►
From YouTube: Kubernetes SIG Node 20210414
Description
Meeting Agenda:
https://docs.google.com/document/d/1j3vrG6BgE0hUDs2e-1ZUegKN4W4Adb1B6oJ6j-4kyPU
A
Okay,
it's
recording
now
hello,
it's
april,
14th
2021
and
it's
a
signal
ci
meeting
today
we
yeah
let's
get
to
agenda,
we
have
many
people
joining
so
if
you
would
put
your
name
into
a
gender
document
the
great
let
me
paste
agenda
document
in
chat.
B
A
Thank
you,
yeah.
I
put
some
agenda
items
here
as
I
promised
I
wanted
to
get
into
smaller
items.
So
let's
go
through
agenda
first
and
then
we'll
do
triage
of.
A
So
first,
I
wanted
to
highlight
this
coupe
test
update,
there's
a
pr
that
updates
in
coupe
test
to
support
like
we
want
to
switch
to
coupe
test
and
coop
test.
Has
many
benefits
like
it
will
allow
to
run
locally
like
just
take
a
command
line
from
test
execution
and
run
test
locally,
with
skip
test,
which
is
great,
and
this
pr
starts
in
this
migration.
So
it's
has
a
subset
of
features.
I
see
david
porter
already
reviewed
it
so,
but
this
pr
may
need
more
eyes.
C
It
looks
like
it's
also
already
approved,
who
approved
it.
D
Yes,
so
I
can
maybe
give
just
a
little
bit
background.
I
actually
wanted
to
get
other
people's
feedback
here.
So
basically,
the
summary
there
is
that,
right
now
for
all
the
ci
tests,
we
are
currently
using
like
coop
test
one,
and
so
these
people
are
I'm
on
here.
It's
trying
to
move
to
coop
test
two,
the
thing
that
I
wanted
to
bring
up
and
get
feedback
from
other
people
that
I
kind
of
brought
up
on
the
pr
is
that
currently
for
coop
test
one
for
all
of
our
ci
jobs.
D
We
basically
you
use
coop
test
and
and
there's
like
certain
flags
like
dash
dash
node,
test,
args,
et
cetera,
and
so
all
of
our
ci
jobs
are
set
up
to
use
that,
but
for
people
who've
run
the
ete
test
locally.
We
also
have
like
this
make
thing,
so
you
can
run
it
from
make
like
make
node
test
e
to
e,
and
then
you
pass
some
tests
and
stuff.
D
So
there's
like
kind
of
two
interfaces
right
now
for
running
an
end
to
end
node
test,
and
so
the
pr
here
for
coupe
des
2
is
trying
to
move
to
that
make
target.
D
D
So
I
was
just
bringing
up
kind
of
in
the
pr,
if
that,
if
that
makes
sense,
if
we
want
to
standardize
on
that
or
if
we
want
to
kind
of
just
copy
over
kind
of
the
same
interface
as
cube
test
one,
so
that
was
kind
of
what
I
was
describing
on
the
dot.
But
I
I'm
not.
You
know
we
get
to
get
other
people's
feedback
there.
D
A
D
C
D
C
D
C
But,
like
you
know
a
couple
of
paragraphs
because,
like
I
people
keep
asking
me
to
look
at
this,
I'm
like
I
don't
know
what
you
want
me
to
say
like
this
is
probably
the
second
or
third
time
I've
looked
at
this
pr,
but
like
what
do
you
need
from
me?
B
Hey
just
quick
question,
slash
comment
from
my
side
and
I
I
would
like
to
have
a
look
and
learn
indeed
about
more
about
context,
but
just
just
because
and
node
and
not
the
end-to-end
test
are
a
bit
special
meaning
they
they
they
create
their
own
mini
cluster
with
the
ap
server
atcd
they're,
not
really
part
of
the
regular
end-to-end
test.
B
Let's
say
as
far
as
I
know,
so
I'm
really
looking
forward
to
to
I'm
really
curious
to
learn
a
bit
more
and
because
I'm
saying
so
as
a
consumer
of
desktop,
not
really
that
much
into
infrared.
That
looks
like
a
great
opportunity
for
me
to
learn
a
bit.
So
my
question
is:
is
this
part
of
process
to
move
the
node
and
to
interest
toward
to
make
them
more
similar
to
other
android
tests
or
is
completely
unrelated.
D
I
think
the
whole
idea
is
just
an
interface
like
to
launch
the
test.
It's
still
going
to
be
completely
separate,
it's
just
like
a
different
cli
different,
like
arcs,
but
it
eventually
calls
into
the
same
stuff.
So.
A
D
Mean
I
I
can
ask
sure
I
mean
I
yeah.
I
think
I
actually
sort
of
understand
the
the
changes,
but
the
question
is
like
whether
we
want
to
do
it
and
who
wants
to
do
it
on
our
team
or
on
insigno.
You
know
so
that's
kind
of
the
bigger
question
I
think,
but.
A
D
Not
not
to
mine,
no,
I
mean
it
calls
into
the
same
test
as
it
does
today
right.
It's
just
a
different
kind
of.
A
A
Okay,
okay,
francesca,
okay,
you
you
added
the
node
okay,
so
second
item:
I
looked
at
our
test
grid
and
now
we
we
cleaned
up
all
the
seek
ownership
across
all
the
projects
and
before
most
of
this
test
had
two
seeks
like,
for
instance,
this
had
like
this
test
had
like
six
node
and
sig
network.
But
now
we
cleaned
up
specifically
like
detached,
like
only
picked
one
sig
only
in
the
test.
A
So
this
way
you
know
who
to
ask
if
it
fails-
and
for
instance
here
I
see
that
one
of
the
six
storage
thingy.
I
think
this
is
six
storage
was
flaking
like
maybe
next
one-
and
I
I
just
realized
that,
like
this
is
not
signaled,
it's
six
storage
and
the
six
orange
team
storage
may
not
even
know
about
this
test
failing
or
being
flaky,
and
we
are
not
looking
at
it
because
it's
six
storage.
A
So
I
I
I'm
thinking
we
need
to
delete
and
like
move
all
six
storage
tests
out
of
our
test
grid
and
move
it
to
under
seek
storage
and
signature
network
on
the
sig
network.
A
So
the
effort
is
basically
at
the
skip
for
our
test
grid
to
remove
sig
network,
but
before
that
we
need
to
get
an
agreement
with
other
seeks
to
pick
up
the
stats
and
run
them
as
part
of
their
test
grid.
A
A
One
drawback,
I
see
is
this
test
like
right
now
they
execute
all
together
with
a
move.
They
will
need
to
be
executed
separately,
so
we'll
like
waste
a
little
bit
of
resources
on
spinning
up
infrastructure
and
tearing
it
down,
but
I
don't
think
it's
a
big
concern.
I'm
more
worried
that
other
teams
are
not
looking
into
tears
that
belongs
to.
A
So
my
biggest
motivation
is,
I'm
not
sure
whether
they
like
anybody
from
those
six
are
looking
at
this
test
failures.
So
if
tests
will
fail
like,
for
instance,
this
test
that
I
showed
in
118
and
118
will
need
to
be
removed
soon,
but
I'm
I'm
not
sure
whether
this
test
is
yes,
maybe
the
best
example
but
ultimately
like
if
that
starts
failing,
and
it
is
failing,
because
we
picked
a
bad
cherry
pick
like
we,
we
took
a
cherry
pick
for
the
release.
F
A
So,
okay,
so
you.
F
Yeah
I
vote
to
keep
them
and
maybe
remove
all
the
flake
ones
like
the
scpp,
but
this
corn
arcade.
A
A
Okay,
I
will
create
an
issue
to
discuss
it
and
maybe
like
the
way
to
notify
them,
will
be
to
just
create
an
issue
and
focus
them
on
slack
just
yeah.
Okay,
let's
see
I,
I
don't
really
want
to
be
this
person
who,
like
coordinates
all
the
chicks
and
trying
to
do
stuff.
A
A
The
next
big
thing
is,
we
have
this
conformance
tab
completely
broken.
Something
is
infrastructure,
I
believe,
because
nothing
is
running.
Basically,
all
tests
are
failing
and
yeah
that.
A
A
Yeah,
I
think
one
good
thing
about
it
is
some
of
these
tests
are
covered
by
eddie's
blocking.
C
A
How
an
issue
and
paste
it
here
I'll
mention
you
thank
you
and
yeah
another
tab
that
is
totally
red
like
there
are
many
like
flaky
things.
You
know
test
grid,
but
these
two
tabs
are
like
super
broken
and
supposed
to
be
a
very
easy
fix
because
likely
when
something
is
that
broken,
it's
just
infrastructure
issue.
So
serial
tests
are
also
like
all
red
and
it's
also
likely
infrastructure
issue.
E
B
Hey
francesco
here,
if
there
are
no
volunteers,
put
my
name
on
it.
I
cannot
promise.
I
can
never
look
before
next
meeting,
but
I
I
will
try.
B
A
Okay,
I'll
create
issue
then,
and
mention
that
maybe
not
immediately.
Okay
and
last
thing
that
caught
my
attention
is
this
james
is
trying
to
like.
We
have
multi-zone
tests,
we
have
three
of
them
and
I
think
these
tests
are
likely
validating
some.
Like
some
latency
and
network
issues.
We
have
tests
for
like
just
running
all
the
tests
except
disruptive
and
slow
running,
specifically
ingress
tests.
A
That
may
discover
some
problems
with
how
ingress
is
deployed
and
how
it
runs
across
clusters
and
across
zones,
not
clusters
and
container
details.
So
basically,
all
multi-zone
tests
are
currently
broken
and
I'm
trying
to
understand,
like
jim's
suggestion
to
just
remove
multi-zone
completely
and
I'm
trying
to
dig
into
history
whether
those
tests
like
I
bet
those
tests
were
added
for
reason.
It's
not
like
we,
I
mean
somebody
just
from
a
goodness
of
their
heart
to
adding
some
tests
for
multi-zone.
So
I
maybe
they
caught
some
like
regression
or
something
in
kubernetes
before.
C
C
A
Yeah,
yes,
I
yeah
that's
what
I'm
saying
like
they
likely
may
expose
some
like
assumptions
that
may
be
taken
in
kubernetes
around
like
which
nodes
something
is
located
on
which
zone
is.
It
is
running
something
with
networking
like.
Maybe
connectivity
will
be
lost
with
improper
configuration,
so,
okay,
I
just.
A
Okay,
so
I
what
I
was
planning
to
do
is
to
dig
into
who
created
this
test
and
like
why
it
was
clear
like
what
kind
of
regression
record,
and
once
we
figure
out
what
regression
is
called
like
in
this
test
are
important.
Then
we
can
get
somebody
to
work
on
that.
So
is
there
anybody
who,
who
has
context
or
who
wants
to
take
a
look
at
multi
zone
interested
in
multi-zones.
A
Okay,
I
will
give
an
update
next
signature.
Then.
A
A
A
Yeah
good
thing
about
this
dashboard,
it's
very
small,
so,
most
of
the
time
you
I
don't
need
but
yeah.
Thank
you
comparing
to
signal
dashboard
like
product
one.
It's
it's
very
small.
A
C
A
C
E
A
A
Also,
quite
straightforward
to
just
one
new
test.
A
But
it's
all
in
the
same
area
right
so
it's
just!
This
is
a
life
cycle
hunter
test.
A
A
Thank
you
very
much,
and
this
like
is
there
anybody
else
who
wants
to
talk
about
what
was
assigned
to
him
and
or
her,
and
it
wasn't
very
straightforward
to
need
help.
C
You
can
drop
it
if
I
will
take
it
over
thanks,
sergey.
A
And
let
me
stop
recording.