►
From YouTube: Kubernetes kops office hours 20200911
Description
Recording of the kops office hours meeting held on 20200911
A
Hello,
everybody-
this
is
cops
office
hours
today
is
friday
september,
the
11th,
I'm
your
moderator,
facilitator,
john
myers.
I
work
for
proofpoint
a
reminder.
This
meeting
is
being
recorded
and
will
be
put
on
the
internet.
Please
be
mindful
of
the
code
of
conduct,
which
is
basically
to
be
a
good
person.
I
put
a
link
to
the
agenda
in
the
zoom
chat,
which
actually
I
have
not
done
yet.
A
Somebody
else
has
and
please
feel
free
to
put
your
name
on
the
attendees
list
and
add
any
items
you'd
like
to
discuss
to
the
agenda
in
the
appropriate
place
and
with
that
we'll
get
started
with
action
items
from
the
last
time,
and
I
believe
we
did
a
bunch
of
releases
cyprian's
cypriot
in
here.
Yes,
what's
going
on
with
with
alpha
4.
B
C
C
Yes,
I
did
that
cool
yeah.
I
think
it
should
be
let's
yeah
as
long
as
you
test
it
and
it
works
then
yeah.
I
think
that
sounds
great
well,
I
can
test
it
only
after
I
release
it.
B
So
that's
the
catch
yeah,
that's
a
bummer,
but
I
tried
it
with
alpha
three
binaries
and
works.
Okay,
I
set
the
mirror
to
that.
Yeah.
C
And
we
can
always
if
it
doesn't
work,
just
ping
me
and
I
can
do
whatever
the
extra
stuff
is.
The
extra
publishing
steps.
C
Want
I
think
we
have
to
I
we
want
to.
I
think
we
want
to
get
to
artifacts.kate
studio
and
artifacts
up.
Kate
sadeo
will
eventually,
hopefully
have
a
a
mirror
on
s3,
but
we
would,
I
think
my
opinion
is.
We
would
like
artifacts
case
studio
to
be
the
canonical
location
and
it
is
not
yet
the
canonical
location
right
now.
The
canonical
location
is
still
that
s3
bucket,
which
is
why
we
have
to
pull.
We
have
to
publish
them
there.
C
Oh
yeah,
I
don't
know
if
you're
gonna
be
able
to
do
this
because
we'll
see
I
don't.
I
thought
that
some
of
the
chars
had
to
be
pulled
from
the
s3
bucket.
So
I
don't.
B
C
Yeah,
I
think
that's
a
good
point
like
I
think,
there's
no,
it's
no
worse
than
where
we
are
today
in
terms
of
difficulty
publishing
and
there
is
a
much
clearer
path
to
making
a
community
published
as
it
were,.
A
B
I
added
it
because
I
don't
think
peter
is
around
so
he
did
it.
He
wrote
the
issue.
Let
me
paste
the
link
in
chat.
Sorry,
I'm
a
bit
late
to
the
call.
B
So
this
is
the
link
to
the
issue.
I
think
there
were
some
opinions,
but
I
don't
think
there
were
that
many.
B
Well,
so
I
don't
really
know,
I
think
I
missed
the
discussion
last
time,
but
I
don't
know
exactly:
do
we
want
to
do
another
one?
Do
we
want
to
move
this
one
or
is
there
a
purpose
I
mean,
can
it
be
done
or
we're
just
discussing.
C
I
think
we
wanted
to
try
to
find
a
more
inclusive
time
for
people.
I
think
we,
it's
not.
I
think,
there's
some
concerns
over
the
day
like
friday
and
friday
evening
for
europe.
I
think
we
also
don't
want
to
like
it'll,
be
difficult
to
find
a
time,
that's
universally
better,
and
we
don't
necessarily
want
to
sort
of
divide
our
our
things
such
that
people
feel
like
decisions
are
made
in
both
and
like
the
decision-making,
is
less
clear.
C
We
could
try
having
an
additional
meeting
where
we
avoid,
like
any
contentious
decision,
making.
B
A
Do
we
is
a
different
day
available?
I
don't
know
what
other
meetings
we
run
up
against.
C
I
think
we're
always
going
to
clash
with
well
it's
not
if
it's,
if
it's
in
sort
of
off
our
off
california
hours
it,
it
might
not
clash
with
anything
but
we're
likely
to
clash
with
something.
And
the
question
is:
what
are
we
going
to
clash
with
yeah?
We
can
yeah.
A
I'm
sure
that
the
the
nine
pacific
is
prime
meeting
hours,
but.
B
And
I
remember
that
you
couldn't
do
it
earlier
or
you
could
could
do
it
much
earlier.
I
think.
A
Yeah
yeah
one
or
two
hours
earlier
is
difficult
and
before
then
it's
just
getting
up,
but
you
know
household
and
distance
learning
kind
of
knocks
out
the
eight
o'clock
slot.
For
me.
B
A
So
so,
okay,
I
think
we
don't
have
any
resolution
to
do
today.
So
I
guess
we'll
move
on
to
ftd
manager,
support
for
ftd,
three,
four,
thirteen.
C
Yeah,
it's
I
can
give
you're
still
sending
the
pr
to
the
right
place
as
part
of
the
idea
of
sending
of
getting
the
the
images
for
getting
entity
manager
into
the
six
kids
of
io
area
and
publishing
images.
In
that
way,
we
will
at
some
stage
move
over
right.
Now
we
are
mirroring
or
manually
mirroring
into
that
other
repo
and
trying
to
get
everything
else
set
up
so
continue
to
send
prs
here.
If
you
really
want
to
yeah.
C
I
know
please
send
pr's
here
for
now,
because
I
don't
want
to
have
to
like
please
and
then
so.
This
is
good
is
43413,
the
version
that
is
in
120
or
it's
in
1.
C
B
Nine
or
something
and
was
bumped
to
13
a
week
ago,
so
I
think
this
will
probably
be
the
last
dead
cd
change
for
119.
C
B
Okay
and
probably
we
should
get
a
release
for
that,
because
we
need
to
test
in
cops,
I
mean,
even
if
it
won't
be
the
final
release,
at
least
something
to
get
soaked
in
in
here.
A
C
B
Okay:
okay,
john.
A
A
Yeah,
so
I
wanted
to
give
a
rundown
of
my
project
degree
factor
a
bunch
of
stuff,
probably
primarily
around
the
problem
of
you
know
you
change
something
and
it
takes
effect
before
you
do
the
update
cluster,
yes
or
the
terraform
reply,
and
there
are
sort
of
three
categories.
A
One
is
managed
files
if
you're
using
terraform
or
cloud
formation
looks
like
they
get
updated.
A
When
you
do
the
update
cluster
before
you
do
the
apply,
and
there
is
a
new
resource,
type
and
terraform
for
a
file
in
an
aws
bucket
or
in
a
gce
bucket,
and
so
I'm
trying
to
refactor
stuff
to
use
that
instead
for
terraform
running
into
an
import
cycle,
which
I
have
to
figure
out,
what
I'm
trying
to
do
is
get
the
code
for
generating
that
terraform
in
vfs,
which
means
I
need
something
that
can
generate
terraform,
which
does
not
depend
on
a
model.
A
You
know
the
rest
of
the
the
target,
otherwise
everything
gets
in
the
cycle.
I
don't
think
there's
a
solution
for
cloud
formation
because
they
don't
have
a
resource
for
a
file
and
an
sv
bucket.
A
A
I
refactor
a
lot
of
stuff
in
node
up
to
put
stuff
into
the
model
context,
but
there
are
two
things
that
are
all
potentially
large
and
might
not
fit
in
the
user
data,
and
so,
if
I
can
get
the
that
managed
files
well
in
terraform
properly,
then
I
might
create
a
manage
file
per
launch
template
version
so
that
the
user
data
can
pull
it
from
the
appropriate
managed
file.
A
D
A
Other
is
the
pr
to
for
cops
controller
to
take
the
node
labels
from
the
cloud
tags.
Instead
of
from
reading
from
the
instance
group
spec,
and
I
was
wondering
yeah.
C
On
the
creation
of
a
file
per
launch
template
version,
how
is
how
does
that
compare
to
like
versioning
the
instance
groups
and
keeping
a
history
of
the
instance
groups,
but
they
like
the
raw
the
input
yaml,
rather
than
the
transformed,
yaml
or
transformed.
A
Data,
well,
it
seems
you
know
whatever
we
do,
it's
going
to
be
per
launch
template
version,
because
every
time
you
know
basically
every
time
that
changes
right.
You
need
a
new
one
of
those
and
then
you
need
to
do
some
sort
of
garbage
collection.
So
when
you
remove
the
launch
template
version,
you
need
to
remove
that
file.
C
B
B
B
It
doesn't
update
the
version
in
itself,
so
it
always
wants
a
new
name
at
the
moment.
That's
one
problem
and
then
how
to
actually
well
at
when
you
render
it,
you
should
always
use
the
latest
version.
Whatever
everything
was
before
I
mean
whatever
someone
chooses
there,
so
I
guess
anyway,
so
before
we
do
more
workarounds
around
launch
templates,
probably
we
should
try
to
use
them
correctly.
A
Yeah
they're
not
not
sure
what
how
difficult
that
refactor
would
be,
but.
B
B
A
Especially
if
we've
got
other
associated
files
with
them.
B
A
So
anyway,
to
finish
up
so
yeah,
so
I'm
I'm
wondering
suggestion.
You
had
some
concerns
about
the
what
was.
C
It
wasn't
a
concern
it
was.
It
was
along
the
same
lines
where
I
was
going
there.
It
was,
I
feel
like
there
are,
if
the
if
the
motivation
is
like
there
are
other
things
that
come
from
the
instance
group
and
so
sure
we
can
get
the
node
labels
here.
But
are
you
sort
of
kicking
the
can
down
the
road
and
we're
gonna
have
to
deal
with
the
instance
group
version,
and
I
mean
it
sounds
like
you
are
dealing
with
it
and
that's
yeah,
I'm
trying
to
fix
everything,
but
yes,
and
what
I'm?
C
What
I'm
wondering
is
like
the
the
the
more
obvious
to
me
way
is
to
keep
the
history
of
the
instance
group
groups
themselves
right
and
that's
sort
of
what
we
built
in
cluster
api
with
like
where
a
machine
set
is
a
version
of
the
instance
group
and
a
machine
deployment
is
an
instance
group.
C
A
C
The
huge
advantage
of
that
is,
it
might
create
more
or
it
should
create
more
accurate,
rolling,
update
logic
right,
like
we
can
better
detect
whether
a
change
actually
like
yeah
caused
a
real
change
or
was
some
random
thing
that
was
like
in
a
different
place.
That
would
be
good.
We're
gonna
start
bumping
up
against
the
16k
limit
right
of
metadata
yeah,
which.
C
Yeah,
I
mean-
I
guess
I
guess
that
makes
so
from
that
perspective.
That
makes
sense
so,
like
I
think
what
you're
saying
to
paraphrase
is,
you
are
doing
instance,
group
versions,
but
at
the
same
time
you
are
also
saying,
like
it's
incorrect
to
actually
do
instance,
groups.
We
should
be
doing
versions
of
the
node-up
model
context
and
that's
a
better
thing
to
do,
and
we
might
well
hit
a
16k
limit
and
so
we'll
put
those
into
s3
and
we
will
delete
them
along
with
the
launch
templates
with.
C
C
C
That's
sort
of
where
I'm
stuck
right
now
I'll
have
a
look
at
that
that
input
loop
as
well
with,
along
with
before
I
close
this
machine.
The
that
sounds
really
good
to
me.
I
I
very
much
like
the
that
it
sounds
like
we
will
end
up
with
a
very
clear,
explicit
specification
of
exactly
what
is
going
to
run
on
a
particular
machine
with
minimal,
like
node-up
logic,
which
is
always
like.
That
makes
me
happier
the
less
known
uplogic.
C
We
need
the
the
better
because
it's
sort
of
opaque
paint
attached
they're
made
yes
and
so,
like.
I
think,
that's
I
think,
that's
good
and
yes,
I
we
will
hit
the
16k
limit
and
we
will
have
to
put
it
in
s3
or
vfs
and
yeah.
If
we
really
do
that,
that
sounds
that
to
me
sounds
good.
I
don't
know
if
anyone
else
has
an
opinion.
A
That's
what
I
have
on
that,
so
I
guess
we'll
move
on
to
docker
binary
install
for
noda.
E
Great
hi,
michael
andrews-
this
is
talking
about
pr
7974
and
really
this
pr
was
a
replacement
for
the
pr
I'm
about
to
send
in
in
this
chat
here.
Six
nine
five
six
you
know
downloading
docker
is
is
is
not
allowed
for
our
company,
so
we
have,
oddly
enough,
we
can
download
it
through
our
artifactory
mirror,
but
we
can't
hit
it
directly
docker
hub.
E
So
we
have
tried
to
to
do
a
pr
here
for
specifying
where
to
go,
get
docker,
and
that
was
the
first
pr
6956,
where
we
overrode
the
docker
slice
in
docker
go
at
line
89
and
we
provided
a
way
to
specify
the
slice
you
want
in
your
in
your
cluster
spec
and
talking
about
the
16k
limit
on
user
data.
E
Once
we
moved
to
19
and
we
had
to
put
all
those
dependencies
in
there
right,
we
we
hit
that
pretty
quick,
so
it
it
also
doesn't
allow
us
to
do
heterogeneous
clusters.
Obviously
so
we
opted
for
pr7974
here,
install
docker
from
a
static
binary
in
which
we
could
host
an
artifactory,
and
that
seems
to
work
well
for
us,
although
it's
still
not
probably
ideal
right
so
in
earlier
office
hours,
austin
moore,
who
had
done
these
pr's
unfortunately
has
broken.
My
heart
and
left
left
the
group.
E
So
he's
not
here
today,
but
there
was
talk
in
earlier
office
hours
about
just
rethinking
this
whole
way
of
downloading
docker
and
installing
docker
on
the
nodes
and
all
that
so
we're
kind
of
here
just
to
ask.
If
is
that
a
priority
to
refactor
how
we
do
docker
install
and
if
it's
not,
can
we
get
this
pr
merged
or
looked
at
installed
from
a
static
binary.
B
Can
I
speaker
so
right
now
all
the
installs
are
done
from
static
binaries.
I
just
didn't
expose
the
option
of
installing
it
from
you
know.
There
is
no
option
in
the
cluster
spec.
B
I
am
working
on
something
to
move
it.
Let's
say
in
the
applied
section
so
that
when
you
run
apply
cluster
together
with
all
the
other
artifacts,
you
will
also
see
the
docker
binary,
so
it
will
make
it
docker
archive,
so
it
will
make
it
easier
to
to
install
a
custom
one
as
long
as
it
looks
like
the
one
from
docker
hub
docker,
repo.
B
It's
not
hard,
so
I
already
have
something
working.
I
didn't
push
it
as
work
in
progress,
but
I
don't
see
why
it
won't
be
part
of
119
as
long
as
you're,
okay
with
running
119.
B
E
That's
that's
really
great
to
hear
we've
been
cherry-picking
our
pr
on
top
of
on
top
of
your
releases
and
hosting
cops
internally
to
to
get
around
this
issue
it'd
be
great
to
use
the
open
source
project
instead
of
having
to
host
our
own
cops.
So
just
so,
I
can
understand
this
when
you
say
everything's
installed
now,
with
the
static
binary.
E
Previously
we
saw
we
saw
a
big
slice,
a
docker
slice
in
docker
versions
in
the
docker
go
at
line
89,
for
maybe
I'm
just
wrong
on
that.
But
hey
that's
changed.
Let
me
let
me
show
you
what
I'm
talking
about,
and
maybe
I'm
just
not
this
guy
here.
E
I
just
pasted
that
into
the
the
chat,
so
so
yeah.
So
is
this
what
you're
you're
talking
about
it's
been
kind
of
redone
here.
B
E
So
this
is
from
this
is
from
the
pr
obviously,
so
that
makes
sense.
Okay!
Well,
that's
really
good
news
and
yeah.
We.
We
can
definitely
wait
to
119
on
this
and
just
to
be
clear,
there
will
be
a
way
for
us
to
specify
where
to
go,
get
the
tg.
C
Yeah
we
should,
I
think
it
should
go
through
asset
remapping,
ideally
in
the
long
term,
but
there's
also
a
valuable
use
case
in
being
able
to
specify
a
tgz
in
that
suppose,
like
it
looks
like
the
latest
version
we
have
is
1903
11.
docker,
supposedly,
let's
suppose
they
released
like
1903
12.
like
it
would
be
nice
to
be
able
to
just
like
drop
that
in
there
and
sort
of
override
easily.
So
I
think
that's
an
important
reason
to
allow
arbitrary.
B
It's
not
a
it's,
not
a
problem.
My
only
question
so
right
now
there
is
no
versioning
there.
So
if
docker
tomorrow
decides
to
change
the
way
the
structure
is
inside
the
archive,
we
will
have
to
modify
and
add
versions,
but
as
it
is
now,
it's
a
pretty
simple
model
and
we
just
have
to
copy
everything
in
one
directory.
B
I
think
so
we're
good
to
go
just
that
you
will
have
to
wait
another
one
month
or
two
something
until
it's
released.
Maybe
you
can
drop
me
a
message
on
slack
so
that
I
can
ping
you
and
you
can
maybe
test
it
when
it's
ready.
I
can
also.
E
Great
really
good
news
guys.
Thank
you
very
much.
This
is
kind
of
the
last
piece
for
us.
You
know,
so
so
it's
good
to
hear
that
that
it's
getting
addressed.
Thank
you.
Hey.
B
Yeah
I
tried
various
things
but
depends
on
credentials
and
I
don't
have
credentials
to
see
if,
let's
say
manifest
tool
works
with
whatever
we
put
well.
The
cloud
build
environment
provides
to
generate
the
manifest
for
sure
it
works
with
docker,
because
I
think
most
projects
do
that.
B
I
saw
it
in
nginx,
so
my
question
is:
can
I
do
some
prs
to
see
if
it
works?
You
know
and
revert
them,
or
is
there
a
better
way.
C
You
should
you
should
put
up
a
work
in
progress
pr.
I
may
be
able
to
get
an
answer
for
you
on
whether
or
not
it
will
work.
If
we
have
high
confidence,
we
can
merge
it.
The
other
way
would
be
to
create
a
second
build
and
like
do
that.
C
I
think
if
you
have
access
like
there
shouldn't
be
anything
special
about
cloud
build
about
our
cloud
build.
Ideally,
so
if
you
have
access
to
gcp-
and
I
don't
know
if
you
have
like
you
know-
you
can
just
run
a
cloud
build
on
your
own
account.
I
can
also
do
that.
Obviously,
on
my
own
account
just
like
as
a
user
of
gcp,
and
so
we
can.
We
can
get
confidence
that
way.
C
If
you
put
up
a
work
in
progress
pr,
I
think,
and
then
we
can
make
the
call,
I
think,
we're
only
technically
using
it
for,
like
the
latest
releases,
so
like
the
119
alpha
releases.
So
it's
not
the
end
of
the
world.
If
we
temporarily
break
it,
I
don't
think,
but
I'd.
B
B
So
I'm
talking
about
arm
in
particular,
if
we
want
multi-arch,
we
can
do
it
even
later,
because
we
don't
really
care
about
it
being
multi-arch
right
now
right
we
care
only
that
it's
built
for
two
architectures.
We
can
do
the
same
thing.
We
do
for
the
kubernetes
images
and
just
replace
md
with
arm
when
we
deploy.
C
B
I
did
an
update
to
that
also
so
a
bit
further,
but
there
is
no,
as
I
didn't
find,
a
way
to
do
the
manifest
with
basil.
There's
still
the
question
of
how
to
do
that
part.
So
with
docker,
it's
simple,
but
it
depends.
C
B
C
Okay,
so
if
we
let's
at
least
get
regardless,
that
feels
like
the
first
step
to
get
the
the
separately
named
images
right
and
we
can
get
that
going
and
yeah.
So
that
was
great,
okay,
so
I'll.
Just
I'll.
Look
at
that
pr
and
make
sure
we
can
merge.
It
is
that
right.
C
Yeah,
if
I
recall
correctly
like
there
is
no
two
to
one
build
of
yeah,
it's
it's
just
not
gonna
happen
so
like
there
will
be
no
two
ncd2
support
on
arm64,
which
I
don't
think
will
upset
anyone.
B
C
A
C
Alpha-4
yeah
I
mean
we
did
117
2
and
118.1.
C
I
guess
scd
manager
like
getting
that
built
into
kcgcrao
would
be
good.
Getting
the
scd
manager
arm
builds
going
would
be
good.
B
Okay,
olly
marcus,
sent
the
pr
you
merged
it
a
few
weeks
ago
or
something
okay,
but
that
we
haven't
questioned
it.
B
A
Okay-
and
we
talked
last
time
about
deciding
what
we
want
to
do
with
the
release
119
release
branch
this
week-
I
kind
of
get
the
idea-
we're
probably
not
ready,
for
it
is
anyone.
What
do
people
think.
C
B
B
C
And
I
don't
see
any
massive
things
I
mean,
there's
the
we're
using
future
flags.
I
think
to
be
able
to
put
things
in
that
aren't
ready
to
necessarily
go
in.
I
don't
think,
there's
anything.
That's
too
problematic
from
that
point
of
view,
if,
like
the
launch
template
launch
template
version,
thing
came
in
maybe
we'd
want
to
branch
at
that
point,
but.
B
A
B
C
The
oh
yeah,
the,
but
anyway
it
sounds
like
we're
not
ready
to
cut.
One
thing
I
actually
want
to
think
about
is
not
one
one
flip
side
of
not
exporting
the
cube
config
at
all
is
that
now,
like
the
flow,
you
have
to
specify
the
cluster
name
a
lot.
So
I
don't
know
if
we
can
somehow
address
that.
I
am
playing
with
my
I'm
trying
to
break
playing
with
and
trying
to
break
the
the
thing
I
did
where
you
have
the
plug-in
model,
so
maybe
we
can
make
the
plug-in.
C
I
don't
know,
I
don't
necessarily
make
a
plug
in
the
default,
but
I'm
trying
to
think
of
a
way
where
we
can
export
like
have
the
coupe
config
exported,
so
that
we
have
some
notion
of
a
current
cluster,
which
is
otherwise
makes
the
workflows
a
little
bit
painful.
C
A
C
Right
but
yeah,
but
yes,
and
then
the
problem
is
to
not
export
the
credential.
I
have
to
activate
a
mode
which
we
were
like
well.
This
is
a
this
is
an
experimental
like
thing
for
the
future.
Don't
worry
about
it,
and
now
it's
like,
oh,
no,
it's
not
so
experimental
so
that
wasn't
the
intent.
Okay,
so.
A
C
I'm
in
favor
of
doing
something
I
need
to
learn
more
about
what
oktoberfest
is
like.
Am
I
allowed
to
contribute
to
it
as
a
googler,
but
you
know
like
I
am
I'm
in
favor
of
doing
something
personally,
if
other
people
are
as
well
that
that
would
be
great,
but
we
can,
I
think
we
can
yeah.
The
question
is
how
we
can
continue
the
conversation.
C
Yeah,
I
don't
know
if
we,
those
the
if
it's
people
coming
in
and
fixing
small
things,
I
think
that's
going
to
be
harder
for
people
to
contribute
to
right.
It's
it's
a.
I
hope.
If
we
have
a
small
issue
in
the
tracker
that
it's
been
suitable
for
tagging,
that
it's
been,
we
might
have
already
been
fixed.
C
A
few
tagged
that
way,
I
I
think
the
worry
is
always
that
they
are
tagged
that
way,
but
they
are
actually
like.
Oh
actually,
when
you
try
to
fix
it,
you're
going
to
find
this
and
then
this
and
then
this
sort
of
thing
I
feel
like
we
could
try
to
identify
some
appropriately
scoped
projects
that
could
be
done
in
a
hacktoberfest
time
frame
of
like
realistically
a
week
of
of
the
evening.
So
like
four
or
five
evenings.
C
So
shall
I
propose,
let's
open
an
issue,
a
separate
issue,
to
discuss.
Let's
look
at
the
issues
which
we
in
that
issue,
which
might
be
good
issues
and
do
that
and
decide
whether
we
have
anything
that
actually
is
a
good
opportunity
for
people.
G
One
of
those
people
that
wants
to
get
involved
great,
I'm
david
muckle
from
sonos
awesome
and
what
I
work
with
mike.
Oh.
C
G
That's
a
good
question
because
my
perspective
as
someone
who's
new
to
this
project
in
terms
of
contributing
I'm,
I
would
like
to
do
whatever
would
best
get
me
exposed
to
as
much
of
the
project
as
possible.
So
to
that
end
I
think
that
would
be
a
lot
of
smaller
issues.
That
being
said,
I
also
see
merit
in
taking
like
a
whole
month
to
do
one
really
big
thing
right,
maybe
not
a
whole
month,
but
like
two
weeks.
C
Yes,
exactly
like
it's
described
as
a
month,
but
yeah
okay,
that's
really
helpful
tonight,
like
yeah
the
a
good
first
issue
where
you
go
and
like
change
percent
v
to
percent
percent.
W
is
going
to
in
theory,
touch
all
the
lines,
but
not
actually
like
give
you
any
real
exposure
to
what's
going
on,
for
example.
So,
like
that's,
that's
it's
good
to
understand
your
sort
of
motivation
and
yeah.
That's
helpful.
Thank
you.
C
A
F
F
I'll
drop
in
an
issue
into
the
meeting
notes
as
well.
Oh
thank
you.
B
B
G
B
Had
a
question
about
azure
support,
it's
one
of
our
top
requests,
at
least
from
the
past,
has
about
100
votes,
but
we
don't
have
well,
not
that
we
don't
have
anyone,
but
I
don't
know,
do
we
have
some
account
with
asia
or
something
like
any
idea.
If
we,
if
we
want
to
start
doing
ourselves,
you
know
trying
to
start
the
support,
would
it
be
possible
or
it's
just
something
that
someone
has
to
be
involved
with
asia
or
you
know,
use
it
to
actually
make
sense.
C
C
I
I
think
this
is
where
it's
why
it's
stayed
open
for
so
long
is
or
like
hasn't
been
started,
is
because
there
is
a
chicken
and
egg
thing
here
where
I
don't
think
anyone
that
that
knows,
azure
wants
to
has
felt
motivated
to
do
it.
I
think
there
is
cluster
api
for
azure,
so
maybe
that's
a
way.
We
all
get
there
one
day,
but.
C
You
know
we
don't
have,
we
don't
have
a
ton
of
demand
for
it.
So,
oh
sorry,
we
don't
have
a
ton
of
people.
We
don't
have
a
ton
of
contributors
that
want
it.
As
far
as
I
know,
I
could
be
wrong
but
like
they
actually
wanted
themselves,
and
so.
C
I
didn't
know
whether
someone
was
I'm
okay
with
that
yeah,
so
I
mean
if
we
have
a
contributor
that
wants
wants
to
to
do
that.
I
think
that's
great
if
someone
wants
to
learn
and
do
it
as
a
project,
someone
that
knows
azer
wants
to
like
say
if
someone
knows
azure
once
once
added
to
cops,
that's
wonderful
and
we
can
like
work
with
them
on
that,
oh
david,
do
you
know
azer.
C
Oh
okay,
I
was
like
be
careful
things
not
to
do
at
an
auction
or
yeah
right,
the
so
yes,
but
until
we
get
that
I
feel
like
you
know
we
we,
I
would
be
very
wary
of
anyone
that,
like
I
don't
know
as
her
and
I'd
be
very
wary
of,
like
I
shouldn't
be
doing
that
coding.
B
C
Mean
the
opposite:
people
did
great
work
and
I
think,
but
I
think
they
were
motivated
to
get
it
to
work,
and
then
we
we
we
didn't,
have
to
do
a
lot
of
help,
but
we
like
we
helped
them
do
that.
And
if
someone
we
need
someone
that
wants
azure
support
well
enough
to
implement
it
and
sort
of
own
it.
And
then
we
can
help
them
with
the
sort
of
cop
side
of
things.
B
B
A
Well,
I
suppose
thank
you
for
all
coming.