►
From YouTube: Kubernetes kops office hours 20200117
Description
Recording of the kops office hours meeting held on 20200117
A
Hello,
everybody
today
is
Friday
January
17th.
This
is
cops
office
hours.
I,
am
your
moderator,
facilitator,
Justin,
Santa,
Barbara
I
work
at
Google
I
reminded
this
meeting
is
being
reported.
What
we
put
on
the
internet
and
is
also
subject
to
our
careers
kind
of
conduct,
which
boils
down
to
please
be
a
good
person
respectful
of
others.
We
have
an
agenda,
the
link
to
which
is
in
the
calendar.
Invite
and
I
will
paste
it
into
our
chat.
There
is
a
fairly
large
amount
of
or
a
large
number
of
things
on
there.
A
A
As
we
looked
about
last
week,
we
had
some
releases
ever
planned,
I
didn't
get
through
all
of
them,
but
I
think
where
he
I
can
discuss
them,
but
I
did
manage
to
do
basically,
as
the
time
ticks
closer
to
this
meeting.
I'm,
like
there's
no
point
in
doing
one
in
the
in
the
dying
hours
of
the
of
the
thing
before
talking
to
everybody,
but
I
did
push
up
some
a.m.
eyes,
which
was
interesting
because
it
found
a
issue
or
two
and
also
highlighted
the
fact
that
we
aren't
using
Debian
Buster
yet,
which
is
the
latest.
A
We
are
so
using
stretch
which
I
don't
recall
exactly
when
it
goes
out
of
support.
But
it's
not
too
long.
The
nice
things
about
there
was
a
blocker
to
using
Buster
in
terms
of
IP
tables
that
should
have
been
fixed
as
of
kubernetes
117.
I,
don't
know
if
anyone
has
formally
tested
that,
but
we're
actually
running
that
and
how.
B
A
That's
super
a
good
point
and
thank
you
for
that.
Like
real
world
experience
as
well,
I
think
yeah
I
think
what
I
can
do
is
I
can
build
a
sort
of
the
same
way
as
when
we
went
from
Jesse
to
stretch.
We
had
a
period
when
we
were
building
both
images,
and
so
users
can
just
change
the
word
well
in
the
past
it
was
Jesse
the
word
stretch
and
now
it
will
be,
they
will
change.
The
word
stretch
the
word
Buster
and
they'll
be
able
to
choose
to
use
the
Buster
image.
A
The
Buster
am
I
it
sounds
like
we
can't.
We
shouldn't
move
the
default
until
we've
resolved
most
of
the
cni
or
like
the
vast
majority,
this
unites
issues.
So
it
feels
like
it's
unlikely
to
happen
in
117,
but
we'll
see
or
four
cops
117,
but
we
can
we
can
see,
but
we
can
at
least
have
it
as
an
option
for
people
that
are
running
AC
and
Ida
is
compatible
with
well.
A
The
one
that's
installed
by
default
is
EB
tables,
don't
know,
I
mean
that
I
mean
NFC
balls,
no
sorry,
NF
tables,
yes,
and
so
anyway,
EC
and
I
that
or
the
C
and
eyes
that
either
use
NF
tables
or
have
the
sort
of
auto
detection
mode
will
just
work.
And
then
we
can
sort
of
see
what
happens
whether
we
should
do
more
I,
don't
know
if
that
sounds
like
a
reasonable
approach,
but
essentially
creating
an
image
and
making
it
optional.
So
parallel
to
our
stretch,
image
for
a
period
of
time
until
things
resolve
themselves,
yeah.
B
B
A
C
A
A
Anyway,
I
think
yeah
I
think
we
we
want
to
get
there.
We
we're
still
in
I,
think
catch-up
mode
and
I
think
we
I
think
John.
You
have
the
next
item
on
the
agenda,
which
we
can
come
to
in
a
second
which
is
like
progressing
on
the
catching
up,
and
then
we
can
sort
of.
We
don't
yet
have
to
worry
about
releasing
too
early
because
we're
not
yet
in
that
boat
mm-hmm.
A
B
Peter
thinking
about
adding
tests
for
each
network
plugging
in
periodic,
so
then
we
will
see
which
one
works
and
which
one
doesn't
I
would
say
that
1.17
is
a
bit
like
one
month
or
two
months
away,
all
right,
so
it
may
be
that
we
are
able
to
set
it
as
a
default,
then
say:
hey.
If
there
is
one
Network
plug-in
that
doesn't
work,
we
can
add
an
exception
or
something
the
other.
The
other
thing.
A
Is
I
feel
like
it
from
a
release
sort
of
stability,
point
of
view
we
should
have
a
release
which
has
them
optional
right,
so
people
can
like
try
it
mm-hmm
and
say:
oh
this
didn't
work
because
of
thing.
I
wasn't
expecting,
rather
than
like
switching
with
no
going
back.
Is
it
Mike
doesn't
matter?
We
can
still
like
to
change
the
default
and
still
have
a
way
to
go
back,
but
anyway,
I
get.
B
A
D
Yeah
and
there,
and
before
this
morning
there
hasn't
been
a
lot
of
progress,
is
like
three
things
landed
last
two
weeks,
so
some
of
them
are
peers
that
are
destined
for
1/16,
but
haven't
even
gotten
in
a
master.
Yet
so
so
I
am
thinking.
Those
are
eight
to
forty
eight
to
100.
Are
we
okay
on
the
approach
we've
gotten
for
dealing
with
priority
classes
of
flipping
the
feature
switch
on
for
the
older
versions,
I.
A
B
E
C
A
I
think
tallies
with
what
John's,
so
if
I
and
my
fix
was
your
fix,
so
you
fixed
it,
you
found
the
issue
and
fixed
it,
which
was
that
we
have
this
empty
block
and
I
was
like.
Oh,
we
should
stuff
that
ever
happening
again,
so
my
one
sort
of
like
just
sort
of
caches
and
under
the
carpet
yeah
and
you
like
actually
fixed
it.
So
if
we
can,
if
we
burns
your
fix,
that
also
fixes
it.
But
yes,
there
was
only
one
nine
I
can.
A
A
D
A
A
D
A
B
B
A
A
It
and
we
we
have
a
deprecation
policy
and
we
are
trying
to
narrow
that
window
and
so
I
think
yeah.
This
is
I,
think
that's
respawn.
We
can
put
in
like
the
the
release
notes
again
like
just
like,
although,
although
we
have
in
theory,
this
is
in
the
deprecation
window
like
this
is
not
a
good
idea.
All
night
is
not
a
good
idea
in
is
like
not
currently
passing
a
to
you
for
exactly.
A
D
A
D
A
D
Okay
yeah
so
now
the
core,
DNS
CDE
I,
think
that's
destined
for
1/6
and
that
needs
to
get
in
that's
assuming
we're
willing
to
go
ahead
of
kubernetes,
encore,
DNS.
D
C
That's
my
feeling,
so
we've
been
running
that
version
with
cop
with
kubernetes
111
internally
and
since
since
that
got
announced,
and
we
haven't
seen
any
issues
and
I
I
understand
they're.
They
call
so
I've
been
out
of
action
for
three
days.
So
I
responded
to
any
of
the
comments
on
it.
I
I
can
see
the
argument
for
not
jumping
ahead
and
potentially
documenting
it
instead
and
just
saying
you
could
take
advantage
instead
of
the.
C
A
A
B
A
A
When
it
comes
to
your
security
fixes,
we've
always
said
we
do
security
fixes
like
ideally
if
Cordy
and
s
had
like
I,
see
an
accordion
in
the
one
six
manifesto
used
for
later
than
one
six
and
less
than
twelve,
like
we
go
from
131
to
166
and
ideally
coordinates
with
about
1/3
to
right
would
have
something
in
the
1/3
frame
yeah.
That
is
not
proved
to
be
the
case
in
the
ecosystem,
so
we
we
don't
have
great
choices
here.
Yeah.
G
There
have
been
times
where
we've
patched
this
before
and
then
people
from
the
core
DNS
team
have
come
in
a
repo
and
reverted
back
because
they
said
we
should
stick
with.
What's
in
KK
last
year
happened
once
I
found
the
PR,
but
I
mean
this
is
a
big
enough.
Cee
ie
I'm,
just
saying
that's
what
what
has
happened
in
the
past,
so
we're
changing
that
I'm.
Okay
with
that
I
think
in
this
case
I
think
we
just
need
to
be
aware
of
that.
That's
all
do
we
know
why
KK
is
not
gone
to
166.
F
D
A
Great
okay
and
so
then
I
guess
we
can
add
to
our.
If
all
goes
to
plan,
we
can
do
one
16-0
beta-1
this
two-week
period.
Hopefully
anyone
think
ever
reason.
Why
not
all
right?
We
have
a
list
of
blockers
for
that
and
it
seems
like
that
is
good.
If
anyone
comes
up
with
more
blockers,
please
to
like
add
it
to
the
list
of
blockers
and
I'm
gonna
I'm
gonna
start
the
release
plans
section
and
while
I'm
doing
that
John
you
have
the
next
item,
which
is
around
rolling
update
concurrency.
Yes,.
D
A
A
One
which
labels
I
think
was
my
I
had
like
five.
Second
look
at
that
in
terms
of
like
thinking
about
that,
I
just
don't
know
whether
we
shouldn't
think
about
whether
that
needs
to
be
kubernetes
like
should
we
propose
something
that
can
be
used
outside
of
sorry,
a
lot
of
cops
yeah,
not
just
like
kubernetes.
Why,
like
some
sort
of
general
annotation?
On
the
other
hand,
we
can
just
recognize
both
of
them.
D
A
Think
we
should
preclude
it,
I,
don't
think
it
precludes
it.
It's
just.
Do
we
want
to
support
that
I
guess
if
there
is
so,
if
someone
wanted
to
build
such
a
thing,
why
no
I
believe
if
people
have
built
such
a
thing,
if
I
wanted
to
put
such
a
thing
into
cups,
for
example,
I
would
ask
whether
it
like
needs
to
be
in
cups
like,
or
should
it
be
like
an
external
controller
that
can
be
like
is
there?
Is
it
tied
to
cops
right,
yeah.
C
A
Okay,
that
is
cool
I,
will
take
a
look
at
that.
Thank
you
for
that
context
there,
and,
but
if
we
do
both
for
your
PRS,
that's
that's
good
yeah
and
then
another
I'm,
more
controversial
and
a
deprecating,
the
UN
alpha-1
API
and
you
mean
deprecated.
You
mean
removed
John
well,.
A
The
removal
I
feel
like
we
probably
should
be
more
like
gradual
about,
although
it
is,
we
also
have
acknowledged.
We
have
a
lot
of
removals
coming
from.
We
were
other
applications
and
which
I'm
super
excited
to
see
like
a
lot
of
the
cleanups
on
that
are
gonna,
be
great
I,
don't
know
if
we
said
we've
had
to
pace
ourselves
a
little
bit,
maybe
I
don't
know
and
like
keep
users
from
too
much
churn.
A
D
A
A
A
A
A
A
D
A
Out,
that's
true.
Alright,
let's
say
about
this:
yes,
deprecated
sherry
picked
that
back
to
1/16,
which
we
haven't
released.
Yet
people
will
see
it
say,
we're
gonna
release
it
remove
it
in
118,
and
if
people
do
come
up
with
reasonable
high
objections,
we
can
always
move
it
backwards.
Okay,
you're
right,
I'm
persuaded
all
right.
Everyone
is
anyone
happy
with
that.
I
should
say:
okay,
Rodrigo
custom
config
for
Etsy
de.
Thank
you
jumping
up.
I
custom
confit
Rodrigo
custom
conflict
for
ed
to
be
using
a
CD
manager
that
way
to
do
so.
Currently,
yeah.
H
H
A
Five
minutes
the
okay,
the
there
is.
Certainly
we
don't
we're
in
the
same
boat.
We
need
to
expose
options
sort
of
individually,
so
we
sort
of
manage
them.
There's
no
like
generic
sheet
passed
through.
So
we
have
a
cheat
pass
through
like
on
API
server,
where
we
can
pass
any
flag
and
we
don't
have
so,
but
we
don't
have
either
one
today
so.
H
I
was
thinking
of
just
setting
so
digging
through
at
city
manager
code.
Currently,
it
just
looks
at
and
sets
a
bunch
of
em
variables
for
the
actual
sed
process.
I
was
thinking
in
just
passing.
A
key
value
pair
to
EDD
manager
said
it
in
the
pod
config
so
that
it
gets
passed
in
as
a
variable
to
the
actual
container
and
sed
can
grab
that
and
pass
it
down.
Yes,.
H
A
I
think
that's
reasonable
as
I
like
a
parallel
to
like
the
flags
where
we
have
a
sort
of
pass
through,
that's
not
guarantee
compatible
and
we
effectively
like
allow
you
to
specify
things,
give
you
control
and
then,
if
we,
if
this
is
a
and
use
case,
we
can
wear
I'm
sure
there
will
be
some.
We
can
like
identify
the
common
patterns
and
expose
them
as
managed
fields
that
have
more
guarantees
from
them.
Yeah.
A
D
A
Can
if
@cd
removes
the
n,
VAR
or
renames
the
n
VAR,
then
it's
it's
on
you
yeah
and
just
sort
of
like
I
guess:
I
would
mirror
the
power,
mirror
the
pod
thing
with
n
name
and
value,
or
we
should
do
key
value
pairs.
I,
don't
know,
I,
don't
know
where
we
are
today
on
that.
H
Okay,
when
it's
at
least
set
that
or
just
key
value
pair
and
you
can
set
whichever
ones
you
want
or
sorry
can
you
repeat
that
do
we
want
to
limit
which
ones
folks
consent.
A
Or
just
let
them
go
at
it,
I
think
I
I
would
imagine
you
should
be
able
to
set
whatever
n
VARs
you
want.
It'll
get
set
in
the
etsy
D
manager.
I
think
will
tell
at
C
D
manager
to
pass
down
at
least
a
subset
of
them.
Whatever
it's
doing,
and
if
there's
a
conflict,
I
don't
know
what
we
should
do,
I
presume
we
should
go
on
the.
H
H
A
A
Anyway,
so
there
is
a
new
entity
manager,
release
that
will
likely
land
fair
release
or
will
try
to
put
in
fairly
soon
but
yeah.
We
should
also
get
this
in
there
as
well
good.
A
B
B
Container
SELinux
I
think
it's
something
required
by
selinux
to
make
any
of
these
to
work.
Frickin
container,
okay,
yeah.
So
it's
and
the
problem
is
not
that
one.
The
problem
is
that
is
built
only
for
CentOS
and
requires
sin.
Toasts
version
dependencies
at
some
point,
I
think
Amazon
decided
to
drop
the
versioning
from
Santos,
so
things
don't
work
anymore.
I
mean.
D
B
A
B
B
A
A
B
B
A
B
A
B
A
B
Us
through
that
right,
so
right
now
we
are
using
kubernetes
masters
to
test
our
PRS
and
pretty
much
everything
to
me
seems
not
such
a
good
idea
like
we
are
not
developing
kubernetes
features,
and
we
are
anyway
at
least
three
months
behind
Abernathy's
so
would
make
sense
to
at
least
to
use
stable,
latest
or
I.
Don't
know
some
stable
something
instead
of
the
master
branch,
maybe
less
annoying
again
yeah.
A
I
think
so,
by
way
of
history,
we
used
to
run
cops,
pre
submits
on
every
PR
in
KK,
and
it
was
sort
of
the
way
AWS
test
or
testing
for
eight
of
us
was
done.
We're
generally
moving
away
in
KK
from
the
idea
of
testing
on
cloud
providers
on
every
PR
moving
to
periodic
since
Ted
and
like
more
mocks
and
things
like
kind
things
like
that,
precisely
to
get
the
same
stability
there
like
it,
it
was,
it
was
hard
on
users,
who'd
be
like
I,
don't
know
anything
about
idos.
A
Why
is
my
tests
like
I,
am
touching
it
as
my
test
flaking
on
any
of
us
or
why's,
my
just
faking
I'm
GCP,
otherwise,
my
test
braking
on
whatever
else
like
scenario.
It
is
just
because,
like
when
you
have
enough
scenarios
that
it
just
happens,
I
agree
with
you
that
for
most
of
our
periodic
tests
like
when
we're
testing
the
different
scenarios,
we
probably
do
want
stability
and
so
like
moving
back
a
little
bit
would
be
fine.
I
think
we
do
also
want,
though,
to
catch
when
there's
a
regression.
A
When
there's
a
breaking
change
made
in
KK
that
breaks
a
WS,
we
want
to
know
about
it
or
breaks
cops.
We
want
to
know
about
it
as
soon
as
possible,
so
we
want
some
periodic
s--.
Ideally
that
run
pretty
fast.
It
are
tied
to
the
master
branch
and
they
also
give
some
good
signal
on,
like
you
know
some
of
the
tests
that
are
in
there
much
flakier,
so
we
can
like,
but
we
don't
so
we
should
have.
We
should
have
some,
but
I
asked
agree
if
you
want
to.
A
Older
versions,
don't
need
to
be
well
whatever
it
is
if
you
want
to
back
off
a
little
bit,
but
we
do
so
that
you
have
more
stability.
That
seems
like
a
reasonable
approach.
We
do
also
want
to
know
we
do
want
to
give
signal
on
the
master
branch
as
well.
We
surround
the
head
of
the
branch
as
well
so,
for
example
like
if
they're
thinking,
you're
planning
of
new
1/16
release
like
it
would
be
nice
to
tell
them
yeah
go
look
at
the
the
cups,
the
cups,
this
particular
coffee
shop,
to
see
where
it
passes.
H
A
There
are
no
pre
submits,
there
are
no
cake,
hey
pre,
submits,
oh
I,
see
what
you're
saying.
Yes,
sorry,
yes,
and
then
there's
the
cops
test
as
well.
Yes,
yes
well,
but
we
also.
We
also
want
to
know
if
upstream
breaks
things,
so
it
often
happens.
The
upstream
will
make
a
change
that
does
break
things.
Yes,.
A
B
A
B
B
B
A
And
I
think
if
you
know
like
I,
know
that
people
are
adding
you
and
Peter
and
others
are
adding
like
tests
for
like
different
scenarios,
those
I,
don't
those
I,
don't
have
a
particular
preference
on
whether
they
should
keep
the
review
on
whether
they
should
be
like
more
stable
or
not
and
I
think
I
think
as
I
think
John
and
you
suggested,
like
the
cops
CI
tests.
I
think
it's
fine
for
those
to
use
yeah
some.
A
Whatever
you
say
like
the
the
release
table,
that's
the
last
stable
release
of
kubernetes,
which
will
still
be
ahead
of
where
we
are
right.
It
will
still
be
like
today
it
will
be
118,
zero
alpha
one
I!
Think:
oh
yeah,
that's
a
different
question.
So
it
should
it
should
it
doesn't
need
to
pin
to
master
and
I,
don't
know
whether
it
should
pin
to
like
118
zero
alpha
one
right
now
or
117.
What
I
think
they
just
true
release?
I,
don't
know
if
anyone
knows
there's
a
view
on
that.
I
think.
A
A
F
Yes,
I
put
a
link
to
the
test
grid.
We
added
a
bunch
of
jobs
for
different
distributions.
Pac-Man
helped
out
that,
and
we
identified
a
few
issues
got
those
fixed
so
now
sent
out
seven
needed
a
newer
ami
for
the
newer
instance
types
and
BME
na
we
have
the
rel
that's
passing
above
to
16
and
18.
We
had
a
core
OS
and
flat
bar
as
well,
so
those
are
all
under
test
now.
He
also
added
older
kubernetes
versions
for
what
we
support.
This
is
where
we
identified
that
one
eye
is
broken.
F
I
was
talking
to
the
cig
release
folks
couple
days
ago
and
the
staple
one
staple
to
staple
three
markers
that
they
use
have
been
too
problematic
and
they
want
to
switch
to
more
version
specific
markers,
so
I'm
gonna
update
those
to
be
page
116,
117,
118
and
I'll
use
the
the
latest
staples
stuff
that
we
just
talked
about
and
then
we're
talking
about
adding
new
jobs
to
test
core
DNS
and
as
well
as
the
remaining
CNI
plugins,
that
we
don't
have
tests
for
that's
kind
of
our
update
wow.
That
is
super
impressive.
Thank
you.
A
And
I
have
a
I
have
a
little
theory
that
the
the
next
item
is
gonna
deal
with
some
of
the
flakiness.
So
we
interested
to
see
whether
that
theory
proves
right.
I'm,
sorry
so
happen.
You
dropped
for
a
moment
and
we
I
think
I
think
we
agreed.
We
basically
said
what
you
what
you
were
gonna
do:
a
race
are
gonna,
have
the
RCI
is
gonna
be
on
the
latest.
The
cop
CIA
is
gonna,
run
against
the
latest
release
version
of
Cooper.
That
is
tagged
version
of
kubernetes.
A
Maybe
that's
the
correct
word
so
it's
moment
of
you
1:18
0,
K,
so
stable
laters,
I
can
say
but
yeah
the
last
we
were
debating
like
117
one
versus
118,
zero
alpha
one
right
now,
but
we
went
with
a
more
aggressive
one.
Eighteen,
zero
alpha
one,
but
still
still
a
fixed
version
that
shouldn't
change
too
much.
And
if,
if
something
goes
wrong
we
can.
A
Maybe
they
have
a
real
name
for
it.
You
get
to
get
a
name.
We
are
running
on
time,
so
gonna
try
to
rush
us
through
the
last
three
things
we
have
on
our
agenda
and
leave
one
minute
for
the
release.
Funny
I,
guess
hack,
when
you
just
did
changing
the
default
instance
types
which
I
think
is
a
good
idea
that
I
think
had
a
lot
of
Plus,
Ones
and
LG
TM.
Some
I
think
we
just
want
to
sort
of.
Oh
I,
guess
I
feel
this
is
a
different
topic.
No,
that's
the.
A
A
B
A
A
Whichever
one
you
prefer,
we
just
need
someone
to
create
three.
Maybe
I
instead
want
to
start
creating
release,
notes
like
once
we
once
we
have
the
top
it'd
be
creative
in
general.
It
would
be
good
if
each
of
the
each
of
the
pair's
had
the
release
know
in
its
wicked,
like
discuss
them
together.
Okay,.
B
A
A
The
the
logic
behind
putting
the
package
dependencies
is
that,
if
you
were,
if
you
were
not
installing
doctor
right,
then
you
wouldn't
need
to
like
pull
in
the.
So
we
sort
of
keep
the
what
you
wouldn't
need
to
pull
and
you
wouldn't
need
necessary
need
to
install
those
other
system
packages
OS
packages,
so
by
my
fund
by
by
putting
them
next
to
each
other.
Then
we
sort
of
don't
just
end
up
with
this
big
list
of
packages
where
no
one
really
knows
what
they're.
A
A
F
Is
a
very
old
PR,
almost
a
thousand
PR
told
basically
adding
support
for
a
file
system
state
store
which
would
not
be
used
in
practice
only
used
in
CI
environments
as
a
read-only
process.
Basically,
the
workflow
would
be
you
copy
your
real
estate
store
down
to
a
local
directory.
You
could
point
your
cops
binary
at
this
local
state
store
with
whatever
changes.
You're
gonna
make,
whether
it's
a
change
to
the
cluster
manifest
or
a
new
version
of
cops
and
then
you're
able
to
do
a
cops.
You
know
for
place
and
cops
update
and
it'll.
F
Allow
you
to
preview
changes
without
changing
any
files
in
the
state
store,
because,
right
now,
if
you
update,
cops
and
then
run
cops,
update,
it'll
update
the
files
in
the
state
store,
even
though
it
hasn't
updated.
Your
cloud
provider
resources,
which
means
new
instances
that
are
launched,
are
gonna,
be
pulling
down
new
versions
of
files,
even
though
you
haven't
actually
brand
cops
update.
Yes,
so
I.
A
D
Not
for
why
cuz
you
have
to
explicitly
change
your
state
store
anyway,
yeah
for
us
as
well.
We
have,
we
have
s3
buckets
just
for
temporary
states.
Doors
are
our
flow.
Is
we
have
our
own
cluster
spec?
We
then
do
a
cops,
create
cluster
dry,
one
into
a
temporary
state,
store,
delete
the
state
store
and
then
take
the
resulting
manifest,
modify
them
and
then
do
a
replace
okay.
A
F
A
Right,
like
I
just
worry
about
like
I,
don't
think
I
think
if
you
export
the
variable
I,
don't
have
any
way
to
look
to
know.
We'd
have
to
actually
like
ask
like:
did
you
set
your
cup
state
store
to
a
file?
I,
don't
know
made
us
I'm
sure
you'd
see
like
the
URLs.
That's
all
right,
but
yes,
I
think
we
can
I
think
we
should
sounds
like
there's
a
valid
use
cases.
We
should
merge
this.
We
might
need
to
add
a
warning
later.
Give
her
a
message:
okay,.
A
And
we
are,
or
we
can
merge
this
and
then
I,
don't
know
I
guess
you
can
add
to
that.
Why
me?
Yes,
we
are
at
time
I.
Basically
I
copied
the
release
plan
for
the
previous
two
weeks
into
this
week,
which
is
basically
a
big.
The
headline
I
think
is
the
116
0
beta
1
per
John's
list,
and
then,
ideally,
we
can
roll
to
117
0
alpha
2
I.
A
Yes-
and
we
can
also
do
I-
think
we
have
a
bunch
of
PRS
that
are
queued
up
and
we
should
do
a
140
and
a
114
patch
release
for
those
but
I
sort
of
am
using
the
beta
1
as
the
driver,
because
that's
sort
of
whatever
we
get
into
the
beta
1.
We
can
then
say
like
do
we
cherry
pick
it
back
as
sort
of
that's
sort
of
the
way
I
think
about
it?
Oh
yeah,
we're
still
until
we
released
116
zero.
B
We
are
getting
many
questions
on
slack
like
when's.
The
next
release
when's
the
next
release,
when's
the
next
release.
It's
like
a
daily
question.
Maybe
it
would
be
a
good
idea
to
have
a
policy
for
this
like
say
one
week
after
the
kubernetes
releases
or
something
not
sure
if
possible,
but
would
address
the
questions.
I
think.
A
B
A
Don't
why
don't
we
work
we're
over
time
ready?
Why
don't
we?
Why
don't
we
tell
people
to
come
in
two
weeks
and
discuss
this
topic
you're
like
in
general,
and
we
can
like
just
keep
going
I
think
also,
if
we
like
make
more
progress
on
the
releases
that
tends
to
answer
some
of
these
questions,
no
problem.
Thank
you
Thanks.
If
there's
anything
else,
if
everyone
needs
to
get
on
with
their
day,
I
can
see
Vall
rotating
in
their
chairs.
Looking
at
the
lunch
time,
all
right,
here's
everyone,
I'm
very
happy
two
weeks
and
season
thanks.