►
From YouTube: Kubernetes SIG Multicluster 2020 Mar 24
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).
B
C
A
D
A
D
A
E
So
basically
I
added
custom
metrics
from
q3
for
those
operations
we
identify
the
could
be
given
as
pointers
about
where
scalability
motivates
my
my
D,
and
also
just
for
for
having
a
future
Q
fed
base
for
financeable,
so
yeah
I
agree.
Tyranny
shall
be
out
with
those
and
also
I
touching
the
same
tiara
cap
with
the
intention
of
the
matrix.
So
if
anyone
can
have
a
visual
review
to
the
piano
sure.
A
So
I'm
giving
it
a
once-over,
it
seems
pretty
reasonable
to
me
so
far.
I
know
you
had
a
couple
of
things
that
you
wanted
to
clean
up
about
it,
so
I'll
definitely
give
you
a
review
later
on,
but
you
know
if
anybody
would
like
to
has
experience
with
metrics
and
wants
to
take
a
look
I'm
sure
that
would
be
appreciated.
Thank
you.
A
I
think
we
should
decide
how
we
want
to
record
things
that
I
view
I
would
maybe
call
pre
API
principles
that,
like
aren't
part
of
any
like
code,
API
so
far,
but
they're
they're
fundamental
to
to
capture
explicitly
so
that,
like
we
have
a
explicit
list
of
principles
that
we
reason
from
so
I
wondered
if
folks
had
an
opinion
on
what
the
best
way
to
record.
That
is
not
on
the
their
opinion
on
the
question
itself,
which
I'm
very
happy
to
talk
about,
but
like
initially,
let's
just
talk
about.
C
Yeah
I
mean
github
seems
to
make
sense.
Maybe
github
pages
is
I,
don't
know
if
there's
a
lot
of
precedents
for
this
ton
of
non-clumping
API
statement.
A
F
A
F
F
C
F
A
There
are
like
sig
folders
under
community
yeah
at
one
point,
the
markdowns
that
lived
in
those
things
were
like
people
wrote
them
by
hand,
I
think
they're
generated
now.
I
think
there
might
be
a
place
under
the
SIG's
damo
file
in
the
root
of
that
repo,
that
we
could
at
least
put
a
link
in
there
to
something.
A
One
idea
that
I
had,
which
is
like
extremely
cheap
and
like
doesn't
doesn't
necessarily
put
the
cart
before
the
horse
in
terms
of
officiality
would
be.
What
if
we
make
like
we've
got
this
like
very
well-known
Google
Doc
Tim
that
you
wrote
up
about
your
musings.
Could
we
write
a
Google
Doc
that
was
like
record
of
pre
API
principles.
F
We
could
I,
don't
like
I
mean
I
love,
Google
Docs,
don't
get
me
wrong,
I,
don't
like
the
lack
of
history
and
tribute
ability
I'm
looking
at
community
repo
again
so
there's
like
under
community
contributors.
There's
contributors,
there's
design
proposals
which
is
mostly
defunct
and
replaced
with
caps,
but
under
contributors
is
also
devel,
like
docks
for
how
to
contribute
to
the
project,
not
exactly
what
we're
thinking
about
here,
either
yeah.
A
What
I
liken
the
idea
of
like
putting
a
mark
down
under
community
/,
sig
multi
cluster?
More
and
more,
it
gives
us
the
attribute,
attribute
ability.
It
gives
us
the
PR
process
and
like
review
and
just
normal
way
of
working
with
things,
and
if
there's
a
point
in
time,
when
there's
an
API
that
reflects
these
principles,
it
can
become
part
of
the
kept
that
defines
that
API
and.
F
F
A
What
about
the
actual
evaluation
of
that
principle,
like
there's,
been
a
really
healthy
discussion
on
the
mailing
list
since
I
have
the
bully
pulpit?
I'll
start
with
my
own
opinion,
which
is
that
I
think
personally,
the
principle
of
like
namespace
sameness
across
clusters
within
at
least
a
flotilla
or
a
group
of
clusters
is,
is
very
reasonable.
I
think
it
will
capture.
A
F
F
F
Yeah
just
hand
waving
that
there
exists
a
group
and
you
have
to
decide
for
yourself
what
the
scope
of
that
group
is.
When
you
look
at
use,
cases
like
what
Sandor
was
proposing,
I
felt
like
I,
didn't
get
to
respond
to
his
last
messages,
but
I
feel
like
what
he's
proposing
is
actually
a
lot
of
groups,
not
a
single
group
yeah.
C
That
was
my
I
actually
did
respond
this
morning,
and
that
was
coming.
That
was
my
takeaway
as
yeah.
It's
it's
a
lot
of
separate
groups
which
is
totally
fine.
You
could
you
could
even
decide
that
you
wanted
your
group
to
be
individual
clusters.
You
know,
there's
nothing
stopping
right.
It
just
means
you
don't
take
advantage
of
whatever
we
decide
to
do
with
this
sameness
or
take
advantage
of
all.
A
F
The
only
sorry
to
argue
with
ourselves,
the
only
question
I
have
is
for
those
people
who
have
already
implemented
something
that
isn't
that
what
does
it
mean
like
I,
don't
intend
to
invalidate
other
people's
work?
Obviously
we
don't
have
that
authority.
Does
it?
Does
it
mean
that,
like
they,
people
have
to
abide
by
this
in
order
to
build
sig
approved
features,
or
does
it
mean
that
this
is
a
sort
of
something
you
keep
in
your
mind,
but
it's
really
more
of
a
suggestion.
A
A
In
the
same
way
that,
like
you
know,
we
we
at
one
point
had
like
a
template,
feature
accepted
into
kubernetes.
There
wasn't
one
dove
being
implemented
in
Cube,
but
it
doesn't.
The
presence
of
that
thing
does
not
invalidate
help
right
and
like
the
the
community
is
so
large,
and
people
have
just
such
a
wide
variety
and
volume
of
use.
Cases
that,
like
I,
don't
personally
want
to
be
in
the
position
of
saying
something
that
you
went
on
your
own
and
did
is
invalid,
because
it's
your
choice
right.
F
F
F
A
A
A
C
A
A
A
So
I
think
lazy.
Consensus
is
like
important
as
a
value
in
our
community.
I,
don't
want
to
go
making
a
decision
in
a
sig,
kneading
and
I
know
that
I
have
yet
to
like
respond
to
this
thread,
but
I
wonder
like
what
the
right
way
to
drive
this
down
is
like
I'm
I
am
NOT
hearing
a
strong
like
objection
to
accepting
that
principle
in
this
meeting
and
I'm,
not
necessarily
even
like
reading
it
when
I
skim
through
the
thread.
It's
just
that
some
people
have
already
made
a
choice.
That's
different
than.
F
So,
given
the
extenuating
circumstances
of
the
world
today,
procedurally
I
feel
like
we
should
give
this
breathing
room.
Would
it
make
sense
to
like
Jeremy
I'll,
throw
it
back
on
you
craft
this
into
a
MD
file
against
community
and
leave
it
open
for
some
defined
period
of
time
to
try
to
accumulate
any
other
naysayers,
feedback,
wordsmithing,
etc.
F
D
A
A
C
Do
it,
and
just
basically
that
Doc's
been
around
for
a
while?
It's
got
a
lot
of
attention
from
from
this
sake,
sig
Network
and
some
people
from
sto
of
Commons
as
well.
There's
still
some
open
questions
and
things
you
need
to
figure
out
but
I.
If
there's
no
objections,
I'd
like
to
make
it
a
little
more
real
and
I
guess
trying
to
figure
out
what
the
best
way
to
do,
that
would
be.
Would
it
be
a
cap?
Would
it
be
a
big
project?
I
think
that
ok,
yeah.
A
F
C
D
D
F
So
it
may
be
sorry
to
sort
of
break
the
human
connection.
Connection
I
can
throw
one
small
item
on
towards
the
agenda
topology,
so
in
the
face
of
multi
cluster
topology
becomes
important,
especially
if
we're
gonna
try
to
blend
services
across
clusters.
It
is
pretty
important
that,
if
we're
going
to
try
to
do
anything
automatic
that
that
automatic
considers
cross,
zonal
cross-continental
connections
as
maybe
less
ideal
than
local
connections,
this
is
not
news.
F
The
Signet
side
actually
had
a
kept
for
a
topology
API
and
that
API
actually
went
to
alpha
as
an
API.
It
makes
some
pretty
serious
trade-offs
and
it
forces
the
user
to
make
some
pretty
serious
trade-offs
and
I've
been
contemplating
that
for
a
while
and
I
wrote
a
note
to
sig
network
week
before
last
saying,
maybe
we
should
put
them
brakes
on
that
topology
API,
because
the
trade-offs
are
so
egregious.
F
Personally,
it's
something
I'd
really
like
to
explore.
I
just
one
of
those
engineering
projects,
I'd
really
like
to
roll
up
my
sleeves
and
get
into
pragmatically
I'm,
not
gonna
have
enough
time
to
really
drive
on
it.
If
anybody
else
here,
this
is
not
a
like
I.
Don't
know
anybody
to
commit
to
me
right
now,
but
if
anybody
is
interested
in
that
topic,
let
me
know
reach
out
to
me
and
I
would
love
to
just
brainstorm.
A
I
think
I
heard
Tim
is,
is
basically
I
evaluated
what
you
said
in
the
scope
of
like
multi
cluster
services,
and
so
what?
What
I
heard
in
just
in
terms
of
like
the
specifics
of
what
you
were
thinking
about
from
like
a
technology
standpoint,
is
that
in
in
a
cluster
that
consumes
a
multi
cluster
service
to
bias
toward
Network
links
that
are
like
in
the
highest
priority
cluster
level.
F
I
I
think
sorry,
jumping
from
SEO
perspective.
It
would
be
useful
to
have
some
policy
or
some
some
control
over
how
they
are,
how
to
better
know,
customer
feedback
on
multi
cluster
cool.
It's
clearly
something
we
want
also
in
the
Gateway
had
it
so
I,
don't
like
a
gateway
is
part
of
the
networking
sync
or
not,
but
it
has
similar
requirements
in
terms
of
apology
and
yeah.
I
More
thing
because
I
have
to
go
at
then
sorry,
don't
forget
multiple
networks,
because
that's
one
of
the
biggest
prongs
in
East
you
part
of
topology
or
whatever,
since
the
ipv4
is
so
restricted
uncommitted
oak
as
well
part
of
this.
We
need
to
include
some
information
about
how
to
reach
remote
cluster
if
it's
not
network.
Yes,.
F
I'm
a
little
bit
down
on
the
service,
topology
API
in
part
and
I'm
happy
to
provide
a
link.
If
anybody
wants
to
read
the
thing
I
wrote
I've
sent
on
the
cig
network
man
list,
but
I
think
that
would
be
great,
basically
I'll.
Do
that
the
the
real
problem
is
that
forcing
the
user
to
express
it
means
that
they
have
to
take
on
these
trade-offs
that
are
sort
of
egregious
trade-offs.
The
topology
API,
as
expressed,
was
really
soon
biased
towards
easy
implementation
and
not
easy
use
and
I.
F
So
I'm
coming
back
and
revisiting
sort
of
my
own
decision
there
and
I
think
I
was
wrong
and
I
think
we
need
to
explore
it
automatically
and
I
think
the
best
we
can
do
in
really
the
truest
sense
of
the
best.
The
best
thing
we
can
do
is
a
heuristic
that
helps
the
traffic
to
try
to
stay
local
but
doesn't
necessarily
guarantee
it
and
then,
if
that's
not
sufficient,
then
we
bust
out
the
API.
H
Yeah
yeah,
that
would
that
make
sense
to
me
years
off
and
in
particular
so
I
can
see.
It
would
be
very
nice
to
have
some
automatic
nice
defaults
and,
at
the
same
time,
we're
looking
at
use
cases
and
submariner
where
you
do
want
to
be
able
to
say
that
you
have
unusual
requirements
like,
for
instance,
there
are
scenarios
where
you
want
to
explicitly
have
direct
traffic
towards
a
cluster
in
your
flotilla.
That's
not
in
your
current
availability
zone
because
you
want
to
use
it
for
replication
and
disaster
recovery.
Things
like
that.
F
H
Well,
we'd
want
to
be
able
to
say
you
know
you
have
this
service
and
obviously
like
you're,
saying
it's
nicer
to
keep
traffic
local
or
get
your
results
quicker
and
so
on.
But
there
are
some
scenarios
where
we'd
want
to
ensure
that
some
specific
requests
go
to
this
particular
service,
but
on
another
cluster
and
Lopaka,
so
that
we're
easier.
A
That's
narrow,
hold
on
my
bias
in
general
is
two
word.
Simplicity
of
use
like
I
will
just
put
my
hand
up
as
somebody
that,
like
early
on
and
cube,
I
definitely
feel
like
I
was
doing
some
features
for
ease
of
implementation
versus
ease
of
use,
like
many
things,
involving
secrets
that
we,
like
eventually
relaxed
or
changed,
etc,
etc
and
yeah.
That's
I
learned
a
lot
from
those
things.