►
From YouTube: Agones Community Meeting - October 2022
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
C
Let's,
let's
take
it
off
Ravi
you
want
to
talk,
I
wasn't
sure
if
you'd
make
it
so
I
stuck
my
name
on
the
first
one,
but
feel
free
to
replace
it
with
yours.
The
127
release
just
got
cut
so
release
candidate
was
last
week.
The
release
got
finished
yesterday,
so
yay
it
was
the
first
time
Ravi
cut
a
release.
We're
all
very
excited
went
very
smoothly.
C
And
let's
see
I'm
just
reminding
myself,
what's
new
in
case
people
haven't
gone
and
read
the
release,
notes
we've
upgrade
to
a
version
of
go
so
for
developers
maintainers,
we
can
start
using
fun.
New
features
go
finally
added
support
for
generics,
which
is
kind
of
cool
I,
know.
We've
made
lots
of
other
improvements,
so
we've
got
image.
Streaming
is
now
enabled
by
default
on
gke.
If
you
follow
the
getting
started,
guides
or
use
the
terraform
has
been
provided.
C
It's
also
been
turned
on
on
the
ee
cluster
until
the
tests
are
running
through
that
configuration,
we've
added
a
feature
where,
if
you
want
to
use
a
cluster
Auto
scaler
to
do
cluster
compaction
by
evicting
your
pods,
you
need
to
make
sure
you're
setting
graceful
termination
properly.
But
you
can
now
do
this
before
we
prevented
you
from
doing
it.
So
this
is
great.
If
you
have
a
use
case
where
you
know
that
your
game
servers
can
be
evicted
with
with
some
time
to
shut
down
safely
and
I
think
are
the
the
biggest.
B
A
C
Yeah
so
I
guess
for
other
people,
it's
it's
kind
of
buried
by
Design
right.
This
is
a
very
Advanced
feature.
You
should
use
it
if
you
know
what
you're
doing
and
if
you
know
what
you're
doing
you're,
probably
looking
to
figure
out
how
to
do
it
and
then
you'll
come
across
this
section
and
yeah.
So
basically
you
can
add
an
extra
annotation
previously.
C
If
you
tried
to
add
this
annotation,
we
would
actually
overwrite
it
now
we're
not
going
to
overwrite
it,
which
means
that
if
you
are
specifically
adding
this
annotation
to
your
pods
that
then
now
they
can
get
preempted.
C
So
I
I
think
we
kind
of
mentioned
it,
but
it
actually,
we
should
probably
add
it
to
the
sample
now
that
I'm
thinking
about
this,
usually,
if
you
add
this
annotation
you're,
going
to
also
want
to
add
a
corresponding
graceful
termination
claws
to
your
pot
as
well,
so
I
think.
Actually
that
would
be
something
we
should
add
to
the
documentation
to
show
like
hey
if
you're
adding
this
part.
You
should
also
add
this
part.
A
B
B
So
I
used
to
double
check
and
go
look
at
the
code,
but
I'm
fairly
sure
our
friends
that
we're
looking
at
this
looked
at
the
code
for
the
cluster
Auto
scaler
and
the
cluster
Auto
scaler.
Well,
while
this
is
not
documented,
if
you
look
at
the
code,
it
basically
can
look
at
a
pod
and
be
like.
Oh
if
this
is
attached
to
like
a
deployment.
I
know
what
that
is
so
I
know
it's
safe
to
evict.
B
But
if
I
see
a
pod
that
I
don't
know
if
it's
what
it's
attached
to
is
like
a
grouping,
then
I
think
it
isn't
safe
to
evict.
I'm
gonna
are
on
the
side
of
caution
and
not
evict
it
so
where
we
have
this
in
our
talks
and
we
we
say
here
like
when
using
packs.
We
say
like
that
that
we
add
this
save
to
a
big
false,
which
I
think
is
good,
I,
think
it's
gonna,
be
it
functionality.
C
C
C
But
yes,
you're
right,
the
the
behavior,
the
cluster
Auto
scalar
is
going
to
look
at
the
game.
Server
pods
think
that
they
are
not
sort
of
replicatable,
but
they
don't
have
an
underlying
controller,
that's
going
to
restart
them
and
it
will
not
evict
them
unless
you
specifically
add
The
annotation,
that's
shown
in
the
documentation
that
tells
the
Clusters
autocla.
No,
no,
we
we
know
what
we're
doing
with
this
one
you
can,
you
can
remove
it
if
you
want
to.
C
Awesome
so
next
on
the
agenda
is
there
have
been
a
couple
of
AWS
specific
issues
that
have
been
filed
since
our
last
community
meeting
I
linked
to
both
of
them
here,
I
think
in
both
of
these
cases,
it's
it's
hard
for
a
lot
of
the
sort
of
core
maintainers
of
the
project,
who
don't
really
have
a
great
way
to
test
on
AWS,
to
reproduce
or
and
and
to
diagnose
or
understand
the
underlying
problem.
C
So
I
want
to
kind
of
put
a
call
out
if
there
are
folks
that
are
knowledgeable
on
AWS
we'd
love,
some
help
with
AWS
specific
issues
that
arise
both
on
GitHub
and
in
slack
I
do
feel,
like
the
ones
in
slack,
tend
to
get
more
notice
and
get
more
responses.
C
So
I
guess
maybe
that's
another
thing
to
mention
to
folks.
If
you
are
looking
for
help
it's
great
to
file
an
issue,
but
it's
you
know
oftentimes,
especially
if
you're
on
you
know
an
on
Google
platform,
where
it's
harder
for
people
that
are
monitoring
GitHub,
more
often
to
reproduce
right
if
you're
on
digitalocean
or
Amazon
or
Microsoft
or
bare
metal.
You
know
if
you
post,
something
in
slack
I,
think
more
people
are
going
to
see
it.
C
C
But
if
you,
if
you
are
in
this
situation
and
you
file
an
issue
and
you're,
not
getting
traction,
I
would
highly
suggest
also
posting
in
slack,
so
I
guess,
I
would
suggest
doing
both
file
an
issue
for
persistence
and
then
posting
in
slack,
especially
if
you
don't
get
traction
on
the
issue
to
maybe
get
see.
If
that
helps
get
an
answer.
I've
definitely
seen
cases
where
people
have
posted
in
both
and
the
issue
kind
of
sits
without
a
lot
of
updates
and
then
there's
a
that
was
like
Hey.
C
This
got
solved
in
slack
and
if
you
look
in
slack,
there's
like
a
thread
with
like
20
responses
and
people
debugging
and
working
together
and
getting
through
things
so
I
just
want
to
put
a
call
out
there.
I
don't
know
if
anybody
on
the
call
is
a
is
an
AWS
expert
or
would
know
how
to
fix
these
issues.
I'm,
certainly
not,
but
you
know,
maybe
somebody
who's
watching.
The
recording
will
see
it
and
go
jump
in
and
help
out.
Yeah.
B
Yeah
we
had
one
in
slack
that
had
to
do
with
whether
the
network
was
set
up
as
ipv4
IPv6
a
genuine
something
and
something
wasn't
working.
They
worked
it
out,
it
would
yeah
I.
Just
I
was
like
please
submit
a
PR
to
let
us
know
what
the
thing
should
be.
B
C
Yeah
I
think
the
last
time
I
checked
the
the
answer.
Was
we
needed
to
wait
for
eks
to
bump?
Okay?
Because
if
we
look
at
our
our
official
policy,
it's
that
we
won't
move
to
a
version
that
you
can't
run
on
one
of
the
the
three
major
Cloud
providers
right,
so
yeah
Amazon
doesn't
like
right
now.
Amazon
only
supports
20
21
to
2
and
23
of
kubernetes,
so
I
and
I
think
we're
at
23
right
now
as
well.
C
Well,
yeah,
but
if,
but
if
we
bump
our
project
up
to
124,
then
that
means
that
our
anybody
that
wants
to
use
the
officially
supported
version
combination
is
prevented
from
doing
so
on,
Amazon,
which
either
means
they're
going
to
run
a
SKU
that
we
don't
haven't,
tested
and
don't
support,
or
they
can't
run
the
newest
version,
which
is
going
to
put
pressure
on
us
to
backport
fixes
which
we'd
prefer
not
to
do
oh
well.
C
The
Amazon
eks
release
for
1.24
is
expected
in
November
of
2022.,
which
is
coming
up
next
week.
So
we
can
keep
an
eye
on
that.
I
think
if
we
get
I'd
say
more
than
like
halfway
through
this
release
cycle,
and
it's
not
there-
it's
probably
too
late
to
bump,
and
we
should
do
it
right
after
the
next
one,
which
is
sort
of
very
early
December
foreign.
A
C
Yeah
yeah,
so
the
roof
counted
for
eks
they.
They
actually
say
that
that
upstream
was
May
3rd
2022
for
1.24.
It
looks
like
yeah
Microsoft
is
saying
that
it
was
in
April,
which
is
kind
of
funny
that
they
don't
agree
on
when
upstream
was
cut,
I'm
sure
we
can
go,
find,
find
the
right
answer,
but
I'm
not
particularly
inclined
to
do
that
right
now,
either
way.
I
think
where
we're
at
right
now
is.
You
can
run
124
on
gke.
C
You
can
run
1.24
on
Azure
and
we
expect,
within
the
next
month,
you'll
be
able
to
run
it
on
eks,
so
I
think
the
answer
to
your
question
is:
if
we
want
to
be
ahead
of
the
curve,
we
can
bump
it
later
during
this
next
release
cycle.
Since
we've
got
six
weeks,
if
we
see
it
show
up
on
Amazon
and
we
have
enough
bandwidth
or
we
can
I
think
pretty
safely,
do
it
for
the
following
release:
yeah,
that's
the
very
beginning
of
the
release
cycle.
C
A
C
128
comes
out
the
beginning
of
December,
like
December,
9th
I,
believe
yeah
and
the
following
one
sort
of
in
the
second
or
third
week
of
January,
so
that
kind
of
gives
people
through
the
end
of
this
year,
with
the
the
current
version.
I.
Think
that's
that
will
give
people
plenty
of
time
for
1.24
to
have
soaked
on
the
cloud
providers
and
possibly
be
getting
closer
to
the
stable
Channel
by
the
time.
A
C
B
B
I
agree:
kubernetes
itself
is
definitely
moved
to
like
a
quarterly.
Is
it
quarterly
now.
C
I
think
it
was
quarterly
now
they're
doing
three
per
year
last
I
checked
three
per
year
because
that's
the
fourth
of
those
courses
like
super
compressed
because
yeah,
it's
like
all
of
the
holidays
at
the
end
of
the
year,
and
that
like
release
cycle,
was
always
kind
of
small
and
awkward,
and
they
had
to
pick
dates
really
carefully
yeah,
so
they've
reduced
it
to
I
think
three
per
year,
and
so
we
used
to
basically
go
every
other
version,
because
we
were
in
six
weeks,
Sunday
ordering
yeah
exactly
double
that
every
every
three
months.
C
C
C
There
are
certain
cases
where
staple
sets
sort
of
fit
the
use
case,
and
there
are
lots
of
cases
where
the
behavior
of
staple
sets
isn't
quite
what
you're
looking
for
I
know
like
I've,
definitely
heard
Mark
describe
this
really
well
multiple
times,
but
I
can't
find
anything
in
our
documentation
that
actually
describes
a
difference
and
why
you'd
pick
one
or
the
other.
C
So
this
is
mostly
a
reminder
that,
like
Mark,
should
write
this
up
since
he's
got
a
good
answer
to
this,
and
if
anybody
else
has
has
thought
through
this
or
especially,
if
anybody
else
is,
you
know,
writing
a
game
where
they've
chosen
to
use
staple
sets
like
it
would
be.
Awesome
like
I,
don't
want
to
say:
Agonist
is
not
always
the
right
solution
right.
There
are
times
where
staple
sets
are
better
and
I
think
we.
C
We
definitely
need
to
highlight
when
you'd
pick
each
one
and
make
sure
we're
steering
people
in
the
right
direction,
and
so
I
was
gonna
say,
especially
if
you're
actually
like
using
staple
sets
or
you're
using
both
staple
sets
and
a
goodness
providing.
You
know
if,
if
you're
your
gamer
texture
is
proprietary,
and
you
don't
want
to
share
exactly
what
you're
doing
providing
sort
of
like
the
guidance
that
you
would
give
to
somebody
else
when
to
choose
each
one
would
be
awesome.
C
So
I
think
we
can
take
a
first
cut
at
this
and
by
we
I
mean
Mark.
Can
I'll
keep
poking
him
to
do
that.
But
if
other
people
want
to
contribute
we'd
love
to
to
hear
from
people
who
are
who
are
making
this
choice
and
why
they're
doing
so
and
sort
of
feed
that
into
our
documentation
to
help
everybody
else.
C
Don't
worry
all
right,
so
the
next
one
I
took
on
here
was
another
issue.
That's
been
filed
since
our
last
meeting,
which
was
somebody
had
suggested
that
we
change
the
namespace
for
our
protobuf
for
allocation.
This
is
for
the
grpc
endpoint,
because
the
generated
code
for
the
current
protobuf
doesn't
make
it
clear,
like
if
you're
looking
at
your
own
code
and
you're,
linking
into
it
that
this
is
like
an
agonis
allocation
protobook
that
you're
talking
to
it's
just
called
allocation.
C
Not
it's
not
sort
of
namespace
with
a
gun
at
the
beginning
and
it'd
be
it'd,
be
you
know
clear
like
for
us.
It's
it's
not
a
problem,
because
all
of
our
code
is
a
goodness,
but
if
you're
linking
like
you
know
the
clients
into
your
other
code,
it
makes
it
really
obvious
that
this
is
like
a
Proto
that
relates
to
Agonist
and
relates
to
allocating
things
that
to
me
sounds
like
a
great
usability
Improvement.
C
My
concern
is
about
compatibility
and
what
happens
if
you
change
the
namespace
of
a
Proto
I
think
it
will
stop
working,
so
we
actually
I
think
have
to
do
a
reasonable
out
of
work
to
basically
expose
on
the
server
side,
both
the
Old
and
the
new
namespace
as
sort
of
two
drpc
endpoints
that
are
both
registered.
They
can.
They
can
sort
of
fall
into
the
same
code
right
because
the
underlying
you
know,
encoding
of
the
Proto,
would
remain
the
same.
C
C
We
want
to
make
sure
we
don't
break
people
which
is
the
biggest
concern
here,
and
so
we
have
to
figure
out
a
how
long
we
want
to
run
both
in
parallel
and
B
to
make
sure
we
actually
can
run
both
in
parallel
and
then
that
quarter
leads
into
the
question
of,
like
is
the
usability
Improvement
worth
that
effort
and
the
maintenance
burden
of
doing
that,
and
so
I
raised
those
questions
on
the
issue.
B
Only
other
thought
I
had
is
I
tried
to
find
the
documentation
for
the
tool
that
generates
the
code,
which
I
can't
find
because
reasons
I
was
wondering
if,
as
a
user,
you
could
change
the
package
for
protobuf
and
still
have
it
be
packed
like
Backward
Compatible
is
like
a
command
line
option
or
something
I,
don't
know.
If
that's
a
thing
I
was
like
that
would
seem
like
a
thing
that
would
possibly
be
in
there
yeah.
C
I
know
the
proto-authors
have
gone
to
Great
Lengths
to
allow
you
to
modify
the
protocol
over
time
while
retaining
compatibility
right
yeah
at
like
the
field,
and
you
know
like
the
resource
level
like
the
rpcs
and
the
fields
and
everything
I
don't
know.
If
they've
done
the
same
thing
for
packages
right
like
it
might
be
like
you
pick
your
package
and
you've
got
it
forever,
yeah,
so
yeah
we
can.
We
can
do
some
investigation
there.
Maybe
they've
come
out
come
across
ways
to
help
people
through
this
sort
of
transition.
B
Yeah,
there's
got
to
be:
there's
got
to
be
a
path
through
there,
but
I
was
also
wondering
like
if,
if
you
are
proto-seat,
that's
the
thing
I'm
looking
for
you
feel
like
oh
I,
depending
on
the
language
you're
using
as
well
like
it
like.
If
I'm
using
go,
I
can
generate
proto-seeker
I
can
generate
go
code
and
basically
put
it
in
whatever
directory
I
want
and
that
that
that
packages,
my
client
but
I,
don't
know
if
the
same
is
true
for
other
languages
or
maybe,
if
you're
trying
to
include
it.
C
Yeah,
like
the
generated
Java
code,
I
know
you
there's
an
option
like
a
job
option,
which
it
generates
a
specific
Java
package
right,
so
you're
kind
of
stuck
with
the
name
of
the
generated
code.
I
think
the
same
was
true
for
for
us
a
couple
other
languages
like
C,
sharp.
C
I
think
the
other
thing
this
this
reminds
me
is
that
if
we
are
adding
new
Proto
stuff
in
the
future,
we
should
probably
make
sure
we
bike
shed
sufficiently
about
the
name,
because
we're
going
to
potentially
be
stuck
with
it
forever.
C
I
I,
don't
recall
exactly
when
this
Proto
went
in.
It
was
maybe
before
I
started
working
on
the
project
and
I.
Don't
know
how
much
discussion
there
was
about
the
name
or
or
sort
of
considerations.
It
wasn't
impact
right
so,
but
especially
if
we
find
during
this
investigation
that
we're
stuck
with
it
forever,
we
should
make
sure
that
you
know
in
the
future.
We
are
cognizant
of
that.
A
A
C
All
right
so
I'm
gonna
move
on
to
the
next
one
Mark
Googling.
So
the
next
thing
was
one
of
our
friends
filed
a
bug
for
improved
documentation
for
people
that
are
using
IPv6
Mark.
You
responded
on
the
ticket
with
like
hey,
we
can
do
this
and
we
can
do
this
and
I
know
that
floating
around
in
various
people's
heads.
C
There
are
some
answers
here
and
we
do
have
a
little
bit
of
information
on
our
FAQ
about
using
the
new
feature
that
recently
was
promoted
stable
for
using
the
you
know,
DNS
name
of
a
node
as
part
of
your
allocation,
so
your
clients
can
connect,
and
you
can
use
I
think
that
with
IP,
it's
like
four
to
six
tunneling
to
reach
your
node.
C
It
would
be
I
think
nice
to
have
like
a
single
documentation,
page
talking
about
iqv6
connectivity
or
we
kind
of
discuss
the
different
options
right.
So
there's
different
ways
that
we've
seen
people
do
this
and
we
kind
of
put
those
all
in
one
place.
I
think
that'll
improve
Discovery,
but
I
think
also
they
are
not
all
written
down,
necessarily
right,
I
think
some
of
them
you
can
find.
C
If
you,
you
know,
maybe
search
your
tickets
or
search
through
slack
history,
some
of
them
are
documented
in
different
places
on
our
website,
but
it
would
be
nice
to
have
sort
of
one
page
for
people
that
are
are
looking
for.
Iqv6.
B
No
I
agree
there's
a
lot.
It
feels
like
there's
also
a
lot
of
confusion
about
what
it
means
to
be
IPv6
compatible
and,
and
a
lot
of
people
go.
Oh,
it
has
to
have
an
IPv6
address,
but
it
doesn't
it
I
mean
if
you're
looking
at
it
needs
to
be.
It
needs
to
work
on
the
on
the
IPv6
stack,
which
is
a
very
different
conversation.
C
Right,
yes,
and
that's
also
something
that
we
can
clarify
too
right,
even
if
your
cluster
doesn't
have
IPv6
addresses
it
can
be
connected
to
over
IPv6.
It
doesn't
mean
that
your
clients
can't
be
running
after
these
things.
Foreign
I
think
I'm
going
to
flip
the
next
two,
because
I
think
the
next
one
is
going
to
be
really
quick
and
then
we'll
I
wanted
to
save
enough
time.
Mark
for
you
to
go
through
your
issue,
so
I
did
want
to
mention.
C
There
was
a
ticket
filed
probably
about
a
year
ago
asking
about
running
I
got
us
autopilot,
which
we
closed.
This
won't
fix,
because
when
GE
autopilot
was
first
launched,
it
had
a
huge
number
of
restrictions
in
terms
of
all
the
things
you
couldn't
do
on
GK
autopilot,
including
quite
a
few
core
things
that
August
needed
like
validating
and
mutating
web
hooks
among
lots
of
other
things,
so
over
the
course
of
its
its
life.
C
The
last
major
blocker
is
host
ports,
which
is
currently
being
worked
on,
and
so
one
of
the
guys
on
our
team
is
starting
to
look
at
trying
to
close
the
rest
of
the
gaps
to
actually
allow
a
Gunners
to
run
on
gke
autopilot,
so
I
just
wanted
to.
Let
people
know
there's
a
tracking
issue
here.
If
that's
something
you're
interested
in
feel
free
to
follow
along.
C
All
right,
Mark
now
we'll
throw
it
to
you
for
I,
think
what'll
be
the
meaty
topic
of
the
day.
Oh.
B
Yeah
I
wasn't
I,
didn't
think
it
was
I
didn't
think
it
was
that
maybe
I
was
just
gonna
I
mean
it's
it's
media-ish,
depending
on
how
far
people
want
to
go
into
it,
but
basically
I
primarily
wanted
to
say
I.
Think
I
think
we've
gone
through
enough
iterations
on
it
that
we
have
a
pretty
stable
design.
So
just
to
do
a
quick
recap
as
well
arbitrary
counts
and
lists.
B
If
anyone's
played
with
the
player
tracking
stuff,
where
we
can
have
like
lists
of
players
and
how
many
do
we
have
at
a
capacity
for
players,
it
became
increasingly
obvious
that
having
essentially
counters
and
lists
of
items
were
incredibly
useful
for
a
variety
of
use.
Cases
not
just
player
tracking
and
people
were
hacking
player
tracking
to
do
other
things
other
than
track
players,
because
it
was
useful
for
other
stuff
so
wanted
to
sort
of.
Oh
god,
I've
been
feeling
pretentious.
B
I
wanted
to
elevate
that
to
a
higher
higher
functionality,
I've
been
watching
way
too
much
pretentious
cook,
Tick
Tock
and
build
out
we're
basically
replace
the
player
tracking
functionality
with
something
that
was
more
generic
and
and
could
handle
different
types
of
use.
Cases
I
also
commonly
refer
to
this.
As
how
many
ways
can
we
abuse
that
CD?
B
But,
yes
to
my
point,
I
think
I
think
it.
The
design
is
in
a
happy
place
that
I'm
pretty
happy
with
my
next
task
on
this
is
to
actually
write
like
a
checklist
of
individual
items
that
can
be
taken
on
independently,
so
we
can
sort
of
split
out
some
of
the
work
and
and
and
hopefully
work
on
it
as
a
group,
but
also
a
big
thanks
to
everyone
who
has
provided
feedback
and
there's
been
a
lot
of
it,
and
it's
been
great
I.
A
I
think
that's
probably
where
I
would
leave
that.
C
Yeah
I
think
I
consider
this
one
to
be
me
because
I
know
that
there's
there's
lots
and
lots
of
details
in
there.
There's
lots
of
discussions
that
have
happened
around
you
know
whether
things
should
have
duplicates
or
not
have
duplicates
how
this
works
with
the
auto
scaler
What
fields.
The
auto
scaler
is
looking
at
in
terms
of
you
know,
sizing
and
what
it
means
to
size
up
or
down,
and
how
we're
how
we're
inferring
that
so
there's
if
people
have
questions
about
those
things,
I
think
there's
lots
of
places.
C
We
can
go
pretty
deep
in
discussion
on
these
things.
If,
if
folks
are
ambivalent
or
have
sort
of
already
read
and
rocks
the
issue,
then
that's
cool
too.
B
A
B
Feel
free
to
read,
also
asynchronously,
write,
notes
and
stuff
like
that.
Yeah
I
need
to
I
need
to
break
this
down,
and
some
chunky
work
put
my
to-do
list
as
well,
but
I'm
excited
for
this
I
think
it's
gonna
be
really
useful,
we'll
make
sure
we'll
also
have
some
sort
of
guide
or
something
of
like
how
do
I
move
this
from
play,
if
I'm
using
player
tracking.
Now,
how
do
I
move
this
over
and
we'll
make
sure
that
we
do
that
in
a
considerate
way
for
people
who
are
using
player
tracking.
C
That's
a
good
point
and
I
guess
also
along
those
lines,
because
player
tracking
is
Alpha
I
think
the
current
plan
is
basically
to
do
a
knife
switch
of
One
release.
You
can
have
player
tracking
and
the
next
release
is
not
it's
going
to
have
this
instead,
which
will
be
similar.
You'll
have
the
same
functionality,
but
all
the
names
and
stuff
will
be
different,
so
that
will
be
a
case
where,
if
you
are
dependent
upon
the
current
player
tracking
apis,
that
you
will
not
have
sort
of
a
seamless
upgrade
between
icon
instructions,
foreign.
B
Yeah
we
can,
we
could
do
a
mild
overlap,
like
you
have
a
version
where
you
have
both
and
then
the
next
version.
It
goes
away,
that's
true,
actually
that
which
would
also
give
a
good
opportunity
for
people
to
to
migrate
and
provide
feedback
in
case
they're
like
oh,
this
I
did
this
thing
with
player
tracking
and
now
I
can't
do
it
anymore
for
whatever
reason,
but
yeah
no
one
started
work.
Yet
so
don't
panic.
C
Okay,
so
I'll
put
a
note:
I
think
this
kind
of
captures
what
we
both
said,
which
is
unlike
things
like
the
protobot
namespaces
for
for
grpc,
we're
not
planning
on
providing
either
a
seamless
transition
or
sort
of
a
long
maintenance
period.
C
B
C
B
B
A
All
right,
let's.
A
A
B
Because
I
think
this
is,
they
were
running
mini
Cube.
A
B
C
Close
that
I
mean
it's
it's
there.
We
can
always
reopen
it.
The
information's
all
there
it
just
will
be.
B
A
It's
necessary
possibly.
A
C
B
Think
the
answer
to
that
is
what
we're
gonna
do
is
like
run
it
on
like
an
actual
Windows
machine,
as
part
of
our
like
automated
testing,
to
confirm
that
it
actually
does
what
we
think
it
does,
because
there
are
different
paths
for
like
if
you're
on
Windows
versus,
if
you're
on,
like
Linux.
B
A
That
closes
all
the
stale
stuff,
yep
all
righty
issues,
sort
least
recently
updated.
Okay,.
C
The
first
one,
oh
so
that
was
from
our
friendly
C
plus
plus
expert,
who
is
saying
that
we
shouldn't
have
checked
in
generated
protophiles
for
protos,
that
we
don't
own.
A
B
C
A
I
would
agree
with
that.
Let's
take
that
off
foreign.
C
C
Namespace
is
we're
going
to
tell
a
gun,
is
to
manage
so
I
think
that
part's,
okay
and
the
second
part
is
we
have
to,
and
we
we
probably
already
do,
telegonus,
which
ones
to
look
at
or
maybe
not.
B
C
A
B
A
Since
we're
already
telling
runners
in
which
new
spaces
yeah,
it
has
that
Constitution.
B
Moderate
this
seems
very
topical
right
now
separate
deployment
for
gunners
allocator.
B
Yeah
I
think
this
is
yeah
here
we
go.
This
is
what
you
said.
The
crds
are
always
installed
no
conditionals,
because
you
need
yeah,
you
need
those.
Those
are
important,
I,
guess.
Here's
an
interesting
question:
if
you
had
a
gun,
is
installed
in
your
cluster.
C
C
Don't
know
who
made
this
one
it's
under
AWS
samples
and
it
has
a
nice
little
picture,
so
you
can
throw
this
up
on
your
screen
if
you
want
they're
doing
something
similar
except
they're,
actually
running
a
gke
cluster
with
just
the
allocator
service
and
using
multicolesterol
agent
from
that
to
do
multi-cluster
allocation
so
like
they
call
it
a
routing
cluster,
so
it
doesn't
run
any
game
servers.
It
doesn't
need
a
controller
in
it.
C
It
just
has
an
allocator
service
and
that
it
uses
MCA
to
talk
to
the
other
clusters
in
the
same
way
that,
like
the
cloud
run,
allocation
endpoint
can
talk
to
the
other
clusters
and
be
a
proxy
yeah.
So
this
I
think
would
be
the
use
case
for
doing
it.
If
you
wanted
to
set
this
sort
of
thing
up,
then
in
that
routing
cluster
you
need
the
crds.
You
need
the
allocator
service,
you
don't
you
need
the
agonis
controller
you're,
not
running
any
game,
servers
and
yeah
it
should.
C
It
should
work
right
because
if
you've
got
the
you're
going
to
crds,
then
you
can
use
that
to
sort
of
plumb
the
information
for
the
allocator
service
on
how
to
talk
to
it's.
It's
friendly
neighborhood
clusters
that
are
actually
running
game
servers,
yeah,
so
I
think
this
is
literally
the
use
case
for
it.
I
don't
know
if
this
was
filed
by
somebody
who
was
working
on
this
or
or
this
is
sort
of
an
independent
question,
but
this
is
where
you'd
use
it.
A
B
Don't
know:
okay,
okay,
that
makes
sense
all
right.
So
it
sounds
like
you
need
to
okay.
So
it's
what
was
I
gonna
say:
not
a
bad
thing
doesn't
need
to
be
stale
at
some
point.
Someone
could
do
it
if
we
wanted
to
correct
okay.
So
probably
just
leave
this
one
alone,
and
it
also
sounds
like
really
it's
about
having
an
option
not
to
install
the
controller.
B
I'm
trying
to
think
if
there's
anything
else
or
you
may
not
want
to
install
like
the
or
you
could
already
turn
off
like
the
Ping
service.
If
you
want.
C
A
C
B
C
Think
it's
a
it's
a
small
Helm
change,
yeah
yeah
I
mean
it
could
be
a
good
first
issue
if
this
is
somewhere,
something
that
that
people
want
to
do.
I
guess
again,
like
I
from
personal
perspective,
I'm
I'm,
not
gonna,
encourage
someone
to
do
this
because
it
doesn't
it's
not
something
I,
don't
know
when
I
think
about.
B
C
When
you
look
at
like
the
helm,
charts
themselves,
they
turn
into
this
templated
method,
like
you,
can't
actually
tell
what's
going
on
unless
you
have
like
a
very
deep
understanding
and
global
view
of
them,
which,
which
starts
to
become
really
ugly
over
time,
like
we're,
maybe
already
there,
and
we
should.
You
know,
think
about
like
what
that
means.
I
don't
know,
I
was
watching
a
video
that
Brian
Grant
had
created
about
a
year
ago
about
how
awful
these
things
were
made
me
think
and.
C
If
somebody
sent
a
PR
for
this
because
they
really
wanted
it
I
would
say:
that's
fine,
I'm,
not
sure
I
want
to
encourage
somebody
to
to
get
extra
complexity
to
the
helm
chart
when
we
don't
have
somebody
we
know
would
use
it
I.
Guess
that's
really
the
point.
If
somebody
says
I
need
this
I'm
going
to
use
it.
That's
different,
then,
let's
add
complexity
in
case
somebody
might
want
to
do
something
that
they
asked
about
two
years
ago,
yep.
C
B
To
do
like,
maybe
one
more
yeah,
this
is
a
fun
one.
I
I,
like
I
I
kind
of
want
to
keep
this
one.
So
this
is.
This
is
having
automation
such
that
like,
if
you
allocated
a
game
server
and
your
Matchmaker
didn't
actually
send
players
to
it,
or
maybe
all
the
players
happen
to
just
like
not
connect.
Is
there
an
automated
way
of
being
able
to
detect
that
and
respond
accordingly,
player
tracking
was
gonna
help
with
this.
C
This
is
also
something
that
game
servers
themselves
could
address
right.
They
can
say:
hey
I've,
been
allocated
start
a
five
minute
timer
if
nobody
connects
just
go
ahead
and
exit
or
something
right
soluble
from
the
game,
server
side
yeah.
So
the
question
is:
is
there
something
we
can
do
on
the
infrastructure
side
that
is
generally
applicable
sort
of
reusable
across
use
cases?
That
makes
it
so
you
don't
have
to
write
that
that
custom
code
in
every
game,
server,
yeah.
B
And
I
think
there's
also
some
useful
things
for
like
if
you're
thinking
about
Lobby
servers
and
stuff
like
that,
where,
if
it's
like,
oh
I,
haven't
had
players
for
the
last
15
minutes
so
like
kill
it
off
and
having
that
be
sort
of
generic
as
part
of
our
Council
list
stuff,
so
I
think
I
think.
B
Yes,
let's
keep
it
I
think,
but
I
think
it'll
end
up
getting
moved
once
we
have
like
lists
and
lists
and
accounts,
in
which
case
then
it
just
becomes
a
case
of
like
how
long
and
which
list
or
count
do
you
want
to
use
as
your
I
have
players
right.
Let
me
actually
make
a
note
of
that.
C
That's
like
Cyril
back
in
2019
said
the
same
thing:
I
did,
which
was
game
server,
could
just
shut
down
yeah.
B
C
Yeah
somebody
said
this
issue
can
occur
in
a
Rite
of
games.
It'd
be
nice.
To
be
honest,
would
do
it
right,
that's
kind
of
what
we're
saying
is.
Yes,
you
can
do
it
yourself,
but
it
would
be
great
if
a
bonus
could
provide
a
way
to
do
this,
for
you
automatically
through
a
simple
configuration,
yeah.
B
It
has
not
come
up
recently,
actually
I'm
almost
tempted
to
Market
as
stale
as
like
a
non.
A
gun
is
problem,
but
a
kubernetes
problem
and
sort
of
be
like.
Let's
shove,
this
Upstream.
C
Yeah
the
last
comment
I
put
on
here
was
a
while
ago
and
the
link
is
broken,
which
is
always
fun.
A
C
B
B
It
does
also
feel
like
I
know,
a
lot
of
people
who
are
just
like
I'm,
just
gonna
make
a
big
Docker
image
or
or
they're
gonna,
be
like
really
sophisticated
about
it
and
sort
of
do
like
chunked
asset
management
and
like
diff
changes
and
like
that
kind
of
stuff.
B
So
stay
learned
unstable.
How
do
we
feel.
C
I
think
we
should
Market
a
stale
I
think
I
found
the
new
cap.
It's
called
volume
populators.
A
C
Okay,
volume
populators
will
be
implemented
as
out
of
tree
controllers.
Entry
might
do
some
validation,
so
it
sounds
like
kubernetes
itself
is
saying.
This
is
not
going
to
be
solved
in
tree
as
part
of
your
kubernetes
installation,
but
there
might
be
other
tools
you
can
install
in
your
cluster
that
will
Auto
populate
volumes
for
you.
B
C
A
B
We'll
do
we'll
do
the
outro
well,
thank
you,
everyone
for
joining
us
yet
again
for
another
episode
of
the
agones
community
podcast,
it's
delight
and
a
pleasure
to
hear
from
you
all
and
we
appreciate
all
the
listeners
that
have
tuned
in
both
live
and
listening
to
the
recording.
You
can
see
us
again
next
week
next
week
next
month,
where
we'll
be
joining
once
more,
you
can
follow
us
at
Agonist,
Dev
on
Twitter
or
go
to
agonist.dev
on
the
website
and
see
all
the
community
contributions
that
you
can
make.
It's
see
you
next
time.