►
From YouTube: Kubernetes SIG K8s Infra - 20230412
Description
A
A
Thank
you
before
we
jump
into
the
different
item.
Do
we
have
new
people
in
the
car
that
want
to
introduce
themselves.
A
Which
is
a
huge
Improvement
science,
the
beginning
of
the
year
and
I?
Think
with
that
we
basic
I
think
we
can
officially
announce
we
have.
We
now
have
to
cause
fire
and
then
control.
We
may
not
go
out
of
project
by
end
of
the
year
if
we
are
not
doing
anything
wrong
or
accidentally.
A
C
C
A
A
Oh,
no,
sorry,
no!
It's
basics!.
A
A
It
will
see
the
difference
like
so
last
month
in
I
would
say
13
days.
We
were
basically
going
over
one
hundred
and
thousand
dollars
just
that,
and
now
this
one
just
11
days
including
weekend.
We
just
go
by
half
of
that
and
that's
like
that's
what
I'm
saying
yeah,
because
that's
a
huge
win
for
us,
so
I'm
assuming
by
end
of
the
month,
will
be
around
150
000
of
monthly
expense.
A
Starting
this
month
and
going
for
the
end
of
the
year,
so
that's
basically
keep
us
under
budget
by
I,
think
30
or
something
like
that.
We
need
to
work.
I
didn't
do
the
calculation
so
I
need
to.
We
need
to
come
back
on
that
first
meeting
of
May
so
as
like
doing
analysis
in
April
and
see
how
we
can
have
a
forecast
for
the
rest
of
the
month,
but
I
think
we
ballpark
calculation
I,
think
we.
We
are
fine
right
now.
D
Online
building,
if
we
just
you're
asking
about
where
we're
going
to
go,
if
we
skip
over
to
the
graphs
according
to
the
calculation,
if
we
keep
burning
at
the
rate
that
we've
earned
last
week,
we
will
be
at
2.2
by
the
end
of
the
year.
D
A
We're
going
to
go
even
even
with
this
man
we
are,
we
are
we're
going
to
be
under
2.2,
Millions
right
end
of
the
year,
yep,
so
I
think
I
think
I
would
say
the
burning
red
which
is
changing
every
month.
Basically,
so
this
is
like
based
on
Walgreens,
but
I
think
we
should
just
have
a
calculation
based
on
the
month.
D
A
B
A
Okay,
okay,
so
I
think
we
can
drop
to
the
open
discussions.
The
first
one
is
me,
so
just
go
in
quick
and
say:
I
put
a
documents
for
honorary
Pro
feel
free
to
put
any
data
you
have
on
that.
A
A
Okay
next
next
is
next
item
is,
like
me,
also
I
forgot
to
put
my
name,
but
it's
some
kind
of
foi
about
this.
We
have
a
walking
session
next
week
during
computer
Summit,
so
April
18,
it's
us
by
Jiffy
is
the
one
running
that
I
put
the
location,
their
title,
I'm,
not
sure
I'm
gonna
be
there
because
I'm
doing
operation
for
computer
Summit,
so
I'm
supposed
to
run
another
room.
A
B
Yeah
I
figure
truly
worst
case
laptop
and
zoom
and
then
we'll
yeah,
we'll
make
do
I
did
send
an
email
out
to
the
Kate's
infra
mailing
list
as
well.
So
if
you're
there
and
you're
not
doing
anything,
show
up.
C
Thank
you.
Yes,
so
I
actually
got
pinged
by
someone
from
cloud
run.
Who
was
interested
in
how
like
hybrid
devrel,
who
is
interested
in
how
we
are
using
registrykits.io
and
I,
think
they
wanted
to
publicize
that,
and
it
occurs
to
me
that
we
do
not
currently
publish
any
of
those
metrics.
So
I
was
wondering
if
we
should
publish
some
of
those
metrics
into
like
the
safe
ones
into
which
is
basically
not
the
logs
publish
the
metrics
into
bigquery.
So
we
can
report
on
them
from
looker.
There's
no
easy
integration.
C
It
basically
is
going
to
be
a
a
Cron
job
or
a
cloud
scheduler
job,
but
there
are
some
samples
and
my
feeling
is:
if
we
eventually
decide
that
we
like
something
better
than
bigquery
and
looker,
which
I
I
am
eager
to
see
that
day,
then
we
should
fairly
be
fairly.
It
should
be
fairly
simple
to
move
to
that
that
thing,
but
yeah
I
wanted
to
know
whether
people
would
be
want
to
see
these
metrics,
whether
we
think
we
should
be
publishing
them.
That
sort
of
thing.
A
I
think
I
think
the
the
the
quick
answer
is
yes.
Now
the
question
is
about
how
I
mean
there
are
measurable
option
here.
I,
don't
have
a
strong
opinion
about
this,
and
it's
kind
of
weird
like
Cloud
running,
is
asking
for
this,
because
I've
I
have
a
lot
of
feature.
Requests
about
metrics
related
to
crowd,
run
so
well,.
C
I
mean
that's,
that's
I,
think
the
the
great
thing
is
if
we
open
this
conversation,
that
would
be
a
good
conversation,
a
good
channel
to
have
open
this
is
devrel.
So
you
know
it's
not
like
it's
the
the
cloud
run
PM,
but
it's
at
least
like
one
step
closer
to
having
that
that
channel,
open
and
yeah
it
would.
It
would
either
be
a
cloud
that
we
have.
C
Google
has
published
a
thing
where
Cloud
scheduler
sends
a
pub
sub
message
to
a
job
running
in
Cloud
functions
and
it
seems
a
little
ornate
and
the
obvious
answer
would
be
a
kubernetes
pod,
but
honestly
either
one
would
probably
work
so
I,
don't
know
how
people
feel
I
could
also
just
do
a
one-time
export
of
some
safe,
some
safe
subset,
and
then
we
could
see
what
what
we
want
to
do
from
there
type
thing.
A
C
A
C
A
Yeah,
if
we,
if,
if
you
need
to
expose
the
cloud,
run,
metrics
the
ground
service,
metrics
again
see
who
can
have
an
exporter,
yeah
graph
and
app,
can
basically
use
JCP
as
an
exporter
and
we
can
use
graphene
response.
Metrics,
that's
my
easiest
solution.
Now
we
can
figure
out
how
we
can
have
one
job
interacting
with
Cloud,
run
API
getting
those
metrics
and
putting
that
bigquery.
C
A
E
Hi
I
I
wanted
to
discuss
one
of
the
issues
I've
been
working
on
it's
about
enabling
enabling
Tech
policies
for
the
new
AWS
organization,
so
I
wanted
to
know
what
what
are
the
next
steps?
E
I
have
created
a
organizational
unit
called
policy
staging
and,
in
that
organizational
unit
an
AWS
account
to
test
the
tech
policies.
So
what
have
what
I've
been
working
on
is
in
creating
Tech
policies,
plus
service
control
policies
to
m4s
tagging
at
creation,
time
of
the
resources
that
we
want
to
to
enforce
tagging
so
far,
I
have
ec2
some
some
resources.
E
Primary
C2
is
too
instance
volumes
and
bpcs
eks
clusters
from
Ikeas
and
repositories
from
ECR
and
S3
buckets,
and
that's
all
because
I
wanted
to
start
small
and
to
test
them
first
and
I.
Guess
that
the
the
the
the
the
tags
that
we
are
going
to
enforce,
we
have
to
enable
cost
allocation
for
those
stacks.
E
I,
don't
know.
If
do
you
have
any
I,
don't
know
some
suggestions
or
you
want
some
more
details
about
what
I'm
doing.
A
Yeah
I,
just
I'm
gonna,
gave
context
here
is
basically
open
initial
that
tagging
policy
on
AWS,
because
I
wanted
to
be
I.
I
wanted
us
to
have
basically
tag
on
any
resource
we
create,
so
we
can
easily
establish
ownership
at
the
group
level
and
also
other
individual
level.
Right
now
we
are
focused
on
group
levels
basically,
which
seek
or
working
group
home
is
specific
resources,
so
gabra
has
been
working
on
that
for
I.
Think
two
months
now
he
came
up
with.
If
you
check
the
issue,
you
will
see.
E
Yeah
and
and
I
have
I
I
I'm.
Sorry
I
should
have
mentioned
this
earlier
tagging
throughout
AWS
Services,
it's
not
very
consistent,
I
mean
S3
has,
for
instance,
S3,
for
example.
You
can't
enforce
tax
or
tagging
a
bucket
at
creation
time,
because
I
don't
know
well,
I,
don't
know
why.
But
you
can't,
with
with
the
tag
policy
enforcement
or
with
service
control
policies
ec2,
for
instance,
you
can,
and
with
with
attack
policy
enforcement
and
plus
the
service
control
policy.
E
You
can
make
enforcement
at
creation
time
and
you
will.
You
will
get
an
error
that
it
doesn't
really
it
it.
It
isn't
related
to
tagging,
but
you
will
get
an
error
and
in
newer
services
like
ECR
or
eks,
you
will
get
an
error
related
to
that.
You
will
get
an
error
that
that
some
tag
is
needed
or
that
maybe
the
value
of
the
tag
that
you're
trying
to
to
set
it's
wrong.
E
So
it's
a
little
difficult
to
create
those
policies,
because
it's
it's
not
consistent
through
the
AWS
services.
So
that's
why
I
wanted
to
start
small.
A
Problem
with
that,
because
it's
basically
the
current
state
of
AWS
API,
they
are
not
I,
think
there's
a
lack
of
consistency
at
some
point
about
specific
resources.
I
don't
have
a
problem.
That's
why
I'm
saying,
basically,
if
possible,
I
want
to
emphasis
of
if
possible,
if
it's
a
possible,
that's
fine!
The
good
thing
is.
We
have
AWS
folks
working
with
us,
so
we
can
just
report
that
then
see
what's
happening
as
well.
A
We
can
Target
resources
and
we
have
a
type
of
play:
I'm
fine
with
that
I'm
finding
that
if
it's
not
possible,
it's
also
fine.
We
can
try
to
be
universals
about
this.
Maybe
we
get
special
treatment.
I,
don't
know,
but
I
mean
the
end
game
here
is
basically
have
any
risk
of
style.
That's
what
I
want
now
we
can
have
that
as
consistent
that
we
need
to
be.
A
It's
already
confusing
for
JCP,
so
I
would
like
to
basically
not
let's
make
the
same
mistake.
We
have
I,
don't
think
it's
a
mistake.
It's
more
like
sagging
was
not
present
at
the
beginning
of
this
working
group
and
at
the
beginning
of
the
infrastructure.
So
now
we
have
the
opportunity
to
end
to
to
introduce
that
approach.
Let's
do
it.
A
Account
yes,
okay,
yes,
I
will
follow
up
on
that
in
the
in
the
issue.
E
This
this
will
I
will
want
to
say
that
this
will
maybe
affect
the
well.
It
will
affect
the
the
way
that
some
of
us
create
resources,
because
if
the
tag
is
enforced
enforced,
maybe
you
you
will
get
an
error
and
something
that
that
bothers
me
is
that
some
some
services,
like
ec2,
doesn't
give
an
error
related
to
the
what
what
is
going
on
so.
A
That
that's
okay.
As
long
we
can
iterate
and
basically
see
what's
happening
if
we
cannot
create
a
resource,
because
a
tag
is
missing,
it's
a
good
thing,
that's
mean.
Basically,
we
can
someone
we
reach
out
to
one
of
the
one
of
the
org
admin
to
basically
see
what's
happening,
I
mean
that's
good.
If
a
tag
is
not
present,
you
can
investigate
that.
Maybe
we
can
detect
that
with
like
a
world
only
the
best
config
or
something
like
that
and
have
basically
highlighting
based
on
that
I.
A
C
Yeah,
so
just
to
confirm,
like
a
tool
like
chaos,
will
have
to
add
this
tag
to
all
of
our
jobs.
C
A
Yes,
if
you
check
the
pull
requests,
you
will
see,
we
only
have
froze.
We
only
have.
A
E
I
do
have
another
question:
how
how
are
we
supposed
to
communicate
that
we
that
we
have
to
set
these
texts.
A
Mailing
list
possibly
basically
use
the
official
Communication
channel,
so
mailing
lists
and
slack
messages.
Okay
anywhere.
A
We
have
I
think
right
now.
Right
now
is
like
the
next
step
is
basically
create,
maybe
two
or
three:
oh
you
already
test
that
in
the
policy
staging
yes
front,
I
want
to
directly
see
more
tests
in
one
of
the
non-production
unit
and
see
what's
happening
whether
to
brush
about
this,
because
we
still
in
let's
say
we
still
in
the
bootstrapping
phase,
especially
regarding
Bosco.
So
we
are
not
in
brush
about
this.
A
Let's
do
the
maximum
test
we
have
before
we
start
communicate
this
mostly
oh
I
would
say
crops
and
impact
by
this
is
going
to
be
those
creating
e-tweek
tests.
So
that's
why
I
think
that's.
Why
Chelsea
has
this
question
I'm
thinking
about
also
six
capability
and
other
groups
mainly
use
aw,
yes,
AWS
for
tests,
so
I'm
not
very
worried
about
communication,
but
this.
If
there's
something
failing
at
the
creation,
people
will
reach
out
to
us,
we
I
would
say
where
I
used
to
firefighting
CI.
So
I
don't
see
this
as
a
problem.
A
B
This
is
really
just
for
the
recording
but
Friday
afternoon
found
out
that
the
cncf
AWS
account
is
going
to
run
out
of
credits
which,
why
am
I
bringing
up
the
cncf
AWS
account
and
not
the
Kate's
infer
account,
because
it
turns
out
that
there
were
a
bunch
of
jobs
that
were
running
from
kubernetes
but
referencing
or
still
under
the
root.
Cncf
account
the
two
real
bait,
not
real
big,
but
the
two
main
action
items
are.
B
There
are
two
two
AWS
accounts
that
need
to
be
migrated
under
the
Kate's
in
for
bucket,
that
is
in
Flight,
I,
believe
and
then.
The
second
thing
is
a
bunch
of
chaos
or
cops.
Ci
jobs
were
referencing,
the
old,
not
old,
but
the
cncf
AWS
management
account,
which
kind
of
makes
sense,
because
that
was
like
the
start
of
all
of
this.
B
Was
that
cncf
account?
So
we
need
to
shift
those
jobs
to
point
to
the
Kate's
in
for
bucket.
That's
it
again,
like
I've
talked
about
it
in
slack,
I'm,
really
just
stating
it
for
the
notes
and
for
the
record.