►
From YouTube: Kubernetes SIG K8s Infra - 2021-10-27
Description
A
So
hi
everyone,
we
are
october
25
27,
and
this
is
the
kitchen
from
meetings.
I'm
anno
and
I
will
be
your
host
today,
so
command
reminder
to
everyone
that
this
meeting
is
under
the
code
of
contribute.
So
I
will
invite
everyone
to
be
kind
to
each
other.
I
see
new
face
in
this.
B
Yeah
hi,
I'm
tom
stewart
I
just
kind
of
saw
everybody
at
the
cncf
kubecon
and
thought
I'd
try
and
get
get
involved.
B
C
We
actually
just
had
a
donation
from
a
company.
I
think
it
was
cox
that
has
edge
hardware
that
they're
wanting
to
iot
type
stuff.
So
it
might
I
mean
it
obviously
needs
to
be
a
cncf
project
that
goes
through,
but
it
might
be
interesting
to
see
what
they're
offering
that
might
overlap
with
your
skill
set
and
see
what.
D
E
I
suppose
I'm
a
new
face
as
well.
My
name
is
joshua
hoplet,
I'm
a
devops
engineer
at
rubin
observatory.
I
am
here
so
that
I
do
not
have
to
think
about
buggy
igmp
implementations
for
at
least
an
hour.
E
I've
been
a
kubernetes
user
since
very
early
on
about
2015..
It's
a
big
part
of
my
daily
life.
I
have
been
wanting
to
get
involved
in
the
kubernetes
community
for
years,
I've
been
sort
of
contributing
around
the
edges.
I've
contributed
to
things
like
the
terraform
kubernetes
provider
and
other
stuff
around,
but
I
haven't
been
involved
in
the
central
project
and
it's
a
really
crying
shame
that
I
haven't
so.
E
I
would
really
like
to
engage
somewhere,
and
I
think
this
is
actually
a
really
great
place
for
me
to
get
started
and
sort
of
build
up
some
relationships
with
folks.
So
here
I
am,
how
can.
A
Welcome
again
so
for
every
new
person,
I
always
suggest
to
basically
win
the
charter,
because
the
charter
will
provide
a
good
guidance
for
everything
we
are
doing
right
now
and
also
for
the
next
year.
So
I
put
the
charter
in
the
in
the
chat,
feel
free
to
read
and
ping
me
on
slack
about
some
things
you
don't
really
understand
or
how
you
can
do
this
start
to
get
involved.
A
A
A
G
Yes,
is
this
something
that
we
are
still
wanting
to
discuss
without
aaron
and
tim,
or
I
thought.
A
D
A
C
C
G
G
C
To
tell
if
it's
right,
do
you
mind
if
I
share
screen
or
not.
C
Res
to
come
in
as
a
screen
share
the
zoom.
This
is
a
funny
one.
C
So
the
background
here
I'll
give
a
little
intro
and
then
I'll
hand
it
over
to
caleb
and
I'll,
be
his
hands
on
the
on
the
screen,
we're
needing
to
set
up
registries
that
different
providers
so
that
it
doesn't
all
come
from
a
few
providers,
and
I
just
had
an
interesting
conversation
this
week
around
a
company
from
china
that
wants
to
be
able
to
also
be
a
part
so
that
people
in
china
can,
without
going
to
google,
be
able
to
use
kubernetes,
and
so
not
it's
not
always
just
about
the
cost
savings.
C
Sometimes
it's
about
making
our
communities
artifacts
available
to
all
humans
on
the
planet.
C
So
as
we
look
at
this,
keep
in
mind
what
this,
what
requires
is
for
us
to
know
where
people
are
coming
from
and
where
it
is
best
to
send
them
to
and
yeah
that's
basically
it
and
at
the
moment,
we'll
do
some
cost
stuff
later,
where
it
would
be
really
good
to
start
using
more
and
not
having
it
just
be
all
the
the
sole
burden
of
of
google
to
to
show
that
we're
closer
we're.
C
Looking
forward
to
having
more
members
of
our
cloud
native
community
provide
infrastructure,
we
have
a
new
cloud
native
credits
program
that
priyanka
launched
at
kubecon
if
you're,
not
aware
of
that
and
that's
beyond
it's
kind
of
encompassing
what
we
do
within
kate's
sig
cates
infra,
but
for
the
rest
of
the
kubernetes
community
as
well.
C
G
So
this
has
been
something
that
we've
been
discussing
for
a
little
bit
of
time
and
said
the
the
idea
of
how
it
should
work
is
matured
on
the
the
side
of
what
the
way
that
it's
implemented
and
the
technologies
used
and
various
things
around
that,
and
so
I
believe
it
will
end
up
being
registry.catestodaio.
G
Are
you
inside
of
a
cloud
provider
and
if
you
are
then
go
back
to
the
cloud
providers
registry,
if
they're
a
provided
solution
and
if
you're
anyone
else
just
go
to
the
default,
so
we
can
make
a
few
rules
around
how
that
will
end
up
working
and
so
there's
been
two
implementations
of
this
in
the
past
that
have
put
together
for
the
purpose
of
this.
Is
the
technology
stack?
Has
it
working
in
very
basic
form?
G
Eventually,
it
will
end
up
being
based
on
source,
ip
blocks
and
asns,
going
back
to
companies
and
all
kinds
of
stuff
like
that.
But
for
now
it's
just
is
your
ip.
This,
then
we'll
take
you
to
here
and
so
to
try
and
get
the
the
fastest
idea
out.
Here
are
two
more
options.
We
have
the
envoy
wasm
and
the
nginx
njs,
which
was
inspired
by
ricardo's
post
that
he
shared
to
me.
G
Thank
you,
ricardo,
and
so
I
made
these
more
extensive
and
put
in
some
more
information
about
it
on
each
of
them,
and
these
are
the
each
implementation
works
if
you
just
go
through
each
code
block
in
a
pair
instance
pair.sharing.io.
G
But
if
you
want
to
bring
it
up
yourself
using
all
of
this,
you'll
just
need
to
modify
it
ever
so
slightly,
but
this
is
a
deployment
for
artifact
server.
Artifact
server
was
written
by
justin
santa
barbara,
and
I
made
a
modification
to
it
to
add
some
configuration
and
to
make
it
do
more
dynamic,
redirects,
because
right
now,
it's
just
to.
I
think
it's
cops
or
something
and
so
here's
it
being
built
and
deployed
in
such
a
way.
And
then,
if
you
go
down
to
the
the
bottom,
where
there's
notes.
G
G
The
this
implementation
that
I've
put
together-
isn't
it's
not
the
greatest
I
I
can
definitely
improve
it
further
to
make
it
be
more
consistent
because
it
seems
to
have
some
bugs
in
its
responses.
But
one
thing
I
think,
would
be
a
good
benchmark
regardless
is,
is
it
performance
and
does
it
seem
to
do
the
job
at
least
50
of
the
time
in
the
state
that
it's
at
at
the
moment?
G
A
Yeah
how
I
can
test
that,
because,
basically
I
I
think
I
I
went
over
all
your
full
requests,
but
it
was
basically
unclear
to
me
how
I
can
basically
test
that
in
some
kind
of
station
environment
or
how
we
can
deploy
that
between
different
continents.
C
There's
we
can
walk
through
it,
but
for
us
we
just
bring
up
a
pair
instance,
and
then
we
walk
through
our
walkthroughs
and
bring
it
up,
and
then
that
actually
gives
you
an
ingress
to
test
like
a
public,
https,
kubernetes
ingress
that
goes
to
the
raw
hardware
on
equinix
metal.
That
runs
all
of
these
solutions.
G
A
G
The
doc
right
here
doesn't
have
the
source
code
in
it
directly,
but
it
does
have
all
of
the
ways
to
check
out
the
the
branch
and
how
to
put
it
together
and
I've
got
some
manifests
for
deploying
an
impair,
at
least,
and
I'm
happy
to
spend
some
time
with
you
to
bring
that
up
in
a
different
environment,
because
it's
it's
not
too
much
to
have
to
change
there.
The
oh,
that's
a
thing
yeah
I
yeah.
I
can
give
you
a
tour
of
that.
If,
if
that's
what
you'd
like.
C
What
we
can
do
is
we'll
go
ahead
and
bring
up
parents
and
bring
them
bring
them
all
up
and
give
you
access
to
that
instance,
so
that
you
can
pair
it
and
break
it
down
and
do
whatever
would
be
great
and
we'll
we'll
get
that
for
you
today,
okay
and
hand
it
over.
And
that
way
when
you
need
us
we're
here,
but
you
can
do
it
asynchronously
and
feel
free
to
destroy
or.
H
Okay,
I
just
do
we
have
any
more
thoughts
or
decisions
around
how
we
get
the
artifacts
to
these
other
places,
because
when
hippie
you
and
I
spoke-
we
had
talked
about-
maybe
having
sick
release
published
to
certain
places
and
then
aaron
seemed
to
prefer
the
idea
of
having
prow
jobs
set
up
to
just
sync
them
to
different
places.
H
C
Definite
parallel
thing
that
needs
to
happen
at
the
same
time,
and
I
think
it's
kind
of
up
to
you
and
me
to
hang
out
a
bit
to
actually
do
that
side,
because
I
I
know
that
there's
preferences
on
on
the
way-
and
I
think
we
just
need
to
go
ahead
and
bring
up
some
prototype,
brow
job
and
a
prototype.
C
You
know
parallel
instance
of
a
promoter
that
pushes
to
two
places
and
just
to
to
to
do
it
and
then
point
to
that.
It's
working
and
then
because
that's
how
you
vote
and
open
sources
with.
H
C
A
I
think
I
think
this
is
kind
of
a
kept
because
I
feel
like
you
need
to
talk
about
with
cigarettes
about
how
we
can
optimize
synchronization
between
the
the
different
bracket,
because,
if
I
use
my
heart
as
a
race
manager
associate,
I
would
say,
is
going
to
be
impactful
in
the
release
process.
So
I
think
it's.
A
A
A
document
express
how
we
basically
distribute
the
binaries
or
anything
related
to
the
release
between
different
cloud
provider,
because
until
now
the
release
process
is
only
focused
on
one
cloud
provider
which
is
gcp.
If
you
start
to
basically
distribution,
s3
packet,
ngc
packet
and
or
a
cloudflare
bucket
is
gonna,
be
impactful
for
the
release
process.
A
H
So
I'm
sorry
I
had.
I
had
talked
to
stephen
about
this
a
few
times.
Steven
definitely
wants
to
see
some
kep,
the
the
two
options
that
again
we've
discussed,
and
I
can't
speak
for
aaron,
but
aaron
seemed
to
favor
the
idea
of
having
proud
jobs
that
we
could
set
up.
That
would
basically
make
sure
that
our
community
owned
bucket
is
synced
to
all
of
the
downstream
buckets,
but
again
that
that
begs
the
question
of
policy
around
who
owns
those
buckets?
Do
the
clouds
provide
credentials?
A
I'm
okay
with
everything
any
discussion
happening
right
now,
but
I
want
I
want
for
us
to
be
transparent,
because
this
is
private,
one-on-one
conversation.
So
that's
why
I'm
I'm
asking
for
a
document
that
is
that
we
can
basically
present
to
community
and
say
this
is
what
we
we're
gonna
do
for
cube
adm
for
cube
cuddle
for
every
binary
release
every
time.
So
I'm
okay
with
all
the
options
we
have
now,
but
we
need
to
present
this
at
the
community
and
say
this
is
our
option
and
we
need
to
make
a
technical
decision
about
this.
E
So
I
think
the
important
thing
is
that
there's
some
level
of
control
within
the
project
that
you
can
turn
things
off
if
they
go
wrong.
So
you
could
say
it's
important
that
the
project
has
central
control
of
all
of
let's
call
them
mirrors
and
that
it's
a
cloud
provider
and
they
give
credentials
and
the
project
operates
all
those
mirrors
so
that
there's
centralized
control.
E
You
could
also
say
we're
going
to
take
like
the
fedora
or
ubuntu
or
linux
distribution
model
and
the
mirrors
are
going
to
operate
independently,
but
then
there
has
to
be
some
sort
of
central
mechanism
to
say
that
mirror
is
malfunctioning
in
some
way.
You
haven't
updated,
you're
out
of
sync
it's
broken
and
we
have
to
have
some
way
to
turn
it
off,
so
that
people
are
not
getting
a
bad
mirror.
A
Yeah,
I
agree.
I
agree
with
that.
The
the
the
problem
we
have
right
now
is
about
governance.
It's
not
technical!
It's
about
governance,
it's
about
it's
about
how
we
basically
deal
with
we
deal
with
the
different
way
we
ask
the
artifact
for
for
the
record.
We
use
credit
donation
right
now,
which
means
we
rely
on
only
one
cloud
provider
until
we
get
funding
from
the
csf
cloud
credit,
because
until
now,
it's
not
very
official.
A
If
I
basically
want
to
use
another
cloud
provider
except
gcp
and
amazon,
we
need
to
request
a
credit.
So,
for
the
moment
we
rely
on
just
two
cloud
providers.
If
we
want
to
distribute
that,
we
need
to
contact
to
the
different
other
cloud
providers
like
microsoft
and
say:
oh,
we
we
need
to
mirror
those
binaries
inside
a
version
of
microsoft.
A
How
we
can
do
that-
and
we
tell
in
the
conversation
about
this-
and
I
agree
with
you-
the
main
idea
is
to
replicate
what
the
foundation
are
actually
doing
right
now,
because
they
even
went
to
the
point
they
have
a
central
cdn
for
all
the
mirrors.
So
now
they
use
one
city
and
they
have
a
contract
with
fastly
and
they
basically
now
use
a
system
for
that
I
think
is.
We
can
replicate
that
for
us,
but
we
still
don't
have
the
full
capacity
to
do
that.
Right
now,
yeah.
E
Process
yeah,
I
mean
the
thing
you
run
into
as
soon
as
you
have
more
than
one
administrative
domain,
people
that
are
subject
to
different
priorities.
Right
and
okay,
it's
release
day,
and
somebody
else
has
got
something
more
important:
there's
a
fire
in
their
production,
environment
and
they're
gone
for
two
weeks,
and
they
don't
care
what's
important
to
you,
and
so
you
have
to
have
some.
You
have
to
have
some
way
of
dealing
with
that.
Maybe
okay,
we
don't
deploy
for
their
cloud
and
we
have
some
way
of
protecting
our
end
users.
E
So
they're
not
impacted
they're,
not
you
know,
they're,
not
stuck
because
of
that.
I'm
speaking
from
you
know
my
day.
Job
as
as
we
have
lots
of
these
international
agreements
where
people
run
on
different
schedules,
it's
just
reality
and
it
will
happen.
A
E
Basically,
I
would
just
say
make
it
a
statement
of
principles
and
just
say
the
the
number
one
priority
is
that
we
make
sure
that
end
users
are
not
impacted
by
these
sort
of
I
mean,
essentially
it's
a
cost
optimization
right.
So
obviously
we
have
to
stay
within
a
cost
envelope,
but
we
want
to
make
sure
that
end
users
are
not
stuck.
That's
that's
the
number
one
goal
and
then
the
second
goal
is
to
try
and
optimize
the
cost
and
performance
and
then
operate
within
that.
E
A
A
I'm
not
sure
I
think
we
just
have
to
this
is
the
main
point
of
the
conversation,
because
the
rest,
I
don't
see
man,
let's
see
what
I
say
in
the
call,
so
we
can
spend
some
times
over
this
and
trying
to
have
some
conversation
about
the
re,
the
other
implementation,
because
I
think
you
add,
I
think,
two
options
for.
G
The
registry-
that's
right,
there's
the
envoy
go
wasm
implementation,
which
is
new
and
then
there's
also
the
nginx
njs
implementation
and
each
of
those
have
their
own
ecosystems
around
them
and
would
have
their
own
benefits
to
each
different
implementation.
G
The
the
main
logic
that
is
quite
cool
is
down
at
the
bottom
here
in
this
checking
the
headers.
So
if
the,
if
the
header
for
the
remote
ip
matches
a
value,
then
we
just
change
it
and
rewrite
the
the
body
and
then
send
you
off
on
a
new
direction
with
a
302
and
this
one's
very
fast
like
it's,
it's
pretty
fast
and
it's
also
really
small.
G
Once
it's
compiled
to
just
compile
it
into
wesm,
with
the
command
below
right
there
and
that's
using
tiny,
go
and
then
here's
a
envoy
configuration
listening
on
a
standard
port
and
then
for
this
section
right
here
is
pretty
much
saying
on
any
request:
that's
http,
and
then
the
next
section
is
the
configuration
for
wesm.
The
only
really
important
things
here
are
the
code
local
file
names,
we're
saying
we
want
to
use
this
wesem
file,
then
we're
also
passing
through
some
environment
variables
and
we're
also
setting
one
and
then
that's
the
important
stuff
out
of
there.
G
This
one,
of
course,
is
templated,
so
that
variable
comes
from
pair
instances
and
there's
some
notes
at
the
bottom
good
references
there
and
yeah.
I
quite
like
this
one.
This
one
seems
really
good,
because
it's
in
the
ecosystem
envoy
and
go
we
we
in
in
this
community
we
quite
like
go
and
envoy,
is
also
another
cncf
project.
So
it's
pretty
good,
though
there
wasn't
really
much.
You
don't
actually
need
to
understand
how
wesm
works.
G
You
just
need
to
right,
go
and
understand
where
things
are
coming
from
and
where
things
are
going,
which
is
nice,
and
I
don't
think,
there's
really
any
major
things.
You
need
to
change
with
the
onboard
configuration.
So
that's
pretty
good
there
again
happy
to
answer
any
questions
about
these
implementations
at
any
point.
Just
raise
your
hand
so.
E
Where's
where's
the
asm
data
come
from.
G
The
asn
data
will
eventually
be
coming
from
sources
like
pi,
asn
and
potaru,
and
we've
used
those
in
a
separate
thing.
If.
G
Not
that
it'll
be
it'll,
be
in
the
test
and
for
roots
no
case
today.
Oh
sorry,
kubernetes,
slash
case.io,
but
we've
used
the
this
asm
data
from
these
providers
in
a
different
way,
which
will
be
quite
related.
If
you
go
view
code
down
there
right
in
the
center.
G
Sorry,
and
then,
if
you
go
to
images
and
then
you
go
to
public
log,
asn
match
it.
This
is
how
we're
currently
using
the
asm
data
and
we'll
be
using
it
quite
similarly,
when
it
comes
to
the
redirecting
and
there's
a
whole
lot
of
information
about
this.
If
you're
interested
in
that,
I
can
also
share
a
bit
about
that.
E
A
full-size
like
global
asn
table
in
there
that's
being
looked
up
against
great.
I
I,
where
I
was
going
with,
that
is
just
if
it
was
being
tested
against
a
full-size
table
for
performance.
G
Yeah,
I
I
have
some
ideas
about
implementing
things
for
speed
and
in
building
a
production,
dns
cluster
that
runs
on
kubernetes,
with
in
a
previous
place
that
I
was
at
with
some
others
using
some
in-memory
stuff,
which
might
be
quite
fun
to
look
at.
G
And
then
the
other
implementation
is
the
nginx
njs
one.
This
one
is
does
not
contain
a
lot
of
code,
it's
just
not
a
lot,
which
is
that
could
be
favorable.
I
don't
know
this
is
the
nginx
configuration
some
of
it
is
because
I
like
to
make
things
a
little
bit:
production
e
so
having
the
things
point
to
temp,
but
anyway,
that's
that's
a
bit
of
a
detail
we
have
upstream
host,
that's
been
rewritten
to
and
that
one
comes
from
the
the
function.
That's
beneath
it.
This
is
the
function.
G
Yep,
that's
right
there.
So
this
function,
which
is
a
javascript
file,
it
was
just
checking
for
what
the
remote
address
is.
This
would
be
passed
through
to
some
library,
because
njs
supports
modules
like
us,
webby
folks
are
used
to,
and
so
we'll
just
be
able
to
include
some
modules
that
do
exactly
what
we're
wanting
it
to
do
and
then
return
the
correct
host
based
on
the
ip
after
processing,
a
bunch
of
data
and
same
thing.
You
can
write
it
in
docker
this
way
and
you
can
also
run
it
in
kubernetes.
G
D
G
Very
good
so
yeah,
if
we're
wanting
to
those,
are
all
noises
we're
if
we're
wanting
to
do.
Yes,
everyone
gets
a
honk
if
we're.
If
we
wanted
to
do
things
in
javascript-
and
this
is
the
right
path
for
that,
so
people
could
be
very
familiar
with
it.
I
I'm
not
sure
about
how
quick
it
is.
G
But
those
are
my
thoughts,
but
I'm
I'm
very
interested
to
see
which
one,
oh
it
eddie
does
javascript
good
yeah
so
happy
to
hear
any
thoughts
about
any
of
the
implementations,
and
I
will
be
bringing
them
up
very
soon
and
providing
you
around
anybody.
C
This
is
not
far
from
our
mind
josh,
but
at
this
point,
we're
just
trying
to
figure
out
the
where
to
where
the
handles
are
on
this
thing.
With
regards
to
the
definitely
like
the
next
iteration
is:
okay,
now
where's
ipv6,
both
with
the
the
ingress
like
how
it
hits
the
the
reg
that
case
that
I
owe
where
that's
running
and
what
we're
handed
off
to,
because
I
think
we
just
finally
got
integrated
name
integrated
networking
this
that
was
announced
at
kubecon,
so
give
us
a
moment
to
breathe.
H
A
So
the
first
option
was
a
pull
through
cache.
If
I
remember
currently-
and
this
one
is
basically
some
kind
of
load
balancer
between
the
trip
based
on
the
ipad
just
of
the
of
the
origin-
or
am
I
missing
something
so
I
think
go
ahead.
A
C
C
It
into
the
to
the
channel
and
nobody
stepped
up
to
poke
around
with
it
yet
so
this
is
I'm
glad
so
we'll
we'll
do
that
and
invite,
because
what
we,
I
think,
did
you
post
it,
and
then
we
cuz,
I
I
don't
think
so,
not
posting
it
directly.
I
think
we
sent
it
to
tim
and
a
few
others.
So
if
you
want
access
just
raise
your
hand-
and
let
us
know.
C
C
So
thank
you
for
allowing
us
to
go
into
the
technical
areas
of
how
to
hold
the
the
packets
as
they
pass
through,
to
find
out
how
everybody
that
gets.
Kubernetes
hits
this
code,
and
I
I
am
I've,
had
it
on
the
action.
I
think
it's
been
a
kind
of
a
coordination
with
me
trying
to
find
the
right
people
to
coordinate
with
in
the
process.
I
do
think
a
cap
and
integrating
with
sig
release
and
I'll
I'll
take
the
the
lead
on
that
and
anybody
that
wants
to
help
me.
C
Please
raise
your
hand
and
come
come
meet
me
in
the
in
the
in
the
channel
or
or
an
uneye
to
eye,
thank
you
or
arnod
for
noticing,
as
your
release
manager,
that
there's
going
to
be
an
impact
on
that
process
and
that'll.
Allow
us
to
go
through
and
figure
out
how
to
get
the
stuff
out
to
these
different
providers
via
the
prowe
or
promoter,
and
thank
you
josh
for
lots
of
your
feedback
and
also
ricardo
for
the
implementation
of
the
mj
in
the
java
stuff
and
arnaud.
C
Thank
you
for
hosting
us
today.
No
problem
at
all.
E
Just
a
general
comment:
without
having
looked
at
the
technical
details
of
it
all,
I
would
expect
that
the
the
tricky
part
is
just
getting
the
past
lookups
of
of
matching
an
ip
to
the
asm
and
not
issuing
the
redirects.
I
mean
this
is
to
right.
It's
just
say:
okay,
I
just
want
to
you
know
flatten
that
ip
and
do
the
asn
look
up
as
fast
as
I
possibly
can,
and
that
may
be
something
as
was
suggested,
that
maybe
you
know
turning
it
into
a
dns
service
or
something
else.
A
It's
possible
to
do
that,
but
it's
a
future
in
preview
inside
the
gcp.
So
basically
they
can
amazon.
D
C
So
it's
not
necessarily
asn
data.
That's
why
this
big
process
for
creating
the
list
pulls
from
lots
of
data
sources,
because
the
public
asn
stuff
we're
not
actually
routing
packets,
we're
routing
decisions
on
who's
the
best
and
so
there's
some
logic
in
there.
That's
not
directly
delegateable
to
pre-made
solutions
like
regional
dns
or
you
know
we're
responding
for
geo-location
data
from
a
standard
dns
service.
E
C
Question
is
probably
a
longer
thing,
like
I've
got
thoughts
on
that
for
doing
bgp
with
ips
owned
by
this.
You
know
linux
lf,
right
where
we
have
this
type
of
service,
but
where
this
is.
This
is
like
a
through
two
to
five
year
goal
for
conversation,
starting
because
right
now
the
need
is
that
we're
spending
heaps
of
money
and
we
need
to
plug
the
holes
in
the
ship
so
that
the
boat
can
keep
going
before.
We
start
any
casting
from
all
the
ship.
All
the
all
the
boats.
A
The
other
problem
with
any
cats
is
we
need
to
own.
I
mean
it's,
not
we
need
to
own.
Is
the
the
current
cloud
provider?
Don't
really
allow
any
cast,
because
I
don't
think
it's
something
you
can
do
with
gcp
and
I
don't
think
it's
something
you
can
really
do
with
amazon.
I
think
that's
one
of
the
issue.
If
you
basically
I'm
not
sure,
I
need
to
check
again
well,
it's
all.
It
means
well
yeah.
A
E
Right,
but
you
just
you
just
needed
to
bounce
to,
like
whatever
your
registry
is
right,
you
just
you
have
to
be
able
to
outsource
it
to
whatever
is
hosting
registry
dot.
You
know
whatever
it
it
bounces
at
that
level.
So
you
could.
You
could
outsource
whoever
is
hosting
just
the
dns
to
point
into
the
first
level
of
registry,
so
you
get
listed.
C
C
C
G
G
Because
everybody
cares
about
their
own
thoughts
and
that
just
makes
everyone
feel
a
lot
less
pain.
D
G
A
A
D
Was
yes
now
now?
The
reason
why
we
want
to
do
it
for
now
is
to
plug
the
holes
in
the
ship
and
save
some
money.
So
this
is
just
a
screenshot
of
a
data
studio
report
that
used
all
the
underlying
data
for
since
april
this
month
and
interesting
enough,
the
trend
is
pretty
steady.
You
get
the
same
thing
everywhere.
You'll
see
april
is
a
little
low
for
everybody
and
october,
because
it's
not
full
months,
but
the
rest
of
the
trend
is
roughly
the
same.
D
So
basically
just
explain
what
you're
seeing
on
the
screen
the
top
of
the
the
top
graph
hundred
percent.
That
is
everybody
that
is
all
the
providers
together
amount
of
gigabytes
of
traffic.
So
I
didn't
go
into
money.
This
is
the
amount
of
gigabytes
of
traffic,
then,
for
those
that
have
not
been
in
any
calls
where
we
discussed.
These
trends
is
basically
the
top.
D
So,
there's
not
a
large
amount
of
providers
that
we
have
to
to
go
to
to
do
this
redirects
and
we
can
save
the
payments
we're
trying
to
keep
it
as
anonymous
as
possible.
We
can
save
somewhere
between
40
and
50
of
the
cost.
If
we
redirect,
we
keep
some
within
the
within
the
google
community
infrastructure,
that's
being
funded
that
way
with
four
kubernetes
and
the
rest,
the
other
bits
we
take
to
their
service
providers
and
they
do
their
own
hosting.
Then
we
can
save
the
other.
D
Two
providers
costs,
so
it's
assuming
without
assuming
who's,
who,
we
can
just
say
somewhere
between
40
and
50,
of
cost
saving.
If
we
can
fully
redirect
and
as
we
said
earlier
in
the
call
for
from
episode,
it
is
money
is
nice
to
save
and
we
do
have
a
problem
with
using
too
much
at
the
moment
with
infrastructure
growing
and
demand
growing
up
which
we
do
really
like.
But
it's
also
about
access
for
people
that
do
not
have
a
method
to
get
to
our
artifacts,
with
the
current
structure
so
get.
D
D
The
other
topic
is
the
next
one.
I
have
is
a
very
much
an
iron
question.
He
did
a
great
diagram
of
the
committed
kubernetes
layout.
He
was
talking
about
needing
help.
D
So
the
question
is
just:
is
this?
Okay?
Does
he
still
need
help,
if
not
cool?
I
just
wanted
to
bring
it
here
to
discuss
with
him.
A
So
I
I
met,
I
made
basically
a
a
I
added
an
issue.
So
basically,
what
irwin
was
looking
he's
still
looking
for
is
basically
is
an
open
tool
solution
to
generate
this.
That
maintain
and
generate
this
diagram,
so
help
is,
is
still
needed
for
this.
D
D
A
C
We
have
some
stuff
that
we
use
inside
of
of
org
mode
and
emacs
that
uses
a
similar
structure,
and
I
think
we
talked
about
looking
into
it.
Maybe
after
we
get
the
ingress
controller
stuff,
the
ingress
is
for
these
four
different
solutions
up.
We
can
give
an
example
of
that
to
see
if
it's
worth
pursuing
more
because
that
pipeline
is
used
a
lot.
I
think
it's
a.
C
A
Okay
feel
free
to
to
ping
aaron
directly
or
drop
something
in
the
in
the
hri
post
to
see
how
we
can
improve
the
maintenance
process
for
this
diagram,
because
I
think
it's
still
and
complete
on
some
aspect.
And
we
may
have
to
update
that
in.
A
Okay,
I
have
like
a
last
question
for
last
request
for
your
ep,
so
I
think
I
saw
an
issue
where
you
were.
C
C
C
So
our
repair
instance
can
we
target
the
pair
instance
for
this
other
project.
Maybe
yeah
we'll
look
into
it,
but
we'll
give
you
some
urls
and
we'll
make
sure
and
I'll
make
sure
we
get
added.
I
think
it's
via
email
address.
So
if
you
go
ahead
and
send
me
the
your
packet
username,
which
is
usually
an
email
and
then
I
will
get
them
to
add
the
chairs
of
the
sig
cates
infra
to
that.
C
C
List
right
because
it's
law,
it's
credentials,
it's
logging
in.
So
if
we
want
to
do
that
style,
that's
I
don't
think
we
should
do
that
style,
because
we
need
accountability
for
who
held
that
token
to
get
in
right
and
the
thing
about
the
thing
about
with
with
with
equinix
is
there
is
no
privileges?
You
either
have
full
access
or
no
access?
A
G
C
Account
I
will
make
sure
that
your
admin
on
there
and
also
it's
aaron
right-
is
the
other
chair.
Your
co-chair.
A
C
C
E
Cool
yeah
thanks
guys
nice
to
meet
everyone.
I
would
it
would
be
great
for
me
as
a
new
guy,
if
I
could
schedule
like
an
hour
a
week
as
sort
of
a
co-work
with
somebody
to
work
on
some
new
project.
I
don't
particularly
care
what,
but
you
know
like
an
open
zoom
session
or
something
where
we
just
work
on
something
of
of
your
interest.
I
I
would
find
that
very
helpful
to
kind
of
help
me
get
centered
and
kind
of
learn
the
lay
of
the
hand.
A
Something
I
want
that
for
now.
Yes,
but
I'm
not
sure
that's
my
problem,
I'll
stay.
C
E
A
Okay,
see
you,
everyone
have
a
good
week.
We
can
and
see
you
in
two
weeks.