►
From YouTube: Kubernetes SIG Node CI 20230315
Description
SIG Node CI weekly meeting. Agenda and notes: https://docs.google.com/document/d/1fb-ugvgdSVIkkuJ388_nhp2pBTy_4HEVg5848Xy7n5U/edit#heading=h.2v8vzknys4nk
GMT20230315-170529_Recording_1842x1120.mp4
A
A
A
A
Okay
and
I,
also
added
couple
agenda
items,
just
small
announcements
and
just
FYI
kind
of
things.
First,
is
we
discovered
very
late
in
a
release
cycle
that
some
tests,
like
some
scenarios,
were
failing
with
the
new
features?
Specifically
when
you
enable
In-Place
vertical
Port
scaling
and
one
of
the
storage
feature
case?
Standalone
agent
will
start
failing.
A
Typically,
I
mean
it's
a
one
of
the
beta
features,
so
it's
critical
another
one
in
place.
Update
is
Alpha
feature
so
trait.
Equally,
you
can
just
disable
this
feature,
but
our
typical
a
threshold
for
this
kind
of
functionality
that
the
oga
features
are
supposed
to
be
working.
The
oga
features
are
supposed
to
be
working,
even
if
Alpha
features
are
enabled.
That's
why
we
are
running
not
conformance
on
Alpha,
with
all
Alpha
feature,
Gates
enabled.
A
The
problem
is
that
we
have
a
set
of
functionality,
Standalone
kubolid,
Standalone
mode,
when
Google
transfers
out
API
server.
So
basically,
it's
only
capable
to
run
static,
pods
and
in
this
mode
you
don't
have
Coop
client
object,
it's
new
and
whenever
you
try
to
access
it-
and
there
are
many
reasons
you
want
to
access
it
into
a
panic,
so
we
caught
two
pionics
and
we
have
no
way
to
validate
it,
except
manually,
run
covet
and
Standalone
mode.
It's
not
too
hard.
A
I
mean
you
just
like
have
a
binary
and
you
run
it,
but
the
same
way
you
don't
want
to
do
it
on
every
PR.
So
we
created
this
Standalone
mode
test
they
already
merged,
but
they
are
failing
right
now
because
of
some
problems,
so
the
pr
first
pair
of
fixing
them
to
make
them
green
and
second
PR
agent
test
keys
that
will
catch
a
regression
if
in
place
vertical
pod
scaling
will
some
of
that
in
place
vertical,
but
scaling
a
cost.
A
A
Yeah
and
historically
we
never
tested
it.
I
think
I
was
told,
I,
don't
know
the
history.
I
call
whole
history,
but
I
was
told
that
Standalone
mode
was
always
on
a
back
burner
like
testing
set
and
mostly
it
wasn't
tested,
because
some
of
the
end-to-end
executions
are
running
master
components
with
a
Kubota
Standalone
in
Standalone
mode,
meaning
that
we
tested
at
least
somehow,
like
we
have
some
basic
tests.
The
problem
is,
we
don't
enable
Alpha
feature,
for
instance,
when
we
run
so
yeah.
A
A
C
A
A
A
This
kind
of
test
and
I
think
we
need
more
of
those
more
edge
cases
and
I
hope
that
our
test
framework
Will
Survive
and
handles
it
very
well
so
and
yeah
I
I,
looked
at
this
flake
today
and
I
saw
limit
and
10
and
like
resources
I
like
okay,
it's
what
resources,
probably
but
yeah,
not
the
same
test.
So
some
tests
is
flaking
right
now,.
C
And
just
one
question,
which
is
probably
silly
but
still
looking
at
the
test
grid,
I
see
that
the
test
kit
is
telling
us
that
the
test
is
failing
because
it
reads
it
reached
the
CPU
limits
exactly
so
I
I
wonder
if
this
is
an
artifact
of
the
test.
You
know
of
the
machine
of
the
configuration
I
mean
the
machine
we
we
have
for
for
running
the
tests
as
some
limits,
of
course,
or
it
is
some
some
like
a
red
bearing
I'm,
really
wondering.
A
A
I
know
that
there
were
issues
with
even
with
serial
tests
and
cleanup
of
previous
tests
took
too
long,
but
I
think
we
still
run
them
seriously.
We
don't
want,
we
won't
run
it,
but
there
may
be
some
leftovers
that
we
forgot
to
clean
up
or
something
like
that.
Evictions
are
very
sensitive
in
the
sense.
A
D
A
B
B
D
A
B
B
A
A
F
D
B
A
A
A
D
B
A
A
E
A
A
A
F
A
D
B
A
B
A
B
B
B
A
A
A
D
A
A
A
E
C
E
B
A
F
A
A
F
A
F
C
A
D
C
Yes,
Sam
I
want
to
look
at
that
I
didn't
time
too
yet.
So
this
is
why
I'm
chasing
myself,
you
can
even
assign
me
to
that.
If
no
one
else
wants,
but
I
still
need
to
find
time
at
glance
looks
legit,
but
this
is
why
I'm
still
not
to
treasure
accepting,
because
at
the
glance,
looks
legit
I,
think
it's
legit,
but
it
didn't
have
time
to
to
look
into
that.