►
From YouTube: Kubernetes SIG Node CI 20230816
Description
SIG Node CI weekly meeting. Agenda and notes: https://docs.google.com/document/d/1fb-ugvgdSVIkkuJ388_nhp2pBTy_4HEVg5848Xy7n5U/edit#heading=h.2v8vzknys4nk
GMT20230816-170353_Recording_1920x1200.mp4
A
Okay,
hello,
everyone-
this
is
signal.
Ci
today
is
August,
16,
2023.
and
yeah.
Let's
get
started,
okay,
so
first,
let's
go
on
the
agenda.
We
have
a
couple
of
PRS
from
thought.
Apparently
all
of
them
are
already
more
anything
else.
You
want
to
add
on
this,
but
yeah.
C
Yeah
I
see
interesting
discussion
on
busy
box
cherry
pick.
What
is
the
final
decisions
that
are
like?
Are
we
cherry
picking
previous
versions.
B
B
B
Yeah
I
think
it
should
be
okay
to
cherry
pick.
I
think
it
was
someone
from
red
hat
that
that
wanted
it
for
one
of
the
other
architectures.
A
C
Is
it
what
timer
fixed?
E
I
think
they
came
out
was
for
serial
tests
and
they
were
not
the
police
blocking
ones.
They
were
the.
C
B
F
A
This
one
I,
don't
know
why
it's
like
one,
but
it's
plastic.
Now:
okay,
after
counters,.
A
Closet
I
will
update
this
image
this
week,
so
Jesus
was
better
she's
not
released.
D
A
A
E
F
E
E
C
A
F
A
A
It
looks
yeah,
it
looks
good.
Those
celebrations,
I,
don't
know.
A
C
Previous
drop
down
into
our
runtime.
C
Yeah
yeah
this
one,
two
down
cool,
I,
think
right.
E
Also
recently,
there
are
the
changes
in
secret
security,
accounting,
so
actually
we'd
expect
a
little
bit
more
memory
because
there
wasn't
dating
any
content
correctly.
F
C
I
think
it's
a
good
action
item,
maybe
not
like
immediately
Mike,
if
you
can
put
it
on
agenda
to
check
what
performance
is
measured
because
it's
configured
for
P1,
C
group,
E1
or
C
group
video,
because,
ideally
we
need
to
start
switching
things
to
V2
and
have
V2
as
a
default
everywhere.
A
Yeah,
just
from
the
current.
A
D
A
C
B
E
B
A
On
this,
I
couldn't
find
any.
A
Couldn't
find
any
related
PR,
maybe
I
can
move
this
to
in
progress.
Oh
don't
just
move.
B
A
B
A
A
This
it's
just
bumping
the
images
of
few
things
and
and
yeah
I
think
that's
it.
A
C
No
move
it
to
Archive,
it's
typically
like
so
Global
changes
is
demographic
testing.
So
we
don't.
A
B
Yes,
this
was
a
failing
on
a
ec2
cluster.
Anyone
interested.
A
A
A
A
A
A
A
B
Yes,
that's
a
cryo
failure,
I
believe
yeah.
F
And
that's
the
pr
Wild.
A
A
A
G
There
was
a
very
old
issue
in
here
that
is
linked
and
this
issue
is
similar,
so
the
symptom
is
remote.
Volumes
of
the
Pod
are
becoming
local
Mount
point
after
rebooting.
The
nose
and
the
issue
is
triggered
by
this-
create
a
cluster
create
a
deployment
with
one
replica
shut
down
the
mode
you
delete.
The
forward
option
first
bring
back
the
node,
the
node
pod
is
running,
but
it's
actually
using
the
node
local
network.
G
There
was
it
was
mentioned
somewhere
that
the
issue
might
be
fixed
in
this
and
was
asked
to
retry
127
128,
but
apparently
there
is
still
that
issue.
That's
why
they
created
this
three-day
support.
C
Things
ignored
here
is
just
nominally
it's
mostly
handled
by
zip
storage.
G
C
We
will
Jason
Fran
and.
C
C
G
G
C
And
then
say
like
six
storage
kind
of
this
Dash
first.
C
Just
just
a
text.
G
Okay,
this
one
pods
in
image
full
back
off,
gets
stuck
there
if
image
fails
to
pull
for
long
enough,
so
they
have
a
part
that
is
stuck
in
imageable
back
off.
Initially,
this
was
caused
because
the
Pearl
Secret
was
not
existing.
We
collected
that
issue,
however,
by
the
time
it
was
corrected.
Much
time
has
passed
and
now,
even
though
the
secret
Secrets
exist,
the
image
is
still
stuck.
F
C
Universal
124,
which
is
other
support,
but
yeah
I,
don't
remember
any
changes
in
this
space.
I.
G
Think
somebody
mentioned
that
you
should
ask
the
support
at
the
manage
GK
first
and
then
log
it
in
Upstream
in
case
there
is
above.
G
How
can
we
reproduce
see
what
happened?
I
think
those
events
are
container
where
they
have
just
mentioned
that
reproduced
with
whatever
information
is
here
like
not
have
secrets
first
and
then
are
those.
C
I
think
there
are
three
pieces
like
if
it's
an
open
source
and
we
need
Reaper
on
a
latest
version
like
later
than
124,
and
that
video
I
didn't
need
good,
look
and
steps
to
your
produce.
I,
don't
feel
that
there
are
enough
steps
to
reproduce
yeah.
G
You
can
ask
for
more
information
here
and
ask
for
steps
to
reproduce
and
also
if
this
is
reproducible
in
later
later
versions.
Right
is
this
just
a
text
message
that
I
need
to
write
yeah.
G
G
The
next
one
is
propagated,
sit
down
signals
resulting
till
after
30
seconds,
although
the
termination
grace
period
is
120.,
they
have
a
card
with
processes
orchestrated
from
shells
liquid
system
into
the
quality
propagated
to
these
processes.
I
find
that
these
processes
occurred
in
30
seconds.
Although
the
grace
period
is
120,
the
board
itself
is
still
in
120
seconds.
C
And
wait
for
graceful
termination
period
to
finish
so
I
think
it's
expected.
G
E
Yeah
interesting
terminology
and
the
follow-up
like
they
say
that
the
process
shut
down,
but
then
the
Pod
is
not
killed.
So
I
think
that
they're
expecting,
like
some
life
cycle,
tie
to
the
rest
of
the
Pod
like
if
the
single
container
dies
and
the
rest
of
it
should
be
like
reported
or
shut
down
which,
like
maybe
they
I,
don't
know,
I'm,
not
sure
exactly
what
the
Pod
is.
A
All
right,
I've
seen
several
cases
where,
if
you
send
the,
if
you
send
the
secret
room
to
a
part
and
it
doesn't
reply,
then
they
kill
something,
but
sometimes
the
ball
is
not
cleaned
up
and
I
I,
don't
know
if
this
is
like
expected
Behavior
by
now,
but
yeah
everything
it
should
kind
of
expect.
The
pottery
clean
after
a
big
kill.
G
C
Look
at
the
zip
file
and
zip
file
only
contains
shell
files
running
inside
the
docker
image,
so
I
would
suggest
to
ask
for
post
Tech
as
a
information
that
we
need
and
after
Prospect
we
can
decide
what's
next
to
us,
but
yeah
I
agree
with
Mike.
After
we
have
a
port
spec,
we
need
to
understand
what
is
work
in
the
termination
of
God.
G
D
D
Potentially
yeah,
so
this
is
like
this
is
something
that
we
as
a
community
would
have
to
discuss
and
see.
If
that's
something
we
want
to
invest
resources
into,
but
I
I,
don't
think,
there's
any
immediate
action
item
from
our
point
of
view.
If
someone
wants
to
actually
pick
this
up,
that
would
make
sense.
But
maybe
we
can
just
say
like
it's
a
kind
feature.
D
C
C
Yeah,
remove
kind
back
and
I'm
like
and
for
policies
like
that.
Do
we
simply
go
through
care
process
or
we.
C
Yeah
and
do
remove
this
kind
bug.
G
G
Next
one
kubernetes
post
start,
who
doesn't
show
even
since
125,
so
we
upgraded
from
195
to
126
and
after
that,
every
life
cycle
group
doesn't
show
the
error
on
or
describe
for
older
versions,
which
show
the
error
basically
for
the
new
versions.
The
same
code
shows
as
this
expecting
as
older
versions.
How
can
we
reproduce
this?
Okay.
G
G
I'm
not
sure
so
I
think
like
the
gist
of
this
is
the
behavior
change.
So
we
were
showing
errors
before,
but
not
now.
C
Yeah
I
I,
don't
these
are
part
of
some
PR.
Can
you
look
at
PR
description?
Oh,
it's
just
normalizing
life
cycles.
Okay,
so
I
think
what
happened?
Is
you
failed
to
include
message
inside
the
let.
C
And
we
can
do
Post
start
hook,
failed
semicolon
and
older
message
like
previous
models.
E
B
D
E
G
C
I
think
we
can
close
this
expectedly.
C
G
I
looked
at
this,
they
were
mentioning
that
the
parameter
is
not
set
why
the
config
file
specified.
But
then
the
last
comment
says
that
it
is
working
as
expected.
I
think
are
not
reproducible.
B
G
G
G
E
G
It
looks
like
like
they
mentioned
here:
Cube
proxy
should
be
part
of
Readiness
check
is
what
they
are
set
lasting.
C
You're
making
this
change
as
a
feature
request.
I
think
this
may
be
a
duplicate
of
some
older
box
and.
G
G
Okay,
I
think:
what
does
what
Antonio
is
suggesting
here
that
if
they
make
you
proxy
as
a
part
of
node
Readiness
check,
then
it
would
involve
a
lot
of
checks.
I
think
it
would
require
all
the
static
ports
as
a
part
of
mode
Readiness
check.
This
would
solve
scheduling
problem,
but
it
will
impact
mode.
Startup,
readiness.
C
So
the
issue
with
static
ports
and
other
ports
message
is
well-known
problems
that
we
have
for
many
years.
So
I
would
say
just
accept
this
box.
We
may
need
to
do
duplicate
it
towards
some
other
box,
but
I
think
at
this
stage
it
may
not
be
on
a
bucket,
maybe
some
feature
request
to
change,
how
we
treat
Northern
regions,
conditions.
C
Yeah,
but
it
doesn't
fix
anything
completely
right
so
right
and
it
really
depends
on
how
long
it
takes
cook
proxy
or
anything
Network
related
to
Startup
yeah.
G
I've
added
an
actionable
items
where
I
can
deruplicate
this
to
another
two
other
bugs
that
are
related.
B
G
G
It
returns.
Failure
under
some
condition
was
met
once
the
condition
is
met.
It
returns
success
when
upgrading
to
123.6
the
failure.
Reason
of
message
returned
from
the
soft
admit,
Handler
seem
to
jump
between
four
phases
that
is
transition
from
pending
to
running,
despite
this
of
admit,
Handler.
Subsequently,
returning
success.
G
C
So
many
things
around
with
this
bug,
I
think
123
is
long
out
of
support
the
first
message
and
then
we
need
to
understand
what
the
soft
admit
handles
mean
because
we
don't
allow
any
plugability
there.
C
Thank
you
very
much
Mike
and
Dixie
for
driving
this
season
that
have
a
good
rest
of
the
day.
Bye
thanks.