►
From YouTube: Agones Community Call May 2020
Description
No description was provided for this meeting.
If this is YOUR meeting, an easy way to fix this is to add a description to your video, wherever mtngs.io found it (probably YouTube).
A
A
A
C
Root
so
one
point:
six
point:
Sarah
went
out,
I
think
looking
at
everyone's
faces
here,
we're.
D
C
C
C
Lots
and
lots
and
lots
of
enhancements
a
big
one
for
me-
is
all
the
player
tracking
stuff.
So
that's
under
an
alpha
feature
flag.
So
if
you
want
to
track
which
players
are
connected
or
disconnected
or
how
many
players
you
have
or
any
of
that
kind
of
stuff,
please
enable
that
feature
and
take
it
for
a
test
run.
We
definitely
love
feedback,
see
if
there
are
any
bugs
any
performance
issues.
C
Anything
like
that,
just
to
see
how
it
how
it
works
and
that
it
does
the
things
that
we
hope
that
it
does
note
kubernetes
115,
so
we
moved
everything
up
to
115
now.
You
will
also
note,
as
of
the
last
request,
we've
changed.
I
think
was
it
this
one
was
the
last
one
I
think
it
was
the
last
one.
We
changed
it
so
that
the
N
minus
1
for
the
kubernetes
version
was
only
for
the
majority
of
kubernetes
versions,
supported
across
all
cloud
providers.
C
D
Just
to
talk
about
upgrades
we've
been
on
yeah
I
came
this
recently
of
every
other
release
of
ours,
we'll
pick
up
a
new
crew,
Nettie's
version
which
has
made,
which
makes
sense,
because
are
we
released
exactly
twice
as
fast
already
so
that
means
we're
now
keeping
up
I.
Think
the
next
Q
Bernette
release
got
delayed
due
to
the
Kobe
situation
and
they're
slowing
down
their
release.
D
Cadence,
so
I
think
we
may
not
keep
that
cadence
up
so
and
there
may
be
a
couple
of
releases
before
he
moved
to
1.16
yeah,
but
I
think
you
know
if
things
were
we're
all
normal,
then
that
feels
like
about
the
right
speed
at
which
we
should
upgrade
to
stay
sort
of
current.
If
you
will,
with
kubernetes
releases
I
mean.
C
I
also
like
says
to
your
point
previously
Robert
in
previous
meetings,
this
this
also
means
that
we're
not
just
skirting
that
edge
of
like
supported
kubernetes
versions
on
platforms,
especially
especially
on
GA
kits
Google
project.
But
so
that's,
that's
a
that
makes
me
feel
home
where
it's
not
more
comfortable,
I.
Think.
D
C
C
F
C
You'll
note
I,
don't
know
if
it's
at
all
interest,
but
the
the
Alpha
and
if
you
haven't
looked
the
Alpha
and
Beta
SDKs
have
their
own
produce
and
swag
O's,
so
they're
actually
completely
separate.
You
can
for
pretty
much
all
the
languages.
You
can
point
that
there
they're
all
served
to
the
same
API
endpoint
but
they're
just
have
separate
separate
proto's,
so
you
have
to
generate
them
separately.
There's
an
example
of
that
in
the
Indigo
stuff
and
the
swagger
stuff.
If
you
need
to
do
that.
D
Don't
want
me
writing
JavaScript
one
one
thought
about
that
market
is
Newton,
we
have
lots
of
different
SDKs,
many
of
which
are,
you
know,
primarily
not
maintained
by
you.
I
mean
are
sort
of
shorter
on
the
community
going
for
is
we're.
Implementing
new
features
like
this
with
player
tracking
or
we
are
SDK
I
think
it
would
be
nice
if
we
had
a
way
to
sort
of.
D
D
That
you
were
pushing
for
that
like
so
they
could
also,
like
maybe
slide
their
PRS
and
beforehand
instead
of
hearing
at
this
meeting
like
by
their
release,
but
not
in
your
SDK
yep
I,
don't
know
if
we
have
a
good
way
to
capture,
like
you
know,
like
kunai's
uses
the
owners
files
right
work
with
prowl,
which
we're
not
completely
using
it,
but
if
we
sort
of
captured
like
here
so
the
primary
maintainer
is
for
each
SDK
that
we
could
sort
of
you
know
ping
those
people
or
create
issues
to
let
those
people
know
that
a
new
feature
is
gaming.
D
D
Know
I
yeah,
I,
guess
hey,
you
know
we
should
add
the
player
tracking
support
to
the
node
SDK.
You
know
CMC
Steven
like
if
you
have
time
like,
and
this
is
something
you're
interested
in
you
know.
This
is
something
you
could
be.
You
should
be
aware
of
a
head
of
time.
Instead
of
sort
of
that's
a
good
idea.
That's.
C
D
Yeah
and
I
don't
know,
process
wise,
it's
easier
to
create,
like
one
master
tracking
ticket
with
a
bunch
of
check
boxes
yeah,
so
we
can
track
in
one
place
like
here's,
where
it
is
implemented.
Here's
where
it's
missing,
because
you
know,
as
we
also
talked
about
graduation
to
GA
and
beta
like
it
seems
like
we
shouldn't
graduate
something
to
GA,
certainly
without
supporting
all
the
SDKs
awesome
video,
it
would
be
an
open
question
of
like.
Is
it
a
majority?
Is
it
all
of
them
implemented
before
we
graduate
to
beta
I?
Think
alpha?
C
Yeah
that
was
sort
of
a
question
I
was
gonna.
Ask
you
on
the
player
tracking
stuff
like
at
what
point.
Do
we
close
that
ticket
as
done
and
I?
Think
that's
a
that's,
maybe
maybe
we
also
put
a
note
on
that
ticket.
Saying
hey
it's
released
as
alpha
and
then
we
can
also
put
a
note
on
saying:
hey
it's
now,
it's
released
as
beta
and
like
so.
We
can
track
that
through
I.
Like
that
idea,
a
lot.
C
Reason
I
asked
that
more
than
it
is
like
I
kind
of
have
a
mental
map
of
like
who
does
what
generally
like
I
know.
Stephen
usually
does
the
nodejs
stuff
and
whatnot,
but,
like
my
memories,
only
so
good
I,
don't
know
if
we
want
I,
don't
know
if
we
want
to
go
down
that
path,
I'm,
just
thinking
more
for
like
if
one
of
us
is
working
on
a
feature
and
like
they
want
to
go
look
up
who
those
people
are
probably
useful.
D
D
As
as
owners
for
nodejs,
both
are
the
examples
and
for
the
SDK
I
think
those
are
the
only
two
dive
owner,
so
a
couple
of
them
have
owners,
but
the
other
ones
it
would
be.
It
would
be
nice,
even
if
those
owners
are
the
same
people
that
are
the
top
level
owners
files
just
so
it
was
obvious
like
who
the
primary
maintainer
is
for
yeah.
C
D
Are
those
are
pretty
big
in
different
parts
of
the
code
tree
right
like
if
you
think
about
like
there's
a
whole
bunch
of
like
the
controllers
and
everything
where
you
say
like?
Okay,
like
I'm,
a
project
maintainer
like
I
kind
of
all
that
stuff?
But
then
those
always
like
different
languages
and
tiers.
D
D
One
yeah
so
I
stuck
this
on
here,
Eric,
could
actually
sort
of
paying
this
this
ticket
a
couple
of
days
ago
and
I
thought
it
would
be
a
good
forum
to
discuss
it
quite
a
while
back
I
created
a
ticket
to
create
a
proposal
for
creating
the
unites
operator
so
right
now
sort
of
the
lifecycle
management
for
it
donates.
Is
you
install
via
hell
or
animal,
and
then
mark
wrote
up
some
upgrade
instructions
about
how
you
get
it
from
version
a
to
version?
D
Eight
plus
one-
and
you
know
you
can
upgrade
you
can't
sort
of
delete
everything
manually
and
reapply
install
but
ya
know
in
either
case
there
is
the
recommended
process
of
draining
your
cluster
from
active
or
clothes
so
effectively.
If
you
look
at
sort
of
the
manual
steps
and
the
upgrade
documentation,
one
question
would
be
is
whether
we
should
try
to
automate
that
with
something
like
an
operator
right.
D
So
you
take
those
steps
you
encode
them
in
encode
and
software,
and
then
that
sort
of
takes
the
uncertainty
away
from
people
and
it
becomes
something
that
you
can
test,
and
so
that
was
sort
of
the
idea
behind
operator
was
like.
Could
we
provide
some
software
that
would
help
people
for
the
upgrade
process?
Now,
if
you
look
at
the,
if
you
step
back
and
look
at
the
upgrade
process,
the
recommended
upgrade
process
is
actually
like.
D
Creating
new
crust
cluster,
it
already
green
deployment
move
to
the
new
cluster
and
part
of
the
reason
for
that
is
that,
then
you
aren't
having
to
deal
with
in-place
upgrades,
which
can
be
rather
tricky
and
if
you're
doing
in-place,
upgrades
and
part
of
that
procedure
is
to
drain
all
of
your
workload
anyway,
like
unless
you
have
planned
downtime
for
your
game,
which
a
lot
of
modern
games
are
trying
to
avoid.
Then
you
have
to
have
a
second
cluster
anyway.
D
So
is
it
just
as
easy
to
have
a
second
cluster
and
do
a
red,
green
and
move
over
as
it
is
to
have
a
second
cluster
drain,
the
first
cluster
and
under
in
the
first
cluster?
That's
kind
of
an
open
question,
but
so
we
kind
of
let
this
issue
on
hold
is
that
that
was
the
mandatory
procedure
as
the
product
matures
I
think
it
becomes
maybe
more
interesting
to
go
back
and
talk
about
what
in-place
upgrade
might
look
like,
possibly
without
having
to
drain
your
cluster
and
I.
D
Think
that's
where
the
operator
would
be
so
so
my
initial
sort
of
thinking
by
an
operator
was
maybe
we
get
rid
of
helm
and
noose
operator
for
installation
and
upgrades.
Another
idea
which
Eric
put
in
the
ticket
was
having
the
plate
a
wrap,
helm
and
used
to
be
like
the
helm
operator
so
that
we
keep
helm.
So
people
want
to
someone
can
do
that
an
operator
sort
of
automates
the
things
you
would
do
at
home.
I
think.
The
one
thing
I'd
like
to
hear
from
folks
here
is:
do
people
like
helm?
D
Do
people
want
to
keep
using?
How
do
people
think
that
in-place
upgrades
or
something
it's
important,
or
are
people
okay
or
comfortable
with
the
use
of
new
cluster
sort
about
great
strategy?
So
Eric
I
don't
know
if
you
want
to
say
anything
also
since
this
this
piqued
your
interest
recently
and
and
sort
of
where
you
were
coming
from
aye,
sir.
C
G
I
cannot
answer
that
question
as
far
as
wrapping
the
using
the
elm
freighter
I
think
it
works.
Okay
for
a
minor
version,
great
I'd,
I,
don't
know
enough
about
operators
if
grading
a
major
version
is
something
that
is
supposed
to
be
supported.
If
we
want
to
support
that
use
case,
the
current
operator
is
the
KLM
wrapping
wouldn't
work,
and
it
also
doesn't
deal
with
like
in
place
migration
of
CRD.
So
if
we
need
to
do
that,
I
know
see.
G
C
C
C
D
Yeah
so
when
I
first
found
the
ticket
that's
I
was
trying
to
get,
was
I
thought
that
the
main
pain
points
were
installation
and
up
and
intercept
phrase,
yeah
and
I.
Think
having
worked
on
the
project
for
a
while
now,
like
I'm,
not
hearing
a
lot
of
people
complaining
about
installation
right,
it
seems
like
it's
all.
I
am
Allah
is
like
the
quick
and
easy
get
stuff
up
and
running
installation,
and
if
you
wanna
have
a
little
bit
more
controlled
and
helps
you
to
work.
Okay,
it
seems
working.
D
Having
took
to
that,
and
so
that's
kinda
what
I
was
trying
to
start
with
was
it
felt
like
there
was
sort
of
two
critical
user
journeys
that
identified
initially
that
the
operator
patent
seems
to
solve
and
I'm,
not
sure
either
of
those
are
necessarily
critical
to
this
customer.
This
audience
right,
that's
what
I
was
trying
to
get
is
like
are
those
things
that
people
are
having
pain
points
with
right
now
or
are
those
things
that
were
sort
of
sighs,
dinging
through
process
and
best
practices
where
it's
not
important
to
code.
Those
things.
F
Yes,
there's
we
do
have,
we
do
maintain
two
clusters
and
we
will
drain
drain
one
cluster
down
to
zero
to
do
our
again
as
updates,
but
yeah
we've
considered
the
more
like
the
blue-green
style
and
system
as
well.
Just
so
very
you
know,
a
temporary
second
cluster,
yeah
I
think
on
the
surface
seems
like
in-place
upgrades
to
be
valuable.
But
then
there
is
this
kind
of
the
other
kubernetes
up
upgrade
thing
as
well.
You've
got
to
make
sure
the
game
will
still
stay
up
while
that's
happening.
So
it's
yes
and
it's
all
about
uptime
uptime.
G
D
Sort
of
adjust
the
number
of
nodes
in
each
cluster
so
that
your
total
capacity
is
approximately
peak
capacity
plus
a
little
bit
of
surge
capacity
during
the
migration
and
not
you
know,
have
to
X
your
sort
of
number
of
CPUs
or
spend
during
the
upgrade
process
right
and
given
that
ago,
nice
really
since
every
six
weeks,
this
upgrade
process
is
gonna,
be
something
that's
gonna
happen
frequently
and
so,
like
I,
think
that's
the
other
idea.
The
operator
is.
D
If
this
is
happening,
all
the
time
is
it
it's
something
we're
gonna
want
or
people
are
gonna
want
to
audit
to
some
degree
right.
So
the
question
is
like:
at
what
level
do
we
provide
automation?
Because
we
don't
want
people
to
stop
upgrading,
because
it's
too
hard
right?
That's
the
most
important
thing
is,
you
know
we're
not
back
porting
a
lot
of
bug
fixes.
We
very
could
really
go
back
and
release
patch
versions
on
old
releases.
We
were
asking
more
kissing,
stuff
ahead
and
asking
people
to
upgrade.
So
we
want
people
to
upgrade.
E
C
One
of
the
things
I
would
like
to
tackle
soon
is
probably
upgrading
us
to
helm
three.
It
seems
like
everything
works
on
home
three
but
I'd
like
to
move
like
all
our
developer
tools
and
make
that
a
tested
path
and
write
docs
for
that,
because
I
think
I
think
minorly,
there's
a
small
like
command-line
change.
I,
don't
think
anything
else
really
changes
for
that.
C
Actually
did
some
digging
into
some
of
the
command-line
tools
actually
I.
Think
that's
really
cool
I'm,
just
cuz
a
user
experience.
A
lot
of
those
are
actually
just
wrapping,
helm,
libraries
and
using
helm
behind
the
scenes
to
do
that.
Work
which
I
think
is
quite
cool
because
it
handles
a
lot
of
upgrade
experience,
type
stuff
that
you
don't
have
to
do
yourself
necessarily,
but
yeah
I
figured
that
was
possibly
a
good
first
step.
G
That's
how
we
do
install
I
go
nice
on
custard
on
our
side,
we
have
a
control
plane
that
uses
L
memories
to
install
agonize
on
communities
cluster
as
well.
The
downside
to
that
I
would
say
is
if
installation
fails,
or
you
know
times
out,
it's
it's
hard
to
pick
up
again
where
it
was
with
an
operator
and
the
reconciliation
joke.
I
mean
it.
You
know
we
can
patch
stuff
and
fix
it,
and
then
it
will
end
up
installing.
G
D
G
C
E
C
Be
I
would
I
would
definitely
also
love
just
like
if
you're
getting
like
regular
timeouts
on
Helmand
stalls,
I'd
love
to
see
a
like
all
the
steps
that
you're
doing
to
do
that
and
see.
If
we
can
replicate
that
in
some
way,
maybe
there's
a
dock
stakes
or
maybe
we
can
like
if
switching
tooling
or
something
I,
don't
know
what
the
right
answer
is.
We
look
to
look
into
that.
H
I
C
D
J
Yeah,
the
other
is
a
problem
when
we
have
ronk
bottom
plates
back
and
then
it
becomes
infinitely
creating
a
new
game
servers
and
it
realize
it.
It
scanners,
great
board
for
game
servers
and
each
create
fleet
and
healthy
game
server,
and
then
do
this
again
and
an
idea
behind
this
ticket
can
we
add
some
kind
of
feature
gate
or
two
mm
have
exponential
back-off.
All
these
great
in
new
game
servers
I
mean,
for
example,
Z
sure
was
that
normal
fleet
could
not
be
handled
well.
J
C
C
J
C
J
C
J
C
J
C
D
J
Yeah
yeah
actually
currently
it's
or
started
automatically,
and
this
is
the
reason
in
the,
for
example,
you
cannot
get
events
or
even
in
simple
UDP.
For
some
reason.
I
am
not
sure
why
it
took
more
than
30
seconds
to
connect
and
then
for
other,
as
a
normal
fleet
now
I
described
some
various
best
scenarios,
which
could
happen
so
I
can
replicate
it.
C
J
C
C
That
makes
sense,
I
think
if,
if
we
are
moving
a
game
server
to
error,
then
we
probably
shouldn't
restart
it,
because
that
is
that
is
meant
to
be
like
a
catastrophic
state
like
something
horrible
has
gone
wrong
and
it's
not
going
to
recover,
so
we
should
leave.
We
should
probably
actually
just
leave
that
alone,
so
people
can
see
it
I
think
that's
a
reasonable
bug
fix
and
would
make
make
things
good
for
people.
Okay,.
J
C
J
B
So
basically,
what
it's
starting
here
is
that
if,
during
the
deployment
of
a
fleet
the
bottom
plate
is
invalid,
I
can
we
can
use
the
the
validation
package
from
kubernetes
to
validate
the
whole
pods
back
so
basically
can
check
every
single
field
that
could
be
presented
in
the
port
template
and
in
case
this
is
not
valuable.
You
could
just
return
back
to
the
I,
don't
know.
B
Maybe
the
controller
saying
that
this
is
not
the
value
template
or
whatever,
so
this
fleet
will
net
would
never
be
created
actually,
but,
as
you
can
see,
that
brings
quite
a
lot
of
the
extra
dependencies
and
I'm
not
sure.
If
that's
the
right
way,
the
PR
is
not
full.
Yet
it's
not
completed
yet
I'm,
just
demonstrating
how
that
could
be
done,
but
I
think
that
the
validate
function
could
be
part
of
the
admission
controller-
maybe
just
not
here
yeah.
B
C
C
B
C
C
C
C
B
C
I
guess
the
only
question
I
have
is:
is
there
concerned
that
there
are
like
if
somebody
takes,
this
version
wants
to
run
it
on
an
unsupported
version
and
they
want
to
use
a
pod
feature
that
is
new
in
that
version?
Suddenly
it
won't.
It
probably
won't
validate
that
new
thing
in
that
pod
I
guess
we
don't
really
care,
because
we
only
support
a
given
version.
Any
point
in
time.
That's
my
only
thought,
I
thought
of
that
I
think
this
is
awesome
right.
B
D
B
J
B
When
I
was
investigating
this
issue,
I
saw
many
different
people
doing
pretty
much
the
same
with
the
replace
block
from
different
sort
of
projects
and
I
don't
see.
I
haven't
seen
anyone
from
kubernetes
complaining
about
using
the
version
or
the
package
straight
from
the
kubernetes.
They
what
they
say
is
that's
not
it's
not
suggested
to
use
these
packets
externally
they
were
pretty
much
conceived
it
for
internal
use,
but
in
terms
of
cop
I,
don't
copy
rights
or
something
like
that.
If
you
I,
don't
I,
don't
know
exactly.
C
C
Yeah
I
wonder
if
we
can
I
mean
look,
I
think
was
I,
gonna,
say
I
think
we
can.
We
can
work
on
tiding
that
up
and
like
that's
fine
I,
think
that's
that's
a
solvable
problem.
This
is
cool
I'd
like
to
I.
Think
probably
also
like
we'd
be
good
to
do
this
at
the
fleet
level,
as
well
as
the
like
it.
All.
The
levels
are
probably
actually
more
at
the
fleet
and
game
server,
because
if
I
create
a
fleet-
and
then
it
validates
at
that
level
before
it
starts
game
servers
that
would
be
awesome.
B
C
You
have
that
make
sense
that
make
something
right
or
each
other
thought
I
had.
It
probably
makes
sense
to
I
would
say,
make
this
an
alpha
feature
with
a
feature
gate
just
because
I
don't
want
to
put
out
a
thing
where
suddenly
people
can't
roll
out
their
versions
anymore,
because
something
validates
and
doesn't
work
but
outside
yeah.
What
I
was
trying
to
cover.
C
B
Was
a
yeah
and
I
was
adding
the
apply
default
so
in
case
folks
are
not
providing
at
least
the
minimum.
Spec
icon
assist
its
adding
the
at
least
the
minimum
required
fields
for
the
validation
to
pass
so
yeah,
so
I
figure
out
which,
which
are
the
fields
that
required
for
the
validation
to
pass
the
minimum
state.
And
if
it's
not
part
of
the
manifest
we
can
just
apply
the
false
see
yeah
pretty
much
it.
Thankfully,.
J
B
H
D
Quick
suggestion
for
your
PR
Daniel,
if
you
can't
squash
your
commits
into
like
one
commit
of
dependencies
and
one
commit
of
the
things
you
actually
want
people
to
take
a
really
close
look
at
maybe
a
lot
you
get
to
review
then
either
having
124
files,
where
I
have
to
figure
out,
what's
sure
vendored
automatically
what
I
should
look
at
or
having
14
commits.
As
you
are
sort
of
working
for
the
process,
that's.
B
J
F
Yeah
somebody's
came
out
of
this,
the
slack
Channel,
but
it
seemed
like
there
were
cells
as
well
with.
There
are
multiple
ways
to
allocate
a
game
server.
Actually
looking
at
the
documentation,
there's
is
a
preferred
way,
which
is
the
game
server
allocation
to
create
that
resource.
But
perhaps
there
is
some
clarification
on
why
we
can
do
multiple
ways
and
is
there
a
downside
anyway
and
also
press
as
a
follow-up?
Should
we
expand
on
the
documentation
around
this
I.
C
I
One
of
the
ways
through
the
API
supported
API
server
expansion
and
that's
the
original
way.
It
basically
uses
the
OCONUS
controller,
which
is
on
this
one
part
to
do
allocation,
meaning
that
data
plane
and
control
a
nice
mix
into
the
one
solution.
There
is
an
gunas
allocator
service
which
basically
extends
introducing
service.
These
scales
are
separate
from
the
control.
That's
basically
a
data
plane
for
allocation
and
the
API
is
table.
I
So
we
can
document
and
clarify
if,
for
example,
you
to
allocate
from
external
to
the
cluster
or
if
you
want
to
scale
up
when
you
have
higher
now
you
know
higher
traffic
for
allocating
nameservers,
then
the
service,
the
alla
khair
service,
is
the
preferred
solution.
I'm,
the
dickerson,
sorry,
for
the
background
noise,
and
if
you
have
your
matchmaker,
for
example,
on
the
same
cluster
and
your
socialist,
doesn't
need
to
scale
up
for
allocating
game
servers,
then,
basically,
you
can
use
the
IQ
extension
server
on
the
same
classroom.
No.
C
The
only
thing
I'll
add
to
that
is
like
the
logic
behind
game
server,
allocation,
CRD
and
the
logic
behind
the
GOP
cm
point
are
exactly
the
same.
So
that's
totally
fine,
it's
just
a
question
of
which
one
you
want
to
use.
There's
still
the
SDK
allocate.
If
you
need
that
for
specifically
allocating
a
specific
game
server,
that's
not
going
anywhere
I
do
yeah
and
to
that
thread
on
slack
like
I,
want
to
get
rid
of
that
go
sample.
That's
in
there
I
think
it's
gonna
quickly,
start
even
of
like
writing.
I
C
Yeah
there's
some
weird
stuff
in
it.
Yeah,
if
you
need,
is
if
you
need
a
service,
we
have
one
for
you.
So
if
you
want
to
like
a
G,
RPC
or
actually
a
GOP
cm
point
which
I
think
we
were
talking
about
making
a
rest
on
point
as
well
at
some
points,
if
you
need
a
service,
we
have
one.
So
you
can
use
that
if
you
want
to
use
the
Zod,
that's
also
totally
fine,
and
if
you
need
the
SDK,
that's
cool,
but
yeah
we
can
put
some.
We
can
put
some
more
words
around
that.
F
C
So
we
cover
some
of
that
in
the
there's,
a
there's,
a
doc
that
talks
about
like
scheduling
and
like
that
kind
of
stuff
and
like
that's,
why
we
talk
about
for
free
I
from
anacondas
point
of
view.
We
will
always
prefer
if
you
use
the
our
allocation
service,
so
either
the
GCM
point
or
the
game
server
allocation
CRD,
because
we
do
a
lot
of
work
to
pack
stuff
together
and
it's
gonna.
Do
it
like
you're,
not
gonna,
have
that
Swiss
cheese
problem
nearly
as
much
because
we
control
that.
So
we
recommend
that.
C
F
Yeah
yeah,
that's
really
useful,
I!
Think
as
well
like
it's,
it's
clear:
how
to
use
the
SDKs
and
on
the
game,
server,
side,
I!
Guess
it's
a
wrap
when
it's
around
this
from
the
matchmaker
I
mean
the
other
one.
If
Rosie's,
like
you
know
listing
listing
the
game,
servers
and
I
think
there
is
some
information
about
this
as
well,
but
Russ
we
could
add
more
value
by
thinking
like
yeah.
What
can
we
provide
on
the
matchmaker
side
from
outside?
Do.
F
In
our
case,
I
guess
we
might
have
multiple
games
running
and
they
don't
like
for
them,
and
the
games
never
ends.
So
we
want
to
write
place
players
onto
a
particular
ready
server
where,
but
even
there
really,
we
just
want
to
prioritize
for
scheduling
purposes
in
our
case,
we're
prioritizing
the
newest
one,
but
you
know
that
might
not
be
right
yep.
F
C
This
also
sounds
like
it
leads
into
some
of
this
stuff
we've
discussed
in
tickets
around.
How
do
we
like
player
tracking
leads
into
this,
but
like
how
like,
if
you
were
like
I,
want
a
game
server
that
has
room
for
three
people
and
then
we'll
do
the
prioritization
for
you,
so
that
we
pack
things
as
tightly
as
possible.
There's
that
would
that
sort
of
solve
a
problem
and
the
problem
in
a
different
way?
Yes,.
F
C
We
literally
put
that
on
the
Eric
Garner's
project,
which
is
pushing
game
server
events
into
pub/sub,
so
you
could
actually
have
a
centralized
place,
so
you
can
use
pub/sub
to
track
what
game
servers
you
have
and
what
data
they
have
on
them
from
your
matchmaker
by
tracking
your
pub/sub
events,
which
is
I,
think
kind
of
awesome,
actually
so
Daniel.
If
you
weren't
about
that
at
all,
it's
awesome,
I,
just.
B
Want
to
add
one
extra
thing
that
another
project
I've
been
working
recently
using
that
same
pub/sub
strategy,
but
it's
about
exposing
all
the
game.
Servers
via
a
REST,
API
and
I
can
also
post
the
link
on
the
shaft
cool,
it's
really
early
stages,
but
it's
basically
using
the
broadcaster
core
functionalities
and
exposing
game
servers,
fields
or
States.
Everything
that
you
can
add
to
like
a
a
more
restful
way
so
basically
can
consume
these
endpoints
using
any
sort
of
it.
Rest
client,
yeah,
it's
really
early
stages,
but
it's
it's
quite
functional
right
now.
B
So
it's
basically
listening
to
events
from
Ghana's
controller
or
actually
Cooper
needs
API
and
push
into
a
broadcaster
which
is
implementing
the
HTTP
back-end,
but
that
could
kind
of
so
could
kind
of
help
on
this.
Finding
server
states,
clear
labels
and,
in
future
I'm
planning
to
added
the
graph
QL
endpoint,
so
folks
could
filter
and
whatever
they
want
in.
B
I
C
Possibly,
what
I
think
in
my
ideal
world
and
I
work
fit
for
all
scenarios
I'd
like
to
see
the
allocation
system
have
enough
flexibility
to
handle
a
lot
of
different
matchmaking
scenarios
such
that
you
don't
need
to
list
game
servers
across
all
the
things
you
just
assume
that
a
goddess
is
going
to
do
the
right
thing
for
you,
that's
not
there
yet.
So
this
is
a
this.
Is
a
nice
feature?
C
That's
that's
what
I'd
like
to
see
I'd
like
to
you
just
be
like
give
me
a
game
server
with
these
requirements,
because
I
have
this
type
of
game
and
it
just
goes.
Here's
the
best
one
for
your
scenario.
That
being
said
in
the
short
term,
this
is
also
useful
and
not
a
bad
thing.
Prune
a
were
you
thinking
like.
We
should
be
like
pull
this
into
core
or
something
is
that
sort
of
that's?
What
do
you
thought?
I.
I
I
C
C
Oop!
Five
minutes
left!
Oh
my
god,
I'll
drop,
an
extra
thought
there
is.
We
do
have
a
section
in
the
organize
ducts
called
third-party
libraries
and
tools.
So
if
anyone
who's
watching
this
or
is
here
who
has
third-party
tools
because
there
are
a
few
on
github
I've
done,
searches,
please
feel
free
to
add
them.
There,
I'm
gonna
not
put
other
people's
tools
on
there,
because
I
think
if
you
want
to
put
your
tool
on
that,
that's
want
your
your
approval.
Oh
my
god,
we've
got
a
let's
talk
about
today.
C
I
I
had
this
proposal
to
actually
disable
and
TLS
authentication.
There's
built-in
gonna
solitaire
service,
that's
partially
to
enable
scenarios
like
ingress
that
it
supports
the
tills,
authentication
and
easier
integration
with
cloud
solutions
like
ingress
in
the
cloud.
So
my
proposal
is
to
provide
a
flag
to
enable
disable
MPLS
and
then
provide
the
samples
like
an
ingress
and
how
you
can
actually
use
that
to
set
up
your
you
know:
we've
not
port,
set
up
your
English
ingress
allocation.
A
C
Only
part,
I
didn't
understand
and
I
put
a
comment
on
and
I
just
I
think
I'm
just
being
able
thick.
How
does
the
cross
cluster
communication
still
work
with
the
SSL
cert
stuff
with
it
like?
If
I
turn
off
em
tell
you
us?
Is
there
still
gonna
be
like
a
TLS
certain
that's
worked
work
that
does
the
cross
cluster
a
bit,
but
a
non
non
encrypted
version.
That's
exposed
like
I'd,
like
that's
the
bit.
That
confuse
me
outside
of
that
I
think
this
is
a
good
idea.
I
think.
I
For
example,
ingress
can
support
multiple
authentication
approaches,
so
if
you
want
to
set
up
them,
your
multi
cluster
allocation,
your
thiele
of
you,
should
basically
provide
an
MPLS
authentication
from
the
ingress
and
set
up
your
certificate
so
that
the
clusters
can
talk
to
each
other.
But
from
your
matchmaker
perspective,
if
you
want
to
use
a
different
authentication
solution,
then
you
can
basically
use
a
different
authentication
provided
by
ingress.
Okay,.
C
Okay,
so
what
was
I
gonna
say
so,
basically,
if
you
turn
off
MPLS,
then
it's
on
you
to
sort
out
how
you
want
to
do
your
encryption
between
clusters
or
not.
Have
it
at
all,
that's
entirely
up
to
you,
but
if
you
do
turn
it
on
then
like
will
we
have
steps
for
how
to
set
up
all
that
stuff.
Okay,
got
it
right.
I
This
is
basically
for
more
complicated
scenarios
that
we
want
to
control
your
own
traffic,
so
the
original
goal
for
alloc
air
service
is
simpler
as
much
as
possible
possible.
So
that's
it.
Ganas
installations
actually
get
installed,
but
I
think
would
be
valuable
and
I
don't
know
if
this
is
actually
a
requirement
by
the
customers,
but
it
would
be
valuable
to
actually
being
able
to
disabled
and
then
handle
traffic
yourself
right.
C
And
then
you
could
use
like
a
managed
service
or
something
if
you
want
to
seduce
that
that
doesn't
make
sense
to
me
cool,
sweet,
I
guess
the
only
thought
I
have
on
that
is.
It
would
probably
be
good
to
specify
either
specify
whether
MTL
SR
is
on
or
not,
and
what
service
type
you
want,
but
outside
of
that
I
don't
have
anyone
else.
So
I'll
stop
talking.
C
I
H
J
Getting
my
PR,
which
is
about
I,
didn't
want
multiply
end
to
end
test
into
a
pipeline,
it's
ads
for
me.
Four
minutes,
I
mean
into
every
cloud
build
in
the
know.
Currently
we
don't
test
on
stable.
We
test
on
future
gates
enabled
which
is
not
so
good.
Yeah
I
mean
probably
we
can
edit
this
additional
form,
eosin
cloud
build
and
the
later
do
it
in
pro
job
or
how
in
in
what
prayer
prayer
we
would
do
this
Robert.
D
Yeah,
the
idea
with
trowel
is
that
we'd
be
able
to
run
multiple
things
in
parallel
right.
You
can
add
your
web
hooks
or
build
it'll.
Do
multiple
tests
on
a
PR
or
multiple
things
after
PR
submitted,
we
haven't
quite
gotten
it
hooked
up
and
I.
Think
you
know,
Mark
was
something
I
was
like
once
we
get
one
thing
in
proud:
we
have
sort
of
an
example.
What
it
looks
like
it'll,
probably
be
really
easy
to
add.
D
More
I
will
have
that
sort
of
first
like
Trailblazer
config
in
there
to
get
the
first
like
test
up
and
running,
and
so
I
think
that's
if
you've
got
stuff
working
on
cloud
build
and
we
won
in
right
away.
That's
that's
gonna,
be
the
quickest
way
to
get
it
in,
even
if
adds
a
little
bit
of
extra
time,
because
we
have
to
do
think.
Serially
I
would
love
to
get
the
prow
stuff
that
I've
finished.
Just
it's
just
a
matter
of
time
and
priorities
right
and
listings
are
hard
to
come
by
these
things.
Okay,.