►
From YouTube: Kubernetes SIG Apps 20190603
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
Welcome
to
kubernetes
cig
apps
for
Monday
June
3rd
2019
before
it
started,
I'll
go
ahead
and
paste
the
meeting
minutes
URL
in
to
chat,
so
anybody
can
follow
along
with
the
minutes
or
the
agenda,
and
you
can
see
what
we
have
going
on
the
hi
I'm
Matt
Farina
I'll
be
chairing.
Today
we
don't
have
any
announcements
this
week
we
are,
although
I
do
think
code
freeze,
moved
I
didn't
grab
the
update
this
morning
to
when
that
happened,
but
we
are
in
the
midst
of
1:15.
A
B
B
B
B
B
Some
think
that
people
tripped
over
the
power
Warren
died
for
coating
and
then
also
for
partnership.
This
is
basically
just
taking
that
give
me
this
text
in
cable
and
I
tries
to
generate
the
diesel
same
for
probably
shield
and
then
say
today
to
that's
really
using
that
vs
a
cutie,
it's
a
Gradle,
plugin
I,
don't
know
if
you
guys
are
Java
based
on
only
go,
kill
myself.
The
copy
is
very
close
to
the
Java
world.
B
That's
why
trail
I've
been
a
committed
against
also
it's
very
extendable,
so
you've
got
keeping
to
oneself
and
on
the
job.
Also
configuring
or
reversed
parties
in
the
DMZ
John
just
extending
easy
to
extend,
and
in
the
end
what
QP
does
is,
allows
you
to
define
these
sources
with
an
annotation.
And
then
you
just
have
the
inner
function
and
it's
just
you're
the
kind
Noam
those
years
on.
B
B
If
it's
just
you
get
a
small
fragment
in
the
built,
that's
like
where
you
say
I'm
using
to
the
next
version,
this
or
official
version
this,
then
you
get
the
the
DSL
in
New
Yorkers.
You
don't
really
have
to
specify
much
to
get
going.
You
should
look
at
the
narrative
structure.
The
kids
are
shouting
outside
I'm,
going
to
close
it
off.
B
B
B
B
You
can
have
files
that
are
encrypted
or
properties
just
value
this,
for
example.
Here,
if
you
initiate
the
more
money
you
also
get
year,
you
guys
were
here.
L,
don't
get
dinner
in
that
caucus
file.
For
that
dining
and
you
can
add
thousand
creative
I'm
just
using
a
very
simple
strategy
here.
It's
just
physics,
violence
and
come
up
with
anything
better.
Just
how
can
you
be
sure
to
have
direct
detection
of
what
is
the
secret
annex
part?
One?
It's
a
it's
just
a
piece
which
apparently.
B
B
B
A
A
B
A
B
B
B
B
A
B
But
we
we
have
like
a
few
beckons
and
do
some
times
and
everybody's
on
there
and
the
department
beyond
just
by
ones
and
like
you
start
all
over
and
there
one
last
one
and
the
solution
of
that
and
just
use
a
function
to
wire
to
continually.
They
only
have
one
specification
of
that
seriously.
Yeah
copies
like
five
times.
E
A
All
right,
the
next
thing
we
have
on
our
agenda
is
to
talk
about
coop
con
unless
other
folks
had
questions
about
and
feel
free
to
put
them
into
the
chat
as
we're
going
along.
So
if
there
were
any
other
questions
about
this
or
if
you
want
to
go,
take
a
look,
there
is
a
link
to
the
demo
kubrick
in
today's
meeting
minutes
and
feel
free
to
answer
any
questions
and
text
or
possibly
reach
out
offline.
The
next
thing
we
have
it
is
coop.
A
F
Yeah
I
think
he
went
okay,
yeah
I
mean
we
booked
out.
It
was
like
an
hour
in
an
hour
or
so
session
a
bit
longer,
but
it
ended
a
bit
earlier
than
that,
as
the
conversation
kind
of
petered
out
quite
quickly,
I
think
there
was
a
I
think
most
of
the
people
there
weren't
that
full
of
questions
really
so
having
the
kind
of
audience
drive.
The
conversation
didn't
didn't
work
that
well
potentially,
and
maybe,
instead
of
just
being
it
as
bored
as
just
like.
What
do
you
want
to
talk
about?
F
G
G
G
D
A
So
talking
to
Adnan
about
this
one
of
the
things
that
he
noticed
was
it
was
the
same
problem
and-
and
we
talked
a
little
bit
about
it,
such
as
when
we
first
started
doing
these
years
ago.
One
of
the
big
ideas
was
to
get
people
together
to
have
conversations
and
you'd
have
conversations
on
a
variety
of
things.
In
fact,
I
remember
back
in
I
think
it
was
Seattle,
not
the
last
Seattle
one,
but
the
one
before
that,
where
we
had
different
people
at
different
tables.
Discussing
different
topics.
A
The
kubernetes
community
that
shows
up
and
has
engaged
at
these
conferences
has
radically
changed.
Since
then,
though,
right
people
who,
who
are
kind
of
at
the
core
of
things,
are
far
more
stretched
and
extended
than
they
used
to
be.
They
don't
have
as
I
when
I
attend
I,
don't
have
nearly
the
free
time
that
I
did
back
in
the
day
to
have
those
side.
A
Conversations
and
a
lot
of
people
are
in
those
same
positions,
and
so
the
ability
to
have
those
conversations
and
to
show
up
may
not
be
there
plus
there's
a
whole
lot
of
brand-new
people
who
are
trying
to
learn
stuff
either
about
you,
know,
workloads
or
how
to
get
involved
with
a
particular
sig
or
something
like
that,
and
so
it
might
be
a
different
audience
than
we
were
targeting.
Even
some
of
the
guidance
says
to
kind
of
target
with
the
way
we
approach
these
and
so
Adnan
was
suggesting.
B
G
There
was
a
question
with
regards
to
how
to
contribute
to
sig
apps
per
se.
I,
remember
pointing
them
to
the
presentation
that
I
was
giving
some
time
ago
about
fighting
controllers
and
to
the
sample
controller
that
we
have.
Maybe
one
of
the
options
could
be
a
code
walkthrough
just
to
explain
how
a
sample
controller
could
look
like
how
does
it
look
like
and
bait
and
and
then
maybe
some
kind
of
a
poll
at
the
beginning
of
the
meeting?
A
Yeah
I'm
not
sure
if
the
cig
apps
mailing
was
folks
would
mirror
the
people
who
would
attend,
but
at
the
same
time,
getting
kind
of
a
pull
would
be
a
good
idea.
Adnan
suggested
something
similar
with
a
code.
Walkthrough
of
may
view
controllers,
and
so
there
might
be
something
there.
Quite
frankly,
if
two
of
you
are
suggesting
it,
that's
a
good
idea
to
keep
in
mind.
Thank
you.
A
G
B
A
A
Twice
three
times:
alright,
then
we'll
move
on
to
the
next
topic
and
the
next
topic,
I
kind
of
wanted
to
throw
an
idea
out
there
and
I.
Don't
know
whether
this
is
a
kubernetes
idea
or
an
ecosystem
idea,
but
just
an
idea
to
throw
out
there
I'm
coming
back
to
the
application.
So
I'll
share
my
screen
here,
real
quick.
A
A
Alright,
so
I
went
to
our
API
Docs
and
this
is
for
1.13
I
did
it
before
when
da14
came
out
just
before
and
I
hit
print
to
PDF,
and
then
I
thought
my
computer
was
broken
and
about
five
minutes
later
a
PDF
showed
up
and
it
turned
out
to
be
over
2,000
pages
to
print
our
API
doc,
site
and
I
said
our
header
and
I'm
gonna
pull
out
everything
deprecated,
because
we've
got
lots
of
things
right.
You've
got
deployment.
It's
you
know
from
the
extensions
API.
A
You've
got
lots
of
duplication
in
there
and
it
reduced
it
down
a
lot
to
over
1,300
pages,
and
this
is
all
of
the
stuff
that
is
the
current
non
deprecated.
They
said,
alright,
if
someone's
just
gonna,
deploy
an
app
or
work.
Look
I'm
gonna
get
rid
of
everything
that
that
you
may
not
need
for
that
stuff.
Right,
there's
stuff
with
nodes
and
stuff
like
I,
don't
need
those
objects,
so
I'm
gonna,
just
reduce
it
to
stuff.
A
I
may
use
when
doing
apps
and
they
came
in
just
shy
of
1,200
pages
when
I
print
to
PDF,
and
that
was
the
first
inclination
that
maybe
we
have
a
gigantic
surface
area
for
the
kubernetes
api
that
you
need
to
work
with
right
because
I
don't
know
about
you
to
me,
that's
a
huge
surface
here
and
that's
just
core
kubernetes
right.
That's
just
the
core,
kubernetes
and
so
I
started
poking
around
which
of
the
kubernetes
or
the
helm.
Community,
charts
and
I
said
all
right.
What
do
we
got
here
and
there's
336
charts?
A
I
said
all
right:
we've
got
deployment
stateful
sites
daemon
sets,
but
how
do
people
use
things
like
pod
disruption
budgets?
Because
if
you
have
a
huge
surface
area,
how
much
of
it
are
they
digging
into
and
I
started
digging
into
some
of
these
things
and
and
really
with
all
those
deployments
and
all
those
stateful
sites,
especially
the
deployments
I,
would
have
expected
more
horizontal
pod
auto-scaling
right,
because
those
are
the
kinds
of
things
that
are
perfect
to
auto
scale
and
quite
often
you
can
get
basic
stuff
with
kubernetes
out
of
the
box.
A
But
it
didn't
see
a
lot
of
people
using
these
and
I
was
surprised
by
even
though
they're
they're,
newer
objects
and
and
I
started
to
wonder
what
does
it
have
to
do
with
the
surface
area?
Is
there
a
point
where
people
hid
in
their
surface
area?
They
say
I've
done
enough
to
get
it
going
now,
I'm,
not
going
to
move
on
to
that
complexity
and
even
went
onto
github.
These
estimates
are
scoped
to
Hamel,
but
you
see
similar
things
and
I
included.
A
Volume
claim
templates
in
here,
along
with
persistent
volume
claim,
because
a
lot
of
people
are
gonna
use
the
volume
claim
templates,
which
are
not
an
object
of
their
own,
but
you
end
up
using
him
and
you're.
Looking
here
and
I
say:
okay,
look
at
how
many
deployments
are
out
there.
I
would
have
expected
more
secrets,
definitely
more
pod
disruption,
budget
and
horizontal
pod
autoscaler.
How
are
people
doing
these
things?
What
does
this
mean
to
the
API
service
area
when
it's
so
big?
And
so
maybe
they
use
just
the
basics
to
get
things
done
right?
A
So
I
started
looking
at
this
and
just
go.
It's
big
it's
bothersome,
and
this
is
of
course
just
kubernetes.
What
happens
if
you
add
SEO
or
cross
place,
the
other
crt-based
things
out
there?
Those
just
expand
the
surface
area
that
somebody
needs
to
know
to
work
with
the
API
and
as
I've
gone
around
just
talking
to
people
I'm
finding
as
I
bring
up
you
know,
I
might
talk
to
a
group
of
five
or
six
people,
like
my
co-workers,
or
something
and
I'll
ask
him.
You
know
what
about
this
object
in
this
object.
A
They
don't
know,
and
they
don't
know
about
these
fields
on
them
and
they
realize
just
how
big
the
surface
area
is
for
things.
They
don't
know,
and
you
know
there's
just
so
much
right-
I
mean
once
you
add
sto
in
you're,
probably
back
up
near
2000
printed
pages.
If
you
were
to
print
it
out
again,
just
to
give
that
idea
and
so
I
asked
the
question
the
smaller
yam
will
help
and
for
those
I
looked
at
multiple
tools
out
there
that
have
tried
to
do
this.
There's
Kej
qej
is
a
Red
Hat
project.
A
It
hasn't
had
any
activity
in
over
a
year,
but
they
did
shrink
down
the
amount
of
PMO
to
do
things,
and
they
had
examples
on
that.
Then
there's
a
project
called
SCI
cube
who
tried
to
do
the
same
thing,
but
they've
had
stagnant
activity
and
it
didn't
really
take
off,
and
so
you
see
this
idea
of.
Let's
just
shrink
down
the
infrastructure.
Yeah
mo
didn't
work,
so
smaller
infrastructure
amel
did
not
take
off.
A
In
fact,
when
I
start
looking
at
this
I,
don't
I
can't
find
projects
that
were
really
successful
in
trying
to
shrink
the
amount,
and
that
led
me
to
the
question
of
what
about
is
it
kubernetes
surface
area?
Is
that
the
problem,
application
developers
and
operators
now
need
to
go,
learn,
kubernetes
isms
and
it's
huge
right.
It's
just
huge
onboarding
tasks.
I
was
talking
to
somebody
I
used
to
work
with,
and
we
said
she
said
everybody
should
go
through
kubernetes
training
and
they
said
how
long
should
the
training
be?
A
A
This
is
one
of
the
things
they
did
they
developed
and
you
can
see
up
here.
This
was
the
quote
from
their
blog.
They
didn't
want
their
app
developers
to
have
to
learn
kubernetes
internals,
so
they
created
a
yam
obey
system
to
do
their
cron.
This
was
the
cron
system
that
they
wanted
to
do
and
you
didn't
have
to
go
to
learn.
Kubernetes
isms.
They
went
and
generated
now
their
second
take
off
of
it
was
actually
sky
config
right.
They
did
this
afterwards,
where
they
kind
of
backed
off.
A
They
need
to
not
know
kubernetes
internals,
I'm
curious.
How
that
went.
I
haven't
had
a
chance
to
chance
to
talk
with
them
recently,
but
coming
back
to
this
original
idea
caught
my
attention,
because
you
don't
have
to
know
kubernetes,
they
can
go
generate
the
kubernetes
objects
based
on
this,
then
it
also
got
me
thinking
about
Heroku
and
cloud.
A
Foundry
has
something
quite
like
this,
where
they
have
a
very,
very
simple,
app
JSON
that
you
can
define
your
application
with,
and
it's
focused
on
the
app
in
the
App
sneed
Cloud
Foundry
is
another
example
that
doesn't
have
talked
rented.
These
are
passes
and
so
they're
gonna
do
a
lot
more
for
you.
But
the
idea
with
these
two
was
to
focus
on
the
application
rather
than
piles
of
infrastructure
stuff.
Now
you
get
some
of
it
in
there
like
domain
names
and
things
like
that.
A
But
it's
it's
a
lot
less
right.
It's
a
smaller
amount
of
stuff
and
I
know
because
I've
worked
with
Cloud
Foundry
stuff
a
lot.
In
fact,
if,
if
rumors
are
true,
they
may
even
be
building
Heroku
now
on
top
of
cloud
foundry
software
or
looking
at
that,
because
they
are
hiring
cloud
foundry
or
kubernetes
people,
but
the
essential
these
right
is
that
they're
not
describing
the
infrastructure
they're,
describing
the
application,
and
from
that
you
can
generate
what
you
need
right.
A
Let's
see,
that's
the
core
essential
part
here:
you've
got
app
people
focusing
on
app
stuff,
and
so
imagine
for
a
moment
you
have
an
application.
Developer
could
define
a
configuration
for
an
application,
not
the
infrastructure.
It
needs
a
description
of
the
application
and
what
the
application
needs
and
then
you
could
generate
kubernetes
infrastructure
stuff
from
that
whatever
it
is
for
maybe
you've
got
to
pass
on
some
stuff
about
the
environment
or
ops,
but
you
could
generate
it
think
about
it
generation
or
compiling
just
that
whole
idea.
Now
imagine
if
somebody
else
came
on
kid.
A
You
know
what
we
want
to
go
run
the
same
application,
but
we're
good.
Instead
of
using
ingress,
we're
gonna
use
SDO
and
you
could
say,
okay
from
that
basic
can
fake
instead
of
generating
an
ingress
and
that
set
up
for
nginx
ingress.
We're
gonna
generate
everything
for
SEO
and
you
didn't
have
to
change
anything
else.
The
app
developer
needed
to
know,
in
fact
they
could
have
had
its
small,
concise,
and
then
you
generate
this
stuff
right.
A
That
idea
just
imagine
that
for
a
moment,
because
it
puts
the
onus
on
app
people
needing
to
know
app
stuff,
and
then
you
generate
what
you
need
for
the
platform,
and
so
the
question
that
I'm
left
wondering
is:
is
this
an
experiment
for
the
ecosystem
or
an
experiment
for
kubernetes
itself
to
take
on
I?
Don't
know
if
we
have
everybody
here
to
answer
that
question,
but
I'm
curious
what
you
all
think
of
this
idea
and
yeah.
C
C
Now
it's
generating
ingress
with
nginx,
or
it's
doing
something
with
if
you
could
just
swap
that
out
and
you
shouldn't
care
as
Def
and
yeah.
That's
also
something
I
think
my
my
gut
feeling
is
that
as
a
platform,
you
won't
be
able
to
cope
with
every
case
out
there.
That's
also
I
am
where
I'm
targeting
coach
and
not
llamo,
because
it's
way
more
flexible
to
do
stuff
and
people
can
just
take
the
DSL
and
build
stuff
on
it.
A
So
so
one
of
the
things
that
came
to
mind
for
me,
especially
when
I
saw
Skyy
config
right
I,
mean
with
that
they
talked
about
things
like
deployments
and
stuff
like
that
which
are
kubernetes
objects,
whoa,
and
so
you
do
need
to
know
a
certain
amount
of
the
kubernetes
internals,
but
they
are
a
Python
shop
or
they've
got
a
lot
of
Python
knowledge,
and
so
you've
got
a
lot
of
people
who
know
Python
there.
So
it's
a
subset
of
Python,
but
I
thought
you
know
where
I'm
at
we
aren't
a
big
Python
shop.
F
C
A
Yeah
and
I
get
it
and
I
work
with
a
lot
of
people
who
don't
have
my
background,
so
I'm
always
trying
to
figure
out
how
do
I
help
people
who
have
got
different
backgrounds
from
mine
that
that's
one
of
the
things
that
I've
been
trying
to
think
about
with
this
a
lot
lately.
Anybody
else
have
any
other
ideas
on
this
I'm
curious.
If
there's
any
other
takes
on
this.
F
They
really
don't
want
to
deal
with.
If
we've
kind
of
focused
the
effort
on
that
level.
Instead
of
on
the
deployment
level,
we
found
the
deployment
levels,
the
better
they
find
easiest
is
basically
coming
from
somewhere
like
pcs,
whether
used
to
defining
a
task
now
I
define
a
deployment
instead.
Now
a.
F
C
Yeah,
that's
also
something
that
I
see
you.
We
have
a
developer
background
myself,
so
I've
come
up
with
the
exhaustive
and
typesafe
stuff,
but
there's
also
a
larger
project
with
yeah,
enabling
teams
that
do
everything
from
platform
stuff
and
they
they
just
create
some
custom
JSON
and
they
just
have
a
four
line
Jason
to
do
a
deployment,
for
example,
and
everything
comes
out
of
it.
C
C
C
C
A
D
A
I
was
expecting
Ken
to
review.
That
Janet
was
one
of
the
co-chairs
for
coop
kaan,
&
EU,
and
so
she
has
put
out
use
the
other
person
on
it.
I,
don't
think
she's
back
in
the
office
until
today.
Actually,
oh,
so
that's
part
of
the
reason
you
haven't
got
it
from
her
I
was
hoping.
Ken
would
get
a
chance
to
look
at
this.
In
fact
it's
assigned
to
him.
So
we
need
to
go
bug
him
on
it.
He's
the
one
that
I
was
expecting
to
take
a
look.
A
D
A
A
Sounds
great
I,
don't
see
an
issue
with
it
again.
I
want
one
of
the
folks
who's
intimately
knowledgeable
with
the
controller
to
approve
it,
and
so
you
let
me
get
on
that
firing
off
an
email
again.
This
is
where
I'd
go
after
Ken
and
Jana
done
this,
so
I'll
go
chase
them
down
and
see
what
we
can
get
do.
Moving
on
these
I'm
happy
to
fire
off
an
email
on
this.
They
look
good
to
me,
but
I
do
want
that.
The
final
sign-off
from
them
no.
D
B
A
A
There's
just
a
lot
to
learn
and
so
I
scored
the
smaller
llamo
help
well
occasion
psyche
you
both
tried
to
solve
that
and
they
didn't
get
massive
uptake,
even
though
they
did
make
things
smaller
and
they
really
haven't
had
much
activity
either.
They've
been
kind
of
well,
neither
one
of
them's
had
much
activity
and
over
here
so
I
would
say
that
smaller
infra
yamo
didn't
take
off.
But
one
of
the
things
that
caught
my
attention
was
from
stripe
and
how
they
handled
cron.
A
They
didn't
want
their
developers
to
have
to
understand
kubernetes
internals
to
use
the
system,
so
they
came
up
with
a
simpler
format
for
cron
and
granted
their
next
go
at.
This
was
sky
config
which
again
required
you
to
know
kubernetes
isms
and
instead
you
were
doing
it
in
a
subset
of
Python,
but
disco
here
really
caught
me,
because
it
was
the
idea
that
your
app
developers
didn't
need
to
know
the
complexities
of
kubernetes.
And
then
this
drew
me
in
to
looking
at
things
like
here's.
A
Here's,
a
brief
snippet
of
Heroku
zap
JSON
cloud
foundry
has
something
similar.
Although
it's
yamo
and
it's
a
simple
description
that,
for
the
most
part,
describes
the
app
there's
a
couple
things
in
there
like
instances
and
stuff
like
that,
but
the
big
target
is
the
application
and
granted
these
are
passes.
A
But
it's
that
idea
that
okay,
an
app
person
deploying
an
app
deals
with
app
things,
and
so,
if
you
look
at
the
cron
in
the
app
the
essential
part
about
these
is
that
they
describe
the
application
rather
than
the
infrastructure,
which
is
what
we
tend
to
deal
with
with
the
long
form
kubernetes
stuff.
And
this
then
you
know
had
me
imagine
for
a
moment.
You
have
an
app
developer,
who
describes
our
application
in
the
simple
configuration
its
configuration
describing
and
targeting
the
application,
not
the
infrastructure,
then
you
use
something
that
you
know.
A
You
generate
the
kubernetes
infrastructure
from
it
and
maybe
got
to
pass
in
some
environment
and
ops
information,
but
you're
able
to
generate
all
the
long-form
kubernetes
from
that.
And
then,
if
you
had
something
like
Sto
later,
you
can
regenerate
it
with
Sto,
and
so
something
like
an
ingress
then
would
be
replaced
by
the
sto
CRTs
in
the
generated
material.
But
your
application
description
didn't
have
to
change
its
the
generated
material.
It
changed.
D
A
I'm,
finding
that
onboarding
people
to
kubernetes
is
is
difficult,
because
the
sheer
amount
of
information
that
people
have
to
learn
to
onboard
the
kubernetes
one
of
the
stories
I
related
was
I
was
talking
to
somebody
to
work
with
and
she
suggested
get
a
bunch
of
people
in
for
training
and
I
said.
Okay,
this
application
has
a
thousand
people
on
it
and
you're
and
she
was
suggesting.
Maybe
you
know
weeks
worth
of
training
and
I
go.
D
A
D
D
A
And
my
thought
had
been
if
you
could
assume
a
container.
So
if
we
just
assume
containers
is
the
basic
and
I
don't
know
if
that's
right,
but
that's
the
assumption,
and
then
you
have
a
spec
that
defines
what's
in
it.
You
could
have
one
two
three
four
things
that
generate
different
things
out
of
it,
and
then
you
just
have
a
comments
back
and
then
it's
not
a
requirement
that
you
have
a
specific
tool,
implementation
and
so
different
people
could
generate
different
things
from
that
comments
back
depending
on
their
needs.