►
From YouTube: Kubernetes SIG Apps 20170911
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
A
So
before
we
get
into
demos
or
anything
of
that
nature,
we
have
a
few
announcements.
The
first
one
is
for
anybody
who
didn't
know
the
1.8
feature.
Freeze
has
happened
and
now
we're
looking
for
bugs
and
stability
fixes
and
in
the
near
future.
Here
we're
actually
going
to
start
the
1.9
planning,
so
we
can
start
planning
some
of
that
stuff.
Not
today.
A
The
next
thing
that
we
haven't
heard
the
announcements
is
when
it
comes
to
see
gaps
related
tests,
the
ones
that
are
the
flaky
estar,
the
stateful
Sat
ones.
There
is
a
link
in
the
agenda
to
the
test
grid
to
see
some
of
that,
but
it
turns
out
that
stateful
sets
is
where
we're
currently
flaky.
So
somebody
working
on
stateful
sets
has
time
to
go.
Take
a
look
at
that
see
what
issues
are
created,
try
and
clean.
Some
of
that
up.
A
It
would
be
appreciated,
and
then
the
third
thing
we
have
is
there's
a
couple
of
Doc's
out
that
people
are
looking
at
on
how
we
do
application
management
in
kubernetes,
and
some
of
this
came
out
of
Brian
and
so
we're
looking
at.
How
would
you
do
just
manage
applications
which
is
exactly
what
we
talked
about,
and
so
these
are
kind
of
long
reads
and
they
might
be
a
little
bit
painful,
but
if
you
can
take
a
little
bit
of
time
anybody
who's
interested
and
read
them
and
provide
feedback
or
thoughts.
A
A
B
Okay,
great,
there
might
be
a
little
background
noise,
but
quick
question
about
the
code.
Freeze,
so
I
found
a
a
bug
in
one
of
the
helm
PRS
that
I
made
that's
in
two
point:
six
and
it's
well.
You
could
consider
it
a
bug
or
maybe
just
changing.
It
might
be
a
feature
request
and
I'm,
not
really
sure
which
it
falls
under.
It's
about
deleting
the
ability
to
delete
a
key
value
ski
through
him,
with
the
null
value
with
with
the
null
value
that
does
not
properly
work
for
struts,
so
I
would
have.
B
Interestingly,
that
was
my
main
reason
for
doing
it.
But
when
we
wrote
the
test,
we
just
wrote
a
relatively
simple
test
that
worked.
Fine
and
I.
Think
I,
unfortunately
assume
that
that
would
descend
to
all
of
its
children.
But
doesn't
that's
not
the
case
when
it's
destruct,
so
what
would
you
advise
for
I
mean
obviously
opening
a
new
PR?
But
what
would
you
advise
for
defining
that,
as
would
that
be
considered?
Fixing
a
bug
fix
since
it's
already
a
feature
in
there,
but
it
wasn't
fully
worker.
A
D
E
Okay,
yeah
I'm
not
familiar
with
that
feature.
This
is
Adam
here,
one
of
the
helm
in
here,
so
the
the
helm
project
is
not
under
feature
freeze.
If
that's
what
you're?
Oh,
please!
If
that's
what
you're
concerned
about
it's
only
a
good
riddance
grenades.
Oh
you're,
talking
about
this.
B
A
I'm,
sorry
yeah,
that's
a
great
thing
to
bring
up.
This
is
kubernetes
kubernetes,
not
the
supporting
projects.
Those
may
have
their
own
schedules
and
how
has
its
own
schedule
all
right,
and
so
with
that
I
guess
we
can
move
on
to
the
demos
and
the
first
one
is
of
Kubis
the
functions
as
a
service
at
Nanyang,
hey
yeah,.
F
G
G
Whenever
you
add
a
chart
table
or
two
to
repository
or
a
bucket,
it
would
be
nice
if
there
was
a
if
there
was
something
running
that
could
pick
that
change
up,
regenerate
the
index
and
go
and
upload
that
index
again.
So
that's
kind
of
what
I've
built
with
you,
which
is
the
functions,
is
a
service
platform
for
kubernetes,
as
matt
said
so
I'll
just
go
ahead
and
share
my
screen
here.
So
if
you
haven't
heard
of
cube
less,
you
can
find
it
over
it
get
hub
con.
Slash!
G
G
G
Example,
functions
for
cube
this
and
go
ahead
and
install
them
and
I've
just
contributed
this
it's
an
important
era,
so
it
should
be
there
soon
and
this
charts
repo
function.
So,
if
I
go
to
Doc's
here,
it's
pretty
easy
to
get
up
and
running.
The
first
thing
you
need
to
do
is
create
a
a
topic
to
listen
on,
and
this
is
using
cube,
Lissa's
kafka,
back-end,
so
I'm
creating
a
topic
here
called
charts.
G
I've
already
done
this,
so
I'll
skip
over
that
step,
and
the
next
thing
I'm
gonna
do
is
install
cublas,
sorry,
install
menu
and
the
reason
I'm
using
Mineo
is
because
it
integrates
really
nicely
with
with
caf-co
here.
I
just
I
could
just
point
it
to
the
caf-co
running
on
my
own
cluster,
in
a
cube
listening
space
here
and
type
of
topic
to
use
and
I'll
go
and
send
events
to.
D
G
B
G
B
G
Actually,
I
decided
to
not
use
mini
cube
to
to
make
sure
that
this
goes
faster,
but
still
good,
yeah,
all
right
so
I
have
an
IP
address.
I
can
go
ahead
and
this
host
here
I
haven't
configured
Mineo
with
any
specific
secret
key,
so
just
just
using
the
defaults
here
that
the
chart
generates
for
me
and
then
I'm
gonna
go
ahead
and
create
my
bucket
so
mcmd,
my
you
know,
charts
and
because
I
need
to
access
this
over
HTTP
I'm
going
to
give
it
the
download
policy.
G
G
So
please
show
you
what
this
looks
like:
I
won't
go
into
too
much
detail
how
it
works.
You
can
look
at
the
code
later
and
figure
out
figure
it
out
yourself,
but
effectively.
It
connects
to
Mineo,
it
does
a
helm
install
here
and
it
listens
to
so
that
the
handler
gets
run
its
function.
That
gets
run
every
time.
G
There's
a
message
on
that
top
pik
Mineo
is
is
sending
out.
There's
those
messages.
It
then
goes
and
downloads
that
chart
package
and
then
processes
it
and
updates
the
index
dot.
Yama
file
so
pretty
soon
so
go
ahead
and
destroy
us
now
and
we
need
to
provide
here
is
the
repository
URL,
which
I'll
set
to
this
here.
G
H
G
B
G
G
G
Yeah
you
can
as
long
as
you're
running
kubernetes.
You
can
you
can
you
know,
move
it
about
okay,
so
I,
basically
just
uploaded
this
Redis
table
over
here
and
I've
already
got
this
index
element.
You
can
see
it
was
kind
of
been
rated
at
the
same
time.
So
now
I'll
add
this.
To
my
I.
This
is
a
helm
repo.
G
G
D
G
G
You
know,
with
with
the
s
request,
free
or
with
Google
Cloud
Storage
buckets
as
well.
So
it's
just
a
very
basic
idea
of
how
you
might
want
to
manager
to
our
repository.
It
obviously
doesn't
go
further
than
that
and
Advent
occation
and
authorization
which
I
guess
we'll
see
next
we've
with
the
Quai
app
registry.
But
it
gives
you
a
kind
of
simple
idea
of
how
this
could
be
managed.
So.
A
Suffix
on
the
name,
fires
off
an
event
into
Kafka.
That
Kafka
event
is
then
picked
up
by
cube
lists
that
says:
oh
I
know
I'm
supposed
to
fire
off
this
function.
For
it,
it
fires
off
the
function
that
picks
up
the
packages
and
regenerates
the
indexed
imo
file.
That
hell
knows
how
to
read.
Is
that
right,
exactly
yeah,
you
summed
it
up
for
me
pretty.
G
B
B
G
B
Could
be
implemented
separately,
but
it's
a
it's
a
it's
just
a
different
model
than
the
than
the
citrus
repo
I
believe.
This
is
modeled
on
the
charts,
repo
and
the
other
is
just
an
another
way
of
doing
it.
That's
more
more
I,
guess
more
pluggable
and
I'm
just
curious.
If
you
foresaw,
may
be
doing
another
demo
of
this
that
took
advantage
of
that
second
model
in
some
way.
I
don't
I
have.
G
G
Think
we'll
actually
see
it
with
the
next
demo.
I
think
everyone's
gonna
do
a
demo
of
registry
and
what
you'll
find
is
is
Aborigine,
isn't
limited
by
the
index.
What
yeah
more
requirement
that
helm
chart
repos
have
so
app
registry
and
said,
provides
an
API
that
will
then
give
you
the
latest
version
of
that
chart,
because
it
is
server
in
itself.
It's
not
just
a
simple
bucket
with
a
within
XML
file,
yeah
and
you'll
seed.
It
with
the
next
demo
yeah.
A
I
think
a
good
way
to
look
at
it
is
they
the
way
that's
natively
built
into
helm,
and
then
the
app
registry
are
kind
of
two
different
models.
The
hell
methodology
is
built
around
the
way,
something
like
Debian
packages
work,
whereas
the
module
model
with
the
app
registry
that
you
seen
in
quei
and
things
like
that-
appears
to
be
a
little
bit
more
similar
to
something
like
a
docker
hub
model
of
doing
things,
and
so
there's
slightly
different
models.
I
would.
B
G
J
I
J
Soviet
suits
few
feature.
We
don't
get
on
get
so
first
or
so.
It's
agnostic
on
the
format,
so
it
does
not
under
any
chopped
it
can
be
K
so
net
or
any
other
kind
of
package
format
that
we
can
see
in
the
future.
It's
aware
so
to
audit
logs
know,
what's
going
on,
would
deploy
your
application,
how
many
time
etc
on
get.
You
can
discover
easily
it's
like
for
source
code,
so
it
ad,
which
browse
and
search
packages
to
create
like
a
specific
communities.
J
App
Store,
it's
really
so.
The
idea
of
three
diseases
like
it's
immutable
and
it's
based
either
on
semantic
questioning
or
you
can
address,
also
contains
through
digest.
Once
also
thing
is
like
machine
query
we
can.
We
can
see
that,
but
on
gates,
most
most
repo
is
like
milky,
chart
people
and
make
it
very
complex
to
say
to
use
the
tag
mechanism
to
say:
I
want
this
particular
fashion
questions
you
have
24.5,
for
example,
one
so
part
I
think
is
very
interesting
in
a
registry
is
a
concept
of
chairs
we're
introduced
it.
D
J
Can
sink
a
feature
that
are
very
much
more
specific
to
communities
that
we
can
can
have
on
git,
and
there
is
many
many
other
feature
mostly
I
listed
on
run
security,
so
we
can
have
signing
up.
You
can
sign
application
before
deployment
and
check
this
and,
as
you
know,
unquestionable
service
like
vulnerability,
a
detection
on
image,
but
then
we
can
just
inspect
the
resources
and
find
the
image
and
propagate
this
information
to
the
application
level.
J
So
a
quick
note
about
the
hammer
proof
I
can
describe
so
it's
complex
to
to
to
to
manage
or
the
version
there's
no
discovery
and
one
thing
that
people
have
been
reported:
that
app
registry
service
is
particularly
in
the
CI
CD
use
case.
So
in
ham,
you
need
to
index
to
regenerate
index
on
the
client
side.
So
basically,
if
you
have
two
jobs
at
doing
this
at
the
same
time,
they
will
just
overwrite
each
other
and
average.
J
J
One
of
the
idea
is
like
to
unify
this
kind
of
workflow
that
could
be
in
compose
with
a
K
T,
so
net
or
ham
will
repeat
to
begin,
but
it's
at
Russells
IDs
like
we
really
have
like
one
store
or
application
storage
for
all
kind
of
communities,
application
yeah
we
tried
to
do
a
shot
demo
to
see
how
to
use
it.
So
I
will
show
you
is
through
the
EM
plugin.
So
there
is
like
a
up
registry
command-line
tool
and
we
created
also
a
plug-in
that
use
it.
J
So
I
say
you
can
search
I,
don't
know,
let's
say
for
Jenkins
and
say
on
which
we
just
really
want
to
search
ability.
I
say
so.
You
can
find
different
things.
So
here
I'm
using
cuidate
you,
but
as
I
say
it's
it's
open
source,
so
quit
just
an
implementation
and
actually
just
using
the
open
source.
You
can
find
it.
So
you
can
just
run
your
own
on
registry.
I
think
is
already
running
somewhere.
J
J
J
Ok,
so
here
it
was
download
from
the
server
the
chart,
and
so
here
comes
the
streams
that
are
different.
For
example,
if
I
want
to
edit
the
chart-
let's
say
I,
add
a
dimension,
so
this
can
be
made
from
your
CI
I
can
make
this
available,
for
you
want
by
just
say,
M
registry
push
we're
at
I/o
and
which
namespace
I
want
to
push.
J
J
So
we
have
a
description
like
this
is
one
I
said
one
implementation,
but
this
can
be
implemented
also
in
different
UI,
like
molecular
etc.
Everything
is
open,
so
I
really
invite
everyone
to
comment
on
the
API
and
spec.
So
we
can
progress
on
this.
So
you
say
here
the
question
I
just
poured
and
yeah.
So
this
is
a
China,
for
example,
just
a
bad
point
to
the
zero
three
zero.
So
if
I
want
to
border,
let's
say
I
will
show
you
the
encirclement.
J
J
And
that
is
so,
you
see
at
the
moment.
I
just
I
never
use
the
index
document
so
upstream
one.
It's
never
my
look.
I
have
nothing
I,
didn't
fetch
any
repo
or
synchronize
anything,
it's
all
server-side.
So
if
we
cry
clients
at
the
same
time
gets
the
same
question
everywhere.
So
it's
very
easy
to
be
sure
that
you
get
the
latest
version
and
push
and
Nvdia
pipeline,
because
each
client
doesn't
have
any
state
anymore.
I.
J
G
G
B
Have
a
quick
question:
I
posted
it
in
the
in
the
chapter.
Also
admin
that
sounds
fun.
That's
really
amazing.
I
would
love
to
see
that
is
just
to
help
helm
registry
and
upgrading
a
chart
upgrading
it
differently.
I
mean
how
would
one
do
that
because
I
my
hair,
look
less
I've
been
hosting
some
charts
on
Quay
and
my
earlier
tests
in
an
attempt
to
upgrade
all
pretty
much
failed,
so
I've
kind
of
known
and
then
just
at
a
local
chart
employ
rather
than
the
registry
to
do
that.
J
J
J
J
J
A
J
G
J
J
J
G
J
Is
there
is
an
open
issue
about
that?
So,
yes,
Hema
implemented
like
plugin
for
fetching
different
protocol,
but
it's
only
for
the
M
fetch,
so
I
created
a
plug-in
for
M
fetch
something
so
this
will
work,
but
it's
not
yet
working
in
the
requirement
of
German
soldiers
and
she
opened
about
that.
Okay,
this,
the
thing
like
you
see
it's
Mona's
temporary
walk
around
until
this
is
implemented
in
him,
but
the
yeah
really
really
is
that
when
I
say
I
would
like
this
to
be
more
extreme
or
more
integrated.
J
G
F
J
G
Does
that
reduce
you
do
anything
different
with
dependencies
so
Wilhelm
for
example?
What
you're
supposed
to
do
is
fetch
all
the
dependencies
beforehand
and
then
package
and
then
push
it
up.
Basically
bender
the
dependencies
in
does
a
purchase.
You
follow
the
same
model
or
could
app
registry
resolve
dependencies?
When
you
do
a
you
know,
a
helm,
fetch
or
a
Hellman
store.
J
J
B
A
J
A
This
was
great,
thank
you.
Did
you
have
anything
else
all
right
great?
So
then
we
roll
into
the
discussion
topic.
So
we
saw
a
little
bit
about
using
Helms
core
feature
and
you
can
have
your
own
registry
and
then
we
talked
about
the
app
registry
which
is
implemented
in
quei.
So
now
we
are
looking
to
just
kind
of
have
an
open
discussion
on
how
to
run
a
private
chart.
Repository
right,
you've
got
your
charts
for
your
applications.
You're
using
a
charts
may
be
helmet
to
manage
or
your
flow
of
it
deploying
things.
A
What
do
you
use?
How
do
you
use
it
things
like
that?
What
are
the
options,
so
we
looked
at
you
know,
there's
your
your
own
charts
repository,
that's
using
the
helm,
setup
or
home
can
generate
the
index.
Gm
will
file
for
you
and
you
can
do
something
like
that
or
there's
using
app
registry.
Those
are
kind
of
the
two
choices
we
tend
to
talk
about.
Does
anybody
know
of
any
other
choices.
F
B
In
fact,
that's
what
we
do
and
my
mainly
the
choices
that
I've
tried
to
make
is
to
take
seriously
what
was
said
on
the
past:
calls
for
the
charts
to
the
helman
charts
development
columns
that
natal
charts
are
meant
to
be
production
worthy
or
Enterprise
production.
Where
are
they
even
or
small
scale?
B
B
We
definitely
are,
but
that's
more
of
a
little
shift
model
where
we're
we're
just
getting
it
up
as
fast
as
we
can
in
order
to
try
get
our
infrastructure
other
things
onto
Elms,
so
that
they
can
be
packaged
properly.
But
then,
as
we
are
able,
we
try
to
look
for
the
deltas
and
say:
oh:
is
that
application?
Does
it
already
have
a
stable
chart?
If
so,
why
can't
we
use
that
stable
chart?
What
would
we
need
to
add
to
in
a
generic
way
to
allow
us
to
use
it
too,
and
then
we.
A
A
B
C
Yeah
over
here
at
Center
for
open
science,
we
just
took
a
really
basic
approach,
initially
and
just
reusing
Travis
and
some
shell
scripts
to
do
the
helm,
repo
index
and
all
that
and
then
push
it
up
to
github
pages
and
eventually
abandoned
that
because
Travis
our
builds
back
up,
because
we
have
so
much
stuff
running
on
there
all
the
time
and
just
did
it
the
same
thing
essentially
with
Jenkins.
But
it's
just
really
a
really
simple
shell
script.
I
put
a
link
in
the
chat
for
it.
A
K
We
are
looking
at
in
fact,
we
are
already
using
quays
at
registry
for
a
couple
of
our
open
source.
You
know
public
charts,
like
the
alb
ingress
controller.
We
are
very
interested
in
deploying
quai
ee,
weird
tectonic
users,
so
we're
very
interested
in
deploying
quai,
Enterprise,
Edition,
privately
and
then
kind
of
evaluating
that
for
use
instead
of
the
s3
route.
G
K
G
K
A
So
to
jump
ahead
here
a
little
bit
because
somebody
asked
the
question:
what
should
chart
repos
look
like
in
helm?
Three,
and
you
know
you
bring
up
the
idea
of
security
and
there's
these
different
models.
Does
anybody
have
any
feedback
in
what
they
would
like
different
in
charts,
because
I've
noticed
there's
varying
practices
here?
So
I
would
say
you
know
asking
the
question
like
what
are
good
practice
is
I,
see
a
number
of
varying
practices
and
we're
trying
to
solve
some
more
problems,
but
sometimes
we've
got
differences.
G
K
I
think
the
the
index
diamo
approach,
while
you
know
very
very
simple,
yeah
I,
had
says
you
know-
is
also
kind
of
dangerous
you
it's
pretty
easy
for.
You
know
to
stomp
on
that
index
tag
animal
file-
or
you
know,
over
override
it
completely
so
and
I
will
say
we're.
K
One
of
the
reasons
we're
so
interested
in
a
precious
tree
is,
is
because
of
the
idea
of
channels
you
know
being
able
to
have
a
stable
channel
and
an
unstable
channel,
or
you
know
a
beta
beta
channel
to
be
able
to
publish
to
each
and
promote
you
know
between
them
and
then
deploy
from
each
channel.
You
know
to
our
respective
clusters:
I
think
that
has
a
lot
of
potential
live.
I
would.
B
Like
to
ask
how
how
what
the
best
path
for
it
is
for
for
tying
a
registry
for
making
a
registering,
discoverable
and
I
say
that,
because
I
mean
quays
a
really
great
tool,
but
you
know
we
have
charts
infrastructure
charts
for
a
reason.
Now
there
are
centralized.
B
G
Yeah
it
would,
it
would
be
able
some
if
you
could
just
deploy
something
to
your
cluster
and
that
would
just
turn
into
a
chart
repository
and
and
if
they
even
came
of
a
UI.
You
know,
maybe
it's
something
that
we
could
integrate.
Somebody
can
easily
integrate
with
with
monocular,
and
then
you
have
a
UI
plus
chart
repository
just
kind
of
exactly
yeah
that'd,
be
amazing.
J
From
my
point,
like
obviously,
I
would
like
for
v3
to
see
a
registry
more
this
version
or
another
version,
the
default.
What's
the
best
path
to
go
through
so
currently,
I'm
talking
about
a
registry
just
leave
in
a
separate
organization.
Should
we
try
to
incubate
it
or
Oh,
can
be
more
close
to
the
community
and
and
find
a
way
to
be
extremely
ham
or
the
youth.
A
A
Alright,
so
it
seems
like
there's
a
bit
of
interest
in
a
prejudice
tree
compared
to
the
traditional
home
way
of
doing
it.
Is
there
anybody
who
disagrees
with
that?
Who
think
you
know
what
I'd
like
the
way
helm?
Does
it
more
apt
get
debian
style
compared
to
the
registry?
I'm
curious,
because
I
hear
some
poll
in
one
direction,
I'm
curious.
If
anybody
wants
to
give
a
counter-argument
to
that.
A
All
right
silence
mean
there:
isn't
that
much
interest
in
it?
Okay,
that's
that's
a
pretty
good
indicator.
Actually,
so
so
thank
you.
If
somebody
wants
to
disagree
offline
or
doesn't
have
seen
that
Fisher
doesn't
have
strong
opinions,
but
if
anybody
you
know,
has
any
opinions
offline
that
appreciate
hearing
that
I
think
this
is
a
good
thing
to
feed
back
into
the
helm
folks
as
well
so
yeah.
G
There's
not
many
of
us
home
folks
on
the
core.
I
will
actually
say
one
thing:
one
Pro
that
Ford
the
the
current
model
is
that
it
is
just
very
simple
to
set
up
and
it's
a
small
pro
but
I
think
it
is.
You
know
if
you're,
just
playing
something
playing
with
it
and
trying
it
out.
It
can
be
a
lot
simpler
than
having
to
set
it
up
and
effectively
a
registry
for
this.
A
So
so,
since
this
appears
to
be
a
little
dead
in
the
water
direction,
I
do
have
a
question
on
best
practices,
I'm
curious.
How
different
folks
do
this
so
you've
got
charts
and
you've
got
it
in
your
CI,
maybe
even
in
your
CD
setup
right
and
so
a
change
comes
in
and
now
you've
got
a
new
image
tag
for
your
application
and
you're
gonna
update
your
chart
to
correspond
to
that
image
tag
because
you
don't
just
want
to
use
latest
right
because
who
knows
what
latest
is
gonna,
be
you
want
a
title,
a
couple
them?
A
How
are
you
tagging
things
are
using
semantic.
Versioning
are
using
sha.
Hashes
are
using
dates
using
some
combination
of
them,
and
how
do
you
actually
go
about
updating
the
Charter
he
actually
just
manually
like
using
said
and
some
other
bash
script
things
to
update
the
tag
in
the
chart
before
you
push
it?
You
know
I'm
kind
of
curious.
What
are
people
doing,
because
I've
heard
a
number
of
different
ways
on
this
particular
part.
K
Yeah
I'll
go
more
often
than
that.
We
leave
the
tag
either
defaulted
to
some.
You
know
some
value
or
we
leave
it
unsaid
altogether
and
them
either
said
it
directly
from
the
command
line
using
their
set.
Or
you
know
if
we
maintain
separate
values
files
for
each.
You
know,
release
we'll
set
it
in
that
file.
A
B
Yeah
and
I
just
want
to
weigh
in
on
that
one
I
have
actually
been
the
same,
so
it
is
Michael
and
I
we're
very
new
to
this
on
the
production
level
and
right
now,
actually
only
our
infrastructure
support.
More
and
more
of
our
infrastructure
is
being
deployed
to
criminate
easily
on
charts,
not
our
public
facing
projects.
Yet,
although
I
believe
that's
pretty
quickly
coming
and
we've
been
thinking
a
lot
about
this,
as
how
will
we
be
handling
world
acts?
I
think
part
of
the
problem
is
some
applications.
I
think
it
depends
on
the
application.
B
F
You
could
always
do
a
thing
where
you
always
move
forward
and
if
you're
using
a
per
industry,
you
can
have
like
a
temporary
channel
which
is
like
currently
trying
to
roll
back
this
thing
and
it's
not
a
public
channel.
It's
not
what
other
people
are
using,
it's
just
for
your
production
instance
and
temporarily.
So
you
get
to
the
next
stable
version
that
actually
fixes
the
problem.
And
then
you
can
revert
back
into
a
normal,
stable
channel
or
beta
Channel.
A
G
Did
have
one
last
minute
question
kind
of
going
on
from
or
talking
about
how
do
people
actually
version
their
charts,
not
just
the
images
that
go
in
them,
because
this
is
something
that
has
come
up
in.
You
know
the
kubernetes
charts
Reaper
quite
a
bit
and
we've
been
we've
been
kind
of
slowly
trying
to
make
people
version,
charts
separately
to
the
application
versions
and
then
have
a
separate
app
version
field
just
for
the
application
version.
But
but
I
was
curious.
However,
people
are
doing
it.
Yeah.
J
G
F
A
Yeah
those
are
considered
pre-release
tags,
though
not
actual
release,
tags
under
semantic
versioning
and
a
whole
bunch
of
tooling
out.
There
says:
oh,
these
are
pre
releases,
which
means
they're
not
considered
API
compatible
with
it,
and
so
it
gets
a
little
funky
under
semantic
versioning.
For
that
right.
E
A
G
A
A
It
wouldn't
the
way
we
had
the
way
I
had
done
it
in
the
past
is
I.
Would
kid
I
would
do
the
chart
configuration
separately
from
the
application
and
semantically
version
that,
because
your
configuration
could
radically
iterate
over
time,
I
mean
the
way?
Let's
just
take
a
simple
WordPress
setup
right,
you've
got
word
price.
You've
got
your
data
base.
Maybe
you
make
a
major
change
to
the
way
you're
doing
your
chart,
but
yet
your
WordPress
versions
and
your
database
versions
don't
change,
but
you've
come
up
with
a
better
way
to
do
the
chart.
A
G
That's
kind
of
what
we've
been
going
on
going
within
and
giving
at
these
charts
a
lot
of
people
come
to
us
and
say
that
they
want
to
see
the
application
version
more
than
the
actual
chart
version.
But
there
are
those
times
when
you
want
to
see
the
chocolate,
and
you
know
when
something
important
changes
like
there's.
A
bug
fix
and
the
way
ingress
is
configured
for
example,
and
you
need
to
you
know:
an
operator
needs
to
know
that
that
version
has
been
updated
and
rolled
out
a
new
version
out.
So.
A
Yeah
yeah
and
I'll
give
you
a
practical
example
in
Church
repo.
There
are
some
things
in
the
church:
stable
repo
that
don't
work
in
1:7
right
now,
because
of
things
that
had
to
do
with
feature
changes
and
things
like
that,
and
you
might
upgrade
the
chart
to
fix
all
of
those
things.
But
your
application
versions,
don't
change
at
all.
G
That's
exactly
why
we
added
the
I,
don't
I'm,
not
sure
if
everyone
has
moved
over
to
me
yet
do
you
ever
issue
with
that
version?
Is
the
helm,
the
helm,
our
CLI
doesn't
expose
it.
So
if
you
do
a
home
search,
not
gonna,
see
the
apparition,
you're
gonna
see
the
chart
vision,
but
that's
that's
certainly
something
that
we
probably
should
add
to
to
the
helmsley
a
tie
and
make
that
more
exposed
and
I.
Think
that
probably
the
best
way
going
forward
all.
A
Right
folks
were
two
minutes
passed.
Thank
you
for
such
a
healthy
conversation.
I've
seen
most
folks
have
dropped
off,
probably
because
they've
got
the
next
round
of
meetings
to
get
to
so
I'm
gonna
cut
it
off
here.
Folks,
thank
you,
everyone
for
coming
and
having
such
a
great
conversation
today
and
with
that.