►
From YouTube: Agones Community Meeting February 2023
Description
No description was provided for this meeting.
If this is YOUR meeting, an easy way to fix this is to add a description to your video, wherever mtngs.io found it (probably YouTube).
B
Welcome
to
another
episode
of
the
gone
as
community
meeting.
Thank
you
very
much
for
joining
us.
I
threw
some
stuff
on
the
ticket
Stephen.
We
can.
You
can
move
that
up.
If
you
want
you
don't
have
to
be
at
the
bottom.
I
was
just
throwing
stuff
there.
You
were
okay,
where
you
were
don't
nobody
puts
baby
in
a
corner.
Okay,
I.
B
I
was
gonna
mention
that
next
week's
meeting
is
during
GDC,
but
I,
don't
know,
I,
think
out
of
this
crew.
I
might
be
the
only
one
going
in
which
case.
If
somebody
else
wants
to
run
it
I
can't
remember
who's
actually
going
to
be
an
sf4
out
of
out
of
this
crew.
B
Do
you
want
to
run
the
community
meeting
next
week
since
I
volunteer?
You.
B
B
C
Not
that
I've
got
a
timing
issue
I've
met
the
week
before.
So
that's.
B
Brown,
it's
a
big
trip,
totally
understood
very
March,
20th
okay.
So
what
I'm
hearing
is?
Maybe
we
cancel.
B
Careful
all
right,
Robert
Bailey,
who
is
sick
today
with
strap,
is
apparently
everyone
is
sick.
Right
now,
1.30
release
is
next
week
on
Tuesday
and
there
is
no
release
candidate
anymore.
We
go
straight
to
release
so
just
a
reminder.
Feature
Zach
you've
got
a
feature
dropping
next
yeah.
Your
your
baby
is
coming
to
light.
B
Have
autopilot.
D
Support
and
130.
we'll
also
have
as
of
130.
The
split
controller
and
extensions
feature
will
be
in
Alpha,
so
the
we
should
have
highly
available
lagunas
in
Alpha.
At
that
point,
Max
has
a
another
baby,
dropping
two,
which
is
the
multiple
version,
support
kubernetes
version
support.
E
Yeah,
so
from
the
release
perspective,
do
we
need
to
pay
special
attention
at
anything,
since
these
two
are
getting
introduced
in
this
release?
The
gke
autopilot
and
the
multiple
versions
support.
A
Sorry,
for
the
multiple
words
support
the
release.
Tab
is
the
same,
but
for
one
particular
step
you
know
you
have
to
previously.
You
need
to
modify
the
website
content
to
reflect.
That's
the
current
support
version
in
a
single
word.
So
right
now
you
need
to
you
know,
turn
that
to
This
Server
into
the
sport,
but
I
have
that
I
have
a
PR
opening
to
change
the
release
steps
you
just
need
to
follow
that
it
should
be
fine.
E
Okay,
okay,
so
once
your
PR
is
submitted,
medical
I
can
take
a
look
at
it
and
see
what
those
changes
are
and
if
there
are
any
questions
she
can
check
with
you
right
right.
D
D
D
You
asked
about
the
autopilot
side
on
the
autopilot
side.
We
added
it
to
CI,
so
it's
it's
now
in
continuous
integration
and
like
as
of
this
last
release.
Our
our
test,
Matrix
for
CI,
has
actually
expanded
from
like
one
to
I,
guess
five
right,
because
autopilot's
missing
a
version,
but
so
we're
actually
continuously
testing
many
more
configurations
than
even
the
last
release.
E
Yeah,
because
since
we
are
not
going
to
have
RC
this
time,
the
whatever
version
that
it's
going
to
be
the
latest
version
and
not
the
pre-release
kind
of
a
thing.
So
that's
why
I
just
wanted
to
be
a
little
bit
more
cautious
with
new
changes.
B
B
That's
very
exciting.
All
right,
Stephen
excuse
me.
C
Yes,
so
yeah
I
was
just
wanted
to
dig
into
a
bit
about.
We've
got
this
recommendation
on
the
docs
about
running
a
Goonies
on
separate
nodes.
If
there's
any
more
info
on
that
and
then
I
guess,
I
guess
with
auto
pilot
support,
then
no
I,
guess
that's
not
gonna!
Well,
you
don't
really
have
that
level
of
control.
So
maybe
it's
relevant
to
discuss.
B
E
B
More
details
on
that
really
Zach,
you
were
going
to
say.
Let
me
go
ahead.
D
I
was
about
to
say
our
public
docs
actually
do
reference
trying
to
separate
the
ogune
system,
namespace
from
the
Whatchamacallit
from
the
game
servers.
D
D
On
autopilot,
there
actually
is
an
option
for
this
and
I
intend
to
document
it
when
we
actually
get
more
of
the
autopilot
docs
in
basically
on
autopilot,
there
is
an
option
for
a
workload
separation.
D
We
didn't
default
to
it
on
because-
and
this
is
kind
of
a
negotiation
going
on
with
the
autopilot
team
right
now
too,
in
order
to
use
workload
separation,
it
induces
a
half
vcpu
minimum,
basically
like
we
don't
want
people
abusing
autopilot
and
trying
to
like
force
it
into
a
situation
where
it's
like
one
pod
per
node,
but
they're
tiny
or
something
of
that
nature.
D
So
that's
why
it's
there,
but
that
was
probably
a
way
too
long
answer,
but
there
is
an
option
for
it
on
autopilot,
I
intend
to
document
it
soon.
C
And
I
think
there's
actually
a
bit
more
info
on
the
official
docs
on
the
helm.
Install
agoni's
Helm
about
dedicated
nodes
as
well
is
a
bit
more
details.
E
C
E
C
It's
the
it's,
the
node
pool
that
maybe
it
goes
into
a
bit
more
detail
there.
B
Why
don't
we
do
that
twice:
yeah!
That's
fine!.
B
Resiliency
yeah
I
mean
it's
a
good
thing
like
I
feel
like
it.
It
can
depend
right
if
you're
like,
if,
especially,
if
you're
running
a
game
server,
that's
extremely
CPU
sensitive,
you
may
just
be
like
hey
I,
want
to
have
different
machine
types
for
my
controllers,
so
I'm,
like
maybe
I'm,
running,
like
c2s,
for
my
game
servers,
but
I
only
really
need
like
if
I'm
talking,
Google
cloud
like
like
any
E-2
or
something
for
just
like
the
compute,
so
I
want
to
have
separate
types
for
that,
because
there's
no
point
learning
c2s
for
that.
B
But
if
you
know,
if
the
agonized
controller
is
not
gonna
impact,
your
game
servers
they're,
not
that
CPU
heavy,
maybe
you
don't
have
to
which
is
fine.
Generally,
it's
you
know
it's
it's
nicer,
because
it
gives
you
that
that
that
room
kind
of
to
breathe.
But
you
know,
there's
there's
nothing
inherently
like
if
your
system's
working
there's
nothing
like
inherently
bad.
B
B
A
B
Okay,
I
just
want
to
do
a
check
on
a
couple
of
my
PR's
that
are
technically
approved,
but
had
like
we're
like
had
either
just
wanted
to
double
check,
I
think
I,
think
they're,
fine,
I
just
wasn't
sure.
If
people
were
like
particularly
the
make
out
for
Invader
crd
stuff,
consistent
Zach,
you
were
like
I'm
waiting
for
the
optional
thing.
I've
done
the
optional
thing
I
just
wanted
to
check
before
I
was
like
marking
it
to
be
merged.
B
B
And
probably
hit
auto
update
at
some
point
yeah
we
can
just
do
that
and
then
the
other
one,
okay,
I,
don't
think
anyone's
I.
Think
Max,
you
up.
Was
it
you
this
yet
was
you
okay?
So
you
approved
that
one
all
right
I
will
also
enable
auto
merge
on
that.
B
E
B
Seen
this
one,
what
was
that.
B
Was
no.
B
B
D
Yeah
yeah
I
mean
that's
what
that,
if
that's
just
like
you're,
seeing
that's
what
that
indicates,
it's
a
bandwidth
issue.
It
was
kind
of
a
Band-Aid
like
before
this
would
have
been
timing
out
in
a
console
lock.
D
Yeah
about
an
hour
and
a
half
is
pretty
bad.
It
might
be
worth
trying
to
understand
if,
like
our
ede
times,
went
up
somewhere
along
the
way,
but
actually
the
build
times
that
I'm
seeing
on
the
dashboard.
Don't
look
particularly
yeah,
like
the
Builder
succeeded,
all
were
in
the
kind
of
half
hour
range.
So
that's
weird
about
right.
The.
B
Only
other
thought
I
have
not
that
you
can
read
that
is
I'm
assuming
these
are
in
the
right
order
as
well.
B
If
we
look
at
our
history
where's
one
that
like
succeeded,
GitHub
PR
get
up
here
there
we
go
this
one
that
worked
so
end
to
end
it's
fantastic
about
yeah.
It
takes
like
15
minutes
for
a
build
and
10
takes
about
40.
B
D
B
But
if
we
look
at
this
one
here,
oh
that
one
took
like
an
hour
and
36
minutes,
I
suppose
it
also
depends
on
how
long
it's
blocking
for
as
well
yeah
exactly
there
is
there.
Is
that
too?
Actually?
What
was
that
one,
the
other
one
that
I
did
that
was
like
40
here
for
this
one?
We
can
actually
see
okay,
we're
already
okay,
so
that
one
didn't
that
one
didn't
wait.
The
actual
submit
end
to
end
test
was
38.
B
D
In
like
for
a
lot
of
our
internal
stuff,
would
it's
we
run
projects
and
then
create
the
cluster
and
kind
of
start
from
scratch,
but
but
yeah
like
we.
If
we
had
something
like
that,
we
could
obviously
do
it,
but
it
also
requires
a
more
a
broader
infrastructure
investor
investment,
and
there
is
an
open
source
version
of
that
like
I.
Think
it
was
called.
Bosscos
is
what
the
kubernetes
project
uses,
but
I'm,
not
sure
if
that
got
merged
into
prow.
B
D
Well,
it
looks
like
it's
still
active
okay,
this
is
the
project
we
use
in
the
kubernetes
land
and
it's
meant
for
basically
I
have
a
config
I
want
to
rent
that
config
and
then
I'll
return
it
and
I
suspect,
like
even
without
prowl.
We
could
do
something
like
that
if
we
wanted
to
get
like
a
pool
of
clusters
up,
but
it
is
obviously
a
little
more
work
to
figure
out.
Oh
my
gosh.
D
D
B
D
B
D
D
We're
telling
this,
like
careful
line
between
you,
know,
there's
an
argument
to
be
said
to
be
made
that
we
should
just
blow
the
time
out
out
and
be
like.
Let's
just
wait
forever,
especially
for
possibly
for
this
like
one
case,
but
then
then
you
also
then
run
the
risk
of
e
to
E
something
happens
and
it
takes
four
hours
or
whatever
to
complete,
and
so
I
I
don't
know
like
what
the
right
tension
there
is
like
you
have.
Every
test
system
I've
worked
with.
Has
that
issue
where
it's
like
yeah.
B
D
Actually
that
reminds
me
of
an
idea
that
I've
had
at
some
point,
which
was
a
prioritized
approved,
pianist
it
well
that
or
it
actually
wouldn't
be
hard.
One
of
the
problems
we
have
is
that-
and
this
is
super
common
for
me
to
do-
is
I'll
push
the
pr
and
then
I'll
re-review
it
myself
on
GitHub,
because.
B
D
Then
you
realize
that
you're,
like
oh
whoopsies
and
then
I
fixed
something
and
I
push
it
again,
and
so,
like
five
minutes
later,
it's
running
again
and
I
noticed
that
actually
happens
a
lot
on
our
on
our
cues
and
it
wouldn't
be
hard
to
insert
a
build
step
to
basically
delete
anything
that
was
running
with
the
same
reference.
B
D
B
It
doesn't
do
that
either
that
or
just
like
go
back
in
and
cancel
your
old
build,
which
I've
absolutely
done
as
well.
Then,
like
oh
I've
got
a
new
one.
Coming
I'll
just
go,
cancel
my
old
one.
D
Because
that
would
save
like
a
lot
of
these
are
kind
of
nickel
and
dying
things
the
same
as
a
little
bit
of
bandwidth
here
and
there
and
I.
But
I
feel
like
we're
at
that
point.
Where
that's
fine
and
we
don't
need
to
invest
in
like
a
big
project
to
make
this
like
multi
yeah,
multi
infrastructure.
Yet
because.
C
D
Feel
like
occasionally,
we
just
hit
traffic
issues.
It's
like
I
think
right
now
we
have
the
one
lane
Street
and
we're
we're
debating
whether
to
make
it
multiple
lane
or
figure
out
how
to
make
the
traffic
a
little
more
orderly.
B
The
the
easier
answer
may
actually
be
to
be
like
hey.
If
you
find
a
build,
that's
got
this
PR
in
it
then,
like
the
same
Branch.
That
being
said,
that
is
specific
to
how
the
build
system
works
now
with
some
string
and
glue,
but
it
works.
So
it's
it's
not
barking,
you
don't
you
don't
know
how
it
works.
It's
awful!
Don't
don't
don't
look.
B
B
It
runs
a
luck
in
Chrome
and
I
mean
it's.
It
is
worked
flawlessly
for
years.
I
think
the
only
time
it
tends
to
fall
over
is,
if,
like
the
disc,
gets
pulled
on
the
instance.
It's
like
this
tiny
little
instance
and
then
I
think
it
does
some
other
cleanup
stuff
like
cleaning
up
like
app
engine
instances
and
some
stuff
like
that
which,
if
that
ever
fails,
is
not
the
end
of
the
world
but
yeah
at
some
point.
B
It
yeah
moved
to
them
yeah
and
then
there's
some
Cloud
functions
that
you've
seen
for
doing
all
the
GitHub
status
stuff,
but
there's
like
yeah
I've,
set
it
up
properly
when
I
did
quilkin,
which
is
using
there's
a
swing,
go
code
for
doing
notifications.
We
have
like
a
little
bot
thing
that
we
can
run
in
Cloud,
run
and
stuff,
that's
way
easier,
but
or
reasonably
easier,
but
yeah
one
day,
one
day
when
the
pain
gets
big
enough
whatever,
so
one
of
us
will
be
like
all
right,
I'm
doing
this.
B
Are
we
doing
for
time?
Okay,
do
we
want
to
talk
about
the
pr
about
agones,
gke,
quick,
start
autopilot
cluster.
D
I
was
debating
if
we
had
the
Right
audience,
although
this
is
the
obviously,
this
is
the
community
meeting.
So
in
theory,
this
is
the
Right
audience,
but
I
wanted
I
personally
wanted
to
get
ishan
and
Robbie's
input.
Here
too,.
D
D
I
mean
my
two
cents
is
I,
actually
wouldn't
mind.
The
quick
start
pivoting
over
to
autopilot,
except
for
a
couple
like
I,
actually
agree
with
your
points
that
right
now
like
it
relies
on
a
feature
gate
that
is
an
alpha
clearly
and
like
there's
parts
that
are
kind
of
like
I,
think.
Maybe
we
should
wait
on
that
until
until
everything
is
GA
or
at.
B
Least
not
I
mean
there's,
there's
yeah
there
yeah
like
like
I
I,
know
we
I
know
we
want
to
like
we're
like
this
shiny
new
thing,
but
like
probably
sending
setting
expectations
appropriately
and
also
sending
customers
on
the
trade
path
is
probably
the
better
one,
but
also
like
give
them
the
option.
I
don't
have
a
problem
with
that.
D
Fair
I'll
try
to
figure
out
how
to
yeah
I
can
work
with
Shannon
on
on
some
way
to
structure
that
yeah.
D
D
Mostly
yeah,
like
I,
can
tell
you
an
example,
because
I
have
a
PR
out
for
of
a
case
where
one
page
might
be
excessive,
which
is
the
metrics
page.
It's
it's
like
this
weird
swiss
cheese
of
like
three
different
back
ends
and
yeah
installation
instructions
and
like
I,
I,
definitely
understand
the
trade-offs
there,
because
it's
like
I
I,
don't
know
exactly
how
to
structure
some
of
these
things
and
Shannon
will
probably
help
with
that
later.
At
some
point,
most.
B
D
Okay,
mostly
I,
was
trying
to
get
and
like
I
started
to
reference
it
in
the
docs,
but
I
do
want
some
sort
of
landing
page
available
for
the
release
notes.
That's
probably
my
target
by
Tuesday
is
to
get
yeah
somewhere
to
link
off
to.
If
someone
sees
it
on
the
release
notes-
and
it's
like
okay,
this
is
how
how
to
start
absolutely
makes
sense.
B
Does
that
make
sense
all
right
all
right?
Well,
I'll!
Let
you
I'll,
let
you
Pro
catoids!
Listen!
If
we
want
to
sit
down
and
like
offline,
have
a
chat
for
sure.
D
B
Is
true
if
it's
easier
to
explain
stuff
in
voice
I'm,
saying
I'm,
happy
to
yeah
yeah
internal
video
chat,
foreign
cool
I
will
raise
just
given
where
allocation
overflow
is
and
it's
Thursday
it
might
slip.
This
release
I.
C
B
E
So
that
release
we
were
thinking
of
starting
at
eight
in
the
morning,
so
are
there
initial
stoppers
for
that
or
it's
good
that
we
can
start
at
8am
PST.
E
B
E
You
were
thinking
of
getting
some
of
these
full
requests
merged.
So
that's
why
I
was
just
thinking
that,
should
we
wait
for
anything
getting
merged
before
we
start
the
release.
B
Do
we
should
we
say
if
there's
anything
blocking
we'll
we'll.
A
B
B
Oh
all
right,
that's
my
things.
B
A
Yes,
that's
what
I
want
this
morning?
Okay,.
D
B
Cool
thank
you
for
that
I'm
just
going
through
yeah.
Let
me
know
when
you
yeah,
if
you
want
to
check
the
link
into
the
notes.
B
The
other
thing
I
have
is
I'm.
Just
checking
to
see
who'll
be
a
GDC
who
might
be
at
kubecon,
EU
I,
don't
know
Steven
if
you'll
be
a
cute
kind
of
you.
A
B
B
I'll,
let
you
know
Jack
who's,
that
director,
he
is
told:
okay,
yeah
yeah.
C
B
B
C
C
D
Else
actually
I'll
take
this
opportunity.
Since
we
have
a
you
know,
an
actual
customer
around
Stephen
I'm,
curious
I've,
been
playing
with
the
Prometheus
and
grafana
on
autopilot.
Just
I'm
curious,
like
what
your
metric
setup
is
at
and
like
do
you
all
use
our
stock
dashboards?
Did
you
copy
them
at
some
point
and
Fork
them
like
just
kind
of.
C
C
B
A
D
Across
all
platform,
thing:
okay,
honestly
just
wanted
to
hear
from
a
real
life
customer
that
as
to
like
what
people
are
using
I
I,
we
have
all
these
instructions
available
and
honestly,
the
cloud
monitoring
instructions
seem
to
be
one
of
the
less
well-served
and
I
honestly
haven't
tested
that
path
with
autopilot
just
yet.
But
that
might
give
me
an
indication
that
I
should
yeah
so.
C
C
B
It
that's
that's
unless
anyone
answers
anything
else,
I'm
just
trying
to
think
what's
else
is
in.
C
E
B
C
At
some
point
or
no
I'll
just
go
to
delete
I
think
the
reason
I
went
to
we
were
on
we're
actually
on
117
on
fraud
and
I
wanted
to
test
something.
So
I
was
like:
what's
the
what's,
the
smallest
change?
I
can
make
and
still
build
a
cluster,
because
I
can
build
a
I,
think
121
gpe
cluster
cells,
but.
C
Works
I,
I,
yeah
I'll,
just
we'll
we'll
get
there
I,
don't
think
I'll
I,
don't
know
part
of
me.
Yeah
I
should
just
go
through
every
single
version.
The
other
parties
would
just
probably.
D
D
B
I'm
trying
to
I'm
actually
trying
to
think
and
I
I,
don't
remember
anything
or
I,
don't
remember
anything
actually
changing
in
the
last
years
and
now
from
beta
features.
So
you
might
be
fine,
but
those
are
the
only
things
I
can
think
of
I
mean
we
deprecated
preferred
and
required
on
the
allocation
apis,
but
they're
still
there
and
you
can
still
use
them,
so
they
haven't
actually
gone
away.
E
A
B
B
Exactly
right,
so
I
think
everything
else
would
be.
Okay,
yeah,
yeah
I
would
actually
be
really
curious,
but
in
theory
it
should
be
fine.
C
We
never
we
didn't.
We
would
I
think
we're
talking
about
changing
the
EU,
the
rest
API,
but
we
I
don't
think
we
haven't
did
change
that.
Did
we.
C
B
Cool
unless
there's
anything
else,
then
going
once
going
twice
going,
Thrice
all
right.
It's
sold
into
the
meeting
I
think
if
you're
joining
once
again
pleasure,
seeing
you
all
and
I'll
see
you
all
in
a
couple
of
weeks:
awesome,
that's
better
yeah
catch!
You
all
later.
Thank.