►
From YouTube: Agones August 2020 Community Meeting
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
All
right
well
welcome
to
the
august
version
of
the
agonize
community
meeting
everyone
thanks
for
being
here,
mark's
dancing
for
us.
It
looks
like
on
the
agenda.
Mark.
You've
got
the
first
three
items,
something
about
upgrading
to
kubernetes
1.16.
I
think
that's
the
thing.
Do
you
want
to
talk
about
that.
B
I
got
that
first,
I
think
it's
basically
it
mostly
we've
started.
I
think
is
the
number
one
thing
so
the
antoine
cluster
is
now
running
1.16.
It
seems,
like
things,
are
passing
so
good.
B
I've
ticked
off
some
of
the
items
in
that
list
that
have
been
completed.
I
think
that's
up
to
date,
take
a
quick
look
here,
there's
still
a
bunch
of
stuff
still
to
do,
but
one
cool
thing
that
actually
came
out
of
that
is,
since
we
all
the
dev
tooling,
we
now
got
rid
of
deployment
manager,
which
is
super
cool.
Everything
is
running
all
the
same
terraform
scripts,
all
the
way
through
which
is
kind
of
nice.
B
So
we
get
a
lot
more
usage
of
terraform
and
like
if
we
find
bugs-
and
so
I
actually
found
a
bug
in
the
process
which
was
nice
and
fixed
it,
which
is
good,
so
it's
all
using
exactly
the
same
stuff.
So
actually
even
doing
this
upgrade
was
a
much
smaller
footprint.
I
could
essentially
change
the
gke
terraform
versioning
there
and
it
funneled
its
way
through
to
everything
else,
which
was
awesome,
so
yeah
people
want
to
help.
That
is
great.
I
think
a
lot
of
it
outside
of
client
go.
C
Yeah,
I'm
happy
to
help
drive
that
as
well.
I
wanted
to
I
put
some
notes
on
the
the
issue
when
we
opened
it
sort
of
around
the
a
previous
meeting
just
to
let
people
know
that
api
changes
like
more
bigger
ones,
started
happening
in
116.
and
so
they're.
C
I
think
the
only
api
that
there
was
only
like,
maybe
one
api,
that
affected
us-
that
got
dropped,
which
are
some
of
the
the
beta
apis
got
dropped,
but
there
are
a
bunch
of
apis
that
also
went
ga,
and
so
we
need
to
figure
out
when
we
start
moving
stuff
up
to
ga
apis,
because
those
apis,
the
beta
versions,
will
get
dropped
in
a
future
release
of
kubernetes.
C
And
you
know,
I
know
that
we
officially
only
support
one
version,
but
some
people
try
to
install
stuff
on
slightly
newer
or
slightly
older
versions,
and
we
should
be
cognizant
of
when
we
start
relying
or
removing
support
for
specific
api
versions.
So.
B
C
C
F
C
To
those
to
the
ga
version
that
will
mean
that
people
will
not
be
able
to
install
and
run
on
gones
on
versions
older
than
116
right,
because
the
v1
api
won't
be
served
on
the
version,
so
we
don't
necessarily
want
to
do
it
now,
but
but
this
is
when
it
was
part
of
the
kubernetes
release.
Note.
So
if
we
forget
about
it,
we're
not
going
to
remember
until
you
know
119
when
they
completely
drop
support
for
the
old
ones
right,
so
we
want.
C
C
I
just
want
to
kind
of
give
people
a
heads
up,
but
there's
a
little
bit
of
turn
there.
You
know
we
often
talk
about
like
supported
versions
that
we
support
aversion,
but
people
try
to
install
stuff
and
it
generally
works
on
other
versions,
and
this
is
a
case
where
we
know
that
things
would
break
on
other
versions.
E
Sense,
yeah
just
to
make
a
notice
about
this
bug,
which
is
related
to
helm,
plus
g
key.
If
we
update
a
recreated
cluster
gk
cluster,
we
currently,
there
is
no
way
in
terraform
to
update
else
to
helmet
style
of
the
new
release
or
reinstall
agonize.
So
probably
we
need
to
have
additional
step
just
to
make
sure
all
is
fine.
E
I'm
not
sure
I
understood
this
actually
well.
I
have
added
recently
some
comments
about
from
helm.
What
reform
issues
which
this
closely
relates
to
and
if
we
have
two
models,
one
dependent
on
another
traform
would
not
create
a
dependency
c
child
which
is
helm
after
we
create
a
parent
which
is
each
key.
So
it
would
not
recreate
the
child
after
this.
B
E
It
would
not,
you
know,
create
a
gonasia,
no
if
gk
cluster
would
be
recreated,
so
please
take
care
of
it.
I
mean
there
was
some
comment
about
this
white
reform.
Doesn't
do
this
in
multiple
model
structure
when
one
depends
on
another?
So
when
I
used
a
new
terraform
zero
points,
so
team
depend
on
for
modules.
It
also
doesn't
help
so.
E
Say
when
you
just
rename
the
cluster
from
test
cluster
to
test
cluster
number
two
yeah
and
say
you
have
test
cluster
plastic
on
us
on
it
and
when
you
have
testosterone
or
number
two,
you
don't
have
any
a
goodness
when
you
just
perform
apply
with
with
a
new
name,
say.
F
F
F
What
he's
trying
to
say,
if
you
basically
alexander,
if
I
understood
correctly,
if
you
rename
the
cluster
and
you
create
like
a
new
one,
the
workflow
to
the
playagonics
is
not
going
to
work
right.
0,
13.,.
E
I
mean
I
mean
if
you
just
have
to
form
apply
first
time.
It
would
create
all
what
you
need
and,
if
you
just
to
create
say
to
reform,
apply
with
a
different
name
of
the
cluster.
It
would
just
create
a
cluster
for
you
and
you
can
make
the
reform
plan
to
see
what
resources
it
would
update
and
it
would
update
only
a
cluster
and
all
resources
for.
C
E
And
no
argonauts
inside
so.
E
F
Okay,
so
if
you
rename
the
cluster
gketer,
the
gke
terraform
resources
will
recreate
the
cluster
and
for
what
I
understand,
it's
not
going
to
deploy
again
it's
basically,
if
you
already
have
a
cluster
running
with
agonies
in
terraform,
and
you
rename
the
cluster
which
will
recreate
it.
It's
not
going
to
depend
on
instagram.
F
Yeah,
the
problem
lies
in
how
terraform
handles
state
with
gke
cluster
renamed.
Basically,
I
think
it's
a
very
specific
issue.
E
And
the
same
with
the
new
version
of
kubernetes
could
be
the
case,
but
I
have
not
tested
it
yeah.
Let's
say
the
comment
here
not
to
forget
yeah,
just
reform.
Yes,
so
with
new
version
of
kubernetes.
E
Is
there
a
terraform
bug
for
this
yeah?
I.
E
Well,
let's
I
will
add
the
inside
our
meeting
docs.
If.
G
B
H
Yeah,
so
my
question
wasn't:
I
don't
think
it
was
particular
to
this
bug,
but
the
question
was
in
terraforms:
can
terraform
create
crds
for
us
yet
with
the
united
provider,
because
in
that
case,
if
you're
doing
it
with
telephone,
you
can
just
do
full
terraform
and
bypass
helm.
H
Because
I
mean
personally,
I
would
like
crds,
because
the
way
I
logically
think
about
stuff
is
like
the
bootstrapping
of
the
cluster
and
installing
crds
is
part
of
the
bootstrap
into
the
cluster.
For
me,
which
is
terraform,
then
everything
after
that,
which
is
the
applications
in
the
cluster,
is
span
up
using
customize.
B
B
H
B
Have
a
built-in
function
now
to
frickin
yaml
decode
for
converting
yaml
to
hcl.
You
know
fancy
anyway
yeah.
Let's
know
how
that
goes.
Actually
we're
kind
of
curious.
G
B
Sweet,
oh
there's
lots
of
stuff
coming
in
now
excellent.
I
just
wanted
to
bring
it
up
in
case
people
hadn't
seen
it
or
wanted
to
add
to
it
a
conversation
that
happened
in
slack
that
ended
up
in
a
ticket
around.
B
What's
the
prerequisite
knowledge
for
getting
started
with
organize
and
dedicated
game
servers,
which,
I
think
is
actually
a
really
good
nice
piece
of
scalable
like
explanation
type
stuff,
we're
getting
a
bunch
of
people
who
are
into
the
project,
who
necessarily
may
not
know
anything
about
dedicated
game
servers
or
know
anything
about
kubernetes
and
so
having
a
good
understanding
of
what
level
of
knowledge
in
both
areas
are
required
before
getting
involved
with
this
project
or
trying
to
get
started
with
it.
B
I
think
is
super
useful
because
seeing
some
people
who
are
like,
I
don't
know
anything
about
kubernetes
we're
going
to
learn
a
gunners
and
kubernetes
and
docker
and
all
the
things
all
at
once,
which
is
just
genuinely
an
impossible
task.
So
if
you
get
a
chance
to
have
a
look,
I
think
there's
some
great
resources
in
there.
If
there's
anything
else,
you
want
to
add,
please
feel
free.
I
think
we
have
a
pretty
good
thing.
We
just
need
to
write
it
and
I'll
probably
end
up
doing
it
unless
somebody
jumps
on
it.
First.
C
B
Meeting
notes,
oh
in
the
in
the
ticket,
okay,
yeah,
there's
a
bunch
of
stuff,
so
I
outlined
what
I
thought
would
be
the
most
pertinent
concepts
for
docker
and
kubernetes
people
seem
to
be
like
that's
a
good
thing
then
asked
also
what
resources
people
liked
people
pointed
to
the
getting
started
experience
as
well
as
apparently,
everyone
loves
cuda
coda,
which
is
awesome,
as
I
think
it
was
dom.
Actually,
who
also
pointed
out
that
yes,
people
should
know
about
dedicated
game
servers.
B
I
was
like
that
makes
a
lot
of
sense.
Actually
I
didn't
think
about
it.
That
way,
I'm
silly
probably
need
some
good
resources
for
that
I
did
a
silly
video
about
it
if
other
people
have
other
thoughts
on
that.
It's
on
like
how
to
find
a
good
example
of
how
to
explain
that
that
would
also
be
handy,
but
there's
some
good
yeah,
there's
some
good
stuff
in
there.
So
if
people
have
other
thoughts,
this
is
especially
true.
B
If
you
are
actually
a
beginner
and
you
you
want
to
help
in
this
area,
it
is
very
hard
to
go
back
to
being
a
beginner
once
someone
has
been
doing
stuff
for
a
long
time,
so
that
is
useful
knowledge
to
have
so,
if
there's
path
you
took,
please
explain.
C
Yeah
it
feels
like
in
our
sort
of
getting
started.
We
need
like
a
getting
started
from
complete
scratch
versus
getting
like
hear
the
sections
you
should
skip
forward
like
if
you
already
understand
kubernetes
like
skip
past
this.
If
you
already
understand
what
a
dedicated
game
server
is
get
past
this,
but
like
the
the
basic
flow,
should
probably
have
more
information
linking
out
to
like
the
resources
for
people
to
get
up
to
speed
on
those
things.
Yeah.
B
H
I
H
C
I,
like
it
yeah,
that's
especially
true
like
when
I
was
trying
to
update
the
sdks,
the
games,
the
game
engines,
it's
really
difficult
to
figure
out
like
how
do
you
build
and
put
this
thing
in
a
container
and
like
test
stuff
and
get
it
on
onychonates
or
kubernetes
at
all
right
so
like.
I
was
certainly
struggling
with
that,
like
it's
easy
for
me
like
the
simple
udp,
went
up
and
running
right
but
like
yeah.
If
we
have
links
to
help
people
with
like
the
specific
engine
that
they're
going
to
be
using,
I
think
that'd.
B
C
C
I
think
in
to
play
devil's
advocate
to
your
proposal
mark.
I
think
I
wouldn't
put
something
above
the
overview,
because
I'd
worry
that
people
would
skip
that
and
the
overview
kind
of
says,
like
we
do
dedicated
game
servers
on
kubernetes.
It
links
to
both
those
things
and
gives
like
a
background
of
like
what
the
product
is
and
why
we
do
it
and
it
feels
like
what's
next,
should
be
like
as
you
get
started
like,
then
you.
D
B
C
B
B
B
Yeah
this
is
this
is
something
that
comes
up
this.
I
keep
asking
people
this.
It
comes
up
in
slack
a
lot
like
the
common,
so
this
is
having
a
like
a
solutions,
template
or
a
solution
section
on
the
website.
So
the
thing
that
comes
up
on
a
regular
basis
is
usually
how
do
I
do
web
sockets
with
a
gunners.
B
It
probably
comes
up
once
every
three
to
six
months
and
there's
a
few
different
ways
that
people
can
do
it
and
what
I
usually
say
at
that
point
is
people
go
off?
There's
a.
We
have
an
issue
ticket
that
sort
of
describes
some
solutions
and
some
history
that
describes
some
solutions,
but
I
sort
of
go
to
like
it
would
be
great
if
someone
was
like
this
is
my
opinionated
way
of
solving
x
right,
like
websockets,
something
else,
and
so
I
was
thinking
it
would
be.
B
Maybe
it
doesn't
seem
like
anyone
has
taken
me
up
on
the
invitation
to
write
a
solution
which
made
me
wonder
like.
Should
we
write
a
solution
section
and
just
put
it
there?
Should
we
write
a
solution
section,
maybe
have
a
solutions
template
to
sort
of
like
help
people
along?
Does
this
sound
like
a
good
idea?
I
don't
know
I'm
sort
of
just
floating
it
because
it
seems
like
I'm
just
looking
for
ways
to
scale
out
helping
people.
B
B
H
There,
the
only
reason
why
I
ask
is
that
to
me
the
whole
web,
sockets
plus
kubernetes
thing
I
didn't
quite
know
how
they
meet,
why?
Why
was
it?
How
do
websockets
work
in
containers.
B
A
H
So
that
that's
yeah
for
some
reason
I
was
thinking
websockets
within
the
cluster,
because
I've
done
that
before
I
didn't.
I
wasn't
really
thinking
about
outside
the
cluster
but
yeah.
It
makes
sense.
C
A
B
I
did
it
again.
Okay,
I
know
what
the
problem
is.
C
So
mark
I
was
going
to
ask,
I
think
there
are
sort
of
two
ways:
I've
seen
this
in
the
past.
One
is
you
know,
often
times
when
people
build
a
solution
that
they,
you
know,
want
to
write
a
blog
post
on
their
company,
blog
or
somewhere
else
that
we
had
linked
to
right.
So
if
company
x
does
something
with
websockets,
they're
gonna
go
nace
and
they
wanna
write
that
up.
Then
they
can
do
that
and
sort
of.
C
Let
us
know
that
it's
there
and
we
can
link
to
it
right
if,
if
these
are
intended
to
be
opinionated
solutions
like
that's
a
great
place
for
it
to
be
really
obvious
whose
opinion
it
is
and
like
we
can
sort
of
link
out
to
that.
I
think
the
one
concern
I
have
with
putting
it
on
the
ego
next
site
is.
It
starts
to
look
like
it's
like.
C
This
is
the
endorsed
solution
and
might
might
prevent
people
from
like
looking
at
other
ways
to
do
it
or
thinking
that
like
that
is
the
only
way
that
it
should
be
done.
So
if
we
put
on
our
website,
we
just
need
to
make
really
clear
like
here
are
some
ways
people
have
done
this
like
dom
said.
These
are
examples.
These
are
not
like
yeah
the
solution
right.
This
is
not
the
way
to
do
it.
This
is
a
way
someone
has
done
it
right.
H
What
about,
I
think,
a
couple
of
communities
use
the
contrib
repos
and
you
can
have
like
this
isn't
officially
supported,
but
contributions
community.
So
it
could
be
one
of
those
things.
A
A
C
H
Work
exactly
I
mean
obviously,
ecosystem.
C
I
would
say
april:
how
does
that
work
in
terms
of
like
documentation?
Right,
like
I
understand,
like
a
contributor,
if
somebody
wants
needs
to
like
dump
some
code
or
some
configs
or
that
sort
of
thing,
but
like
we
want
to
link
our
website
to
something
right,
would
be
linked
to
a
readme
and
github
would
be
like
put
pages
on
the
website
like
how
does
how
does
grpc
handle
that.
A
A
I
do
think,
like
you
know,
the
big
thing
is
just
setting
that
expectation.
I
would
just
point
to
like
the
repo
and
be
like
here's
a
place
where
these
kind
of
things
are
held.
You
know
use
at
your
own
risk.
I
don't
think
that
we're
there
yet,
but
I
know
like
with
grpc.
We
actually
have
a
forum
process
where
someone
wants
to
be
in
that
repo
they
have
to.
A
A
B
Makes
sense,
I
think,
to
your
point,
rob
I
think
it
whatever
we
do.
It
needs
to
be
explicitly
cleared
that
it's
coming
from,
probably
an
external
company
or
external
user.
It's
not
a
like
it's,
not
official.
It
is
an
opinionated
solution
and
like
we
need
to
make
sure
it's
yeah
yeah
that
I
think
that
makes
100
sense
right.
Yeah.
C
B
And
to
your
point
as
well
like
if
a
solution
stops
working
or
is
no
longer
valid,
or
something
like
that
like
what
do
we
do
with
it
and
should
the
onus
be
on
us
is
actually
a
really
good
question,
and
so
maybe
that's
a
good
point
to
just
say
yeah.
If
you
want
to
do
that,
put
on
your
own
website,
we'll
link
to
it.
A
J
Is
there
is
there
value
in
adding
more
sample
code
for
some
of
these
common
solutions?
Well,
and
do
these
third-party
solutions
exist
that
we
can
even
link
to.
A
B
B
B
Yeah.
We
do
have
a
third
party
section
in
the
website
as
well
like
where
we
point
to
other
people's
projects,
but
that's
the
thing.
Okay,
I
mean
I'm
happy
to
let
it
sit.
I
just
wanted
to
bring
it
up
as
an
idea.
I.
J
Think
we're
really
valuable.
It's
just
like
a
question
like
say
this.
This
web
stock
at
one
like,
is
there
actually
a
solution
that
someone's
prepared
externally
on
web
sockets
in
existence,
or
is
it
gonna
be
like?
Are
they
waiting
for
a
reference
implementation
first
and
then
we'll
start
to
see
the
blog
posts.
A
Well,
yeah
because
we
could
have
a
running
list
like
these
are
the
things
we
want
and
once
we
tell
people
what
we
want
they'll
totally
just
make
it
that's
how
it
works.
I'm
curious,
like
you
know,
where
they're
it
sounds
like
you
know.
Maybe
at
least
a
couple
of
folks
would
be
interested
in
kind
of
maybe
working
with
me
on
our
github
org
organization.
A
We
can
create
like
a
little
a
little
working
group,
our
first
working
group,
anybody
interested
you
don't
have
to
all
jump
together
right
now
we
can
yeah,
but
that
I
do
think
like,
as
the
project
has
grown
like
this
was
always
the
thing
of
like
there's
nothing.
It's
a
lot
easier
to
start
off
when
it's
still
small
yeah
and
to
try
to
pull
these
things
out
into
their
own
repo.
Later.
B
One
thing
I
do
have
this
is
from
the
last
meeting
we
talked
about
pulling
like,
for
example,
the
site
into
its
own,
like
build
docker
image
and
its
own
cloud
build
script
a
then.
B
That
means
we
could
do
like
conditional
testing
so
like
if
your
stuff
doesn't
touch
the
website,
then,
like
we
don't
run
those
tests,
which
is
nice
and
also
lets
us
transfer
over
to
the
actual
official
like
cloud
build
github
integration,
stuff,
which
we
don't
use
right
now,
because
it
wasn't
around
when
we
first
did
it,
it's
never
been
updated,
and
that
could
be
a
nice
first
step
to
getting
that
stuff
into
its
own
repo.
B
If
we
want
to
do
that,
but
at
least
it
then
gets
it
separated,
which
is
nice,
there's
a
ticket,
it's
got
like
help
wanted
good
first
issue,
I
think
on
it,
or
at
least
help
wanted.
A
C
I
I've
seen
very
few
prs
that
like
have
to
touch
stuff
in
the
website
and
the
code
at
the
same
time,
so
I
think
that's
a
pretty
clean
split.
We
also
talked
potentially
about
sdks
and
maybe
move
those
around.
I
think
that's
a
messier
split,
but
the
website
seems
like
a
sort
of
pretty
clean
thing
to
move
out.
A
B
A
Issue
just
a
quick
time
check,
you
still
have
a
lot
to
go
through
dom.
You
have
a
question.
You
want
to
talk
about.
Yeah.
H
So
one
thing
I'm
using
in
the
minute
is
I'm
heavily
using
player
tracking,
which
is
great
and
as
part
of
that,
having
my
ccus
from
a
decent
source
would
be
quite
nice.
So
I
was
thinking
of
putting
something
like,
oh,
if
there'd
be
any
pushback
of
putting
open,
telemetry
or
prometheus
inside
the
side
car,
because
it's
not
there
at
the
moment
and
exposing
the
ccu's
per
game
server
on
that
endpoint.
B
I
I
don't
know
anything
about
how
the
metrics
were,
but
the
question
I
will
ask
that
is
a
vague
drawing,
let's
line
in
the
sand,
since
we
have
that
information,
the
game,
server
level
and
we're
pushing
out
metrics
through
the
controller.
Could
we
capture
those
metrics
there
and
push
it
out
that
way,
rather
than
adding
an
extra
thing
to
the
cycle.
H
B
Changes
and
it's
already
cached,
it's
all
in
a
memory
case,
so
it's
not
a
huge
amount
extra,
and
we
also
already
have
all
that
machinery
in
place
to
watch
for
stuff
and
then
push
out
things
like
what
states
are
things
what
counts
thing
have
so
anyway,
it's
an
option.
We
could
write
it
it's.
Yes,
I
think
the
metrics
are
good.
It's
probably
worth
writing
a
design
ticket
and
going
backwards
and
forwards
on
right,
actual
stuff.
I'll,
probably.
C
B
B
H
Yeah
and
the
the
kind
of
related
then
is
kind
of
is,
is
going
to
be
adopting
open
telemetry
is
that
the
kind
of
route
we're
going
to
be
going
down.
C
Yeah
I
actually
checked
on
this
like
a
week
ago,
because
I
saw
that
ticket
was
like.
Oh,
I
wonder
where
open
telemetry
is.
I
heard
somebody
on
the
podcast
talking
about
it
and
when
I
checked
it
looks
like
they
are
still
maybe
just
got
to
beta,
like
they're,
not
quite
v1,
and
I
think
we
were
sort
of
waiting
for
them
to
hit
p1
before
we
did.
The
migration
work.
H
E
K
C
Where
it
says,
overslumber
doesn't
seem
to
be
available
yet
so
we
should
use
open
census
and
open
tracing
for
now,
and
I
think,
rather
than
instrumenting
with
tracing
we
were
just
like,
since
nobody
actually
like
started
working
on
that-
and
I
think
at
this
point
I
was
kind
of
looking
at
it
going
open
telemetry.
It
seems
like
it's
getting
close.
We
probably
need
to
switch
to
that
anyway,
so
like
when
we
do
that
we
can
also
add
tracing.
C
Which
also
begs
the
question
of
whether
what
we
have
an
open
census
does
anything
at
all
of
what
you're
looking
for
for
metrics
right
because
like
mark
was
saying,
like
you
know,
like
the
the
metric
stuff,
I
think
tyrol
is
driving
a
lot
of
and
you
know
making
it
work
for
himself
the
way
he
wanted
metrics
to
work.
And
I
don't
think
there
were
a
lot
of
other
people
that
had
really
strong
opinions
on
metrics.
And
so,
if
you
have
opinions
or
ideas,
we'd
love
to
hear
those.
H
Is
that
somewhere
in
the
project
I
mean
in
agonist
project
yeah,
that's
the
controller
stuff
yeah.
I
can
have
a
look
at
that
when
I
jump
on
and
look
at
doing
the
cc
news.
B
H
H
It
seems
as
though,
at
least
from
the
external
viewpoint
is
that's
what
google
do,
and
so
I
thought
that
maybe
it
should
be
something
that
the
project
adopts
for
any
new
apis
and
we
had
a
bit
of
backwards
and
forwards
around
how
that
may
look
for
player
tracking
and
if
we
want
to
do
it
for
player
tracking,
I'm
not
sure
quite
how
much
internal
change
it
would
cause
for
tracking.
B
Did
we
not,
I
thought
we
did.
H
Oh
there's
a
country,
maybe
we
should
we
put.
Maybe
we
should
add
this
link
and
to
contributing
make
it
stand
out.
B
I
think
my
general
takeaway
from
that
was
there
was
some
discussion
around
like
which,
like
some
of
the
finer
details
of
that
yeah,
that
yes,
it
was
a
good
idea,
and
it
just
requires
someone
to
actually
do
the
work.
Cool.
B
That
being
said
at
some
point,
we
will
go
from
alpha
to
beta
and
then
beta
to
ga,
so
we
have
opportunity
to
break
it
along
the
way
or
even
within
within
the
alpha
cycle
and
stuff
of
player
tracking
where's
player
trekking
at
that
is
yeah
yep.
That
is
a
good.
C
B
B
I
think
primarily
most
of
the
change
was
on
the
rest
side,
which
is
a
lot
easier.
I
think
I
think
the
unreal
will
probably
change
the
most
there.
I
think
there
are
some
minor
changes
to
the
grpc
service,
naming
I
think,
from
memory.
C
Oh
yeah,
this
kind
of
flows
nicely
into
steve's
question
next,
which
is
where
are
we
at,
and
the
answer
is
we're
in
alpha.
It
sounds
like
we
potentially
have
some
api
changes
we
want
to
make
that
will
be
breaking
to
do
before
we
promote
to
beta.
Are
you
kind
of
asking
steve
like
when
we'll
get
to
beta
nga,
that's
kind
of
what
it's
gathering.
J
Yeah
yeah
pretty
much
just
to
see
if
there's
any
more
any
more
changes
required.
B
Actually,
thanks
for
bringing
this
up,
one
of
the
things
I
was
going
to
try
and
start
tackling
in
this
sprint
as
well,
is
to
add
some
of
the
features
that
take
advantage
of
player
tracking
to
try
and
also
get
some
more
adoption,
so
allocation
based
on
capacity
and
auto
scaling
based
on
capacity,
because
once
those
are
in
then
like
people
are
actually
going
to
want
to
use
them
and
which
means
we're
going
to
find
bugs
and
we're
going
to
go
back
and
fix
them.
So
that
was
that
was
my
thought.
H
That's
interesting
because
I'm
kind
of
doing
allocation
based
on
player
tracking
in
a
minute,
but
I've
got
my
own
thing
because
there's
no
time
to
tie
to
the
game
that
it
sort
of
could
be
pulled
out,
but
a
lot
of
it's
very
game
specific,
so
yeah.
That's
why
I'm
using
player
tracking
to
see
like
what
game
worlds
are
running
and
who's
in
them
and
then
to
put
people.
B
H
C
B
C
Yes,
yeah
to
marx
to
mark's
point,
though,
like
the
the
goal
of
having
an
alpha
is
for
people
to
use
it
us
to
make
sure
that
it's
good
enough
to
promote,
we
want
to
keep
supporting
it,
and
so
really
the
feedback
we
want
is
like.
Yes,
this
is
a
feature
that
I
really
need.
It
works
great
the
way
it
is
like
we
should
promote
it.
I
think
the
the
feedback
that
don
was
given
earlier
was
like
yes,
the
feature
I
need.
It
doesn't
quite
work.
C
B
In
beta
that
can
probably
get
g8,
I
think
yeah
tuna
has
one
of
those
on
the
list
coming
up.
What
was
I
gonna
say?
Do
we
need
some
sort
of
formal
version
for
like
or
some
some
formal
process
to
be?
B
Like
you
know,
at
the
end
of
every
release
like
during
a
community,
maybe
part
of
a
community
meeting
we
go
through
whatever
is
in
alpha
beta
and
see
if
there's
stuff,
that
should
be
stabilized
or
something
like
that,
so
that
we
make
sure
that
nothing
languages,
because
it
seems
like
right
now
it's
sort
of
just
like
if
somebody
goes
and
looks
and
does
does
a
check,
then
that's
when
it
gets
that's
when
it
gets.
You
know
sort
of
graduated.
J
B
Something
like
something
like
that,
I
don't
know
between
releases,
I'm
not
sure
what
the
right
solution
there
is,
or
maybe
this
is
fine
like
right
now,
because
we
have
it
listed
and
documented
of
what
is
in.
C
Like
the
link
in
chat,
so
this
is
what
it
says
like.
Maybe
it's
just
as
simple
as
looking
at
feature
stages
during
this
meeting
and
saying
like
yeah.
Let's
look
at
feature
stages,
see
the
things
that
aren't
listed
as
ga
and
give
like
a
quick.
You
know
round
the
table
of
are
people
feeling
like
it's
ready
to
be
promoted
or
people
feeling
like
it
needs
more
feedback,
et
cetera
and
at
least
sort
of
touch
base
with
each
of
the
things
that
are
on
that
list.
Yeah.
B
C
Yeah,
I
think
that
that
makes
sense.
I
think
if,
if
the
consensus
in
this
meeting
is
that
it's
not,
then
we
maybe
shouldn't
file
a
bug,
but
if
the
consensus
meeting
is
like
yeah,
we
think
it's
ready,
then
we
should
file
a
bug
to
get
people's
voices
that
can't
make
it
right.
But
if
people
here
are
objecting,
then
it's
probably
not
worth
the
extra
process
of
like
asking.
If
other
people
also
object
right,
there's
also
a
nice
formatting
issue
at
the
bottom
of
that
table.
We
should
fix.
C
K
I
can
talk
about
it,
so
multi-cluster
allocation
been
added
for
a
while
it
was
changed
to
beta.
The
api
servers
for
economist
allocator
is
stable
on
vivon
and
crd
that
is
used
for
multi-cluster
allocation
is
also
unstable.
However,
the
documentation
is
still
documented
as
a
better
feature,
because
we
haven't
received
enough
feedback.
K
I
see
a
lot
of
recently.
A
lot
of
issues
relatively
was
filed
against
that,
and
we
tried
to
address
some
of
them
in
the
last
release
for
1.8
this
release.
There
are
a
couple
of
more
one
is
fixed
for
supporting
the
this
reserved
ip
for
issuing
the
certificate
and
there's
another
one
which
robbie
also
added
to
the
document
for
timeout
on
the
grpc
and
then,
if
there's
no
more
requests,
then
I
think,
based
on
the
use
that
the
feedback
we
get,
we
should
decide
to
make
it
stable.
K
Most
of
the
issues
came
from
improbable
company.
I
don't
know
if
anybody
from
that
company
is
on
the
call,
but
I
would
like
to
hear
if
there
is
somebody
right
now
against
basically
moving
it
to
stable
or
for
moving
it
to
stable.
H
I
think
I
would
have
like
so
I
used
to
work
at
improv
world,
so
you
know
pretty
much
what
they're
doing
with
it.
It
seems,
though,
adam
raised
a
ticket
around
it
around
the
timeout.
H
I
think
from
what
I
remember
can't
really
go
into
detail,
but
it
seemed
fine
for
what
they
were
using
it,
for
I
think
multi-cluster
allocation
in
general
will
be
good
and
I
think
most
people
will
be
running
multiple
clusters
based
on
geolocation.
So
I
think
it's
definitely
a
use
case
that
everyone
will
be
wanting.
H
H
B
C
Yeah,
I
was
also
going
to
say,
like
you
know,
lots
of
other
parts
we're
going
to
certain
ga
and
we
still
find
little
bugs
and
we
fix
them
right.
So
we
shouldn't
let
the
fact
that
we
found
a
bug
like
block
us
from
saying
it's
ready
to
graduate
like
if
we
believe
that
it's
stable
and
it's
something
we
as
a
community
are
going
to
keep
supporting
and
keep
using.
Then
that
also,
I
think,
argues
for
it
being
promoted
to
ga.
C
Sure
I
was
just
going
through
and
trying
to
add
some
recent
tickets
to
the
agenda
at
the
beginning
to
make
sure
that
sort
of
things
got
surfaced
that
people
have
brought
up
recently
at
least
one
of
them.
I
asked
somebody
if
they
wanted
to
come
to
this
meeting.
I
think
the
first
one
was
filed
by
nikhil
who's
working
with
pune
at
google
and
it's
about
installation
and
in
particular
about
do.
C
The
allocator
service,
of
course,
depends
upon
the
crds,
so
you
have
this
sort
of
dependency
graph
of
like
pretty
much
everything
you
always
need
to
crds
and
then,
like
the
other
pieces,
can
largely
be
installed
independently
and
we
have
some
flags
to
turn
some
of
them
on
or
off.
And
you
know
he
was
exploring
a
used
case
where
we
wanted
to
basically
turn
everything
else
off,
except
for
the
allocator,
which
is
not
a
use
case
that
we've
supported
to
date,
and
we
had
a
little
bit
of
discussion
in
the
ticket
about.
C
You
know
how
we
might
be
able
to
modify
the
existing
home
chart
or
create
a
new
help
chart
et
cetera.
So
I
wanted
to
sort
of
just
put
this
out
there
so
that
if
anybody
else
had
similar
use
cases,
they
could
chime
in
on
the
issue
or
if
there
were
other
parts
of
a
donates,
that
people
thought
would
be
useful
to
be
able
to
install
separately
that
we
could
also
chime
in
on
the
issue
there.
H
So
from
my
side,
I'm
using
default
yaml
install
and
find
it
quite
easy
just
to
not
install
parts.
H
H
So
if
we
do
go
modular,
I
would,
I
think,
adding
lots
of
different.
Smaller
demons
can
often
get
confusing.
K
B
Know
I
think,
I
think,
to
your
point
dumb.
What
you're,
probably
trying
to
say,
is
like
by
default.
Have
it
all
worked
together?
If
you
want
to
split
it
out,
rather
than
be
like
hey
make
sure
you
grab
these
seven
different
things
and
put
them
all
in
together
to
make
it
all
work
right
like
by
default.
It'll
work,
yeah
yeah,
so
you
can.
You
can
opt
out
of
things
where
you
don't
have
to
opt
into
everything.
C
Yeah,
I
think
that
makes
a
lot
of
sense.
We
don't
want
to
force
people
to
have
to
run
five
different
installers
to
get
the
whole
thing
right.
I
think
that's
what
you're
saying
dom
is
like
if
seo's
installation
method
is
like
run,
these
eight
different
things,
and
then
you
have
istio
like
that
sounds
not
very
user
friendly
than
just
one
run
thing.
I
think
what
we're
exploring
is
use
case
where
we
don't
need
everything,
so
we
might
as
well
turn
off
the
stuff.
C
We
don't
need,
and
that
would
basically
be
an
opt-out
of
like
I
want
to
install
it
without
these
pieces
right
would
be
sort
of
the
way
you
express
that
through
your
home,
configuration
is
like
hell,
install
dash,
dash
skip
this
skip
this
get
this
or
whatever.
C
Cool,
so
the
other
two
issues
that
I
linked
here
are
very
much
related
and
they're,
both
about
supporting
game
servers
running
in
multiple
namespaces.
So
by
default,
when
you
install
the
donates,
you
get
name
servers
running
in
the
default
name
space
and
you
have
to
tweak
a
number
of
things
if
you
want
to
have
it
manage
game
servers
and
other
namespaces.
C
People
and
people
are
looking
for
solutions
to
make
this
easier,
so
the
most
most
recent
ticket
somebody
was
saying
that
their
their
workflow
is
that
they,
you
know,
add
a
new
person
to
their
their
team,
and
you
know
that
person
gets
a
new
name
space
and
they
want
things
to
just
work
right
and
right
now
it
doesn't,
and
you
know
that
person
gets
frustrated
and
they
have
to
go
back
and
and
realize
what
went
wrong
and
and
go
fix
their
they're
going
to
installation
to
manage
this
new
namespace
they
created,
and
I
think
the
first
the
first
ticket
was
like.
C
Let's
make
it
easier
to
tell
agonies
which
namespaces
are
managed
and
I
think
alex
you
went
back
and
forth
on
that
one
with
robbie
haywood
a
little
bit
to
try
and
figure
out
if
there
was
a
way
to
change
a
guys
to
make
it
more
sort
of
friendly.
And
so
I
think
two
things
have
come
out
of
this.
I
think
one
is
there's.
C
There's
always
some
tension
here
between
security
needs
abuse
right
like
right
now,
our
posture
is
a
little
bit
more
secure,
which
makes
it
a
little
bit
less
friendly
to
people
and
we
can
decide
if
that's
the
right
posture
to
have,
but
the
second
one
is.
I
think
this
is
really
illustrating
the
fact
that
a
lot
of
people
are
using
more
than
one
namespace
or
a
non-default
namespace
for
their
game
servers,
which
you
know
probably
wasn't
the
case
when
the
project
was
started,
and
it
was
much
more
of
a
like.
C
Let's
get
stuff
up
and
running
really
quickly
and
people
just
throw
some
default
namespace,
because
it's
easy
and
you
don't
have
to
specify
extra
flags
to
cuddle,
but
as
people
sort
of
use
things
more
for
production
and
more
seriously,
I
think
the
use
of
name
spaces
is
is
going
to
be
more
and
more
common.
So
I
think
again
this
wasn't
looking
for,
like
let's
design
a
solution,
this
meeting
but
sort
of
get
this
out
in
front
of
people
like
if
you
have
use
cases
where
you
are
in
multiple
namespaces
or
non-default
namespace.
C
It
would
be
great
to
chime
in.
If
you
have
ideas
of
how
to
make
this
easier,
it
would
be
great
for
you
to
chime
in
and
if
you
think
that
our
security
versus
ease
of
use
posture
is
wrong,
like
we'd
love
to
hear
that
too,
so.
B
I
was
going
to
add
a
note
to
that
would
also
love
to
know
if
you
have
different
namespace
needs
for
development
versus
production
and
what
those
are.
So
my,
I
wonder
if
people
are
more
inclined
to
use
more
namespaces
in
development
because
they
might
be
sharing
a
cluster
versus
production,
but
I
don't
know.
B
I
have
again
like
to
ruby's
point
like
that.
Has
different
security
versus
ease
of
use
needs
for
both.
E
B
Sounds
like
a
installation
option
on
the
helm,
template
somehow
or
is
a
feature
gate
or
something.
Is
that
what
you're
going
to
say,
robbie
well.
C
Like
well,
it
comes
back
to
your
question
about
like.
If
this
is
for
non-production
use,
then
we're
willing
to
have
sort
of
a
different
security
posture
for
non-production.
We
could
make
a
flag
on
install,
so
it's
like
hey.
I
want
to
install
this
development
cluster
and
I'm
okay,
using
my
security
posture
to
make
my
life
simpler.
But
when
I
go
install
my
production
cluster,
I
don't
want
to
do
that,
and
so
that
comes
back
to
like
are
people
running
things
differently
in
different
clusters?
C
Are
people
okay,
with
having
their
configuration,
be
different
in
production
and
non-production
or
people
want
things
to
be
consistent,
so
things
are
tested
the
same
way
right,
so
I
think
we
sort
of
need
to
tease
apart.
You
know
how
people
want
to
use
this
and
make
sure
we're
solving
it.
The
right
way.
E
Yeah-
and
there
is,
there
was
a
case
that
we
can
read
secrets
from
every
namespace
by
a
gunner's
controller.
So
that
was
the
case.
B
All
right,
yeah,
I
think
it's
interesting,
I
think,
alex
to
the
the
like
the
two
tickets,
when
you're
mentioning
and
like
robbie's
mentioned
to
both
of
them.
I
think
one
of
them
was
about
making
security
tighter
and
the
other
one
was
about
making
the
ease
of
use
easier
by
making
security
less.
C
Yes,
it
has
an
effect,
I
mean
it's,
it's
not
like
hard
multi-tenancy,
but
it
does
generally
provide
a
separation
of
concerns.
You
can
apply
a
lot
of
the
role-based
authentication
at
the
name
space
level.
So
you
know,
if
you
do
things
correctly,
you
can
say
this
actor
is
allowed
to
access
secrets
and
namespace
a,
but
nowhere
else
right,
and
so
that's
especially.
C
If
you
can
break
out
of
containers,
if
you
can
get
root
on
that
node
you
can,
then
you
know
potentially
inspect
what
the
cube
is
doing
or
advertising
split,
and
you
know
access
things
across
namespaces,
it's
not
trivial,
and
so,
if
you
are
concerned
about,
like
your
code,
doing
the
wrong
thing
like
you
can
use
our
back
to
enforce
your
code,
doesn't
accidentally
reach
across
namespaces,
but
it's
not
a
sufficient
sort
of
sandbox
to
run
untrusted
code
and
assume
that
it
can't
break
out
right.
C
So
you
need
another
layer
of
security
for
that
cool
yeah.
But
it's
great
for
cases
where
you
have
like
a
bunch
of
developers
at
the
same
company
who
want
to
run
stuff
and
not
step
on
each
other's
toes
right
and
I
think
that's
what
somebody
expressed
in
the
ticket
was
like
hey.
We
want
to
give
each
person
their
own
little
area
to
play
with
we'll
give
them
a
name
space.
B
C
Right,
I
mean,
even
if
you
were
running,
like
a
production
system
with
two
games
from
the
same
company
that
were
managed
by
the
same
team
like
that,
would
be
reasonable
to
to
put
in
a
cluster
with
two
namespaces,
because
the
people
who
have
access
to
the
cluster
you
would
want
to
have
access
to
both
anyway
and
it
might
be
less
less
overhead
and
work
for
them
if
they're
all
in
the
same
cluster
right.
So
I
think
there
are
lots
of
legitimate
cases
for
using
namespaces,
but
it's
more
of
like
a
resource
separation.
C
You
can
make
it
more
firm
by
using
things
like
chains
and
tolerations
and
preventing
things
from
scheduling
on
the
same
nodes
and
that
sort
of
stuff,
if
you're
trying
to
to
really
really
go
down
the
multi-tenancy
route,
but
even
there
like
there's.
There
are
a
lot
of
people
in
the
kubernetes
community
that
are
trying
to
improve
the
multi-tenancy
posture
and
realizing
that
it's
it's
not
at
a
a
hardware,
dependency
level
right.
H
C
A
I
did
oh
well
done
since
we
pretty
much
are
at
time.
I
would
just
say
you
know
we'll
continue
conversations
and
slack
and
github
and
everything
else
thanks
everybody
for
joining
and
having
such
a
great
discussion
and
then
we'll
see
y'all
again
in
september.