►
From YouTube: Kubernetes Federation WG sync 20180905
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
A
C
Sure
so
currently
we
we
have
installed
latest,
which
I
was
somehow
confused
I
for
some
reason,
I
thought
that
was
gonna
be
I,
want
to
install
the
latest
committed
state
and
it
actually
means
installs.
Some
arbitrary,
like
latest
pushed
release
because
it's
tied
to
the
image,
and
so
really
we
have
to
update
the
image
with
the
tag
latest.
At
the
same
time,
when
we
update
install
latest
IMO
I
find
that
super
confusing
I
think.
C
Ideally,
we
would
have
like
one
path
to
install
the
latest
committed
state
and
probably
a
separate
path
per
released
or
pushed
version,
and
so
those
tags
would
be
like
a
1
there.
Oh
1
alpha
1
alpha
2,
whatever
it
is,
but
for
me,
like
I'd,
like
to
just
have
some
clarity
around
how
you
install
a
given
version
or
something
make
sense.
Yes,.
A
There
seems
to
be
some
confusion
around
the
state
off
from
what
we
have
in
the
reports
of
now
and
the
way
it
is
linked
to
the
images.
So,
first
one
point
which
is
not
term
probably
of
we
can
come,
not
talk
it
as
part
of
our
that
separate
issue.
I
think
there
is
all
the
reason
if
you
are
mounted
that
is
one
confusion
is
because
of
the
weight
tags
are
put
on.
The
image,
so
latest
is
actually
the
tag
which
is
on
the
latest
release
version
rather
than
the
head
of
Master.
C
A
Agreed,
that's
the
normal
thing
like
latest
all
responds
to
the
let
us
commit
it
is
not
a
release,
ideally
yeah
and
other,
or
the
remaining
mechanism
of
installation.
I.
Think
there
already
is
a
is
some
progress
which
is
towards
having
a
chart
for
this,
rather
than
using
I
am
L
here
we
can
in
fact
maintain
both,
but
I
believe
it
might
make
sense
to
duplicate
Yama's
in
favor
of
a
chart
which
can
has
had
been
intended
me
to
the
person
yeah.
C
D
Just
gonna
say
like
I:
I,
have
kind
of
mixed
feelings
about
Helmand
general,
but
I
I
think
that
a
chart
is
probably
a
very
sensible
way
to
to
install
Federation
for
now,
especially
as
we
get
to
the
point
where
we
want
any
of
these
things
to
like
serve
TLS
or
like
a
metrics
endpoint
or
you
name,
it
Gamal
is
gonna.
Just
a
raw
Gamal
is
gonna
fall
short.
C
A
No
I
did
not
mean
that
you
we
won't
have
to
manually
make
changes.
Probably
a
charge
would
be
the
right
from
Campbell.
So
that's
what
I
meant
by
this,
that
we
can
for
now
maintaining
both
the
Yambol
and
the
charts
as
part
of
the
bit
unico.
Unless
we
have
a
better
mechanism
for
having
an
interview
they
can
been
taking
at
our.
D
C
C
A
B
C
D
B
And
also
with
all
the
ham
part,
maybe
not
able
to
let
customer
to
upgrade
auto
downgrade,
and
it
has
no
need
to
need
to
get
different.
You
know
templates
and
then
to
do
the
upgrade
on
to
the
Deaton
great,
but
you
feel
healthy
ham
chart
and
then
the
Casper
can
just
can
be
related
to
Warren
and
then
do
the
thumb
grid
of
upgrade.
C
C
I
guess
maybe
these
are
two
different
things
and
I
shouldn't
be
trying
to
conflate
them.
The
one
issue
is,
we
want
to
be
able
to
have
canonical
way
of
installing
a
given
version
and
then
there's
another
sort
of
like
another
topic,
which
is
how
do
we
make
sure
that
you
can
upgrade
between
versions
I'm,
not
sure
we
should
consider
the
upgrades
until
we
have
just
the
ability
to
install
a
given
version?
E
Yes,
that
makes
sense,
let's
focus
on
getting
a
way
of
installing
the
latest
master
and
a
versioned
Federation
v2,
one
thing
about
helm,
I
think
and
I
think
they're
working
on
it
and
someone
might
be
able
to
inform
us
on
the
latest
status.
But
the
thing
I
don't
like
is
that
you
have
to
actually
install
tiller
and
you
have
to
get
a
deployment
for
helm
working
before
you
can
actually
install
Federation
v2.
E
C
C
My
suggestion
would
be
is
that
we
solve
the
versioning
for
just
the
install
EML
generated
by
queue
builder
and
if
I
mean,
if
we're
just
going
to
continue
on
the
home
chart,
that's
fine
and
we'll
have
kind
of
a
parallel
way
to
do
it.
But
in
my
mind
at
least
right
now,
we'd
be
ideally
having
install
llamó
as
one
way
to
do
it
and
helman
basically
repackage
that,
in
a
way
that
maybe
is
more
preferable
to
some
people
and
the
way
like
we
kind
of
kill
two
birds
with
one
stone.
A
C
All
I
said
I
mean
basically
like
if
we
get
if
we
have
install
well,
for
you
know
installing
head
and
also
every
version
that
we
want
people
to
care
about,
and
then
we
have
helm
charts
that
consume
that
install
oh
and
provide
it
in
a
way
that
some
people
might
find
preferable,
but
effectively
we
would.
We
would
get
the
ability
to
have
sort
of
a
straight
forward
path
without
helm
and
helm
would
kind
of
just
inherent
inherit
the
work
we
do
for
the
install
oh
yeah.
A
So
that's
what
I
was
trying
to
derive
it,
so
the
choice
of
helm.
For
me,
it's
sort
of
a
little
important,
because
it's
a
super
popular
tool
or
a
it
has
a
path
for
users,
so
that
alone
gives
me
a
reason
to
probably
have
a
helmet
art
installation
for
Federation,
but
need
not
be
necessarily
right
now
or
that's
not.
That's
super
important
or
necessary
to
be
done.
As
of
now.
It's.
A
So
what
what
my
my
solution
would
be
so
right
now
we
actually
started
off
this
discussion
because
of
issue
that
we
face
relating
the
mapping
of
the
correct
image
to
what
we
have
in
stone
latest,
which
was
probably
maimed
wrong
itself
in
stone
latest,
was
actually
installing
a
given
version
or
the
last
release
version.
So
this
naming
mismatch
is
one
solution
for
sure
we
have
to
solve
that.
That's
an
immediate
solution.
A
We
also
have
to
have
a
maybe
a
path
forward
like
how
exactly
are
we
going
to
map
the
release
versions
to
the
CI
CI,
always
test
the
latest
o
CI
test
the
release
version
also
or
verifies
the
release
version
also
and
verifies
the
master.
Also
something
like
that
and
we
post
that
we
have
some
mechanism
where
we
can
probably
have
Auto
generation
of
a
chart
from
the
yamen
that
is
generated
by
the
cube
builder.
Maybe
a
script
can
do
that
I
mean
I'm,
not
sure.
A
Maybe
some
people
are
already
are
coming
up
with
a
rule
like
that.
Does
that
make
sense.
A
What
I
was
saying
is
there
are
two
different
things
we
are
talking
here.
So
one
is
the
solution
to
a
little
bit
of
problem
that
we
have
with
the
youngins
and
the
mapping
of
the
image
to
that.
That
is
one
different
problem
and
should
be
solved
right.
The
other
is
having
a
help,
chart
installation
also
for
Federation.
We
do,
and
the
issue
that
we
see
with
it
right
now
is
that
we
might
have
to
hand
curate
the
charts
for
each
update
that
we
do
in
v2.
C
A
C
That
is
as
to
why
we
can't
just
have
a
big
collection
of
things
and
helmet.
Just
install
is
not
an
option.
Yeah
I
mean
how
long
consists
of
a
tiller,
and
it
applies
configuration
optionally
templatized.
But
in
our
case,
if
we
just
have
like
the
big
collection
of
configuration,
couldn't
adjust
speed,
install
you
mo
to
the
helm
chart
and
have
it
work.
A
D
I
want
to
point
out
I
think
there
are
like
a
few
different
logical
things
that
are
in
the
installation,
die
yeah
Mille
and
those
different
things
are
going
to
change
at
different
paces.
So,
for
example,
there's
the
there's
the
resources
necessary
to
deploy
like
the
Federation
controller,
and
then
there
are
the
CR
DS
for
the
like
core
Federation
resources
like
federated
type,
config,
federated
cluster,
propagated
version,
and
then
there
are
CR
DS
for
federated
types
like
the
federated
config
map,
etc.
D
And
then
there
are
federated
type
configs,
like
the
actual,
like
federated
type,
config
resources
that
we
need
to
create
to
configure
Federation
out
of
the
box
for
the
kubernetes
api,
and
those
are
all
going
to
change
at
different
paces.
Like
I'd,
expect,
probably
that
the
thing
that
will
change
the
least
often
is
are
the
resources
to
run
the
federation
controller
and
that
the
things
that'll
change,
probably
more
often
are
like
the
the
the
shape
of
CR,
DS
or
instances
of
those
CR
DS.
D
So,
for
example
like
if,
if
we
think
things
are
gonna
change
like
if
there's
a
group
of
resources
that
we
think
is
the
the
group-
that's
gonna
change
all
the
time.
That's
probably
the
most
important
for
like
an
automated
generation
path
to
keep
up
to
date
and
for
things
that
aren't
gonna
change
all
the
time.
D
D
Can
we
really
fit
this
all
into
a
single
home
chart
because,
in
order
to
create
custom
resources,
for
example
like
in
order
to
create
a
federated
type
config,
the
federated
type,
config
CRD
has
to
already
been
in
be
installed,
and
so
I
guess
one
one
question
is
how
how
are
you
handling
that
in
the
chart
currently
I
haven't
had
a
chance
to
look
at
it?
Yo.
B
B
C
A
C
D
C
C
Those
are
currently
manually,
curated,
I,
think
I
mean
unless
we
just
say
we're
only
gonna
start
like
to
use
how
I
think
would
make
sense
to
have
those
in
a
place
that
can
be
consumed,
though
manually
and
through
him
I.
Don't
think
those
are
really.
Customers
are
customized,
they're,
more
or
less
static,
God.
C
D
B
D
Think
if
we
could
have
Q
builder,
generate
the
c
rd
definitions
in
a
single
yamo
file,
which
I
believe
that
you
can
do
I
think
I
actually
added
that
feature
to
Q
builder.
If
I
remember
correctly,
oh,
you
should
be
able
to
get
Q
builder
to
put
all
the
custom
resource
definitions
into
a
single
file
and
I
believe
that
you
can
tell
Q
builder,
where
you
want
that
file
to
live.
A
C
D
B
C
B
D
E
One
of
the
ways
queue
builder
might
be
recommending
to
make
alterations
to
their
generation
is
by
using
customized
and
having
like
a
template
version
of
what
fields
you
want
to
customize
that
it
can
apply
after
the
fact,
I
know
that
Cuba
are
starting
to
use
customize
for
certain
parameters,
so
we
may
want
to
look
at
that
and
I
know.
Maru
and
I
also
discussed,
potentially
using
customize
as
a
way
of
providing
different
parameters
and
not
having
to
use
these
sort
of
set
alterations
around
yeah.
D
C
All
this
sounds
fine
to
me.
I
mean
at
this
point.
I
guess
home
is
the
de-facto
at
some
point.
Do
we
imagine
cutting
over
to
customize?
Doing
that
I
mean
the
main
advantage
of
helm.
In
my
mind,
is
that
it
allows
customization
I
get
facilitates
it.
The
other
templatized
temporizing
whatever
but
I
mean
to
me.
Is
there
other?
Are
there
other
values
to
home
other
than
that,
because
customize
can
provide
the
same
traumatization
region,
yeah.
D
That
I
have
found
to
be
extremely
convenient
about
helm
in
the
past
and
I'm.
Not
sure
if
customized
allows
us
or
has
a
similar
feature
is
that
helm
has
a
way
for
you
in
a
chart
to
generate
a
certificate
so
that
you
can
serve
TLS
I'm,
something
that
you
install
in
a
chart
which
is
very
helpful
since,
since
there
is
no
official
way
to
have
a
serving
certificate
generated
for
you
in
Cube.
Yet
it's.
D
So
in
in
openshift
we
have
a
feature
called
service
serving
certificates
that
you
can
annotate
a
kubernetes
service
with
something
that
basically
says.
I
want
to
have
a
service
serving
certificate
created
for
this
thing
and
put
into
this
secret
in
this
namespace,
and
then
you
can
consume
it
in
a
pod.
In
that
namespace
and
there's
a
there
is
an
operator
for
service
signing
certificates
that
at
one
point
we
were
very
interested
in
pushing
the
dysfunctionality
upstream
for
the
service
signing
certificates.
D
C
D
B
D
B
D
A
A
D
C
Mean
I'm
just
I'm
worried
about
this,
getting
confusing
we're
already
having
problems
with
install
Amal.
The
way
it
just
seems
like
it's
not
gonna,
make
it
easier.
I
just
want
some
clarity
on
like
I
mean
to
me
the
baseline,
for
this
isn't
do
we
have
a
helm,
sure
it's
water
strata?
What's
our
strategy
for
allowing
people
to
install
stuff?
As,
like
the
agenda
item,
we
led
with
I
think
a.
A
I
was
saying:
this
is
because
I
heard
somewhere
that
using
a
chart
might
make
a
little
make
life
will
be
more
complicated
for
developers
and
like
having
one
more
dependency
for
that
so
now,
and
chart
makes
a
lot
of
sense
for
users
like
rather
than
having
to
install
the
way
it
is
installed
right
now,
if
there
is
an
official
chart
for
say
a
particular
release
of
vacation,
it's
well
convene
thing
for
users,
so
is
it
okay?
If
we
keep
these
two
separate
or
they
should
match
this
I?
B
A
A
C
To
me,
the
the
challenge
is
that
there's
there's
two
sort
of
use
cases:
one
is
CI
and
developers,
and
one
is
users
and
how
chart
maintained
and
released
periodically
is
totally
fine
for
users,
but
I
mean
the
value
of
installing
ml.
Today
is
that's
what
we
use
the
test
and
CI.
We
actually
generate
the
ammo
based
on
the
state
of
the
tree.
We
use
that
employee
Federation
to
a
mini,
Q
cluster,
and
then
we
run
into
each
other,
so
I'm.
A
C
So
I
guess
I
guess
the
suggestion
is
then
like
we
can
continue
using
Q
builder
until
we're
happy
with
helm
and
then
maybe
we
could
even
like
for
the
configuration
and
missed
all
you
know.
Eventually
we
could
probably
just
cut
over
to
helm
and
the
near-term
we'll
just
keep
generating
and
probably
what
we're
gonna
be
replacing
is
install
latest,
so
install
latest
can
maybe
go
away
and
we're
no
longer
committing
install
Amal
to
the
tree
in
favor
of
a
help
chart
once
that's
merged.
Does
that
make
sense?
Yes,.
C
B
A
B
A
A
E
C
Still
not
entirely
clear,
though,
on
how
we're
gonna
be
like
if
I,
if
I
want
to
support
if
I
wanna
find
install
0:02,
there's
gonna
be
home,
chart
that
we're
gonna
release
or
publish
or
something
that's
gonna,
presumably
allow
for
that.
Where
is
it?
It
simply
gonna
be
a
matter
of
when
you're
using
the
Helen
chart
you're
supplying
a
version,
and
that
version
will
pick
up
the
appropriate
image
like.
What's
what's
the
strategy
for
relating
the
chart
to
the
version
to
the
version
image
so.
C
Think
I
mean
you're
still
fuzzy
on
this,
and
maybe
maybe
this
is
a
song.
I
should
stop
wasting
everyone's
time
and
we
can
be
offline,
but
there's
still
not
clarity
in
my
mind
of
how
the
chart
and
our
repo
is
gonna
be
related
to
because
the
chart
in
the
repo
is
going
to
basically
be
updated
periodically.
It's
gonna
bruise
on
the
future
release
date.
Does
my
confusion,
making
sense
to
people
yeah.
C
A
D
A
A
A
D
B
A
Saying
is
that,
like
Mara
mentioned
that
we
can
create
a
milestone,
and
then
we
can
tag
items
to
this,
but
the
chart
for
this
particular
release
would
not
be
part
of
this
milestone.
Oh,
would
it
be
like
the
chart
has
to
be
created
once
the
release
is
cut
out
right,
yeah
and
I
made
this
portion
well.
C
We
have
to
update,
we
have
to
put
the
chart
into
the
tree
and
make
it
current
and
maintain
it
manually
and
then,
at
the
time
we
cut
a
release,
we're
essentially
transferring
that
chart.
Referencing
a
published
image
into
some
char
repo
I
think
creating
the
chart
is
part
of
the
milestone.
Publishing
the
chart
is
kind
of
like
post
milestone,
Joe.
B
D
Now
I
I
did
some
follow
up
on
that
and
at
one
point
we
released
to
a
chart
repository.
What
what
happens
now
is
that
we
released
like
a
Google
storage
bucket
so
for
for
each
version
of
the
chart.
We
like
make
a
new
directory
in
this
Google
Storage
bucket
and
put
the
chart
there,
but
that
does
not
mean
that
I
necessarily
think
we
want
to
do
that
for
for
Federation
and
and
I'm
very
interested
in
seeing
if
you
can
get
a
federation
v2
chart
releasable
into
the
main
chart
repository.
B
B
B
B
D
A
A
A
B
B
Mauro
for
the
van
so
here
I
just
want
to
focus
on
some
questions
so
yeah.
So
the
first
question
that
I
think
that
we
agree
that
we
should
create
a
new
controller
to
to
do
the
generation
configuration
and
how
questions
of
the
question
is.
Do
it?
Do
we
need
to
do
this
product
Leon
separately,
Apple,
or
we
can
also
do
it
in
the
current
revolution?
We
tweaked
it.
D
B
D
D
It
should
maybe
go
in
a
different
group
and
then,
when
someone
wants
to
add
a
new
like
reconciler
or
propagation
mechanism,
the
the
control
surface
for
that
mechanism
will
likely
be
different
than
at
least
in
some
respects.
Then
what
we
have
for
the
sync
controller,
so
those
could
go
into
their
own
distinct,
API
group
at
the
point,
when
someone
wanted
to
add
them,
I
wonder
what
people
think
about
this
and
I
had
planned
to
write
an
issue
up
about
it.
B
B
There
are
more
templates
that
we
want
to
create
in
the
memberid
clusters,
so
here
I
just
want
reliability
in
controller
to
generate
the
placement
policy,
and
also
to
generate
e
mo
templates
that
we
want
to
create
in
member
cluster
that
we
wanted
to
push
into
it
travel
so
so
here
I
just
want
reliability
in
controller,
because
we
already
have
some
nodded
here.
Yeah
controller.
C
Of
the
underlying
clusters,
for
example
in
the
document-
and
it
has
to
have
a
bunch
of
informers
related
to
that
and
so
like
using
the
sink
controller,
would
just
be
like
muddying
up.
What
is
a
very
simple
configuration
exercise?
You
look
at
the
template,
the
overrides
the
placement
you
generate
yamo
into
the
appropriate
directory.
You
know
the
path
from
the
file
system,
the
corresponding
that
cluster.
C
That
is
a
very
disjoint
Li.
The
only
thing
in
common
is
how
you
propose
the
resource,
but
that
really
is
a
trivial
part
of
the
same
controller.
You
can
lift
that
out
and
make
it
reusable
or
just
copy
directly
with
far
less
effort
than
you
would.
If
you
try
to
take
the
same
controller
and
repurpose,
it.
A
B
A
B
Our
checks
is
passed,
yeah,
okay,
so
and
also
for
the
repo
and
I
think
that
we
we
have
the
agreement
that
we
can
just
talk
the
current
session
with
Rocco
or
we
can
just
create
another
repo
unto
deep
Pratap
there
and
a
34
tap
was
finished,
and
then
we
will
block
Rebecca
and
then
we
can
decide
if
we
want
to
have
this
source
coding
producing
material
or
we
should
have
a
separate.
Both
yeah
make
sense.
Yes,.
A
B
B
C
E
C
D
I
think
it's
easy,
I
think
it's
easy
to
to
get
into
a
situation
where
you
generalize
prematurely
so
I
would
also
I
would
also
treat
like
this
exercise.
Right
now
is
forget
about
any
other
like
pull
reconciliation,
implementation
and,
let's
I
suggest
to
focus
just
on
Cuba
plier
and
at
the
point
where
you
know
there
is
some
other
method
that
someone
wants
to
look
at,
that.
That
would
be
a
better
time
to
consider
what
the
interface
for
an
extension
would
look
like.
Yeah.
B
B
And
the
last
question
that
I
also
got
coming
from
Mauro
and
I
think
that
we
be
on
same
page
and
because
Canada
I
wanted
to
use
the
configuration
condition
controller
to
generate
the
results.
Yet
young
templates
and
also
want
to
generate
a
prison
policy.
And
currently
the
placement
policy
is
just
the
the
classroom
named
where
I
wanted
to
post,
where
I
want
to
create
the
resource.
So
here
are
just
the
pinky
answer,
but.
C
C
C
A
B
Here
you
can
get
at
the
placement
will
only
include
which
classes
that
I
want
Cree
the
resources,
so
so
in
here.
I
just
want
to
crazy
namespace
in
both
class
one
and
two
so
I
will
post
the
young
temp
is
in
two
directories
and
don't
remember:
cursor
can
pull
the
templates
from
github,
ons
and
credit
in
in
the
memory
custom.