►
From YouTube: Kubernetes SIG-Apps 20170227
Description
demo of monocular and kubeapps.com,
discussion on declarative application management by Brian Grant
A
All
right
welcome
to
Vegas.
Everyone
today
is
februari
27th
neocon
agenda.
Today,
just
for
announcements
and
housekeeping
there
are
some
demo
slots
available
in
the
next
few
weeks
weeks.
So
if
you
have
something
to
show
feel
free
to
throw
yourself
on
the
agenda,
our
first
demo
is
by
the
folks
at
bitnami
Miguel.
A
He
is
going
to
talk
about
the
UI
for
the
hearts
repository
the
repository
that
that
codes
lives
in
is
called
monocular
and
all
of
him
and
talk
about
that.
A
little
bit
more.
Our
psychic
donna
I've
got
cancelled,
and
our
discussion
topic
for
today
is
a
brian
grant,
sharing
thoughts
on
declarative
application
management
and
after
that,
if
you
have
any
other
items,
you
want
to
discuss,
feel
free
to
throw
it
under
the
open
discussion
section
or
just
bring
it
up
at
the
end,
and
with
that
I'll
hand
over
to
Miguel.
B
Ie
world
every
night,
I'm
going
to
show
you
a
for
yourself:
I
am
wonderful,
patina
ham
and
is
I'm
working
on
is
full.
After
sudden,
soil
decision,
among
here
we
come,
depends
diaper
rated
that
were
using
their
way
home
realty
month
and
another
fifty
pence.
The
same
stem
cell
for
red
I
have
workers
other
better
I
noticed
those
occasions
or
I
was
a
colon
cancer
and
she
was
just
together
and
find
them.
It
was.
B
B
Sorry
For,
Women
against
it
I
license
revenues
of
the
conversation
you
can
decide
if
we
produced
there
are
others
and
even
convert
a
lien
on
the
body
so
cute.
I
remember
with
just
a
information
from
the
fitting.
My
poster
in
this
is
typical
American
armed
forces
perfidia
give
me
life
and
take
another
way
to
a
jury.
This
is
amazing.
B
Monogram
will
be
happy,
is
related
to
find
faculty
group
presenting
that
matter.
So
there
is
awesome
nasa
in
resented
diagnosis.
Today,
a
monograph
support
computed
by
all
to
use
data
services
which,
which
are
dating
twins
data.
It
has
so
everything
the
vacuum
is
a
viable,
a
dangerous
team.
We
say
that
mean
together
become
a
member,
it
is
the
change
and
we
shall
changes
version,
and
it
will
show
that
in
the
data
in
memory.
B
So
when
we
have
that
information
memory
on
in
this,
we
suppose
that
using
a
determined
yet
under
the
eyes
appear
on
any
other
asking
people
for
some
others,
which
is
occasions
that
we
use
today
from
double
dragon
to
be
high,
of
which
there
is
a
feather
or
low
profile
on
Manhattan
several
reference.
We
probably.
A
B
D
B
So
going
back
to
you
know,
monocular,
basically
an
interface
on
top
of
doing
some
charts,
the
way
they
implemented
using
a
packet
enough
fronting
the
bag
and
getting
information.
30-60
minutes
from
the
facial
chat
repositories
and
indexing
that
meta
data
in
memory
or
in
the
end
is
like
a
local
cache
and
that
made
the
day
that
gets
exposed
via
HTTP.
I
put
the
boot
the
server
that
is
part
of
the
back
in
itself,
and
then
we
have
a
fronting,
which
is
an
angular
single
page,
app.
The
big
bundle
next
to
an
indian,
a
foot
server.
B
And
so,
if
you
go
to
any
one
instance
of
mono
girl,
you
will
download
the
single
page
of
new
processor.
Oh
my
stars
in
a
star
interact
with
the
back
end.
So
does
that's
the
architecture
and
defied
version
of
architecture
diagram
so
how
this
looks
like
once
you
run
it.
Is
this
cube
com?
It's
a
running
instance
of
monocular
the
is
configured
to
to
pull
info
from
the
fidgets
chat
repositories
and,
as
I
mentioned,
we
are
when
the
main
goals
is
to
help
users
to
discover
to
discover
new
one
newham
charts.
B
So
that's
why
you
consider
the
messaging
is
more
generated.
Is
we
are
talking
about
long
launch
great
when
it's
ready
at
various
we're
not
saying
launch,
come
chat
that
way?
What
we're
moving
is
the
hump
chart
processes
so
called
himself
a
different
Asian
detailed
some
way,
so
they
were
with
the
planet.
It's
bigger
people
discover
us,
and
then
we
explain
how
to
study.
So
you
have
a
search
engine
search
input
here
and
then
the
lipids
are
sorted.
My
name
from
both
this
table
unemployed
of
the
poster
face
with
the
latest
version
shown
there.
B
Just
you
can
see.
There
are
many
people
with
different
ads.
We
have
things
like
my
sequel
with
applications.
We
have
many
different
Prometheus
and
then
some
informational
how
to
contribute
a
voice
behind
this
project.
If
you
want
to
do
you
want
to
do
a
mud
run,
filter
and
you
can
both
to
be
able
to
dislike
charge
link,
and
here
you
can
do
filtering
by
very
positive-
you
can
filter
by
one
was
created.
B
The
poachers
scene
now
is
the
is:
what
is
information
about
any
specific
version
of
a
child
and
what
you
have
a
belief
is
all
day
informed.
What
are
the
frequencies,
but
it's
inside
a
charge.
I
will
study
some
configuration
options,
which
is
literally
the
beam
inside.
You
will
find
in
the
github
repository
for
this
seating
chart
and
then
on
the
price
you
have
the
iphone.
I
mentioned
that
we
need
to
process
in
the
back
turn
the
name
lithium.
B
All
this
information
coming
from
the
metal
itself
index
itself
and
the
next
box
is
the
is
the
structure
hospital
how
to
start
how
to
use
this
discharge,
and
I
feel
it's
very
cool
because
in
how?
If
you
try
this
salad,
sharp
that
is
not
on
the
table
official
stable
charge,
you
first
to
add
dead
people
shatori,
and
that
was
monopolize.
Did
you
hear
that
we
need
to
add
the
incubator
repository
and
then
has
been
studied
and,
as
you
can
see,
a
boutique
Todd
also
tells
you.
B
System.Exit
is
impeding
a
specific
version
of
the
church
I'm
talking
about
versions.
We
have
here
different
versions
that
you
can
she
slap.
You
can
switch
versions
that
you,
when
it
was
released
and
changing
persons
means
that
you
get
a
new
route
up
here
that
you
can
share
with
other
people
want
to
hit
the
printed
version.
You
can
modules,
you
get
instructions
on
how
to
start
a
specific
version,
and
even
if
you
actually
take
a
closer
look,
you
will
see
that
the
readme
file
change
and
my
opinion.
B
This
is
one
of
the
most
interesting
features
from
one
of
your
because
you
know.
Sometimes
you
cannot
delete
this
version
of
a
chart
so
having
having
an
easy
way
to
find
the
documentation
of
any
specific
version
which
you
sorry.
This
is
no
signal
is
very
simple
to
do
it
today,
because
you
will
need
to
either
life
find
the
star
the
package
and
then
get
the
pygmy
file
or
go
to
the
github
repository
and
go
back
from
history.
B
B
There
have
been
discussions
from
moving
the
dell
search
capabilities
toboggan
because
today
is
a
very
simple
client-side
search
so
moving
into
the
back
end,
so
other
clients
could
use
it
likessee
allies
they
use
this
search.
I'm
also
add
more
advanced
query
for
the
village
by
indicates
today
is
still
functional.
You
can
search
by
name,
you
can
search
but
I,
you
can
say
by
keyword
or
you
can
even
search
for
my
mind.
B
There
yep
buy
me
dinner
so
today
and
give
up
with
gum
or
any
other
grounding
instead
of
no
brother
68
charts
important-
and
this
is
as
I
mentioned
in
the
diagram
we
have
faced
every
15
minutes
or
any
change
that
anomaly
other
folks
in
the
chart
Tim
make
it
will
happen
here.
So
I
recommend
you
go
to
classes,
come
and
check
it
out.
So
going
back
to
the
repository,
I
would
like
to
point
out
a
couple
of
things.
First,
what
if
you
want
to
continue
and-
and
I
personally
encourage
you
to
do
it?
B
So
what
will,
if
you
want
to
contribute
anyone?
We
want
to
run
this
same
locally,
so
we
hadn't
structures
that
we
may
english
is
playing.
How
to
have
very
big
development
environment
for
both
the
buggy
on
the
front
end
using
a
doctor
composed
film,
and
what,
if
you
want
to
learn
it
yourself?
So
we
also
have
an
article
to
the
starter
guide
to
the
step.
Guide
explains
how
to
plan
monocular
in
your
hood
in
this
cluster
and
monocular.
B
Is
this
package
as
a
as
a
charge
itself
since
tallying
monitor
as
simple
as
and
tell
the
gentleman
over
I'm
actually
eventually
monopoly
and
staff?
In
the
fidgets
set
repository,
we
will
have
no
no
parental
monocular.
Who
knows
what
will
happen,
but
it
is
very
easy
to
trace
it
to
do
it
this
way
and
regarding
the
charge
itself,
I
recommended
that
you
check
out
the
code,
because
we
have
something
gusta
put
stuff
in
the
image
chart.
We
have
things
like
a
server-side
rendering
gear
or
like
a
pre-rendering,
very
pre-rendering
engine
for
for
SEO
purposes.
B
We
all
were
also
rely
on
engine
X
and
maintenance
controller,
and
we
also
computer
charger
support
pls
hospira
super
so
recommend
that
you
check
it
out
vehicle
and
if
you
actually
deployed
it,
and
then
you
want
to
configure
industrial
city,
we
have
isetta
script,
also
be
particularly
acting
wrappers
around
him.
That
might
help
you
to
do
that.
So
here
you
have
on
instructions
and
how
do
we
use
them?
B
And
last
but
not
least,
we
like
to
thank
everybody
involved
in
this
project
especially
mentions
to
deb
francis
from
days
who
let
the
backin
efforts,
my
co-workers
on
and
on
hang
on
shown
and,
of
course,
the
day
harm
in
the
chair
depository,
because
without
the
tool
and
without
the
content,
monocular
probably
wouldn't
make
any
sense
and
please
contribute
continues.
Given
your
pizza,
we
like
to
hear
from
you,
we
have
actually
go
to
the
issues.
Q
and
you
find
some
some
setting
stuff
that
we
are
that's
going
on.
B
We
have
the
pipeline
features
like
computing,
your
own
checks,
deposited
storage
or
for
him
being
able
to
install
shower
just
clicking
1
bottom
once
you
deploy
monocular
into
your
cluster.
So
if
you're
sitting
at
this
is
interesting,
which
it
could
be
good
for
the
community,
please
go
on,
please
no,
no
I!
Think
that's
terrific
for
me.
Thank
you
for
your
time
and
I
have
to
answer
questions.
D
First,
let
me
say
this
is
Matt.
Thank
you.
I've
been
longing
for
a
user
interface
on
top
of
the
registry.
I
think
this
was
fantastic.
It
makes
my
day
to
see
this.
It
was
great.
Thank
you
very
much.
It
does
bring
me
two
to
one
question
them
once
you
can
search
for
things
and
find
them.
What
would
it
take
to
go
that
next
step
and
say?
Oh
I
click,
a
button
that
lets
me
go
install
it
into
my
repository.
Is
there
anything
out
there
that's
talking
about?
D
B
Yes,
there
is,
this
is
something
I
would
have
been
discussing
their
naughty.
So
what
was
well?
They
won't
install
a
minute
Lissa.
He
starts
mon,
a
grinder
and
you're
not
supposed
to
30
or
cluster,
and
you
compute,
your
own
home
chart
repository
or
using
this
table
1
the
fact
that
we
could
connect
to
either
the
dealer
mr.
hung
button
or
use
a
similar
effort.
Some
folks
didn't
touch
word
to
communicate
with
himself
directly.
B
We
think
that's
possible
and
that
put
will
offer
an
experience
that
I
think
similar
to.
We
are
mentioning
which
we
will
get
to
that
WordPress
details
page
and
we
have
a
button
that
says
in
stock
right
and
that
under
needwood
do
all
the
magic.
So,
yes,
that's
one
of
the
wanted
features
that
we
have.
We
have
in
mind
that
we
work
in
our
opinion,
should
work
with
another
one
which
is
being
able
to
compute
your
own
charts
depositories.
D
A
B
A
B
B
A
A
B
Personally,
I,
don't
know,
I
think
it's
a
way,
I
believe
that
it
would
be
good
to
have
one
place
that
has
as
many
apps
as
possible,
so
people
can
find
what
if
we
charge
cycle,
is
up
there
from
where
they're
coming
from.
But
I
don't
know,
let's
open
the
discussion
offline,
I
would
like
to
hear
the
I
would
agree
with
the
pain
on
the
official
charged
maintenance
and
she
excited
this
case.
How
does
just
looks
like
if
they
have
no
quality
for
being
next
to
them
of
I
mean
coming
back
to?
B
What
do
we
need
to
do
in
the
front
end
before
sorry?
Before
we
add
more
charts
there,
we
might
need
to
implement
some
features,
2
to
Mark
some
charts
official,
to
be
sure
that
we
can
have
a
huge
amount
of
content,
but
that
contact
can
be
easily
differentiated
by
a
computer.
I.
Think
you
dated
a
come
to
that
and
incubated
so
I
interest.
It's
very.
D
A
A
C
C
Floating
there,
yes,
okay,
so
I'm
going
to
talk
about
the
clear
there's
application
management.
I
previously
wrote
a
dock
that
some
people
may
have
seen
a
white
box
coming
off
the
shelf
application
management.
This
content
is
locally
taken
from
there.
There
have
been
a
number
of
tools,
helm
included,
which
provides
some
degree
of
declarative
application,
management
and
Trinities,
but
still
feels
like
they're
gaps.
So
I
want
to
talk
about
a
solution.
C
That's
inspired
by
a
number
of
partial
solutions,
I've
seen
in
this
space
and
maybe
try
to
drum
up
some
interest
in
filling
the
gaps
that
exist
so
yeah.
My
the
problem
statement
is
I.
Think
urban
any
still
lacks
in
twin
cooling,
for
the
creative
application
management
by
white
box.
I
mean
that
the
user
can
see
and
understand
the
configuration
of
the
application
and
all
the
two
NH
resources
involved.
C
That's
to
distinguish
it
from
black
box
where
it
would
just
be
ok
canned.
They
might
not
know
anything
about
the
underlying
implementation,
so
I
have
a
few
requirements
in
mind
and
I'm
going
to
talk
about
most
needs
in
subsequent
slides
I'm,
not
going
to
talk
about
too
much
of
them.
Here,
except
that,
I'm
going
to
point
out
that,
however,
you
know
someone
writes
a
yamma
or,
however,
they
express
the
declarative
configuration
of
the
community
resources
and
of
the
application.
It
still
needs
to
leave
space
for
automated
prosecutors
to
do
some
amount
of
automatic
configuration.
D
C
Scaling
where
you
want
the
number
of
replicas
of
your
deployment
to
represent
a
replication
controller,
what
have
you
to
be
automatically
controlled
by
an
auto
scalar,
but
there
are
lots
of
other
cases
as
well.
You
know
we
encourage
people
to
ride
controllers
that
do
things
automatically
to
create
resources
also,
so,
whatever
solution
we
come
up
with
needs
to
allow
for
that
and
not
have
the
declarative
configuration
conflict
with
automation
and
vice
versa.
C
So
the
first
requirement
is
it:
the
knees,
be
possible
and
more
impossible.
Ee
D
to
manage
the
declarative
configuration
under
version
control
configuration
is
typically
human
authored
and
it
should
be
human
readable.
So
that's
very
different
than
binary
build
artifacts.
You
think
up
for
traditional
packages
that
are
stored
in
a
package
repository
which
your
gal
put
of
some
complex
build
process
and
no
one's
going
to
disassemble
them
in
leavey
assembly
code.
That's
just
not
useful,
you
know
if
you
write
a
deployment
manifest
or
service
manifest
or
a
config
map.
C
C
Also,
you
know
the
when
the
human
is
expected
to
be
writing.
Reviewing
changing
configuration,
that's
very
different
than
an
automated
system
like
an
autoscaler
in
part,
because
if
something
changes,
the
time
scales
involved
in
noticing
that
something
has
changed
and
responding
to
that
change
is
very,
very
different.
It's
not
like
an
auto
scalar,
which
continuously
observes
the
current
states
and
rapidly
takes
press
of
action.
So
if
you
like
blow
away
the
number
of
replicas,
an
automated
process
could
potentially
repair
that
very
very
rapidly.
C
Whereas
you
know
if
a
human
housecoat
recreate
it,
they
might
not
understand
what
what
the
configuration
should
be.
It
will
take
them
some
time
to
figure
that
out
so
keeping
the
configuration,
safe
and
discoverable
and
understandable
are
all
pretty
important
for
those
sorts
of
scenarios.
Keeping
configuration
and
version
control
is
considered.
The
industry
stay
best
practice,
industry-wide
best
practice,
all
kinds
of
areas,
whether
it's
chef
or
puppet
config,
or
your
engine,
X
config
or
whatever
it
is
just
people
have
decided
that
that
is
a
good
approach
to
take.
C
It
is
provides
a
familiar
workflow
for
making
changes
and
controlling
the
changes
for
viewing
the
changes
auditing.
The
changes
you
can
revert
changes
pretty
easily
using
standard
verse
control
tools.
If
you
need
to
reproduce
the
same
configuration
again
for
testing
or
the
different
cluster,
or
something
like
that,
then
they're
stinking
again,
that's
a
pretty
straightforward.
Thank
you
by
running
off
the
same
configuration
or
by
cloning
it
or
14,
branching
it
etc.
C
So
just
sort
of
a
brief
aside
on
customization.
You
know,
obviously
for
configurations
be
reusable.
It
has
to
be
customizable.
The
same
configuration
doesn't
work
for
everybody
in
every
current
scenario,
especially
even
it's.
The
same
person
or
team
is
instantiating,
multiple
different
instances
of
a
configuration.
Those
different
instances
typically
need
to
be
customized
in
some
kind
of
way.
The
most
typical
approach
the
people
take
is
parameterization,
and
you
see
this
in
open
shifts
templates
you
get
home,
charts,
hidden,
pretty
much.
C
C
That's
going
to
be
more
and
more
useful,
I'm
going
to
mention
the
service
broker
later,
but
what
once
we
have
service
brokers
for
things
like
helm
and
other
other
solutions?
That's
one
example
of
where
parameters
need
to
be
automatically
surface,
but
it
also
be
super
cool.
If
the
command
line
tools
can
query
these
or
four
parameters
to
fill
in,
so
that
they
don't
actually
have
to
go,
we
find
a
file
and
weep
and
meet
us
and
figure
out
what
what
values
they
need
to
provide
into
one.
So
a
common
scenario.
C
Again,
as
I
said,
is
people
want
to
instantiate
the
same
configuration
and
many
contexts,
so
the
structure
of
the
configuration
needs
to
lend
itself
to
that
like
in
the
helm
case,
an
example
would
be
providing
multiple
files
full
of
parameters
forcing
a
little
charts.
You
can't
can
quote
on
the
whole
thing.
Bob
people
have
a
resistance
to
doing
that.
There's
an
inherent
desire
to
choose,
specify
things
only
once
that
need
to
be
specified
at
once,
and
then
why
you
could
dip
all
you
know
multiple
different
copies
of
a
configuration
against
each
other
words.
C
Keep
them
in
sync,
that's
a
lot
more
tedious
than
just
keeping
one
copy,
and
then
you
know.
Finally,
just
like
you
want
a
version
control,
the
parametrized
templates.
You
want
to
verse
control
the
parameter
values
as
well,
typically,
so
not
just
necessarily
providing
them
on
my
family
and
although
you
could
check
in
a
script,
does
that's
why
I
devised
I'm
expand
mine,
but
people
really
want
a
declarative
file
checked
in
under
version
control,
so
they
can
manage
it
the
same
way
they
manage
the
temp
like
this,
but
then
there
are
also
many
other
customization
strategies.
C
Parameterizations,
not
the
only
one
and
you
can
see
parameterization
becomes
much
less
attractive
and
certain
scenarios.
If
anybody
is
taking
a
look
at
the
Mesa
spheres,
universe
packages,
some
of
those
packages
have
dozens
or
even
hundreds
of
parameters,
because
they
not
only
specify
every
parameter
that
is
needed
to
customize
the
configuration
for
marathon,
but
also
of
the
application
itself.
So
every
application
parameter
is
mirrored
in
a
message
for
universe
package
parameter,
and
you
know
what
that
sort
of
scenario.
C
C
C
Every
single
template
to
permit
that,
especially
for
an
uncommon
kind
of
changes.
14
works
much
better,
some
kind
of
patch
or
overlay
mechanism
k
p.m.
I
believe,
provided
some
sort
of
patch
mechanism.
It's
similar
to
forking
without
actually
forking,
specifying
a
patch
against
an
API
resource
can
work.
We've
actually
had
projects
previous
config
projects
internally
to
google
that
were
shut
down.
That
attempted
to
provide
stacks
of
stacks
and
patches
that
you
can
apply,
and
this
is
also
being
discussed
with
respect
to
way
Justin
Keith
control
apply.
C
It
gets
pretty
complex,
pretty
quickly
and
similar
to
docker
image
layers.
If
you
change
one
of
the
base
layers
and
invalidates
all
the
layers
above
so
that
create
some
fragility
that
can
be
really
hard
to
manage,
so
touching
works
well,
I
would
say,
is
fairly
general
purpose.
Patching
works
well
restricted
cases,
but
you
have
to
be
careful.
Generation
can
be
good,
really
useful
for
static
decisions.
I'm
a
big
fan
of
this
and
I
think
we
could
build
a
lot
more
chewing
in
this
regard.
C
For
example,
if
you're
building
a
java
application-
and
you
set
certain
parameters
which
respect
to
the
JVM
I,
there
have
been
some
tools
to
do
this,
for
you
know,
outside
the
container
space,
especially
for
platform-as-a-service
scenarios,
but
also
I,
think
the
successor
to
vagrants
did
some
amount
of
this
when
that
was
still
being
developed.
But
you
know
looking
up
at
the
application
and
automatically
generating
some
configuration
could
make
a
lot
of
common
cases
a
lot
simpler
and
works
very
well
with
the
rest
of
the
whole
flow.
C
If
you
just
do
it
on
the
front
end
and
use
it
to
fill
out
a
bunch
of
boilerplate,
similarly
composition.
This
is
where
you
know,
I
think
a
better
approach
to
changing
application.
Specific
parameters
is
to
just
provide
a
way
to
form
the
whole
application
configuration
in
as
opposed
to
parameter
parameterizing.
Every
individual
field
of
the
application,
configuration
and
config
map
and
keepers
are
intended
to
permit
this
and
Eddie.
C
So
you
can
potentially
provide
a
whole
config
map
to
an
application
rather
than
plumbing
those
individual
parameters
out
automation,
again,
I
nada
scaling
is
the
most
Cortana
lot
of
feelings,
the
most
common
example
of
this
there.
If
people
aren't
aware
and
in
the
community
repo
there's
a
proposal
for
a
vertical
pod,
auto
scaling,
I
think
that
would
be
super
useful
tool.
C
So
we
don't
have
to
expose
CPU
and
memory
parameters
for
every
home
chart
or
or
provide,
like
small,
medium
large
t-shirt,
size
kinds
of
separate
templates
for
for
applications,
regardless
of
whether
what
the
templating
scheme
is
and
then
finally
transformation.
If
you
look
at
the
open
shift
templates
or
the
helm
sharks,
the
vast
majority
of
parameters
are
just
customizing
resources
and
label
names,
resource
names
and
legal
values.
I
think
things
like
that,
we
can
just
automate
our
customization
strategy
that
actually
understands
the
needs
resource
format,
because
that
format
is
key
to
kotas
and
stable.
C
So
I
also
also
think
you
know,
coming
up
with
some
common
conventions
is
already
happening
within
the
helm.
Try
repository
for
what
labels
we
recommend
will
simplify
things,
especially
for
new
users,
so
on
a
big
fan
of
domain-specific
transformation,
either
in
the
client
tool,
or
you
know
increasingly
in
the
server
we
there's.
C
You
know
I
won't
go
through
all
these,
but
you
know
some
concerns
are
known
at
development
time
when
the
application
is
written
and
when
the
configuration
is
written,
that's
going
to
be
shared,
but
a
lot
of
things
really
depend
on
the
context
in
which
the
application
is
deployed
and
having
a
mechanism
where
you
can
easily
add
those
additional
configuration
me
without
14
or
parameterizing.
The
core
approach,
that's
going
to
scale
better.
C
There
are
some
key
kinds
of
automation
that
the
mechanism
need
to
integrate
with
I
mentioned
the
service
broker
earlier
and
people
aren't
aware,
there's
a
Service
Catalog
efforts
trying
to
provide
a
Service
Catalog
based
on
the
open
service
broker,
API
that
was
spun
off
as
a
dependent
project
by
Cloud
Foundry.
The
Service
Catalog
fig
has
been
working
on
that
and
a
gay
Authority
has
a
helm
based
service
broker
and
there
are
few
other
prototype
service
brokers
that
exist.
C
C
You
know
for
people
who
are
new
to
the
system.
Typically,
an
in
production
people
I,
don't
really
see
people
managing
applications
through
through
a
GUI,
but
it
is
a
great
demo
mechanism
and
a
great
way
to
learn
the
system
and
then
again
other
other
complex's
configuration
like
secrets
and
application
configuration
that
sometimes
bc
generated
to
compose
together
application,
specific
parameters
and
environmental
ambers.
Since
most
of
the
applications
weren't
designed
to
be
cloud
native
and
have
a
lot
of
parts
of
the
environment,
dynamically
change,
so
universal
universal.
What
do
I
mean
by
that?
C
There
are
a
lot
of
different
dimensions
where
we
want
the
declarative
application
management
mechanism
to
apply.
You
know
for
sure
stainless
applications
which
a
lot
of
platform
as
a
service
systems
have
traditionally
focused
but
also
faithful
applications.
Other
classes
of
workloads,
people
want
to
use
I,
believe
the
same
tools
and
workflow
for
all
the
different
kinds
of
applications,
at
least
the
core
of
the
workflow.
C
Maybe
they
need
additional
components
in
some
scenarios
or
for
some
types
of
work
with,
but
we
sort
of
recorded
it
will
use
totally
different
tools
and
have
a
totally
different
experience
for
different
applications.
They
need
to
deploy
built-in
api's
and
extensions.
So
for
some
systems
like
dr.
compose,
is
a
good
example
of
this
to
create
a
configuration
language
that
is
different
from
the
API.
That's
creates
a
burden
to
actually
extend
it
for
the
additional
kinds
of
api's.
C
That
gets
a
belt
that
you
might
want
to
express
some
configuration
for
and-
and
the
really
hard
case
I
think-
is
extensions.
Super
Nettie's
we're
trying
to
make
more
and
more
extensible
at
silicate,
more
custom
automation
by
its
users
and
those
things
will
be
difficult
to
express
in
a
hand
design
manually
designed
configuration
language.
That's
different
from
the
API
mechanism,
the
Stoke,
an
off-the-shelf
ass.
C
So
but
still
gaps
is
what
I
call
applications
that
people
write
themselves
and
off
the
shelf
is
things
more
existing-home
charts
where
there's
an
existing
application
that
you
will
just
want
to
run
bespoke
applications
are
I
mean
most
people
have
to
run
combinations
of
both
right.
They
have
their
business
logic
in
their
own
stateless
fun.
End,
for
example,
with
the
canonical
scenario
and
then
the
infrastructure
tears
for
caching
storage,
etc,
are
off-the-shelf
components.
C
I
think
people
would
like
to
use
again
the
same
tools
and
workflow
for
both
cases,
but
the
for
the
distinct
application
they
might
want.
Some
help
getting
started,
generating
the
two
benetti's
configuration
rather
than
having
to
copy
pasted
the
ammo
from
somewhere
else
or
write
the
whole
thing
from
scratch
by
hand.
So
I
think
there
are
some
tools
in
the
space.
C
For
example,
openshift
new
app
command
to
keep
control
run
help
with
this
a
little
bit
but
they're
not
really
designed
to
plug
in
to
a
declarative
configuration
workflow
where
you
would
generate
the
configuration
and
maybe
customize
it
and
then
check
it
in
somewhere.
I
also
think
you
know
the
same
core
tools
and
workflow
can
serve
both
beginners
and
power
users.
C
The
beginners
would
just
need
more
more
truly
to
help
them
get
started
from
some
kind
of
generic
inputs
or
examples
and
then,
as
they
want
to
grow
into
more
sophisticated
deployments
with
canary
deployments
or
or
what
have
you,
they
could
customize
their
configurations
as
needed.
And
then
you
know
the
thing
again.
The
same
tool
should
work
on
the
laptop
and
in
production.
But
you
know
if
you're
just
iterating
on
your
laptop,
you
might
not
want
to
have
to
check
in
every
version
of
configuration
file.
So
we
test
it
so
and.
C
D
C
And
patterns
to
help
people
take
those
in
turn
them
back
into
generic
templates,
of
whatever
form
they
are,
is
also
really
important,
because
inevitably
new
users
start
moving
the
iterative,
narrative
style
generation
and
then
down
there.
We
want
them
to
be
able
to
do
declared
it.
So
then
Julian
course,
kitchen,
yep
I,
agree
that
I
have
a
picture
on
the
on
the
next
slide.
I
think
we're
I'll
talk
about
where
I
speed
up
to
the
end,
yeah
so
again,
api
dates.
C
I
think,
because
we
need
to
its
work
for
all
the
communities,
resources
for
extensions
and
so
on.
Whatever
schema
is
used
for
configuration
in
the
general
case,
I
think
there
are
some
specialized
cases
where
things
like
the
composed
tool,
which
generates
to
know
these
configurations
from
the
doctor's
report.
Format
useful.
But
if
you
want
to
use
or
group
potentially
grow
into
the
full
power
of
committees,
what
configuration
schema
needs
to
be
so
literal
api
schema
as
we've
predominantly
used
in
any
ecosystem
so
far,
or
at
least
automatically
derived
comet.
C
For
instance,
I
we're
trying
to
generate
open
API
clips
are
greater
and
greater
fidelity.
That
is
why
someone
from
puppet
hedging
demonstrated
generating
puppet
like
schema
from
the
swagger
spec,
so
something
like
that
could
work
if
you
maintain
it
by
hand
I'm
going
to
work
for
extensions,
that
it
is
hard
to
keep
up
the
pace
of
aepi
evolution.
Go
in
my
experience
again,
anything
by
hand
just
isn't
going
to
work,
but
it
doesn't
actually
be
literally
guys.
C
You
know,
though
you
know
so
I
guess
another
example
would
be
if
you
want
to
write
configuration
in
using
affluence
style,
HD
I
like
the
fabricates,
Java,
client
library,
you
could
eventually
do
that.
I
think
the
configuration
needs
to
be
person
according
to
the
API
version
again,
I
had
to
keep
up
with
the
rate
of
API
key
elements
on
extensible
and
thi
is
extended.
I
also
proceed.
C
So
we
have
been
working
on
cue
control,
apply
as
a
way
of
enabling
declarative
intensity
composed
with
automatically
set
attributes
the
way
to
control
apply
work.
The
people
are
available
busy.
It
keeps
track
of
the
configured
state.
So,
as
user
changes,
the
configured
states
so
patches
can
be
applied
to
the
live
state,
but
the
anything
that
was
never
specified
by
the
user
is
not
considered
me
under
the
domain
of
the
configured
states
and
the
the
runtime
States
doesn't
get
trashed
by
the
tool.
C
Just
you
keep
control,
replace
that
says
everything
if
there
were
dynamically
said
values
by
some
one-time
automation,
those
values
we
get
trashed.
So
eventually,
I
would
like
to
actually
move
support
for
apply
as
the
server
and
they
configured
states
are
the
API
make
this
work
for
all
clients,
but
now
for
various
historical
Wiggins
with
currently
done
in
the
client,
we're
trying
to
fix
the
bugs
that
exist
and
keep
control
apply,
but
I
see.
C
This
is
an
approach
to
synchronize
date
between
different
layers
of
Asia
as
many
places
in
the
system,
for
example,
if
we
ever
do
in
place,
updates
and
deployments
are
one
of
the
other
workload
controllers.
I
would
imagine
we
be
using
this
world
already
have
discussed
doing
this,
using
this
approach
for
demon
set
to
figure
out
what
changes
have
been
made
or
whether
changes
have
been
made
to
the
template
and
if
we
create
any
kind
of
template
resource
similar
to
the
open
shift
templates,
incra
Nettie's
I
would
expect
to
use
this
approach
to
synchronize
between.
C
C
Is
important
to
be
couple
degeneration
and
transformation,
steps
from
orchestration
of
creating
resources
and
what
I
call
reconciliation,
which
is
just
taking
the
declared
configured
state
and
synchronizing
it
with
the
live
state
in
the
server?
And
that's
because
you
know
these
pieces
are
generally
in
Italy
useful
and
many
different
scenarios
and
I
don't
want
to
get
in
a
situation
where
people
feels
compelled
to
use
a
certain
a
tool
or
workflow
just
because
it
provides
the
one
piece
of
functionality
that
they
really
need.
C
We
have
that
problem
inside
Google
in
its
configuration
system
and
it
led
to
all
sorts
of
crazy
scenarios
where
things
were
a
lot
more
complex
than
it
used
to
be.
This
is
why
we've
tried
to
keep
keep
control
working
just
on
vanilla,
API
resources,
for
example,
so
it
can
just
use
the
bare-bones
reconciliation
step
with
the
server
and
more
complex
tools
can
be
built
on
top
of
that.
So
the
idea
does
well.
I
have
in
mind,
looks
like
this.
So
you
know,
regardless
of
whether
it
is
the
spoke
application
or
an
off-the-shelf
application.
C
You
know
people
are
starting
from
scratch
or
you're
in
San
shooting
an
application
that
is
widely
used.
I
imagine
the
configuration
they
start
from
will
be
some
kind
of
commented
set
of
configuration
files
and
some
get
repository
somewhere,
and
then
we
can
provide
to
link
to
discover
those.
Those
applications.
I'm
monocular
looks
great
for
discovering
home.
Charts
is
just
one
example,
but
we
can
provide
some
kind
of
tooling
for
just
for
finding
the
right
starting
point
and
then
but
I
envision.
C
The
first
steps
they'll
do
is
they'll
for
kits
from
the
git
repository
and
transform
it
through
a
combination
of
automatic
cooling,
customizing
super
common
things.
Look,
you
know
like
names
and
labels
or
what
have
you
and
maybe
editing
it
by
hand
both
the
raw
templates
and
whatever
parameter
values
are
expected
to
be
used
and
the
reason
for
a
fork
and
that
step,
especially
if
it's
been
off
the
shelf,
is
again
that
will
make
it
easy
for
them
to
maintain
patches
so
that
they
have
some
reasonable
way
of
consuming
updates.
C
But
if
I
were
using
any
kind
of
off-the-shelf
templates,
I
would
definitely
force
the
thing.
As
the
first
thing
I
did
and
not
just
run
off
of
templates
in
even
if
I
didn't
cry
initially
to
change
anything,
there's
always
the
potential
to
need
to
change
parameters
later.
You
know
again,
especially
things
like
operational
parameters.
If
you
don't
have
automatic
system
for
doing
that,
or
you
know
even
simple
things
like
how
the
the
probing
interval
for
a
liveness
probe
or
something
like
that,
the
need
to
fork
in
real
scenarios
is
extremely
extremely
common.
C
So
I
just
do
that
first,
making
EE
to
maintain
patches
and
then
check
in
the
configuration
into
my
own
repository
somewhere,
not
the
same
as
the
code
repository
but
check
into
my
own
apakah
tori.
So
I
have
control
over
it
and
then
automatically
reconcile
with
the
with
the
communities
cluster.
From
that
you
know,
box
wrote
about
this
in
there
and
their
blog
posts
they
have
an
automatic
applier.
C
C
The
admission
control
extensions
are
one
example
of
this
to
allow
administrators
to
plug
in
their
own
custom
policies,
whether
it's
scheduling
pods
to
dedicated
nodes
or
whatever
it
is.
They
want
to
do
like
the
pod
presets
for
injecting
environment
variables
or
volumes,
but
that
those
posts,
creation,
transformations
I
imagine
are
going
to
simplify
the
configuration
that
users
have
to
write
by
hand.
C
So
you
know
this
is
the
basic
work
flow
data
flow
and
I
have
in
mind
the
additional
tooling
or
transformations
I.
Imagine
ski
at
the
first
step
at
the
bottom
left
you
know,
and
if
we
need
to
build
tools
to
also
generate,
or
fact
you
know
add,
parameters
to
configuration
or
something
like
that
to
make
it
you
here
for
people
to
build
templates
from
configurations
are
running
from
I'm,
all
in
favor
of
that
as
well,
but
yeah
any
custom
generation
or
stings
like
that.
C
D
With
people
have
questions
hi,
this
is
my
free
and
I
have
questions
yep.
Of
course,
I
did
so
so
I
know
you
talked
about
in
the
workflow
here
even
on
this
last
slide
of
you
know,
I
find
another
configuration
and
then
I
for
kit
and
I
check
it
into
my
own.
That's
one
of
the
possible
roads
but
I
know
another
road
that
folks
like
to
have
is
to
say:
okay
I'll
start
with,
maybe
my
configuration
file
and
I'll
declare
my
parents
and
then
I'll
declare
just
my
override
values
might
change.
D
Maybe
my
parent,
I
say:
okay,
my
parent
at
this
version,
or
this
version
rates
right,
so
you
got
versioning
in
there
and
then
I
can
say:
okay
through
my
override
parameters,
for
that
now
go
to
point
and
your
tool
and
go
fetch
the
parent
from
wherever
it's
at
know
that
apply
the
differences
and
then
use
that
state
rather
than
forking
it
it's
a
different
style
workflow.
Then
this
get
workflow.
What
do
you
think
of
something
like
that
in
this
flow
as
well.
C
C
Not
parametrize
in
the
specification
you
consumed
from
the
original
source,
you
know,
then
you'd
have
to
go
fork
it
at
that
point
and
that's
a
lot
riskier
than
having
your
own
Fork
that
you
have
control
over
or
another
scenario
is
you
know
if,
like
the
whole
business
with
dr.
tags
right
where
people
push
additional
images
at
the
same
tag,
if
you
have,
if
that
is
a
possibility,
where
the
version
you're,
depending
on
to
change
without
your
awareness
that
could
potentially
put
your
production
deployment
at
risk.
C
D
You
know
we've
done
that
too.
We
have
done
both
ways
over
the
course
of
just
the
last
five
years,
and
so
I've
experienced
the
positives
and
then
negatives
to
both
sides
of
the
equation.
It's
more
of
the
idea
of
having
that
freedom
built
into
the
flow
to
do
some
of
these
things,
so
both
sides
are
enabled
that's
kind
of
where
I
am
asking
this
question.
Not
so
much
to
say
you
should
do
one
or
the
other.
D
You
know
I
think
people
can
make
our
own
business
decisions,
at
least
it's
good
to
convey
the
positives
and
negatives
of
both,
but
to
enable
that
freedom,
sure
yeah
I'm,
all
in
favor
of
freedom
and
then
the
other
question
that
I
had
was
more.
If
I
were
going
to
use
helm
today
in
a
CI
CD
workflow.
What
would
need
to
change
to
enable?
Maybe
the
idea
of
your
philosophy
here
to
work.
C
Well,
I
think
people
could
fork
the
chart
repository
right,
it's
a
little
bit
because
of
limitations
of
kits
and
gab
a
little
bit
inconvenient
to
you,
so
I'm
actually
getting
kicked
out,
but
you
could
potentially
use
it
in
this
fashion.
I
think
they're.
You
know
if
you
want
to
instantiate
charts
in
multiple
different
contexts.
I,
don't
think
that
values,
file,
charge
of
structure
is
really
set
up
to
facilitate
that
so
I
think
there's
some
minor
tweaks
that
need
to
be
made.
Okay,
all
right,
yeah.