►
From YouTube: Kubernetes SIG Node CI 20230809
Description
SIG Node CI weekly meeting. Agenda and notes: https://docs.google.com/document/d/1fb-ugvgdSVIkkuJ388_nhp2pBTy_4HEVg5848Xy7n5U/edit#heading=h.2v8vzknys4nk
GMT20230809-170507_Recording_2560x1322.mp4
B
Okay,
hello,
everyone-
this
is
signal.
Ti
meeting
today
is
August
9
2023
yeah,
let's
get
started,
let's
see
so,
let's
say
we
have
agenda
items.
A
Yeah
I
don't
see
swatchy
I
think
yesterday,
I
was
trying
to
reopen
this
PR,
but
maybe
I
failed
I.
A
B
A
C
A
Okay,
I
guess:
we
need
to
move
again
on
multinoma,
so
people
want
to
do
some
type
of
OG
like
Resource
Management
changes
and
Numa
tests
are
really
available
before,
as
you
may
remember,
most
of
our
changes
were
tested
on
some
Red
Hot
environment
first
and
Upstream.
We
just
taken
test
results
as
granted.
I
hope
we
can
start
testing
upstream
and
get
more
contributions
across
the
different
companies.
B
Yeah
I
agree
on
that.
Okay,
let's
go
back
to
the
next
one,
then
and
second
agenda
item
tell
you
in
a
call:
yeah,
okay,.
D
It's
really,
it
was
just
sort
of
like
how
to
handle
this
issue
document.
There
ends
up
being
like
two
different
ways
to
prevent
this,
and
it's
just
more
of
like
yeah.
If
you
sort
of
let
pods
create
Federal
volumes.
C
E
D
D
In
general,
yeah,
I
think
General,
but
for
these
memory
backgrounds
that
you
can't
put
a
size
on
them
on
there,
if
you
do
it's
sort
of
it
prevents
this
from
occurring,
and
if
you
a
supply
memory
limit
on
your
pod
or
your
container,
will
also
it
prevent
it
from
occurring.
B
D
And
so
I'm
saying
there's
some
scenario
where,
basically
you
consider
memory,
you
know
faster
than
respond
I'm
assuming
it's
not
that
he
could
somehow
walk
his
note
up
entirely
and
I
tried
it
several
times.
It
was
never
able
to
reproduce
across
a
bunch
of
nodes.
B
B
Yeah
I
have
no
other
input
on
this.
Besides
that,
do
you
have
any
recommendations
here?
Oh
I,
I,
don't
I,
don't
know.
If
there's
anything
else,
we
can.
A
A
Will
make
our
recommendation
even
stronger
here
and
second
is
I
wonder
if
we
have
any
way
to
detect
the
situation?
Is
it
is
it
detectable
because
it
was
detectable
and
we
have
a
like
a
way
to
set
an
image
and
it's
not
an
issue.
If
we
have
problem
with
either
of
them,
it
may
be
a
feature
request
to
improve.
B
Do
you
have
the
cap
for,
for
the
empty
dealer
limit
like
I,
can
maybe
put
that
here.
B
B
D
B
B
C
B
C
B
A
Stage
of
release,
I
I
would
really
want
to
concentrate
on
failing
tests,
not
hiking
tests.
A
I
mean
at
this
stage
of
release,
we
probably
wouldn't
fix
any
flakes
yeah.
B
C
C
B
Don't
know
if
this
was
a
flick
because
it
was
the
last
run.
We
can
wait
for
the
next
one,
so
this
is
green
in
a
bunch
of
flicks
for
cubelet
cereal
containerdy.
B
B
B
Then
all
right,
the
other
I,
have
changed
something:
okay,
oh
yeah,
they're
right,
that's
good
enough!
I,
guess,
conformance
okay!
This
is
conformance
server.
Green,
we'll
need
to
create
a
127
conformance
Branch
afterwards,
I
can
I
can
work
on
that.
After
one
thing,
I
discovered.
A
B
Okay
yeah:
this
is
legendary
I'll
work
at
this.
After
the
meeting,
let's
equivalent
potential
provider,
we
have
failed,
but
lutrin
green.
B
Oh
you're
right:
we
could
just
go
there.
Okay,
we
have
this
one.
A
For
a
while
right,
yeah
smart
effects,
as
if
you're
trying
out.
B
B
Yeah,
let
me
let
me
see
I'll,
do
a
quick
search
in
in
testing
for
it
to
see.
If
we
find
anything
about
the
swap.
E
E
E
C
B
E
Okay,
oh
goodness,
we're
getting.
There
flock
test
totally
yeah,
so
this
one
I
actually
don't
know
what
the
state
of
the
tests
are
here.
I
know
we
had
mentioned
that
it
like
could
kind
of
have
to
do
with
the
swap
enhancement.
I,
don't
think
we
explicitly
did
anything.
Anything
was
explicitly
done
for
that.
A
Tests
fail
to
start
and
I
believe
the
suggestion
was
to
take
a
look
at
maybe
some
bootstrapping
configuration
stuff
like
that.
So
it's
not
like
test
are
failing.
It's
not
fair.
The
table
can
start.
B
C
E
E
E
Well,
for
the
I'm,
seeing
the
ones
for
cryo
or
also
they're,
still
like
they're,
sometimes
coming
up
and
then
sometimes
not,
which
is
strange
so
still
an
issue
with
teams.
So
I
think
the
issue
has
kind
of
changed
a
little
bit.
I
think
it
needs
some
needs,
some
more
love,
so
I
I.
We
can
keep
it
assigned
to
me.
I'll
try
after
this
meeting
to
take
a
look
and
update
the
issue
and.
A
B
A
B
C
C
C
B
A
And
this
one.
C
C
B
It
looks
fine
to
me
I
think
all
of
these
are
there
is
nothing
critical
about
it.
Besides
the
swap
test,
probably
we
can
skip
those
I
know
what's
going
on
with
this
one,
so
I'll
just
keep
it
for
now.
B
Okay,
yeah,
it
looks
like
the
only
the
only
critical
you
have
is
the
Swap
and
and
I'll
create
the
release
Branch
jobs
from
127
128.,
okay,
yeah.
With
that
we
have
finished
test
triage
yeah,
let's
go
Peter.
Do
you
want
to
take
over
Buck
trash.
E
Yeah
I
need
to
log
out
and
log
back
in
so
give
me
like
one
minute.
D
E
E
C
E
Okay,
let's
share
the
screen.
E
Okay,
let's
see
the
background
of
my
desktop
and
now
I'm
pulling
up
okay,
so
pulling
up
the
node
board
with
the
archery
Ash
and
you
all
can
see
that.
B
Yeah
yeah,
if,
if
it,
if
anybody
else
wants
to
to
drive
a
bug
triage,
if
not
I'm,
happy
to
drive
it
as
well,.
E
B
No
worries
yeah.
Anyone
wants
to
drive
back
to
Yash.
C
C
C
B
B
They're
using
the
particular
AWS
server
form
template,
but
this
looks
like
more
of
an
issue
in
I.
Don't
know
if
this
is
something
more
like
related
to
terraform
binary
itself.
A
And
clearly,
we
need
to
apply
news
information
tag,
the
only
question
what
to
ask-
and
there
are
too
many
questions-
comes
to
mind.
C
B
Is
this
a
lot,
is
this,
it
doesn't
feel.
Like
knows
what
should
the
limited
knows.
A
If
it's
memory
baked
empty
series
may
be
related,
okay,.
B
A
C
C
B
Okay,
let's
go
with
the
first
one:
empty
Theory
directory
is
worth
not
cleaned
after
Paul
deleted.
Okay,
no
problem
with
the
interior
volumes
is
deleted,
just
after
it's
great
it's
possible
that
they
need
containers
to
the
right
content.
E
See
yeah
I
wonder
if,
like
run
time,
didn't
unmount
that
ball
that
or
something
like
that
and
another
sometime
they're
using,
but
that
might
be
relevant
as
well,
but
I
agree.
We
should
check
those
anywhere
kubernetes
for
trying
that
hard
at
it.
B
So
yeah,
let's
ask
for
that.
It's
kind
of
hard
to.
C
A
Yeah,
can
you
check
on
time
runtime.
C
A
It
wouldn't
be
removed,
it
will
stay
as
needs
information,
okay,.
B
B
Okay,
next
one,
why
is
my
bots
that
container
stats
logs
use
white,
continuous
right,
119?
Okay,
that's
an
old
version.
B
B
B
B
B
Okay,
but
this
looks
like
a
regular,
it
doesn't
looks
like
a
like
an
EK,
a
specific
problem.
Okay,
so
you
have
the
thing
not
ready,
then
you
create
a
path.
The
part
should
not
be.
A
So
position
who
controller
manager
that
doesn't
update
things
fast
enough.
B
B
What's
is
there
any
mitigation
in
play?
I
guess
if
you
schedule
the
poll
Again
by
then
I'm
expecting
key
controller
manager
to
now
know
that
the
tent
is
not
in
place.
C
B
This
should
be
an
accepted,
but
I
want
to
know
European
or
not.
E
I
also
wonder
if
the
the
Pod
life
cycle
rewrites
will
fix
this
I.
Don't
they
haven't
gotten
in
yet,
but
or
some
of
them
haven't,
but
yeah
checking
on
a
newer
version
will
verify
because.
B
B
This
logic
read,
write
was
done
in
125
right.
There.
E
There
was
one
set
that
was
done
in
145,
but
there's
like
more
that's
happening
that
I,
don't
think,
has
merged
yet
I,
don't
know
if
David
is
on
the
call,
but
I
think
he's
significant,
but
he's
closer
to
it,
but
I
think
that
there's
a
specifically
related
to
like
the
cube
of
reporting,
the
accurate
pod
life
cycle
to
all
of
the
entities
that
need
it
like
which
the
tape
manager
would
so
like
they
would
affect
the
hot
eviction.
In
that
case,
okay,.
B
C
B
B
B
B
Thank
you
right.
Let's
see
you
next
one
wrong
container
memory
working
set
bytes
for
the
end,
I,
don't
know
what
this
is.
B
B
All
right
there's
an
issue
for
that
I'll
take
a
quick
look
into
it.
B
C
B
B
B
B
E
C
C
E
Totally
did
they
say
you
know
if
they
said
what
platform
they're
on
if
they're
on,
like
you
know,
a
cloud
provider
or
like
if
they're,
building
an
in-shop.
B
A
Mike,
it's
depends
on
network
from
kubert.
I
was
answering
the
question
Peter's
question
about.
E
E
E
C
E
Francesca,
do
you
think
I
mean
you're
a
little
closer
to
VRA?
Do
you
think
that
this
is
expected
or
known
about.
F
Hey
at
glance,
I
would
say
totally
no,
but
this
is
99
of
my
gut
feeling.
I
need
to
Deep
dive
to
see
if
there
is
a
legit
case
for
this
won't
be
code,
but
it
seems
really
considering
the
issue
and
everything
I
would
say
it
should
be.
Could.
B
Okay,
they
want
to
take
this
Francesco.
A
B
Organized
policy
here
I
cannot
edit
this
okay
I.
B
Oh
nice,
nice
feature:
okay,
no
matter
the
state
of
the
container
or
the
results.
You
should
be
restarted.
E
Place
resizing
then
I
actually
wouldn't
expect
to
restart,
but
I
would
yeah
I
guess
the
qubit
state
machine
is
probably
not
the
two,
the
like
resizing
piece
and
the
Crash
Loop
piece
or
not
talking
to
each
other
so
like
I'm,
guessing
what
they
want
if
they
want
the
the
crash
Loop
timeout
to
reset.
If
a
resize
comes
in.
E
C
D
E
Yeah
I
don't
know
if
vinay
is
still
working
on
this
I
could
see.
Would
you
I
can
see
if
our
teammate
Joel
can
look
at
it?
He's
been
looking
at
some
CPA
things
so
well,
you
can
assign
me
and
I
can
triage
it
to
him
if
it
feels
like
it's
right.
E
E
B
It's
a
rolling
content:
okay
yeah!
This
is
an
alpha
I'm
guessing.
It
makes
sense
that
it
was
an
edge
case.
We
didn't,
we
might
have
not
catched
all
right.
Let's
see,
we
have
two,
no,
never
mind
we're
done.
B
Okay,
we
have
two
minutes
for
and
in
the
meeting,
I
think
we
can
call
it
for
now.