►
From YouTube: Kubernetes SIG Node 20220503
Description
Meeting Agenda:
https://docs.google.com/document/d/1j3vrG6BgE0hUDs2e-1ZUegKN4W4Adb1B6oJ6j-4kyPU
B
Me
so
good
morning,
everyone
it's
the
first
signal,
the
weeknd
meeting
in
the
may.
C
B
Today
is
the
may
3rd,
so
we
are
going
to
first
continue
to
finish.
Our
1.25
planning
now
riven
and
manu
is
handed
off
to
you
too.
D
Yeah,
let's
continue
the
kept
planning.
I
think
this
is.
We
were
last
time.
So,
let's
yeah
just
continue
here.
Add
note.
Logo
plug-in
support
for
automated
handler.
D
Yeah
surround
right
provide
pronounced
correctly,
so
I
don't
see
an
owner
listed
there.
E
A
Yeah,
I
don't
know
if
he
has
followed
up
on
that,
but
I
would
guess
we
should
poke
him
again.
Okay,
so.
E
So
this
one
doesn't
have
an
owner
right
now
and
we
may
not
be
able
to
get
to
it
unless
we
have
books
signed
up
yeah.
B
D
Let's
move
on
pot
level
resource
limits.
E
So
this
one,
I
think,
like
david
porter,
you
were
interested
in
this
one.
Do
you
think
we
have
enough
time
to
tackle
this,
or
we
should.
F
B
I
think
it
may
be
more
makes
sense
for
us
to
understand
the
powder
part
level.
Overhead
first
arrives
so
when
the
limit
power
level
of
the
resource
limit
is
more
makes
sense
for
the
user,
so
the
power
overhead
is
not
complete
down
and
and
also
not
many
people
start
using
yet
so
we
haven't
have
the
much
of
the
feedback
for
that
one.
So
I
I
think
it's
reasonable.
We.
B
D
All
right
so
david,
I'm
just
gonna,
put
your
name
here
too
kind
of
gather,
more
information.
G
I
can
get
in
touch
with
kevin.
I
didn't
get
the
chance
to
to
do
that
until
now.
Sorry
about
that,
I
see
I
need
to
talk
with
him
to
see
if
he
is,
he
wants
to
do
it.
Last
time
we
spoke
it.
We
were
not
sure
about
the
timelines,
but
I
will
get
in
touch
with
him.
H
G
D
Remember
this
is
for
beta
okay.
The
target
is
better
graduation
in
125.
A
So
we
have,
we
have,
like
a
general
feature,
gate
right
now.
I
thought
for
all
these
policy
knobs
on
this,
so
I
don't
know
if
we
don't
want
a
way
to
group
them
but
like
when
I
was
looking
at
this
feature,
I
didn't
think
we
actually
needed
a
net
new
feature
gate
per
se.
It
was
like
it
would
just
be
in
that.
A
I
This
one
was
alpha
in
terms
of
the
policy
option
itself,
so
we'd
have
to
graduate
this
particular
or
essentially
move
this
option
so
that
it's
enabled
by
default.
G
D
B
And
also
derek
is
derek
also
last
last
time
we
also
decomposed
and
we
break
this
into
several
pieces,
so
try
to
accelerate
progress
and
let's
follow
up
with
the
camera
and
see
what's
going
on
now
we
can
iterate
this.
J
E
D
Okay,
since
there's
no
owner
listed,
is
patrick
here
on
the
call
or
whoever
has
contacts
on
this
one.
H
I
don't
see
patrick
on
recall,
but
I
know
what
he
is
working
on.
The
prototype
on
all
of
this
and
kevin
also
is
also
working
on
a
way
how
to
integrate
it
into
device
manager
and
how
to
do
a
prototype
of
protocols
like
how
to
plugin
can
do
immigration,
art
and
few
other
people
working
on
this
thing
as
well.
B
So
this
one
is
kim
talked
to
us
a
couple
times
and
I
just
want
to
say
that
this
one
will
continue
move
forward
but
looks
like
we
don't
have
the
target
delivered
in
1.25
so
far.
That
means,
though,.
D
Okay,
so
this
sounds
like
well,
we
are
actually
actively
working
on
this.
Can
I
just
upload
an
owner
here
if
we're
planning
to
do
this
in
125.
H
Owner
is
patrick
or
if
we
need
some,
somebody
from
sick
leads
to
to
be
owner
or
approver.
For
that,
when
we
need
to
find
someone,
I
don't
know
who
will
be.
B
Okay,
I
I
will
be
the
approver,
but
I
wouldn't
need
to
find
something
viewer
for
this
one.
I
I
don't
think
I
have
the
good
time
to
review
this
yeah.
We
need
some
reviewer
to
help
earlier
with
you.
D
K
Hey
so
this
is
deep.
We
had
a
couple
of
rounds
of
review
during
1.24
and,
like
the
concerns
have
been
addressed
in
the
cap
so
far,
just
waiting
for,
I
guess
another
on
the
reviews.
It's
a
fairly
small
feature.
E
I
I
think,
like
one
outstanding
question
that
derek
had
added
was
like:
do
we
need
this?
If
we
add
tracing,
I'm
not
sure
whether
that
that's
answered.
A
A
K
Yeah
we
we
identified
some
use
cases
where,
like
you
know,
the
controller
use
cases
where
tracing
may
not
quite
help,
that's
kind
of
like
the
main
update,
and
then
we
simplify
the
conditions
a
whole
lot
too.
So
it's
just
like
a
single
condition
that
just
comes
with
a
sandbox
ready
state.
A
Yeah,
so
that's
fair,
so
I
think
for
myself.
I
just
want
wanted
to
to
see
if
deep
in
the
other
folks
who
gave
review
on
this
were
like
aligned.
But
my
initial
feeling
was
the
argument
for
was
maybe
too
narrow.
A
But
sounds
like
there's
additional
feedback
that
maybe
we
can
all
take
a
look
at
and
see
if
we
all
agree
would
be
useful.
L
About
this
about
this
cap,
I
had
one
comment
on
the
on
the
order
of
this
new
sandbox
creation
status
and
the
current
initialized
status.
I
was
saying,
maybe
there's
a
chance.
We
can
make
it
right
that
the
initialize
the
status
is
what
will
be
reported
after
this,
a
new
sunburst
creation
yeah.
I
put
that
comment
in
the
cap.
L
I
I
was
asking
elena
to
take
a
look,
but
she
is
out
now.
Maybe
someone
else
can
help
see
if
it
makes
sense.
A
Is
in
total
transparency
is,
I
kind
of
view
the
sandbox
as
an
implementation
detail
of
the
cubelet,
and
it
kind
of
upsets
me
that
we
have
to
expose
the
concept
into
the
end
user
api.
So
at
like
some
almost
like
spiritual
level,
I
feel
like
we
messed
up.
A
If
we
have
to
do
that,
I
know
we
already
exposed
the
concept
of
sandbox
elsewhere
and
like
error,
messages
or
condition
reasons,
but
that's
personally
one
of
the
things
I
struggle
with
on
this
one
is
how
much
of
the
end
user
api
should
have
to
know
about
the
implementation
detail.
B
Yeah,
I
I
agree
with
you
derek
I
also
worry
about.
This
is
kind
of
the
open
of
the
sky
of
the
wolf,
like
the
people,
trade,
the
standard
box
as
instant
as
the
the
the,
because
it's
implement
detail
as
part
of
a
power
concept
right.
So
I
worry
about
this.
One
move
become
to
should,
I
say
so,
become
to
like
become
to
like
the
smaller
scheduling
target
in
standard.
Like
the
object
we
want
to
schedule.
The
part
is
supposed
in
this
object.
B
B
This
is
my
concern
and
if
we
expose
this
too
much
to
the
customer,
so
that's
why
I
want
to,
but
I
also
say
that,
because
sometimes
the
partner,
the
centerbox
level
of
the
problem
make
customer
support
much
harder.
So
I
want
to
find
the
middle
ground
this
and
this
the
middle
ground
and
how
we
are
going
to
solve
the
support
issue
and
give
the
bubble
up
to
certain
things.
But
at
the
same
time
not
expose
this
as
the
detail
of
the
object.
Customer
can
manipulate.
K
Yeah
great
point,
so
I
I
called
out
several
use
cases
where
we,
as
a
sort
of
like
end
user
of
kubernetes,
found
this
pretty
useful
and
yeah
like
basically
like
downstream
customers.
Very
concerned,
like
you
know,
what's
up
with
certain
plugins
like?
Are
they
actually
like
plugins
and
cni
or
csi
plugins?
Are
they
preparing
everything,
that's
necessary
for
the
pod
at
the
right
time,
and
what
we
found
is
the
existing
initialized
condition,
as
I
think
just
mentioned,
does
not
reflect
that
correctly.
K
So
this
was
sort
of
one
way
of
surfacing
this,
both
from
a
stats
perspective.
That's
something
like
cube.
State
metrics
can
surface
as
well
as
something
that
very
specific
controllers
can
also
consume
to
determine
what
they
need
to
do.
When
a
plot
they're
managing
is
failing
at
a
certain
state
and
so
to
optimize
the
set
of
things
they
they
can
get
away
with
doing.
A
Yeah
that
makes
sense
deep,
so
I
think
just
have
to
catch
up
on
the
latest
iterations
you
have
there,
but
maybe
we
refine
this.
We
can
find
ways
to.
M
C
My
reader
I've
been
reading
through
it,
looks
to
me
like
it's
more
of
a
you
know
they
just
left
the
pod
part
out
of
it
right.
Just
it's
just
not
saying
pod
sandbox.
C
Yeah,
I
don't,
I
don't
think
it's
really
trying
to
drive
into
the
sandbox
nature
right.
You
have
the
implementation
detail.
I
think
it's
just
saying
when
when
the
pod
has
been
started,
other
you
know
metrics
are
available,
such
as
networking.
Before
any
you
know,
other
containers
are
created
or
you
know
I
think
that's
what's
going
on
here
exactly.
K
This
gap
is
basically
geared
towards
like
exploring
an
alternative
condition
that
tries
to
fix
some
of
the
observations
around
initialized
that
we
found
were
sort
of
inaccurate,
but
maybe
they
had
a
reason
for
the
way
it
is.
A
Yeah,
it
makes
sense-
I
I
think,
I'll
review
it
just.
I
think
it's
sometimes
useful
to
like
talk
about
the
emotional
response
to
a
term,
and
that
was
just
my
emotional
response
of
like
man.
I
wish
we
weren't
exposing
the
concept
of
sandbox
versus
like
the
attributes
of
the
pod
that
are
present
at
that
point
in
time,
and
the
other
thing
like
in
my
emotional
response
would
be
like
sandboxes
can
get
teared
down
and
brought
back
up
in
the
life
of
a
pod.
B
A
For
that
it
was
like,
oh
in
my
own
mental
state,
that's
an
overloaded
concept
versus,
like
you
know,
very
prescriptive
or
less
overloaded
terms
for
me
personally,
others
can
feel
differently,
but
I
just
that
that
was
one
of
my
struggles
with
that
deep.
So
maybe
we
can
iterate
on
it.
Yeah.
B
B
Should
we
we
think
we
want
to
move
forward?
I
mean
not
move
forward,
it's
not
mean
that
we
definitely
want
to
hit
the
alpha
target.
It's
just
continue.
Review
continue,
iterate
the
proposal
and
the
implementation.
We
need
to
find
the
reviewer
and
and
approval
right
for
this.
One.
A
Yeah
we
were
both
ilana
myself
gave
initial
feedback,
so
we
can
okay.
B
Can
we
also
have
some
other
people
volunteer
to
review
besides
the
hysterical,
also
still
overloaded
so
and
direct
deafness
will
be
the
poor,
and
but
can
we
help
community
grow
more
people
and
chutong?
Since
you
already
look
into
this
one
before,
can
you
take
the
ownership
of
the
reviewer,
the
first
round
of
the
reviewer
and
feedback.
D
E
Okay,
all
right
yeah,
I'll
I'll
poke
swank
on.
D
D
Because
I
know
vancouver
created
this
one,
but
I'm
not
sure
about
the
other.
B
H
Okay,
this
is
quite
quite
minimal
change
to
existing
device
looking
api,
so
I
think
kevin
has
almost
implemented
her
okay,
okay,.
D
D
Right,
okay,
so
that's
all
of
the
caps
cut
from
1.24.
So
now
we
can
move
on
to
the
new
caps.
Okay.
First
one
username
space
support.
E
D
Okay,
cool,
so
does
that
mean
we're
good
on
this
one?
So
we
can
move
on
talk
about
the
yeah,
okay,
class-based
resources
in
cri.
O
O
So
there
were
all
both
good
questions
and
there
are
open
questions
so,
but
in
terms
of
the
implementation,
but
correctly
is
in
the
scope,
it's
it's
pretty
minimal,
so
as
as
soon
as
and
if
we,
if
we
just
get
an
agreement
on
the
terminal
terminology
and
how
how
how
it
should
work,
then
I
I
don't
think
the
implementation
is
too
big
really.
But
it's
probably
more
of
the
fundamentals
of
time.
O
E
Yeah
exactly,
I
think
we
yeah.
We
need
a
follow-up
to
answer
some
of
the
questions
that
were
raised
in
the
last
one.
So
maybe
we
look
forward
to
the
yeah
update
yeah.
B
So
do
we
do
we
want
to?
We
agree
about
the
class
based
resource
first
and
and
all
we
want
to
decouple
this
one
and
make
sure
we
have
that
ci.
O
K
Yeah,
this
is
a
a
cap
that
we've
been
pursuing
quite
a
bit
with
sig
storage
with
some
level
of
signaled
involvement.
K
The
idea
here
is
for
cases
like
kata
we
are
exploring,
which
is
micro,
vm,
we're
exploring.
If
there's
a
way,
csi
plugins
can
coordinate
handing
off
the
mounting
of
the
volumes
to
the
runtime
itself,
which
leads
to
a
much
better
security
posture
as
well
as
better
performance.
K
That's
kind
of
the
high
level
overview
we
initially
started
off
with
like
a
fairly
grand
mission.
I
think
we
have
managed
to
kind
of
whittle
it
down
quite
a
bit
and
like
essentially
just
need
a
couple
of
fields
that
we
are
looking
for
in
the
runtime
class
that
we
can
like
cubelet
can
just
pass
down
to
csi
plugins.
K
So
that's
mainly
the
sig
node
aspect
that
there
that
we're
looking
at
where
the
runtime
can
basically
advertise
that
hey.
I
do
have
this
ability
to
mount
volumes,
so
I
think
during
wonder,
24
I
made
a
presentation
and
we
got
some
feedback
that
what
is
not
desired.
Is
this
runtime
class
parameters
to
actually
affect
any
behavior
in
the
pod
life
cycle,
and
some
of
the
updates
we
are
making
to
the
cap
is
to
ensure
that
does
not
happen.
K
It's
just
a
bunch
of
fields
that
just
needs
to
be
passed
down
to
the
csi
plug-in
and
the
csi
plug-in
makes
decisions
based
on
those
that's
kind
of
like
where
we
are
and
want
to
take
this
forward.
E
Probably
needs
a
review
right,
yeah.
B
One
concern
I
have
is:
there's
the
csi
plugin,
so
man
so
many
versions.
We
run
into
a
lot
of
our
production
issue.
I
believe
people
using
the
if
you
for
the
user,
they
are
using
one
csr
plugin,
they
maybe
don't
have
the
problem,
but
for
the
vendor
provider
to
support
for
cs
that
are
piloting,
because
customers
can
using
whatever
css
parking
so
so
the
hand
off
more
to
the
css
parking
and
then
kubernetes
have
less
visibility
on
that.
B
What
we
are
doing
and
which
will
affect
of
the
power
life
cycle
management
and
so
that's
kind
of
what
we
are
struggling
right
now,
even
like
the
some
of
the
production
issues.
We
already
see
that
so
so
I
remember
I
remember
you
came
talk
to
the
signaled,
and
so
we
we
need
to
review
those
use
cases
more
detail
and
why
we
need
to
hand
off
more
correct.
K
Yeah,
like
essentially
the
context
is
like
csi
plugins
today,
have
this
ability
to
basically
read
the
whole
pod
spec
from
the
api
server
when
they're
handling
it
and
that's
how
we're
kind
of
working
around
this
so,
but,
unfortunately,
like
this
pattern
of
having
the
csi
plugin
like
go
up
to
the
api
server
and
basically
request
the
whole
pod,
spec
and
parse
it
out
to
figure
out
what
it
needs
to
do
is
you
know,
is
not
a
pattern
that
the
six
storage
community
encourages
and
so
basically
having
this
capability
that
the
runtime
can
assist
with
mounting
and
having
it
pass
down
to
the
csi
plugin,
just
kind
of
smoothens
it
out
and
allows
things
to
be
more
compliant
with
with
like
the
overall
storage
recommendation.
K
E
M
E
B
E
Yeah,
I
I
think,
with
sally
from
the
red
hat
side
and
if,
if
we
can
get
a
reviewer,
maybe
yeah
I'm
happy
to
take
a.
B
E
E
Okay,
so
I
guess
we
can
check
with
zimching
if
he's
interested
in
taking
it
to
beta.
D
Okay
windows
once
virg
containers
no
owner
listed
here-
let's
see
oh.
M
That's
me
mark.
Yes,
we
may
look
at
going
trying
to
go
to
stable
in
125.
I'm
still
we're
gonna
give
it
another
week
or
two
to
see.
If
that's
what
we
want
to
do,
okay,
I
don't
think
there's
any
major
changes
needed.
B
So
mark
can
we
just
in
this
case
it's
because
we
do
have
all
those
requests
right,
so
we
just
want
to
have
the
more
user
input,
let's
just
try
to
target
1.25.
This
one
graduation
is
that
okay,
okay.
M
I
E
P
P
B
E
G
D
D
All
right,
let's
go
on.
Okay,
we
have
launcher.
D
E
B
D
Oh
smoke,
oh
I'm
sorry,
none
were
so
no
shut
down.
B
E
Right,
I
can
ask
zing
if
she
wants
to.
Q
Yeah,
this
is
ryan
from
red
hat.
I
don't
think
we're
going
to
work
on
this
this
release,
and
so
maybe
take
it
out.
D
C
This
might
I'll
take
a
look
at
it
see
if
there's
anything
we
can.
We
need
to
do
in
sub
second
probes,
okay,.
O
D
D
Thanks:
okay
pod
priority,
based
in
note
group,
so
shut
down.
E
F
E
B
D
So
not
active
and
probably
we
don't
have
capacity
for
it.
G
Maybe
I
can
appear,
I
think,
the
we
need
to
clarify
the
ga
graduation
criteria
and
I'm
not
sure
we
met
for
also
for
the
reason,
I'm
not
sure
we
met
them.
I
don't
think
we
need
something
for
this
cycle
for
this.
D
Okay,
so
no
clear
ga
criteria.
G
D
All
right,
moving
on
demonstrate,
support
max
search.
E
B
B
I
agree
we
need
to
support
the
max
search,
but
I
don't
think
about
the
signal
again.
This
one
also
is
signal
is
not
driver
here
right,
so
this
is
more
like
the
controller
have
to
yeah
yeah.
So
we
definitely
support
and
like
the
to
support
the
max
search.
We
need
to
have
some
like
the
feedback
loop,
so
they
can
make
decision,
but
I
think
the
control
control
plan
api,
mercenary
and
we
are
maybe
driving
this
the
things
we
basically
just
support.
B
E
No
robert
robert
isn't-
and
I
don't
think
we'll
have
the
capacity
to
take.
B
J
B
D
Right
kublai
resource
matrix,
endpoint.
B
Yeah,
this
is
the
older
thing
we
did
some.
I
think
the
ana
actually
summarized
what
we
having
done
last
many
years
and
the
the
problem
is
all
those
things
we've
been
discussed
and
a
talk,
but
we
didn't
finish,
though,
so
that's
kind
of
a
problem
today
here.
D
B
I
don't
think
that
david,
okay,
all
right
so
david,
did
help
the
alala.
To
summarize
that
talk
share
with
the
signal
the
community
recently
right.
So
so
so
maybe
we
need
peter
or
maybe
david
another
favorite
part.
Can
we
follow
up
on
this
one
and
then
we
can
see
at
least
we
can
follow
up,
but
I
don't.
I
don't
think
we
have
boundaries.
B
See,
what's
the
what
kind
of
things
we
need
to
complete
right,
so
the
the
csi
metrics?
That's.
Actually
it
is
the
one
we
all
agree.
We
want
to
move
forward
the
current.
We
are
just
missing
some
testing
data
performance
data,
but
then,
in
that
doc
we
already
do
several
things,
but
none
of
them
is
really
finished.
B
F
B
Originally,
this
is
one,
it
is
because
we
kubernetes
have
the
monitoring
endpoint
right,
so
we
are
using
that
monitoring.
End
point
together.
We
start
the
control
plan,
specific
the
metrics
together,
so
in
the
output
we
have.
B
So
all
the
monitoring
pipeline
will
be
blocked
by
our
development.
But
meanwhile
we,
our
enterprise,
also
have
the
too
many
datas.
But
then
we
are
not
a
certified
customer.
The
the
customer
metrics
all
those
kind
of
other
requirements
for
the
monitoring
purpose.
So
there's
so
many
duplicate
metrics,
which
is
make
kubernetes
and
not
kubernetes.
No,
the
management
overhead
is
too
big,
so
original
we
have
want
this
kind
of
things.
Is
we
want
to
control?
B
Like
you
see
here,
it
is
dedicated
for
the
control
plane
and
include
of
kubernetes
using
and
reduce
those
amount
of
the
metrics
there,
and
in
the
past
we
call
we
we
have
the
bad
name
called
the
core
matrix,
but
so
we
want
to
expose
the
new
separate
of
the
monitoring
specific
metrics
from
the
control
plane
related
endpoints.
B
So
that's
the
initiate
we
start,
but
so
you
can
see
that
we
did
some
of
the
work,
but
then
we've
not
never
finished
anybody's
thing,
but
this
situation
changed
because
when
we
first
discussed
that
one
a
lot
of
the
cs
say
cri
metrics,
all
those
kind
of
things
is
not
there
yet
so
so
things
have
changed.
So
this
is
why
we
need
to
revisit
our
decision
direction.
B
F
It
makes
sense
make
sense
yeah
we
can.
We
can
review
this
one.
I
mean,
I
think,
what
one
one
just
comments
is
like.
I
believe
this
research
metrics
api
is
already
being
used
by
autoscaler
and
all
those
things
right.
So
maybe
we
need
to
separate
out
kind
of
like
the
api
perspective
of
the
research
metrics
api
and
see
how
we
can
graduate
that
and
kind
of
have
a
separate
discussion
for
how
those
metrics
are
actually
populated
right,
which
is
more
of
the
cri
stats
work
right.
D
All
right,
it's
gonna
do
a
quick
time
check.
Okay,
we
don't
have
jimmy
left.
Oh,
let's
just
go
through
them.
Ephemeral
containers.
E
I
see
a
recent
comment,
so
maybe
we
have
to
take
this
to
ga
right.
I
I
see
yeah
okay,
so
maybe
we
target
it
for
g,
then,
okay,.
J
D
Okay,
so
back
to
this
fine
grain,
couplet
api
third
edition
see.
O
E
E
C
P
C
D
Okay
moving
on
comp-
and
we
have
one
for
enabling
enables
that
company
for,
let's
see
so.
E
This
one,
I
think,
like
yes,
we
are
interested
in
taking
it
forward,
so
I
think
with
sasha
sasha
should
be
the
owner
for
sitcom
by
default.
E
Okay,
I
can.
I
can
approve
review
if
anyone
else
wants
to
help
out
happy
to
work
with
them.
E
D
And
we're
not,
do
you
know
if
the
one
above
it
the
sec
com
is
the
same
thing
or
it's
a
different.
J
E
D
Okay,
no
recent
discussion,
no
capacity
to
work
on
this
yeah,
no
racing
update.
B
D
Oh
john
I'll
put
your
name,
as
you
mentioned,
to
follow
up
on
that
one
yeah
thanks.
E
J
No,
I
haven't
heard
from,
or
I
mean
they
targeted
1.25,
so
I
mean
I
could.
O
E
Okay,
probably
needs
okay,
so
so
tim
removed
the
stale
label.
So
is
he
intending
to
work
on
it
or
have
someone
work
on
it?
Maybe
don
a
question
for
you.
B
E
E
D
All
right,
I
guess,
if
there's
a
link
to
any
cap
or
issue,
maybe
we
can
just
paste
the
link
here
as
well:
yeah,
okay,
cool
moving
on
graduate
cpu
manager,
device
manager
to
ga
francesco
you're
the
owner
for
both.
G
Yes,
so
basically
those
two
features
are
beta
from
ages,
so
I
think
it's
time
to
propose
and
discuss
if
we
want
to
move
to
ga.
I
I
believe
from
myself
that
accept
guarantee
that
we
have
end-to-end
test
coverage
and
test
coverage.
It
should
be
straightforward,
so
here
we
are
device
manager
is
so
old,
it
doesn't
even
have
a
kept
proper
cap.
It
seems
to
me.
G
B
Yeah,
so
so
francisco
or
alexandra,
can
you
help
to
follow
up
and
then
come
back
to
yeah
just
for
the
cleaning
up?
We
maybe
want
to
call
a
closure
on
those
open
and
so
yeah.
D
All
right,
I
guess
that's
all
of
the
well
caps
we've
finally
finished
all
of
them,
and
there
are
some
general
proposals
and
I
don't
think
we
have
time
today
to
go
through
them.
So
I
think
we
can
just
we'll
call
it
a
day.
Yeah
I'll
yeah
after
the
meeting
I'll
send
an
email
just
to
well
quickly.
Recap
the
planning
and
let
people
know
if
they're
and
let
them
know
if
there's
any
action
items
on
them.