►
From YouTube: Kubernetes kops office hours 20191206
Description
Recording of the kops office hours meeting held on 20191206
A
Hello,
everyone
today
is
Friday
December
6th
2019.
This
is
cops
office
hours.
I.
Am
your
moderator,
facilitator,
Justin,
Santa,
Barbara,
I
work
at
Google
reminder
this
meeting
is
being
recorded
and
will
be
put
on
the
Internet
and
to
be
mindful
of
our
code
of
conduct
which,
for
us
sound
to
be
a
good
person.
I
have
pasted
a
link
in
the
chat
to
our
agenda
document.
A
Please
feel
free
to
add
your
name
and
any
items
that
you
would
like
to
talk
about
to
that
agenda.
There
is
a
lot
on
the
agenda,
so
we
will
I'll
try
to
keep
us
on
track
and
please
do
add
things
to
make
sure
we
get
have
time
to
get
to
the
more
track
that
we
don't
for
next
time.
Otherwise,
I
suggest
we
jump
right
in
Peter.
You
have
the
first
item
on
the
agenda.
Oh
I,
don't
see?
Oh
yes,
oh
you're,
sorry
I
thought
I
saw
yes,
hello,
Peter.
B
So
we
ran
into
an
issue
where
gaming
sets
get
updated
on
a
cop's
update
cluster,
but
sometimes
they
have
dependencies
that
would
only
be
met
on
a
new
node
that
would
occur
during
a
rolling
update.
The
specific
issue
that
I
ran
into
it
before
was
went
and
abling
the
PWS
Authenticator.
It
said
it
requires
source
or
some
config
file.
The
volume
mounted
to
the
host-
and
that
gets
placed
on
instance
launch,
and
so
when
the
daemon
set
is
added.
Deep
pods
are
in
a
crash
loop
because
they
don't
have
the
old
pot.
B
A
It
does
sound
reasonable.
It's
not
it's,
not
perfect
right
in
that.
If
a
pod
well,
what
does
Anjali
do
if
a
apart
if
I
know
a
container,
if
the
process
crashes,
does
it
update
or
not
or
tell
me
if
it's
a
new
but
synced
only
if
it's
a
new
pod?
Okay?
So
it's
actually
so
it's
good.
Okay,
like
we've,
talked
about
strategies
involving
multiple
Damon
says.
It
feels
like
a
reasonable,
easier
thing
to
do
until
we
get
there
I
guess.
A
B
A
I
mean
sound,
actually
yeah,
it's
not
terribly
problematic
cuz,
you
just
like
create
a
new
one
with
a
new
name,
but
it's
certainly
ugly.
The
I
think
we
should
probably
try
this
out
with
the
one
you
hit.
The
address
I
intend
to
cater
and
get
some
miles
on
it
that
way
in
the
118
branch
or
the
118
target,
so
I
think
that's
reasonable,
and
then
we
can
see
how,
if
you
tried
this
yourself.
A
This
is
all
theoretical,
so
far,
okay,
so
yeah
I'd
be
I'd,
be
in
favor
of
getting
you
on
to
like.
For
those
don't
know,
overcooked
con,
we
had
a
wonderful
meeting
of
our
maintained
errs
and
some
I
think
we
saw
some
of
it.
Someone
some
of
the
other
people
there
as
well
so
great
to
see
everyone,
and
we
welcome
some
new
maintainer.
So
congratulations
and
welcome
and
thank
you
to
the
camera.
Peter
was
and
I
actually
thought
so
he
did
already
was
it
Peter
and
Ryan?
Were
you
not
officially
maintainer
previously
Curt.
B
A
Did
we
promote
you
Eric?
You
were
a
maintainer,
but
we
see
tap.
We
made
a
maintainer
in
his
in
absentia,
just
just
we,
we
forced
him
to
be
a
maintainer,
so
thank
you.
Russia
is
all
of
them,
but
anyway
we
also
did
a
bunch
of
releases,
and
so
we
are
basically
master
is
I
believe
caught
up
at
this
point
and
we
are
targeting
master
to
118.
Now
switch
is
in
three
ish
months,
I!
Imagine
so
we
we
have
a
lot
of.
A
We
have
a
lot
more
flexibility
and
breathing
room
to
put
things
on
the
master
branch.
Then
we
have
previously
had
in
the
past.
We
should
probably
communicate
that
to
our
users
that
may
be
running
off
master,
who,
in
the
past
of
had
a
rough
ride
because
of
needing
to
build
their
own
versions
and
now
may
have
rough
ride,
because
it's
gonna
be
fairly
on.
A
A
So
we
should
look
at
that,
one
as
well.
I
I
thought
it
as
the
we
are.
We
are
I'm
hoping
we
will
be
able
to
get
more
of
the
open
PRS
in,
but
yes,
I
apologize
for
the
backlog
today,
so
yeah,
that's,
but
it
we
should.
Probably
it
doesn't
conflict
with
Peters
idea
of
10
years
to
undelete.
Do
you
think
John
or
no.
A
So,
let's,
let's
give
it
a
gallon!
Thank
you,
Peter
Z,
tab,
I,
don't
know
if
you
are
here.
I
don't
see
you
here,
but
there
is
a
PR
that
I
need
to
review,
which
I
will
review
yes
and
so
back
to
Peter
about
our
which
actually
is
going
to
help
with
I
think
a
future
conversation,
but
yes
go
for
it.
Yes,.
B
During
cubed
on,
we've
decided
that
we
should
have
this
discussion
on
formalizing
a
deprecation
policy,
there's
a
few
different
things
that
we
need
to
decide
on
forming.
You
know
few
components
of
this
related
to
versions
of
cops
that
we
can
cherry-pick
back
to
persons
of
kubernetes,
that
we
should
support
flags
to
the
kubernetes
components
that
get
removed
those
sorts
of
things.
So
there's
a
discussion
about
the
issue
and
I
listed
a
few
Peters
that
would
be
impacted
by
those
but
yeah.
A
D
D
A
E
I
think
my
only
worry
about
marriage
in
those
pull
requests
before
we
have
a
formalized
application.
All
I
see
is
it's
something
I've
seen
in
discussions
from
sig
instrumentation
before
in
terms
of
their
move
to
a
structured
logging
of
going.
We
can't
we
can't
really
do
this
before
we
actually
notify
people
at
least
notify
people,
because
I
think
that
there
is
a
risk
that
someone's
been
running
for
two
years
with
that
flag
and
it's
not
broken.
Yet
it's
not
necessarily
warn
them.
Yet
that's.
F
A
Now
and
it
won't
affect
them
for
quite
a
while.
We
also
technically
feature
flags
are
not
a
flag.
It's
like
a
little
bit
of
a
cheat,
but
yes,
but
feature
flags.
I've
always
been
I've,
always
been
I'd,
say
more
like
a
an
odd
category,
so,
but
they
don't
even
have
to
follow
the
same
duplications.
But
yes,
I
think.
Yes,
it's
a
good
point.
If
we,
if
we
announce
it
pre
announce
it
now,
I
think
that
will
be
fine
in
terms
of
like
turning
off
the
feature
flag.
That
makes
sense.
A
A
The
cops
in
theory
supports
the
very
first
version
of
kubernetes.
We
ever
ran
so
like
there's
still
code
in
there
for
one
least
one
five
for
sure
yeah
there's
at
least
one
there's
three
one
five
and
while
that
is
good,
I,
don't
think
any
I.
Don't
think
we
want
to
be
encouraging
anyone
to
be
running
one
four
one
five
and
we
need
to.
We
shouldn't
I.
A
C
Technically
we
have
the
same
support,
but
that
means
I
think
what
we,
how
we
have
it
word
it
is.
We
support
what
kubernetes
versions
are
supported.
Therefore
you
know
so
with
our
160,
so
we
support
116
115
14,
which
means
we
support
cops
on
all
those
versions
which
114
then,
if
that's
the
same
policy
that
supports
three
versions
lower
or
something
like
that
and
so
I
think,
no
matter
what
we
need
to
clarify.
C
A
lot
of
that,
and
and
in
my
opinion,
also
new
contributors
who
come
in
and
look
at
the
code
anytime,
they
see
the
greater
than
kubernetes
1.5
in
the
code
or
my
favorite
is
the
one
that
I
think
Eric
was
referencing
that
the
ones
that
we
have
like
free,
1.5,
it's
like
okay,
it's
fine
and
I-
think
it's
fair
for
us
to
also
say
it's
fine
to
run
this
older
stuff.
If
that's
your
requirement
and
your
choice
and
all
this
stuff,
we
just
think
that
we
can't
provide
you
the
testing
and
support
that
you're.
C
A
I
feel
like
the
answer
of
like
double
the
window
feels
like
when
we
forget
with
the
windows,
3
or
4,
feels
like
the
right
answer:
I'm,
not
sure
that
the
the
intent
of
like
the
cops
like
sports,
three
versions
back
and
we
support
three
versions
back
with
comps.
So
therefore
that's
I,
don't
know
that
was
deliberate,
but
it
certainly
feels
like
a
reasonable
number
I
guess
we
should
circulate
that
policy
and
see
how
people
feel
about
it.
I
don't
know
with
that.
So
that
would
so
concretely
today.
A
Over
yes
for
crew
Denny's,
alright,
so
in
theory
cops
114
is
going
out
of
support,
so
115
with
us
voted
one
in
cops
from
15
sports,
115,
114
113,
so
we're
talking
about
dropping
support
for
112
at
the
end
of
the
year.
So
the
question
for
the
people
on
this
call
is:
would
that
hurt
anyone
like
112?
Is
anyone
still
running
112
or
earlier?
Oh,
yes,
yes,
yeah!
It's
certainly
a
very.
A
B
A
Get
there
and
like
you,
do
yeah
pick
a
number
like
1
8,
1
9
like
another
another
3
and
we'll
definitely
start
there
and
then
we'll
gradually
say.
Look
like
you
shouldn't
be
running
more
than
and
release
his
back
and
I.
Imagine
when
you
go
to
4,
it
sounds
like
we're.
Gonna
go
to
4
because
it
is
really
awkward
around
now,
and
so
that
would
give
us
if
we
did
double
that
would
be
8,
which
is
a
lot
like
2
years.
That's.
D
B
B
A
But
then
yes,
I
said
that,
doesn't
we
then
have
to
support
that
version
of
cups
as
well,
for
which
we
sort
of
do
anyway,
for
that
sort
of,
like
a
bug
in
upgrade
would
be
a
bug
that
would
be
cherry
picked,
but
this
is.
This
is
indeed
complicated.
I'd
suggest
what
we
start
with
is:
let's
pursue
some
form
of
strategy
on
this
issue
and
pick
a
pick,
a
number
that
is
not
gonna
cause
anyone.
Any
problems
like
1/8
1/9
pick
some
threshold
on
10
like
anything
well.
A
Is
it
like
that
and
be
like
I
presume
all
right,
I
presume
1/8
would
not
cause
anyone
any
problems
because-
and
that
also
means
we
get
rid
of
the
1
6
1
5
stuff.
That
will
actually
clear
a
lot
of
stuff
I
think
so
it's
a
it's
a
fairly
easy
starting
point
to
say
1
8,
and
then
we
can
push
the
window
at
all.
I
feel
like
when
it
is
too
far,
and
but
we
can
also
then
night
when.
B
It
makes
a
big
difference,
though.
It
definitely
allows
us
to
trim
a
bunch
of
stuff,
and
that
gives
us
stack
deaths
in
the
right
direction.
So
I
think
that's
good,
plus
I
feel
like
anyone.
That's
running
kubernetes
1/8
is
probably
not
running.
Cops,
116,
correct,
they're,
pretty
not
funny
if
the
master.
B
E
C
A
Are
you
saying
even
that
might
be
too
tight,
I'm,
not
saying
anything
more
than
that?
Okay,
one
six
is
the
one
that
has
the
big
payout
of
payback
right.
So
we
could.
We
could
be
more
conservative
and
do
one
seven
or
something
and
it'll
say
look
at
us.
We
can
change
it
to
ones
drop
support
for
less
than
one
seven,
which
would
also
mean
we
smaller.
A
H
B
B
The
default
route
starts
with
a
zero
zero
zero
decimal,
and
so
we
can
update
the
naming
pattern
for
terraform,
but
the
problem
will
be
that
when
they
go
to
apply
it
because
the
resource
name
changed,
terraform
is
gonna,
try
to
delete
the
route
and
recreate
it
which
could
be.
You
know,
problematic
for
their
network.
So
we
need
a
way.
B
A
A
Well,
in
that
case,
I
mean
I,
guess
I
as
I
understand,
people
are
moving
to
terraform
0.12
like
it.
It's
not
like
people
are
sticking
on
0
11.
We
could
continue
to
support
without
too
much
trouble.
This
is
here
yet
another
future
deprecation.
Okay,
first
of
all,
like
terraform,
0.11
style,
I,
don't
know
if
that
feels
like
it
might
be
important
to
people
yeah.
A
Yeah
there
is
the
I'm
just
worried
that
people
by
default,
most
people,
are
going
to
be
using
terraform
0
to
12
to
endowment
terraform
they're
gonna,
like
notice,
doesn't
work
like
what
does
this
serve?
Cops
doesn't
work,
and
so
we
should
probably
make
it
work
with
that
scenario.
I
do
agree
that
we
should
have
yeah
a
feature
flag
would
be
great
to
support
it.
One
way,
the
other
without.
B
B
A
B
A
That
feels
like
yeah
doing
that
and
feels
helpful
and
then
a
release,
note
and
and
yeah
I
think
and
then
we
could
also
change
it
to
when
we
write
terraform
0
11,
the
older
form,
if
we
I
don't
know
if
we
can
put
a
less
than
12
requirement,
then
the
message
we've
got
a
comment
above
it
and
then
hopefully
it'll
be
very
obvious
to
users
like
that.
It's
a
like
why
the
problem
arises.
Are.
A
You
look
at
the
cluster
I
guess
and
with
the
naming
of
the
resources.
Is
that
the
idea?
Well?
They
should
be
the
same,
but
I'm
wondering
if
we
could
first,
we
could
like
pave
the
way
right.
We
could
start
labeling,
like
am
I
using
Terra
for
my
using
something
else
somewhere.
The
other
thing
we
could
do
is.
A
So
that's
possible.
We
could.
We
could
certainly
log
a
warning
if
you
are
big
all
morning,
if
you're
doing
terraform,
we
make
terraform
0:12
that
a
full
customer
table,
we
don't
have
the
version
every.
What
I
was
like
I
was
gonna,
say,
but
only
like
log
whenever
that
happens,
but
then
that
would
have
people
that
just
went
0.12
I'm
just
a
normal
update,
so
we
have
to
why
don't
we
start
with
the
new
behavior
which
feels
like
we
should
do
it
and
then
we
can
add
appropriate
release,
notes
and
see
how?
A
B
There's
nothing
really
to
say
here
other
than
I'm
interested
in.
This
is
more
of
a
statement
of
intent
than
a
discussion
at
this
point.
We
don't
have
enough
time
for
discussion,
but
my
intent
is
to
to
move
the
GC
stuff
to
do
it.
Ga
I
mean
in
tackling
that
I
wanted
to
to
talk
more
about
that,
but
I
have
I
thought
we
might
have
more
time
to
talk,
but
I,
don't
think
that's
gonna
happen
today.
So
I'll
catch
up
with
Justin
different
Pamela.
That's.
D
B
A
B
F
So
last
time
you
were
talking
about
adding
support
for
container
D
I,
think
I
started
doing
some
other
thing:
I
wanted
to
split
container
D
from
occur,
because
at
the
moment
they're
kind
of
two
separate
things,
and
even
the
discussion
about
the
tarball
distribution
would
not
work
now
because
container
disservice
wouldn't
be
part
of
that.
I
think
you
don't
manage
it
in
any
way.
At
the
moment
right.
The.
A
F
This
is
how
I
started
so
after
that
one
thing
led
to
another:
I
asked
on
dev
chat
if
it
would
be
interesting
to
try
and
continue
and
right
now
it
works
to
me
it
doesn't
look
very
bad
and
there
weren't,
many
workarounds
or
I,
don't
know,
but
it
became
kind
of
huge
PR
61
files,
something
yes
I.
10
of
them
are
important,
but
I
don't
know
if
this
is
interesting
and
I
think
it
should
well
it's
good
quality
and
should
be
included.
A
F
I
Also
interested
in
this
oil
also
review
it.
It's
one
of
the
things
I
mentioned
that
you've
got
one
of
the
things
I
just
skimming
through
the
PR.
That
I
probably
would
want
to
ask
real
quick.
We
want
to
keep
adding
support
for
separate
sea
and
ice
in
with
separate
structures,
or
do
we
just
want
to
create
a
separate
structure
that
says
hey,
support
and
sea
and
ice,
and
we
build
everything
under
that.
Otherwise,
we're
gonna
have
the
docker
container,
the
next
cutemon
or
whatever
else
someone
else
wants
all
in
there.
A
I
A
I
I
I
A
It's
a
good
question:
let's
have
a
look
at
the
PR
and
try
to
figure
it
out,
so
I
mean
I.
Think
the
I
it's
great,
that
there's
the
most
looking
like
there's
this
container
runtime
string,
which
is
a
how
it's
called
like
a
selector
or
something
selector,
but
the
idea
of
a
string
that
tells
you
which
of
a
set
of
alternatives
is
the
active
one.
That's
really
important
for
apply
semantics,
that's
great!
We
don't!
A
I
Think
there's
quite
a
few
people
trying
to
clean
up
a
docker
one
as
well
like
there's
folks
trying
to
move
how
it
gets
installed
because
of
all
the
dependency
issues.
We've
had
there's
folks
trying
to
install
it
in
different
ways,
so
it
might
not
be
a
bad
spot
that
we're
in
or
just
start
some
of
that
up.
I
guess.
A
The
other
thing
is,
my
understanding
is
that
docker
itself
is
going
to
would,
rather
that
we,
as
of
a
couple
of
months
ago,
I,
don't
know
that
this
has
changed.
The
docket
stuff
would,
rather,
that
we
used
container
d
on
servers,
and
so
docker
is
itself
a
rat
or
something
some
sort
of
wrapper
around
container
D.
This
is
just
my
understanding,
please
feel
free.
Anyone
to
correct
me
if
they
are
I
am
wrong,
but,
and
so
we
should
be
like.
F
I
think
it's
if
we
get
it
into
some
version
at
the
moment
by
default,
it
still
installs
docker
and
everything
as
it
is
not
so
we
can
experiment
with
it
and
see
if
it
actually
works
on
servers
without
too
much
trouble,
because
it's
a
bit
different
I
mean
loading
stuff.
There
are
different
features
in
there,
but
I
don't
know
why,
regarding
the
upgrades,
I
also
sent
a
PR
for
pulling
packages
with
dependencies
instead
of
using
rpm
and
Deb
to
install
them
in
the
startup.
There
are
still
some
caches
in
that
function.
A
A
J
Yeah
so
I
just
I
feel,
like
I've,
seen
a
few
PRS
and
work
around
the
rolling
updates
and
I
kind
of
just
wanted
to
see.
If,
like
we
can
come
to
a
consensus
or
get
something
going
because
I
know
most
people
I
talked
to
a
cube.
Connor
also
rapping
cops
to
do
more,
I,
guess
I'll,
say
saying
or
safe
rolling
updates
and
what
I
mean
by
that
is
specifically
not
reschedule
things
more
than
once
and
I.
Think
70
407
is
the
PR
that
does
kind
of
what
I
want
to
do
and
pretty
sure
somebody
for.
D
E
D
J
A
Would
just
I
think
both
pr's
definitely
addresses
sort
of
the
issues
before
that
we
had
where,
like
they
were
boiling
huge
oceans
in
terms
of
size,
so
now
they're
most
they're,
now
much
more
like
reviewable
I
think
John.
Thank
you
for
the
the
one
with
a
test
and
all
that
I
think
yes,
I
will
I
will
try
to
look
at
both
cuz
I
agree.
We
should
we
should
we
should
get
this
rolling
I.
Think
we
were
debating,
whether
like
block
on
Cluster
API
I'd
still
want
to
get
clusters.
I,
don't
but
I.
A
A
A
Mean
I
think
the
strategy
idea
is
actually
super
like
super
useful
as
well,
because
we
probably
also
want
the
idea
in
future,
not
necessarily
the
PR
of
like
strategies
for
like.
Do
we
back,
in
instance,
group
with
sort
of
cops,
auto
scaling
group
of
behavior,
or
do
we
want
to
back
it
with
a
machine
development
feature?
So
this
is
all
great
stuff
and
great
line
count
ratio
of
like
long
shadows,
line
count
so
I
think
this
is.
We
should
definitely
get
one
or
protect,
perhaps
both
of
these
in
depending
on
I.
Don't
know
that
my.
D
A
A
J
You
yeah
yeah
I,
mean
I,
think
the
key
thing
that
I'm
saying
is
just
you
know
right
now,
if
I
detach
all
the
instances,
let
the
ASG
need
to
detach
and
cordon,
but
they
is
she
bring
up
new
ones
or
taint
them
whatever
and
let
new
things
go
to
the
new
and
then
start
draining
one
at
a
time.
I
get
the
sanity
that
I
want
pretty
sure
that
those
PRS
support
in
some
combination.
A
A
A
I,
double
negative
myself!
I,
don't
see
you
here,
but
I
presume
you
are
watching
on
the
video
which
I
will
post
diligently.
There
are
two
PRS
which
I
will
take.
It
looks
like
you're
just
asking
about
CPR,
so
I
will
take
a
look
at
those
two
PRS
and
I
apologize
for
having
or
some
earlier.jon
new
features
in
v1
alpha
one.
If
you
want
our
friend
V
1
alpha
one
API.
Yes,.
D
A
Think
it's
so
I
think
it's
a
good
idea.
The
the
challenge
is
this
gets
into
like
deprecation
and
it
gets
into
the
API
machinery
rules
where,
if
we
have
to,
if
we
have
to
number
two
API
versions
that
are
both
active,
my
understanding
is,
they
have
two
round
trips
between
them,
the
so
in
other
words,
you
should
be
able
to
go
back
and
forth
between
the
two
versions
and
not
lose
data,
and
this
is
important
for
reasons.
A
A
So
that
feels
like
a
natural
time
to
do
this,
so
I
think
we
should
I
think
we
should
they.
Oh
you
were
there
to
an
half
years
ago.
We
should
we
should
deprecated
you
enough
of
one
but
I,
don't
know
that
we
can
actually
stop
requiring
I
think
all
my
tests
will
break
if
you
only
had
a
feel
to
view
one
else
too.
I
hope
all
my
tests
will
break.
If
you
tell
me
how
to
feel
that,
you
want
up
with
you
that.
A
A
Shall
we
let
should
we
pursue
this
on
the
on
the
deprecation
like
Peters
deprecation,
this
like
issue
and
try
to
figure
out
the
right
I
think
we
should
deprecated.
V
went
out
for
one
like
I'd
like
to
get
to
point
where
we
remove.
If
you
went
off
of
one
I
would
like
to
follow
the
rules
I.
As
you
point
out,
we
met
people
to
deprecate
it
now,
in
whatever
the
Alpha
is
whatever
it's
like
one
16s,
you.
C
A
Must
be
careful,
the
the
API
guarantees
our
have
in
the
past,
like
provided
stability
for
in
communities,
have
provided
stability
for
users
as
so
they're
old,
so
their
existing
yeah
moles
never
broke
or
broke
in
a
controlled
way.
So
there's
plenty
of
notice
and
so
that
it
is
super
tempting
and
we
have
seen
people
that
have
gone
into
communities
and
like
just
removes
all
the
old
versions
and
said
Wow
I
have
removed
all
the
old
versions
and
then
smaller
right
and
that's
great.
But
it's
not
really
the
kubernetes
way.
A
So
the
cumin
is
way
is
just
too
deprecated
and
like
follow
the
procedure.
I
agree
with
you,
it
is
annoying
and
I
do
think
we
should
or
it's
there
is
a
cost
to
it.
I
think
I
think
we
should
deprecated
and
follow
the
policy
and
I
think
we
should
use
the
open
versions
to
make
that
less
painful
on
ourselves,
but
I
also
like
to
not
break
the
rules
and
if
I
break
the
rules
I
get
in
trouble.
So
let's
not
break
the
rules
but
I
think
also.
A
A
A
Okay,
thank
you.
That
is
I
mean
it
is.
It
is
important
at
the
same
time,
I
appreciate
your
raising
the
issue.
It
is
important
to
like
make
development
efficient,
but
we
also
want
to
those
rules
are
part
of
what
makes
communities
nice
and
perhaps
a
dozen
tree.
What
we're
doing
here,
but
like
it
any
one
off
one
is
not
necessarily
like
the
the
bit
where
it
really
matters,
but
that's
in
general,
like
we
should
try
to
follow
the
rules.
A
B
It's
just
really
quickly.
Our
cherry
picks
to
116
have
a
and
end
test
that
I
added
recently
and
they're
failing,
which
means
we
can't
cherry
pick
anything
to
116
right
now.
It's
because
of
the
volume
permissions
bug
they,
you
Justin
fixed
in
the
kubernetes
repo,
but
it
only
made
it
into
117
and
not
116.
So
I
opened
the
cherry
pick
in
there
and
they
need
to
decide
whether
to
accept
it
or
not.
If
they
do,
it
should
pass
I
think
if
they
don't
accept
it,
we
should
just
disable
the
and
add
job
so.
A
One
okay!
Thank
you!
Okay!
Well,
anyway.
I.
Thank
you
for
that.
In
this,
that's
a
good
good
heads
up
and
yes,
I
agree.
If
we
can't
get
it
merging,
we
should
just
make
sure
we
have
whatever
test
coverage
we
can
have,
rather
than
no
test
coverage
at
all.
Oh
you're,
saying
to
just
you
would
disable
the
whole
key
to
each
up
or.
H
Yeah
well,
when
we
were
upgrading
our
clusters
to
1.14,
we
hit
this
issue.
It's
an
issue.
I
actually
I,
think
we
fixed
in
1.12
and
then
a
cherry
pick
in
114
to
fix
a
different
issue
reverted.
So,
if
you're,
using
terraform,
currently
in
1.14
and
using
mixed
justice
policies,
rolling
updates
force
on
the
roll
every
time,
it
always
thinks
that
there
is
a
version
change
so
that
cherry-pick
I
the
PR
has
been
merged
to
master.
To
fix
that.
H
A
Or
on
the
side
of
recommending
a
rolling
update
or
rec
of
doing
that,
but
I
agree.
It's
certainly
disruptive.
It
feels
certainly
like
it
should
go
into
115.
The
only
debate
is
whether
it
should
go
into,
or
the
debate
in
my
mind
is
whether
to
go
into
114
cops
114,
given
we
have
115
0
out,
but
it
is
a
relatively
recent
release,
so
I
wouldn't
be
opposed
to
putting
it
into
114.
Some
other
people
feel
I'm
doing
a
114.
3
can't
remember.
C
H
A
I'm,
not
I,
wouldn't
say
I
think
opposed
is
too
strong,
but
just
questioning
whether
it's
suitable,
yeah,
critical
criticality
to
go
back
to
like
when
we
have
a
different
latest
version.
This,
but
yeah
I,
think
it
that's
at
least
put
it
on
the
branch
well,
but
then
I
view
that
we're
sort
of
like
it's
a
bit
weird
to
never
release
it.
A
Why
don't
we
do
it?
People
are
probably
still
running
114
cops
114.
It
doesn't
feel
here's
an
idea,
let's
put
it
on
114,
but
also,
if
there's
anything
else,
you
want
to
get
in
there
like
the
terraform
zero
one,
two
warnings
we
could
like
consider
cherry-picking
that
at
the
same
time,
but
yes
I'm,
let's
put
it
on
114
it
feels
given,
are
in
the
status
of
our
rolling
update
not
being
the
fastest.
It
feels
sufficiently
annoying
that
we
should
yes
and
then
last
issue.
Last.
I
Topic,
Rodrigo
I
know:
we've
discussed
this
before,
but
importing
bin
data
that
go
I
know
we
currently
them
commit
it,
and
anyone
that
tries
to
import
cops
as
a
library
fails
miserably
because
of
that.
So
how
I
do
and
I
believe
some
of
the
folks.
This
call
do
it
as
they
generate
it
locally
after
they
clone
to
get
repo
I,
don't
know
if
there's
any
way
around
that
due
to
the
nature
of
been
data,
I'm,
not
sure
we
really
want
to
be
committing.
I
I
A
We
certainly
commit
other
generating
code
like
every
scenery
code.
Been
data
is
in
theory,
easier
to
recreate,
but
it's
certainly
like
any
any
pick
up.
There
is
pro
Matic
and
there
is
a
general
policy
that
go
code
should
be
go
buildable
or
gettable
and
we're
not
go
gettable.
So
that's
not
great
I,
don't
know
of
a
way
around
this
like
it
would
be
great,
so
yeah,
no
problem.
We
don't
even
need
to
generate
this
at
all,
but
I
don't
know
if
such
a
way
I
would
therefore
be
in
favor
of
I
believe
something.
I
A
We
also
I
I
think
this
is
a
good
thing
to
do.
We
also
I
think
have
an
outstanding
issue
which
we
have
sort
of
shied
away
from
today,
because
we've
been
focusing
on
catching
up,
which
was
to
update
our
version
of
Gobind
data,
something
more
more
recent
version,
I
think
it's
also
changed
conical
locations.
Now
we
could
also
reevaluate
whether
there
is
a
as
I
record,
also
like
optimizes
I
can
compresses
it
or
something
so
it
like
optimizes,
like
that,
the
output
is
a
little
bit
weird
and
it
or
that
the
go.
A
C
A
B
A
That
way,
yeah
that'd
be
great.
Thank
you
and
I.
Might
we
can
also
just
tell
Ben
data
not
to
compress
or
something
if
that
is
in
fact
the
issue?
Okay,
great!
Thank
you.
We,
unless
someone
added
something
in
the
last
minute
we
reached
the
end
of
our
agenda.
We
did
some
releases.
We
have
I
put
some
holding
like
some
standard
like
we
should
look
at
Kate's
versions
and
do
some
am
eyes,
perhaps
as
the
release
plan
for
the
next
two
weeks.
A
I
think
probably
the
real
goal
should
be
to
I
guess
enjoy
Christmas
time,
but
also
to
get
some
PRS.
The
PR
backlog.
Now
that
we
have
put
up
a
little
bit
releases
cover
criminalize
releases,
I,
think
getting
through
the
PR
backlog
would
be.
The
next
focus,
I
would
guess
and
cherry-picks
sorry
pics.
We
can
actually
have
the
117
out
released
come
out
in
a
timely
fashion,
but
otherwise
we
are
at
the
top
or
almost
the
top
of
the
hour.
I.