►
From YouTube: Kubernetes SIG Node 20210202
Description
Meeting Agenda:
https://docs.google.com/document/d/1j3vrG6BgE0hUDs2e-1ZUegKN4W4Adb1B6oJ6j-4kyPU
B
Hello,
it's
a
signal
meeting,
it's
2nd
of
february
2021,
so
pr
status,
we
have
increase
of
prs,
which
is
based
on
pure
number,
but
I
analyzed
what
was
created
and
there
are
a
lot
of
pr's
for
every
small
component,
converting
the
logs
from
unstructured
to
structured
logs.
So
hopefully
we
can
bulk
review
the
sprs
and
get
rid
of
them
and
it
will
get
us
into
green
zone,
so
it
will
be
less
pr
created
than
closed
and
something
so
yeah.
B
So
this
increase
is
not
unnaturally
just
because
somebody
opened
a
lot
of
small
prs
for
a
structured
attraction.
Logging.
D
So
does
anyone
want
to
just
like
volunteered
to
act
as
the
primary
shepherd
for
all
these
structure,
long
ones.
C
I've
been
kind
of
doing
it
because
I
have
to
via
sig
instrumentation,
so
like
cerrathius
merrick
at
google
owns
this.
I
have
not
been
able
to
get
an
update
from
him
in
terms
of
what
we
are
doing
for
121.
so
like.
I
think
we
want
to
target
beta
in
121,
which
means
we
have
to
migrate
everything
and
then
set
up
set
up
static
validation
to
ensure
that
people
aren't
reintroducing
the
old
log
formats.
D
C
Yeah,
I
guess
the
the
biggest
issue,
so
I've
been
reviewing
them
obviously,
but
I
don't
have
approver
so
I
need
to
like
pester
people
to
get
the
I
mean,
there's
no
real
urgency.
They're
they're,
not
really
clogging
things
up.
I
don't
think.
E
C
Oh
great
caps,
I
would
love
to
talk
about
caps,
so
I
put
an
item
on
here
to
talk
about
caps,
because
the
cap
freeze
is
fast
coming
and
a
lot
of
people
wait
until
the
very
last
moment
to
try
to
get
their
kept
merged
and
then
have
to
go
through
an
exception
process,
because
it
takes
time
to
get
everything
done.
That
needs
to
get
done.
C
So
I
wanted
to
make
sure
you
know:
we've
got
a
week,
it's
coming
up,
it's
the
ninth
and
I
wanted
to
make
sure
everybody
knows
what
needs
to
get
done.
So
we
have
a
planning
doc
that
mourinho
put
together,
which
I
linked
and
one
of
the
things
I
did
yesterday
was.
I
went
through
that
doc
and
I
ad
I
checked
every
single
thing
that
was
in
there
and
I
added
to
do's
before
freeze.
So
some
of
them
don't
have
anything
to
be
done
before.
C
Freeze,
for
example,
like
the
swap
stuff
we're
just
working
for
towards
the
cap,
there's
nothing
to
be
done
there
for
those
that
had
outstanding
prs
that
needed
to
merge.
I
linked
those,
but
for
a
lot
of
them
like
there's.
No
pr
that
needs
to
merge.
We
just
need
to
update
the
cap
to
include
the
newest
milestone,
and
that
may
actually
also
include
some
prr
review
when
we
do
that.
C
So
all
of
the
to-do's
are
there,
but
in
order
for
things
to
be
worked
on
as
a
cap
in
this
release,
they
have
to
have
a
tracking
issue
in
kubernetes
enhancements.
They
have
to
have
that
issue
in
the
121
milestone.
They
have
to
be
implementable
in
the
cap.yaml
and
they
have
to
target
the
correct
release
and
stage
in
the
kept.yaml
like
that
is
the
new
process
for
the
release
team
and
they
want
the
sig
chairs
to
take
all
of
those
caps
and
bring
them
to
them
to
track
them.
C
So
it's
not
like
previous
releases,
where
they
have
been
doing
all
of
the
work
of
tracking
and
shepherding
things
and
whatnot.
So
if
you
are
not
on
that
list-
and
you
don't
have
a
tracking
issue
or
like
any
of
these
things,
it
is
unlikely
that
it's
on
the
agenda
for
121s,
so
you
should
be
working
towards
fixing
that
I've
been
getting
some
pings
on
random
prs.
Where
I
had
asked.
Please
file
a
tracking
issue
and
it's
like
you
still
need
to
file
the
issue
that
that
requirement
is
not
going
away.
D
So
I
tried
to
take
two
really
simple
features
that
we
had
had
in
alpha
and
guinea
pig,
the
pro
review
process
to
get
them
to
beta
last
week,
and
I
know
we
have
a
week.
My
experience
is:
it's
like
a
five
day,
turnaround
time
on
pr
reviews,
which
I
assume
hopefully
will
go
down
as
as
more
eyes
get
put
on
it.
But
the
one
thing
that
stood
out
to
me
was
in
the
current
process.
D
The
bot
on
prr
reviews
only
verifies
the
change
to
the
cat
pad
the
corresponding
change
in
the
prod
readiness
tree.
If
you
had
created
the
file
in
the
prod
readiness
tree,
and
so
one
thing
I'm
anxious
about
as
we
go
through,
these
is
just
ensuring
that
everyone
does
go
and
create
the
file
and
the
prod
readiness
tree
where
there
wasn't
one
already,
and
I
don't
know
if
anybody
knew
if
there
was
anything
happening
in
that
pr
verified
job.
That
says.
C
Yeah
there
is,
I
I've
hit
this
a
bunch
with
the
instrumentation
stuff.
So
basically,
as
soon
as
you
target
the
latest
milestone
and
you
set
the
phase
or
whatever,
if
there
isn't
a
corresponding
thing
in
the
pr
readiness
review,
the
verify
will
fail
and
it
won't
let
you
merge
so
then
you
have
to
add
that
file
and
then
the
the
prr
approvers,
like
approval,
is
required
in
order
for
the
pr
to
get
merged.
C
It's
a
little
bit
like
flaky
in
the
sense
that,
like
I,
have
seen
some
where
it
sort
of
like
skips
under
the
hood,
and
the
reason
is
because
it's
not
requiring
all
of
that
metadata
to
be
filled
out.
So
if
you
just
don't
include
it,
it
will
ignore
it.
So
that's
the
concern
that
I
have
on
some
of
these,
which
is
why,
like
on
almost,
I
think
everything
in
the
spreadsheet
I
said:
yeah
we've
got
to
get
the
metadata
updated
because
that'll
force
you
to
go
through
the
prr
review.
A
E
A
We
are
not
engaged
to
promote
or
deliver
anything
right.
So,
okay,
that's
more
clear
because
for
this
one
we
do
have
the
reviewer.
We
have
everything
owner
identifier.
There
are
other
things
like
the
windows
privilege
container.
I
understand
the
mac,
hope
markets
here
and,
and
we
are
missing
off
the
approver.
I
think
I
think
we
do
have
the
reviewer
so
because
I
think
that
both
ninten
and
the
yuji
actually
jumped
into
provide
the
review.
I
will
chasing
those
two
and
make
sure
that's
the
whole
review
process.
A
It
is
the
throughout
this
one
and
so
so
mark
you
can
update,
put
their
name
as
the
reviewer
and
we
haven't
identified
approval
there.
Yet
if
there's
no
one,
I
can
be
the
approver
so
then,
which
means
I
also
have
to
review
that
one.
So
just
just
make
sure
we
are
everything
and
then,
in
that
case
we
might
have
to
add
a
milestone
for
this.
One
also,
hopefully
mark
is
here
today.
F
Yeah,
I
think
I
need
to
update
the
milestone.
I
just
realized
that
that
was
still
targeting
120..
I
will
go
ahead
and
add
lantau
and
yuju
as
the
reviewers.
If
anybody
from
signored
would
like
to
act
as
the
approver
here,
please
let
me
know
definitely.
A
And
then
there's
the
other
several
cap
on
the
feature
we've
been
talking
about,
like,
for
example,
cisco's,
I
think
even
like
the
owner
is
unclear,
has
the
question
mark
and
they
are.
I
think
that
they
attended
the
signal
meeting
last
week,
but
they
didn't
clearly
say
they're
going
to
be
owners,
so
we
also
didn't
clearly
identify
the
review
and
approval,
so
we
have
to
reach
out
for
them
to
them
and
make
sure.
A
I
personally
want
to
go
through
that
one,
because
that's
kind
of
the
more
you're
out
here,
right
and
and
we
should
at
least
have
the
owner
who
is
the
owner
and
and
who
is
going
to
be
the
reviewer.
So
if
that
one
is
not,
you
know,
I
think
about
until
today
is
unclear.
I
think
there's
no
way
we
are
going
to
deliver
that
in
the
near
term
of
the
milestone.
A
A
F
Maybe
it
would
be
good
to
go
through,
and
at
least
the
ones
that
look
like
they
are
going
to
make
progress.
This
milestone,
make
sure
that
they
have
owners,
reviewers,
approvers,
especially
a
lot
of
the
new
enhancements,
and
then
because
I
think,
if
we
wait
till
next
week,
there's
not
going
to
be
enough
time
to.
A
Like
this
idea-
yes,
maybe
yeah-
maybe
at
least
mark
as
the
color
called
animated
community
notes,
and
if
people
disagree
anything
will
be
speak
out
and
then
we
can
require.
So
then
they
can
provide
additional
information.
Cool,
okay,
let's
go
yeah,
okay,
so
yeah.
Let
me
give
you
the
co-host
so
you
can
share.
Then
we
can.
C
E
E
B
There's
also
a
windows
improvement
like,
but
it's
probably
additive
for
security
context.
F
Yeah,
I
think
we
addressed
some
of
the
security
context,
improvements
actually
in
the
windows
privilege
container
cap,
there's
some
suggestions
for
new
cri
fields
that
I
think
one
tau
thought
looked
good.
D
E
A
E
Or
I
can
just
say
like
to
update
yeah,
okay,
yeah.
C
C
Runners
group,
I
think,
like
we
just
basically
graduate
this
one
yeah.
It's
it's
basically
ready
to
go.
It
just
needs
to
be
added
to
the
even
I
think
we
have
the
issue
in
the
milestone
we
just
need
to
like.
E
Okay,
all
right
so
no
graceful
shutdown,
so
david
and
I
david
sergey
and
I
met
with
clayton.
He
had
some
feedback
on
it
last
week,
so
we
met
on
friday.
So
I
think
is
david
on
the
call.
E
David,
do
you
think
we
you
have
the
bandwidth
for
this
in
121.
H
Yeah,
I
think,
yeah
I
think,
moving
like
amy
to
move
this
debate
in
when
121
makes
sense.
I
have
some
there's
been
someone
who's
been
helping
out
with
et
testing
of
the
feature
which
has
been
really
awesome
to
get
her
help
with
I'm
the
only
concern
I
have.
H
I
think
it's
fine
to
aim
it
for
121,
but
we
might
need
some
more
time
to
incorporate
other
people's
feedback
if
we
get
some,
so
I
think
it's
a
good
goal,
but
if
not
we're
going
to
move
it
later,
so
I
think
it's
it's
good
to
aim
it
for
121.
For
now,.
I
D
E
A
E
E
A
I
can
answer
questions
you
know:
do
we
do
we
have
any
data
about
today's
cube,
kubernetes
resource
management
dependency
on
the
sql
pro
version?
One?
Do
we
have
any
of
those
data?
I
never
saw
that
when
people
say
oh,
we
have
this
dependency
or
maybe
we
don't
even
those
dependencies.
Oh
it's
been
covered
by
sick
version.
Two
or
it
is
there's
authoritative
way
we
could
achieve
the
single.
I
I
do
see
that
in
your
original
type
mention
something,
but
it's
not
like
the
concrete.
E
E
The
I
think
the
resource
management
dependencies
direct
like
how
they
are
using
it
currently
and
if
any,
if
they
need
to
change
anything.
Somebody.
D
C
So
in
that
case,
I
guess
no
change
is
required,
because
we're
not
going
to
graduate
at
this
release
we're
just
going
to
continue
working
on
it
is
that.
E
C
D
D
No,
the
minimum
that
it
remains
in
alpha's.
You
know
one
release,
but
it
there's
there's
plenty
of
precedent
for
something
to
go
out
in
alpha
in
one
release
and
immediately
become
beta.
The
next
release,
the
beta,
the
ga,
can
be
elongated,
depending
on
on
the
nature
of
the
feature.
E
D
It's
like
automatically
two
releases
of
being
on,
but
off
by
default
on
the
cubelet
to
make
sure
that
everyone's
upgraded
to
the
level
that
could
be
on
it,
so
typically
that
that
guidance
will
be
covered
in
the
kep
review
or
pr
review.
When
talking
about
version.
Skew.
E
All
right
so
moving
on
to
the
next
one,
so
username
spaces.
So
there's
been
a
lot
of
comments
from
tim,
hawkins
and
michael
toffin
on
the
cap
and
so
right
now
I
I
think,
like
we
haven't
reached
agreement
and
I
don't
think
we'll
be
able
to
in
a
week.
So
maybe
the
goal
should
just
be
to
finalize
the
cap
during
121
like
just
keep
working
on
it
and
agree
on
the
on
the
design
and
what
what
stages
it
should
be
done.
C
So
for
all
of
these
ones,
that
I
are
that
we're
not
gonna
change
the
the
the
stage
or
like
target,
I
think,
for
the
release
team,
I'm
just
throwing
n
a
on
there
yep.
Yes,
that
sounds.
C
C
Oh,
I
think
it's
yellow
it
actually
gets
merged
because
it's
not
yet
merged
and
implementable.
E
D
Actually,
just
to
make
sure
on
that
last
one
like
I
had
reviewed
the
pr,
so
we
just
wanted
any
objection
to
just
oh,
I
know
at
least
clayton
myself
had
reviews
sergey,
but
just
to
we
should
probably
identify
clear
honor,
I'm
happy
for
that
to
be
me
or
or
anybody
else.
D
A
B
Yeah
but
official
approver
called
john
biller
mark,
at
least
in
in
in
the
pr
it
says,
john
de
la
merica.
B
C
C
C
Oh,
that
was
that's
not
for
the
cap.
That's
for
the
prr.
A
Yeah,
okay,
so
I
I
do
I
just
think
about
this:
it
should
be
should
be
yellow
and
it's
target
for
the
better
okay.
E
E
Auto
sizing
system
reserved,
so
we
have
it
on
the
agenda
for
today
and
depending
on
how
that
goes,
we
can
color.
It.
C
A
I
A
No,
I
think
about-
I
just
saw
that
and
I
guess
maybe
we
cannot
really
target
for
this
one
right.
So
we
we
people
will
move
on
with
the
cap,
but
I
don't
think
there's
the
goal
for
the
for
this
release.
A
E
C
So
I
put
alpha
there
as
the
goal
for
this
release,
so
the
idea
is
the
owner
wants
to
put
it
to
alpha
in
this
release.
But
it's
red
because,
like
the
p.
K
Yeah,
actually,
we
want
to
consider
it
for
1.21
if
it's
possible.
A
E
Okay,
so
the
next
one
is
this
cuddle
so
dawn.
I
know
you
started
with
this
one,
but
you're
thinking
this
is
red
because
we
don't
have
a
clear
owner.
E
C
They
were
very
excited
about
finding
something
to
work
on,
and
this
is
probably
the
one
thing
that
we
don't
have
an
owner
for.
I
think
this
used
to
be
s-t-c-s
and
then
who
disappeared,
sort
of
and
said
it
needed
a
new
owner.
There
was
like
a
comment
chain
which
I
resolved
because
patois
said
he
wanted
to
work
on
it.
So
I
mean
we'll
see,
I
guess
really.
D
D
I
could
be
perfectly
fine
if
we
just
said
as
a
group
that
improvements
to
the
user
experience
not
yet
identified
is
just
a
separate
cap.
A
shocking
number
of
the
world's
applications
are
taking
advantage
of
this
capability
now,
and
so
maybe
we
should
treat
this
a
bit
like
cri,
where
the
the
perfection
is
in
some
ways
the
enemy
of
progress.
D
D
And
if
that
means,
if,
if
this
individual
doesn't
want
to
shepherd
that,
I'm
happy
to
then
take
on
moving
this
to
ga
myself.
But
I'd
want
to
appear.
A
I-
and
this
is
the
good
way
to
also
and
then
also
thanks,
alana,
to
work
with
them,
because
I
think
it's
hard
for
the
different
time
zone
and
we
want
to.
We
want
to
have
more
contributors,
yeah.
C
Think
that's
fair.
I
will
try
to
given
what
derek
has
said
aggressively,
slash
down
scope
to
make
sure
that
this
is
feasible
within
like
the
next
week.
L
Yeah,
so
there
is
no
enhancement,
but
I
found
an
original
design.
Docs
I've
been
sort
of
converting
that
over
to
an
enhancement.
I
haven't
opened
an
issue
yet
because
I
thought
we
had
to
open
both
in
parallel,
but
I
finally
opened
the
issue
first
I'll.
Do
that
today
and
get
the
enhancement
out.
E
Okay,
so
I'll
call
it
yellow
if
folks
don't
have
objections
and
see
how
it
goes.
C
C
How
can
it
be
beta
if
it's
is
this
one
of
those
things
that
we
implemented
without
a
cap?
Yes,.
A
This
is
part
of
the
ci.
I
think
that
people
will
know
this
is
the
part
of
the
style
see.
I
actually
define
a
lot
of
things
that
I
define
the
security
monitoring
and
the
logging
and
and
also
sandbox
every
many
many
things,
and
they
are
also
actually
in.
In
my
original
mind,
I
even
want
to
be
defined
of
the
image
management
we
haven't
done
that
yet
so,
which
is
on
purpose,
leave
that
out,
but
to
define
many
many
things
through
this
ai.
So
that's
why.
A
E
A
Circuit,
can
you
also
help
on
this
one?
I
want
to
make
sure
there's
no
any
production
potential
issues
on
this
one,
even
I
thought
about
that
one,
but
still
I
want
to
make
because
in
the
past
we
hold
it
this
way.
It
is
because
openshift
do
have
production
concept.
So
then,
when
we
hold
this
for
a
long
time
and
I'm
not
sure
all
the
other
production
evol
may
maybe
may
cause
some
issues,
but
I
don't
know
just
make
sure.
E
Yeah,
but
it's
okay
as
well,
so
the
next
one
is
memory
manager
so
terek.
I
think
it's
waiting
on
review
from
you
and
clayton.
D
Yeah
so
we
can
mark
the
screen
and
we
just
need
to
go
and
get
the
cat
minute
after
kep
metadata
updated,
but
the
pr
was
was
really
far
along
and
just
needed.
Final
review.
C
E
D
Yeah,
so
is
that
yeah
is
the
cap,
metadata
and
stuff
updated
all
right
and
I'll
I'll
take
a
look,
but
if
it.
D
Yeah,
basically,
this
one
and
the
next
one
from
francesca
we're
just
trying
to
get
the
pr
and
all
that
stuff
worked
out.
So
yeah.
C
I
saw
there
was
a
thread
in
sig
node
from
francesco,
so.
M
D
E
D
Well,
it
goes
back
to
like
the
overarching
question
here,
we'll
have
to
tackle
as
a
sick,
and
maybe
we
can
spend
a
minute
talking
about.
It
is
additive
enhancements
to
an
already
stable
api.
Endpoint
are
not
really
things
that
we
can
track
within
a
feature.
Gate
concept
right
now,
so
like.
D
I
it's
probably
the
best
way
I
can
frame
this
right
now,
I'd
be
like
how
would
we
track
future
versions
of
apis
that
get
added
to
the
cri
after
it
goes
ga
like
they
don't
get
gated
behind
feature
gates,
because
it's
kind
of
a
cubelet
to
another
end
consumer
api,
endpoint
versus
we
just
have
to
ensure
these
things
are
backwards,
compatible
open
to
people's
thoughts
on
that
honestly,
but
we've
been,
we've
been
very
varied
on
that
approach
like
when
we
did
the
numa
alignment
stuff
like
we
changed
the
cri
just
ensured
it
was
backwards
compatible
and
we
didn't
like
gate
every
change
to
the
cri
as
a
feature
gate
or
anything.
D
A
I
think
that
we
cannot
just
add
other
feature
for
either
feature
where
you
have
to
think
about
proper
production
layout
to
some
degree
right
so
in
other
cases,
feature
gate
or
backward
compatible
is
all
important
detail.
But
do
we
just
need
to
think
about
when
we
draw
out
to
the
future
and
what's
the
way
it
could
be
really
harsh
way
and
it
could
have
been
like
the
really
smooth
silly
back
water
compatibility,
it's
just,
but
we
need
a
at
least
to
tell
the
user
here.
D
D
D
A
So
when
I
talk
about
the
production
production
possibility
lot,
it's
not
just
for
the
api
backward
compatibility
honestly,
for
example,
for
since
we
mentioned
that
know
the
size
right
so
so,
when
we
first
introduce
of
the
node
allocable,
that's
really
small
things
to
add.
So
it's
like
concept
training
is
pretty
simple,
but
roll
out.
You
have
to
think
about
more.
A
You
cannot
the
reason
at
the
end,
there's
no
api
backward
compatible
issues
because
all
is
implement
detail,
but
we
didn't
force
those
kind
of
things
it
is
when
you
roll
out,
you
have
to
think
about,
it
could
be
potential
customers
all
of
a
sudden.
Their
code
is
being
their,
their
workload,
is
being
evicted
and
they
don't
have
the
enough
know
the
the
cluster
level.
Even
don't
have
the
capability
to
go
out.
So
that's
the
basic.
Is
the
production
outage
for
some
of
the
user
right?
A
So
then
we
have
to
carefully
think
about
how
to
introduce
that,
of
course,
the
way
it
is
pretty
straightforward.
So
so
we
don't
set
any
default
value.
We
give
some
suggestion,
but
don't
set
that
one
and
then
we
we're
doing
some
other
thing
and
also
feature
kit.
I
just
want
to
say
that
it's
not
just
it's
not
just
simply
like
the
api
backward
competitor,
that's
the
one
of
the
things
we
have
to
think
of,
but
the
other
things
for
rule
out.
We
have
to
think
about
from
high
level.
It
needs
to
give
me
yeshua.
D
E
D
D
But
that.
C
C
D
C
C
C
Basically,
according
to
at
least
the
cap
metadata
for
this
one
like
it's
supposed
to
be
targeting
stable
in
121,
and
it
has
a
feature
gate
so.
D
A
So
so
derek
is,
do
you
think
well
that
can
be
rule
out,
be
part
of
the
ga
requirement
right
right,
the
grpc
and
the
point-
and
that's
still
is
not
a
part
of
the
ga
plan.
So
then
you
want
to
separate
the
ga
you.
You
still
want
to
shoot
for
the
ga
and
leave
some
items
for
later
right.
A
D
A
D
D
I
mean
I
I
will
read
the
cap,
maybe
give
some
recommendations
but
like
if
we
added
a
new
cri
method
right.
How
would
we
can
gate
that
is
basically
the
same
basic
question
and
in
the
past
we've
just
added
the
method:
we've
never
gated
it
or
done
anything
special.
A
A
E
All
right,
so
the
next
one
is
pod
overhead.
So
sergey
you
are,
I'm
not
sure
you're
marked
as
the
reviewer
or
the
owner
for
this
one.
So.
B
I
graduated
runtime
class.
I
don't
think
there
is
a
big
rush
for
pod
ever
had
to
be
graduated.
B
E
E
E
E
All
right
windows,
privilege
containers
should
be
green
right.
F
I
think
yellow
so
I
have
alignment
with
lentau
about
or
from
sig
node,
I'm
still
waiting
for
an
api
review.
We
there
was
an
initial
one,
but
some
details
have
changed
since
some
reviews
from
lantau
and
also
sigoth.
I
would
like
to
take
a
look,
I'm
bringing
it
to
their
meeting
tomorrow
and
I'm
still
not
entirely
clear.
Does
that
mean
that
john
you'd,
like
youtube
to
just
be
the
approver
or.
F
E
All
right,
so
the
next
one
is
cubelet
credential
providers.
Andrew,
are
you
on
the
call.
N
Yes,
I'm
gonna
call
that
is
on
my
plate
and
yes,
hoping
to
get
this
to
data
for
121.
E
Okay,
direct
dawn,
any
concern
starts.
N
Fun
because
it
kind
of
overlaps
a
lot
with
dig
off
jordan,
the
git
was
the
one
who
approved
the
the
alpha
pr,
so
I'm
hoping
jordan
will
have
bandwidth
to
be
there
for
beta.
But
of
course,
if
don
donner
derek
can
also
help
review
that
be
great
confidence
level
wise,
I
think
I'd
put
a.
N
I
would
put
a
green
on
this.
The
follow-up
graduation
criterias
we
added
for
beta
are
are
mostly
just
like
caching,
optimizations
and
just
making
sure
you
know.
We
get
folks
testing
this
before
we
bump
this
data.
So
I
I
personally,
I
would
agree.
N
So
derrick
approved
the
the
initial
pr,
so
I'm
inclined
to
say
derek,
but
I
don't
know
yeah
I
mean
I'm
I'm
happy
to
finish
this
out.
We.
D
I
spent
a
lovely
afternoon
learning
about
things
I
didn't
know
we
had
when
he
first
presented
jordan
and
I
so
it
takes
time
to
educate
someone
else
on
on
the
unintended
the
unintended
treasures
the
cubelet
had.
So
I'm
happy.
N
F
I
can
talk
briefly
about
this
one,
so
this
one,
I
know,
is
coming
in
a
little
bit
late.
This
is
basically
an
enhancement
to
expose
both
journalctl
through
cubectl
and,
more
specifically,
add
functionality
into
the
cubelet
to
be
able
to
get
windows
node
logs
without
journalctl,
which
doesn't
exist
for
windows.
F
These
are
changes
that
have
been
in
openshift
for
quite
a
while
now,
and
some
folks
from
red
hat,
have
just
kind
of
finished
working
with
sig
cli,
to
get
some
to
kind
of
get
consensus
on
how
to
implement
that
in
cube
ctl,
if
possible,
it
would
be
good
to
it'd
be
nice
to
get
a
reviewer
from
signate
here.
I
think
that
most
of
the
work
is
in
cube,
ctl,
and
but
there
is
a
little
bit
of
work
in
the
cubelet
so
for
signal
to
look
at.
A
Oh,
I
I
think
this
is
the
features
useful.
So,
but
let
me
figure
out:
can
we
find
the
reviewer?
Unless
someone
stand
up
say
they
can
be
the
reviewer?
Otherwise
can
we
mark
this
is
right.
Is
that
okay
for
the
temporary
mag
right
until
we
figure
out
who
is
the
reviewer
as
soon
as
possible,.
A
F
F
We
wanted
to
get
yeah.
This
one
is
also
very
like
it
touches
a
lot
of
cigs,
so
I
think
we
got
six
cli
alignments
on
an
approach
and
with
that
also
sig
off
a
lot
of
the
off
kind
of
concerns
were
kind
of
coupled
with
six
cli
or
the
qctl
access.
But
there
are
other
sigs
that
we
want
to
just
be
aware
of,
and
also
have,
reviews
from.
E
A
E
A
M
B
Yeah
for
removal-
I
just
put
it
as
wide
right
now,
because
we
don't
do
anything
in
this
release.
I
have
a
document
in
agenda
explaining
the
current
status.
Some
works
that
we're
doing
and
proposal
to
shift
the
removal
of
the
equipment
from
entry
for
at
least
one
release.
So
maybe
I
mean
it's
only
five
minutes
left.
I
I
don't
know
whether
we
want
to
handle
it
now
or
we
can
discuss
our
next
meeting.
A
We
do
have
five
minutes,
but
do
we
have
some
most
of
things
just
carpool
call
out,
so
I
don't
need
to
go
through
that.
One
is
that
okay,
give
everyone
the
the
the
time.
So
the
first
one
is
the
dynamic.
You
know
the
sizing
for
kubernetes.
We
just
touch
base
a
little
bit
any
things
that
you
want
to
call
out
here.
K
Yeah,
so
this
was
modeled
after
the
cubelet
club
credentials
provider
where
we
were
talking
to
derek
and
we
wanted
to
be
able
to
cubelet
to
determine
the
system
reserve
and
cube
register
value
because
we
consistently
seeing
having
them
indirect,
inadequate
values
ends
up
the
nodes
being
frozen
so,
but
I
think
elena
has
some
comments
on
that.
Elena.
If
you
can
call,
you
want
to
highlight
them.
Oh.
C
Yeah
well,
my
concern
was
basically,
if
we're,
adding
an
endpoint
for
a
privileged
binary
to
like
specify
another
arbitrary
binary
to
exec
like
there
are
cons.
There
are
security
considerations,
particularly
when
all
we
really
want
is
like
a
tiny
bit
of
metadata.
So
my
suggestion
on
the
cap
is,
I
I
think
we
really
need
to
get
into
like
the
exploration
of
alternatives
with
like
why
not
you
know
have
like
just
a
web
endpoint
that
we
could
secure
and
like
post
the
values
to.
G
C
So
we're
talking
about
the
the
cubelet,
the
dynamic
sizing,
so
the
the
kep
as
written
suggests
adding
like
a
new
flag
to
the
cubelet,
so
that
the
cubelet
can
exec
a
binary
to
get
the
values,
and
my
suggestion
was
like
as
an
alternative.
Why
would
we
not
have
this
as
an
http
endpoint
because
of
the
security
considerations
of
having
the
cubelet
exec
and
arbitrary
binary.
D
C
A
I
think
that
we,
we
only
have
the
two
minutes,
and
can
we
carry
this
discussion
out
of
the
cab
and
I
I
I
agree
with
elena
and
we
may
have
some
authority.
We
figure
out
how
to
do
this.
One
so,
and
I
also
start
to
review
this
one.
So
we
just
discussed
this
one.
We
find
the
review
and
the
approval
and
the
police
will
come
to
meet
that
cap
and.
D
Just
just
to
be
maybe
a
devil's
advocate
a
little
bit.
I
I'm
going
to
wager
that
if
I
look
at
the
c
advisor
code
base
or
other
aspects
of
the
cubic
code
base,
there's
no
shortage
of
places
where
we're
executing
arbitrary
binaries
on
on
that
host,
I
mean
we're
calling
df
all
over
the
time,
so.
A
That's
not
that
that's
it
erica
that's,
but
we
also
have
to
understand.
Kubernetes
has
been
here
for
more
than
six
years
like
for
you
that
I
could
know
the
allocatable.
That
makes
the
note
the
perfect
capacity
I've
been
proposed
five
years
ago
right.
So
everyone
said
no,
a
lot
of
things.
D
Executing
a
binary
and
whether
or
not
it
should
be
an
http
endpoint
that
the
cube
execs
binaries
all
the
time
calling
du
and
df
and
other
things
on
on
the
host.
That.
A
Yeah
but
df
is
a
little
bit
different
from
arbitrary
of
the
battery
right,
so
that's
kind
of
you
may
be
end
of
the
kubernetes
to
run
something
but
the
other
hand
like
the
node
problem
detector.
We
also
allow
them
to
run
arbitrary
of
the
script.
Even
it's
discouraged,
because
we
think
about
that's
only
eventually
everything
it
is
rudeness
for
the
work
node
majority.
D
In
my
view
is
not
a
bad
thing
and
then
arbitrary
binary,
it's
up
to
us
to
to
narrow
the
arbitrariness
of
it
and
the
information
model
of
it.
But
I
I
don't
think
we
should
just
blankly
reject
things,
because
you
exactly.
A
So
we
find
out
the
time
so
can
other
type
owner
and
also
quad,
because
the
rest
are
for
related
to
type
owner
except
mpd
can
just
quickly
see.
If
you
have
some,
you
want
to
call
out
next
one.
It
is
the
sad
weather,
less.
G
Yeah,
hey
so
yeah
really
quickly.
There
have
been
a
couple
of
comments
on
the
cap.
It
wasn't
in
this
list
because
it
was
one
of
the
newer
ones,
but
I
I
would
like
to
target
it
at
121
and
then
I'll
work
with
alana
and
legroom
and
all
I'm
getting
it
on
that
list
and
judge.
G
I
would
probably
deem
it
to
be
read
at
this
point,
but
I
yeah
so
I'm
just
looking
for
feedback
like
lantau
said
that,
actually
three
years
ago
we
were
trying
to
deprecate
the
set
summary
api,
which
would
mean
that
all
of
this
work
would
be
for
naught.
So
I
think
I
I
think
it's
possible
for
us
to
come
up
with.
You
know
fish.
You
know
with
no
time
it's
not
really
a
good
time
now
to
discuss,
but
I'm
curious
for
feedback
on
the
cap.
A
Sure
so
can
you
please
update
that
document
for
planning
document
like
this
one
mark,
that's
right
and
david
david,
the
porter.
Can
you
be
the
one
of
the
reviewer
since
you
actually
look
at
those
sales,
weather,
refactory
and
all
those
kind
of
things
sure.
H
E
A
B
Yeah
I
mean
for
document
duplication,
maybe
a
little
bit
longer
discussion,
so
I
I
would
rather
responding
to
the
next
meetings.
A
So
I'm
so
I'm
sorry
for
today,
because
we
don't
have
much
enough
time
for
everyone
else.
We
go
through
the
one
that
21
cup
and
but
I
think,
that's
the
worst
the
time
to
go
over
since
a
lot
of
the
crap
being
touch-based
a
little
bit.
But
if
you
have
more
concern
and
more
question
and
or
you
want
more
time
to
discuss
that
in
detail
at
the
signal
community,
please
come
back
next
week
and
we
can
discuss
and
put
it
into
the
agenda.