►
From YouTube: Kubernetes SIG Node 20220119
Description
Meeting Agenda:
https://docs.google.com/document/d/1j3vrG6BgE0hUDs2e-1ZUegKN4W4Adb1B6oJ6j-4kyPU
A
Okay,
hello,
everybody,
it's
january,
19
2022,
it's
a
signal.
Ci
weekly
meeting
welcome
everybody.
We
have
first
agenda
item
about
moving
to
another.
B
Thank
you.
Basically,
we
have
some
job
using
custom
project
inside
the
google
organization
and
because
we
get
involved
supposed
to
migrate
all
the
jobs
to
the
community
infrastructure.
We
want
to
stop
doing
that.
We
start
with
this
issue.
I
think
was
in
october.
So
there's
just
one
basically
the
last
one
and
gets
cry
control
energy
we
need
to
remove,
but
those
chops
push
the
logs
in
specific
bracket.
B
So
I
don't
know:
why
are
we
doing
this
for
those
specific
tests?
If
is
it
possible
to
drop
this
bucket.
C
Is
just
because
there
hasn't
been
necessarily
a
ton
of
continuity
and,
like
I
think
people
in
this
group
possibly
would
not
know
because
we
didn't
set
this
up,
so
I
mean
I
think
it's
fine
to
like
go
ahead
and
just
try
it
and
see
what
happens
and
hopefully,
if
you
know
if
it
breaks
all
the
tests,
then
referred
it.
I
think
that
would
be
fine.
Like
honestly,
I
think
that
that's
even
if
it's
a
little
bit,
you
know
less
sophisticated.
C
It's
probably
the
least
work
to
just
go
ahead,
migrate
things
one
by
one,
see
what
breaks.
If
it
breaks,
revert
it,
and
then
we
try
to
debug
it,
because
you
know
anything
other
than
that.
I
just
I
think
it's
gonna
be
a
lot
of
energy
and
a
lot
of
effort.
B
B
C
Yes,
sorry,
I
think
I'm
completing
the
two
so
like
there
are
kind
of
two
things
we
need
to
do.
One
is
we
need
to
stop
running
stuff
on
the
google
cluster.
C
We
need
to
run
that
on
the
cncf
cluster
and
two
is
we
want
to
make
sure
if
there
are
any
custom
projects
that
we
try
to
get
rid
of
those,
and
I
think
that
some
of
the
cleanup
has
been
happening
so,
like
you
know,
we
can't
move
stuff
to
the
proud
cluster
if
the
if
the
project
doesn't
exist
kind
of
thing,
but
I
would
say
we
could
probably
do
both
at
the
same
time
and
again
see
what
breaks
I
wasn't
involved
in
setting
up
any
of
the
projects.
A
C
Any
of
this
stuff
was
set
up
so,
okay,
I
think
it's
fine
to
try
it
and
if
it
doesn't
work,
then
to
revert
it
and
then
we
can
go
and
debug
and
get
more
specific
and
we're
really
early
in
the
release
cycle
right
now.
So
it's
not
like
we're
going
to
destroy
signal
by
doing
horrible
things
or
something
like
that.
We
just
need
to
make
sure
that
we
don't
like
let
things
stay,
broken.
C
B
B
The
general
rules
is,
we
have
a
pool
of
projects
for
bosco
prefix
with
getting
friend
name,
and
you
just
use
that
you
have
to
care
about
specific
a
specific
project.
So
bosco
will
take
care
of
the
rest.
But
if
you
have
a
specific
need,
it's
like,
I
would
say,
for
example,
bar
metal
instance
with
swap
enable
you
may
find
a
request
for
this.
But
for
the
moment
anything
running
inside
kits
and
fra
don't
need
its
custom
project.
B
A
Do
you
think
you
can
help
look
into
that
and
troubleshoot.
A
C
C
A
Yeah
another
thing
that
just
came
to
mind
looking
at
npt
test,
I
also
saw
similar
something
complaining
about
ssh
as
well,
so.
A
Yeah,
it
may
not
be
related.
I
don't
know.
C
C
So
it
is
possible
for
what
it's
worth
just
because
it's
in
test
grid
doesn't
mean
that
we're
running
it
in
kubernetes
like
it's
possible
to,
for
example,
configure
test
grid
with
a
custom
bucket
to
just
go
and
look
at
that
bucket,
assuming
it
has
like
the
permissions
and
like
pull
test
results
out
of
it
so
like
it
could
be
run
by
you
know.
I
don't
know
little
green
men
or
something
like
really
anything,
and
we
will
display
it
so
long
as
the
results
are
in
a
bucket
in
a
form.
C
B
A
Okay,
thank
you.
C
B
D
B
A
Okay,
let's
take
a
look
at
other
failures.
That's
great!
So
we
have
a
bunch
of
greens.
There
is
no
progress
in
this
couplet
couplet.
What
is
this.
A
It's
still
flaking
and
we
have
a
swap
ubuntu
item
mike.
You
wanna
call
right.
E
Yes,
yes,
I'm
taking
a
look
into
this
item.
In
particular,
I
guess
right
now,
I'm
trying
to
run
this
test
with
more
verbosity.
If
I
don't
detect
anything
I'll,
just
bomb
the
timeout
period.
Okay,
it
should
work.
E
A
A
A
Okay,
david
says
that
there
is
some
change
in
path
and
he
says
he'll
be
looking
at
that
state
on
the
call.
A
C
Like
danielle
was
looking
at
them,
but
she
is
currently
out
sick,
so.
A
Yeah,
I
don't
want
to
assign
without
her
present,
so
I
would
rather
wait
if
you
believe,
certainly
the
best
person.
A
Any
other
takers-
okay,
let's
hang
for
another
week,
and
then
we
need
to
find
somebody.
A
Yeah
we
had
this
green
without,
like
this
one,
no
problem
detector
failing
and
then
we
get
back
into
like
all
failed
yeah.
I
will
I'll
figure
out.
I
mean
we
still
need
to
ask
costume
if
they
interested
in
this
tab
and
what's
this,
whether
we
need
it
at
all,
maybe
just
remove
it
and
similar
story
here,
flaky,
no
changes,
it's
still
failing
yeah
my
changes,
yeah,
let's
ping
them
again.
A
Yeah
for
npgs
there
was
some
progress
is
changed
on
default,
but
the
issues
that
I
just
showed
about
the
mismatch
of
ssh
keys,
yes
advisor
is
fine
and
cryo
has
a
bunch
of
new
failures.
So
last
time
it
was
just
graceful
termination
that
was
failing
because
missing
tools,
it
was
fixed
by
paca
and
now
there
are
a
bunch
of
new
feathers
alpha.
C
Oh,
I
see
on
the
alpha
features
tab,
I'm
not.
F
A
I
think
we
should
because
he's
so
conformant
yeah
yeah.
A
A
I
mean
yeah
it
just
like
guessing
but
yeah
anyway,
this
one,
it's
a
new
failure,
so
eviction
on
container
d
failing
this
is
a
disk
pressure
and
something
else.
C
A
This
one
is,
I
think,
it's
a
new
failures,
yeah,
it's
new
failures.
Oh.
C
C
C
And
that
one
is
a
disruptive
test,
so
who
knows
previously?
It
looks
like
it
was
getting
skipped
all
the
time
and
now
it's
failing,
so
I'm
not
sure
what
changed.
I
wouldn't
worry
too.
A
C
A
A
Okay,
so
we
need
bugs
for
two
of
them
and
yeah.
Everything
else
is
green,
so
we
good,
I
think,
we're
making
progress.
Hopefully
I
don't
know
this
format.
I
I
filled
it
up
myself
this
week
and
I
think
format
may
be
improved.
Maybe
you
can
break
it
down
by
individual
tests.
I
will
try
it
out.
It
may
be
too
many
of
them.
If
so,
then
we
will
need
to
figure
out
what
to
do,
but
I
hope
we
can
just
do
something
said:
okay,.
A
A
Okay,
so
peter
you
you're
on
a
call.
You
heard
that
I
don't
know
it
will
help
is
looking
into
that
right.
It's
assigned
to
you
correctly.
A
Maybe
it's
wise
to
wait
till
after
weekend.
A
C
So
this,
I
don't
think
this
is
test
related.
No,
this
is
like
an
api
change.
I
think
this
could
be
archived
for
that
board,
but
I
think
we
do
need
someone
to
look
at
it
from
node.
Maybe
I
need
to
look
at
this
for
instrumentation
reasons.
I
think
so
we.
C
Yeah
I
I
reviewed
that
one
already,
so
I
think
I'm
waiting
on
matthias
is
he
on
the
call.
A
C
Well,
if
matthias
ever
watches
this
recording
yeah,
basically,
I
just
suggested
that
we
don't
we.
We
can't
touch
any
release
branch
jobs,
like
generally.
E
C
They're,
like
totally
broken
once
the
release
gets
cut.
We
leave
those
completely
alone.
They
should
stay
in
that
state.
So
I
just
asked
him
to
make
sure
that
we're
not
touching.
We
don't
want
to
remove
docker
shim
jobs
from
previous
branches,
because
docker
is
still
supported
there
and
you
know
we
don't
want
to
remove
our
coverage
for
it.
A
Absolutely
okay
update
cryo
commit
to
something
I
have
no
context
here.
That's
homemaster.
A
A
Yeah,
I'm
sometimes
lost
like
there
are
so
many
clean
up
pr's
and
the
intercept
intersect
with
the
charger.
F
A
G
D
C
A
A
This
should
be
done,
of
course,
itself.
Okay,
we
have
26
minutes
for
box
triage.
Is
there
anything
last
minute
for
test
related,
otherwise
we'll
just
go
to
box
yeah.
Let
me
check
really
quick,
so
good
anything
related
to
this
going
once
going
twice.
C
Yes,
I
mean:
maybe
we
should
triage
accept
it
just
because
we
know
we're
gonna
research.
It
do
we
have
francesco
today.
F
C
A
F
G
A
Anybody
wants
to
take
a
look
at
this
test.
Filler.
C
F
C
A
If
you
have
some
freaking
checks
but
yeah,
so
something
like
please
use
form.
G
A
A
What
which
one
is
that
it's?
What
does
it
do.
A
It
sounds
like
a
legit
issue
and
this
person
sounds
to
be.
I
mean
yeah.
This
area
pro
is
quite
sophisticated.
You
need
to
insert
some
sleep
and.
C
You
can
assign
this
one
to
me.
I
don't
know
if.
A
C
C
Although
it's
hard
to
say
because
I
guess
they're
talking
about
the
end
to
end
tests
or
whatever
for
a
local
invocation.
C
A
A
I
mean
cleanup
is
a
little
bit,
I
mean
yeah
black
is
better
than
cleanup.
I
think.
A
C
It's
ignored
because
we
have
our
the
the
node
e
to
e
tests.
Are
our
own
special
maintained
thing,
yeah
bespoke.
A
Okay,
I
don't
know
if
anybody
wants
to
take
it.
Please
assign
yourself.
C
That's
on
119,
which
is
not
supported
anymore
and
they're,
talking
about
versions
from
114
to
117,
so
I
think
only
120
plus
is
supported
right
now.
A
C
C
Pods,
don't
get
rescheduled
pods
immortal,
so
you
can.
Let's
see
if,
if
that's
an
issue,
I
would
say
it's
an
abs
issue,
not
an
us
issue.
C
C
Could
be
a
bug,
but
I
don't
think
it
is
a
bug
so
yeah
just
to
sign
it
to
me
and
I'll
I'll
write
to
them.
A
Let
me
turn
this:
is
it
yeah,
that's
fair,
because
if
they
expect
the
other
part
to
be
attacked
here,
it's
one
thing,
but
users
expect
this
port
to
be
rescheduled.
Yeah.
C
That's,
that's
not
how
that
would
work.
What
should
happen
is
the
controller
for
whatever
that
app
is
I'm
assuming
that's
a
deployment
should
schedule
a
new
pod
on
a
new
node
like
that's
pods
are
mortal
once
they
get
evicted,
they're
dead,
they're,
never
coming
back.
So.