create
issues,
but
you
are
not
authorized
to
delete
it.
So
we
just
add
an
annotation
on
the
resource
to
trigger
that.
So
it's
very
useful
to
be
able
to
to
add
some
information
around
the
resources.
The
last
part
are
the
dependencies.
B
So,
for
example,
goes
depend
on
my
square
and
you
can
just
override
the
default.
Value
goes
quite
easily,
my
god.
So
when
your
package
is
ready
ready,
the
only
thing
you
have
to
do
is
to
get
them
to
the
cap
and
push
command
and
if
we
just
do
a
double
of
your
package
and
send
it
to
the
registry
of
your
choice
so
I've
released
on
that.
B
Ok,
we
have
a
package.
We
just
are
poor
so
now
to
deploy
it.
It's
very
simple:
you
just
keep
an
explosion
of
the
package
you
can
select
in
which
an
in
space
and
you
want
to
deploy
it
and
we
just
create
a
p.I
to
get
to
package
and
the
resources
to
deploy,
and
we
just
aren't
deploy
it
with
all
the
dependencies.
So
one
st.
is
a
for
example.
You're,
aware
of
our
current
status,
so
is
a
created.
That's
if
we
deploy
it
again,
it
would
just
detect
that
there
is
no
change
to
do
so.
B
Everything
is
grainy,
say:
ok,
I
done
nothing.
So
this
is
how
we
can
just
improve
iteration
of
your
application.
For
example,
if
you
change
in
value
going
to
manifest
they
say
the
best
earth
which
is
going
to
change
the
config
map.
Ok,
so
we
update
the
version.
We
push
it
again,
so
we
have
our
new
version
available,
so
everyone
can
deploy
it
and
if
we
deploy
it
again
in
the
same
namespace,
mexicanos
change
and
indigo
just
a
package,
we
have
a
change.
So
you
see
the
coming
back.
B
I've
been
updated,
so
you
can
very
easily
iterate
on
your
abs.
You
can
also
a
word
back
to
any
previous
state
with
division,
so
you
can
put
another
control
back
or
change
the
version
number
okay.
So
it's
very
easy
to
push.
But
what,
if
you
want
to
to
have
some
secret,
so
I
will
show
you
something
as
so.
I
created
and
such
one
private
ghost
package,
where
I
want
to
put
my
my
secret,
so
I
have
a
new
dress
with
my
in
grass
with
some
certificate
is
so
different.
B
You
don't
want
that
to
go
to
the
period
space
right,
so
I
created
package
for
that
so
adjust
the
resources.
The
secret
in
grass
is
any
depend
on
my
square
and
ghost
ok.
So
I
will
push
on
that
in
my
pre
rate
registry
yep
and
push
the
same
come
on.
You
just
just
select
the
register.
You
want-
and
I
have
my
private
like
edge
here,
so
I'm
going
to
try
to
deploy
it
in.
B
Ok,
so
what
is
going
to
happen?
It's
even
if
the
name
of
the
package
is
different,
ok,
so
the
and
such
one
parade
goes.
It
will
just
detect
that
the
resource
are
the
same.
So
we
do
nothing.
So
it's
not
about
the
package
name
that
I
can
collide
it
just
above
the
resources,
so
it's
pretty
safe
to
deploy
our
sings
with
KP
app.
So
also,
if
you
notice,
so
we
don't
stall
the
ghosts.
Epic
down
the
ghost
package
interpreter
history,
so
we
have
an
ID
from
Brandon
free
from
Korres
to
implement
a
package
discovery.
B
So
oh,
it's
work.
You
just
say:
ok,
the
package
is
not
in
my
in
my
internal,
our
registry,
so
we
prefix
with
the
domain
name.
So
I
will
call
KPM
the
desert
to
ask
a
do.
You
know
where
the
ghost
package
can
be
found
and
if
there
is
a
reply
we
use
that
registry.
So
I
think
it's
very
own,
very
nice
way
to
connect
together
different
registry
and
use,
provide
and
tribute
once
that
something
else
can
add
two
to
manage
and
you
provide
and
debate.
Ok.
B
So
this
is
already
a
deploy,
but
this
is
a
lk
x
type,
so
I
use
a
rating
to
whether
as
a
broker.
I
do
you
need
a
sixer,
some
configuration,
some
diamond
set,
etc.
Cuba.
Now
so
you
see
it's,
you
can
just
manage
our
quite
complex
deployment
and
in
such
case,
for
example,
if
you
want
to
upgrade
on
these
Cuban
are
it
can
be
done
very
easily?
We
just
faced
say:
okay,
this,
the
this
is
a
resource
to
be
updated.
B
Okay,
so
I
was
talking
about
persistency
and
how
to
deploy
or
so
on,
twister
application.
I
thing
is
something
also
nice
from
on
k
p.m.
I
will
show
you,
so
I
have
a
vacationing
RabbitMQ,
which
is
really
predict
because
a
variables
on
a
data
volume
so
by
default,
is
just
mountain
until
here,
which
is
just
for
chemo
and
what
you
want
to
when
you
deploy
or
some
gesture
applications,
you
want
to
apply
different
multiple
times
with
catch
on
controller,
but
with
some
variation,
not
a
different
volume.
So
we
added
a
kid
named
Sherrod.
B
B
B
So
this
pocket
is
just
a
dependent
series
with
we
don't
deprive
resource
from
the
package,
so
the
manifest
we
plug
that
we
skip
the
first
part
from
now
so
I
say:
okay
I'm
depend
on
the
rabbitmq
package
we
just
deploy,
but
instead
of
using
one
child
I
will
go
these
three
of
them.
So
the
name
you
put
the
name.
We
want,
for
example,
Oh,
community
and
also
I,
don't
want
to
use
anterior.
B
This
is
not
something
you
want
for
prediction,
so
I
change
the
default
variables
with
a
system
where,
in
trade
of
course,
you
have
to
deploy
also
your
work
Jerry.
So
we
have
a
package
for
that
and
I
say:
okay
will
deploy
a
triggering
train
with
that
name.
Also
about
this
feature,
each
shot
can
add
its
own
vipers,
so
you
can
just
tweak
on
survivors.
You
want,
for
example,
if
the
last
we
don't
claim
you
want
50
gigabytes.
B
B
So
we
want
it's,
we
will
use
it,
it's
a
side
project
and
we're
looking
for
contribution
and
become
more
community
project,
because
its
packaging
are
very
easy
to
do
and
is
it
to
share
so
I
know.
For
example,
there
is
a
Sebastian
from
Sky
box
may
join
the
team
k
machine,
so
we're
very
much
looking
for
feedbacks
and
just
try
on
KPM
and
just
go
back
to
us
and
that's
basically
all
further
that
you
know
do.
A
A
E
Perfect
thanks
so
I'll
keep
this
actually
pretty
fast.
There's
not
a
ton
to
say,
but
I
want
to
make
people
aware,
just
in
general,
because
we've
sent
a
couple
of
email
lists
and
the
there's
been
an
issue
open
for
a
very
long
time,
like
uber
Nettie's,
where
we
desired
to
add
a
portable
serialization
to
the
API.
That
gives
us
some
nice
performance
benefits.
E
That's
the
primary
goal
of
it
is
for
internal
use
in
the
short
term,
and
so,
if
people
out
there
see
this
and
I
think
oh
my
gosh,
you
guys
are
generating
ideas
and
now
I
can
go,
generate
client
objects
against
them
and
be
a
little
cautious.
We
plan
to
evolve
the
serialization
and
we
certainly
an
experimental
state.
So
we
make
no
guarantees,
you
won't
break
it
or
change
it
until
it
shows
up
in
a
in
a
proper
beta
form
in
a
release,
but
under
underneath
the
covers.
E
We
have
started
to
add
this
into
the
code
path.
So
I
just
want
to
make
sure
that
if
people
are
contributing
that
and
they
hit
something
that
is
per
debuff
related
to
open
an
issue
into
CC
the
sig
api
machinery,
a
list,
we
are
hoping
to
turn
this
on
fairly
soon.
You
know
in
the
next
couple
of
weeks,
but
it
we're
still
in
the
process
of
finishing
up
some
of
the
deliverables.
E
A
A
F
Yeah
so
we
sort
of
said
that
at
the
beginning
of
1.3,
we
wanted
to
come
back
to
the
community
meeting
every
couple
of
weeks
as
coding
goes
on
words
and
sort
of
sync
back
up
as
a
group
on
the
top
several
features
so
I'm
going
to
go
over
I
guess
the
the
four
highest
profile
features
for
1.3
and
let
folks
know
where
they
are
a
lot
of
the
work
for
these
is
being
done
through
cigs.
So
if
you've
attended
sig
meetings
and
you're.
F
On
top
of
this,
you
probably
know
more
than
I
do
but
sort
of
the
general
update
for
the
larger
group.
So
first
off
the
first
big
feature
is
uber
Nettie's,
and
there
was
a
big
discussion
in
last
week's
cig
Federation
meeting
about
the
right
scoping
for
1.3,
we're
down
to
order
three
and
a
half
four
weeks
left
of
coding,
and
so
the
decision
was
made
for
1.3
will
get
cross
cluster
service
discovery
in
that
looks
more
or
less
on
track
and
like
it
will
land.
F
F
So
they
did
sort
of
a
in
person
virtual
hackathon,
amid
a
lot
of
good
progress
just
last
night.
So
that's
great
to
hear
that
that
that
stuff
is
moving
forward.
So
that
is
the
sort
of
uber
Nettie's
high
level.
The
next
one
is
pet
set,
so
pet
set
is,
is
essentially
on
track
for
alpha.
So
the
API
group
in
the
Alpha
resource
PR,
is
merged
in
in
the
controller
PR
is
in
flight
and
being
worked
on
and
the
initialization
containers
proposal.
F
F
G
A
A
All
right,
well,
I,
will
give
a
very
short
update
there.
There
seems
to
be
a
lot
excitement.
There
was
a
great
demo
yesterday,
no
I'm,
sorry
Tuesday
from
the
core
OS
team
about
running
containerized,
open
stack
on
top
of
Coober
Nettie's.
I
think
it
based
on
twitter.
It
seems
to
have
broken
a
lot
of
people's
brains
about
how
you
might
conceive
of
this
and
I
know
that
the
OpenStack
group
has
been
working
very
carefully
to
just
to
tease
apart
the
different
ways
open.
A
Openstack,
community
and
space
will
write
up
or
you
can
look
at
the
google
group
of
an
OpenStack
the
OpenStack
stake
in
order
to
find
links
to
information
about
the
khala
OpenStack
and
call
us
Cooper,
Nettie's
opportunities
and
then
there's
a
third
bit
which
I
know
that
Mike
spritzer
has
been
talking
about
and
working
on,
which
is
getting
Cooper
Nettie's
to
work
with
some
OpenStack
components
and
I.
Think
there's
been
a
fair
bit
of
discussion
about
this
on
the
sig
networking
list
and
also
on
the
sig
OpenStack
list.
A
So
they've
gone
ahead
and
tease
this
apart
into
three
distinct:
three
distinct
project
paths
for
lack
of
a
better
description
and
we'll
be
working
on
them
differently,
so
lots
and
lots
of
buzz
around
to
burn
eddies
and
OpenStack
11
Stephanie
Cooper
Nettie
is
at
the
OpenStack
summit.
People
were
pretty
excited
about
it,
so
that's
the
sig
update
and
the
city
I
love
the
notes.
Openstack
summit
may
have
eaten
a
let's
hope
not,
but
we
can
get
a
longer
update
from
him
next
week.
Alright,
so
we've
got
any
questions
on
that
No.
A
H
Just
a
quick
note,
I
was
reminded
by
an
email
that
I
got
today,
there's
a
lot
of
proposals
that
got
merged
that
are
either
no
longer
relevant
or
obsolete
or
didn't
get
worked
on
or
whatever
so
I'm
going
to
take
a
run
through
and
start
sentence
and
PRS
to
remove
stuff.
If
I
don't
know
how
I'll
make
sure
that
the
original
author
of
the
proposal
is,
you
know
added
into
the
pr
to
remove
it.
H
H
I
mean
there's
a
bunch
of
cruft
in
there
that
the
example
was
somebody
found
like
the
workflow,
a
workflow
doc
that
I
don't
think
we're
going
to
work
on
I,
don't
think
we're
going
to
add
workflow
as
an
object
to
the
Cure
benetti's
API
and
they
were
like
wait.
I
was
going
to
do
this
and
you
already
in
us.
I
think
it's
important
that
we
make
sure
we
realize
that
people
are
out
there
discovering
our
stuff
via
web
search
engines
and-
and
it's
very
very
hard
for
them
to
tell
the
difference.
I
guess.
C
All
docs
are
proposals
we
need
to
cool
up
in
general
yeah.
We
talked
about
merging
the
two
directories,
clearly
labeling
the
state
of
the
proposal
with
whether
it's
you
know
not
not
done
in
progress,
obsolete,
etc,
obsolete
things.
We
should
probably
just
delete
on
workflow,
specifically,
please
sync
up
with
air
tune,
but
yet
yeah
we've
also
talked
about
for
the
docs
Rico
docs,
putting
some
metadata
in
each
talk
about
who
the
owner
is
when
it
was
last
updated
and
reviewed,
and
things
like
that
I
think
the
same
thing
would
be
useful
for
design
docs
yeah.
H
D
D
Think
it'd
be
great
to
actually
scrape
metadata,
like
then
generate
some
sort
of
index
page
for
that
designed
I
page
so
because
right
now,
it's
an
arbitrary
alphabetical
order.
Right
and
there's
definitely
there's
definitely
in
order
to
those
things
and
there's
definitely
some
that
are
more
central
and
core
than
others
and
so
yeah.
Many.
C
A
H
A
Thank
you.
The
only
other
notice
I
have
is
again.
The
May
fifth
meeting
will
be
the
Asia
friendly
5
p.m.
pacific
time,
and
we
will
see
how
that
goes,
and
if
we
want
to
continue
with
the
first
meeting
of
the
month
to
be
in
Asia
friendly
time,
depending
on
how
many
people
we
get
attending
this
one
next
week,
thanks
sweet
holy
cow,
okay,
beyond
that,
do
we
have
any
open
I
mean
this
is
this
is
open
mic
time
now,
because
we've
got
extra
time
for
the
first
time
in
a
while.
A
A
Is
the
10am?
Yes,
that
is
something
that
I
have
to
fix
today,
because
we
have
hit
a
peculiar
limit
with
google
calendar
we're
in
when
we
have
more
than
200
invitees.
We
can
actually
no
longer
use
groups,
so
I
need
to
do
a
bit
of
rejiggering
with
this
and
rewrite
a
bit
of
documentation
and
mike,
I
will
add
you
to
the
calendar
item
so
that
you,
you
are
getting
that
as
well
and
yes,
it
will
get
moved
this
week.