►
From YouTube: Kubernetes SIG Node CI 20221214
Description
SIG Node CI weekly meeting. Agenda and notes: https://docs.google.com/document/d/1fb-ugvgdSVIkkuJ388_nhp2pBTy_4HEVg5848Xy7n5U/edit#heading=h.2v8vzknys4nk
GMT20221214-180629_Recording_1736x1120
A
Hello,
hello,
it's
a
signal.
Ci
weekly
meeting
today
is
Wednesday
December
14th
2022..
Welcome
to
the
last
meeting
of
the
year
for
CI
group,
we
have
a
very
short
agenda
as
I
observe
most
people
trying
to
wrap
up
things
for
the
year.
Close
aprs
do
some
cleanups.
A
B
A
Okay,
I
see
a
couple
items
here
before
you
go
there,
Swatch
I
I
know
you
created
this
GitHub
issue
and
I
just
replied
there.
Do
you
want
to
give
small
update.
C
Yeah
I
think,
since
you
last
pointed
out
that
we
have
these
two
machines,
I
looked
into
AWS
Alternatives
as
well
and
they
seemed
expensive,
so
I
think
the
best
option
might
be
so
far
to
go
with
the
one
you
proposed
into
the
standard.
32
Dooms
was
helping
me
out
a
bit
with
equinix
systems
and
there's
someone
from
equinix
looking
into
it
as
well.
C
But
I
haven't,
got
any
update
so
so
far,
I
think
and
to
the
standard.
32
is
the
way
to
go.
I
guess
but
I
haven't
got
any
kind
of
Acts
from
West
infra
group
per
se.
A
I
think
what
they
mentioned
is
they
leave
it
for
us
to
control,
but
they
will
come
back
if
you
will
be
too
expensive,
I
guess:
okay,.
C
C
Okay,
so
like
do
you
think
it
would
be
appropriate
to
maybe
create
a
PR
to
update
the
jobs
for
resource
managers
to
run
on
the
system.
A
Yeah:
let's
try
that
let's
try
to
do
the
minimal
change
and
see
how
far
this
jobs
run
and
if
it's
reasonable,
we
can
just
tell
those
are
numbers
and
we
need
to
test
the
system.
So
if
numbers
are
not
like,
if
they're
not
satisfied
with
that,
we
can
run
this
jobs
a
little
bit
rarer
than
other
jobs.
Maybe
once
in
three
days
or
once
in
a
week,
something
like
that.
C
D
About
running
the
tests
on
actual
normal
machines,
I
was
thinking
I'm,
mostly
thinking
aloud.
Now,
since
we
moved
to
to
Ginkgo
B2,
we
can
start
adding.
Maybe
labels
Ginkgo
labels
more
aggressively
and
label
tests,
which
really
makes
sense
on
multinoma,
which
is
pretty
minimum
and
safe
change,
because
you
literally
just
did
an
annotation
on
the
tests
and
then
it's
up
to
us
reviewers
to
make
sure
this
list
is
reasonable
and
not
growing
I
mean
unbounded,
and
then
we
will
have
a
very
easy
way
to
run.
Just
the
test
with
require
actually
required.
A
I
would
suggest
to
start
with
just
trying
out
like
that
works
and
a
job
is
not
being
skipped,
and
then
we
can
adjust
pricing
for
that
by
doing
this
extra
labeling.
What
do
you
think.
D
The
reason
why
I'm
mentioning
labeling
right
now
is
because
I
I
I
believe
the
shop
it
will
make
this
kind
of
benchmarking
easier,
but
it's
actually
not
needed.
I
mean
if
it
helps.
Let's
I
can
help
with
that.
Let's
do
that.
Otherwise,
yes,
we
can
totally
start
with
existing
code
with
no
changes
at
all.
C
Yeah
and
I
think,
like
my
initial
attempt,
would
be
to
just
focus
on
the
jobs
that
related
to
Resource
managers
and
then
I
think
the
next
pass
would
be
if
there
are
the
things
that
require
such
systems,
and
we
we
add
labels
to
specify
that.
A
Perfect
and
even
easier
to
like
have
Ginkgo
Focus
select
specific
test
by
name
yeah.
A
There
were
taking
notes:
okay,
let's
move
into
PRS.
E
E
A
B
B
A
A
B
E
A
B
A
E
A
Okay,
are
we
done
with
that?
Let's
try.
I
clean
up
needs
approver.
Earlier
today
waiting
on
author.
It
is
a
failing
test.
A
Some
bog,
I-
don't
think
it's
belonged
to
here.
I
will
clean
it
up
work
in
progress,
something
big
work
in
progress.
Work
in
progress
failing
test
needs.
The
base,
contextual
working
login-
is
not
related
to
us.
Okay,
it
feels
that
the
original
author
is
clean.
Anybody
wants
to
discuss
what's
assigned
to
them.
A
Okay,
we'll
need
to
clean
up
that.
So
maybe
we
can
take
a
look
really
quick
since
it's
end
of
the
year
and
the
cleanup
is
generally
encouraged.
So
first
Mike,
like
you,
have
this
I
I
know
you
had
a
PR
you're
on
the
call
right.
F
I
think
this
is
yeah,
so
the
issue
well
I
I
sent
a
PR
for
this
I
think
it
should
be
emerging
right
now.
F
We
realized
that
even
to
all
the
cost
stops
are
failing,
we
don't
have
duplicate
jobs
for
most
of
them,
so
it's
probably
better
to
keep
them
in
otherwise,
we'll
have
a
signal
for
for
those
jokes
at
all
the
one
that
I'm
deleting
yeah
that
one
should
be
emerged
process.
Okay,.
F
F
I'm
not
sure
about
what
the
failure
about,
but
this
if
the
failure
is
related
to
cost,
we
could
try,
updating
or
create
or
using
new
images.
A
Yeah
yeah,
when
I
looked
at
some
of
them
say
well,
weren't
cost
related
like
soak.
For
instance,
it
was
related
to
actually
test
being
running
for
a
long
time,
I'm
not
sure.
What's
what's
failing
now
in
this
test?
A
B
A
A
Okay,
I
think
yeah
I,
don't
see
anything
criminals
that
needs
to
be
resolved.
B
E
G
Okay,
so
I
guess
sorry
I
was
on
mute.
The
the
issue
that
you
closed
earlier
was
it
under
cryo,
tab,
yep
and
the
job
he
said
mood.
Did
we
check
in
the
cryo
tab
in
yeah?
This
is
on.
Oh,
it's.
Okay,.
A
Yeah
just
click
on
this
and
it
didn't
get
me
inside
the
job.
So
maybe
it's
because
of
GMC
names,
I
think
it's
called
pool
now.
A
I
mean
jobs
are
not
here
any
longer.
I
don't
want
to
keep
the
bug
around
when
it
points
to
unexisting
jobs.
Do.
G
A
Yeah
I
mean
we
can
reopen
bug
again,
make
sense
thanks,
but
thank
you
for
checking.
E
A
I
think
we
have
a
soak
test
now
so
I,
let
me
Mike,
you
were
looking
at
the
stuff
with
soak
test.
Do
you
want
just
to
be
assigned
to
you
just
to
double
check,
yeah.
F
You
can
you
can
see
him
either
just
to
make
sure
that
I
didn't
find
a
soap
test,
but
it
wasn't
using
containerdy.
A
B
A
Injection
I:
will
anybody
interested
in
that
you
can
take
it
over
for
me,
but
if
not
I
will
be
I
will
be
looking
into
that
seems
like
I
I
vaguely,
remember,
assign
it
to
myself
and
thinking
that
it's
exciting
thing
to
clean
up,
but
I
never
get
to
that.
A
Okay,
this
is
still
a
problem,
so
we
have
a
skip
logic
in
this
test
and
this
keep
logic
is
not
very
efficient,
so
it
skips
too
much
of
a
test.
So
you
have
very
little
coverage
for
a
runtime
handy
right
now.
I
will
try
to
see
you
for
David
gonna
work
on
that.
A
Okay,
I,
don't
know
the
status
of
it.
Maybe
I
will
keep
it
around.
A
Nobody
else
cool
yeah.
If
you
want
an
item,
just
go
here
and
try
to
find
something.
There
are
many
unassigned
items
and
even
if
it's
assigned
and
still
interesting
to
you,
you
can
try
to
take
it
over.
A
Okay,
if
you
will
fail
to
find
something,
let
me
know.
B
A
Let's
move
forward
we'll
go
to
back
trash
right
now.
If
you
hear
just
for
test,
you
can
drop
off
no
hard
feeling
at
all.
We
will
go
into
box
Dash
right
now,.
E
A
A
Here
in
CI,
I
think
issues
to
do
triage
column
is
empty
right
now,
so
you
can
take
issues
to
do.
Okay,.
A
H
You
know
I
updated
the
documentation
for
this
I
think
it
was
a
yeah
I
can
Brian
R
Phillips
I'll,
add
a
link
to
it.
H
A
B
B
A
A
A
B
B
A
So
the
the
linked
box
is,
we
looked
at
the
last
triage.
It
says
that
some
OS
has
a
issue
with
how
they
created
memory
stats,
so
I
don't
think
we
need
to
solve
this
problem,
but
I
will
be
surprised
this.
How
much
if
Amazon
has
the
same
problem.
E
E
A
Interesting
one,
that's
interesting,
because
I
believe
many
sidecars
are
doing
that.
H
The
I
think
this
is
a
valid
test
case,
though
it's
not
mounting
in
the
the
host
host
file.
Oh,
it's
using
the
containers
host
file.
E
I
thought
the
container
would
share
the
container.
A
A
B
B
H
Is
the
bug
reporting
that
the
machines
post
files
I.
A
So
they
want
Ubuntu
container,
which
is
regular
container
to
get
the
update
from
init
container.
B
B
A
E
B
B
B
A
It's
duck
change,
duck
change
to
make
it
less
more
clear.
A
E
B
B
B
A
B
A
Okay,
so
the
issue
is
basically
when
you
do
Forest
deletion
once
and
then
you
run
Force
deletion
again
because
you
are
something
got
stuck
then
second,
first
edition
wouldn't
taking
effect
before
first
edition
completed.
So
there
is
a
virtual
queue
of
deletions
and
the
suggestion
here
is
to
not
queue
deletions
but
otherwise
deletions.
So
if
somebody
else
said
that
I
want
to
delete
faster,
then
this
will
be
respected
and
I
think
there
is
a
PR
for
that
already.
So
I
will
go
ahead
and
assign
this
to
this
person.
A
I
think
Json,
the
pr.
A
B
A
Okay,
we're
done
with
the
ash
and
we
almost
out
of
time
as
I
mentioned
in
the
beginning.
This
is
the
last
meeting
this
year.
We
can
keep
talking
offline
asynchronously
in
the
chat
and
feel
free
to
bring
me,
but
other
than
that.
Let's
see
you
in
January
thanks.