►
From YouTube: Kubernetes SIG Node 20210622
Description
Meeting Agenda:
https://docs.google.com/document/d/1j3vrG6BgE0hUDs2e-1ZUegKN4W4Adb1B6oJ6j-4kyPU
A
Good
morning,
everyone
today
is
the
june
22nd
and
it's
our
regular
weekly
signal
meeting,
both
derek
and
sergey,
cannot
participate.
Elena,
do
you
want
to
kick
out
meeting
with
your
announcement.
B
Yes,
so
announcement
there
will
be
a
bug
scrub
that
we're
running,
starting
depending
on
where
you
are
in
the
world
from
my
perspective
tomorrow.
B
But
it's
going
to
be
running
the
24th
and
the
25th
of
june,
starting
at
midnight
utc
and
going
through
the
end
of
the
25th,
so
we're
sort
of
divided
up
into
three
time
zones
to
ensure
that
we
can
support
contributors
all
across
the
world
and
if
you
have
not
had
a
chance
to
get
involved
in
sig
node
we're
gonna
try
to
go
through
every
single
bug
and
there
are
450
of
them.
So
I
highly
recommend
you
show
up.
We
will
be
meeting
on
slack.
B
B
Yes,
I'm
looking
forward
to
being
able
to
hang
out
all
waking
hours
for
the
rest
of
the
week.
It'll
be
fun.
I
haven't
run
a
bug
scrub
in
a
while.
So.
A
A
C
C
B
B
D
Yeah,
I
don't
know
if
you
can
share
that
link.
I
can
share
that.
B
D
All
right,
do
you
see
it.
B
Yeah,
so
if
you
click
on
like
stage
or
something
like
that,
we
can
filter
them
down
to
just
the
sig
node
and
then
there's.
I
think
it's
in
data
in
the
menu.
B
B
B
Okay
yeah,
so
if
you
scroll
to
the
right,
the
enhancements
team
went
and
put
all
of
the
pr's
that
they
know
of
in
kubernetes,
kubernetes
more
to
the
right.
B
A
B
D
Okay,
all
right,
so
maybe
we
can
go
through
the
ones
that
don't
have
a
pr
and
see
where
things
are
and
then
go
through
the
status
of
the
other
ones
on
who's
who
is
reviewing
them
and
so
on.
D
B
E
D
E
E
Yeah
and
it's
one
pr.
E
Which
carries
both
the
api
and
the
cri
change
together
and
I'm
debating
whether
I
should
do
a
separate
pr
for
the
core
implementation
or
just
have
it
in
the
same
year
as
another
commit
which
kind
of
gets
big,
if
it's
overwhelming,
so
I'm
not
sure.
What's
the
best
way
to
bring
this
in.
But
for
now
I'm
thinking
I'll
just
start
a
new
commit
for
the
review
of
purposes.
G
E
A
separate
pr,
then
we
have
to
make
sure
that
it's
both
or
none
so.
E
Separately,
it's
a
little
easier
to
review.
Okay,
you
have
the
api
change
and
the
cri
change
independently,
which
is
tested.
It's
sort
of
not
really
not
it's
almost
non-consequential.
When
you
enable
the
feature
gate,
then
you
see
some
resources
show
up
in
the
status
field.
Otherwise
it
doesn't
make
a
difference
to
the
current
execution,
but
that
in
itself
other
than
adding
you
know
support
for
ability
to
call
update
container
resources
for
windows.
It
doesn't
really
bring
a
whole
lot
of
value
so.
E
Yeah,
so
the
change
is
going
to
be
a
separate,
commit
right
now.
The
way
I
have
since
we're
already
on
top.
I
think
mine
is
on
the
next
and
the
list.
So
why
don't
we
let
renault
finish
up,
go
through
the
list
and
then
I
can
go
into
details
of
this
one.
E
E
B
E
E
Okay
and
what
would
be
the
right
place.
B
The
issue
in
the
enhancements
repo
so.
A
So
this
is
my
indicator:
actually
there's
the
map.
If
you
look
at
the
the
dock,
the
new
keep
and
actually
the
ti's
their
status
is
more
clear.
I
guess
this
is
the
enhancement
more
care
about
the
cab
documentation,
all
those
beyond
signals,
each
sake
right,
which
is
make
sense
like,
for
example,
in
this
spreadsheet,
didn't
really
indicate
who
is
the
reviewer
only
have
the
the
rr
reviewer
approver
there,
so
we,
but
actually
we
are
working
on
the
implementation,
reviewer
and
the
design
reviewer.
A
So,
on
the
signal
perspective,
it's
not
linking
on
the
prr
reviewer,
because
that's
that's
kind
of
the
process
to
take
care
and
there's
a
collaboration,
but
we
are
really
keen
on
the
on
our
signal
because
that's
where
cause
the
feature,
how
feature
being
implemented?
How
feature
being
designed.
B
So
don
the
the
reason
that
I
suggested
we
look
at
this
is
because
the
the
prr
team-
oh
sorry,
not
the
prr
team,
the
enhancements
team
went
and
like
grabbed
all
of
the
relevant
prs
for
every
cap
that
we're
working
on
these
are
the
only
caps
that
they're
tracking.
So
I
know
that
we
have
like
a
bunch
of
other
caps
in
our
list.
A
bunch
of
them
didn't
make
it
so
this
is
like
from
in
terms
of
like
who's.
Gonna
come
pester
us
about
deadlines.
B
This
is
the
this
is
the
list
that
they're
gonna
pester
us
about,
and
this
is
the
list
of
like
what
they've
seen
so
if,
for
example,
caps
don't
have
prs
in
this
list,
then
they
need
to
be
listed
here.
Asap
because
the
enhancements
team
will
say:
hey,
there's
no
code
for
this,
and
we
need
to
make
sure
that
they
are
aware
of
it.
So
they
don't
kick
things
out
of
the
milestone.
Basically,.
A
But
there
are
another
cap,
for
example,
there's
a
certain
the
cap
merge
in
the
1.21
and
because
copy
is
merged
though,
but
the
implementation
is
not
ready,
so
the
basically
people
start
working
on
the
implementation
and
this
quarter.
I
can
quickly
scan
through
this
one
just
from
the
screen.
I
noticed
that
at
least
the
two
is
meeting
here.
D
Okay,
which
one's
done.
A
A
A
A
Yeah,
it's
not
in
this
list
and
also
it's
not
that
least
but
there's
another
one
is
in
this
list
in
this
nist
neck.
The
informal
container
actually
also
emerge
that
type,
but
the
the
implementation
is
not
writing.
Actually,
implementation.
Already.
Writing,
but
then
there's
the
debate
on
the
api
change,
so
we
cautiously
decided.
Okay,
let
this
settle
down.
A
Was
the
api
changed
a
couple
of
times
and
for
the
last
many
witness,
so
we
just
say:
oh
since
there's
another
new
concern
about
the
api,
let's
punch
that
one
and
the
move
to
this
one
is
we
we
do
include
in
here,
but
in
that
one
because
copies
didn't
change,
so
so
so,
actually
in
that
spreadsheet
they
didn't
include
it.
So
this
is
the
two
in
my
mount.
I
just
quickly
scan
so
maybe.
D
Like
maybe
we
use
this
and
like
on
the
side,
try
to
get
the
links
from
the
spreadsheet
into
the
stuff.
B
That
sounds
good
to
me.
I
can,
I
can
send
them
or
in
fact
I
can
just
paste
them
right
in.
D
Okay,
yeah,
if
you
can
keep
helping
me
update
that,
then
we
can
go
through
this
list
and
I
think
that
way,
yeah
all
right,
so
huge
paid,
storage
medium
size.
I
bet
I
saw
a
pr
for
this.
Let's
check
here.
D
It
and
elana
you
and
derek,
are
on
the
review
or
on
the
hook,
for
reviews.
B
B
Sorry,
the
link
that
they
put
in
isn't
actually
a
valid
link.
So
let
me
try
to
look
at
it
for
status
yeah.
The
pr
has
an
lgtm
on.
It
is
just
waiting
for
an
approval.
There's.
B
A
Okay,
I
we
can
either
ask
ask
a
direct.
Take
a
last
look
if
there
directly
stupid.
I
can
take
a
look
and
approve
this,
so
I
think
this
is
not
the
unrisked
right,
so
yeah
yeah.
A
Okay-
okay,
so
can
you
can
we
link
up
the
conform
test
link?
Then
we
can
see
how
we
can
process
ask
yeah
for
this.
One.
D
Okay,
we
do
that
support
to
size
memory,
backed
volumes.
D
B
Yeah
that
one
merged
okay,
that
one
was
accidentally
landed
during,
I
think
code,
free's
last
release
or
something
yeah.
B
I
saw
one
pr
which
was
doing
some
cleanup,
but
I
have
not
seen
the
second
one.
I
D
Could
you
add
the
links
here
so
we
can
track
them.
I
A
Thanks
looks
I
think
this
is
to
me
this
one
stephanie
is
in
danger
right
for
this
meet
that
night,
but
it's
okay.
We
can
continue
looking
up.
I
Oh,
I
did
a
link
in
chat
I'll
update
the
doc
later.
D
Okay,
thanks
dynamic
cubelet
configuration.
I
know
sergey
opened
up
here
and
reflectively
being
reviewed,
so
I
think
it's
looking
good.
A
C
I
don't
think,
there's
someone,
I
think
this
property
screen
yeah
windows,
privilege
container.
Yes,.
A
D
Okay,
all
right
so
see
cra.
Graduation,
like
I
don't
know
if
folks
cls,
but
we
have
a
lot
of
pull
requests
like
a
lot
of
modifying
cri
so
like
in
places.
One
swap
is
another
peter
and
david's
changes
for
the
cri
stats
also
has
cri
changes,
so
we
are
waiting
for
all
of
those
to
get
merged
and
I
think
more
importantly,
we
are
also
waiting
on
container
d.
Like
I
know
mike
brown
pinged
me
earlier
this
week.
D
D
So
in
place
vertical
scaling,
we
just
talked
about
it.
There
is
a
pr
open
where
there's
anything
is
there
anything
more
to
discuss?
Do
you
think
we
have
enough,
like
I
know
chen
from
red
hat
ibm
is
helping
there
and
joel
is
also
going
to
be
reviewing
it.
Do
you
need
more
help?
There.
E
No,
I
think
that
is
sufficient.
I
have
lantau
reviewing
the
core
implementation
as
well.
I
think
I'll
need
some
help
with
tim
tim
hawkin.
If
he's
available,
I
haven't
pinged
him,
yet
I'm
working
him
to
review
the
api
changes
so
tim
and
I
worked
closely
on
the
api
updated
api.
The
cri
hasn't
changed.
E
Much
actually
has
not
changed
at
all
from
the
past
implementation
that
we
did
so
it's
mainly
the
api
change,
which
tim
morgan
or
john
licket
if
they
can
review
that
and
sign
off
on
that,
if
you
could
please
ping
them
I'll
ping
them
separately.
But
if
you
can
ask
that's
the
most
important
piece,
if
they
sign
off
on
that-
and
it
looks
good,
there
is
one
small
change
that
needs
to
come
in
for
the
api,
which
is
the
addition
of
resize,
which
is
summary
status
of
the
state
of
the
resize.
E
That's
requested,
it's
not
a
deal
breaker
as
in.
If
it's
not
there,
things
will
still
work.
I'm
planning
to
finish
that
in
the
next
few
days.
I
think
the
most
important
thing
to
get
right
here
is
the
accounting
the
how
the
pod
resources
are
accounted
in.
The
scheduler,
when
it
schedules,
sends
new
watch
this
way
and
then,
when
restarted,
cubelet
deals
with
bots
that
I
added.
So
those
are
key
things
that
need
to.
We
need
to
get
right,
but
it
can
be
reviewed
now.
E
The
api
and
the
cri
and
the
core
implementation
is
coming
in
a
few
days.
A
Okay,
great,
I
talked
to
the
mental
this
morning.
He
will
he
committed
this
week
to
spend
more
time
on
reviewing
this
one.
It
won't
change.
If
I
could
understand
your
question,
the
new
api
change
actually
is
not
required
for
alphara.
If
we
could
get
in,
we,
we
will
definitely
getting.
If
it's
not.
I
can
ping
the
team
hacking
of
this
meeting.
A
A
E
Going
to
be
part
of
the
without
the
api
change,
because
this
is
a
change
to
the
pod,
not
pod,
spec
yeah.
There
is
a
change
to
the
part
spec
which
we,
which
we
looked
at
in
the
previous
implementation.
I
don't
think
there
were
any
major
comments
about
that
other
than
renaming
the
policy
names.
E
We
move
the
resources
allocated
from
the
pod
spec
into
status
and
then
checkpointing
all
that
is
working,
the
resizing
that
tim
wanted
per
container.
We
agreed
that
we'll
do
it
as
a
summary
for
all
the
containers
in
the
part
and
he's
agreed
to
that
at
that
time.
So
it's
not
a
big
change
from
what
I
can
tell.
I
believe
the
only
other
thing
that
is
different
from
previous
implementation
is,
as
elena
pointed
out,
we
needed
to.
D
Great,
so
the
next
one
is
liveness
group
timeout.
You
know
I
knew
you're
working
with
someone.
B
Yeah,
my
intern
is
working
on
that
one
and
got
a
little
bit
delayed,
putting
a
pr
up,
but
that
should
be
done
pretty
quickly
and
yeah.
I
think
it's
still
on.
D
A
D
All
right,
so,
while
the
line
does
that
so
next
one
is
the
sqd
and
it's
marked
as
done.
B
D
So
for
the
next
one
like
david
and
I
hope
to
review
an
outstanding
pr,
but
it's
it's
close,
there's
a
pr
already
there
for
the
priority
based
graceful
shutdown,
hopefully
david.
We
can
review
it
this
week
in
closing.
D
For
c
groups
v2,
so
we
merged
in
the
kept
changes.
So
I
think
that
the
next
step
here
is
to
get
that
ci
job
like
running
illinois,
so
the
one
that
herschel
has
been
working
on
and
basically
we
are
testing
and
fixing
bugs
as
they
are
coming
up
like
giuseppe
and
kerr
and
others
are.
They
fix
a
bunch
of
issues
already.
A
D
D
D
D
So
so
the
next
one
I
post
like
a
change
into
this
into
the
cri
earlier
today
and
I
believe,
there's
also
another
pr
open
from
tim.
I
think
david
and
I
were
tagged
on
it
so
that
we
can
probably
review
okay.
A
I
also
want
to
make
sure
this
is
also
don't
have
commitment
right,
so
just
keep
moving
forward
because
secret
version
to
the
feature
it's
quite
different
from
other
one.
It's
no
easy
way
for
people
disabled
not
to
disable
and
define
the
problem.
So
that's
why
we
want
to.
We
are
working
on
the
testing
right,
yeah,
all
those
kind
of
things,
but
we
want
to
also
make
progress,
so
we
just
keep
them
over,
but
no,
absolutely
not.
Okay,
cool.
D
Yeah
agreed
on
totally
away
yeah,
so
the
next
one
is
memory
manager,
rtm
francesco.
Do
you
guys
want
to
talk
to
this
status
of
this
one
waiting
for
claire
kevin's
reviews.
L
Hi
folks,
so
the
pod
resources,
api
pull
request
was
merged.
Today,
thanks
to
kevin
for
review
and
approve,
and
in
general
it's
a
single
request
that
was
mandatory
for
the
relation
to
better
like
it
can
be
wonderful
if
you
can
also
handle
it
contains,
but
again
it's
not
mandatory,
so
I
just
pinned
to
humor
now
and
to
lana
under
the
moving
to
the
bed,
to
the
better
pull
request.
Others
moving
to
the
better
pull
request.
If
you
can.
D
B
The
pr
is
up,
seth
gave
it
an
initial
review
and
I
think
he's
happy
with
it.
There
were
some
changes
in
terms
of
like
terminology
that
we
want
to
make
to
both
the
cap
and
the
api,
but
other
than
that.
As
far
as
I'm
aware,
I
think
it's
good,
like
we
have
some
pr
jobs
running
that
enable
swap-
and
at
least
one
of
them
is
working.
B
The
other
one
seems
to
have
an
issue
right
now,
so
I
would
say,
like
it's
probably
mergeable,
as
is
I'm
just
waiting
on
api
review.
Okay,.
D
M
Francesco
here,
hi
everyone
so
right
now
we
have,
we
have
split
the
pr
to
because
it
needs
api
changes
because
it
touches
cubelet
config.
It
adds
a
new,
a
new
flag.
So
we
split
those
changes
and
kevin
approved
them
thanks
to
him,
which
is
resulting
a
lot,
but
we
need
approval
from
ap
review
groups.
Basically,
it's
everything
is
in
the
pr
I
shared
the
the
the
the
here
I'm
talking
about
in
the
chat,
because
this
pr
is
is
a
spin-off
of
the
main
one.
M
It's
the
same
content
but
focused
on
the
api
changes.
So
it's
easy
to
review
for
ap
api
reviewers.
So
we
basically
we
need
that
and
then
we
can
keep
moving
and
the
changes
pertaining
to
signaled
kevin
did
a
few
passes
already.
So
thanks
to
him
again,
but
not
looks
good
to
me
yet
so
it's
progressing
and
having
some
attention
from
api
reviewers
will
make
the
progress
even
faster.
So
this
is
it
thanks.
M
D
H
D
B
B
D
Okay,
great
the
next
one,
peter
and
david.
I
know
you
all
opened
the
pr
yesterday
for
the
cri
changes.
You
want
to
talk
what's
happening
here.
G
Yeah
there's
a
pr
open
for
the
cri
changes
and
then
also
a
separate
pr
open
for
adding
the
functionality
in
the
cubelet
so
that
that
will
just
need
review.
I
need
to
test
it
and
actually
do
some
of
the
you
know
changes
on
the
cri
implementation
side,
but
I
wanted
to
open
that
it's
basically
at
proof
of
concept
stage,
just
showing
that
the.
D
K
Yeah
I
mean
the
plan
is
to
add,
like
all
the
cri
fields,
basically
for
the
for
the
summary
api
that
we
outline
in
the
cap
and
then
add
those
to
the
cri
and
then
have
another
pr,
as
peter
mentioned,
have
kubelet
use
those
fields
and
then
later
probably
I'm
guessing
we'll
actually
go
out
to
all
the
you
know:
cryo
container
d
implementations
and
implement
those
those
new
fields.
K
G
I
think
there
does
exist
a
couple
of
things
that
a
couple
of
to
do
is
in
the
cri
change.
Pr
that
I
think
we
need
to.
I
need
to
you,
know,
save
it
and
we
need
to
figure
out
as
a
group,
but
then
once
those
are
done,
then
I
I
think
it
could
make
sense
to
merge
the
cri
changes
to
unblock
cri
v1,
as
well
as
enable
us
to
start
going
to
the
cri.
G
And
getting
those
changes
in
okay.
D
All
right
so
yeah,
I
know
like
we
have
a
dedicated
time
set
up
tomorrow,
to
discuss
this
more
so
maybe
we
can
update
signaled
after
we
talk.
D
So,
second,
by
default,
like
there
were
some
changes
to
the
cap
based
on
feedback
from
jordan-
and
I
know
like
derek
approved
them
earlier
today,
so
the
kept
changes
are
in
and
now
the
pr
should
be
ready.
So
I
think
this
one
should
be
okay,
it's
not
a
big
change.
D
D
Docker
shim
removal
same
status.
I
guess
communication
and
transition.
D
So
probably,
I
think
we've
hit
a
line
under
which,
like
all
these
things
are
yeah
probably
will
fall
out
to
the
next
one
status:
dot
host
ip
added
to
pods.
D
D
D
Open
don,
do
you
know
if
sergey
is
going
to
reveal
this
one
or
you
need
someone
to
sign
up.
A
Yeah,
but
I
don't
think
he
committed
to
to
do
this,
I
will
ask
him
and
he's
sick
ask
him
so,
but
I
don't.
E
A
D
A
That's
the
only
things
that
I
think
we
are
missing
from
the
yeah.
We
didn't
talk
about
the
username
in
space,
but
do
we.
A
D
Okay,
all
right,
that's
it!
I
guess
like
we
like
the
everyone
needs
to
keep
reviewing
and
help
get
these
in
before.
A
A
B
B
So
just
make
sure
that
you're
communicating
like
put
in
cross
references,
because
you
know
the
the
person
who's
working
on
the
cap-
that's
probably
the
most
important
thing
to
them,
but
for
everybody
else,
they've
got
like
10
000
things
going
on,
so
make
it
easy
for
us
yeah.
E
In
my
case,
I
just
updated
the
issues
with
the
pr.
I
also
was
chatting
with
tim
on
slack
and
he
he's
aware
of
the
pr
it's
in
this
queue
he'll
get
to
it.
It
looks
like
so
we're
going
to
do
that
thanks.
E
Oh
okay,
it's
all
pretty
much
covered.
I
believe.
Let
me
just
go:
go
through
the
go
through
the
notes
here
and
then
see
if
I
missed
anything
so
yeah
the
pr102884
is
out
for
review.
Tim
is
going
to
look
at
it
and
also
commented
on
it.
Saying
he'll,
look
at
it
this
week
he's
making
time
for
it.
There
are
two
the
core
changes
and
the
cri
changes
are
in
two
separate
commits
a
third
commits
coming
in
a
few
days
which
will
carry
the
core
implementation.
E
The
status
of
that
is
essentially
no
checkpointing
is
working.
The
basic
end
to
end
results
are
working
with
the
new
design.
What
we
need
is
the
resize
status,
which
is
really
nice
to
have,
but
it
doesn't
break
code
implementation.
E
The
most
important
thing
that's
coming
after
this
is
the
e2e
tests
that
we
already
have
chen
wang
did
a
lot
of
work
on
the
previous
iteration
of
this,
and
she
signed
up
to
port
this
and
to
adapt
it
to
the
updated
design,
and
it
looks
like
we
should
be
on
track
for
july,
9th
cool.
B
Yes,
so
in
the
tac
md
for
the
bug
scrub
tomorrow,
so
I
wanted
to
just
briefly
talk
about
this.
So
there's
a
link,
so
you
can
see
what
I've
written
there,
but
basically
a
lot
of
people
effectively
file,
help
requests
in
kubernetes
kubernetes
and
then
they
tag
sig,
node
and
kubernetes.
Kubernetes
is
not
where
you
should
be
go
to
asking
for
help.
We
don't
currently
like.
If
you
try
to
file
a
support
request
there,
it
will
send
you
a
link
to
the
discord.
B
Sorry,
not
the
discord,
the
discourse
we
have
like
a
forum
for
helping
people
with
issues,
so
I
just
wanted
to
make
sure
everyone
is
okay
with
this
that,
as
like
sort
of
a
sig
node
policy,
if
people
ask
signod
for
like
a
support,
request,
basically
like
there's
something
wrong
on
my
cluster,
and
I
don't
understand
what
it
is
that
we
will
be
closing
those
issues
and
telling
them
to
take
it
to
the
forums.
B
A
I
like
that
yeah
in
the
past,
we
did
do
a
lot
of
support
when
and
but
a
lot
of
our
angelo
engineer
or
any
move
on.
So
I
know
I
noticed
that
yeah
exactly
so
many
people
found
the
issue
not
like
necessary
as
they
support
cases.
They
just
think
are
we
have
some
issue
or
missing
such
there's.
B
B
This
is
the
upstream
like
repo,
so
there's
a
lot
of
that
sort
of
thing,
and
I
don't
think
that
that's
controversial
or
anything
like
that,
but
just
generally
for
folks
who
are
even
asking
for
help
with
the
upstream
kubernetes
stuff,
the
issue
tracker
in
kubernetes
kubernetes
is
not
supposed
to
be
used
to
track
that
that's
what
the
forum
and
slack
is
for.
B
So
I
just
basically
it's
one
of
those
things
where
it's
like
you
don't
want
to
be
rude
or
mean,
but
it's
not
the
right
place
for
people
to
be
asking,
and
if
they
ask
there,
then
they
just
get
ignored
and
that's
not
a
good
experience
either.
So
I
just
as
long
as
we're
all
clear
that
that's
what
we're
going
to
be
doing
with
these.
I
don't
think,
there's
any
problem
with
that
and
we
can
maybe
get
some
folks
from
signo
to
try
to
help
out
and
with
questions
on
the
forums
or
whatever.
B
Cool,
I
think
so
we
will
be
closing
those,
and
I
think
that
will
help
us
get
the
bug
count
down
as
well.
Totally
yeah.
A
J
J
B
Properly
but
other
than
that,
I
think
once
once
the
code
changes
merge,
then
we
can
add
the
periodics,
so
they'll
be
running
all
the
time.