►
From YouTube: Kubernetes SIG K8s Infra - 20220817
Description
A
A
C
C
A
D
A
A
A
B
B
A
E
E
E
A
D
E
A
A
E
E
A
E
A
E
E
A
B
A
So
it's
at
the
moment
we
talked
to
them.
It
was
not
part
of
that,
so
we
need
to
go
back
to
them.
So
I
think
the
with
cloudflare
we
want
to
finish
basically
migrate
to
the
migrate,
a
few
domains
we
have
currently
and
once
we
finish
with
that
migration,
we
can
talk
about
using
cloudflare
a2,
but
I
don't
think
we
need
to
move
to
that,
because
there
is
no
specific
value
again
on
that.
A
A
Basically
is
the
same
thing
because
we,
I
think
the
the
the
difference,
because
we
have
an
enterprise,
so
we
have
in
limited
bandwidth.
We
are
not
limited
by
bandwidth.
Okay,
so
moving
to
r2
is
more
like.
Oh,
we
want
to
leverage
our
to
feature,
but
we
already
use
cloud
for
network
use
with
the
enterprise
plan.
So
there's
no
okay,.
F
A
Yep
and
egress
is
always
most
free.
I
think
the
the
minor
difference
is
the
the
acres
costs
coming
from
cloudflare,
because
if
we
use
cloudflare
we're
going
to
pay
for
the
traffic
between
the
origin
and
cloudflare,
that's
what
we're
going
to
pay
and
we're
going
to
have
75
percent
of
rejection
on
that.
A
F
Yeah
and
the
benefits
for
r2
I
mean
it
makes
sense
to
store
the
blobs
in
r2.
If
you,
if
you
but
the
thing
I
was
worried
and
I
never
suggested
looking
into.
That
is
because
we
started
lots
of
lots
of
historical
globes
if
r2
charges
you
for
those
images
which
are
simply
there
for
historical
reasons,
because
we
never
purchased
them
and
we
would
be
paying
for
nothing
really
right.
A
A
F
A
So
it's
like
you
don't
need
to
do
anything.
You
just
need
to
shield
your
basically,
your
your
point
of
distribution
like
dlcs.io
ads
focus.io,
the
the
endpoint
for
installing
meaning
the
system
packages.
I
don't
remember
you
share
that
with
cloudflare.
There's
no
point
to
basically
add
extra
logic
to
push
to
s3.
A
A
What
I'm
saying
yeah
for
2022,
we
should
basically
focus
on
finish
setup
and
cloudflare
finish
with
s3
and
stuff
like
that,
and
when
we
start
on
23,
we
can
come
back
and
basically
say
exactly
also
based
on
the
conversation
we
have
with
cloudflare
because,
like
I
said
at
the
moment,
we
reach
out
to
them.
R2
was
not
the
thing
and
was
not
part
of
enterprise
plan.
A
F
F
Yeah,
so
so
yeah,
so
just
I
think,
I'm
kind
of
seeing
the
benefit
of
of
it.
Maybe
it's
the
same
that
mohammed
is
it's
also
contemplating
so
as
as
we
so
the
the
plan
as
we
are.
We
are
currently
improving,
tooling
on
the
promoter
to
push
to
ir
and
we're
also
building
the
the
aws
infrastructure
mirroring
system,
so
those
are
going
to
be
built
anyway,
and
the
path
that
we
are
working
on
right
now
is
adding
complexity
and
time
to
the
promotion
process.
F
E
A
That's
what
I'm
I'm
proposing
when
I
reached
out
to
cloudflare
was
the
entire
plan
force.
2022
is
focused
on
absorb
the
core
for
what
is
existing
right
now,
so
mostly
the
binaries
and
the
system
packages,
and,
to
be
honest,
we
are
two
was
something
we
wanted
to
look
into,
but
it
was
private
beta
and
I
never
got
came
back
to
them
and
basically
said:
oh,
let's
try
it
out.
So
we
can
talk
about
that
next
here.
If
you
want.
B
A
2023
is
more
like
to
be
defined
because
we
still
don't
know
what's
going
to
happen
in
the
next
six
months
about
image
promotion,
so
even
for
next
20
23
is
to
be
defined,
like
I
don't
know,
what's
happening
with
information
in
december,
so
I
know:
we've
been
talking
about
image.
Promotion
basically
get
rid
of
a
lot
of
system
packages,
but
I
saw
it's
a
different
issue
so
about
that.
I
think
we
we're
clear
about
about
this.
A
A
C
D
E
C
But
yeah
I
I'm
not
sure,
but
I
will
be
revisiting
that
and
then
I
have
been
working
on
this
from
the
start
and
currently
running
into
terraform
bottlenecking.
How
I
can
implement
this.
Those
are
some.
I
am
things
that
I'm
trying
to
get
right,
but
pretty
much.
What
this
pr
is
is
to
add
a
resource
that
does
the
sinking
between
all
of
the
regions.
C
Incident
would
just
put
into
the
us
east
2-1,
and
then
it
would
figure
its
way
out
to
the
rest
of
them,
but
yeah,
sometimes
terraform
can
make
things
a
little
more
difficult
and
running
into
something
with.
I
am
that
I
need
to
figure
out
as
well,
but
been
on
my
mind
since
the
start
still
figuring
it
out.
C
My
thought
was
that
aws
can
take
care
of
it
for
us,
which
means
that
there's
no
management
around
running
a
job
somewhere,
and
it
also
would
mean
that
the
the
traffic
is
likely
free
kind
of
thing.
If
we
want
to
do
any
syncing
and
make
it
definitely
more
on
the
the
cost.
Saving
side,
then
we'll
need
to
make
sure
it's
running
in
aws
somewhere.
D
A
A
Because
the
one
thing
we
we
basically
can
promote
oci
proxy.
If
that's,
if
on
the
bucket,
are
not
in
sync,
that's
that's
my
my
main
concern
here
like
because
I
think
la
last
week
I
had
to
basically
sing
from
gcs
to
to
the
marketing
is
to
because
there
were
like
some
blob
missing
like,
for
example,
the
bluffs
or
the
pose
images
were
missing.
C
B
C
C
However,
I
think
you
might
have
only
done
it
to
one
region,
but
I
I
have
some
scripts
to
do
it
manually
for
gcs
to
s3
and
s3
to
the
remainder
of
the
bucket,
which
would
in
fact
do
the
minimal
cost
saving
of.
However
many
blobs
that
would
be
transferring
over
there,
but
I
what
I'm
trying
to
say
is:
yes,
I
think
automated
sync
is
great.
I
don't
think
it.
C
B
No,
I
think
ben
would
prefer
if
it's
up
and
running,
but
what
I
I
understood
exactly
the
same
from
ben
is
so
basically
he
keeps
log
of
what
is
synced.
He
check.
Is
it
there
and
then
there's
a
a
local
stored
log
of
the
things
that
he
did
find
there
and
if
he
does
not
find
it,
it's
not
in
the
log
it
just
not
don't
redirect.
So
it's
just
a
loss
of
cost
saving
that
we
have,
but
the
end
user
will
not
miss
the
the
artifact,
but.
A
B
Absolutely
absolutely
that's
why
this
issue
is
important
and
is
a
priority,
but
it's
not
a
blocker,
so,
okay,
that
is
the
thing
that
we're
working
on
now.
So
I
don't
think
it's
blocking
us,
but
you're,
absolutely
right.
It
is
a
priority
to
get
it
up
sooner
than
later
and
also
get
the
syncing.
Automated
inside
we've
been
talking
quite
a
bit
with
aws
to
try
and
get.
A
E
A
C
C
I'm
not
sure
about
an
eta
for
it,
but
there
will
be
cost
saving
if
we
go
on
to
production
for
the
new
assets
downloaded
using
the
domain.
Of
course,
when
new
new
images
become
available
and
the
the
blobs
need
to
be
accessed,
the
cost
saving
won't
be
immediately
in
for
those
until
it's
synced,
but
that
that
is
quite
a
minimal
thing.
A
minimal
amount
of
data
to
transfer
compared
to
all
of
the.
B
Now,
from
our
last
discussion
with
ben,
he
said
he
would
love
the
automated
sync,
but
if
we
can
work
out
some
method
to
at
least
ensure
that
there
is
syncing
at
some
stage,
if
even
if
it
has
to
be
manually
up
to
now,
but
it's
not
a
blocker,
so
yeah
we
can
work
out
what
we
could
do
on
outside.
If
we're
really
blocked,
we
can
say:
okay,
we're
going
to
have
a.
E
B
F
E
B
Yes,
that's
the
one
plan
and
the
other
plan
is
hippie
already
send
out
a
notification
on
the
mailing.
Let's
say
be
aware:
this
is
what's
going
to
going
to
happen
within
the
next
30
days.
We
are
working
on
it,
but
we
won't
swap
out
before
the
next
30
days
so
to
give
people
a
heads
up,
because
what
we're
concerned
about
is
that
it
could
be
a
breaking
change
for
people
so
properly.
Just
after
release
has
been
cut,
would
be
a
good
time
to
do
this.
A
C
D
A
D
F
A
E
E
E
E
A
B
A
D
F
F
It
so
some
something
I'm
not
sure
of
the
specifics,
but
I
can
I
can
get
them
if
you,
if
you
need
me
to
apparently
the
the
new
redirection
schemes
broke,
the
sneak
client
so
sneak
the
security
tool.
You
can
point
it
to
an
image
reference
and
and
it'll
scan
it
for
you,
so
the
new
redirection
broke
all
the
scanning
of
the
kubernetes
images.
I
don't
know
why
exactly
and
they
already
fixed
it,
but
I
was
curious
if
you've
heard
about
other
tools
breaking
because
they
can
handle
the
redirection.
C
F
A
We
never
had
does
it
after
all,
the
commission
we
made.
I
think
we
need
to
enhance
the
communication
and
that's
why
we
flag
that
change
as
a
as
a
breaking
change,
because
the
reason
is
going
to
be
breaking
change
once
we
do
that,
we
should
basically
see
people
reach
out
about
things
breaking,
but
if,
like
snake
people
can
reach
out
to
us
and
basically
share
with
us
what
happened,
what
kind
of
issue
they
face,
that
would
be
great.
So
we
can
basically
address
that
now.
A
A
F
A
F
A
Yep,
so
if
they
can
open
an
issue
anywhere,
we
might
be
able
to
investigate
that
and
see
what's
happening
so
because
there
are
too
many
options
here,
like
maybe
they
use.
The
v1
client
of
the
docker
client
is
doing
requests
using
the
v1
version
of
the
api
server
and
that's
why
there's
an
issue
must
be
some
something
in
the
blobs
many
options.
F
Yeah,
so
this
is,
I
mean
it's
basically
what
you
were
saying
we
let
people
know
and
when
it
broke,
they
knew
the
reason
and
fixed
it,
but
to
fix
it
is
if,
in
the
future
we
see
other
other
tools
breaking.
It
would
be
helpful
to
know
that
certain
pattern,
pattern
of
consumption
and
breaks
lines
and
in
order
to
more
easily
advise
people.
F
A
B
A
A
E
E
A
A
A
D
D
A
This
is
on
me.
I
was
supposed
to
get
back
about
this,
so
my
concern
is,
if
we,
I
think,
the
front.
The
first
concern
I
expressed
is
like
we
should
not
do
public
bracket
because
we
gonna
be
almost
be
the
people
handling
the
traffic
for
other
people
interest,
but
by
the
data
you
want
to
pull.
I
think
danielle
suggests
we
use
private
packet,
I'm
fine
with
that.
A
D
A
There's
no,
basically,
if
you
have
a
gcp
project,
it's
fine,
even
of
your
personal
icon.
I
can
add
you
and
see
what's
happening
because
we
don't
have.
We
are
not
specific
about
you
uses
of
gcp
all
the
instances
of
dynamic
public
ip
address,
so
I'm
wondering
why
I
can
using
three
different
two
different
gcp
projects
pull
the
data
with
a
a
curl,
but
in
that
image
it's
failing.
I
think
that's
my
concern.
I
understand
because
at
work
I
use
a
gcp
project
and
I
can
pull
that
data
using
conventional
infrastructure.
D
A
Next
week,
because
I'm
a
little
focused
on
belize
125
is
really
it's
happening,
tuesday,
yeah.
I
have
a
lot
of
things
to
do
before
that,
so
so
for
anyone
interested
in
the
conversation,
there
is
a
job
pulling
that
information
and
which
is
the
public
data
set
and
us
by
the
american
university.
So
we
have
an
image
pulling
that
data
and
approach
of
using
that
image
failed
to
download
that
we
basically
got
the
connection
timeout.
D
A
E
A
Why
that's
not
a
great
idea
but
yeah?
That's
why
there's
a
policy
planning
it
then?
Okay,
it's
like
policy
banning
is
more
like
I
know
in
the
past.
We
had
issue
with
that
so
because
I
would
say,
for
example,
the
the
cluster
sub-report.
I
had
a
lot
of
binaries
and
we
got
rid
of
a
lot
of
stuff
in
that
repository.
A
D
E
Yeah,
so,
as
you
know,
the
infrastructure
is
up.
Kpro
is
a
bug.
I
know
that
it
torpedoed
the
1.25
release
the
other
day.
Unexpected
things
we're
I'm
investigating,
k,
promo's
logic.
I
know
what
the
problem
is.
Just
gonna
work
out
how
to
fix
it
and
then
the
other
thing
is
there.
I've
got
a
couple
of
jobs,
running
to
sync
buckets,
sorry
to
sync,
the
artifact
registry
repositories
and
I'm
running
two
quarter
problems.
E
A
E
A
E
Yeah,
okay,
that's
interesting!
I
I
wasn't
able
to
see
this.
I
have
no
idea
how
well
it's
progressing
if
we're
expecting
a
terabyte,
it's
actually
pretty
good.
The
jobs
are
running
every
two
hours,
some
of
the
quarter
resets
every
two
hours.
Some
sets
resets
once
a
day,
but
okay,
at
this
rate,
I
think
it'll
be
done
by
the
end
of
the
week.
A
E
A
D
E
E
A
E
E
E
A
A
Okay,
so
that
change
is
only
for
like
google,
but
even
in
october
one.
It's
gonna
be
the
same
pricing
and
I
think
because
let
me
see
if
you,
which
is
free
pricing
stuff.
I
think
I
saw
that
somewhere.
Where
is
the
egress
network?
Egress
yeah
ethernet
egress
is
based
on
premium
tier
yeah,
so
it's
mostly
so
there's
no
really
for
everything
outside
of
the
google
cloud.
There's
no
change
because
it's
going
to
be
the
same
price
yeah,
that's
what
I'm
saying
moving
to
at5
registry
is
only
a
benefit
for
us
as
we
have
infrastructure.