►
From YouTube: Kubernetes SIG K8s Infra - 20230816
Description
A
Out
hi,
everybody
today
is
August
16th,
and
this
is
the
Sid
Kate
Cinema
meeting
Arno
is
out
on
vacation
this
week,
so
I'm
here
to
assume
the
host
duties.
So
is
there
anybody
who
can
take
some
notes?
Please
thank
you.
Let
me
share
my
screen
and
we
can
get
started.
A
A
B
B
Yeah,
actually,
just
I
just
wanted
to
learn
more
about
this
stage
and
getting
involved
in
it.
So
though,
it's
1
30
in
my
time
it's
feeling
a
bit
sleepy
but
still
thought
about.
Let's
get
started
enjoying
the
meeting.
Yes,
I
am
here.
Thank
you.
Well,
I'm,
looking
forward
to
contributing
this
space,
yeah.
A
Sounds
good:
where
are
you
located.
B
A
A
E
D
A
B
A
A
A
Anything
concerning
here
we
we
haven't
switched
on
anything.
We
haven't
really
Switched
Off
anything
either
Rick
has
been
working
on
moving
the
stuff
and
he
posted
some
statistics.
A
A
Okay,
last
28
days:
okay,
let's
go
to
the
next
one
then
daily,
spend
by
project
anything
interesting
here.
F
The
I
think
there
was
one
more
page
left,
not
this
one
Cloud.
These
two
are
quite
new,
but
yeah
this
one
here.
A
In
the
probability,
okay,
that's
about
it,
we
don't
have
folks
who
can
show
us
the
AWS
cost.
So
that's
fine,
okay,
I'm
gonna
stop
now,
unless
people
have
some
specific
things
that
they
want
to
look
at
one
twice:
okay,
okay,
now
any
action
items,
reviews
I,
don't
see
any
action
marked
from
the
previous
meeting.
We
skipped
one:
did
we
Skip
One?
No,
we
didn't
Skip
One.
A
Now,
let's
go
to
the
next
one,
which
is
open
discussion
who
wants
to
start
off
Patrick.
You
have
the
floor.
D
Okay,
I
can
just
briefly
discuss
the
topic,
so
I
left
a
red
some
time
ago
on
slack
on
our
Kate
sequeline
Pro,
and
it
was
about
how
do
we
want
to
manage
githubs
in
between
our
clusters?
So
basically,
we
have
two
clusters
calorie
cluster,
in
which
we
test
things
and
the
production
one
which
is
a
build
cluster
that
is
connected
to
our
Pro.
D
So
basically
we
don't
expect
to
have
more
clusters,
so
we
are
not
looking
for
anything
highly
scalable,
but
we
want
to
yeah.
We
want
to
have
everything
in
GitHub,
so
I
listed
here
free
Solutions,
how
we
want
to
keep
our
manifest
so
that
we
can
yeah.
Basically,
the
first
one
is
about
duplicating
the
the
Manifest,
which
is
like
the
the
ugliest
one,
but
at
the
at
the
same
time
the
the
easiest
one
to
extend,
because
it
doesn't
require
any
knowledge
of
tools
like
customize
or
helm,
the
other
one.
D
This
is
the
one
I
started
with
the
customized
I
had
some
PR,
but
then
we
were
discussing
with
Arnold
that
that
might
be,
but
to
confusing
for
people
I
mean
we
were
losing.
The
idea
of
you
know
having
it
very
simple
because
yeah
this
this
requires
basically
knowing
how
customers
work
and
the
reaper
structure
is
specific
and
so
on
and
the
last
one
it
was
not
implemented.
But
we
could
also
leverage
like
an
umbrella,
Helm
charts
to
wrap
everything
and
then
provide
dedicated
values
for
each
of
two
clusters.
D
So
yeah
I
mean
I,
don't
have
strong
opinions
about
any
of
those
options.
I
think
the
first
one
is
the
easiest
one
for
someone
to
contribute
to
start
contributing
because
it's
as
simple
as
like
creating
file,
and
then
it
should
be
deployed
after
it's
merged.
The
question
is
about.
D
G
D
Basically,
we
introduced
it
because
we
were
testing
different
things,
but
before
we
we're
deploying
it
I
mean
this
is
the
first
time
we
were
working
with
eks,
so
we
wanted
to
have
something
to
test
and
we
didn't
want
to
use
production.
That's
why.
D
It's
convenient,
the
cost
is
not
that
high.
We
don't.
We
didn't
connect
to
brow.
So
it's
also
not
very
useful
in
terms
of
testing
jobs,
but
for
testing
some
eks
specific
things.
It's
pretty
convenient,
but
I
mean
we
could
probably
scale
it
down,
but
still,
if
we
want
to
have
it
be
able
to
deploy
it
at
any
time,
we
would
still
need
to
have
some
place
for
those
manifests.
G
Yeah
I
I
think
right
now
we
have
things
either
like
Cube
cuddle
applied
by
a
proud
job,
with
the
appropriate
info
for
things
that
are
running
in
the
cluster
or
we
have
terraform,
hey,
I,
don't
know,
I
I
feel
like
we
should
be
able
to
get
confident
enough
with
eks
to
just
use
it
and
to
roll
out
changes
to
the
CI
cluster
and
we
shouldn't
be
rolling
out
a
lot
of
changes
other
than
like
version
upgrades
eventually
or
maybe
new
node
pools,
I.
G
A
So
Ben,
whatever
you're
saying
I'm
interpreting
it
as
we
don't
need
to
do
two
or
three
we
can
just
do
one
and
like
and
we
we
can
even
go
further
and
say
we
can
just
tear
down
the
Camry
at
some
point.
G
I'm
not
as
familiar
with
eks
but
I
think
we
have
folks
from
eks
here
and
naively
I
would
assume
we'd
be
able
to
get
it
to
a
point
where
it's
similar
to
the
gka
cluster,
where
we
like,
we
have
to
do
version
upgrades
or
occasionally
we
tweak
something,
but
most
of
the
time
it's
just
running
and
we're
not
making
substantial
configuration
changes.
A
A
I
can
add
so
Arno
here.
D
If
I
can
add
something
to
the
complexity,
so
also
I
mean
you
mentioned
upgraded
so
on,
but
we
also
leveraged
terraform
to
provision
it.
So
testing
transform
script
itself
was
also
nice.
I
mean
it
was
nice
to
have
category
to
test
changes
before
applying
or
production.
Yeah
I
know
that
terraform
is
now
like
a
hot
topic,
but
let's
leave
that
aside.
A
Yeah,
it's
it's
yeah,
don't
want
to
digress
into
the
license
stuff,
but
you
know
for
now:
let's
say
we're
going
to
do
with
do
go
with
one
right:
we're
not
going
to
change,
we
can
do
a
diff
and
we
can
apply
things,
but
there
is
duplication
and
let's
have
a
separate
conversation
about.
A
When
can
we
turn
down
the
calorie
cluster
and
how
much
is
it
costing?
We
can
have
that
as
a
separate
discussion.
Okay,.
G
G
We
have
a
lot
of
existing
automated
deployments
and
things
I
I'm,
not
sure
I
would
want
to
encourage
adding
another
CI
tool
like
we
already
have
to
maintain
proud
for
ourselves,
and
we
already
use
terraform
and
Bash
or
whatever,
and
to
the
point
of
terraform
being
a
Hot
Topic.
It
is,
and
we
have
lots
of
existing
usage
that
like
if
that
becomes
a
problem.
You
know
this
wouldn't
be
any
different.
We
have
a
bunch
of
other
places
that,
if
that
needs
addressing
it
will
get
addressed.
G
I
think
we
should
probably
try
to
standardize
the
Sig
a
little
bit.
Maybe
there's
some
argument
for
an
experiment
it,
but
I
would
say
only
if
we're
expecting
to
like
move
everything
to
that
and
treat
it
as
such.
Otherwise,
we
should
just
use
what
we've
been
using.
A
Yeah
sometimes
I
feel
then
that
you
know
the
people
doing.
The
work
should
make
those
calls
too.
So
you
know,
let's
hold
off
on
that
till
Arno
gets
back
and
we
can
talk
about
it.
Next
time
is
that
okay,
Patrick
yeah.
A
For
now
you
have
you
know,
let's,
let's
not
do
the
two
and
three
and
let's
go
with
you,
know
the
simplest
option:
okay,
okay,.
F
Eks
has
a
lot
more
knobs
and
extra
things
you
have
to
do
on
that
cluster.
This
is
how
it
goes
right.
So
if
you
lose
the
canary
cluster,
you
just
have
to
do
things
on
podcaster.
That.
F
Okay
and
back
to
Cluster
management,
really,
we
should
be
using
get
Ops
the
current
process
that
we
do
on
the
Google
own
clusters
and
the
community
cluster
on
gke
is
in
Gray.
A
F
So
the
the
current
approach
that
we
use
to
manage
cluster
State
on
the
GK
clusters
isn't
great.
We
really
should
adopt
flux
and
use
the
same
approach
doing
keeps
his
applied.
Merges
it's
not
really
cool
and
yeah
I.
Think
we've
already
explained
why
that's
bad.
G
Sure
I,
just
I
just
wanted
to
suggest
that
we
probably
shouldn't
grow
more
options
like
if
we
think
that
this
is
a
good
path
for
then
we
should
be
looking
to
to
fix
to
apply
that
everywhere,
based
on
this
going
well,
but
just
that
then
we're
going
into
this,
expecting
that
it's
an
experiment
and
that
we
plan
to
standardize
on
it
if
it
goes
well,
so
we
don't
wind
up
maintaining
like
in
ways
of
installing
to
the
cluster
or
whatever.
A
Okay,
let's
call
it
an
end
to
that:
okay,
who's
up
next.
H
Yeah
hi
everyone.
This
will
be
quick
because
I
just
want
to
bring
this
document
your
attention.
H
Our
backgrounds
I've
been
having
some
conversations
with
GV
last
couple
of
months
and
I
joined
some
of
the
kubernetes
game
from
meetings
and
in
addition,
that
I've
been
looking
for
a
big
relief
Shadow
program
as
well
in
the
past.
I
actually
want
to
apply
to
it
and
during
my
conversations
with
TV
we
said,
would
it
make
sense
to
have
a
conversation
around
possibility
of
having
a
similar
program
for
the
kubernetes
Sig
infra
as
well,
and
then
the
easiest
way
we
said
like?
H
We
can
perhaps
create
a
proposal
and
share
it
with
the
community
to
get
a
preliminary
feedback
see
if
it
is
a
viable
path
forwards,
and
if
it
is,
we
can
perhaps
put
more
time
on
it.
But
today,
I
just
want
to
share
the
comment
with
you
and
get
your
initial
feedback
coming
weeks.
And
then
we
can
bring
this
topic
again
and
have
a
longer
conversations
if
it
is
deemed
valuable.
A
I'm
very
supportive
of
this.
We
need
to
do
this
and
you
know
the
information
and
knowledge
gets
pulled
and
it's
it's
hard
to.
We.
We
end
up
relying
on
a
few
people
to
do
this
for
various
reasons
car
you
know
earning
Karma,
but
also
you
know
making
sure
that
only
a
few
people
can
touch
the
really
sensitive
stuff
and
things
like
that.
But,
yes,
we
need
to
grow
the
pool
of
people
who
who
can
come,
come
on
board
and
help
us
out.
A
You
know
like
I'm,
really
thankful,
for
you
know
Mario
Patrick
and
folks,
like
that
who
have
shown
up
here
and
are
doing
the
work
and
keeping
things
going
so
I
think
we
should
learn
from
their
experience
as
well
to
see
you
know
how
was
their
onboarding
and,
like
you
know
what
kind
of
hiccups
you
know
how
many,
how
many
times
did
they
have
to
wait
on
X,
Y
or
z,
to
you
know
and
like
what
what
was
missing
and
what
did
they
have
to
learn
the
hard
way.
A
I
think
there's
a
lots
of
lessons
that
that
we
can
learn
from
their
experience
that
we
can
roll
into
this.
So
I'm
definitely
supportive
from
from
hey.
We
need
to
do
more
of
this.
We
need
to,
and-
and
one
one
caveat
that
I
have
is
people
the
people
who
are
coming
in
need
to
know
that
this
is
like
a
long-term
commitment
and
not
like
they
can't
just
vanish.
A
They
can't
do
hey.
This
is
a
three-month
program.
I'm
gonna,
be
here
for
three
months
and
then
I'm
gonna
We
invest
time
and
effort,
and
it
is
only
useful
if
people
are
going
to
be
sticking
around
for
a
long
time.
Yes,
individuals
might
have
so
this
that's.
Why,
like
I,
hesitate
to
do
like
a
LFX
mentoring
program
around
this,
because
then
that
is
like
calendar
time
specific
time,
and
you
know
people
are
gone
as
soon
as
that's
on
same
with
the
release
team.
A
All
right,
there's
absolutely
no
guarantee
that
people
are
going
to
be
around
so
I.
Don't
know
how
to
get
commitment
from
folks
that
they're
they're
going
to
be
around
that
that
and
we
will
have
to
expect
attrition
as
well.
Just
because
you
know
life
comes
at
you
fast
right
and
you
know
things
change
as
well,
so
I'll
stop
talking
and
let
others
have
a
turn
go.
Go
ahead.
Bob.
I
I
I
dropped.
This
comment
like
in
the
chat
but
like
for
at
least
the
first
pass.
You
want
to
do
something
I
highly
recommend,
limiting
it
to
org
member,
only
that's
at
least
past
some
bar
of
getting
into
the
project,
and
it
shows
some
level
of
commitment.
I
This
will
probably
be
less
of
an
issue
too
in
terms
of
potential
people
that
aren't
as
committed,
because
we're
going
to
be
being
more
aggressive
with
the
like
org
cleanup
and
we've
gotten
more
aggressive,
I
shouldn't,
say
more
aggressive,
but,
like
you
really
have
to
demonstrate
more
of
a
commitment
before
becoming
an
org
member.
Now.
G
Yeah
I
I
share
those
concerns
and
thoughts.
I've
participated
in
a
bunch
of
different
iterations
on
kubernetes
trying
to
Mentor
groups
before
and
usually
the
problem
is
that,
if
you
start
with
inviting
totally
net
new
people
in
you
put
a
lot
of
energy
into
it
and
then
they
disappear.
G
I've
thought
about
this
problem
a
lot
and
it
seems
like
more
or
less
like
what
Bob
is
saying.
The
thing
that
the
thing
that
seems
to
work
is
just
sort
of
observing
people
that
have
been
repeatedly
showing
up
and
then
starting
to
like
help
them
move
up.
I
think
we
could
do
with
more
of
a
program
for
in
a
structured
way
like
you've,
been
people
who've
been
participating
in
the
Sig.
G
We
want
to
like
guide
them
into
helping
own
some
of
the
things,
but
I
I
do
think
that
we
should
probably
limit
it
to
folks
that
that
have
been
showing
up
and
for
people
that
are
new.
You
can
start
like.
G
We
certainly
want
you
to
come
and
help
with
things,
and
we
should
try
to
try
to
leave
some
things
to
do,
but
until
until,
like
I've
seen
the
commitment
that
you're
gonna
keep
coming
and
helping
it's
hard
to
set
aside
time
to
to
like
one-on-one
meet
with
people,
I
think
most
of
the
people
that
do
own
these
things
today
are
like
dims
and
have
extremely
busy
calendars
and
there's
a
big
trade-off
and
where
that
time
gets
spent.
B
G
Think
that
if
you
can
get
people
to
to
to
show
that
they're
like
committed
and
invested,
that
we
definitely
have
a
gap
in
following
up
with
those
people
and
and
making
sure
that
we're
moving
them
into
to
owning
things
and
whatnot,
and
there
is
room
for
actually
mentoring.
Sig
testing
did
open
call
for
chairs,
but
we
kind
of
did
somewhere
in
between
this,
where
we
didn't
really,
it
was
kind
of
a
very
low
time
commitment
thing
until
folks
had
been
participating
for
a
while.
I
One
other
quick
comment
on
actually
hitting
org
Members
First
for
potential
candidates.
The
most
recent
customized,
like
Mentor
cohort,
was
limited
to
order.
Member
only
and
there
was
like
I
think
it
was
like
around
100
people
applied
and
they
got
it
down
to
I
think
it
was
around
25
candidates,
but,
like
all
of
those
actually
looked,
you
know
pretty
solid
for
going
into
this,
and
most
of
those
people
do
understand
the
type
of
types
of
things
being
asked
of
them.
H
It's
just
it's
some
of
the
comments
you
made.
Actually
these
are
things
that
came
up
during
my
conversations
with
GP
as
well
the
old
member
topic
and
I.
Think
like
again,
even
though
I'm
learning
about
Community,
it
makes
a
lot
of
sense
to
me
because
infrastructure
things
are
sensitive.
Things
like
it
requires
people
to
you,
know,
have
access
to
certain
resource
and
which
might
have
you
know
risked
or
Community,
as
I
saw.
That
is
something
I
can
work
with
GV
to
incorporate
into
the
proposal.
H
In
addition
to
that
about
the
time,
commitment
and
this
proposal
was
heavily
influenced
by
the
release
Shadow
program,
but
because
of
that
reason
we
actually
put
a
minimum
two
release
Cycles.
As
you
know,
the
term
for
the
shadow
shuttles
that
can
be
adjusted
as
well
to
see
you
know,
the
first
cycle
could
be
to
learn
the
community,
the
people
who
are
doing
good
work
and
complete
on
going
up
and
upcoming
work,
so
they
can
become
there
with
all
these
things
and
during
the
second
release
cycle,
they
can
actually
perhaps
start
becoming
more.
H
You
know
into
the
details
and
start
and
some
contributions
because
it
is
like
challenging
to
you
know
even
to
understand,
what's
happening,
what
is
used,
what
is
required
and
so
on.
So
that
is
what
we
put
there,
but
that
can
obviously
be
changed
in
addition
to
that,
we
put
something
called
area
pokes
based
on
the
conversations
we
have
been
observing,
like
public
clouds,
reporting
and
metrics
they
to
cost
management,
GIC
icdm
testing.
H
So
this
could
be
different
types
of
roles
based
on
you
know
the
interesting
individuals
and
their
you
know,
competence
and
experience,
but
what
I
am
hearing
this
made
sense,
so
we
can
continue
working
with
this
and
obviously
we
need
to
sort
of
conversations,
but
it's
great
to
hear
positive
feedback
about
this
I'm.
Sorry,
mentioning
I,
see
your
Hunter.
G
I
want
to
point
out
a
little
bit
more
that
like,
for
example,
you
mentioned
release
Cycles,
but
it's
worth
noting
that
for
a
lot
of
our
infrastructure,
we
aren't
even
necessarily
touching
it
on
a
day-to-day
basis
and
certainly
not
coupled
to
the
release
cycle
other
than
maybe
we
might
avoid
touching
things
when
it's
late
in
a
release.
G
The
release
Shadows
really
has
the
benefit
of
there's
like
very
structured,
regular
things
that
need
to
occur
versus,
like
you
know,
back
when
we
were
still
manually
applying
DNS.
That
was
just
ad
hoc
as
needed
when
there
is
a
new
entry
and
now
that
something
like
that
is
automated.
The
few
people
that
have
access
to
it
are
rarely
touching
it
I
I'm
a
really
big
proponent
of
in
most
cases.
G
If
you
want
to
document
something
like
how
to
work
on
a
project,
you
should
you
should
document
it
instead
of
instead
of
having
meetings,
because
then
everyone
can
benefit
from
it
and
it's
a
it's
a
Perpetual
resource
so
like,
for
example,
in
kind
when
I
wanted
to
grow.
Maintainers
I
spent
my
time
on
writing
a
contributor
guide
and
documenting
some
issues.
G
H
Yeah,
that's
yeah.
You
need
to
put
something
there,
so
we
kind
of
told
like
okay,
if
you
can
sync
this
program's
applications
with
release,
Cycles
done
yeah
avoid
you
know,
onboarding
people
when
the
release
comes
in,
so
the
applications
start
after
the
release
is
done.
It's
that
way.
It's
not
just
like
happening.
Vanderlies
happens
to
the
initial
starts,
it's
kind
of
to
protect
the
release
work
by
not
introducing
or
unnecessary
changes
to
infrastructure,
but
yeah
that
that
is
up
for
conversation.
C
Would
it
not
also
be
possible
to
have
like
that?
You
say
we
have
these
mental
programs
that
that
are
around
and
are
kicking
up
like
once
a
year,
or
something
like
this-
that
you
basically
use
this
as
a
as
a
generalized
time
frame,
so
to
decouple
it
basically
from
the
from
the
release
cycle,
because
the
S
Ben
already
said
I
mean
if
the
Elisa
is
coming
up.
C
Nobody
wants
to
change
anything
right
so,
but
I
mean
we
have
those
those
Mentor
cycles,
and
maybe
it's
better
to
align
it
to
this
part,
so
that
we
have
because
I
agree
that
we
need
to
have
like
a
regular
reoccurring,
possibility
to
put
people
into
the
shadow
program
right.
A
Right
other
than
you
know,
documenting
some
stuff
that
is
in
your
head,
if
at
all,
there's
anything
left
if.
G
Someone
comes
to
me
and
says:
hey
I,
don't
understand
what's
going
on
here,
there
isn't
enough
info,
then
what
I
want
to
do
is
write
down,
whatever's
missing
and
in
that
particular
case
I
think
it's
pretty
well
written
down,
but
I
know
there
are
other
parts
of
the
org
that
aren't
and
maybe
there's
other
parts
of
sick
case
and
for
where
people
are
doing
things
really
frequently
like
maybe
some
of
the
cluster
work
or
something
but
the
bits
that
I
participate
in.
G
A
Yeah
we
allowed
to
figure
out
I
think
we
need
we'll
need
to
start
with
figuring
out
like
what
kind
of
documentation
that
we
need
done,
which
is
missing.
You
know
we
we
had
to
go
back
and
revisit.
You
know
activities
that
we
have
left
on
the
wine
that
we
need
to
pick
up.
A
You
know
before
they
come
in
right
like
so.
We.
A
Unprepared
right
like
when
we
know
so
I'll
roll
this
out
to
people
like
they
show
up.
You
know,
I
I
need
a
set
of
things
that
we
can
distribute
to
people
and
like
have
some
timing
commitments
from
the
folks
on
the
call
here
who
can
help
with
coaching
and
talking
those
kinds
of
things
as
well.
G
Yeah
I
feel
like
the
biggest
thing
that
I
need
like
going
back
to
the
registry.
Again
is
just
so
it's
just
some
isn't
necessarily
something
to
teach
people
I
think
most
people
have
enough
that
work
in
a
space
have
enough
experience
to
poke
at
these
things
or
or
it's
documented.
It's
the
building,
the
trust
with
people
that
just
requires
regularly
showing
up
I'm,
not
sure
how
to
do
more.
For
that.
G
But
wherever
possible,
we've
tried
to
automate
things
and
and
not
have
people
taking
manual
actions,
because
that
scale's
better
and
that's
something
that
I've
even
have
feedback
for
the
release
process
like
we.
Finally,
automated
the
release
fast
forward
like
building
things
so
that
there
aren't
so
many
humans
and
that
it
is
just
things
and
git
and
anyone
can
can
read
and
interact
with
that,
and
we
just
need
a
few
people
with
break
glass.
We
probably
don't
actually
want
to
substantially
grow
that
group
because
of
the
sensitive
access.
A
Here,
okay,
I
think
I.
Think,
like
you
have
some
feedback
that
you
can
incorporate,
then
we
can
talk
again.
We
can
even
do
an
offline
review
of
the
doc
one
more
time
and
and
get
something
for
you
and
GP
to
think
about
more
okay.
In
any
other
open
discussion.
F
I
want
to
add
something
to
that.
So
let's
say
we
approve
this
plan
right
and
we
find
someone
that
wants
to
do
something
from
the
first
of
September
right.
Do
you
have
an
engineering
backlog
of
things
for
them
to
work
on
yeah.
A
That's
what
I
was
saying
right
like
what
are
the
activities
that
we
need
to?
If
we
have
to
make
that
so
we
can,
we
can't.
We
can
definitely
not
do
September
right,
like
the
earliest
we
can
do
is
probably
you
know
when
all
of
us
come
back
from
the
Christmas
New
Year
vacations,
so
best
case
scenario
would
be
second
half
of
Jan
right
like
that's.
A
When
we
can
start,
then
the
question
is
like:
what
do
we
need
to
do
before
that,
so
when
people
do
show,
when
we
do
roll
out
a
program
and
announce
to
people
that
hey?
Secondly,
you
know
second
half
of
January,
we
are
going
to
do
applications
and
then
February
we're
gonna
get
started
with
this
program
and
like
so,
we
need
to
figure
out
like
how
many
people,
what
can
we
do
with
them?
What
are
the
activities,
so
we
have
to
go
about
thinking
about
this.
A
In
you
know,
holistic
fashion,
I
think.
A
Also
I
think
we
had
a
bunch
of
things
and
you
know
I
think
we
are
wrapping
up
a
bunch
of
things
now
right,
like
so
yeah,
so
we'll
probably
have
to
think
about
like
what
is
the
next
set
of
things
that
we
want
to
think
about
right
in
some
ways
you
know
when
Mario
and
Patrick
came,
it
was
like.
Oh,
we
are
doing
eks,
which
is
new,
so
it
was
easier
to
you
know
let
them
free
on
some
new
areas
where
we
needed
work
done.
A
A
So
I
do
know
we
have
anybody
else.
Has
anything
else
to
talk
about.
E
Yes
sure,
hello,
everyone,
I'm
the
open
source
program
manager
at
AWS
and
I'm
the
one
in
charge
of
the
karate
program,
probably
I,
should
bring
good
news.
First,
the
20,
the
250k
credit,
has
already
been
loaded
to
the
account.
So
we
really
appreciate
the
great
work
that
whole
team
is
doing
for
migrating.
E
The
kubernetes
infrastructure
to
Amazon
and
then,
like
I,
would
love
to
work
with
the
community
closer
to
really
understand
our
whole
plan
of
the
workload
to
AWS,
so
that
we
can
have
better
understanding
about
budget
and
then
that's
the
reason.
I
put
together
the
estimation
that
I
sent
out
one
two
weeks
ago,
and
then
there
were
like
3.5
pointed
out
and
I
think
the
biggest
puzzle
to
me.
There
actually
is:
do
you
want
me
to?
Does
it
have
like
a
question?
Does
anyone
has
have
you
already
seen
the
email
I
shared.
A
I
think
you
should
do
you
want
to
screen
share.
E
I
can
do
that.
Okay,.
A
So
there
was,
she
has
done
some
analysis
on
the
numbers
that
we
have
and
the
trends
that
we
have
and
things
like
that
so
hoping
we
can
have
a
little
bit
of
a
discussion
around
that.
E
Oh
like
it,
if
we
look
at
this
area,
there's
the
raw
data
from
January
to
June,
that's
the
detail,
invoice
information
I
put
there
and,
like
I,
said
all
those
invoices
they
were
sent.
I
think
that's
to
the
infrastructure
account.
So
if
you
want
to
look
into
the
invoices,
you
should
have
access
somewhere.
A
Yeah,
so
these
numbers
are
from
What
the
invoices
yeah,
the
GP
and
hippie.
They
have
access
to
these
numbers
and
you
got
it
from
them
right.
Yes,
okay,.
E
And
then
well,
like,
like
I,
said
I
think
the
biggest
puzzle
to
me.
Actually
is
this
number
the
ec2
AWS
ECT
amount
because
it
was
tripled
kind
of
from
March
to
February,
and
then
it
dropped
back
in
April
and
then,
like
the
consistency,
actually
continues
from
a
protein.
So
to
me,
this
one
is
the
biggest
puzzle
and
then
I
wonder
where
or
what's
the
factor
that
drives
there
and
also
it'll
be
great
if
everyone
can
shine
some
lights
here,
as
we
do.
The
analysis
like.
E
A
Yeah,
unfortunately,
hippie
is
and
Company
they
are
not
here
today,
so
we
might
not
be
able
to
have
that
conversation
yet.
But
anybody
have
thoughts
on
like
why
number
look
a
little
skewed
on
March
for
the
ec2
I,
don't
remember
what
we
were
trying
or
why
it.
C
Went
so
high,
I
I
think
this
was
the
time
when,
when
Marco
and
Patrick
started
moving
the
pro
and
we
we
we
played
around
with
different
machine
types,
because
we
had
issues
there
and
I
know
that
we
that
we
scaled
up
different
different
machine
sets
so
I
think,
and
then
we
did
some
cost
optimization
where
we
said
like
we
can
use
specific
machine
types
better.
That
fits
the
cluster
and
this
was
around
March
to
April.
If
I
ever
remember
correctly,.
A
Okay,
so
that
that
that
is
the
answer
yeah,
so
you
had
two
more
questions
right
like
can
you.
E
Yeah,
the
other
two
action
is
across
around
like
if,
if
that's,
if
it's
necessary
for
us
to
continue,
the
other
two
is
about
ECS
and
cost
optimization
on
call
on
cloud
watch.
A
A
A
Can
you
show
us
what
the
numbers
were
for
ECS
like
how
the
numbers
ECS.
E
Ecs
actually
grows
very
fast,
like
a
start
from
around
30,
but
in
June
the
cost
is
5.6
K,
okay,
overall.
A
A
E
That
was
confusing
and
then
next
is
about
Cloud
watch
like.
A
Yeah,
do
we
use
cloud
launch
metrics
or
Prometheus
metrics?
This
is
for
Patrick
and
Mario.
D
I
think
at
some
point
we
enabled
container
insights
to
investigate
the
node
issue,
so
I
mean
I,
don't
think
it's
like
that
much
yeah
I
mean
if
it's
too
much,
we
can
disable
it
at
any
point.
Yeah.
A
The
number
is
not
that
much,
so
we
should
be
fine,
I.
Think
the
the
total
itself
is
like
you
know,
not
too
much
yes,
but
we
we
can
watch
the
trend
and
see
where
it
goes.
But
in
to
your
question
what
is
coming
next,
we
have
two
two
sets
of
things
coming.
One
is
the
scalability
test.
The
scalability
tests
are,
we,
we
are
still
stabilizing
it.
We
have
one
CI
job
that
is
doing
500
node
testing
and
that
is
I'm
gonna.
A
Once
we
have
it
ready,
then
we're
going
to
turn
it
on
periodically
and
that's
when
that
money
is
gonna,
it's
gonna
burn
up
money,
so
that
is
one
the
other
one
is.
We
do
have
a
new
test
Suite
that
we
are
experimenting
with,
but
that
might
come
later.
So
scalability
is
going
to
come
first
and
then
some
other
things
are
going
to
come
after
that.
A
So,
like
I,
think
Arno
mentioned
that
we
are
constantly
turning
things
on.
So
we
do
expect
this
to
go
up
by
year
end,
but
it
is
still
under
a
million
for
the
year.
So
you
know
I'll
be
happy
if
you
push
it
to
just
past
million,
but
you
know
that
that's
probably
not
going
to
happen
I'd,
but
you
know
we
have
to
check
again
in
a
few
months.
E
A
No,
it's
some
of
the
work
is
being
done
by
the
eks
team
folks,
but
then
we
we
are
bouncing
between
the
community
and
the
eks
team
and
the
AWS
support,
because
we
are
tweaking
you
know
the
quotas.
We
are
tweaking.
A
A
Thanks
a
lot
for
sharing
the
information
and
coming
and
talking
to
us
like,
let's
try
this
one
more
time
when
other
people
are
hippie
and
those
folks
are
GV
and
other
folks
are
here
as
well.
So.
B
E
A
Thanks
any
other
topic
of
discussion
from
anybody
else,.
G
I
I
there's
one
thing
that
we
need
to
look
at
in
the
future,
so
we
rolled
out
the
fastly
right.
I
just
want
to
see
these
people's
minds.
We
should
not
have
a
long
discussion
here
and
and
we're
not
going
to
solve
it,
but
something
for
people
to
be
thinking
about,
and
maybe
helping
at
some
point
when
our
note
gets
back,
our
no
primarily
launched
the
fastly
CDN
and
front
of
the
release
bucket.
G
However,
best
I
can
tell
it's
basically
just
a
pull
through
proxy
right
now
that
isn't
caching,
anything
that
is
going
to
have
to
be
fixed
if
we're
going
to
actually
migrate
to
fully
Community
bucket.
Behind
that
we
will
not.
We
will
want
that
to
actually
be
cached
from
what
I
can
tell
it's,
not
being
cached,
so
that's
going
to
need
fixing
and
then,
when
that's
fixed,
we
still
need
to.
G
G
Currently,
what
I've
discussed
with
Folks
at
Google
is
just
making
it
read
only
and
leaving
it
in
the
interest
of
not
breaking
people,
but
it's
something
we
should
do
at
some
point
and
I
took
a
little
look
and
I
think
we
can't
because
I'm
pretty
sure
all
of
the
traffic
is
just
getting
passed
through
still
yeah.
A
Let's,
let's
open
up
an
issue
that
it
is
just
being
passed
through
Ben,
and
so
we
don't
forget
and
when
Arnold
comes
back
he
has
some
something
to
chew.
On
I.
G
Don't
I
do
not
have
good
visibility
and
I
can't
tell
you
like
how
much
of
that
is
because
it
because
people
are
just
hitting
the
bucket
directly
but
yeah
like
squinting
at
the
at
what
graphs
I
do
have
I,
don't
think.
We've
cut
the
bandwidth
at
all
and
just
the
behavior
of
the
end
point
and
the
discussions
around
the
size
of
things
so
I'll
go
file
an
issue
in
case
IO,
but
that
might
be,
that
might
be
a
fun
project.
To
look
at
I.
G
Do
want
to
wait
until
our
nose
back,
because
pretty
much
everything
we
deployed
there
is
is
right.
A
And
I
do
want
to
see
like
you
know
how,
if
people
are
using
the
new
Deb
and
RPMs,
and
you
know
how
is
that
going,
you
know
we
will
have
some
numbers
by
then
right,
like
the
release
was
just
yesterday
so
Rick,
do
you
want
to
share
some
some
of
your
numbers?
Do
you
want
to
screen,
share
and
show
us
how
far
you've
gotten
and
if
you
have
any
blockers-
or
you
know
something
to
talk
about.
A
Okay,
because,
probably
not
there
right
now,
so
if
there
isn't
anything
else,
thanks
a
lot
everyone
and
see
you
next
time,
bye,
bye,.