►
From YouTube: Cloud Foundry Community Advisory Board [April 2017]
Description
Video from the monthly Cloud Foundry Community Advisory Board (CAB) meeting. Learn more at https://www.cloudfoundry.org/.
A
A
A
You
know
what's
interesting
is
just
so
you
guys
know
is
my
internet
here
in
Beijing
has
been
working,
fine
cool
all
the
time
and
then
today,
right
now
on
one
of
my
computer's,
so
anyways
alright.
So
let's
get
started
I
think
we
have
a
full
agenda
and
you
guys
know
already
so
we'll
do
basic
presentations
so
updates
from
teams
and
then
we'll
go
to
three
I.
Think
really
cool
updates
and
presentations
from
people.
So
we
have
David
savetti
from
pivotal.
A
C
So
for
the
purpose
of
brevity,
because
I
think
we've
got
a
lot
of
really
interesting
stuff
to
hear
about
today,
two
major
things
you
might
have
seen
some
noise.
We
made
around
the
training
and
developer
certification
product
you're,
going
to
start
to
see
them
actually
open
up
for
beyond
just
kind
of
early
enrollment
people
will
start
to
flow
through
the
the
course
that's
going
out
on
on
EDX,
which
will
be
a
free
introductory
class.
C
We've
had
some
great
course,
authors
from
start
can
wait
and
working
on
that
we
have
the
e-learning,
which
is
a
bit
more
bit
more
robust,
very,
very
labs,
driven
class.
Then
we're
offering
online
that's
been
built
by
Steve,
Greenberg
resilient
scale
and
then
a
few
folks
like
Daniel
from
engineer
better
in
the
UK
that
will
be
opening
up
in
early
May
and
then
we've
got
the
Cloud
Foundry
certified
developer
exams,
which,
if
all
goes
to
plan,
we'll
we'll
be
able
to
also
open
that
up
for
early
access
in
May.
C
All
three
of
those
things
are
kind
of
coming
in
for
a
landing
right
now,
there's
a
lot
of
work,
that's
gone
into
it,
but
we're
pretty
excited.
We
know,
there's
also
a
lot
of
demand
and
user
companies
were
training
that
kind
of
portable
across
the
different
providers,
as
well
as
being
able
to
have
individuals
be
certified
as
knowing
that
they
that
we
they
have
skills
to
build
applications
on
tapasya,
which
leads
me
to
the
other
point
that
Devon's
head
of
marketing
would
be
very
upset.
C
If
I
didn't
remind
everybody
that
we
of
course,
have
our
Silicon
Valley
summit
coming
up
in
June
roundabout
June
13,
we
are
going
to
be
doing
in
person,
certification
exams
at
the
event,
and
you
get
a
150
bucks
off
the
the
summit
registration
if
you
also
sign
up
to
do
the
as
a
certification
exam
which
should
be
kind
of
neat.
Most
of
this
was
intended
again
for
the
end
users
of
CS
right.
C
A
C
The
commitment
for
time
is,
is
varied,
I've
taken
it
probably
about
20
times
written
so
III
cheat
we're.
What
we're
telling
people
right
now
is
that
it
should
be
less
than
four
hours
right
and,
as
we
have
people
go
through
the
the
beta
process,
to
give
us
some
enough
statistical
data
to
figure
out
what
the
cut
score
is.
Basically,
it's
with
what
the
passing
score
would
be.
We're
also
asking
them
to
tell
us
how
long
it
took
it.
C
Well
sure
yeah
I
mean,
if
you're,
if
you're
interested
becoming
interested
in
coming
to
the
summit,
obviously
get
registered
any
of
your
your
customers
or
the
users
you
interact
with.
You
should
remind
them
of
it
and
in
particular,
if
you're,
working
with
any
end-user
developers
of
CF,
now
point
out
that
that
there's
a
certification
exam
that
that
can
help
them
as
an
individual
because
we're
starting
to
see
a
pretty
significant
uptick
in
number
of
open
positions
for
that
require
or
look
for
class
on
your
skills.
A
A
C
We
struggled
with
this
a
little
bit.
It's
gonna
be
a
two-year.
That's
gonna
be
valid
for
two
years.
The
granite
CS
tends
to
move
forward
a
little
bit
a
little
bit
progressively
quickly,
but
we
think
two
years
is
about
right.
It's
it's
a
performance-based
exam,
so
you
actually
have
to
do
the
things
that
you're
asked
to
do
and
then
the
scoring
process
interrogate
the
the
running
environment.
C
But
you
have
access
to
to
make
sure
that
you
know
you're
able
to
start
the
apps
that
might
have
had
troubleshooting
problems
to
sort
out
where
the
you
know
the
code
you're
asked
to
modify
is
now
doing
what
it's
supposed
to
be
doing
all
right.
So
it's
a
it's.
A
pretty
good,
pretty
good
test
on
the
basics
and
two
years
seems
about
right.
So
that's
what
we're
there.
A
D
C
Can't
answer
that
because
I
don't
know
that
seems
like
a
very
reasonable
approach.
On
the
other
hand,
the
we
take
a
lot
of
our
guidance
from
the
the
Linux
Foundation
certification
team.
They
do
it
for
they've
been
doing
it
to
the
Linux
itself.
For
a
long
time,
they've
been
launched,
form
around
OpenStack,
so
they've
got
a
pattern
for
this
fit
will
probably
follow.
C
C
C
Yeah
especially
reminder
to
anybody:
that's
a
contributor
in
any
way
right,
so
if
you've
helped
with
docs,
you
stand
up
for
question.
You've
submitted
a
bug.
Any
of
the
projects
use
that
contributor
code
that
I
put
out
on
the
dev
mailing
list.
It
again
started
a
couple
times
now
get
registered.
We
want
to
know
you're
coming.
We
think
we'll
probably
get
about
3000
this
year,
which
is
a
good
number.
A
E
A
D
A
B
E
F
We've
been
working
on
boss
DNS
for
some
time
now,
for
you
know,
maybe
a
close
to
month
we're
getting
pretty
close,
we're
going
to
be
trying
to
roll
out
some
of
those
changes
to
figure
those
web
services.
To
you
know,
trial
run
sound
functionality
pretty
soon.
Hopefully
this
week
you
know
we're
continuing
to
work
on
of
the
the
config
server
integration.
That's
closing
down,
it
will
be
that
will
be
kind
of
announced
in
the
next
boss
release,
which
will
be
also
cut
fairly
soon.
F
B
A
Great
I
mean
I,
think
I
think
you
know
this
is
gonna,
be
a
great
release
in
terms
of
you
know,
streamlining
things
I
was
asking
about
this
g8
or
CLI
v2,
because
people
I
work
with
are
asking
other
people
want
to
know.
When
is
it
ready?
Can
we
use
it
I'm
like
yes,
you
can
use
it
right
now
and
people
say
well,
but
is
it
version
1.0.
D
A
D
C
A
D
F
B
B
D
A
A
F
D
F
Correct
we
will
be
providing
additional
functionality
to
allow
users
to
set
multiple
runtime
can
fix
it's
really.
It's
going
to
be
a
little
bit
easier
to
manage
your
kind
of
add-ons
that
we
recommend
/
add-ons.
There
are
people
installing
via
the
DNS
v2
version
would
be
expected
to
be
installed
as
an
add-on.
D
F
A
To
highlight
one
more
thing
is
and
I
think
it
might
be
obvious
to
most
people,
but
some,
maybe
not
is
this
Bosch
effort
is
actually
been
going
on
for
a
while
and
part
of
the
reason
to
do
this.
Besides
helping
Diego
remove
the
user
control
is
also
not
depending
on
how
a
DNS,
so,
for
instance,
if
you.
E
A
A
A
Alright,
so
for
extensions,
I
don't
have
too
much
dry
light,
except
that
we
have
a
couple
more
proposals
that
are
in
the
works.
But
I
can't
really
talk
about
them
yet
because
they
haven't
been
put
to
Fisher
Lee
announced,
but
you
see
them
when
they
come
and
it's
gonna
be
I.
Think
good,
hopefully
I
telling.
A
A
I
think
I'm
trying
to
encourage
all
projects
that
are
not
code
of
khaki
or
Bosch
to
be
considered
as
extensions
and
submit
there.
So,
for
instance,
you
tour,
the
unique
project
is
going
on
there
right
as
an
extension.
Officially,
there
is
officially
about
six
projects
that
are
sort
of
going
under
under
review.
It's.
B
A
B
A
A
Within
your
organization,
or
that
is
in
a
separate
Rico-
and
you
want
to
be
part
of
extensions-
that
would
be
good
there'll,
be
more
reason
for
you
to
be
part
of
extensions
I'm,
trying
to
have
some
time
in
the
summit
to
show
people
essentially
a
way
for
us
to
organize
extension
better.
So
that's
also
them.
B
A
A
B
A
D
A
D
A
A
C
A
A
A
Let's
give
now
I
mean
we've
already
almost
used
more
time
than
I
expected.
That's
good,
too
David
sabeti
he's
been
waiting.
You
know,
David
has
worked
on
voice,
part
of
CF,
so
he's
extremely
knowledgeable
and
he's
been
working
on
something
really
cool
policy
up
department.
So
he's
gonna
tell
us
about
it:
David
you
you,
you
can
take
it
and
if
you
present,
then
you
can
I
think
used
to
show
your
screen.
Yes,
it's
good
all.
E
Right
well,
good
morning,
everybody
so,
like
dr.
Mac
said
my
name
is
David
tsveti
I'm.
Currently
the
PM
of
the
release
integration
team
I
wanted
to
just
get
a
chance
to
talk
to
some
people
about
CF
deployment,
explain
some
of
the
contacts
and
the
goals
for
the
project
and
offer
myself
up
as
a
resource
for
people.
We've
got
questions
regarding
state
deployment.
It's
starting
to
get
some
traction
in
use
so
feels
like
the
right
time
to
introduce
myself
a
little
more
formally
and
give
people
point
contact.
So
those
people
aren't
aware.
E
Maybe
I
should
start
with
an
explanation.
What
CF
deployment
actually
is
so
Steve's
deployment?
Is
it's
just
a
Bosch
manifest
at
the
post,
Cloud
Foundry,
but
it
has
some
special
characteristics.
It
doesn't
depend
on
C
efforts
at
all
and
it
doesn't
use
v
templates
that
you
find
in
CF
release.
If
that's
what
you're
used
to,
we
use
a
bunch
of
some
of
the
new
robust
features,
a
way
to
customize
deployment,
configuration
and
and
like
achieve
other
parts
of
deployment
configuration
in
a
much
simpler
way
than
people
are
used
to.
E
So
this
is
safe
deployment
and
the
main
the
main
start
here
is
this
CF
deployment
da
Gama
right
here
and
you
take
a
look
around
you'll,
see
that
it
looks
like
a
pretty
typical
Bosch
manifest.
But
let
me
highlight
some
of
the
things
that
are
pretty
different
from
the
manifest
you
would
get
from
CF
release.
So,
let's
start
actually
at
the
bottom.
If
you
take
a
look
at
this
last
section
here,
you'll
see
that
there
is
a
releases
section,
it's
got
quite
a
few
releases.
E
This
is
this
is
if
I
don't
know
how
familiar
everyone
is
with
watch
manna-fest,
but
this
is
where
I
should
tell
Bosch
like
which
bits
to
upload
and
deploy.
And
if
you
look
through
this
list,
it
won't
find
CF
release
anywhere
in
there.
Instead,
you'll
see
place,
Kathy
or
Diego
or
logger
Gator,
so
each
of
the
individual
subsystems
of
Cloud
Foundry
they
ship
their
own
releases
and
we
use
those
instead
of
some
like
monolith
like
CF
release,
which
is
what
we've
been
using
in
the
past.
E
The
other
thing
you'll
see
here
is
that
we
hard-code
two
version
numbers
for
each
of
these
releases,
so
this
is
also
actually
a
compatibility
matrix
of
sorts.
What
we're
saying
is
this
set
of
releases
have
passed
our
CI
pipeline
and
are
stable.
So,
yes,
it
sort
of
captures
the
notion
of
the
compatibility
between
the
different
releases.
E
What
else?
What
else
is
in
this
manifest?
You
got
this.
You
got
a
variable
section
so
again,
this
is
a
new
boss
feature
if
you've
seen
them
manifest
generated
by
the
CF
at
least
templates.
You
wouldn't
find
a
section
like
this
and
that
manifest
what's
actually
going
on
here.
Is
you
declare
a
bunch
of
variables
up
front
and
you
actually
get
to
describe
things
like
their
type?
E
So
here's
where
we
would
actually
configure
the
blobstore
to
have
this
password
and
here's
where
we
would
configure
the
cloud
controller
to
use
that
password
to
access
the
blobstore.
So
it
looks
a
little
bit
like
a
spife
template.
You
see
these
double
friends
in
whatever,
but
what
makes
this
really
nice
is
that
you
who
we
actually
have
like
considerably
better
tooling
around
supplying
these
values
then
is
fished
up.
So
the
and,
like
I,
said
you
actually
get
to
specify
the
type.
E
So
this
doesn't
just
take
some
arbitrary
string
like
you,
the
tooling
will
actually
guarantee
that
you
get
a
certain
type
for
this
value.
So
what
I
was
going
to
explain
how
how
these
that
variables
get
generated?
So
so,
if
I
come
back
to
the
front
page
here,
you
can
actually
see
what
the
deploy
command
looks
like.
E
So
so,
if
you're
using
the
new
box-t
a
lot,
you
can
use
this
VAR
store
flag
to
actually
generate
those
values
so
that
whole
list
of
variables,
you
don't
actually
have
to
generate
any
of
those
values
yourself.
So
if
you've
ever
struggled
with
coming
up
with
a
bunch
of
passwords
or
worse
configuring,
you
know
about
100
TLS
certificate
to
make
sure
that
they're
also
signed
by
the
correct
certificate
authority.
This
actually
captures
all
of
that.
E
You
don't
do
anything
like
the
bus,
so
I
just
generates
those
certificates
for
you
and
is
signed
by
the
correct
CA
and
all
that
jazz.
So
the
only
the
only
configuration
parameter
you
currently
have
to
supply
as
the
system
to
me.
So
this
is
what
a
deploy
command
actually
looks
like
right
here
like
that,
right
that
command
right.
There
will
give
you
a
function
in
Cloud
Foundry,
so
so
all
of
the
noise
around
manifest
generation
that
you're
used
to
that
gets
really
radically
simplified.
E
Let's
see
what
else!
What
else
is
there
to
love
about?
Cf
deployment,
if
you
take
a
look
around
this
manifest
you'll,
find
that
there's
not
really
a
lot
of
I
a
specific
configuration.
We
leave
that
to
boss
in
the
form
of
a
cloud
config,
so
so
manifest
becomes
a
bit
more
generalized.
Another
thing
you'll,
if
you
look
around,
is
that
there
is
no
global
properties
section.
E
Every
job
has
the
properties
that
meets
immediately
right
there,
so
we're
not
including
some
global
namespace,
and
it
actually
enables
releases
to
do
some
nice
things
with
the
structure
of
their
properties
and
stuff.
Like
that,
what
else
we've
got
striping
so,
instead
of
creating
a
bunch
of
identical
jobs
with
little
suffixes
that
describe,
which
is
either
instead
we're
gonna
use
Bosch
to
just
go
ahead
and
do
that?
E
What
else
do
we
have?
You
might
even
notice
that
some
some
configuration
seems
to
be
missing
from
this
manifest.
It's
not
a
human
saying,
we're
just
using
Bosch
links.
So
so,
as
teams,
the
work
on
the
links
is
going
to
be
a
little
bit
slower
because
that
has
to
actually
get
implemented
by
the
release.
Change
releases
have
to
provide
and
consume
the
links,
but
the
goal
is
to
sort
of
try
to
shrink
this
manifest
a
bit
as
so
that
we
don't
have
to
constantly
be
providing
things
like
you
know,
match
password
everywhere.
E
Instead,
what
these
things
can
get
consumed
as
a
link
and
then
lastly,
customizations
to
the
deploy
manifests,
are
made
not
in
the
form
of
some
stub,
but
rather
in
what
are
called
what
we're
calling
ops
files
and
if
you
take
a
look
and
see
if
deployment
in
the
Operations
directory
appear,
you'll
see
a
list
of
off
styles
that
the
relation
aggression
team
is
currently
supporting.
So
these
things
make
small
changes
to
your
manifest
so
that
you
can
deploy
it
under
different
circumstances
or
with
slightly
different
configurations.
E
What
else
is
there
to
say
about
this?
I
guess?
Maybe
some
of
the
goals
of
the
project
like.
Why
are
we
bothering
to
do
any
of
this
stuff?
You
want
to
actual
the
driving
force
behind.
This
is
actually
just
being
able
to
unify
Diego
into
a
normal
cloud.
Finder
deployment
you
know
prior
to
this
you'd
have
to
deploy
CF
and
Diego
as
separate
deployments,
but
CF
deployment
includes
all
the
Diego
components
by
default
and
in
fact
you
can't
deploy
Diaz
once
you
have
to
plummet
like
we
officially
walked
away
from
those
and
CF
deployment.
E
Configuration
and
documentation
of
configuration
should
hopefully
be
a
lot
easier.
If
you've
ever
tried
to
look
at
the
docs
on
platform,
you
got
org
for
deploying
could
be
a
little
bit
confusing
and
hopefully
those
will
be
really
simplified
too
and
then
a
lot
of
the
just
like
confusion
and
craft
about
manifests
like,
as
you
know,
these
new
configuration
parameters
came
in
and
became
a
little
bit
harder
to
reason
about
your
manifest.
The
spiff
templates
also
leave
a
decent
amount
of
cruft
in
there,
like
all
of
that
hopefully
gets
gets
reduced
considerably.
E
So
there's
no
longer
a
problem,
and
one
of
our
goals
for
the
manifest
is
for
it
to
be
sort
of
canonical
reference
like
we
wanted
to
be
legible
and
for
people
to
be
able
to
look
at
it
and
understand
any
of
everything
that's
going
on
in
there.
So
we're
working
really
hard
to
keep
it
as
tight
as
possible
so
that
we
don't
end
up
with
moist,
and
the
last
thing
I'll
say
about
this
is
like
people
have
asked
me
how
ready
CF
deployment
is.
E
I
would
say
that
it's
definitely
ready
for
things
like
test
and
development.
Most
of
our
mostly
internal
development
teams,
the
CloudFactory
foundation,
the
OSS-
are
our
using
CF
deployment
in
some
part
of
their
pipeline.
So
it's
ready
to
use
for
those
those
scenarios.
As
far
as
production
environments
go
we're
still.
E
The
release
integration
team
is
still
working
on
two
big
things.
The
first
is
developing
a
migration
path
from
CF,
at
least.
If
you
have
deployment
so
everybody
who
currently
has
their
CF
or
at
least
deployed,
we
want
to
make
sure
that
you
have
a
decent
experience
migrating
this
deployment,
it's
ready
and
then
the
other
thing
is
we
want
to
make
some.
You
still
needs
a
recent
investigation.
It
said
what
would
essentially
act
at
up
time
during
deploys.
We
haven't
really
been
fond
of
testing.
B
E
There
are
decent
number
of
differences,
so
links
are
one
example
like
I
said
the
other.
The
big
one
is
that
Diego
gets
inlined
into
your
deployment.
So
so,
instead
of
having
one
to
put
like
two
deployments,
one
for
CF
and
one
for
Diego,
this
is
good
going
to
go
the
other
way
and
say:
let's,
let's
put
everything
in
the
same
deployment.
D
So
in
the
manifests,
why
do
you
guys
go
with
the
so
most
people
know
like
variable,
interpolation
and
stuff
like
that
as,
like
you
know,
if
you're
familiar
shell
scripting
stuff
like
that,
you
might
use
like
a
dollar
sign
bracket
bracket
inside
there?
Why
did
you
go
with
the
double
parens?
It
seems
really
awkward
and
weird.
That's.
E
F
So
there
isn't,
why
is
it
is
a
much
more
well
City
Mo
friendly,
actually
writes?
We
were
considering
issue.
Do
the
Curly's
to
the
promised
curl
is
you
have
to
put
double
quotes
all
over
the
place,
because
the
actions
that
have
kind
of
a
gill?
So
you
can
it's
almost
an
invalid
yeah
I,
think
it
is
one
yeah.
So
double
friends
are
agnostic
to
that.
If
you
will
and
they
don't
cause
you
to
have
to
go
in
and
fight
with
llamo,
you
know
line
by
line.
F
You
know
some
of
the
other
characters
might
have
been
maybe
possible
as
well,
but
it
seemed
like
double
parents
offered
that
and
a
certain
level
of
visibility
versus
trying
to
hunt
down
where
is
they
and
in
the
beginning
of
the
date,
abilities
like
that
and
also
I
believe
their
summer
bash
friendly
as
well.
So
just
just
Sam,
you
know
what
I
feel
interoperability
issue
right,
yeah.
E
E
F
But
we
will
strongly
encourage
people
to
use
kratom
because
it
will
provide
a
manageable
and
secure
way
to
store
your
credentials.
Now,
if
you,
you
know,
try
to
extend
credit
to
you
know,
save
the
credentials
in
your
favorite
credential
store.
You
know
that
will
be
also
fine,
but
from
a
perspective
of
Bosch
will
definitely
be
hoping
that
people
use
it
because
then
Bosch
would
have
you.
No
raw
passwords
and
certificates
in
its
database
is.
D
F
A
D
D
E
The
difference
between
the
release,
candidate
branch
or
the
master
branch
is
some
notion
of
like
creating
a
version
tag
so
release
candidate
gets
automatically
updated,
whereas
master
gets
manually
updated
when
we
add
a
new
version
and
as
far
as
bruising
he
goes.
One
of
our
other
goals
for
CF
deployment
is
start,
semantically,
learning
so
being
able
to
say
CF
deployment
version,
1.0,
1.1
1.2,
you
like.
Why
actually
do
some
entacle
version
prison
incorrectly,
so
major
version
bunks
would
be
somehow
a
breaking
change
in
your
deployment
of
US,
like
in
my
in
my
mind.
E
If
the
deployer
is
the
primary
personal
concern
here,
then
then
a
major
version
but
means
the
deployer
is
gonna,
have
to
stop
what
they're
doing
when
they're
deploying
and
like
figure
something
out
like
there
are
manual
steps
involved
or
some
some
other
stuff
they'll
have
to
take
part
of
their
employ,
so
will
actually
try
to
be
really
explicit
about
those
business.
Those
expectations
as
part
of
her
today.
D
E
Here,
I'm
suffering
my
screen
right
yeah.
So
if
you
look
here
in
the
transition
directory,
you'll
see
some
of
the
stuff
we're
building.
This
is
all
super
early
I
wouldn't
take
a
look
at
it
just
yet
we're
we're
just
just
getting
started.
We
probably
picked
up
the
story,
maybe
a
week
ago,
so
where
we
started
on
the
track
of
work
about
a
week
ago.
So
it's
also
pretty
I.
Wouldn't
this
is
the
place
to
follow
this
transition
directory.
So
you
have
to
come
and
like
send
some
updates.
E
As
I
said,
we
because
we
made
progress
so
maybe
once
you
get
some
of
the
tooling
updated
lots
of
people
like
hey
this
thing
like
you
ready,
take
a
look
at
it,
give
it
a
shot,
probably
not
ready
to
migrate
yet
but,
like
maybe
there's
something
viable
as
far
as
experimenting.
If
you
want
to
like
learn
about
CF
deployment,
so.
A
A
A
A
D
A
F
Turbulence
is
a
release
that
provides
an
API
server,
that
does
chaos,
monkey
functionality,
kind
of
a
disturbance
in
your
environment
right
so
for
people
who
don't
know
for
people
who
don't
know
chaos
monkey
is
something
that
Netflix
made
you
know
years
ago.
That's
the
way
they
do
stability
or
I,
guess
kind
of
a
randomized
testing
in
their
environments.
What
it
does
it
kills
off
particular
VMs
in
their
AWS
environment
or
maybe
because
of
particular
Izzie's
and
regions.
F
Even
so,
the
concept
is,
very,
you
know
generic
right,
throwing
some
kind
of
a
failure
into
your
environment
and
see
how
your
environment
behaves.
Now.
You
know
some
people
may
decide
to
you,
know
kind
of
a
put
it
in
a
pipeline
for
testing
purposes.
Maybe
I
don't
know.
Let's
say
for
me,
I
mean
I
had
CD
cluster
and
you
see
that
HCD
cluster
recovers
shortly
after
that
right.
Other
people
just
decide
to
do
a
randomized
injection
of
failure.
F
Maybe
like
once
a
week,
there
is
some
random
failure
is
injected,
and
you
see
that
the
system
recovers.
Okay
and
over
time,
of
course,
you
gain
confidence
that
your
piece
of
software
that
she
deploying
can
resend.
You
know
a
variety
of
different
failures,
and
you
know
it
gives
you
a
little
bit
of
peace
of
mind
that
when
a
real
failure
hits
you're
not
completely
unprepared
right
now,
of
course
not
all
software
is
really
able
to.
You
know,
withstand
all
kinds
of
different
failure
injections.
F
So
you
know
you
may
find
out
that
the
software
that
you're
running
doesn't
actually
deal
well,
if
an
entire
easy
start,
and
that's
that's
partly
the
reason
why
you
would
want
to
use
turbulence
to
understand
what
is
the
business
impact?
You
know
that
will
happen
if
something
happens
to
your
environment.
So
turbulence
like
I
was
mentioning
it's
an
API
server,
but
it
also
has
an
agent
agent
can
be
co-located
everywhere
at
the
smallest
possible
or
the
simplest
possible
functionality
level.
F
All
it's
doing
is
just
telling
the
director
hey
delete
this
VM
without
trying
to
do
anything
Bashi
so,
for
example,
doesn't
try
to
drain
anything
or
anything
like
that.
You
can
simulate
that
behavior
yourself
by
just
calling
borsch
delete
vm
and
that's
pretty
much.
What
turbulence
does
it
calls
that
API
call
with
a
specific
set
of
instances
now,
in
addition
to
you,
know
the
up
termination
or
unexpectedly
I'm
termination?
Rather,
it
also
can
impose
different
I
CPU
RAM
IO
load.
It
can
do
Network
partitioning
with
IP
tables.
F
F
You
know
what
kind
of
procedures
can
be
done
so
a
top
level.
Primitives
and
turbulence
is
an
incident,
so
you
can
create
an
incident
and
each
incident
can
have
a
set
of
tasks
to
perform
and
a
selector.
So,
for
example,
you
can
select
30%
of
your
Diego
cells
to
be
stressed,
for
you
know,
10
minutes
and
only
stress
it.
You
know
stress
the
CPU
for
it,
for
example,
right
and
of
course
you
can
select,
you
know
deployment
name,
and
you
know
instance,
group
and
maybe
even
easy
right.
F
So
some
of
the
other
things
that
you
guys
might
be
interested
in
is
there's
a
schedule
to
incidents.
So
arguably
you
know
if
someone
is
really
boring
into
concourse
and
you
know
seeing
and
cheating
then
you're
a
div
CI
system
that
can
do
some.
You
know
scheduled
delivery
of
some.
You
know
action,
so
you
could
just
be
you
know,
making
a
curl
command
to
the
incidents
API
endpoint.
However,
if
you're
not
using
a
CI,
are
you
just
not
bothered
by
it?
F
You
know
for
a
particular
reason,
then
you
can
configure
a
scheduled
incident
and
it
will
do
it
on
a
timer.
Not
you
know
not
the
most
advanced
scheduler,
but
deals
with
some
of
these
things
very
decently.
There
are
some
of
the
explicit
incidents.
You
know
you
have
a
kill,
kill
of
the
VM.
You
have
a
kill
process,
so
it
can
kill
a
random
process
or
maybe
it
can
kill
a
process.
That's
matching
some
kind
of
a
process
name.
F
F
B
F
F
So
you
would
actually
you
know,
have
a
agent
window,
something
like
that,
and
you
know
it's
the
same
code,
because
it's
you
can
go
for
example,
but
there's
definitely
code
paths
that
will
trigger
windows,
specific
things,
so
slightly
different
job
was
more
or
less
the
same
content
on
the
job,
and
you
know
other
releases
in
CF
or
you
know
in
other
places,
have
taken
up
the
same
pattern.
So
you
know
cuz,
you
need
different
jobs.
F
You
need
advice
manifest
you
will,
for
example,
for
the
agent
you
will
configure
there's
a
different
type
of
add-on
right
add-ons
are
allowed
to
be.
It's
actually
applied
selectively
right.
So
there's
a
include
and
exclude
section
for
add-ons,
so
you
can
say
that
particular
job
should
be
particular
add-on.
With
this
set
of
jobs
could
be
applied
to
only
particular
operating
system
or
maybe
excluding
opera
exist.
So
that's
kind
of
that's
how
you're
able
to
say
you
know
do
one
thing
for
Windows.
F
Do
another
thing
for
centers:
do
nothing
for
you
bloom,
let's
say
yeah,
but
you
know,
of
course,
if
you're
so
base
might
has.
Actually,
since
you
mentioned,
base
manifests
right,
let's
take
a
look
at
that
real
quick
base.
Manifest
is
fairly
simple.
All
it
takes
in
is
a
director
of
credentialing
thing
and
it
generates
the
rest
of
the
things
you
know
for
the
certificate
for
the
API
server
and
it
exposes
a
link
for
the
agents
to
connect
to
automatically
and
that's
purge
it.
F
You
know
it
deploys
the
server
the
server
I
have
a
little
deployment
over
here.
I
know
I'm
doing
on
time.
So
stuff
me
whenever
you
want,
but
you
know,
if
you
look
at
turbulence
deployment
over
here,
there
is
a
server
at
this
IP,
so
we
can
real
quickly
go
in
and
hit
that
server.
There
is
HTTP.
You
is
fairly
simple,
actually
there's
a
screenshot
of
the
UI.
It
looks
something
like
this
and
you
know
you
can
schedule
the
tasks.
It
looks
like
I'm
out
of
time
for
doing
the
presentation.
F
A
F
Teams,
I
believe
console
team
used
it
to
verify
some
of
the
you
know
behaviors
when
console
you
know
to
test
to
make
sure
the
console
can
failover
and
things
like
that
right.
There's
few
other
teams,
I
believe
with,
and
people
that
use
it
to
test
some
of
the
other
things
I'm,
not
using
it
to
test
any
of
my
things
on
periodic
or
one
continuous
basis.
But
here
and
there
I
tried
to
throw
her
in
and
see
what
happens.
B
B
Obviously,
the
last
few
years
Dmitri,
his
team
have
been
going
through
the
blush
notes.
What
is
version
two
almost
killed
that
writer
a
year
ago,
but
this
is
kind
of
my
looking
at
what
applies
to
it.
It's
somewhat
about
these
are
the
things
that
you're
not
going
to
be
using.
Like
you,
don't
use,
passionate
use,
Bosch
preamp,
you
don't
use
much
light
use
blush,
create
any
with
the
VirtualBox
CPI,
and
then
we
have
the
new
repose.
B
B
We
have
fabulous
documentation
and
you
know
someone
who's
been
around
for
five
years.
I
got
surprised,
abducts
and
it's
wonderful
to
read
them
so
this
this
pattern,
that
sort
of
today
by
David
the
CF
deployment
actually
spans
our.
So
there
is
the
Bosch
deployment
and
you
might
probably
before
use
Bosch
deployment,
try
the
jump
box
deployment
to
bring
up
a
job
box.
So
after
see
of
deployment
as
an
example,
there
is
a
dock,
a
broker
deployment
which
will
bring
up
a
service
broker
that
runs
and
one
or
more
services
in
a
docker
containers
and
okay.
B
So,
as
you
start
to
build
your
own
Bosch
releases
or
perhaps
to
upgrade
your
Bosch
releases,
have
a
think
about
this
pattern,
or
at
least
have
a
manifest
folder
in
your
release.
With
that
same
idea,
and
when
you
get
started,
we
started
to
write
blog
posts,
obviously
there's
more
and
more
dog
communication.
At
least
this
is
done
with
a
sense
of
humor
and
because
it's
pretty
exciting
and
and
I
thoroughly
enjoyed
the
last
month
of
upgrading
everything
I've
touched
and
as
Sir
David
said,
the
gist
of
it
now
is.
B
B
B
B
So
just
to
go
back.
There's
the
bars
line.
They
need
to
remove
that
and
deploy
I
find
this
irritating
and
I.
Don't
it's
like,
amongst
all
the
beautiful
things
we
have
this.
This
story
to
me
doesn't
make
as
much
sense
and
then
in
Bosch
bootstrap,
it's
the
same
flags,
but
you
use
the
Bata
creative
with
with
whichever
CPI
file,
and
then
it's
just
that.
It's
really
nice
like
it's.
If
you
learn
to
use
Bosch
one
way
that
pattern
rolls
out
and
that's
part
of
why
I
think
you
know
getting
rid
of
favorite
up
story.
D
F
F
Now
there
is
a
feature
called
multi
CPI
that
was
introduced
in
a
hidden
way
in
a
previous
release
of
in
a
current
bas-reliefs,
and
then
there
will
be
a
more
formally
announced
in
the
next
boss
release
and
that's
the
multi
CPI
support
such
that
you
can
configure
a
director
to
use
multiple
CPI's
and
really
a
CPI,
then
maps
to
an
AZ.
So
your.
F
F
F
Once
we
have
that,
then
technically,
you
would
be
able
to
deploy
it
already
by
just
you
know,
calling
create
and
multiple
times
make
sure
that
they
point
to
the
same
database
same
blobstore,
and
then
then
you
set
really,
but
you
know,
I
won't
promise
you
that
by
the
end
of
the
year
we'll
say
this
is
the
production
cerebral.
You
should
be
running.
This
show.
B
F
To
speak
to
that
you
know
the
the
goal
is
to
not
kind
of
a
make
worst
or,
as
you
know,
as
as
dominant
as
it
is
right
now,
perhaps
just
because
we
have
much
better
integration
with
screwed
up
for
real
full
deployments,
because
you
know
at
the
Korean
create
and
time
there
isn't
a
credible.
You
can
talk
to
it,
sir.
You
can't
really
save
the
credentials
anywhere,
but
you
know
during
a
board
deploy.
We
can
say
that
hey
there
is
credits.
B
Just
want
to
this
is
my
experience.
I
mean
this.
This
happened
again.
This
is
the
era
I'm
looking
at
whilst
this
is
this
is
my
generated
config
and
it's
got
all
the
variables
in
it's
like.
Oh
that's,
right,
I
need
to
you
know.
This
is
it's
just
another
very
nice
experience,
and
so
without
trying
to
say
what
the
solution
is
to
never
let
variables
get
unpopulated
on
the
way
through
as
a
successful
deploy,
would
be
one
flower.
C
A
B
You
go
so
we
talked
about
that.
One
of
the
things
I've
been
looking
at
is
trying
to
get
to
this
default.
Config
pattern
that
Demetri
educated
beyond
mostly
through
taxes
to
the
CF
deployment,
repo
and
yeah.
There's
the
URL
which
I
pasted
it
into
the
channel
for
learning
about
the
code.
Patch
stuff
I've
been
playing
all
those
stuff
for
a
month
and
it's
pretty
lovely.
So
it's.
A
B
A
A
Thank
You
Wayne
all
right,
so
this
is
it
for
for
this
time
we
are
running
just
a
little
bit
over,
but
we
started
at
12:05.
Thank
you.
Everybody
for
presenting
we'll
have
one
more
call
next
month
and
then
it'll
be
summit.
Thank
you.
Those
things
are
recorded
by
the
way.
When
guarantee
is
that
thanks
guys,
bye,
kiss.