►
From YouTube: Kubernetes SIG Node CI 20230823
Description
SIG Node CI weekly meeting. Agenda and notes: https://docs.google.com/document/d/1fb-ugvgdSVIkkuJ388_nhp2pBTy_4HEVg5848Xy7n5U/edit#heading=h.2v8vzknys4nk
GMT20230823-170558_Recording_1920x1034.mp4
A
B
Hello,
everyone
today
is
August
23
2023.
This
is
signal
CI
and,
let's
see
percent
in
the
screen.
B
C
Well,
I
think
I
wanted
to
mention
just
briefly
I,
don't
know
if
everybody
is
watching.
This
unfolding
thought
is
trying
to
understand
the
root
cause
of
some
ports,
not
reports
in
killed
when
they
were
unkilled
and
I.
Think
we
get
down
to
continue
the
issue
and
contain
energy.
Has
a
nice
Foundation
of
region
kill
message
from
C
group
so
like
right
now
we
I
think
we
know
the
root
cause
already,
so
sometimes
kill
container
will
just
be
reported
as
failed,
but
we
don't
know
how
to
fix
it.
C
So
if
you're
interested
there
is
a
bug
track
in
that
reach
out
to
me,
if
you
wouldn't
find
it.
B
Thank
you,
yeah
I
I
know
about
this
issue
between
surreal
pain.
It's
good
to
know
that
somebody's
investigating
it
thanks.
C
Yeah,
it's
very
critical
for
job
success.
For
instance,
like
jobs
like
there
is
a
improv
like
caps
that
make
job.
That
is
start
based
on
conditions
and
typically
people
want
to
do
different
things,
whether
it's
killed
or
failed,
and
we,
if
you
don't
report
killed
correctly,
then
this
jobs
may
not
restart
if
they
wanted
to
be
restarted.
So
yeah,
that's
a
problem
right.
A
B
E
C
C
So
in
yaml
definition
you
can
do
decorate
through
and
I
think.
Apparently
it
will
like
run
jobs
in
a
little
bit
differently
and
so
we'll
have
more
logs
collected
from
this
job,
so
it'll
be
easier
to
troubleshoot,
but
I'm
not
sure
we
can
experiment
Francesca.
If
you
want.
E
Yeah
I
we
need
to
because
in
this
case
the
exact
login
which
is
being
shown
there.
It
seems
to
be
this.
There
is
a
CNN
issue,
but
other
than
that,
so
yeah
I
will
totally
try
that.
F
Only
thing
I
thought
the
issue
was
with
the
job
configuration
and
maybe
decorate
through,
could
probably
help
us
troubleshoot.
This
I
tried
to
compare
this
with
already
running
jobs
as
well
like
the
configuration
of
those
with
the
running
job.
But
like
again,
it's
exactly
the
same.
So
it's
I'm
happy
that
actually
Francesco
is
taking
a
look
at
this.
F
F
There
is
an
issue
for
CP
management,
great.
B
Okay,
so
this
failing,
let's
continue
with
the
other
test
Fedora
the
last
time
it
was
running.
It
was
in
February.
B
Ubuntu
is
clean.
Now,
no
I
don't
want
to
serial
failing
on
home
killer.
B
B
Different
tests,
in
each
case,
home
killer,
is
one
common
culprit
here
pretty
much
both
these
will
have
the
ones
that
I've
been
seeing
failing
in
several
test
grids.
D
B
I'll,
probably
just
go
to
the
failing
ones:
2012
fitters.
F
C
B
G
Yeah
this
one
I
was
looking
into
it
morning.
I
created
a
issue
for
this
test
to
to
follow
up
with
this
test.
Do
we
have
an
invented
plague,
no
d2e
for
container
d
as
well,
this
test
just
to
be
quite
this
job
rather
used
to
be
quite
stable,
but
lately
it's
flaking,
so
I'm
just
concerned
whether
it's
a
cubelet
or
cryo,
so
I
wasn't
able
to
find
an
equivalent
job
in
container
d,
but
yeah.
This
there's
an
issue
tracking
this
the
failed
job
that
you
see
for
event,
triplet.
A
B
G
Mike
sure
you
are
just
showing
the
cryo
cubelete
serial
job.
That
was
failing.
Oh
yeah,
let
me
go
back
okay,
yeah,
and
could
you
compare
the
failures
with
the
continuity
serial
job
I
think
that
was
fail
in
terms
responding
to
see
if
we
are
failing
for
the
same
this,
let.
B
G
With
we
just
decided.
A
B
B
B
Pushes
looks
fine
yeah,
it
looks
like
all
the
the
current
issues
are
known
issues
and
some
people
are
working
already
on
them.
Yeah
I
think
we
can
finish
with
this
reaction.
Photo.
C
Play
before
I
switch,
can
you
go
to
GitHub
dashboard
yeah.
B
C
No
I
meant
that
test
board.
C
No
project:
what
is
it,
let
me
put
it
in
a
chat:
okay,.
B
B
B
Okay,
let's
see
first,
one
support
checking
integrates,
enables.
B
Yeah
they're,
basically
adding
a
flight
that
allows
you
to
check
if,
from
from
the
test
itself,
whether
or
not
a
feature
gets
enabled
does.
C
D
A
D
E
Is
meant
to
be
closed
as
soon
as
I
figure
out?
What's
going
on,
okay,
yeah,
yeah
I
agree
thanks.
C
D
B
A
B
Okay,
oh
looks
like
we're
clearing
the
triage.
A
circumventional
will
probably
go
through
it
to
do
really
quick
to
see
how
things
are
and
if
we
need
to
unassign
something
we'll
do
that.
B
Okay,
another
one
cleanup
configuration
and
rename
changing
directory.
B
Oh,
never
mind.
Daniel
is
the
sign
of
this.
If
she's
still
working
on
this,
so
I'm
thinking,
I'm
going
to
unassign.
D
B
Okay
umbrella
show
rmci,
if
you're
not
into
in
testing.
This
was
creating
six.
B
C
B
Oh
did
I
recall
this
one,
so
this
is
for
the
cause
of
jobs.
I,
don't
think
I'm
gonna
work
on
this
anytime
soon.
B
We
don't
have
a
okay,
I
guess:
I'll,
just
keep
it
into
no
point
in.
B
Looks
like
we
don't
have
any
Honor
on
this
well
I'll!
Keep
it.
B
This
one
doesn't
have
any
other,
actually
I,
think
I.
Think
I'll
just
go
through
the
ones
that
don't
have
an
owner
that
have
an
owner.
G
There
was
a
there
was
a
particularly
just
scrolled
past.
It
says,
make
all
please
submit
jobs,
make
all
pre-submit
jobs
run
on
some
cluster
I
forgot,
it's
I.
B
B
Yeah
the
different
special.
Let
me
remove
the
owner.
G
B
Right
yeah
I
think
I
saw
that
now
in
a
couple
of
failures
as
well
yeah,
let's
hope
somebody
it,
let's
see
in
baseball
vertical
scaling.
This
is
scientific
name.
B
B
B
D
C
B
Pure
unicode,
yes
I'm
here,
so
do
you
plan
to
still
work
on
this?
It's
not
like
an
assign,
so
somebody
else.
D
G
If
barcode
is
not
available,
Maybe
the
person
he
sees
to
item
Mark
I
hope
one
zero.
One.
B
Okay,
crushing
test
get
container
arms
or
just.
B
B
C
C
B
Yeah,
this
is
a
known
issue
related
to
some
of
the
info.
I'll
have
to
think,
probably
with
van
dealer,
to
get
the
proper
permissions
on
the
on
the
on
the
project
for
now,
I'll
just
actually
just
change
the
triage.
B
Let's
see,
coordinate
piece,
I
have
perfect
support,
template
hash
level.
A
D
B
A
So
I
think
I've
come
to
the
conclusion
on
this
one.
It's
a
container
D
issue,
okay,
I
think
that's
the
case.
A
B
Probably
yeah,
probably
one
thing
involved
with
English
the
company
or
the
community
to
get
this
I'll
leave
it
as
it
is
for
now.
Okay,
yeah
I
think
we
went
through
the
whole.
The
whole
board
I'm
gonna,
go
now
to
the
signal
PR,
which
should
be
okay
right
now,
I,
don't
know.
B
Okay,
no,
not
authorization
enhancement,
use.
Cubelet's
client
can
only
view
resources
along
things
on
earth.
B
I
don't
know
if
this
is
something
we
need
to
I
mean
I.
Guess
we
need
to
review,
because
no.
A
A
PR
to
like
I
think
during
the
wrong
time:
yeah
we
need
like
a
cap
for
this
moment.
Okay
right.
B
B
B
D
E
G
G
Destroy
say,
group
parts
we
could
demand
more
logs
in
a
sense.
Okay,.
B
G
B
A
B
I'm
not
sure
if
they're
using
like
0p1
or
we
do
the
lock
and
those
three
case.
But
honestly
I,
don't
I
don't
drink
as
much.
D
Maybe
I
mean
yeah
that
that
could
help,
but
I
think
it
is
worth
asking
what
their
system
reserved
is,
because
it
might
be
the
case
at
systemd
that
doesn't
have
enough
CPU
to
be
able
to
do
all
the
stuff
that
they're
asking
it
to.
B
D
B
Okay,
we'll
wait
for
a
commands
background
job
to
finish
when
it's
at
the
topology
containers.
B
C
B
B
E
What
I'm,
not
sure
about
and
I
literally
mean
I
need
to
review
the
documentation
myself
is
that
it
is
supposed
to
behave
like
this
like,
for
example,
it
seems
to
be
that
the
user
want
to
be
everywhere,
which
seems
to
be
like
SSH,
but
it
is
supposed
to
be
like
it
is
claimed
to
be
in
the
first
place,
or
it
was
like
you
know,
whatever
Docker
did
that
leaked
across
the
layering
I
I.
B
G
B
B
Because
yeah
I
don't
know
if
at
least
from
that
I
don't
have
any
conflicts
on
this.
B
A
B
Who
owns
this
this.
A
B
It
has,
as
we
need
to
mentioned.
G
The
details
with
the
error
last
comment
where
it
says,
though,
no
the
in.
B
Okay
looks
like
that's
it
for
foreign.
We
have
four
minutes,
I
think
we
can
finish
now
if
there
is
no
further
items,
but
yeah
I
think
with
a
computer
job
today,
we'll
revisit
that
other
issue
next
week
for
another
one
for
them,
it
was
like,
first
of
all,
but
yeah,
okay
have
a
great
week.
Everyone
thanks
thanks
for
joining.
Thank.