►
From YouTube: Kubernetes SIG Apps 20170501
Description
May 1, 2017
Demo of Armada - tool for collecting helm charts,
Dicussion - moving examples/ in kubernetes repo to its own repo
A
B
B
What
we
wanted
to
do
with
Armada
is
to
enhance
our
helm
experience
by
being
able
to
predefined
multiple
charts
prior
to
deployment
and
well
and
be
able
to
drive
those
deployments
via
either
a
local
directory
or
a
or
a
git
repo,
and
how
we
would
extend
to
it.
How
will
be
deployed?
It
also
is
through
either
a
CLI
or
an
API
endpoint.
B
So,
to
show
up
first
and
Armada
is
still
very
much
using
the
helm
tiller.
So
we
need
to
make
sure
that,
prior
to
the
pointing
anything,
we
need
to
make
sure
that
our
tiller
is
up
and
running
so
to
demonstrate
they
are
not
API,
and
we
can
see
that
our
tiller
server
is
active
and
to
show
you
a
little
bit
about
the
arrow
on
the
armada
CLI,
since
it's
a
little
bit
more
advanced
right
now
and
I
want
to
go
through
a
simple
deployment.
B
B
So
we
have
our
get
locations
here
and
the,
and
this
is
how
we
build
them.
If
charts
need
dependencies,
all
we
need
to
do
is
define
them
under
the
the
appendices
key
name
here.
So
the
Maria
DB
has
a
dependency
on
our
home
toolkit,
and
that
is
pretty
much
it
and
you
just
define
all
the
other
charts
and
dependencies
as
needed.
C
One
reason
why
we
did
this
too,
is
because
the
dependency
is
handled
by
helm
today,
if
I
have,
let's
say,
multiple
applications
that
or
kubernetes
manifest
that
rely
on,
let's
say
moriah
DB
we
did
more
several
instances
of
moriah
DB
or
we
wanted
helm.
We
wanted
Armada
to
handle
that
dependency
a
little
bit
better.
So,
for
instance,
in
our
case,
we
can
deploy
either
multiple
varieties
or
we
can
use
a
single
variety,
be
seen
separate,
namespaces
and
so
to
avoid
those
collisions.
B
B
B
You
can
see
here
that
if
there
are
no
changes
to
edit
the
git
repository
or
the
values
files
we
don't
actually
deploy,
so
we
prevent
any
deploys
that
doesn't
mean
either
difference
in
the
repo
or
in
the
values,
and
we
can
see
here
that
there
was
a
difference
in
the
rabbit
and
we
were
able
to
redeploy
it
and,
as
you
can
see,
changes
have
been
made
to
the
browser.
Now
we
only
have
one.
B
C
So,
what's
nice
about
this
too,
is
during
let's
say,
upgrade
or
if
you're
removing
a
chart
chart
it'll,
also
removal
and
clean
up
all
the
jobs
as
well,
and
that
was
one
one
thing
we
really
needed
out
of
helm
because
there's
a
lot
of
job
dependencies
and
opens
back
and
pounds
will
it
if
you
guys
I
saw
that
really
quickly.
Helm
toolkit
is
basically
it's
not
really
making
a
deployment
so
much
as
it's
a
set
of
functions
for
to
enable
the
jobs,
for
let's
say,
creating
a
DB
user
or
creating
an
endpoint.
C
D
B
C
D
E
E
C
So
so
the
goal
hey,
this
really
spends
with
a
project
that
we're
doing
which
is
OpenStack
helm.
So
OpenStack
is
a
an
extremely
complex
application,
as
many
of
you
know,
and
we
wanted
to
use
helm
for
the
deployment
and
life
cycle
throughout
to
do
that.
What
we're,
what
we're
doing
is
we're
doing
a
tuple
and
running
that
through
a
hash.
So
each
time
let's
say,
there's
a
config
change
in
in
the
Etsy
directory.
Essentially,
what
we're
doing
is
we're
noticing
that
that
hash
would
change
and
we're
redeploying
I
think
Armada
is
following
the
same
logic.
C
C
Exactly
so
I
mean
this
is
really
really
new
and
we're
looking
for
more
help
from
the
community,
because
we
think
it
would
have
wider
wider
adoption
in
more
use
cases
right,
but
that's
something
that
we're
going
to
have
to
look
at
because
we
obviously
don't
want
to
disrupt
the
pods
that
are
that
are
actually
standing.
Actually,
the
redeployment
has.
E
C
The
goal
is
to
redeploy:
that's
why
we're
doing
a
hashed
across
the
deployments
for
any
changes
in
what
would
be
what
would
be
at
sea
and
and
then
bin
as
well.
If
you
look
at
the
project,
but
I
want
to
make
sure
that
we
don't
like
actually
lose
things
as
well,
because
it
looked
like
we
were
in
a
state
where
there
was
zero,
replicas
and
I.
Don't
want
that
so
alright
and
thank
you,
it
was
great
on
yeah,
so
I
mean.
Hopefully
you
guys
find
a
little
bit
of
additional
use.
C
F
F
I
need
to
go
back
and
look
and
see
if
there
are
a
lot
of
things
that
do
make
sense
to
bubble
back
up
to
Alma,
but
I
see
why
it
makes
a
lot
of
sense
as
a
separate
tool
that
can
really
specialize
in
neat
cases
like
the
OpenStack
case,
where
the
charts
would
be
are
just
unbelievably
complex.
Unless
you
do
something
to
manage
that
complexity,
yeah.
C
Definitely
one
two
things
to
note
for
us:
anyways
was
the
get
driven
aspect
which
we've
talked
about
before
right.
The
other
was
cleaning
up
jobs.
You
know
you
end
up
with
all
these
jobs
everywhere
and
we
need
to
clean
those
up
when
we
do
when
we
do
read
them
employs
and
so
forth.
So
those
were
two
of
the
big
drivers.
Yeah.
F
C
Be
cool
yeah,
let
us
know
if,
if
there
are
some
features
that
you'd
like
to
see
in
helm
or
whatnot
right
now,
it's
written
in
Python,
we
had
discussed,
writing
and
go.
We
just
have
a
lot
of
Python
experience
in
our
crew
since
we're
in
OpenStack
Earth,
obviously
so,
but
if
we
needed
to
do
that
or
if
you
know,
if
you
wanted
to
talk
a
little
bit
further
about
that,
then
please
let
me
know,
but
thanks
for
the
time
good
thanks.
B
A
Alright,
so
if
there
are
another
questions,
you'll
move
on
to
our
discussion,
topic
or
just
discussion
topics,
so
the
first
one
is
about
the
proposal
to
move
things
in
the
examples
directory
and
then
main
kubernetes
repo
to
its
on
examples,
repo,
so
a
mess
are
you?
Are
you
here
to
talk
about
that
and
then
I
pronounce
your
especially.
H
A
H
H
Especially
there
are
a
lot
of
storage
examples
here,
which
are
not
particularly
relevant
to
the
main
repository
and
the
problem
with
those
things
are
mostly
in
most
of
the
cases
we
don't
have
clear
ownership
for
those
and
they
have
not
been
maintained
for
a
very
long
time,
and
so
ideally,
we
would
like
to
move
some
of
these
out.
So
the
proposal
I
sent
out
basically
suggests
that
we
should
get
rid
of
the
third
party
examples
gracefully.
We
should
encourage
the
maintainer
of
these
examples
to
create
their
own
repositories,
to
maintain
their
own
communities.
H
Examples
and
you
possibly
create
home
charts
and
serve
them
as
the
examples
to
the
community,
and
some
of
them
are
just
out
of
date,
so
I
think
we're
better
off
just
cleaning
them
up,
and
we
have
some
maintained
examples
here,
if
I
scroll
all
the
way
down.
We
have
some
example
like
guestbook,
WordPress
and
Cassandra
they're,
basically
showcasing
various
features
of
create
and
they
have
different
complexity
levels,
and
we
can
extend
these
examples.
But
the
thing
is
that
these
are
the
canonical
set
of
examples.
H
We
maintain
from
version
to
version,
and
we
make
sure
that
these
actually
work
and
some
of
them
are
actually
used
in
in
the
test
suite
and
tested
thoroughly.
But
the
thing
is
that
some
examples,
the
rest
of
the
examples
here
are
not
tested
at
all
and
some
of
the
mics
with
calls
in
continuous
test
failures
for
for
the
past
six
months.
Really,
so
that's
that's
a
little
bit
of
a
problem,
and
for
that
we
actually
created
a
repository
called
crazy,
slash
examples,
so
we're
planning
to
move
those
there.
H
So
I
really
appreciate
your
comments.
Similarly,
we
have
a
similar
repository
called
contrib,
and
this
is
kind
of
all
turned
into
a
dumpster
of
example
that
people
just
you
know
they
have
a
software.
They
want
to
show
how
it
works
on
clear
days,
so
they
just
sent
a
PR
here
back
in
the
day
and
some
examples
in
both
repositories
actually
are
kind
of
documentation,
like
the
volume
plugins
for
instance
day.
H
Actually,
if
you
click,
if
I
click
volume
here,
you're
going
to
see
that
this
is
actually
like
the
reference
documentation,
sort
of
how
it
is
used,
if
you,
if
I
click
on
something
random,
that
I'll
you'll,
see
that
these
are
the
fields
that
I
should
be
providing
and
so
on.
So
I
think
these
should
ideally
move
live
under
the
documentation
website,
so
probably
will
move
those
to
the
documentation
repository
over
time,
but
yeah
that's
pretty
much.
It
I
have
a
exhaustive
list
of
examples
here
and
what
I
think
of
them.
H
It's
basically
like
you
know,
for
instance,
it's
a
parking
example
is
kind
of
out
of
state
and
it's
third-party,
so
we
should
probably
get
rid
of
this.
There
are
some
maintained
examples
and
so
on.
So
if
you
have
any
opinions
on
this
topic,
please
chime
in
I
really
appreciate
help
so
I'm
still
trying
to
figure
out
what
to
do
with
all
these
things,
but
most
likely
most
of
these
will
go
away
and
I
don't
want
to
just
delete
them.
We
should
totally
make
it
graceful,
that's
pretty
much.
It.
I
We've
been
thinking
about
this
within
my
circle
of
co-workers,
and
then
we
have
this
problem.
I
want
to
find
out.
If
it's
the
same,
we
want
to
write
like
canonical
examples
of
how
to
run
complex
systems
like
a
Kafka,
for
example,
which
is
a
pretty
tricky
to
maze
integration,
and
we
don't
think
the
helm
charts
repository
is
the
right
place
for
all
the
docker
files
and
like
go
code.
That
makes
the
admit,
containers
that
bind
everything
together,
and
so
we
were
thinking.
I
We
need
a
new
place
to
keep
that,
and
one
thought
we
had
was
actually
create
an
entirely
new
org,
because
that
makes
better
use
of
get
hubs
tackle
system.
So
you
can
have
repose
with
very
small
numbers
of
Elinor's
in
them
rather
than
having
like
per
directory
ownership.
I
was
wondering
if
you,
if
you
think
these
two
use
cases
can
be
merged
and
if
you
thought
about
using
a
new
Ora
grab
in
the
new
repo.
H
So
I
think
I'm
trying
to
think
if
we
can
actually
merge
it.
So
the
set
of
maintain
examples
here
actually
serve
a
documentation
purpose
and
they
actually
woke
the
user
through.
You
know
how
to
run
I,
don't
know
WordPress
and
my
circle
on
a
persistent
disk
on
coronaries
step
by
step,
so
Kafka
probably
would
be
more
complicated,
but
if
you
think
that
we
can
make
it
a
VOC
through
because,
ideally
the
plan
is
to
actually
move
these
to
a
website.
Basically,
we
have
a
samples
way,
Pulliam
talks
samples.
H
So
basically
we
have
a
part
of
the
other
section
on
the
website
that
says
samples
and
these
samples
are
actually
just
going
to
github
right
now,
and
there
are
just
bunch
of
things
on
the
left-
and
these
are
the
you
know,
also
unmaintained
stuff.
So
at
the
goal
here
that
I
am
trying
to
achieve
is
make
sure
that
we
have
a
maintained
and
tested
set
of
examples
from
release
to
release
and
they
don't
break,
and
this
is
how
kind
of
how
we
teach
Carre's
other
people
in
terms
of
you
know
solid
projects
that
actually
work.
H
Kafka
is
great.
It's
just
like
it's
a
complicated
system,
but
the
thing
is
that
if
you're,
not
if
you're
not
planning
to
maintain
and
if
there's
a
lot
of
overlap
with-
let's
say
Cassandra,
then
what
am
I
there
might
not
be
necessarily
a
lot
of
value
in
maintaining
both.
But
if
you
think
we
should
totally
have
Calcutta
and
we
can
totally
have
a
organization
created
pulse
tree
in
there
just
for
Kafka,
and
you
know
just
maintain
the
example
there
I'm
not
too
concerned
about.
You
know
where
the
existing
code
should
go.
H
J
I
I,
just
I
want
I
want
understand
what
example
means,
because
one
case
is
like
alarm:
trooper
Nettie's,
so
I'm
gonna
need
an
example
of
an
app.
That's,
not
the
app
I
actually
want
to
run,
but
it
teaches
me
something.
Another
case
is
like
I
know:
I
want
to
run
this
particular
app
on
Kyra,
Nettie's
and
I
want
the
community's
best
practices
to
be
distilled,
I'll,
probably
fork
it
and
edit
it,
but
I
still
just
want
to
watch
that,
like
any
updates
or
security.
Things
that
happen
to
that
I
want
to
see
that
stream.
J
That's
what
I
was
thinking
because
I'm
trying
to
understand
who
it's
for
and
what
the
intent
is
and
if
it's
we're
going
to
take
something
and
follow
its
upstream.
It's
almost
really
creating
like
a
package,
and
you
want
a
package
manager
for
that
or
is
it
an
example
meant
to
teach
somebody
how
to
go?
Do
something.
Those
two
different
contexts
I
think
matter
here
and
maybe
where
some
of
the
confusion
is
like
when
I
go.
Look
at
the
examples,
I'm
thinking
this
is
an
example.
This
isn't
something
I
should
just
necessarily
grab
and
use.
I
I'm,
actually
thinking
about
it
as
a
third
case
I
guess
it
could
be
two
things.
One
is
people
that
want
some
people
might
choose
to
use
something
like
the
cube
sinker
rather
than
helm,
where
they
have
their.
They
just
take
config
this
gun,
templated
or
kit
into
their
revision
control
and
manually
template
it.
Some
people
prefer
that
style.
The
other
option
is,
this
can
be
a
downstream.
It
can
be
downst--.
I
You
can
have
the
helm
chart
be
downstream
from
this,
but
I
noticed
with
the
home
chart
says
the
focus
is
like
on
ease
of
use,
so
some
people
prefer
to
like
you
know,
just
tell
me
what
the
values
are
and
I'll
plug
them
in
and
good
go
and
I'm
good.
Other
people
want
to
like
really
get
into
it,
understand
it
and
tweak
it
which
values
does
not
yeah
Mille
doesn't
really
do
the
power
do
and
when
we
work
with
with
helm,
we
saw
like
we
need
like
a
hundred
different
values
and
values
like
I'm.
I
H
Guess
that
makes
sense,
so
the
audience
of
the
samples
is
that
you
know
I
want
to
deploy
this
thing,
but
I
don't
really
want
to
run
it
like.
It's
I'd
expect
that
it's
not
production
grade
I'm,
just
trying
to
learn
how
the
kubernetes
concepts
fit
together.
How
can
I
use
secrets
within
my
deployment?
How
can
I
you
know,
create
stateful
sites
and
stuff
like
that?
So
that's
I,
think
that
is
the
main
goal
of
the
samples
and
it's
not
meant
to
be
like
this
is
how
I
run
elasticsearch
or
caca.
H
This
is
how
you
should
be
running
it,
and
this
is
guaranteed
that
we're
gonna
update
it
or
anything
like
that,
because
that
operation
doesn't
seem
to
scale.
But
there
might
be
third-party
repositories,
as
you
said,
could
be.
An
organization
could
be
somewhere
else
or
could
be
just
some
person's
repository
that
they
maintain,
but
helm
would
be
a
better
packaging
solution,
at
least
in
general,
but
yeah
you're
right.
So.
K
K
H
Yep,
that's
the
main
goal.
Basically,
you
know
I,
just
listen.
I
just
you
know,
read
the
introduction,
documentation.
I
did
the
interactive
tutorial
now
I
want
to
actually
bring
up
something
myself,
so
I
can
just
go
here
and
click
the
guestbook
on
reddit
I
can
follow
this
tutorial.
This
is
basically
just
a
service
that
has
you
know
it
has
inter-service
communication.
It
tells
you
about
that.
It
tells
you
about
little
bit
of
a
persistent
disk
and
how
can
you
use
secrets
and
environment
variables
and
so
on?
H
So
that's
basically
the
way
I
like
to
see
the
scope
of
the
example
not
like
anything,
real
anything
serious
telling
you.
This
is
how
you
should
be
running
everything
but
close
to
production,
ready
as
much
as
possible,
but
not
necessarily
production
ready
that
you
should
deploy
in
your
company.
So
I
have
a.
L
L
H
Well,
so
today
is
the
thing
we're
trying
to
fix
is
basically
that
problem
that,
as
a
person
who
just
learned
about
queries
by
clicking
storm
here,
I'd
actually
don't
learn
a
lot
other
than
the
fact
that
this
is
how
I
I
guess
how
I
wrong
so
keeper
on
kubernetes.
But
it
doesn't
explain
me
what
I
you
know
what
a
deployment
is
or
what
a
services.
It's
just
like.
It's
telling
me
that,
ok,
there
are
these
yellow
files
or
JSON
files.
H
L
Don't
see
like
there's
no
examples
coming
out
of
this,
where
I'm
like
I'm,
actually
learning
something
about
kubernetes,
and
you
know
it's
a
bunch
of
hi
and
a
replication
controller,
and
it's
not
even
like
calling
out
here's
how
you
use
a
replication
controller
early,
yeah
I,
don't
feel
like
I.
Take
anything
away
from
this
in
terms
of
how
to
better
use
kubernetes
exactly.
I
So
I
think
that,
and
maybe
we're
all
agreeing
I
think
the
use
case,
the
set
of
people
who
are
the
best
at
teaching
people
how
to
come
up
that
learning
curve
on
kubernetes
is
not
the
same
as
set
of
people
who
are
have
the
most
experience
with
a
single
running,
a
single
app
in
production
on
kubernetes,
and
so
we
should
and
organizations
and
repos
center
around
people
that
having
experiences
so
we
it
seems
like
teaching
examples
and
like
for
kiss
and
run
in
production.
Things
are,
should
be
separate
places.
I
I
do
think
I
think
using
an
organ
makes
sense.
If
you
think
you're
going
to
have
a
bunch
of
different
contributors
that
have
different
specializations,
then
you
probably
want
to
use
an
org
I'm,
pretty
sure
that's
the
case
for
like
production,
config
I,
don't
know,
oh,
if
that's
the
case
for
examples
or
not
I.
Don't.
H
I
can
follow
three
tutorials,
but
not
15,
tutorials,
probably,
and
we
can't
maintain
15
tutorial
simultaneously
at
the
same
time
that
there's
that
too
so
I
think
we
can
keep
the
examples
in
a
single
repository
and
for
the
rest,
we
can
create
an
organization,
as
you
described,
and
probably
encourage
people
to
know
things
over
there.
But
that
will
be
a
good
move.
I
think
so.
H
H
H
So
there
is
actually
the
guidelines
under
the
examples
I
agree
which
only
the
main
thing.
The
examples
follow
and
the
rest
just
doesn't
follow
that.
So
it
actually
says
that
what
is
an
example
and
what
is
not
here
but
I
assume
that
this
is
probably
a
little
bit
out
of
date
and
as
it
basically,
this
says
that
example
should
have
a
clear
section
saying
this
is
what
you're
going
to
learn
and
it's
going
to
walk
you
through
all
these
steps.
H
J
H
Ish
the
problem
with
what
I
see
with
the
org
is
that,
let's
say
if
we
have
an
example
of
and
I'll
just
click
storage
and
then
pick
on
hazel
caste.
If
this
is
not
being
updated,
if
it
doesn't
have
the
maintainer,
the
organ
does
not
solve
that
problem.
It's
the
will
still
have
examples
that
don't
have
meat
equality
in
the
first
place.
We
can
reject
the
example
that
don't
meet
the
bar
in
the
first
place,
but
that
doesn't
necessarily
mean
that
people
will
keep
maintaining
them.
I
Was
thinking
that,
if
you're
a
company
that
builds
something
about
the
criminals,
you
should
probably
just
host
out
on
your
own
github
I
was
thinking
about
the
repo
I
was
talking
about
as
being
for
people
who
work
full-time
on
the
turbinates
open
source
and
whose
primary
job
is
to
show
how
to
do
application
integrations
getting
a
place
to
house
stuff.
It's
not
really
an
advertising
space.
H
So
I
wonder
how
many
people
we
have
that
works
on
the
open
source
project
and
want
to
show
people?
How
do
you
spark
of
God
Sarah,
because,
ideally,
we're
kind
of
evolving
all
these
things?
How
much
are
John
I
know
a
few
people
who
are
actually
trying
to
work
on
that
and
I.
Don't
know
how
many
people
actually
want
to
keep
maintaining
these
things,
because
both
of
these
are
basically
not
maintaining
I
click
through
every
single
one
of
them.
Only
a
Fievel
can
are
actually
maintained
in
the
past
six
months
or
so.
I
H
K
K
What
are
the
issues
they're
just
doing
that,
so
we
have
all
of
this
and
there's
a
lot
of
good
material
in
there
for
an
example
of
because
otherwise
my
concern
is
we
either.
You
have
a
large
divide
between
very
basic
examples
and
just
going
into
the
piece
of
discriminative
occupation,
documentation
to
figure
out
how
to
fill
an
appropriate
application
using
the
directive
guys
and
then
there's
this
large
chasm
to
helm,
charts
right.
K
H
Totally
right,
I
think
some
of
these
are
actually
I,
don't
want
to
just
delete
them.
That's
why
I,
probably
I
think
we
should
keep
them
here,
at
least
for
a
while
and
gracefully
migrate
them
to
another
repository.
The
thing
is
that
half
of
the
stuff
that
you
just
described
are
actually
in
examples.
I,
three
and
half
of
the
other
stuff
are
in
the
contrapositive,
and
there
is
kind
of
a
lot
of
overlap
between
these
two
things
and,
as
you
said
most
of
these
things,
are
they
not?
H
They
might
be
the
only
example
that
actually
show
up
on
the
search
results
today
when
I
search,
I,
don't
know,
corn
is
a
spark,
so
we
can
move
them
somewhere,
I'm,
just
not
sure
where
at
this
point,
so
if
you
have
any
ideas,
feel
free
to
chime
in
in
the
talk
or
you
know.
Definitely
let
me
know
if
you
have
a
preference.
H
Right
now,
I
have
no
idea.
Both
are
basically
just
data.
They
turned
into
the
into
a
dumpster
of
fortune.
This
says,
do
not
add
any
projects
of
this
repository.
I
am
assuming
that
this
is
an
except
a
lot
of
PR
lately
and
the
other
one
is
kind
of
in
the
same
shape.
I
mean
they
don't
accept
new
content
here
as
far
as
I
know
so,
I
don't
know
the
historic
difference.
I
think
one
of
them
used
to
have
more
like
documentation.
Type
of
things
like
this
is
how
you
run
cube
DNS.
I
A
It's
all
right,
I'd,
like
to
just
chime
in
here
with
one
last
thing:
I
do
like
a
way
that
chart
the
chart.
Format
allows
you
to
like
version
things
and
you
don't
necessarily
have
to
use
the
values
file
you
can
just
use.
You
can
just
put
raw
brownies
in
the
templates
directory,
so
I
don't
know,
I,
think
that
would
be
interesting
to
explore
an
idea
and
I
understand
that
not
everybody
uses
home
and
that's
totally
fine,
but
explore
the
idea
of
having
a
repo,
which
also
contains
like
home,
charts
that
are
not
temple.
A
It's
not
template
and
just
how
the
mo
file,
because
that
way,
it
makes
it
easier
for
a
person
to
update
and
version
and
kind
of
keep
track
of
like
this.
This
chart,
which
is
not
completed.
These
said,
OPM,
yellow
files
basically
work
on.
You
know
this
version
of
kubernetes
and
so
on,
and
there's
like
an
automated
way
to
make
sure
that
happens.