►
From YouTube: Kubernetes SIG Testing Meeting for 20220111
Description
Kubernetes SIG Testing Meeting for 20220111
C
C
C
C
That's
awesome
did.
How
far
are
you
from
the
ii
office?
Now
I
still
don't
12
000
kilometers,
I'm
still
in
south
africa.
C
That
didn't
happen
that
I
just
moved
to
the
south
this
coast
in
south
africa,
so
this
is
my
south
african
coast
and
now
I'm
busy
inviting
everybody
that
ever
that
says,
I
want
to
go
south
africa
and
my
answer
is
yes,
I
have
an
agenda.
This
is
how
you
do
it
fly
here,
so
I've
got
a
whole
schedule,
so
anybody
interested
come
visit
me
I'll,
show
you
around
and
tell
you
what
to
do
after
you
leave
for
my
house.
C
A
D
D
I'm
also
using
the
web
client
just
because
it
works
fine.
For
me,
I
think
the
zoom
client,
specifically
there
were
like
security
concerns.
D
D
I
don't
think
we
had
an
agenda
ahead
of
time.
Looking
at
the
dock.
B
Yeah,
that's
the
first
question.
I
think
one
editing
trying
to
say
is
which
we
we
want
at
some
point
to
have
a
conversation
about
traumatization.
I
think
that's
one
thing
we
really
need
to
think
about
this
year.
D
D
So
I
don't
know
when
the
last
like
said:
kate's
inframating
was,
but
I
actually
had
a
meeting
yesterday
with
tim
about
this
topic
and
the
gist
of
what
I
got
was
that
we
need
to
control
spin
first,
our
spend
is
out
of
control
and
we
are
not
going
to
be
able
to
keep
moving
things
until
we
rethink
yeah.
B
D
I
think
if
we
introduce
some
different
artifact
hosting
or
something
it's
going
to
take
a
long
time
to
actually
get
everyone
that's
consuming
to
move
over,
and
I
think
it
could
be
a
good
while
before
we've
kind
of
actually
reigned
in
the
spend
on
the
existing
stuff.
Just
because
of
the
delay
from
when
we
implement
something
to
when
it
actually
pays
off.
B
I
don't
know
what
conversation
you
had
with
team.
So
let's
have
this
conversation
in
kids
in
front.
D
Oh
I'm
planning
to
follow
up
we're
just
discussing
that.
Like
potentially
you
know,
google
is
looking
at
staffing
that
problem
okay,
so
tim
was
getting
me
up
to
speed
on
what
has
been
happening,
kids
and
for
because
I
haven't
been
there
in
a
little
while
it's
so
yeah
it
that's
my
on.
That
is
that
I
I
would
suspect
that
we
probably
actually,
I
think,
is
probably
actually
a
feature
at
the
moment
that
most
of
the
spend
is
on
the
like.
Essentially
unmetered
billing
account.
B
D
Nodes
right,
I
don't
think
we
can,
I
mean
like
technically,
it's
maybe
feasible.
I
think
policy
isn't
like
one
thing
that
we
have
is.
I
don't
know
how
much
talk
about
this,
but
one
of
the
things
is,
we
have
pretty
heavily
locked
down
the
ability
to
like
export
service
account
keys.
D
It's
that's
like
a
legacy
feature
inside
google.
At
this
point,
so
it
is
increasingly
challenging
to
have
like
new
control
over
the
old
projects.
I'm
would
need
to
talk
about.
A
D
B
D
Right,
which
is
where
most
of
the
jobs
are
now
so
we
could
like.
There's
enough,
I
think
a
couple
of
gcs
buckets
are
like
blockers
for
moving
certain
like
build
jobs
over.
We
already
have
moved
things
like
kubernetes
pre-submits.
I
don't
think
there's
a
lot
blocking
moving
more
stuff,
but
it
prowl
is
structured
such
that
you
really
only
want
one
service
cluster
handling,
like
pre-summits
and
so
on,
for
a
repo,
the
things
that
actually
interact
directly
with
the
repo,
as
opposed
to
like
periodic
jobs.
D
So
the
google
services
cluster
is
executing
jobs
in
a
community,
build
cluster
and
preparation
for
migration,
but
because
a
couple
things
like
build
have
been
blocked
on
that
being
trickier
to
move,
we
don't
have
all
the
pre-submits
moved,
so
we
also
had
to
move
the
services
cluster
because
you
kind
of
want
to
like
you
want
to
cut
over
the
services
cluster
once
you
don't
you
don't
want
to
have
mixed,
you
can
have
mix,
build
clusters
or
like
mix
ci
jobs
or
something
between
them.
That's
fine!
D
Last
I
saw
the
problem
was
the
jobs
that
push
builds?
Those
buckets?
It's
really.
You
can't
just
like
move
a
bucket
between
projects
and
there's
things
like
consuming
them,
so
I
don't
know
where
we
wound
up
on
that
effort,
but
I
think
those
dominoes
are
how
we're
at
the
state,
where
we
have
some
stuff,
moved
to
scheduling
in
a
community
cluster,
but
it's
being
scheduled
by
the
the
google
services
cluster.
A
I
feel
I
gotta
dig
back
in
because
I
I
really
thought
that
we
had
the
community
cluster
was
able
to
kick
up
jobs
and
the
compute
was
done
on
the
google
owned
cluster
and
for
context.
The
reason
I'm
bringing
this
up
is
because,
when
arno
and
I
talked
about
it,
we
want
the
same,
I
know
wants
the
sigs
to
be
responsible
for
moving
their
own,
proud
jobs
and
that's
going
to
be
a
really
really
long
process.
D
I
think
we
expected
the
also
the
bigger
gains
to
be
from
moving
the
compute
cluster.
I
mean
it
has
cost,
which
is
a
problem
for
us
right
now,
but
that
allows
you
to
like
access
and
debug
the
actual
execution
environment
whereas,
like
what
code
is
running
in
the
services
cluster
and
how
it's
configured
is
already
auto
deployed
from
test
infra
and
we
are
like,
except
for
one
or
two,
you
know,
except
for
a
small
group
of
people
that
need
to
actually
keep
the
pro
running.
D
We
don't
want
to
be
granting
like
every
sig
access
to
the
actual
services
cluster,
because
that
has,
like
god,
keys
over
everything,
but
then
we're
back
to
like
the
execution
cluster
is
where
all
of
the
cost
is.
So
I
don't
know.
I
I
my
assumption
right
now.
Is
that
we're
it's
long
of
a
process?
Is
that
gonna
be
we're?
Gonna
have
to
pump
the
brakes
for
a
while
until
we
can
figure
out
the
budget.
D
D
For
example,
I
believe
most
of
the
cost
is
coming
from
the
docker
images
and
those
right
now.
People
are
downloading
from
a
domain
owned
by
like
the
gcr
team,
even
if
we
had
something
else
that
could
rework
that
it
is
going
to
be
a
very
interesting
conversation
to
consider
trying
to
move
that
domain
over
once.
D
We
have
something,
I'm
not
sure,
that's
even
allowable
I'm
gonna
have
to
dig
more
into
the
policies
and
if
it's
not,
then
we're
gonna
have
to
get
all
the
clients
to
move
to
some
new
endpoint,
which
I
think
we
want
to
do.
Anyhow,
I'm
not
sure
the
history
behind
that,
but
I
mean
I
think
we
have
kate's
daddy.
D
Oh,
we
should
move
people
to
using
kate's
that
I
o
to
indirect
all
of
the
community
things
which
should
include
artifacts,
but
given
how
gcr.io
google
containers
to
case.gcr.io
went
that's
going
to
take
like
six
months
a
year
or
more,
to
see
any
substantial
shift
in
consumption.
People
are
usually
shipping
stable
releases,
a
couple
versions
back.
B
D
B
B
D
Is
out
for
the
moment,
yeah.
B
D
Right
we
are
but
you're
talking
to
tim
about
it.
I
wouldn't,
I
think,
it's
possible.
I
wouldn't
place
bets
on
that
happening,
okay,
it.
It's
definitely
going
to
look
like
a
questionable
move
from
like
a
policy
perspective,
I'm
not
sure
if
we
can
get
that
or
not
and
tim-
and
I
both
think
seems
probably
unlikely
we're
definitely
going
to
ask,
but
we
also
it's
going
to
be
difficult
to
ask
before
we
have
something
more
concrete
on
like
what
does
that?
D
Look
like
okay,
so
it's
possible
if
we
can
ship
something
that
works
and
get
it
up
and
running
and
then
manage
to
redirect
the
existing
domain
that
we
could
cut
traffic
at
that
point,
but
I
think
we're
also
going
to
want
to
be
really
careful
doing
some
kind
of
cut
over
like
that
on
a
very
production
endpoint
that
tons
of
clusters
depend
on.
D
I
yeah.
I
don't
expect
this
is
going
to
be
super
soon
that
that's
done,
and
I
think,
if
there's
a
very,
very
real
possibility
that
case
that
gcr
diogo
is
kind
of
just
deprecated
in
place
like
we
did
with
google
containers
and
that
we
have
to
get
clients
to
cut
over
to
the
new
thing,
and
I
think
that
if
we
have
to
go
that
route,
it's
gonna
take
even
longer
so
yeah.
D
I
I
don't
know
if
there's
other
possibilities
like
maybe
we
can
just
increase
the
budget,
it
sounds
like
that's,
probably
not
going
to
happen.
I
think
tim's
had
enough
friction
on
the
existing
one.
D
I
also
I'll
need
to.
I
don't
know.
You
probably
know
that
data
than
me.
I
was
thinking
this
through
tim.
If
there's
like
anywhere
else
that
we
can
try
to
reduce
it
sounds
to
me,
like
you
know,
artifact
hosting
is
dominating
at
the
moment.
B
Is
the
only
cost
problem
we
have
right
now
so.
B
B
D
Yeah
I
just
I
said:
I
suspect
that
there
are
reasons
that
you
can't
do
that
on
like
official
google
domain,
okay,
but
we'll
we'll
see,
I
I've
worked
a
little
bit
with
domains
at
google
before
because
of
pro
instances
and
in
the
past
the
like
rules
around
that
were
like
really
strict
so
we'll
see.
I
we're
definitely
gonna
ask.
D
But
I
don't
think
we
don't
have
like
direct
ownership
over
that
and
like
if
I
were.
If
I
were
that
team
on
the
other
side,
I
would
not
be
thrilled
to
like
basically
run
a
proxy
pointing
at
something
we
don't
control
on
our
like
official
domain.
B
D
Point,
but
even
then,
I
think
it's
going
to
take
a
while
before
we
have
the
serving
infrastructure
in
place
and
mirrors,
and
things
I
mean
given
that
we've
literally
been
talking
about
this
possibility.
Since
the
first
meeting,
when
tim
announced
that
we
had
the
credits
and
as
far
as
I
know,
there
hasn't
been
any
major
progress.
B
D
I
would
expect,
hopefully
tentatively
assuming
that
likely
to
work
on
this.
I
imagine
we
can
get
something
up
and
running
this
quarter.
That
like
looks
about
right,
but
I
wouldn't
be
surprised
if
it
takes
at
least
another
quarter
to
like
start
really
using
it,
and
then
I
don't
know
how
long
until
we
can
get
all
the
client
load
shifted.
So
I
think
this
is
sort
of
like
probably
the
main
thing
we're
going
to
be
doing
this
year
for
migrating
right
until
that's
done,
and
I
wouldn't
bet
on
it
being
a
quick
exercise.
B
D
D
A
B
B
It's
we
have
a
cdn
as
a
service
since
I
go
for
inside
the
organization
we
run
right
now.
We
just
have
to
enable
the
city
and
in
front
of
a
proxy
and
the
cdn
will
basically
distribute
the
traffic
for
you
and
it
will
catch
the
traffic.
D
Interesting
well,
we'll
have
to
follow
more
about
this,
but
I
said
I
suspect
this
is
going
to
take
a
while.
In
the
meantime
I
don't
know,
I
mean
other
possibilities
are
that
we
could
spin
up
like
I
said
we
could
spin
up
like
a
compute
cluster
on
another
provider,.
B
The
thing
is
for,
if
you
want
to
basically
use
that,
for
example,
amazon
we're
talking
about
eddie
about
this
amazon
as
what
they
call
a
public
content
registry-
and
we
just
need
to
push
that
to
this
content
registry,
but
we
need
to
request
funding
for
that.
At
the
moment
we
push
to
that.
We
certify
the
registry.
We
push
artifact
on
this
new
version
inside
amazon,
amazon
client
can
consume
that
registry.
A
So
yeah
this
is,
we've
been
talking
about
this
for
a
bit
and
I
was
talking
to
tim
about
it
as
well.
There's
still
a
lot
of
policy
that
we
need
to
figure
out
in
terms
of
who
owns
that
account
is,
do
we
do
this
on
the
cncfs
aws
account?
Do
we
stand
up
right
because
I
could
I
said
before
like
I
can
easily
stand
up
and
employee
owned
account
inside
aws
it
is
free
99.
A
D
E
Sorry,
I
I
didn't
mean
to
like
interrupt
the
discussion,
but
I
don't
even
know
if
it's
the
right
forum
but
yeah,
I
already
sent
out
an
email
yesterday
on
a
couple
mailing
lists
for
this
group
and
also
for
the
vet
kubernetes.
E
Basically,
tldr
is
that
I
want
to
work
on
improving
the
proud
documentation,
basically
in
the
style
of
what
kind
has
done
with
like
statically
generated
docs,
because
this
looks
really
nice.
I
just
think
crowd
deserves
one
two
I
haven't
heard
any
big
like
we
shouldn't.
Do
this
type
of
comments
yet
so
it
sounds
like
it
will
probably
happen,
but
yeah,
please
feel
free
to
leave,
notes
the
doc
I've
shared
it
with
those
groups.
E
So
if
you
have
access
to
those
mailing
lists,
you
should
be
able
to
comment
on
them
or
in
the
github
issue.
If
that's
a
an
issue
yeah,
that's
all
I
have
to
say
thank
you.
D
Thanks
linus
also
sorry.
Normally
we
would
have
these
in
like
items
time
box,
but
they're
also
ordinarily,
for
meetings
throughout
the
rest
of
the
year.
We'll
ask
that
there
is
a
meeting
to
cut
up
time
on
the
community
calendar
for
adding
agenda
items
the
day
before.
So
we
know
what
we're
getting
into,
but
right
yeah
didn't
happen
this
year,
so
we're
trying
to
start
better
on
the
next
meeting.
Sorry
about
that,
and
thank
you.
D
I'm
a
huge
pro
of
those
it.
We
had
to
do
quite
a
bit
of
figuring
out
for
like
how
was
that
allowable.
D
D
D
Sig
docs
runs
all
the
like
static
site
like
the
account
and
then
the
rest
is
category
for
the
domains
right.
B
I
would
just
want
to
add
a
clarification.
Basically
you
need
to.
If
you
want
to
use
netlify,
you
need
to
talk
to
say
country
backs
about
this,
because
they
are
the
one
admin
of
the
kubernetes
netlife
account.
B
Kind
of
because,
on
the
caterpillar
side,
we
just
update
the
dns
and
try
to
the
netify
endpoint.
D
B
E
Okay,
I
mean:
are
you
suggesting
that
we
try
to
split
the
code
base?
At
the
same
time,.
E
I
mean
I
wish
I
could,
but
that
sounds
like
a
really
hard
problem.
Okay,
I
mean
not
not
technically
difficult,
but
just
you
know
the
all
the
strings
that
are
attached.
You
know
yeah.
I
haven't
discussed
that
idea,
but
I'll
put
in
the
documentation
or
in
the
design
doc
just
for
people
to
comment
at
least
on
it.
Yeah
we'll
go
from
there.
I
guess.
D
Also
you'll
have
natalify
kind
of
has
like
repo
scoped
config.
So
if
you
want
to
use
in
repo
config,
then
it's
kind
of
like
one
to
a
repo.
D
Okay,
we're
worth
considering
you
don't
have
to
use
in
repo
config,
but
then
some
you
can
only
change
things.
Any
of
the
things
that
you
could
have
changed
within
repo
config
file
can
now
only
be
changed
by
someone
who
has
access
to
the
graphical
interface,
which
is
restricted
so.
E
Yeah,
please
feel
free
to
maybe
I'm
actually
missing
some
parts
of
this,
because
I
haven't
used
that
before
but
yeah.
If
you
could
just
add
a
comment
or
some
or
two
in
there
and
we
can
go
from
there,
I
actually
have
to
run
to
the
meeting.
So
I
gotta
go,
but
thank
you
all
for
the
comments.
E
D
I
think
at
some
point
I
need
to
follow
up
with
eddie
and
arno,
but
that
may
be.
I
don't
know
if
we
want
to
continue
doing
that
now
or.
A
D
Yeah
well,
I
said
I
am
looking
at
that.
Tim
has
said
that
you
know
this
is
one
of
the
most
pressing
things
my
team
is
super
small.
At
the
moment
we
have
some
other
things
and
also
aaron's
out,
but
looks
like
I'll
be
able
to
focus
on
this.
Clearly,
there
are
some
other
approaches
here
that
I
need
to
catch
up
on.
B
D
Yeah
yeah
seems
seems
like
it
just
that
we
had
discussed
this
so
much
and
I
don't
think
we
have
any
other
items
so
we'll
either
call
it.
I
think,
or
continue
that
conversation.
D
Either
is
fine
with
me,
I
have
this
hour
blocked
in
for
this
meeting,
but
and
it
doesn't
look
like
we
have
any
other
items,
but
you
know,
since
we
also
didn't,
have
an
agenda,
I'm
expecting
folks
to
stick
around.
A
If
you
want
to
kill
the
recording,
we
can
talk
more
openly
about
it.
D
Sure
I
think
we
can,
in
the
like
testing
meeting
here
and
just
sort
of
have
our
follow-up
on
this
topic.
So
thanks
everyone
for
coming
to
testing
great
awkward
first
meeting
we
can
can
shoot
for
a
little
more
organized
for
the
next
one
and
we'll
move
on
to
our
just
follow-up
conversation
and,
oh,
I'm
not
actually
sure
if
I
can
from
the
web
client.
D
I
don't,
I
know
how
to
sign
into
the
sig
testing
account
from
my
one
password.
So
awesome.
E
D
I
I
do
actually
have
zoom
up
on
the
other
one
I
just
need
to
log
in
again,
so
it
should
just
be
a
moment.
D
Oh,
I
think
the
password
changed.
D
I
think
this
is
set
up
to
auto
record.
We
might
just
need
to
set
up
a
different
meeting
and
I
guess
we'll
need
to
trim
this
recording
sorry,
y'all.