►
From YouTube: Kubernetes SIG Node 20200921
Description
No description was provided for this meeting.
If this is YOUR meeting, an easy way to fix this is to add a description to your video, wherever mtngs.io found it (probably YouTube).
A
Hello,
it's
a
sig,
note,
ci,
subgroup,
weekly
meeting
it's
monday
september,
21st
hi
everybody.
So
today
on
agenda
we
have
need
to
review
items
from
last
week
and
then
we'll
go
into
new
items.
Some
of
them
was
moved
from
the
last
week.
A
The
first
is
we
need
to
check
if
docker
is
interested
to
support
their
flaky
tests
like
so
it's
not
like
it's
like
it's
never
reported
test,
so
I
think
roy
wanted
to
follow
up
on
that
roy.
Do
you
have
any
update.
B
For
that
one,
it
means
I
I
didn't
get
time
on
that
yet
yeah.
Let
me
give
you
one
one
week.
I
would
like
to
make
sure
yeah.
This
is
why
yeah
mostly
yeah
need.
We
need
to
connect
gku
folks,
yeah
at
least
from
what
was
the
perspective,
because
we
do
know
the
history
yeah.
A
Yeah,
thank
you.
Yeah
next
week
sounds
good.
I
don't
think
there
is
a
super
pressure,
but
it
will
be
nice
to
resolve
it.
So
the
next
one
is
this
flaky
that
amim
was
trying
to
fix.
I
think.
C
C
C
A
A
It's
not
about
moving
test
from
a
folder,
it's
in
a
test
definition
test
grid
definition.
So
if
you
search
for
this
definition
in
test
infra
repository,
there
is
a
line
that
says
like
which
test
tag
to
include
and
which
folders
to
include.
So
in
this
definition,
a
suggestion
is
to
limit
it
to
gc
folder,
rather
than
all
the
folders
or
the
intent.
C
Okay,
I'll
double
check.
I
didn't.
C
A
A
If
you
need
any
help,
just
pin
people
on
slack
and
we
can
figure
it
out.
Okay,
thank.
D
A
And
next
one
is
escalato
signate,
so
we
didn't
have
time
to
discuss
it
on
sicknote
and
the
answer
was
given.
I
didn't
post
it
in
here
so
dawn
send
a
document
that
morgan
sent
to
everybody
before
and
added
short
description.
What
she
believes
that
the
difference
between
not
conformed
and
not
meister
tests.
A
I
still
don't
understand,
like
difference
between
like
running
in
docker
versus
non-writing
and
docker
like
how
is
it
different
and
what
the
definition.
So
I
propose
to
keep
it
keep
the
section
item
and
let's
call
it
again-
and
hopefully
we'll
have
time
tomorrow
to
address-
is
that.
F
F
F
A
Yeah,
I'm
going
to
be
back
by
the
way.
Okay,
do
you
know?
Do
you
have
any
context
here,
because
what
you're
trying
to
figure
out
is
what's
the
difference
between
oh
yeah,
you
created
this
work
item
so
yeah.
So
you
yeah.
E
A
Yeah-
but
we
didn't
get
to
that,
so
I
proposed
to
be
escalated
to
tomorrow
as
well,
so
we
can
discuss
it
tomorrow
to
understand
the
context.
A
And
next
section
item,
so
I
didn't
get
to
any
of
this
action
item
so
like
this
one
is
run.
Timecard
deceptive
test
create
a
new
tab
for
just
one
test.
I
will,
I
will
keep
it
under
my
plate.
Let
me
put
my
name
unfortunately,
this
week
I
was
busy
with
different
work,
and
this
one
is,
as
I
said
last
time,
I
I
yeah
morgan.
A
You
might
be
interested,
so
I
went
to
country
backs
meeting
and
I
know
like
I
mentioned,
that
we
want
this
documentation
to
be
somewhere
visible,
like
kubernetes.therefore
community
repository
and
joel
brought
up.
The
concerns
that
there
is
with
documentation
on
end-to-end
tests
is
that
this
documentation
can
go
stale
quite
easily,
especially
if
you
start
documenting
some
tags
and
stuff
like
that,
so
he
was
asking
if
there
is
any
other
place
and
last
meeting
after
that,
we
discussed
that
there
is
no
better
place,
so
we
need
to.
A
Like
I
mean
we
need
to
take
a
commitment
on
updating
it.
Whatever
place
you
want
and
like.
If
it's
community
repository,
we
need
to
keep
updating
the
community
repository.
So
my
action
item
after
that
was
to
write
up
a
proposal.
What
actual
addition
to
the
communication
want
to
add
and
make
sure
that
nobody
has
a
concern
on
country
backs
repository
or
a
grant
can
rebecca
okay.
So
this
is
you
on
me
to
follow
up?
A
Okay,
any
other
action
item
follow-ups
from
last
meeting.
I
don't
think
so
right,
okay,
then
go
to
agenda
statistic.
Roy.
Do
you
want
to
go
with
course,
81
to
85.
B
Yeah,
I
think,
because
I
on
the
test,
if
I
already
are
complete
and
that's
the
report,
basically
bomb
was
77.
B
A
A
What
you're
saying,
because
of
like
your
voice,
describing
cracking
up,
I
think
it's
connection
issue
is
the
same
for
everybody.
A
B
Yeah
sorry
looks
my
you
can
connect
it
pretty
bad
yeah
I
saw
moga
has
some
basically
bomb
81
to
85
on
the
tesla
infrared
on
for
the
place.
Where
has
immediate
family
are
already
done
and
for
the
specific
image
I
will
do
this
week.
Yeah
the
reason
I
first
basically
image
I
do
this
week
is
because
m85
support
to
mark
people.
B
I
think
early
this
week,
yeah
this
is
the
reason
yeah
for
the
kubernetes
ripple
russia
already
submit
when
pr,
I
think,
is
that
commun
community,
I
think
some
people
from
ibm
is
actually
mostly
after
stable.
I
think
if
that
appears
approved
just
the
last
one
to
yeah.
I
think
that
these
those
two
those
two
reaper
should
be
yeah.
Things
like
in
one
two
weeks,
yeah.
A
So
we
don't
need
to
keep
any
81
family.
We
just
move
every
everything
to
85.
B
No,
basically,
we
if
the
original
77
I
bumped
to
81,
if
the
original
aty
are
bumped
to
85,
as
the
problem
is
whether
we
want
to
maintain
8y
on
yeah.
We
need
to
take
a
look
for
yeah.
That's
one
thing
yeah,
but
for
now,
for
the
sake
of
safety
right
being
safe,
and
we
do
this
traditional
parts,
yeah
fifteen.
E
I
think
there
was
a
worthwhile
catch
from
having
two
of
them
running
recently
from
the
81-85
bump,
the
version
of
python
and
that
caused
a
bunch
of
stuff
to
blow
up
so
just
fyi,
it's
85
uses
53
and
81,
or
it's
like
83
or
something
uses
python
2
and
they
rolled
over.
So
I
think
that
was
a
worthwhile
catch
from
some
of
the
stable
one
stable
to
cause
stuff,
but
we
fixed
it
in
container
date.
So.
B
Yeah,
so
those
two
problems
yeah
these
problems,
like
ripple
water.
What
the
model
is
is
in
community
and
container
dci
report.
We
already
identified
fixed
fixed
posts
yeah,
even
in
the
last
week.
A
Okay,
so
thank
you
for
update.
I
don't
think.
B
A
B
A
Okay,
I
will,
I
would
vote
for
it.
We
don't,
I
don't
think
we
have
1.4
tests
yet
and
this
seems
to
be
a
gap,
but
I
know
there
is
a
work
stream
somewhere.
I
saw
mentioned
that
container
d
is
switching
to
test
infrared
as
well
like
whole
repository,
and
I
wonder
how
much
we
need
to
test
as
part
of
kubernetes
repository
and
how
much
will
be
tested
part
of
container
d
itself.
A
A
So
perhaps
you
can
create
an
issue
and
mention
that
issue
first
and
if
there
is
no
concern,
let's
do
that.
A
Okay,
any
other
topics
about
81
to
85
switch
and
like
the
shiny
new
version.
Of
course,.
A
E
E
I
noticed
that
node
problem
detector
has
its
own
test
grid,
sweet
tab,
whatever
you're
going
to
call
it,
and
I'm
wondering
if
this
test
should
be
not
run
as
part
of
the
normal
test
runs
because
it
seems
it's
been
flaking
for
months
as
far
as
I
can
tell,
and
it
experiences
periods
where
just
it
fails
like
every
time,
but
it
does,
they
don't
seem
to
be.
E
I
don't
I
don't
understand
why
it's
failing
and
so
I've
thrown
out
a
bunch
of
is
it
this.
Is
it
that
what
can
we
do?
And
so
I'm
I'm
just
curious
for
people
to
take
a
look
or
to
you
know
what
what
do
you
think?
Basically,.
E
And
you
know
questions
here
if
happy
to
try
to
answer
but
there's
a
triage
link
from
the
triage
tool,
which
is
actually
very
helpful.
Let
me
think
actually
it
is
very
helpful.
E
It
was
brought
to
me
by
you
know
some
of
the
container
d
folks
who
have
a
build
that
you
know
blows
up
a
lot
on
it
and
oh
there's,
yeah,
there's
one
of
the
things
is
it's
talking
about
counting
numbers
of
events
and
from
my
understanding
of
node
events,
aren't
quote
unquote,
reliable
and
so
counting
events
doesn't
really
make
sense.
So
I'm
not
sure
if
just
the
test
is
poorly
understanding
of
of
how
events
work
or
I
it
could
be
on
misinterpreting
the
event
stream.
So.
A
So
you
said
that
run
as
part
of
two
different
tabs.
Is
it?
Can
it
be
the
throttling
of
events
of
sort,
because,
like
I
mean
it,
would
count
something?
That's
unreliable
and
it's
unrealiability?
It's
not.
I
don't
think
it's
just
an
availability
from
like
us,
not
delivering
it
like
kubernetes,
not
delivering
it.
It's
probably
like
some
throwing
or
whatever.
E
E
You
know
that
tells
to
me
that
there's
some
kind
of
environmental
thing,
but
I
haven't
been
able
to
get
the
attention
of
any
of
the
the
node
problem.
Detector
people,
but
I
I
only
just
created
an
issue
in
their
repo
like
friday,
because
I
didn't
realize
it
was
separate
from
normal
stuff.
I
guess
yeah,
but
you
can
see
there.
It's
it's
10
out
of
10.,
it's!
It
goes
back
for
a
while.
It's
100,
you
know
passing
and
it
would
be
the
ede
node
suite
there.
It's
like
one,
two,
three
four
sixth
one.
E
Sixth
one
down,
but
that
is
the
test
and
it's
it's.
You
know
it
only
runs
literally
exactly
those
tests.
So
I'm
concerned
that
there's
maybe
a
environment
setting
or
maybe
it
should
be
marked
as
serial
or
disruptive,
because
it's
not
it
doesn't
play
nice
with
other
stuff.
You
know
they
might
be
sending
out
a
event
or
something
I
don't
know.
E
A
Okay,
so
I
think
action
item
is
to
like
I
mean
one
action
item
two
just
to
remove
this
test
from
this,
but
I
mean
maybe
remove
from
using
disruptive
feature
or
yeah,
so
you
said
we
need
to
understand
it.
First
right.
E
E
A
Yeah,
actually
I
mean,
if
you
want
to
put
it
in
critical,
then
we
want
to
put
it
in
big
chunks,
like
of
tests
like
yeah,
even
this
runtime
class
disruptive
test
that
I
want
to
add
it's
just
it's
another
like
one
single
test
that
you
run
for
entire
tab
and
like
it
may
be
a
little
bit
too
much,
but
there
is
no
other
option.
A
E
G
E
Right
now
that
that
last
tab
is
literally
the
only
thing
keeping
the
sig
node
cubelet
from
being
all
all
blue,
all
passing-
or
you
know
timed
out,
but
this
one
is
I'm
wondering
if
it's
just
the
the
tests
clobber
each
other
through
the
run
because
it
you
know
it
runs
in
a
different
order,
every
time,
and
so
I
can't
you
know,
I
can
not
reproduce
this
on
sort
of
my
own
machine
or
consistently
with
with
the
the
vm
runners,
and
you
know,
sometimes
some
of
the
tests
run
and
sometimes
they're
successful.
E
E
Serially
like
it
seems
to
run
each
individual
vm
after
the
next,
and
so
you
end
up
with
this
weird-looking
stippled
pattern
where
it's
like
fail,
nothing
fail,
nothing
fail!
Nothing
because
it's
it's
literally
just
it
gives
up
it
times
out
and
it
doesn't
even
try
to
talk
to
the
other.
A
A
E
They
seem
important
they're
the
only
place
that
the
serial
disruptive
tests
run,
so
if
they
are
actually
important,
we
need
to
fix
them
rather
than
delete
them,
because
this
is
the
only
time
these
tests
run
so.
A
Okay,
there
is
no
takers
on
a
call.
Is
there
an
issue
for
that.
H
E
E
No,
I
didn't,
I
don't
think
I
made
one
I'll.
I
can
make
one
I
just
I'm
I'm
out
of
words
for
how
to
describe
it,
but
yeah.
It's
all
red
everything
is
failing
yeah,
I
guess
so.
Okay,
yeah.
A
I
don't
think
we
can
come
up
with
anything
better
and
I
think
it
will
be
something
simple
like
some
like
get
to
the
root
cause
and
it
will
get
better
yeah,
okay,
yeah
we
out
of
agenda.
A
A
So
I
will
take
all
the
action
items
from
this
week
and
move
to
the
next
week,
so
we
can
have
a
agenda
prepared
already
and
yeah.
Let's
I
will
get
to
my
items
this
week.
Sure
thank
you.
E
Wait,
oh,
wait,
wait.
Did
we
all
right
sorry,
did
we
get
that
one
email
thing
set
up
that
that
pr
did
that
get
merged?
I
didn't
check
yet.
A
E
A
H
H
Quick
update
on
that
one,
I
was
hoping
to
see
all
the
note
critical
and
no
no
release
blocking
test
actually
in
there
before
I
actually
proceeded
with
that
one.
I
was
taking
a
look
back
at
it
and
I
I
I
included
jobs
just
based
on
the
tab,
and
I
noticed
that
I
was
including
a
lot
of
jobs
that
were
failing
so
essentially
like
the
mailing
list
was
gonna
be
bombarded
by
prague,
but
gonna
they
closed
out.
It
close
the
pr
and
gonna
update
it
with
the
jobs
all
right.
H
E
H
To
be
honest,
please
don't
tell
this
to
anyone,
but
I
was
following
up
with
the
agenda.
While
I
was
on
the
release
team
meeting,
so
I
kind
of
follow
some
some
some
other
things,
some
of
the
things,
but
I
I
really
feel
like
I
should
say
sorry
for
interrupting,
but
I
think
that
I
think
I'm
following
my
following
up
on
most
of
the
things
and
I
will
make
sure
to
I
to
also
follow
up
on
all
the
tone
of
issues
that
I
have
my
name
on
it.
H
A
And
it
sucked
to
be
double
booked,
keep
keep
the
good
work.
Thank
you,
okay.
I
guess
that's
it
then
thank
you,
everybody
again
and
have
a
good
week.