►
From YouTube: Kubernetes SIG Multicluster 2020 June 23
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
D
A
All
righty,
so
valerie
is
on
the
agenda
with
jojen
and
she
just
let
me
know
that
she'll
be
late.
So
why
don't
we
go
ahead
and
get
started?
Jeremy,
I
think
you're
on
the
agenda.
First
with
super
cluster.
B
Name
yeah,
so
I
shared
out
in
the
in
the
forum
a
sheet
that
a
whole
bunch
of
people
have
joined
in
on
with
with
suggestions
and,
quite
frankly,
I'll
share
it
now,
but
I'm
really
not
sure
the
best
way
to
proceed.
We
got
a
lot
of
suggestions.
B
Perfect
sharing
now.
B
Cool
yeah,
so
a
lot
of
suggestions
here
I
think
we
have
yeah
76
suggestions.
It
looks
like
for
awesome
for
cluster
names
and
you
know
if
anyone
wants
to
open
this
up,
it's
on
the
mailing
list,
it
might
be,
and
it's
on
the
agenda
might
be
easier
to
see
locally.
I
guess
how
should
we
proceed
to
kind
of
whittle
this
down
so.
D
Go
ahead
tim,
so
I
I
participated
in
a
sort
of
naming
exercise
fairly
recently,
and
I
thought
it
was
interesting
that
you
can
actually
approach
this
with
some
amount
of
rigor.
D
I
think
what
we
should
do
is
define
what
we
think
the
top
criteria
are
and
sort
of
like
there's,
probably
some
obvious
ones
that
we
can
just
discard
because
they're
silly
or
they
overlap
with
things
that
exist
already
or
they're
just
too
unrelated.
But
this
is
a
good
brainstorm.
I
mean
the
fact
that
there
was
very
little
filtering
at
the
front
right
then
I.
A
Of
identifying
goals
for
names
that
that
might
be
something
else
that
we
can
like
open
in
the
mailing
list,
but
it
I
wouldn't
also,
I
also
wouldn't
mind
talking
about
it
now
for
a
few
minutes.
D
Yeah
so
like,
let's,
you
know,
let's
throw
out
what
we
think
are
the
the
best
criteria
for
these
things,
and
then
we
can
sort
of
weed
out
some
obvious
ones
that
just
don't
meet
the
criteria
and-
and
I
think
we
can
very
quickly
reduce
this
list
of
70-
something
to
20
something
right
and
then
it's
much
more
manageable.
To
talk
about
the
the
merits.
B
A
Sorry,
let's
see
we
can
let
me
open
up
the
meeting
notes.
A
I
I'm
not
an
amazingly
reliable
scribe,
but
I'm.
A
And
I'm
on
my
phone
but
jeremy,
if
you
want
to
share
the
notes
on
the
screen
like
that
way,
people
can
intervene
in
my
most
hideous
note-taking
errors
as
they
happen.
B
Yeah
yeah
hold
on.
Let
me
bring
it
up.
A
D
A
Understand
the
overlap
yeah,
so
I
I
wouldn't
want
to
to
use
it.
So
federation
fits
actually
both
those
things.
It's
both
an
overlap
with
previous
work,
and
it's
also
something
that
is
like
widely
used
in
a
generic
sense,
but
it
has
so
many
different
meanings.
Meanings
that
it's
like
frequently
feels
to
me
like
it's
counterproductive
to
even
say
the
word
right
like
federated
prometheus,
federated,
you
name
it
like
all
means
different
stuff
right
or
another.
One
I
saw
on
that
list
was
set,
which
is.
D
B
Scan
through
the
list,
it
would
probably
be
nice
if,
and
I
think,
like
I
opened
up-
that
the
the
multi-cluster
zone
could
be
different,
but
it
probably
makes
sense
that
they're
not.
D
Yeah
well,
actually,
we
have
two.
We
have
a
separate
concern,
which
is:
should
we
use
local
local
is
actually
wrong
in
pretty
much
every
sense,
but
we
don't
need
to
conflate
that
with
choosing
the
the
domain.
So,
let's
like
pick
the
term
in
the
domain,
and
then
we
can
argue
about
whether
it's
dot,
local
or
dot,
xx
or
dot
lan
or
something
more
appropriate.
D
So
without
trying
not
to
completely
derail
there
are
a
bunch
of
reserved
tlds
that
we
can
use,
including
a
whole
bunch
of
two-letter
like
reserved
country
codes,
which
may
be
the
best
option,
but
that's
actually
sort
of
a
bigger
discussion,
because
I'd
like
to
include
the
rest
of
sig
network.
If
we're
going
to
talk
about
that,
yeah
like
if
there's
a
v2
dns
for
for
kubernetes
overall,
it
would
probably
go
in
the
same
direction.
So
we
can.
D
D
I'm
tempted
to
offer
not
nautical,
but
that
cuts
out
a
whole
lot
of
decent
options.
D
F
D
Maybe
we
can,
I
don't
know,
encourage
people
to
add
columns
or
comment
on
their
scoring
or
something
like
that.
D
Yep,
okay,
you
want
to
run
through
them
and
and
just
read
them
off
and
then
we
can
spend.
You
know
20
seconds
on
each
one.
B
D
B
B
Okay,
there
we
go.
This
is
perfect
meta
cluster.
B
Yep
seems
to
I
think
that
stays
in
the
running
okay,
cluster
group.
G
Mr
group
is
that
you
may
ultimately
want
to
think
about
collecting
a
set
of
clusters
together,
and
so,
if
cluster
group,
ultimately
it
becomes
a
way
to
deal
with
like
access
control
like
I
want
to
group
together
clusters,
one
through
five.
Do
I
want
to
think
about
that
as
sort
of
one
subdomain
or
kind
of
a
broader
subdomain,
like,
I
think,
there's
other
things
where
cluster
might
come
up
in
the
space.
A
Yeah,
I
think
I
think
I
think
the
name
we're
trying
to
pick
is
the
widest
scoped
collection
of
clusters
that
you
might
deal
with
at
once.
Does
everybody
agree
or
do
what
name
are
we
actually
picking.
B
I
think
that's
right,
yeah
and
then
so
with
cluster
group.
Maybe
this
is
another
criteria,
but
well
I
I
guess
this
fits
not
over
overly
generic,
like
maybe
cluster
group,
that's
just
how
I
would
describe
like
that's
just
plain
english
for
multiple
clusters.
I
don't
know
that
yeah.
I
could
see
that
coming
up
in
other
areas
and
and
being
confusing.
D
Yeah,
I
don't,
I
think
it's
not
obviously
wrong,
though,
and
yeah.
F
D
Reason
is
like
it
is
sort
of
a
natural
term
that
a
lot
of
people
come
to.
I
think
it'll
score
poorly
on
the
overly
generic
dimension,
but
I
would
suggest
we
don't
mix
it
yet.
E
I
have
a
meta
question:
yeah,
are
we
assuming
that
clusters
can
belong
to
more
than
one
group
or
or
whatever?
This
will
be.
B
E
The
choice
of
name
some
names
might
might
imply
that
kind
of
many-to-many
association.
D
D
I
B
All
right,
nix
it
set,
would
be
out
k,
set
getting
colder
yeah.
A
Crew,
isn't
that
the
name
of
another
thing
already
is
it
I
don't
know,
isn't
it.
B
D
B
Yeah,
I
I'm
not
against
clump
category
seems
too
vague,
yeah
and
same
with
catalog
catalog
kind
of.
E
B
E
D
I
think
clan,
for
obvious
reasons
is
just
not
gonna
fly
yeah
yeah
deep
six.
Nobody
even
suggests
the
alternate
spelling
yeah.
B
B
D
B
A
B
Yeah,
I
guess
I'm
stuck
on
the
cluster
super
cluster.
That's
you
know
where
that
came
from.
So
I
don't
know
if
that's
here.
H
D
That's
a
good
question:
yeah,
that's
a
great
question.
Actually
that
was
part
of
the
the
naming
exercise
that
I
did
with
what
they
called
canonical
sentences,
and
you
come
up
with
a
few
phrases
and
where
you,
you
would
use
the
word
and
try
to
drop
it
in,
and
you
see
how
it
feels.
D
A
D
D
B
Yeah
and
I
think
zluster,
I
don't
even
know
how
you
say
that
getting
colder
yeah.
D
B
B
B
D
D
C
C
C
D
C
C
D
B
A
B
G
D
B
A
D
So
I
think
mega
cluster
failed
and
so
I'm
going
to
mix
it.
I
think
it
was
actually
mine.
Macro
cluster
doesn't
roll
off
the
tongue,
but
I
I
to
me
it
would
be
an
acceptable
I
think
multi-cluster.
I
I
actually
kind
of
like
but
macro
cluster
I'd
be
fine
to
nix.
B
We
can
skip
the
only
the
main
issue
I
have
with
multi-cluster
is
like.
If
we
look
at
this
sig,
it
might
start
to
imply
that
we
only
care
about
like
everything
that
we
ever
talk
about,
is
only
about
this
one
like
it.
It's
only
about
the
now
multi-cluster,
which
we've
turned
into
a
specific
concept,
and
maybe.
A
D
D
G
A
D
Maybe
we
need
a
criteria
for
not
a
negative
connotation.
A
B
D
I
I
kind
of
object
to
it
on
the
military
connotation
yeah.
D
I
did
not
put
in
here
flotilla,
which
I
would
strongly
object
to
on
the
silliness
dimension.
J
A
A
C
B
Can
leave
it
multi?
That's
too
generic,
I
think
constellation.
B
Also,
it
requires
the
suffix
to
be
acceptable.
Right,
consolation
doesn't
seem
bad.
C
A
B
A
B
You
do
class
is
too
big,
yeah
cohort
kind
of
fits
yeah.
It.
B
B
B
A
B
D
B
D
D
As
a
next
step,
maybe
jeremy
d,
do
you
want
to
move
all
of
the
removed
ones
down
to
the
bottom?
I'm
not
sure.
If
you
can
sort
by
that
and
then
we
can
add
a
column
and
sort
and
whatever
and
then
take
it
to
the
mailing
list
with
the
criteria
and
ask
people
to
sort
of
pick
their
like
top
five
and
score
them
on
those
four
dimensions.
B
Should
we
do,
should
we
just
do
it
on
the
mailing
list,
like
that,
or
should
we
do
like
a
poll.
D
If
you
want
to
set
up
a
something
more
rigorous
than
a
mailing
list,
then
that's
fine.
You
have
more
free
time
than
I
do.
F
F
D
Criteria
or
on
like
as
a
single
single
dimension,.
D
I
think
different,
some,
like
looking
at
the
criteria
that
we
called
out
in
the
in
the
beginning,
descriptive
shorter,
is
better
no
overlap
with
related
prior
work
and
not
overly
generic,
and
actually
I
was
going
to
add
the
not
negative
connotation.
So
we've
got
five
dimensions.
I
think
some
names
will
score
high
in
one
and
low
in
another.
C
A
D
Right
thanks
and
I
need
to
bounce
out
so
sorry
yeah,
let's
take
to
the
mailing
list,
but
good
filtering
thanks,
guys.
A
All
right
trojan
and
valerie
you're
up
next
with
work,
api.
H
Okay,
shojin
might
be
better
in
a
position
to
give
the
update.
Since
he's
worked
on
a
lot
of
the
api
stuff.
A
Yeah,
so
we
were
talking
about
maybe
talking
through
the
changes
that
are
in
that
pr
to
track
applied
resources.
I
A
I
I
So
this
is
basically
the
apis
that
we
are
currently
working
on
the
values
repo
on
the.
So
there
will
be
two
part
like
in
the
spac
in
the
works
back,
we
define
a
workload
field
and
in
the
workload
field
we
define
a
set
of
the
manifest,
which
is
basically
the
runtime
raw
extensions
as
each
items
as
a
manifest.
I
So
this
is
very
simple
part
that
in
the
spac
you
just
define
a
set
of
the
manifest
to
deploy
it
on
the
small
cluster
and,
on
the
other
part,
is
about
the
work
status.
So
we
currently
have
two
conditions.
One
is
the
conditions
for
this
work.
Api
and
another
conditions
is
the
list
conditions
that
for
each
manifests,
that
is
being
deployed,
so
the
manifest
conditions
will
currently
have
two
part.
I
One
is
identifier
that
is
to
identify
which
resources
by
using
the
gvr
and
the
namespace
and
name
to
identify
what
conditions
for
for
certain
results,
and
then
we
will
have
another
list
of
the
conditions
to
say:
what's
the
current
conditions
for
this
certain
resources
on
the
small
cluster,
so
currently,
we
think
that
there
might
be
several
conditions
that
is
valid
to
to
say
about
the
work
such
as
applied.
It
means
that
we
have
successfully
applied
the
things
on
the
spoke
cluster,
something
like
we
can
say.
I
Progressing
states
is
like
we
are
transition
from
one
state
to
another
state.
All
available
state
means
this
kind
of
the
manifest
resources
is
being
created
and
already
be
running
on
the
small
cluster,
and
maybe
we
will
have
a
degraded
status.
That
means
like,
for
example,
if
we
have
a
deployment
deployed
on
the
on
the
small
cluster,
we
were
able
to
check
whether
the
the
replica,
the
design
replica
matches
the
current
replicas,
and
if
it
is
not,
it
seems
to
be
liking,
degraded
status.
I
B
I'm
just
noticing
on
conditions.
I
wonder
if
it
makes
more
sense
to
make
that
more
of
like
a
status
like
not
not
individual
conditions
but
like
a
status
enum
or
something
because
it
seems
like
those
are
all
mutually
exclusive
states,
or
at
least
you
know,
progressing
available
degraded
are
applied,
I
guess,
could
fit.
A
I'm
wary
of
hooking
into
the
great
conditions
debate
here
yeah,
but
is
it
sufficient
to
just
say
noted
in
about
whether
we
think
these
are
the
exact
right
conditions
to.
A
Okay,
okay,
maybe
we
can
keep
going.
I
Yep
and
yeah:
that's
basically
the
part
of
the
api,
so
I
think
the
another
pr
that
we're
still
working
on
is
about
so
the
control
loops
to
apply
these
resources
on
the
on
the
spoke
cluster.
So
let
me
just
go
to
the
the
control
part.
I
Oh
okay,
so
I
think
currently
they
there.
We
just
use
a
controller,
the
cool
builder,
to
to
to
build
this
controller,
so
we'll
be
like
have
a
reconcile
loop,
and
so
this
is
just
for
the
deletion
of
the
work
such
as
when
we
want.
When
we
delete
the
work
on
the
then
we
should
ensure
that
the
resources,
the
manifest
in
the
work
should
also
be
deleted
in
the
spoke
cluster,
and
then
there
is
another
part
like
when
we
have
give
me
my
minutes.
I
So
then,
then
we
have
a
way
to
apply
the
resources
on
the
on
the
small
cluster.
So
we
just
retrieve
all
the
manifests
in
the
in
the
mana
in
the
work
api
and
then
unmarket
as
unstructured
and
calling
the
something
like
the
a
dynamic
line
to
create
or
update
that
resources.
On
the
many
on
the
spoke
cluster,
and
after
that,
we
will
be
able
to
get
the
results.
I
Give
me
my
minutes,
so
we
will
be
able
to
get
the
result
of
the
reconcile.
I
mean
the
apply
the
result
of
the
resource
supply.
I
So
in
the
results
we
just
put
something
like
the
gvrs,
the
namespace
name,
and
then
we
can
know
that
this
is
it's
a
certain
status
conditions
for
certain
manifest,
and
then
we
based
on
the
apply
result,
whether
it
is
filled
or
successful.
Then
we
can
generate
the
certain
conditions
on
the
conditions
part
in
the
manifest
condition.
I
So
after
that,
we
will
just
merge
this
one
to
the
so.
This
is
just
to
generate
the
desired
conditions
with
in
this
reconcile
loop,
and
then
we
will
use
the
current
conditions
manifest
conditions
in
the
in
the
work
api
to
merge
with
this
one
and
generate
a
new
manifest
conditions
and
update
update
the
apis.
So
I
mean
yeah,
that's
the
basic
things
that
we
are
currently
doing.
I
I
think
the
things
that
are
missing
is
that
we
still
not
to
merge
the
current
one
with
the
desired
one
and
update
back
to
the
api.
So
that's
some
things
you
need
to
do
in
sphere.
H
I
don't
have
anything
too
major.
The
main
thing
that
I've
been
trying
to
think
about
is
how
to.
H
Essentially,
deal
with
reconcile
conflicts
because
we
have
a
lot
of
fields
that
might
be
mutable
as
part
of
like
internal
automation
like
say,
hpa,
editing,
replicas
or,
like
someone's
custom
or
custom
cluster
automation.
I've
noticed
like
in
a
lot
of
the
get
opt
type
stuff.
The
way
it
works
is
basically
it's
an
upfront
apply
once
on
source
change
and
then,
after
that,
it
doesn't
continue
to
reconcile.
H
A
Yeah,
I
don't,
I
think
I
think
that's
a
good
instinct
and
probably
is
durable,
at
least
for
a
couple
beats
of
exploratory
work.
B
H
Yeah
the
my
current
goal
so
far
is
to
make
a
specific
subset
of
resources,
work
first
and
then
start
thinking
about
bigger
picture,
bigger
picture.
That's
definitely
a
concern
is
trying
to
figure
out
how
to
properly
map
intent
here
and
there's.
Probably
some
things
that
I
can
learn
from
the
whole
server
side
apply
change
with
that.
B
A
Okay,
well,
we
are
at
time
for
now
so
folks,
if
you
have
things
you
want
to
put
onto
the
agenda,
the
document
should
be
world
writable
and
we'll
see
you
all
next
week.