►
From YouTube: WG K8s Infra 2019-04-17
Description
GMT20190417 153249 Dims k8s 640x360
A
Coming
and
Tim
is
helping
demonstrate
how
we
adhere
to
the
kubernetes
code
of
conduct
by
raising
his
hand
and
not
being
a
jerk
which
basically,
what
our
code
of
conduct
boils
down
to.
You
can
all
watch
Tim
do
that
later
on
YouTube,
because
we
are
being
publicly
recorded
this
Tim
so
nicely
confirmed
for
us.
A
So
do
I,
believe
I,
recognized,
I,
believe
I
recognized
everybody
here,
so
I'm
gonna
move
past
and
welcoming
new
members
in
attendees.
I'm
gonna
go
straight
to
action
item
review,
but
unfortunately
I
don't
have
any
action
items
directly
in
front
of
me.
Does
anybody
have
anything
they
specifically
want
to
speak
to
or
should
or
would
you
recommend,
I
go
down
last
week's
notes
and
take
some
stuff
out
I.
A
B
Few
minutes
ago,
so
in
the
two
weeks
since
we
have
met
last,
we
have
spent
eighty
five
dollars
on
VMs.
That's
the
test
clusters,
which
need
to
be
finalized
that
sorry
$85
on
compute
$42
on
RAM
and
thirty,
one
dollars
on
SSD
PD
and
those
are
the
those
the
VM
based
charges
$20
on
DNS,
which
comprises
49
million
nine
hundred
and
twenty
thousand
queries.
A
D
A
I'm
reading
those
bills
and
so
I
apologize
for
asking
probably
the
same
question.
I
asked
every
time
I
feel
like
I
want
to
live
in
a
world
where
we
understand
with
each
piece
of
infrastructure
that
we
add
to
this
effort.
We
understand
how
much
that
piece
of
infrastructure
is
costing
us.
Is
that
an
unrealistic
expectation?
No.
B
No,
it
is
not,
and
part
of
the
I
would
say
the
gates
for
getting
the
cluster
turned
on
for
real
is
enabling
the
billing
breakdown,
which
won't
give
us
a
name
space
by
namespace
view
of
what's
happening
right
now.
The
bill
that
I'm
reading
out
here
is
just
the
default
GC
living
report.
It
is
not
a
further
breakdown.
We
have
the
data
in
bigquery,
ready
to
be
analyzed
in
data
studio,
which
someone,
I
won't
name
says,
is
really
easy
and
he
named
himself
that
we
just
need
to
do
like
honestly.
B
A
The
reason
I
feel
silly
for
asking
this
question
is
to
me.
It
goes
to
other
questions
that,
maybe
you
all
have
answered
already,
which
is.
Are
we
considering
doing
different
DCP
projects
to
help
us
shard
this
level
of
building
billing,
and
are
we
considering
one
cluster
for
all
of
our
infrastructure?
Are
we
considering
different
clusters
for
different
piece
of
infrastructure?
I?
Don't.
B
Think
we
have
a
hard
answer
yet
I
would
like
to
have
the
smallest
number
of
clusters
possible
and
to
lean
on
namespaces
and
our
back,
and
you
know
all
the
cool
integrations
that
are
possible
as
much
as
we
can.
We
will
need
multiple
clusters
if
and
when
we
want
to
move
into
multiple
regions.
Obviously
we
don't
just
start
there
we're
not
there
today,
like
the
the
redirector,
for
example,
only
runs
in
the
US,
and
that
seems
to
be
fine,
but
eventually
we'll
want
to
move
it
to
multiple
regions
if
for
no
other
reason
than
availability.
B
A
E
A
B
A
C
So
sorry,
I
have
to
say
that
out
loud,
that's
fair.
So,
just
before
the
call
I
went
looking
for
my
action
items
and
I
have
like
two
sentences
that
I
came
of
it
and
not
more
so.
It
reads
like
this:
the
various
projects
in
Cuba
notice
and
Cuba
96
yet
have
produce
artifacts,
including
release,
binaries
container
images,
etcetera
the
work.
The
working
group
will
help
figure
out
choices,
options
for
supporting
the
creation
and
dissemination
of
these
artifacts.
C
A
Think
that's
fair
just
so
to
me
it's
the
decision
tree
is
like
people
try
to
convince
us
and
if
we
can't
decide,
we
then
escalate
to
steering
I,
guess
right,
I
hate
to
put
steering
so
immediately
in
the
path
of
escalation
there
and
when
I
say
we
decide.
A
C
A
A
They're,
the
people
we've
shared
and
skimming
yeah
I
mean
I.
Can
I
can
brainstorm
some
anti
patterns
down
the
line.
We're
like
everybody,
shows
up
wanting
us
to
pay
to
host
their
artifacts
and
the
system
could
be
gamed,
but
I
also
feel
ok,
saying
we're
just
gonna
have
to
deal
with
that
on
a
case-by-case
basis,
right
to
be
reasonable,
and
if
you
would
like
I
can
take
this
case
by
case
suggestion
up
to
steering
to
let
them
like
fully
delegate
that
to
us
or
something
I.
Don't
know
your
class
III.
C
C
C
A
Let's
just
think
about
what
actually
it
was
related
to
billing.
We
should
do
for
next
week
sounds
like
Justin.
You
still
have
some
or
mysterious
person
who
raised
their
hand
when
Tim
was
not
trying
to
name.
Your
name
sounds
like
you
still
have
some
work
to
do
around
billing
for
storage,
or
you
said
you
had
some
good
news
to
share.
D
F
F
About
that,
okay,
like
that
sounds
like
that
works
as
far
as
it
in
term,
but
I'm
wondering
what
we're
going
to
do
in
the
eventual
future
web.
The
working
group
closes
down
because
working
groups
are
temporary
like
who,
just
something
on
the
back
of
the
mind,
question
to
think
about
like
what
what
the
future
state
of
that
look
like
down
the
road
agree.
Yeah.
A
C
B
A
Don't
think
we're
at
the
point
where
we
need
to
answer
that
question
just
yet
it
could
be.
We
turn
into
a
sig
I
also
feel
like
you
could
make
the
case
that
each
of
these
pieces
of
code
actually
belong
to
different
SIG's
and
whatnot
and
much
in
the
same
way
as
the
steering
committee
has
ultimately
delegated
their
authority
to
individuals.
C
One
more
thing
here,
right,
I,
think
the
first
mission
for
us
is:
how
do
we
get
things
out
of
the
Google
own
stuff
to
see
ncf
own
stuff,
and
that
would
be
like
the
first
thing
that
we
should
do
and
not
trying
to
go
around,
find
where
all
we
can
spend
money
right.
Yeah.
That
would
be
like
a
different
lesser
priority.
Category
I
would
say,
agree.
C
The
other
part
here
is:
we
already
have
like
an
escape
mechanism
for
getting
people
to
talk
to
C&C
F
by
when
we
said
that
in
the
steering
we
said
that
we'll
figure
out,
if
it's
less
than
X
number
of
dollars,
people
can
do
service
less
tickets
and
talk
directly
with
csdf.
So
there
is
one
more
mechanism
there
already.
Yes,
okay.
A
A
B
D
I
was
also
unable,
although
I
did
unhelpfully,
create
a
massive
bike
shed
fest
on
automation.
So
it's
not
necessarily
clusters,
but
maybe
we
can
figure
it
out
on
buckets
and
then
apply
the
learnings
of
the
bike
shedding
to
how
we
manage
clusters,
because
I
would
I
would
like
us
to
have
one
way
to
manage
the
things
whether
that
ends
up
being
bash
cow
or
Pearl
you're.
Trying.
A
D
B
So
Justin,
maybe
in
the
next
two
weeks
we
can
partner
up
for
a
day
or
half
a
day
and
actually
focus
on
trying
to
get
this
over
the
finish
line
with
respect
to
automation
and
access
control.
So
we
can
buy
test
it
with
one
of
these
staging
team
groups,
style
stuff
with
Google,
Groups
and
and
our
back
and
everything
else,
I
think.
D
B
I
was
gonna,
say:
well,
you
don't
have
to
be
here
to
do
that.
But
for
me
the
like
the
Friday
mornings
tend
to
be
great.
I
can
put
aside
like
the
entirety
of
the
Friday
morning
and
even
into
the
afternoon,
and
that's
been
pretty
successful
at
getting
a
big
block
of
time
to
get
things
like
this
over
the
finish
line.
Just
not
this
Friday.
If.
A
A
A
B
That's
a
great
thing
that
we
should
do
it
like
as
soon
as
we
get
the
a
cluster
up.
We
should
have
this
as
one
of
the
very
first
things
that
we
run
in
that
cluster
as
a
proof
of
being
able
to
get
our
backs
and
access
control
and
linking
all
the
pieces
together,
because
we've
been
very
careful
at
curating,
a
group
that
has
just
the
permissions
it
needs
to
be
able
to
audit
and
dump
I
am
stuff,
and
so
getting
a
job
running
in
the
cluster
with
those
credentials,
I
think
will
be
a
good
demonstration.
B
A
B
A
Okay,
and
mostly
since
Christoph,
is
here
and
I've
heard
him
express
opinions
about
the
use
of
G
suite
for
and
it's
wonderful
ap
is
for
management
of
groups.
Where
are
we
on
the
use
of
G
suite
versus
the
use
of
public
Google,
Groups
or
permissions?
Do
we
feel
like
we
have
to
have
this
locked
down
before
we
can
like
really
move
forward.
F
B
F
F
What
groups
we
have
today
and
what
are
kind
of
policy
is
around
group
formation.
What
I
would
like
to
do
like
I,
don't
want
to
block
any
other
pieces
or
any
other
work
on
this,
but
I
would
like
to
look
at
what
what
we
could
do
as
far
as
automation
is
concerned,
so,
like
automating,
generating
those
groups
automating
getting
people
into
those
groups,
preferably
a
get-ups
model,
same
as
we've
done
for
everything
else,
so
that
we
have
kind
of
a
tie
back
loop
between
between
the
visibility
and
auditability
and
all
that
kind
of
stuff.
F
As
far
as
how
those
things
are
generated,
I
think
API
wise.
We
might
need
to
go
with
G
suite
for
that,
because
I
don't
think
we
could
do
that
in
the
public
Google
Groups,
so
we
might
need
to
transition
things
over
from
the
public
Google
Groups.
That
we
have
to
be
sweet
groups,
but
I
can,
if
everybody's
okay,
with
that
I,
can
start
poking
at
that
I
think.
B
B
In
fact,
I
know
that
they're,
not
all
consistent
in
terms
of
who
has
ownership
and
yeah
so
having
an
API
to
that
is
I
think
would
be
great,
but
I
think
we
have
to
move
into
the
G
suite
to
do
that,
and
so
the
only
question
is
you
know
at
some
point
back
when
I
was
on
steering
we
didn't
want
to
do
that.
If
that
general
sentiment
has
changed
or
shifted
in
this
regard,
then
I'm
all
on
board
with
it
and
I
will
helpfully
help.
A
So
I,
first
off
I'm,
happy
to
check
back
in
with
Brandon
and
Michelle
who
I
feel
like
are
looking
at
each
case.
We
proposal
from
sinking
share
backs
to
understand
what
we
can
do.
There
I
have
heard
things
informally
through
the
grapevine
like
there
are
people
at
G
sweet
who
are
horrified
that
we're
not
like
taking
full
advantage
of
the
G
suite,
because,
theoretically,
we
are
a
public,
open
source,
slash
educational
project
and
we
could
have
up
to
a
thousand
users
or
I,
don't
know
so
feels
like
this
is
in
our
wheelhouse.
A
But
so
it's
not
to
block
on
that.
Just
so
I'm,
clear,
I
think
the
the
dumping
of
all
of
the
things
that
you
Tim
and
hippy
have
been
working
on
is
like
half
of
this
equation.
Right
helps
us
at
least
document
all
of
the
people
and
groups
and
what
things
they
have
access
to
you
today,
even
if
we
can't
automatically
change
things
from
that.
So.
A
F
F
B
If
we
can
get
approval
that
we're
going
to
do
this
and
we
can
figure
out
what
the
access
requirements
are,
I
don't
want
to
be.
You
know
one
of
two
people
who's
fielding
with
all
the
tickets
for
all
of
the
G
suite
create
requests
in
perpetuity.
But
if
that's
where
we
start
I,
guess:
I'm.
Okay,
with
that
I'm
happy
to
script
myself
out
of
a
job
to
in
bash.
Okay.
A
Just
so
just
trying
to
to
move
us
forward
and
just
ignore
that
comment
about
batch
entirely,
so
I
feel
like
we
are
saying
we
want
to
come
to
a
yes/no
decision
on
the
use
of
G
suite,
and
that
is
part
of
the
things
we
have
to
solve
before
we
like
open
up
the
gates.
I
feel
like
everything
we
talked
about,
I've
been
trying
to
get
us
to
reason
about,
like
what
do.
A
We
have
to
do
before
we're
saying
like
okay,
we're
kind
of
open
for
business
and
can
accept
more
help
and
talk
about
moving
more
infrastructure
to
clusters
and
whatnot,
and
it's
largely
like
getting
our
billing
right
and
we
have
a
eyes
for
what
to
do
there
and
then
it's
definitely
burning
down
the
first
cluster.
Getting
our
heads
together
to
to
maybe
work
on
like
a
script
to
to
create
a
cluster,
to
prove
to
ourselves
that
that
works
and
to
feel
like.
A
We
have
a
documented
plan
for
whether
or
not
we're
like
documentation
on
what
the
state
of
today
is
from
set
of.
I
am
policies
who
has
the
keys
to
what
which
Google
groups
or
people
have
the
keys
to
what
and
then
how
could
we
use
G
suite
to
make
that
management
more
effective,
have
I
left
anything
out
of
like
critical
path
there,
the.
A
I
did
want
to
like
get
back
to
the
image
promotion
stuff,
but
I'm
just
trying
to
make
sure
like
to
me.
I
keep
coming
to
these
meetings
and
I
feel
like
we're
still
not
at
the
point
where
I
can
say.
Like
okay,
everybody's
start
like,
let's
start
planning
the
infrastructure
to
move
over.
So
to
me,
like
I,
want
to
see
us
get
progress
on
all
those
things
soon.
B
The
topic
I
I,
agree
with
your
list
by
and
large
I
think
we're
ready,
at
least
in
some
fronts
like
some
you
know.
Obviously,
DNS
is
live
right
like
this
right,
nothing
in
the
way
there
and
if
the
G
suite
decision
comes
down
today,
we
could
start
this
week
next
week
on
switching
it
over
to
the
new
group
and
CIO
a
script
in
that
space
anyway,
so
like
nothing's,
blocking
that
the
cluster
is
not
blocking
Network.
The
cluster
is
the
next
most
obvious
step
in
parallel
with
the
GOG
CR
stuff.
Yes,
right.
C
G
B
A
Could
think
of
all
sorts
of
things
that
are
running
on
clusters
around
here?
That
I
don't
want
to
run?
Are
clusters
around
here
anymore,
I
just
haven't
asked
yet
so
I'm
I
want
to
get
to
there.
So
just
whoever
is
taking
notes
if
we
can
put
down
an
action
item
for
me
if
this
is
okay
with
everybody
like
I,
feel
super
uncomfortable
surfing,
a
dock
and
us
having
an
organic
discussion.
A
It's
great
that
we're
all
knowledgeable
enough
to
kind
of
riff
on
this,
but
I
also
have
a
tough
time
keeping
track
of
what
we
did
last
time,
what
we're
planning
to
do
and
how
much
of
it
is
left
to
do
so.
I
want
to
like
turn
all
this
stuff
and
get
other
issues.
So,
instead
of
reviewing
action
items,
I
can
review
github
issues
either
in
a
milestone
on
a
project.
Work
I
feel
like
there
are
other
stakes
or
groups
out
there
who
have
sort
of
tightened
up
their
policies.
A
B
So
DIMMs
opened
a
bunch
of
issues
a
while
back.
Some
were
sort
of
umbrella,
ish
and
some
were
actually
pretty
specific
in
the
original
dock.
For
this
group
we
had
a
list
of
like
40
things
that
we
wanted
to
move
yes
into
community
owned
space.
Do
we
have
it
like?
How
do
we
open
issues
for
every
single
one
of
those
things,
and
are
we
sure
that
nobody's
added
that
list?
Maybe
me
I'm.
A
But,
like
we
talked
earlier
about
how
we
want
to
make
sure
we
probably
or
ties
our
funding
or
moving
existing
infrastructure
out
rather
than
adding
a
whole
bunch
of
new
stuff
and
I,
feel
like
that's
where
we
need
to
express
this
stuff
in
milestones
whatever
so
I'll
take
an
AI
to
take
a
swag
at
that,
and
we
can
review
that
two
weeks
from
now.
I
will.
A
G
So
there
are
some
remaining
items
like
you
know,
like
backups
garbage
collection,
that
kind
of
thing,
but
they're
not
blockers.
So
as
soon
as
somebody
makes
a
PR
against
communities,
flash
Kate's
I/o
for
one
of
those
three
that
we
tracked
currently,
it
should
start
to
work
for
where
it's
worth
the
post
run
ran
successfully.
B
I
wish
I
had
thought
about
it
this
morning.
I
would
have
prepped
and
we
could
have
done
it
live.
Let's.
Let's
put
that
on
the
definite
agenda
committed
for
next
week
to
do
a
real
demonstration,
and
we
can
talk
about
the
details
of
exactly
what's
left
before
we're
happy
to
move
the
entirety
of
the
existing
repo
over.
Is
that
fair,
yeah
yeah.
B
F
A
B
You
know
it
should
work
right
now.
We
are
lacking
some
things
that
I
think
we
want
before
we
go
open
for
business
like
an
automated
way
of
creating
groups
to
enable
news,
Adri
pose
and
confidence
in
the
testing.
We
have
no
real
metrics
about
what
we're
doing
and
Leslie
the
the
new
progress
Oh
like
nobody's
gonna
use
it.
So
it's
not
really
open
for
business
until
we're
ready
to
throw
the
switch
and
flip
Kate's
GCR.
That
I
owe
over
to
this
new
repo.
In
my.
F
B
G
A
Okay,
I
mean
just
just
as
a
thought
throw
it
out
there
I
feel,
like
Linus,
has
been
doing
this
somewhat
under
the
purview
of
the
release.
Engineering
sub
project
of
sig
release.
Sig
elites
has
a
meeting
next
week,
Tuesday
the
23rd
just
to
like
show
them
that
it
works.
But
if
you
feel
like
it's
not
worth
demoing
this
until
we
have
all
our
ducks
in
a
row,
okay,
wait!
No.
B
B
A
E
A
A
D
D
I
have
opened
the
language
holy
wars
and
beginning
to
regret
it,
but
now
have
dug
in
we
could
we
I,
think
Tim
and
I
are
probably
likely
to
figure
out
our
differences
and
settle
on
the
one
true
language
of
school,
to
make
everyone
happy
and
then
rust
the
rust
or
rust.
Please
every
state
high
school
we
will.
We
will
probably
decide
whether
it
will
be
basher
girl
or
both
or
neither
but
the.
D
B
To
achieve
this,
we
may
want
to
actually
rename
some
of
the
scripts
and
repose
and
possibly
write
that
we've
already
created.
So
honestly,
if
we're
going
to
go
to
a
new
groups
process,
let's
just
wait
to
do
that
and
then
rename
it
in
that
transition
right,
because
some
of
the
groups
for
context
include
the
name.
A
A
D
There
are
two
PRS
I
did
link
them
at
the
bottom
of
the
docs.
There
I
think
2
to
2
and
2
to
3.
The
2
to
2
is
ace.
The
bash
side
and
I
I
think
we
I
think
Tim
suggested
and
I
agree
that
I
should
merge
that
into
the
GC
our
script,
which
makes
sense
there's
some
that
just
makes
happen.
The
go
side,
I
think
there's
some
enough.
People
can
still
hear
me.
Yes,
I.
C
A
B
The
question
is
really:
we've
got
a
bunch
of
scripts
for
creating
staging
repos
that
are
literally
bash
scripts
that
call
g-cloud
and
there's
a
little
bit
of
like
test.
If
this
thing
exists
and
if
not
created
and
if
so
set
it
logic,
because
g-cloud
doesn't
have
the
sort
of
declarative
semantics
that
we
want.
B
But
then
it's
not
terrible
and
we've
got
these
scripts,
not
Justin,
maybe
K,
maybe
different
reasonable
Minds
can
disagree
on
what's
terrible
and
what's
not,
there's
a
there's,
a
shell
library,
so
we're
entering
terrible
territory
and
I
haven't
yet
passed
an
array
to
a
function.
Although
I
wish
I
could.
B
So
the
question
that
is
for
creating
something
like
a
cloud
load
balancer,
which
is
significantly
more
complicated
than
creating
a
GCS
bucket
in
terms
of
G
cloud
commands.
Is
it
worthwhile
to
start
thinking
about
a
declarative
way
of
doing
this
as
in
writing
it
and
goes
so
that
we
can
control
the
semantics
a
little
bit
more
closely?
It's
still
gonna
end
up
writing
the
same.
B
D
I
would
also
hope
that
the
go
code
ends
up
being
more
reusable
for
future
things.
That's
more.
What
I'm
thinking
about
cuz
honestly,
like
I,
could
have
written
that
in
bash,
like
the
functionality
that
is
there
today,
I
could
have
just
as
easily
imagined
it
would
be
a
comparable
number
of
lines,
possibly
less,
even
but
you
know,
but
the
it's
more
like
the
future,
like
maintenance
like
richer
reconciliation
or
adding
functionality,
hopefully
using
the
same
code,
and
maybe
this
ends
up
being
something
that
is
used
by
other
projects.
B
Very
special
purpose,
and
and
to
be
fair,
there
is
a
large
amount
of
code
that
already
exists
in
the
Google
cloud
provider
that
is
sort
of
auto-generated
from
SDK
that
we
can
perhaps
lean
on
for
these
things.
So
that's
the
discussion.
I
obviously
have
a
slant,
because
I
wrote
all
the
original
shell
scripts
and
I'm
not
keen
to
rewrite
them.
B
D
A
Okay,
well
I
always
like
to
approach
things
from
the
let's
like
ship
and
iterate
perspective,
so
I'm,
honestly,
fine
with
bash
in
it
being
special
purpose
and
and
then
us
deciding
wait.
This
is
terrible
to
maintain.
Let's
move
it
forward,
but
I
understand
the
fallacy
of
like
once
it
lands
and
once
it's
actually
used
it's
so
much
more,
knowing
and
painful
to
actually
change
it,
and
maybe
there's
value
in
getting
it
right
but
like
to
me
it
works,
works
great.
A
A
A
D
B
D
A
A
F
A
F
There
was
a
discussion
on
an
issue
but
I
kind
of
weighed
in
on,
because
there
was
a.
There
was
at
least
four
part
of
the
net
Liffe
ID
connecting
things
there
was
a
need
for
github
owners
to
be
involved
to
connect
the
net
malai
with,
like
basically
there's
a
couple
steps
that
need
to
be
done,
that
you
need
both
admin
on
a
repo
and
admin,
and
that
will
fly
so
I
know.
F
Sig
dogs
was
talking
about
that
and
like
how
we
can
how
we
can
connect
that
piece
and
have
like
a
documented
process
for
delegating
identify
projects
to
two
repos,
so
I
provided
them.
Some
information
I'll
grab
the
PR,
so
we
can
talk
than
the
the
meeting
minutes
and
stuff
but
yeah.
It's
still
that's
the
lunch,
the
dogs
right
now.
As
far
as
I'm
aware,
okay.
A
B
Can
I
can
start
if
you
want,
so
we
had
a
long,
long-standing
open
to
do
of
moving
the
Godot
kate's
that
IO?
U
RL,
shortening
out
of
the
nginx
config
and
into
something
that
was
actually
designed
to
do
this
and
that
didn't
require
editing
the
nginx,
config
and
restarting
the
nginx
servers.
Every
time
we
wanted
to
add
a
go
URL,
so
Jonathan
proposed
SH
link
and
Michael
came
back
with
there's
this
other
thing.
That
is
even
simpler,
called
Tex
direct,
which
looks
to
be
okay
to
me.
We
there's
an
open-source
version.
B
B
It
uses
text
records
in
DNS
instead
of
a
database,
so
submitting
a
new
go
link
is
effectively
editing
the
DNS
conveying
and
having
somebody
who's
got
DNS
push
access
to
push
it,
which
seems
a
little
fraught,
but
I
think
we
can
delegate
the
entire
sub
zone
of
Godot
kate's
that
IO,
so
that
when
we
push,
those
updates
will
only
be
pushing
updates
to
go
and
not
to
the
rest
of
the
DNS
space.
I
think
so
that's
the
part
that
I'd
like
to
see
us
prove
out.
A
B
Either
I
mean
we
could
run
it
like
the
like
the
mini
motorers
and
run
it
as
a
prowl
plugin
somewhere,
and
then
we
have
to
argue
about
where
prowl
runs
and
and
etc.
Who's
got
access
to
that
or
we
could
literally
run
it
in
a
shell.
Loop
runs
in
a
cluster
somewhere
that
just
watches
get
and
when
they
get
changes
it.
It
updates
like
that.
I
know
how
to
do
that's
trivial
and
we
just
need
a
cluster
to
do
it.
B
So
we've
got
a
bit
of
a
cyclical
dependency,
but
I
would
put
that
in
the
first
first
ten
things
that
I
would
want
to
run
in
the
cluster
I.
Don't
want
a
human
to
be
doing
it
anymore
and
we've.
Actually,
the
script
at
this
point
is
pretty
reliable
and
that
it
it
tests
all
of
the
DNS
records
automatically
thanks
hippie
for
putting
that
together
and
it
retries
and
backs
off.
B
A
B
That's
great,
that's
good
feedback.
I,
don't
have
any
first-hand
experience
adding
plugins.
So
in
my
mind
it's
an
impossible
task,
but
it's
probably
not,
and
while
yes,
making
go
part
of
DNS
will
encourage
more
use
of
it.
I
don't
expect
that
we're
gonna
have
dozens
and
dozens
of
new
go
short
links
like
this
isn't
a
free-for-all.
It
still
has
to
be
relatively
carefully
curated
because
we
only
have
one
level
of
links
as
far
as
I
know,
or
maybe
maybe
text
direct
supports
more
than
one
level,
but
we're
not
currently
doing
that.
So
you.
I
Know
so
I'm
sorry
for
the
delay,
but
so
to
speak
to
text
Iraq.
So
the
general
idea,
and
also
which
allows
for
us
to
still
do
the
manual
thing
in
deploying
the
DMS
is
the
first
step
would
be
to
test
it
out
in
a
canary
fashion,
how
it
actually
works,
and
then
the
next
thing
would
also
be
because
it's
priority
to
self
host
the
open
source
part
so
that
basically
would
block
going
to
production.
And
then
we
still
have
to
figure
out
I.
I
B
Yeah
I
think
that
works
I,
not
sure
with
octave
DNS.
If
we
have
to
create
an
explicit
zone
delegation
or
if
it
can
actually
merge
different
llamo
files.
But
we
probably
want
to
create
the
explicit
delegation
which
tickles
a
memory
that
Google
Cloud
has
some
weirdnesses
with
DNS
and
having
multiple
zone
delegations
of
the
same
top-level
zone.
That
we'll
have
to
try
to
to
work
out.
But
I
agree
with
your
earlier
sentiment
that
basically
running
it.
Ourselves
would
be
the
gate
for
this,
and
that
requires
the
cluster.
I
A
A
Okay,
Tim
and
Michael
I'm
gonna
trust
you
guys
to
carry
this
forward.
I
also
have
to
drop.
So
this
has
been
super.
Productive
I
will
try
to
ping
the
channel
a
week
from
today
to
take
a
look
at
whether
or
not
I
the
what's
on
a
board
or
a
milestone
accurately
represents
everything
we
talked
about
here.
So
thanks
everybody
for
an
awesome,
Wednesday
I'll
see
you
later.