►
From YouTube: Cloud Foundry for Kubernetes SIG [December 2019]
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
I
think
we
should
get
started
so
SWANA,
please
roll
the
recording,
Thanks,
so
welcome
to
this
week's
Cloud
Foundry,
I'm
kubernetes
specially
twist
book
call
just
last
second
check.
So
no
change
in
the
polls
for
the
topic
today,
so
the
topic
will
be
unsurprisingly
irrelevant
for
cloud
foundry
on
kubernetes.
B
A
C
So
I
I
can
mention
that
the
police
integration
project
team
has
been
doing
some
research
primarily
with
the
component
teams
that
they'd
be
supporting
with
component
integration
over
the
past
couple
of
weeks
and
they're
preparing
see
you,
except
on
this
topic
on
Thursday
as
a
team.
So
I
expect,
after
that,
we'll
start
seeing
more
deliberate
activity
from
them
again
focused
primarily
on
supporting
the
component
teams
as
they're
driving
out
their
own
kubernetes
packaging
and
deployment
artifacts
for
their
subsystems.
I.
D
Was
wondering
if
there's
any
way
where
we
can
collaborate
on
QPS
as
well,
we're
looking
to
maybe
pair
with
teams
like
a
Rivini
or
Cathy
to
integrate
our
CI
systems
and
also
to
integrate
on
other
levels
to
like,
but
Irene?
If
you
have
parkland
in
testing
now,
so
we
could
actually
replace
the
bas-reliefs
with
home
chart.
So
yeah
we're
looking
to
pair
we're
looking
to
talk
to
people
about
this
and
connect
the
I
systems
like
when
you
get
an
Iranian
chart.
You
should
also
get
a
new
cube,
CF
bump
things
like
that.
E
I'm
sorry
I
mean
be
there.
Just
your
flag,
I
think
I
can
show
a
little
bit
more
about
the
relevant
section
since
tonichi
and
I
have
been
working
quite
a
bit
with
psy.
So
you
know:
they've
been
working
with
some
excise
been
working
more
with
the
component
teams
in
terms
of
what
he
would
need
and
so
on,
and
then
tonichi
and
I
have
been
working
more
on.
Hey
Marco,
see.
E
Like
how
we
can
fulfill
those
with
you
know
to
change
that,
some
of
which
we
had
explored
as
part
of
our
CFS,
like
life
cycles
sake
that
had
happened
a
while
back
and
some
of
their
tools,
ed
she's,
also
been
working
on
so
I
wished.
I
was
here
to
talk
a
bit
more
about
this,
but
currently
one
of
the
things
that
we're
considering
is
incepting
on
what
it
would
look
like
to
use
the
k
for
Tina's
tooling.
E
E
So
just
each
of
them
had
publishing
their
own
images
and
their
own
config
files,
and
then
us
in
on
the
rel
inside,
putting
it
all
together
and
creating
those
artifacts
and
maybe
also
creating
some
see
I
toss
such
that,
if
you,
if
you
weren't
so
inclined
to
like
use
the
the
tooling
directly
yourself
like
you'd,
so
be
interfacing
with
like
see.
I
toss
such
as
like
a
fresh
deploy
or
an
upgrade,
or
something
like
that,
so
in
light
of
that
or
and
sighs,
maybe
here
no,
he
went
so
okay,
okay
you're
here.
A
B
I
guess
one
thing:
one
thing
that
we
just
crossed
through
play
and
I
guess.
Maybe
we
can
also
talk
a
little
bit
about
some
of
the
differences
between
let's
say
on
template
and
white
sheets
T,
poor
people
or
and
I
really
familiar
with
y
TT
I
guess
how
many
people
are
familiar
with
y
TT
on
the
call
you.
B
I
don't
know
yeah
it's
kinda
hard
to
count
on
zoom,
maybe
that's
a
maybe
that's
a
new
product
right
there,
video
counting
well
I,
see
a
savanna
before
before
I
kinda
show
you
know
a
typical
demo
that
I
show
with
some
of
these
tools.
I
guess
maybe
to
briefly
discuss.
You
know
what
are
the
differences
and
what
is
let's
say,
Y
DT,
so
the
the
workload
that
Luke
is
mentioning
is
based
on
several
tools:
ytt
being
one
of
them.
B
That's
for
templating
and
overlaying
kay
build
is
another
tool,
that's
part
of
the
workload
to
deal
with
image,
references
and
resolving
images
to
lets
it
digests
and
helping
you
build
certain
things
from
docker
images
or
other
ways
of
building
images.
And
finally,
cap
is
the
last
tool
that
is
responsible
for
deploying
all
of
that
stuff
to
kubernetes.
So
the
typical
workflow
really
just
looks,
like
you
know,
run
ytt
Piper
into
okay,
built
Piper
in
choke
app
right,
but
more
specifically
ytt.
B
You
know
what
are
the
benefits
of
having
ytt
read
the
benefits
of
having,
for
example,
capper
fairly
obvious.
You
know,
keep
CTO
is
not.
You
know
dealing
with
a
lot
of
the
you
know,
problems
that
one
would
you
know
would
see,
for
example,
installing
CR
DS
before
installing
other
things
or
installing
namespaces,
before
installing
things
and
maybe
labeling
resources,
so
that
that
that
stuff
is
fairly.
You
know
well
known
within
the
community.
B
The
the
templating
side
is
a
little
bit
more
interesting,
at
least
to
me.
One
of
the
reasons
that
ytt
came
about
is
I
was
playing
around
with
helm
and
using
it
for
a
few
things,
and
it
was
actually
quite
annoying
to
use
for
a
bit
more
complicated
scenarios,
but
even
for
some
of
the
simple
scenarios,
it
was
a
surprisingly
error-prone.
B
So
you
know
what
is
held.
Template
right,
he'll
jump
lit
is
based
on
ghost
jump
text,
templating
package
right.
So
it's
it's
textual
templating!
It's
not.
They
were
a
role
that
your
templating
llamó
now
the
templating
syntax-
that
the
language
itself
that
that
you
using
within
that
ghost
you
know
framework
if
you
will
is
also
not
similar
to
other
I
mean
it's
similar
ish,
but
it's
not
as
similar
and
as
easy
to
use
as
some
other
languages.
B
So
you
know,
given
those
two
things
right,
that
it's
textual
templating,
given
that
the
template
syntax,
is
a
little
weird.
You
know
you
start
running
into
certain
problems
of,
for
example,
you
have
to
manage
the
white
space
manually
right,
so
you
have
to
throw
in
bunch
of
indents,
and
you
have
to
throw
in
I.
Don't
know,
quotes
around
certain
values
of
what
I
write.
So
that's
kind
of
the
main
you
know,
annoyance,
/
challenge
that
I
had
with
was
the
with
the
health
templating
approach.
B
You
know,
aside
from
that,
I
guess,
you
know
just
kind
of
a
you
know,
looking
into
deeper
into
some
of
the
functionality,
so
helm
doesn't
necessarily
offer
any
facilities
to
reuse
some
of
the
configuration,
so
it
doesn't
really
lead
towards.
You
know:
modules
configuration
there
is
a
feature,
for
example,
called
library,
Charles
and
he'll
be
three,
but
it
doesn't
really
improve
that
situation
by
a
lot
and
ultimately
it's
because
the
ghost
ex-champ
learning
libraries
are
not
meant
for
this
kind
of
a
thing.
B
So
so
that's
that's
kind
of
the
the
basic
feeling
around
they
held
template
come
on
right
now.
Ygt
was
designed
specifically
to
address
some
of
those
challenges
right.
So
why
did
he
actually
starts
off
by
being
yeah
malware
right?
The
templating
itself
is
done
on
top
of
the
mo
data
structures.
It's
not
checks
templating
it.
It
doesn't
actually
know
that
any
of
this
is
plain
text.
B
So
it
knows
how
to
quote
things
you
know
by
itself.
You
don't
have
to
do
any
kind
of
crazy
manual
white
space.
Indentation
like
you,
would
have
to
do
in
hell,
because
again
it's
not
textual
templating.
It
just
knows
that
hey
this
set
of
values
goes
after
this
particular
key
white.
Is
he
also
offers
a
Python
like
language?
It's
a
it's!
A
language,
that's
taken
from
basil
the
build
tool
from
Google.
B
They
have
a
standalone
library
for
using
the
language
called
star
lock,
and
you
know
that
also
offers
quite
a
few
benefits
in
terms
of
like
friendliness
in
how
do
you
actually
type
in
some
of
the
template
code,
and
it
also
offers
you
know
a
few
nice
side-effects
that
you
know
your
templating
environment
is
fully
turmeric.
There
is
no,
you
know
unless
you
actually
explicitly.
You
know,
make
it
available
through
the
libraries
or
something
like
that,
but
obviously
I
don't
want
to
do
that.
B
It's
a
you
know
doesn't
have
any
way
to
to
access
network
this
any
of
that
other
stuff
right.
So
the
templating
itself
is
more
or
less
just
a
compute
kind
of
a
task
right
and
I
guess.
The
other
interesting
feature
of
ytt
is
that
there
is
an
overlay
functionality
available.
So
it's
somewhat
competitive
to
you
know
both
shops
files,
customized
but
I-
think
the
syntax
is
a
little
bit
friendlier
and
more
more
flexible
in
some
cases.
B
Comparison
between
those
two
tools,
ytt
also
does
support
a
notion
of
libraries
and
has
you
know
you
know
it's
built
around
some
of
those
primitives
that
allow
better
the
use
of
your
configuration,
so
we're
gonna,
see
I,
guess
some
of
that
functionality
and
play
as
well
so
yeah
so
I
think.
Unless
anyone
have
particular
questions,
we
can
maybe
a
check
out
the
demo.
B
D
B
Monster
office
sublime
over
here,
so
the
idea-
and
you
know
quandary
right-
is
that
there
is
lots
of
different
components
that
interplay
with
each
other
it
which
is
fairly
standard
in
a
you,
know
in
a
modern
application.
There's
multiple
services
right
so
to
represent
that
what
I
have
is
actually
three
folders
over
here:
CC
DD
and
you
a
obviously
this
is
not
like
video
confidently.
So
there's
no
cloud
to
run
or
anything
like
that,
but
it's
representative
of
a
few
components
that
makes
this
thing:
Cloud
Foundry.
D
B
The
idea
behind
some
of
this
layout
is
that
this
individual
components
more
or
less
for
the
same
pattern
right.
There
is
config
folder
that
contains
some.
You
know
templates
you
as
small
as
thing
kind
of
a
thing,
and
then
there
is
also
user
provided
folders.
So
this
is
I
just
do
it
in
for
demo
there,
but
you
know,
let's
say,
there's
some
kind
of
a
database
secret
or
something
like
that.
That
would
come
externally
from
somewhere
else
wouldn't
be
checked
in
your
repo.
B
So
if
we
take
a
look
at,
let's
say,
Cloud
Controller
we'll
see
that
there
is
a
thing
called
value.
Siamo
and
valleys.
Gmo
is
one
of
the
files,
that's
part
of
configuration
and
it's
annotated
through
ytt
syntax
as
data
values.
So
this
indicates
to
ITT
that
this
is
the
input
to
the
system
input
to
these
templates
and
you
can
have
multiple
dos
variables-
data
values
parts.
So
you
know,
as
as
one
would
imagine,
there's
a
few.
You
know
generic
configurations
for
cloud
controller,
maybe
like
a
namespace
where
to
deploy.
B
You
know
in
coronaries
a
you,
a
URL,
that's
by
default
is
empty
because
what's
new,
because
you
know
it
doesn't
know
how
to
connect
to
UA,
but
it
needs
one
what
else
some
database
stuff
right,
so
the
database
URL
username
password
all
that
stuff,
so
the
other
configuration
file.
That's
most
interesting
is
this
guy
over
here.
The
bulk
of
it
is
actually
this
low
section
where
there's
a
services
kubernetes
object.
There
is
a
deployment,
kubernetes
objects
and
there's
a
config
map.
Kubernetes
object.
B
E
B
We're
saying
that
there
is
a
label
CC
and
its
value
empty.
There
is
also
a
config
map
over
here
in
the
config
map.
Let's
say
you
know
as
coming
from
a
boss
world.
There
is
a
lot
of
you
know:
applications
that
consume
Jason
or
no
files,
so
config
map
would
have,
let's
say,
configure
a
monkey
and
under
it
it
needs
to
be
a
full-blown
llamó
configuration
file
right.
B
So
one
interesting
thing
with
ytt
can
you
know
compose
some
of
the
libraries
that
has
built
in
so
here
we're
saying:
let's
yeah
Molly
encode
the
result
of
the
config
file
function
and
a
config
file
function
contains.
You
know,
I
talked
it
over
here
that
dips.
You
know
the
configuration
that
we
actually
care
about,
like
you,
a
URL
GB
or
all
that
stuff.
So
now
over
here,
I'm
also
seeing
still
function
called
value.
That's
actually
something
I've
defined
as
little
shorthand
over
here.
You
know
by
default.
B
You
know
what,
if
you
exposes
this
data
library
data
values
and
pull
out
some
certain
values,
but
I
decided
to
actually
write
in
your
function.
That
makes
the
assertion
the
error
message.
You
look
better
because
I
wanted
to
see
value
is
not
empty,
so
there
is
that
usage
of
that
value
call
for
each
one
of
those
things
so
anyway,
so
that
that's
our
standard
configuration
file
may
look
like,
let's
say
right
and
at
the
end
run
it
and
my
computer's
a
little
bit
slow.
B
So
hopefully
I'm
not
wagging
too
much
so,
let's,
let's
actually
CD
into
config,
which
it
seemed
ABC
and
say
ytt,
config
and
most
likely
all
fail
right
now
and
it's
telling
us
that
hey
look
you
a
URL
is
missing
right.
It
needs
to
be
there
right
and
that's
you
know,
there's
your
stack
trace!
That's
telling
us!
You
know
why
it's
failing
and
because
there's
a
few
lines
of
code,
it's
asking
us
right!
So
I
don't
actually
remember
if
I
populated
this
file
correctly.
It
looks
like
this.
B
So
if
we
take
a
look
at
user
provided
values,
note
that
I've
I've
specified
you
a
URL
to
just
be
here
a
URL
and
then
I've
specified
some
other
diamond
values
for
it
to
database.
Stop
I,
actually
don't
think
clients
is
necessary
for
Club
controller
must
be
assembled
robocopy
face
quite
out
there,
so
yeah
so
note
that
it
only
needed
to
fill
in
the
values
that
are
not
present
in
the
other
configuration
right.
B
So
there
was
value
siamo
right,
so
the
other
values
llamo
specified
all
kinds
of
other
things,
and
but
I
had
a
few
keep
the
missing
and
then
in
this
file
is
specified
only
the
missing
stuff.
Now
you
know,
for
example,
to
throw
in
some
interesting
intrigue
over
here.
Let's
say
if
I
would
have
a
type
over
here
password
instead
of
password,
I
would've
run
the
same
command
over
here
where
actually
would
have
blown
up
telling
us
that
hey
I
don't
know
how
to
apply
this
past
four.
You.
C
B
Configuration
on
top
of
the
other
data
values
file
right,
so
why
did
you
try
to
be
as
protective
of
the
user
configuration
writers?
You
know
so
this
that
kind
of
tries
to
prevent
mistakes
so
anyway,
so
you
know
we
we
seen
that
our
Cloud
Controller,
let's
say
Yama
file,
that
generated
right
and
you
know
that
does
a
fairly.
You
know
simplistic
kind
of
a
template
that
you
know
that
give
us
this
nice
configuration.
We
can
finally
go
ahead
and
deploy
so
I
also
included
this
release.
B
Folder
release
folder
has
this
little
old
snippet
that
tells
kay
build
which
image
to
use
too
bad,
though
that
I
don't
think
this,
because
I
blew
away
my
mini
cube.
This
images,
I,
don't
think,
are
no
longer
true,
so
iam
pushing
to
an
your
registry
so
anyway,
so
I'll
mention
the
I'll
mention
this
in
a
little
bit,
so
we
Clow
controller
uas,
pretty
much
the
same
exact
thing.
You
know
it's
not
that
different.
We
got
some.
We
got
some
configuration
going
on
here.
B
Actually
it
is
slightly
different
and
there
is
my
slightly
different
is
note.
That
else
is
clients
right
and,
as
you
all
know,
you
a
you
know
as
a
bunch
of
clients
that
needs
to
be
configured
for
cloud
controller
and,
let's
say
some
of
the
other
services
to
operate.
So
this
is
where
this
input
would
go
into
rain,
and
if
you
actually
take
a
look
at
the
configure
mo
here,
we
have
a
tip
setup
over
here.
B
We
have
a
config
file
that
that's
ultimately
going
to
be
in
a
config
map
right
and
we're
also
saying
over
here
that
for
each
one
of
the
clients
that's
given
through
to
us
to
the
data
values,
we
want
to
actually
make
little.
You
know
key
value
kind
of
key
map
with
value
kind
of
thing,
and
so
I
was
just
kind
of
trying
to
replicate
what
the
current
you
a
format
looks
like
and
how
that
would
be.
You
know
configured
right,
I
say
otherwise.
There
is
not
really
anything
different
going
on
over
here.
B
One
other
interesting
thing
to
note
over
here
is
that
you,
a
let's
say,
will
be
deployed
to
kubernetes
right,
so
there's
going
to
be
some
kind
of
a
URL
that
you
should
access
you
a
right
and
then
kubernetes
the
urls
or
default
genus,
this
one's
our
namespace
trait,
so
their
names
pasted
under
there.
You
know
namespace.
So
I
just
for
no
reason
I
just
decided
to
put
a
little
method
over
here
called
an
S.
You
know,
obviously
you
can
inline
it
over
here.
B
So
yes,
that's
the
URL,
that's
generated
by
this
UAE
configuration
right
and
if
we
would
have,
you
know,
run
ytg
again
with
this
thing.
You
know.
If
we
go
ahead
and
say
on
a
CD
out
here
in
to
you
a
again,
you
know
it
works,
but
you
know
some
of
the
values
are
empty
because
we
haven't
specified
anything
right.
Our
user
provide
it.
Values
are
empty
over
here,
except
database,
I!
Guess
alright.
So
we
have
these
few
components.
The
database
actually
is
that
me
over
here.
B
It
doesn't
actually
have
anything
just
as
a
config
map,
but
one
interesting
thing
about
databases
is
that
somehow
someone
needs
to
tell
them
that
hey
provisionally
some,
you
know,
you
know
username
password
to
the
database
value
things
like
that
it
so
the
values
llamo
for
this
guy
you
know
takes
a
bunch
of
databases
and
then
for
every
single
database.
It
will
go
ahead
and
you
know,
extract
the
name
username
password
and
make
some
yellow
for
it.
Wait.
So
that
really
depends
how
you
structure
database
right.
D
B
B
It
will
actually
generate
a
lot
of
this
configuration
now.
It
will
generate
the
configuration
for
each
one
of
those.
You
know:
Cloud
Controller,
you
a
database.
It
was
questioned
together
but
note
that
actually
a
few
things
happened
over
here.
So
if
you
look
at
the
Cloud
Controller
config,
it
actually
has
the
you
a
URL
over
here.
You
know
the
proper
namespace.
B
If
you
look
at
let's
scroll,
let's
see
so,
if
you
look
at
the
database,
the
database
actually
has
a
list
of
the
databases
that
needs
to
be
provision
for
cloud
controller
and
for
you
a
now.
If
you
look
at
you
a
config,
it
also
has
appropriate
database
configuration
right.
So
database
has
the
URL
and
particular
username
and
password.
You
know
one
could
also
have
like
a
database
name
or
some
front
array.
B
So
how
did
that
all
come
together
right?
So
there
isn't
really
no
obvious
ways
of
how
to
combine
all
of
that
stuff
and
share
the
information.
So
there's
a
little
bit
of
temporary
needs
to
happen,
so
this
is
a
little
bit
of
sampling
it
through
in
in
the
OL
yeah
mo
file
and
we'll
take
a
look
at
on
a
second
now,
given
that
some
of
this
configuration
requires
some,
you
know,
input
from
the
user.
B
Here's
an
example
Wiley's
file
and
here's
what
a
user
had
to
provide,
for
example,
on
top
of
this
alright,
so
keep
in
mind,
a
ytt
is
not
meant
to
be
used
for
secret
generation.
You
just
accept
some
input
value
C
does
not
many
facilities
to
make
any
network
requests
or
generate
anything,
or
you
know,
there's
not
even
a
way
to
do
any
kind
of
randomized
thing
right.
B
So
if
you
have
to
throw
in
some
username
password,
my
thinking
about
it
is
that
there
is
a
responsibility
for
another
tool
that
knows
how
to
generate
some
secrets
or
interconnect
with
something
right
so
anyway.
So
in
this
particular
case
we
just
have
to
throw
in
some.
You
know
cloud
control,
create
a
username
passwords
for
the
database
that
we
want
it
to
be
used.
B
Alternatively,
if
you
didn't
want
to
use
the
database
that
came
with
this
configuration,
you
know
throw
in
DB
enabled
falls
over
here
and
then
you
know
provide
some
external
database
configuration
right
so
now,
how
does
this
all
Yama
file
look
like
based
on
this
stuff?
So
there's
a
little
bit
of
there's
a
little
bit
of
comment:
we're
not
commenting,
but
validation,
going
on
I
just
wanted
to
kind
of
play
around
with
certain
things
now
one
you
could
always.
B
You
know,
split
up
some
of
this
stuff
into
different
files,
but
I
just
decided
to
dump
it
all
into
one
for
convenience
for
now.
So,
ultimately,
the
end
of
the
the
end
of
the
result
is
this:
is
that
somehow
we
have
a
bunch
of
Cici
configuration
that
we
want
to?
You
know
replace
this
so
table
and
Yama
means
the
document
right,
so
we
want
to
replace
whatever
the
result
comes
in
over
here.
B
We
want
to
replace
this
table
the
document
with
this
set
of
results
that
comes
out
over
here
right
now.
How
did
we
obtain
this
configuration
right?
So
why
Fiji-
and
this
is
developing
feature-
there's
a
few
community
discussions
around
it
and
whatnot,
but
this
is
actually
a
way
to
use
other
ytt
configurations
as
a
library
from
within
another
configuration
right.
So
this
is
why
I
was
kind
of
showing
you
earlier
on
that
there
is
three
folders
ccdb
and
UA
under
ytt
Lib,
and
that's
how
ytt
knows
how
to
find
those
right.
B
So
we're
saying
use
the
library
built
in
module
and
you
know,
grab
the
configuration
from
a
file
system
under
CC
config
right,
so
it
will
grab
this
entire
directory
over
here.
That
has,
you
know,
config
all
that
stuff
great
provide
some
additional
data
values
right.
So
this
is
data
values
that
somehow
you
may
massage
into
it.
B
So,
for
example,
here
we're
saying:
let's
you
know,
let's
throw
in
some
your
you,
a
URL,
it's
doing
some
DB
functionality,
stuff
right
and
then
finally,
let's
call
the
results
on
it
and
that's
pretty
much
it
right,
so
it
kind
of
grabs
it
and
assembles
it
together.
Now
the
benefit
of
this
a
little
bit
more
dynamic
assembling
right
is
that
you
can
actually
throw
in
some
logic
around
it
right.
So
I've
told
you
already
before
controller
needed
the
you
a
URL.
B
So,
for
example,
UA
is
its
own
kind
of
a
configuration
component
right
and
you
can
actually
grab
the
URL
from
the
UA,
so
it
will
actually
execute
this
as
a
function
over
here
from
that
you
a
configuration
set,
and
finally
you
know
you
can
do
some
fancier
things
like,
for
example,
do
you
want
that
DB
to
be
included
in
the
in
the
file
configuration
or
not?
Well,
you
can
kind
of
control
it
through
some
other
data
values
that
you
wanna
inject
right.
So
the
idea
here
is
just
this.
B
These
three
components
need
to
be
kind
of
a
slapped
together.
There
is
some
little
bit
of
exchange
going
on.
You
need
to
massage
it
and
further
in
right,
but
also
you
can
kind
of
see
individual
components
as
big
functions
that
take
in
one
argument.
The
argument
being
the
set
of
you
know,
did
the
data
values
that
they
defied
right
and
based
on
these
data
values
they
produce
some.
You
know
some
some
value
right.
So
that's
pretty
much
what's
happening
in
this
all
diamo
configuration.
B
Yes,
that's
that's
kind
of
it
right.
So,
if
we,
you
know
run
it
this
way
right,
we
can.
You
know
we
can
generate
this
big.
Let's
say
quanta
llamo
there
is
it's
based
on
individual
components
right
so
idea.
There
is
that
individual
teams
will
actually
produce
those
individual
components
right
and
then
there
is,
you
know
someone
else,
let's
say,
maybe
a
relevant
team,
maybe
a
collaboration
between
those
component
teams
where
they,
you
know
bringing
those
different
components
together,
glue
some
of
these
pieces
together
and
offer
it.
B
As
the
you
know,
it's
the
final
templating
artifact
right.
So
one
thing
that
I
also
mentioned
earlier
on
is
I
included.
This
release
configuration
over
here
the
image
ml
or
whatnot
right.
So
the
idea
behind
this
stuff
is
that
your
CI
is
actually
building
an
image
rate.
Doing
some
integration
testing,
whatever
with
this
configuration
and
somewhere,
maybe
in
the
repo,
maybe
somewhere
else,
sharing
that
this
is
the
image
that
got
built
right.
B
Ideally,
a
relevant
team
wouldn't
have
to
be
rebuilding
images
from
source
right
I
mean
they
could,
if
they
wanted
to,
but
ideally
they
wouldn't
have
to
read,
so
that
is
that
you've
actually
shared
this
digest
reference
to
some
location
in
some
registry
here
I
was
just
using
a
local
doctor
daemon.
So
that's
why
this
URL
looks
a
little
funky
and
finally,
when
you
were
when
you're
running
your,
how
you
call
this
when
you're
running
your
templates
right,
you
can
actually
throw
in
these
pre-built
images
into
it
and
then
cable
gladion.
B
You
know
replace
those
images
appropriately.
So
give
you
an
example:
I,
don't
thing
actually
done
this
so
actually
I
did
I
did
actually
have
a
little
script
over
here,
template
pre-built.
So
let's
take
a
look
at
what
the
script
does.
So
this
script
actually
just
runs
ytg
config.
You
know
it
throws
in
the
the
two
pre-build
images
so
for
CC
and
for
you
a
and
then
this
is
just
some
other
configuration
for
the
database.
It's
irrelevant
to
this
discussion,
so
yeah.
So,
let's,
let's
see
what
that
looks
like
right.
B
B
You
know
the
the
cable
tool
will,
you
know,
find
appropriate
image
location
so,
for
example,
this
image
CC
image
will
be
correlated
with
whatever
the
CC
image
is,
and
you
know
we'll
use
that
so
anyway,
yeah
so
I
think
Cooper
has
a
very
good
point,
definitely
spoken
for
a
while
shoulde
ytt
stuff
and
some
of
the
gluing
stuff
going
on.
There
showed
some,
the
component
stuff
how
they
may
look
like.
F
F
F
People
have
been
successful
with
CF
for
years
and
never
needed
the
animal
templating
engine
or
Yama
give
or
take
maybe
four
lines
in
a
manifest
camel
and
people
have
been
successful
with
Bosch
for
years
and
I
mean
they
need
a
Bosch
and
like
a
few
ops
files
to
change
something,
but
they
didn't
need
like
a
full
scale,
programming
language
integrated
into
they
Amal
and
I.
Why
is
this
happening?
F
B
I've
I've
had
a
similar
conversation
with
someone
else
and
what
we're
here,
what
we
have
both
and
the
rap
agreeing
on
is
that
there
is
a
lot
of
so
there
is
this
complexity
floating
around,
and
it's
really
how
you
slice
it
and
dice
this
where
this
complexity
goes,
but
it
doesn't
ultimately
disappear.
So,
for
example,
and
it's
from
my
own
experiences,
you
know
she
of
push
command.
Yes
is
there
and
it
may
look
simple.
B
However,
once
you
start
production
izing
it
the
the
configuration
of
that,
whatever
you
know,
the
app
that
you
pushing
on
whatever
it
still
has
to
come
from
somewhere
and
that's
you
know
typically
not
included
in
that
default
sea-
of
push
experience,
for
example
right.
So
some
of
our
customers
pillow,
for
example,
rate
they
end
up.
Writing
orchestration,
tooling,
around
CF
push
such
that
they
can
do
it.
B
Let's
say
more
declarative
manner,
right
versus
you
know,
having
to
kind
of
what
translated
into
more
effective
tonight
for
borscht,
specifically
I
can
tell
you
that
some
that
complexity
was
pushed
into
the
releases
right.
Some
that
complexion
was
pushed
into
links.
Some
that
complexity,
you
know,
was
left
to
both
shops
files
right.
So
it
just
really
it's
it's.
It
traveling
this
complexity
around
right
and
the
question
is
well
where,
where
is
the
correct
place
for
it
right
now?
B
The
way
I
think
about
it
is
the
correct
place
for
that
complexity
is
really
depends
on
a
system
right,
because
different
systems
have
different
abstraction
levels.
Weight
kubernetes
provides
different,
you
know,
abstractions,
then
boss,
right
in
kubernetes
world,
and
it's
my
opinion,
the
the
primitives
that
it
gives
you
are
more
granular
right.
So
that
means
that
you
have
to
make
that
complex
your
shifts
on
top
of
those
more
granular,
you
know
blocks
right
now.
B
F
So
the
question
my
head
is
is
whether
so,
for
example,
right
let
example
before
K
native,
etc,
existed
right.
So
like
go
back
here,
anyone
who
looked
like
a
deployment
llamo
was
like
this
is
mental.
There's
like
four
places
you
have
to
put
the
same
value
right.
You
have
to
put
the
tag
in
like
four
places.
What's
going
on
and
and
so
people
solve
that
with
these
templating
libraries
so
like
put
the
tag
in
four
places,
but
then
you
seek
in
a
three.
F
F
Are
we
just
missing
a
few
high-level
concepts
with
which
we
could
express
everything
in
yamo
without
actually
needing
to
template
all
the
ammo?
If
you
see
what
I
mean
like
could
come
like
with
the
right
high-level
concepts,
couldn't
we
just
write
pretty
simple
yeah
mo
because
K
made?
If
does
let
me
like,
avoid
a
lot
of
stuff
I
used
to
have
to
write
so.
B
I
think
this
goes
back
to
the
shifting
of
complexity
and
I.
Think
the
shifting
of
complexity
also
results
in
you
know
changes
and
flexibility,
alright,
so,
for
example,
with
K
natives,
because
they've
taken
a
very
opinionated
approach
to
how
they
define
their
higher
level,
limiting
right
now
for
people
who
want
to
let's
say,
add
some
pressure,
I,
don't
know
whatever
it
may
be
right
to
their
configuration
there,
the
effectively
forced
to
go
other
ways
right.
So
you
know
this
is
the
Carver
constant
challenge
that
K
Native
community?
Has
it?
B
How
much
features
to
add
to
you
know
to
expose
particular
functionality
right
and,
if
there's
actually
a
fairly
prominent
thing
that
happens
in
help
community,
where
a
chart
starts
out
with
five
configuration
values
right,
but
over
time
it
becomes
a
thousand
configuration
values
because
everyone
comes
in
and
tries
to
solve
the
particular
problem
or
a
particular
chart
right.
So
I
think
it
goes
back
to
it.
B
There
is,
there
is
different.
There
is
different,
I
guess,
complexity
in
different
places,
and
it
all
depends
how
you
shifted
around
now
to
to
go
back
to
maybe
what
you're
getting
at
is
hey.
Well,
what
if
we
define
some
high
level
primitive
that
maybe
doesn't
in
software
it
not
let's
a
llamo.
Does
it
all
and
go
something
like
that
right?
Well,.
B
Let's
say
you
do
that
right,
it
all
goes
back
to
again
flexibility
right.
So
do
you
define
a
thousand
options
for
all
kinds
of
things
that
that
the
user
may
want
to
do
to
configure
CF?
You
know
I
think,
based
on
our
boss
experience
in
CF
there's.
There
is
a
bunch
of
options
that
people
do
wanna
configure
all
the
time
right
now.
In
my
opinion,
some
things
work
out.
Okay
with
that,
but
some
things
they
don't
necessarily
work
out
as
well.
Is
that
and
so
it
becomes
equation.
D
I
could
interject
a
bit
we're
running
out
of
time
and
I
wanted
to
talk
about
cubes,
yes
as
well.
You
know
I'm
I'm
interested
in
this
effort,
but
I'm
not
I'm,
not
sure
what
the
timeline
is
with
that
and
we've
had
conversations
in
the
past
about
cube
CF
and
how
it's
going
to
be
integrated.
Maybe
with
Roland
I've,
been
looking
to
get
some
support
from
the
railing
team
to
try
and
get
these
cubes
EF
releases
from
relevant.
A
D
Sure
so
a
quick
update
we
are
bumping
on
the
latest.
C
of
deployment
of
the
Soviet
think
we're
12
that
18
chats
are
passing
Iranians
working.
What
else
we
have
quirks
lengths
being
tested
right
now
which
will
allow
integration
of
native
components.
So
if
a
component
team
wants
to
write
kubernetes
native
stuff,
they
can
integrate
with
the
risk
of
the
Bosch
world
on
kubernetes
for
those
mechanisms,
and
we
can
try
and
replace
each
component
as
we
move
as
we
move
forward.
D
A
Really
wonder
about
the
scope
of
this
session
like
on
this
Sun
so
like
really
far
in
the
future
and
philosophical,
and
it
also
looks
like
an
approach
that
has
already
been
tried,
because
SCF
was
also
trying
to
do
all
the
templating
first
and
it
didn't
really
work
out
well.
So
all
that
really
changes
here
is
the
templating
language,
and
then
it's
just
one
deploying
no
active
components
that
pretty
I'm
not
convinced.
This
is
the
way
to
go.
A
B
One
thing
one
thing
that
I
wanted
to
add:
I
think
that
that
equates
to
this
and
to
what
Jules
asked
is
that
you
know
do
we
need
templating
I,
actually
think
that
in
a
lot
of
cases,
most
people
don't
need
tempering
and
maybe
even
eventually,
including
in
this
one.
The
the
challenge
I
see
here
is
that
there's
a
lot
of
software.
You
know
you
a
cloud
controller.
You
know
keep
on
going
down
the
list
rate
that
does
not
go
intermingle
today
nicely
with
some
kind
of
a
more
singular
configuration
right.
B
You
could
imagine
that
you
know.
If
you
shift
some
of
that
complexity
into
those
individual
components,
then
it
would
work
right.
So
this
is
something
that
let's
say,
other
communities
like
Easter
are
slowly
trying
to
figure
out
if
they
can
solve
some
of
their
similar
challenges
to
configuration
complexity,
but
in
a
more
kind
of
a
hidden
way
right.
So
they're
they're
moving
some
of
that
stuff
directly
into
East
EO
software
components
right
to
deal
with
the
configuration
right.
B
So,
for
example,
if
there
would
be
you
know
if
there
would
be
a
concentrated
effort
to
move
some
of
the
let's
say:
how
does
you
know
cloud
controller?
Accept
all
of
that
configuration?
How
does
you
a
accept
all
the
configuration
and
if
it
becomes
a
little
bit
more
I
would
say
kubernetes
native
right,
then?
Actually
a
lot
of
these
templates
really
become
they're
just
they're.
The
template
themselves
are
either
slimming
down
significantly
or
disappearing
right.
B
A
B
But
that
doesn't
necessarily
mean
that
the
components
couldn't
be
referring
to
the
same.
Let's
say:
config
map
right,
not
necessarily
like
all
of
the
configuration
is
to
answer,
but
imagine
if
there
would
be
I,
don't
know
a
TOS
config
map
right
that
specifically
describes
to
us
configuration
for
all
of
the
components
right.
B
D
Is
what
I
would
be
interested
in
interested
in
hearing
what
the
problem
is
that
we're
trying
to
solve
and
whether
it
hasn't
been
solved?
Yet
it
also
feels
like
we
should
have
efforts
in
the
Cloud
Foundry
community
on
the
developer.
Experience
not
just
redoing.
All
this
work
of
deploying
and
redeploying
stuff
I.
E
Think
the
main
problem
that
we
were
trying
to
solve
is
how
do
we
put
together
a
CF
deployment
tight
like
integration
level
artifact,
which
still
adheres
to
as
much
as
possible?
The
boundary
is
set
forth
by
kubernetes
itself,
so
you
know
like
we.
We
understand
that,
like
some
of
the
component
config
and
so
on,
like
there's
still
ways
to
come,
but
in
terms
of
packaging
it
up
and
deploying
it.
E
What
are
what
is
still
happy
like
what
is
the
way
to
like
take
that
and
still
try
to
follow
tooling,
that
is
more
like
kubernetes
needed,
so
that
was
kind
of
the
idea.
There
is
to
try
to
maybe
get
some
of
the
advantages
that
we
got
from
the
Bosch
mode,
but
without
actually
sticking
with,
like
the
exact
concepts
from
there's
like
the
release
concept
or
like
the
Lynx
concept,
for
trying
to
do
things
in
a
more
like
fashion
that
could
work
for
software
that
was
built
for
kubernetes.
E
Yeah
we
did,
we
did
take
a
look
at
the
operator.
I.
Think
one
of
the
things
that
we
were
trying
to
do
was
also
to
break
down
some
of
the
workflow
so
that
it
wasn't.
You
know
just
the
kind
of
the
principle
of
like
take
something
in
and
then
it
goes
through
various
stages
of
life
cycle,
and
then
it
spits
out
like
a
full
form
deployment.
We
were
hoping
to
like
decompose
it
a
little
bit.
Hence
for
hence
some
of
these.
E
You
know
some
of
these
like
boundaries
around
like
okay,
though
just
to
templating,
and
then,
if
you
wanted
to
like
do
overlays,
for
example,
on
top
of
that,
you
could-
or
we
just
want
to
do
like
if
you
didn't
want
to
use
any
of
the
templating.
It's
all
that's
provided
and
and
come
with
your
own
llamó
file.
You
could
and
still
use
staff
and
get
get
value
from
it
so
like
that
was
one
of
the
reasons
we
kind
of
started
thinking
through
it,
which
is
that.
E
D
We
we
have
a
similar
goal
with
the
components
that
we
build.
I
wish
we
we
would
talk
more
and
maybe
find
where
the
gap,
where
you
think
the
gaps
are
in
the
solutions
that
we've
built,
because
definitely
we
would
we
don't
want
to
work
in
a
vacuum
and
we
want
this
foundation
work
to
be
useful
for
everyone.
D
D
B
One
other
thing
to
throw
in,
maybe
is
you
know,
claw
foundry,
even
though
some
people
view
it
as
a
fairly
complex.
App,
in
my
mind,
is
just
a
bunch
of
stateless
micro
services
right.
This
is.
This
is
a
straight
up,
a
use
case
for
okay.
B
Hey
it
is
it
should
it
should
be
pretty
pretty
easy
to
deploy
it
on
kubernetes
without
any
kind
of
a
active
like
component
that
manages
in
lifecycle
right.
You
know
a
lot
of
the
primitives
that
kubernetes
gives
you
know
in
deployment
rolling
updates.
You
know
all
kinds
of
other
features
that
it
has
right
directly
supporting
the
features
that
we
need
so.
D
I
I'm,
in
agreement
with
you,
the
trouble
is
not
on
a
zero
we've
had
cloud
foundry
deployed
using
docker
images
and
just
environment
variable,
so
even
without
a
scheduler,
and
it
works
that
the
problem
is.
When
you
go
through
day,
one
they
end
where
you
need
updates,
you
need
secret
rotation.
You
want
to
make
sure
that
you
don't
have
downtime.
While
you
do
update
yep.
D
B
Is
what
I
think
Ruby
is
getting
at?
Is
that
I
think
she's
she's
she's
talking
about
breaking
down
the
problem
leo
that
you
know
into
into
into
smaller
pieces
right
so,
for
example,
until
you're
talking
about
sound
updating,
stuff
right,
so
a
more
smaller
targeted
tool,
let's
say
like
CAF,
where
I
think
there's
like
I
know,
there's
a
few
other
tools
similar
to
cap
rate
that
solve
a
particular
problem
of
you
know:
converging
a
share
of
resources
under
a
particular
label.
Right,
so
let's
say
you
have
60
resources
and
I
don't
know.
B
Next
morning
you
need
to
have
only
30
resource
rate,
because
someone's
simplify
laundry
right.
Those
kind
of
tools
converge
it
more.
You
know,
you
know
it's
it's
a
tool
that
designed
for
that
particular
purpose
right.
So
I
do
hear
you
that
there
is,
you
know,
day
two
challenges
right,
but
I
think
there
are
tools
that
address
those
challenges
in
a
more
targeted
way
versus
having
a
full-blown
operator
for
this
problem.
D
A
C
A
F
What
is
12
just
I
mean
this
just
feels
just
the
dates
point
it's
in
my
head.
If
you
look
at
how
can
a
TIF
works-
and
you
know
pen,
a
secret
foundry
they're
not
wholly
dissimilar,
if
you
think
about
what
they're
kind
of
trying
to
do,
you
can
deploy
K
native
just
from
the
amiable
right.
Thank
you.
Just
do
the
yet
I'm
from
the
web
site
single
coop
cut
up
like
a
chef
done,
but
there's
also
a
key
native
operator
right.
F
If
you
want,
you
know
the
a
way
of
just
giving
me
a
k-8,
severus
CRD
both
of
those
can
exist
and
and
and
from
for
what
it's
worth
that
feels
like
we're.
We
should
be
trying
to
have.
There
should
be
a
simple
Yama
file
that
relic
gives
me,
which
is
here
is
my
here
is
a
few
net
here.
Is
a
clarifier
of
humanities.
Ql
applied
a
chef
right,
it's
working
if
I
want
to
then
edit
that
and
do
fancy
stuff
with
it.
Maybe
there's
a
way
to
edit
the
conflict.
Maps.
F
B
Think
I
agree
was
that
I
think
you
know
the
this
demo
was
actually
partly
for
you
know
what
would
it
take
to
produce
a
manifest
like
that
right,
but
I
don't
think
this
demo
is
strictly
it's
saying
that
hey
everyone
should
be
running
through
all
those
steps
to
actually
get
the
final
yeah.
Now
there's
going
to
be
advanced
cases
right
and
maybe
those
advanced
cases,
some
some
would
be
less
advanced
than
others
right
and
some
may
be
able
to
use
customize.
B
Let's
say
if
that's
the
tool
of
preference
within
the
club
right
right,
but
I
do
you
know
working
a
little
bit
on
kn8
if
I
do
see
quite
significant
similarity
between
K
native
and
cloud
foundry
and
given
red
to
not
doing
anything.
You
know
significantly
more
complicated
from
a
deployment
process.
I,
don't
see
a
reason
why
I
shouldn't
have
to
should
be
able
to
have
a
single
Yama
that
could
be
you
know,
deployed
through
you
know,
pooling
of
choice.
F
A
A
F
A
F
It
sounds
like
you
might
be
productive
as
wife
Ladd
had
a
time
slot
in
that
meeting
to
take
us
through
some
of
the
overlaps
and
kind
of
maybe
show
is
I.
Think
I
think
one
of
everyone's
main
concerns
is
like
the
the
Boche
bit
of
the
operator,
but
maybe,
if
we
just
like
ignore
the
bush
man
talk
about
some
of
the
components
that
are
there.
That
could
be
available
like
underneath
that,
if
you
see
what
I
mean,
how
they
extend
is
faithful
sets
and
those
other
concepts.