►
From YouTube: Kubernetes SIG Apps 20170717
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
A
Okay,
so
there's
the
link
and
so
I'm
Matt
Farina
I'll
be
walking
us
through
this
stuff.
Today,
the
announcements
if
we
could
use
note
takers
if
anybody
wants
to
take
notes,
please
jump
in
here
the
more
notes,
some
more
details,
the
more
the
easier
it
is
to
find
something
when
we
go
search
for
it
in
the
past
and
to
share
with
others
who
might
come
across
the
videos
or
notes
or
be
looking
for
something
another
announced.
It
looks
like
compose
graduated
through
the
incubator
system.
Congratulations,
that's
another!
One!
That's
come
out
of
say:
gaps!
A
B
So
good
so
yeah
there
might
be
a
little
bad,
but
most
of
them
at
the
airport,
but
just
a
quick
overview
on
the
1.8
schedule.
Is
it
didn't
get
out
of
its
issued
3:05
and
some
of
the
key
dates
I
just
want
to
go
over
with
everyone
who
sega's
I'm
August
first
is
the
feature
freeze
so
make
sure
you
mark
those
August
1st
and
then
September
1st
is
code
freeze
and
then
the
actual
release
date
is
September
27th,
so
the
final
sort
of
code
build
in
September
20th
and
then
the
actual
releases
to
27.
B
B
A
C
Hi
everyone:
this
is
Lana
the
gaps
meeting,
so
it's
great
to
meet
everyone.
The
second
thing
I
wanted
to
just
very
briefly
introduce
is
a
project
that
I've
been
working
on
with
a
few
other
people
in
terms
of
operators.
So
if
you're
not
aware
of
what
an
operator
is,
it's
essentially
a
code
base
which
will
terminate
some
kind
of
manual
process
that
a
user
user
traditionally
does
so
one
example
is
the
STD
operated
by
Corollas,
which
sets
of
a
cluster.
C
So,
yes,
it's
really
a
key,
but
is
everyone
needed
okay,
cool?
And
yes?
So
we
have
a
bunch
of
operators
in
existence
right
now,
core
OS
have
a
few.
We
also
have
rook,
which
is
a
similar
kind
of
thing
to
the
STD
operators,
but
it
manages
the
lifecycle
of
staff
and
we
have
a
few
by
a
giant
swarm
and,
although,
like
all
of
these
projects,
are
super
awesome,
one
problem
which
we've
run
into
initially
is
that
we
see
a
lot
of
duplication
across
code
bases.
C
People
are
solving
the
same
kind
of
problems
and
you
know
at
the
best-case
the
same
way
in
the
worst
case
of
divergent
implementations.
So
we
came
up
with
the
idea
of
having
a
shared
code
base
and
operate
to
the
kit
where
people
have
a
common
library
or
an
SDK
to
write
their
own
operators
and
it
sort
of
establishes
best
practices
and
in
a
coastal
is
common
code,
and
one
of
the
aims
is
to
sort
of
really
promote
people
writing
their
own
operators
at
the
time
and
sort
of
encouraging
them
to
get
involved.
C
C
So,
if
you're
interested
in
reviewing
that
document,
I
can
post
it
in
the
slack,
it's
also
in
meeting
notes
and
so
I'd
like
to
solicit
any
feedback,
whether
you
think
it's
a
good
idea,
whether
you
think
we're
missing
any
potential
features
there
and
then
the
second
thing
which
I'd
like
to
do
is,
if
it
all
looks
good
to
people,
is
to
propose
it
for
incubation,
and
we
already
have
a
bunch
of
interested
people
willing
to
help
out
with
it.
So
coral
s,
giant,
swarm,
rook
and
myself
are
interested
in
creating
this
codebase.
C
So
so
yeah
I
was
speaking
to
some
folks
from
coral
s
and
they
suggested
that
sig
apps
might
be
a
good
fit.
So
if
I
just
want
to
hear
people's
feedback
on
that,
whether
they
think
that
the
gaps
could
be
a
good
sponsor,
it's
not
what
would
be
a
better
fit.
So
in
a
nutshell,
that's
about
it.
If
there
are
any
questions,
feel
free
to
ask
from
now
pinyon
slack
so
yeah.
So.
A
C
So
I
read
the
documentation
on
it
and
the
two
things
that
we
need
right
now
is
a
champion
and
a
sink
responses,
so
Antwon
from
Crowell
Asus
volunteered
to
be
the
champion,
and
he
suggested
that
the
gaps
might
be
a
good
sponsor.
So
and
once
we
have
a
sig
as
the
sponsor,
the
next
step
would
be
submitting.
The
report
proposal
submitted
death,
so
we
need
we
need
that
sig
as
the
next
step.
Basically,
okay.
A
Since
all
the
sig
legs
are
now,
let
me
we'll
take
this
to
the
mailing
list
and
I'm
going
to
run
it
by
Adnan
and
Michele
offline
and
we'll
go
back.
You
I
think
this
is
probably
the
best
place
for
it
and,
let's
see
where
it
goes,
I'm
just
not
going
to
speak
for
lots
of
folks
who
aren't
here
and
in
the
PM's
and
all
of
that
stuff
ahead
of
time,
because
we're
going
to
move
on
so
is
there
any
other,
last-minute
feedback
to
this
any
questions.
D
Can
I
just
ask
a
quick
question,
of
course:
okay
thanks,
my
name
is
Mike,
so
I'm
looking
for
the
proposal
here
and
I'm
just
going
to
kind
of
understand
this.
This
is
the
operator
kit,
maybe
something
to
kind
of
help.
You
automate
deploying
things
into
a
running
kubernetes
through
configuring.
Anything.
C
So
the
operator
kit
is
essentially
like
a
common
library
that
other
operators
would
use.
So
it
wouldn't
you
wouldn't
direct
leave.
So
if
you
just
want
to
deploy
an
app
to
communities
or
you
want
to
deploy
the
control
plane,
you
wouldn't
directly
use
this.
You
would
it's
not
something
that
I
envisaged
sort
of
most
end
users
to
be
using
it's
more
sort
of
people
or
company
to
want
to
write
their
own
operators.
A
All
right,
if
there's
any
other
feedback,
please
head
over
to
the
proposal
or
follow
up
in
slack
afterwards
or
hit
up
on
the
mailing
list.
Thank
you.
I
expect
we'll
be
talking
about
this
in
the
future,
and,
and
so
the
next
thing
we
have
is
a
couple
of
demos,
and
so
the
first
one
is
a
forge
demo.
Do
we
have
somebody
on
Raphael?
Are
you
on
to
do
the
demo
or
somebody
else?
Yes,.
F
Can
everyone
see
that
my
browser,
or
can
anyone
see
it
I,
can
see
it
all
right
awesome.
So
I've
got
a
couple
slides
here
and
I
just
wanted
to
to
give
a
little
bit
of
a
background
just
context,
so
you
guys
can
understand
sort
of
where,
where
forge
came
from
so
results
myself
right
into
a
few
Shlomi
we're
gonna
complete,
called
data
wire
and
you've
been
helping
people
with
micro-services
since
about
2014
and
really
focused
on
what
developers
need
to
build
these
sorts
of
broad
applications.
F
Everything
from
the
enabling
technology
to
you
know
stuff,
like
kubernetes
to
the
organizational
changes
and
the
dev
tooling.
That
kind
of
goes
along
with
all
of
that
and
since
you
know,
since
about
2014,
we've
worked
with
a
whole
bunch
of
different
organizations
at
different
stages
of
adopting
micro
services,
and
we
really
kind
of
discovered
that
that
lead
from
a
certain
perspective
or
fred
a
certain
level.
The
devin
deploy
cycle
remains
pretty
constant
throughout
this
journey.
F
You're
always
doing
this
sort
of
same
you
know,
change,
make
its
change
every
impact
of
that
change
and
diagnose
any
pretender
problems,
so
that
level
stays
the
same.
But
what
you
actually
need
to
optimize
that
cycle
really
changes
as
you
go
through
a
journey,
and
we
found
it
sort
of
helpful
to
think
of
that
in
three
different
stages
that
are
defined
by
sort
of
how
the
definition
of
what
a
successful
employ
means
changes
as
a
project
goes
through
these
different
stages.
So
we
think
stage
one
is
sort
of
rapid
development.
F
You
might
have
a
few
early
users,
but
really
in
deployment
a
successful
pretty
much
if
it
doesn't
crash
and
Stage
two.
You
know
you
start
to
have
a
little
more
users.
These
start
caring
more
about
disrupting
those
users,
and
at
this
stage
you
need
to
take
steps
to
ensure
that
your
changes
minimize
any
potential
disruptions
of
those
users.
This
is
where
you
start
saying:
sort
of
canary
canary
workflows
being
introduced
and
things
like
that
and
then
stage
three
and
not
necessarily
every
service,
to
get
to
this
stage
but
stage
three.
F
It's
not
so
great
at
running
source
code,
I'm
sure
you
guys
all
know
the
sort
of
the
mechanics
of
of
making
a
making
source
change.
You
know
you
need
to
track
down
the
source
code
to
doctor,
build
compute
potential
image
names,
tag,
push
runs
your
manifest
and
do
a
coop
cuddle
apply
when
that
gives
in
the
middle
plus.
On
top
of
that,
you
have
potential
dependencies
for
other
services
to
worry
about,
and
that
really
starts
to
get
in
the
way
of
a
productive
developer.
F
Workflow
pretty
quickly
so,
fortunately,
just
tries
to
to
sort
of
fill
in
the
gaps
necessary
to
use
kubernetes.
In
a
very
sort
of
natural
way-
and
we
don't
try
to
sort
of
high
kubernetes,
we
expect
you
to
know
and
understand
kubernetes
and
the
basic
concepts,
but
we
sort
of
streamline
the
workflow,
the
workflow
around
developing
an
application,
that's
composed
of
many
services.
F
So
how
does
it
work?
It's
pretty
straightforward.
On
every
service
you
have
one
or
more
services
in
a
git
repo.
Actually,
so
you
can
use
mono,
repo
style
or
separate
service.
Our
repo,
if
you
want
works,
just
searches
for
docker
files,
it
will
build
anything
via
those
doctor
files.
It
searches.
Are
you
defined
how
those
doctor
files
are
deployed
via
templated
kubernetes
manifests,
and
it
looks
for
a
service
dog
animal
to
understand
dependencies
and
metadata
of
those
services
and
when
you
run
Forge
deploy
it
gets
those
dependencies
metadata
from
the
service
yamo.
F
It
builds
all
the
services
and
dependencies
based
on
those
doctor
files.
It
pushes
the
updated
doctor
files
in
the
doctor
registry
for
you
and
it
instant
use
the
kubernetes
manifest
with
the
metadata
from
serviced
by
animal
and
from
from
the
actual
build,
and
it
will
run
through
cuttle
apply
to
deploy
all
your
services
and
so-
and
it
will
do
that
intelligently
incrementally,
and
it
will
do
that
in
parallel,
so
you
can
use
it
on.
You
know
many
many
services
and
it
will
be
looking
pretty
efficient
so
to
illustrate
this
I'm
going
to
show
you.
F
This
is
actually
well
make
sure
you
demo
what
force
does
the
application
I'm
showing?
You
is
actually
a
real?
Well,
it
is
a
a
real
sort
of
proof
of
concept
I'm
working
on
and
it
consists
of
four
services.
I
have
an
API
gateway
deployed,
that's
actually
envoy,
plus
a
little
extra
plugin
that
we
have
that
the
delegates
authentication
to
an
external
service
I've
got
an
external
service,
that's
responsible
for
doing
authentication.
It
happens
to
integrate
with
all-zero.
B
F
F
F
So
can
you
guys
see
my
screen
here.
F
F
So
that's
my
for
service
die
animals
and
just
to
see
what
service
back.
This
is
easily
contents
of
service
anymore.
All
it's
really
defined,
for
the
be
a
gateway
is
just
the
name
API
and
the
fact
that
requires
the
authentication
service
to
function
and
then
some
some
additional
metadata
to
get
plugged
into
the
templates.
So
this
is
in
this
case.
This
is
the
target
port
of
the
service
and
every
limits
and
the
CPU
are
the
numeron
CPU
resource
limits
and
you
list
to
the
other
less
to
the
other.
F
F
Force
deploy
and
it
will
in
parallel,
do
docker,
builds
and
build
eight
or
build
push
and
applied,
and
actually
the
reason.
This
is
that
they're
easy
and
that's
pretty
much
done
at
this
point,
there's
actually
pretty
fast,
because
it
one
of
the
things
it
does.
Is
it
detects
based
on
a
hash
of
the
imports,
input,
the
source,
the
doctor
images
or
even
built
for
this,
and
so
it
just
uses
the
existing
cough
damages,
but
my
deployments
should
be
there.
I
can
well.
F
F
F
And
tells
me:
okay,
I,
don't
have
appropriate
privileges
for
for
accessing
a
particular
service,
so
that's
pretty
much
it.
This
is
designed
to
work
nicely
with
sort
of
some
of
the
other
tools
we
have
like
telepresence.
So
this
lets
me
spin
up
my
service
network.
If,
if
I
wanted
to
do
a
more
rapid
thing,
does
something
I
actually
did
yesterday
as
it
was
debugging?
This
I
can
also
do
something
like
because
I
have
the
whole
network
running
in
my
own
cluster
I.
F
F
As
you
get,
you
can
see,
I'm
now,
I'm
now
running
the
running,
either
running
the
officers
locally
on
my
box,
which
means,
whenever
I,
whenever
I
change.
You
know
things
like
this
right.
It's
going
to
live,
reload
and
I,
get
back
right
back
to
attack
to
the
tasks
service.
So
that's
kinda.
It's
kinda.
F
Let
me
go
back
to
my
side,
that's
kind
of
the
story
for
for
what
it
does
to
the
main
use
cases
that
we
develop.
This
for
is
actually
well
essentially
a
couple
different
things.
So,
if
you
think
about
the
main
use,
cases
are
things
like
local
development,
so
you
can
set
up
your
own
issue
like
a
small
team
and
if
you're,
one
team,
where
you
have
a
multi
service
app,
you
can
use
this
to
create
a
really
nice
local
dev
experience
just
by
running
everything
in
mini
to
have
your
own,
isolated
them
environment.
F
It's
useful
in
testing,
so
you
can
use
it
as
an
indie
basis
or
in
your
CI
system
to
install
your
app
from
source
and
run
tests
on
it,
and
it's
actually
useful
because
it's
available
to
dev
tool
as
well.
So
you
can
run
just
one
test
and
exactly
what
your
CI
system
does
and
then
another
thing
that
that
we
use
this,
for
is
knowledge
sharing.
F
So
you
can
kind
of
and
what
a
great
things
about
credit
is
you
can
sort
of
distribute
and
package
of
all
the
operational
stuff
necessary
to
run
all
network
services,
but
when
you
share
it
that
way,
it's
if
it's
you
know,
if
all
the
animal
year
you
pass
around
just
points
directly
to
dr.
images:
it's
not
very
modifiable,
so
we
use
this
for
knowledge.
Sharing.
You
can
share
sort
of
share
something
away.
F
That's
really
really
easy
to
modify
the
source
code
and
sort
of
adapted
to
your
own
needs
and,
let's
say
I,
think
I
I
think
I
covered
teaching.
Oh
yeah,
that's
been
a
plumbing
system,
so
we
can
automatically
apply
certain
transformations
like
the
like
guys
do
like
the
coop
inject
stuff,
we
complied
on
automatically
like
you're,
using
as
your
deployment
templates
need
to
be
need
to
be
modified
in
order
to
integrate
with
Sto.
F
Let's
see
what
else
I
like
I
said
it's
designed
to
be
really
really
fast,
so
it
takes
advantage
of
all
the
existing
docker
images
in
your
repo.
It
does
dependency
managing.
So
if
you're
on
a
larger
sphere,
a
larger
scale
application,
it
knows
how
to
just
pull
down
based
on
the
service
metadata,
the
subset
of
services
necessary
to
stand
up
whatever
it
is,
you're
trying
to
deploy,
and
then
we
use
the
bit
for
the
template
to
use.
F
Let's
see,
I
think
I'm,
I
didn't
actually
show
you
the
templates,
but
so
this
is
the
true
the
nephron
plate
and
it's
just
using
sort
of
standard
Kinjo
templating,
which
we
kind
of
think,
because
that
was
already
in
use
of
sort
of
the
first
organizations
that
we
built
this
for,
but
it's
it's
proven
pretty
useful
in
that
you
can.
Actually,
you
can
actually
eliminate
a
lot
of
boilerplate
with
sort
of
a
ginger
to
tumblers,
and
you
can
enable
another
kind
of
knowledge
sharing.
F
So
there's
there's
a
couple
other
tools
that
are
kind
of
similar
related,
possibly
in
some
ways:
they're
overlapping.
So
we've
seen
we've
seen
you
know
a
lot
of
the
people
we
work
with
use,
docker
compose
for
similar
workflows.
F
F
And
they're
you
know,
force
doesn't
doesn't
have
those
that
sort
of
single
monolithic,
PMO
type
view
the
docker
compose,
has
an
abuser,
your
network
of
services
as
a
network
and
and
tries
to
enable
you
to
stand
up
whatever
subset
of
that
you
want
to
there's
a
Netflix
in
you
which
is
not
which
is
well.
It's
not
see
a
couple
talks
about
it.
F
It's
not
open
yet,
although
I
think
they're
talking
about
when
sourcing
yeah,
it's
very
very
similar
idea
in
a
lot
of
ways
but
built
inside
Netflix
and
because
they're
not
you
know,
because
they
work
with
ec2
and
sort
of
their
own
container
container
platform.
They
are
less
sort
of
their
less
container
centric,
but
very,
very,
very
similar
in
a
lot
of
ways.
Also,
just
because
of
where
it
started
focus
is
a
little
bit
more
on
project
bootstrapping
on
that
aspect
of
the
workflow
draft.
F
So
the
difference
is
they're.
Forge
is
an
entirely
client-side
tool.
You
don't
need
anything.
You
need
anything
on
the
cluster
or
running
in
the
cloud.
There's
no
like
central
server
for
it
just
uses.
You
know
whatever
doctor
register
you
pointing
at
and
whatever
whatever
coming
any
cluster.
You
pointed
at
another
difference
between
employment
route.
There's
direct
produces,
produces
home
charts
and
for
just
uses.
The
doctor
images
include
any
data
directly
based
on
the
change
of
templating
and
I.
Don't
I
don't
know
if
I
don't
know,
if
draft
does
anything
with
dependency
management?
F
A
A
This
was
great,
there's
a
real
problem
space
that
was
being
kind
of
attacked
from
multiple
different
points
of
view
of
how
do
we
go
about
operating
this,
and
especially,
how
do
we
also
develop
along
that
type
of
stuff
and
then
the
separation
of
concerns
between
development
and
operating
and
operating
scale
and
all
those
things?
So
thank
you.
This
is
this
is
useful
and
I
hope
it
was
useful
for
somebody-
and
we
have
another
demo
today,
which
is
cubed
EB.
Is
anybody
on
to
talk
about
that
yeah.
A
G
Yeah
so
I'm
Tamala
from
s
code,
so
Q
DB,
is
a
project
for
running
database
systems
on
culinary's,
so
this
is
actually
based
on.
This
is
actually
a
collection
of
operators
for
videos,
databases.
Currently
we
have
implemented
this
for
postage
and
elastic
search.
So
to
give
you
the
background,
basically,
we
started
running
our
own
search
platform
on
quantities
and
we
needed
you
know
the
way
to
run
our
databases
in
the
beginning.
G
Nonetheless,
and
all
those
kind
of
things
then
once
stiffing
said
become,
you
know
thing
we
started
my
getting
to
do
that
and
and
but
there
are
always
issues.
How
do
you
and
like
take?
You
know
your
data
snapshots?
How
do
you
sort
of
recover
from
those
snapshots
instead
of
the
whole
lifecycle
management,
wasn't
really
easy.
So
we
developed
a
this
operator
based
system.
We
are
calling
this
qtb
so
to
defy
the
system.
Basically,
you
can
go
to
the
CLI
project,
so
it
is
a
open
source
on
github.
G
It
is
actually
under
that
cage
is
DB
github
organization,
because
QDB
was
unavailable,
but
we
call
it
cube
DB.
So
to
use
this
system,
you
can
go
to
the
release
Gabriel
in
Spain
and
actually
download
produced,
see
allies.
So
we
have
my
capable
sale
eyes
for
them
to
the
front
opening
systems.
So
today
I'm
going
to
demo
this
system
using
a
kubernetes,
1.5
cluster
and
running
on
AWS.
G
So
so
this
is
the
my
terminal
to
the
move.
This
cluster,
so
I
have
basically
P
deployed
a
few
things.
Basically,
I
have
a
primitive
running
on
this
cluster
and
I
also
have
a
PG.
My
at
main
running
so
that
I
will
use
to
show
that
altered
a
bit
is
actually
running
and
you
can
connect
to
it
and
actually
know
run
for
different
things
like
that.
So
now,
first,
you
can.
H
G
The
traveler
to
the
CLI
installed
you:
can
you
download
the
CLI
and
copy
it
to
the
basically,
your
pod
folder,
like
either
a
local
being
folder,
and
that's
all
you
need
to
do
so.
Here
are
one
side
on
QB.
You
need
so
in
this
particular
cluster
I
already
have
it
deployed
so
once
it
is
actually
deployed.
G
G
Okay,
now
I'm
going
to
use
a
game
of
mini
space
for
all
these
demo
operations,
so
this
is
the
TPR
that
will
create
a
postcard
database.
So
there
are
like
a
bunch
of
features.
That's
present
here
so
firstly
say
that
this
is
the
version
9.5,
so
we
are
actually
using
community
for
post
less
9.5
database.
So
that's
we'll
just
we're
using
in
our
production
system.
G
So
this
is
actually
dynamically,
create
a
stateful
set
and
it
will
use
a
deploy
DVCS
using
the
dynamic
positioning
stories
positioning.
So
so
we
have
this
first
class
first
class
in
the
first
time,
which
we
call
DP
to
TP
TP
to
space
for
AWS,
and
then
I'm
also
come
back
to
these
sections
later.
So
let
me
phosphate
this
database.
G
So,
as
you
can
see,
so
it
is
not
actually
in
the
stead
of
running.
It
has
created
a
story.
It
is
using
Swiss
closet
in
DB
did
at
this.
We
have
skated
as
far
as
called
p1.
It
has
automatic
allocated
a
dinner,
this
secret
for
the
hostess
routine
super
user,
and
then
it
is
also
actually
configured
a
monitoring
system
using
the
color
is
vomit
evaporator,
and
then
you
can
also
see
the
various
emails
that
are
created.
So
if
I
actually
go
like
you,
video
get
April
said.
G
G
G
G
G
Now.
If
you
go
back
to
the
two
of
restrictions
that
was
in
the
state,
we
had
a
monitor
or
section
which
said
that
I
want
to
create
a
service.
Monitor
which
will
be
created
in
the
lamest
is
deformed
because
my
ponytail
is
running
in
the
Nemesis
default
and
then
it
will
happy.
The
forest
meter
will
have
the
labels
at
Colin
cube
TV,
because
that's
the
that
is
a
service
that
is
the
service,
monitor
selector.
This
particular
community,
like
the
Prometheus,
is
not
using,
as
you
can
see
here
this
part,
so.
G
So
so
now,
if
I
go
to
the
funkiest,
that
is
running
here
for
the
fresh
you'll
see
that
it
has
basically
automatically
ejected
this
p1.
The
data
is
a
bit
of
a
spot
that
is
running,
so
it
is
basically
running
a
sidecar
process,
exporter
and
it
is
coming
from
the
demo
namespace
and
there's
a
bot
p10.
This
is
p1
and.
G
G
This
is
the
job
that
has
been
created
based
on
the
service,
monitor,
ok
and
then
in
the
database.
We
used
one
of
the
thing
which
was
we
I
mentioned.
That
is
going
to
use
a
neat
script,
so
basically,
I
wanted
to
initialize.
The
database
is
some
additional
schema
that
is
not
present
in
the
static
process,
just
to
show
that
we
can
actually
use
a
in
each
script.
G
So
in
this
be
people
there
is
this
round
of
Sh
file
and
this
sh5,
basically
Delta
PL
command
using
the
process
user
with
this
is
running
as
the
localhost,
so
you
don't
really
need
password
this
prison
mode,
and
then
that
actually
cares.
This
allows
this
then
just
demo
excel
file
file.
So
you
can
use
anything.
You
need
for
your
project,
so
this
creates
a
schema.
There's
a
data
schema
that
up
there
and
then
it
also
creates
a
gradual
table
in
that
English
schema.
G
G
G
G
For
this
item,
oh
I'll
show
that
the
cluster
is
running
in
SC.
Buddy
I
can
actually
take
it
back
up
in
there
like
at
the
word
cloud
storage,
so
to
use
Google
Cloud
storage,
you
have
to
create
a
secret
which
is
the
credential
for
the
Google
Cloud
storage,
so
that
secret
has
basically
two
films,
which
is
the
whole
cloud
project
ID
and
then
they
of
the
service
account.
G
Jason
Keyes
is
actually
that
this
sun
file
that
you
can
download
from
the
Google
Cloud
console-
that's
that's
essentially
just
an
encoded
into
this
I
can
actually
show
you
the
so
this
is
position
size.
That
is.
This
is
already
not
valid
anymore,
but
you
can
see
the
demo
so
now,
if
I
go
and
create
this
snapshot
object.
G
G
So
this
is
the
QT
B's
like
a
prefix
that
we
add,
and
then
the
demo
is
the
Minister's
name.
T1
is
a
database
name,
and
this
is
the
snapshot
name,
that
I
came
here
with
Stanford
XYZ
and
then
inside
that
it
has
the
dumper
file,
dot
sequel.
So
this
is
the
PE
dump
command
that
we
then,
with
the
job,
ran
automatically
and
then
just
be
the
output
of
the
process.
So
all.
A
G
Thank
you.
So
if
you're
interested
you
know,
you
can
connect
to
us
on
one
github
project
and
then
yeah.
A
A
Alright.
The
next
thing
we
have
up
is
actually
to
do
some
planning
for
what
we're
going
to
continue
to
work
on
next,
and
this
is
a
follow-up
from
last
week
with
the
kubernetes
core
and
kubernetes
tooling
functionality.
There's
another
document
for
this
planning
and
we'll
get
a
little
bit
of
time.
Left
I
just
dropped
in
a
link
to
the
planning,
and
so
some
of
this
is
filled
in
and
this
is
kind
of
what
we're
planning
for
not
necessarily
what
we're
going
to
do,
and
unfortunately,
I
wasn't
here
last
week
too.
A
I
E
H
A
I
A
Sounds
like
a
great
idea
getting
to
stability
will
influence
the
use
of
it
and
then
we
can
actually
start
to
learn
a
whole
bunch
from
it
being
used,
and
then
we
can
continue
to
improve
on
it
and
discover
we're
breaking
changes
really
need
to
come
in.
But
as
long
as
we're
breaking
things
all
the
time
or
there's
potential
for
that,
that's
going
to
you
know,
cause
people
to
not
use
it
or
treat
it
the
same.
And
so
we
can't
learn
as
much
so
I
liked.
I
So
a
little
more
saying
if
we
want
to
get
some
stability
in
1.9
GA
and
we
really
need
you
most
of
working
1a
and
just
have
one
9
kind
of
is
like
a
soap
period.
I
claim
just
so
that
if
there's
any
last-minute
things,
we
fix
that.
So
I
would
think
that
we
really
need
to
like
look
at
his
list
and
say:
what
can
we
do
in
one
quarter
to
bring
this
to
GA
plus
keep
1/9
clear?
That's
that's
my
take
I.
J
The
other
thing
I
would
say
is
that
there
are
a
few
things
that
we
know
we're
really
confident
and
we
want
to
do
going
forward.
So,
as
we
visit
API
surface
and
try
to
move
it,
it
sort
of
v1,
we
have
to
make
sure
that
you
do
it
in
such
a
way
that
we
don't
Payton,
hold
ourselves
from
implementing
things
on
board.
It
moves
the
core
things
that
we
want
to
do.
We
are
aware
of,
and
we
kind
of
look
at
the
API
and
get
confidence
that
way
we
tend
implement
this
later.
J
A
I
J
So
what
do
you
mean
by
make
changes,
but
do
it
in
a
bag
like
if
we
stabilize
we're
stabilizing
the
API
service?
Like
there's
nothing
to
say
you
can't
add
things
to
it,
that
will
you
know
it
will
that
we're
not
going
to
choose
try
to
deprecated
later?
Basically,
if
we
should
be
confident
that
whatever
we're
going
to
try
to
advance
for
there's
something
that
we
want
to
stick
with,
and
then
you
know
there's
nothing
to
say,
we
can't
also
have
a
J
as
Beta
Alpha
API.
J
I
K
This
is
a
group
of
Christian
Valley,
so
how
about
the
design
review
so
I
know
so
cold
March
is
obviously,
if
you
don't
want
to
do
that.
You
know
it's
part
of
the
stabilization
effort.
It's
there
the
opportunity
to
kind
of
do
the
design
of
user,
so
that
is
if
somebody
wants
to,
cannot
develop
it
internally
and
we
can
go
ahead
and
do
that
as
long
as
we
have
some
kind
of
a
consensus
from
the
community
as
far
as
the
design
is
that
possibility,
though,
is
that
not
a
possibility
of
god.
I
Chris
Paul
I
think
it
can
be
a
guaranteed
time
flies
like
I,
think
everyone.
Hopefully
everyone
who
knows
the
code
best
to
review.
It
is
going
to
be
working
on
bringing
all
four
controllers
to
GA
their
spare
time
at
my
city
design
review,
but
it
can't
be
a
budgeted
thing.
I,
don't
think,
are
you
thinking
of
the
burden
here
is
in
specific,
no.
K
At
the
three
things
we
actually,
we
propose
the
thing
that
the
leap
is
there
on
the
columns
well,
the
default
container
and
that
the
two
other
things
the
updater
rolled
back
and
of
the
automatic
rolled
back,
and
there
was
another
one.
So
what
we
were
thinking
was
if
we
can
get
some
kind
of
a
GTM
on
the
design
review,
then
we
can
go
ahead
and
start
doing
that
internally
or
developments
that
will
kind
of
help
us
find.
J
K
I
So
without
where
I,
rather
than
litigating
specific
issue
in
the
sig
apps
meeting,
maybe
we
can
define
a
process
of
need
to
do
for
contentious
stuff.
So
maybe
we
can
identify
stuff
that
everyone
agrees
needs
to
go
in
soon
and
then
such
as
contentious.
Maybe
we
can
have
time
slots
the
next
couple
so
that
meanings
to
have
that
would
be
basically
an
in-person
design
review
conversation
yeah.
That.
K
A
Okay,
so
how
about
this?
One
of
the
things
that
we
did
in
the
past
was
we
took
these
and
we
put
them
I
think
was
into
a
spreadsheet
and
then
at
the
top
you'll
notice.
We've
got
the
group
codes
for
what
organizations
are
interested.
We
can
put
those
across
the
top
the
features
along
the
left
side,
and
then
you
know
we'll
have
a
space
for
other.
A
For
those
of
you
who
are
not
part
of
one
of
these
organizations
and
then
we
can
Cali
up
which
organizations
are
interested
in
which
features
to
try
to
figure
out
which
ones
we
want
to
get
in
in
a
1.8
time
period,
and
then
we
can
also
talk
about
1.9
will
be
focused
on
stability
of
what
we
have
and
we'll
run
that
by
to
see,
if
we're
still
into
that.
So
by
the
end
of
the
year,
we
can
release
V
ones
of
these
controllers.
Well,.
K
Yeah
I
think
so,
but
we're
not
really
pushing
for
enough.
These
features
to
be
included
in
1.8.
What
we
are
saying
is
we
do
it.
We
definitely
agree
with
the
stabilization
idea:
the
G
airily.
We
agree
with
that,
but
as
long
as
we
can
get
some
kind
of
approval
for
the
design,
then
we
can
keep
developing
it
in
our
internally.
So
that
way
you
know
it
doesn't
really
conflict
with
the
stabilization
at
yep.
A
We
can
still
add
more
features
into
1.8.
What
we
want
to
do
here
is
try
and
prioritize,
which
ones
we
find
to
be
interesting
and
find
out
where
there's
overlap
between
the
different
organizations
and
developers
who
are
interested
to
see
which
features
they
want
to
get
in
and
in
a
spreadsheet
may
help
us
do
that.
Okay,.
A
K
A
K
A
A
So
we've
run
out
of
time
to
dig
into
all
the
kubernetes
tooling.
I
will
say
that
helm
has
some
details
here
and
they
do
have
a
road
map
up
in
the
form
of
a
roadmap
and
a
release
cycle
and
I
see
some
details
in
here
about
monocular
and
the
app
registry.
There
are
a
number
of
others
here,
such
as
track
Testament,
but
open,
compose
composed,
which
graduated
it
should
have
a
road
map
we
can
link
to
in
here,
charts
an
app
controller.