►
From YouTube: Kubernetes SIG Node CI 20221207
Description
SIG Node CI weekly meeting. Agenda and notes: https://docs.google.com/document/d/1fb-ugvgdSVIkkuJ388_nhp2pBTy_4HEVg5848Xy7n5U/edit#heading=h.2v8vzknys4nk
GMT20221207-180550_Recording_1920x1120
A
A
We
have
some
agenda
today.
Well,
I
forgot
to
share
my
screen.
I
can
do
that
and
Swati
if
you
want
to
start
speaking
while
I'm
sharing
my
screen.
B
So
this
kind
of
the
topology
manager
end-to-end
tests,
are
basically
skipped,
especially
the
tests
that
are
multi-numer
related.
They
are
skipped
on
if
the
tests
are
running
on
a
system
that
doesn't
have
multinoma
so
given
that
topology
manager
kind
of
cares
about
multinoma
system
and
we
don't
have
a
proper
tense
test
infrastructure
to
run
those
tests.
I
just
wanted
to
bring
that
up
here,
especially
because
we
were
discussing
topology
manager,
graduation
to
ga
I
intend
to
bring
this
topic
to
the
main
signal
meeting
as
well.
B
But
given
that
we,
we
typically
have
discussion
about,
you
know
tests
and
test
infrastructure.
I
wanted
to
see
if
people
have
thoughts
here
on
how
we
can
handle
this.
If
there
is
infrastructure
or
Hardware
that
we
can
add
to
the
existing
CI
infrastructure,
or
is
there
a
way
that
you
know
like
typically
for
topology
manager,
CPU
manager
within
Red
Hat?
We
have
a
CI
environment
that
runs
these
end-to-end
tests,
but
we
don't
have
signals
from
Upstream
side.
So
I
was
wondering
if
anyone
has
thoughts
or
pointers.
B
I
pointed
to
nodes
in
gcp
that
have
this
capability,
but
so
far
my
understanding
is
the
set
of
nodes
that
we
run
our
tests
on,
they
don't
have
multi-noma.
So
I
was
just
looking
for
thoughts.
Comments
on
this.
C
The
one
thing
I'd
be
wary
of
having
looked
up
the
price
of
the
C2
standard
16.
is
that
we
probably
want
to
talk
to
Kate's
infra,
because
we
don't
have
money
like
we're
perpetually
running
out
of
budget,
and
so
we
can
potentially
spin
up
a
job
that
has
a
very
reduced
test
set.
But
the
cost
of
spinning
up
and
tearing
down
a
CI
node
is
also
fairly
High.
B
B
Yeah
and
for
me
to
talk
to
Sig
in
kubernetes,
and
for
would
it
we
would
have
to
go
to
Sig
testing,
or
would
it
be
someone
else
I.
C
Yeah,
that
meeting
is
like
a
very
awkward
time
but
like
either
talking
to
Dems
or.
C
B
No
I
think
I
can
have
a
conversation
with
them
and
see
if
they
can
help
out
in
this
area,
because
yeah
not
having
tests
running
for
a
ga
feature
or
a
potential
GA
feature,
is
not
ideal,
and
we
would
like
to
solve
that.
A
Do
you
know
what
kind
of
Numa
is
on
red
hat
test
environment?
Is
it
also
like
with
very
many
CPUs,
or
it's
like
more
reasonable
kind
of
machines?
Maybe
new
movies,
Network
or
something.
B
Yeah,
the
typically
the
nodes
that
we
have
it
has
some
Nicks
and
CPUs,
maybe
in
the
range
of
50
to
80
I,
can't
remember
on
top
of
my
head,
but
to
toenuma
nodes
on
each
system
with
next,
and
you
know
the
decent
amount
of
CPUs
like
we
don't
need
a
lot
as
long
as
we
have
multinoma
system
I
like
it,
doesn't
matter
too
much
how
many
CPUs
we
have
as
long
as
you
know,
it
validates
the
checks
that
we
have
in
place
that
okay,
it's
a
multinoma
system.
Now
we
run
the
test.
B
So
again,
I
wasn't
able
to
find
a
lot
of
information
about
gcp
infrastructure
and
the
the
specification
that
gives
us
pneumonode
and
I
just
ran
into
this
particular
article.
That
pointed
me
to
C2
standard
60.
That
definitely
has
two
number
nodes,
so
I
just
put
that
as
an
as
an
example.
But
if,
if
there
are
pointers
on
you
know
where
I
can
get
more
information
on
which
specification
exactly
has
multinoma
and
we
can
go
for
the
cheapest
one
out
of
those.
A
Yeah
I'm
going
to
ask
around
as
well
I
think
I
know
people
who
wants
to
do
Numa
on
GT
as
well.
So,
okay,
we'll
find
information.
A
I
know
wealth
of
a
lot
of
work
on
updating
tests.
We
see
group
mismatch
like
a
c
group
of
continuity,
wasn't
much
in
the
C
group
of
secret
setting
of
kublet
and
it
costs
a
lot
of
problems.
A
I
think
this
work
is
ongoing,
but
I
don't
see
anybody
on
a
call
who
was
doing
this
work.
A
Yeah,
it's
it's
very
strange.
Park
I
I
mentioned
last
time
what
we've
been
discussing
it.
It's
not
easy
to
detect
this
problem.
I
mean
it's
easy
enough
to
detect
it
on
existing,
like
if
you
know
that
you're
between
C
group,
like
between
candidi
and
kubert,
but
it's
not
easy
to
do
it
genetically.
A
Okay,
before
we
go
to
triage,
let's
look
at
performance
just
in
case
I.
B
A
Release
will
be
tomorrow
right,
so
nothing's
supposed
to
be
bad
supposed
to
be
changing
at
least
yeah.
Nothing
is
changing.
Everything
is
stable,
runtime
same
way,.
A
Looks
easy
enough,
I
wonder
how
it's
validated.
A
A
C
A
C
I've
been
meaning
to
play
with
it,
because
they're
one
of
the
things
they
support
is
spin,
which
is
a
thing
we
make.
A
Oh,
this
is
about
timing
of
registration
that
we
discussed
yesterday
on
the
meeting
yeah.
This
is
not
test
related,
so
I
will
archive
it
from
here.
A
At
least
123
to
go,
119
I
think
it's
not
related
to
us.
A
A
A
D
A
Maybe
you
can
find
more
places,
foreign.
D
A
I
wonder:
well,
we
still
have
Quran
here,
so
he
used
to
work
on
our
team,
but
haven't
been
around
for
two
really
two
years
already.
I
need
to
double
check.
Why
why
it's
happening.
A
Okay
and
just
needs
a
review
and
approve
over
here,
first
yeah.
If
you
can
find
like,
maybe
see
if
there
are
other
places
when
the
same
change
needed.
So
you
wouldn't
do
it
one
by
one
sure.
A
B
A
C
Oh
yeah,
that's
helped
just
because
need
to
be
careful
about
them
as
much
as.
A
So
is
the
status
that
do
we
need
it
right
now
or
it's.
A
A
A
A
A
A
A
A
A
A
A
A
D
A
A
A
There
are
still
PR's
waiting
on
author,
but
it's
way
less
now:
okay,.
A
B
Okay,
I
sorry
I
had
a
PR
related
to
device
plug-in
test,
deflating
I,
don't
know.
If
any
of
you
could
take
a
look
at
it.
Brian
had
reviewed
it
and
it
still
waiting
for
an
LG
TM
and
an
approval.
D
At
it,
yeah
I
put
comments
on
it,
I
like
the
intended
fix,
and
there
are
some
slightly
controversial
extra
code
added,
but
I
think
it
was
a
value,
so
I
put
commented
accordingly
in
here.
B
It's
ready
for
review
or
approval
foreign.
A
Okay-
let's
see
maybe
during
qualities,
you
can
clean
up
more.
A
So
we
at
half
an
hour.
If
there
is
no
more
test
related
topics.
I
suggest
we
take
a
look
at
bugs.
We
have
in
three
hours
box
for
a
while
and
I
wonder.
If
anything
came
up
or
is
it
is
crazy.
Cool
company.
A
A
A
D
A
A
A
A
D
A
A
A
A
A
A
A
A
A
A
A
A
A
A
A
A
A
We
have
10
minutes
left
and
we
have
18
bucks
triage.
So
maybe
we
can
go
through
a
couple
more.
Is
there
any
other
topics
that
people
want
to
bring
up
before
we
dive
into
this
box?
No
okay
I
see
shaking
hands.
A
A
A
A
A
A
A
A
A
A
A
Okay,
it
seems
that
people
working
on
that.
A
Yeah
we're
out
of
time.
Thank
you,
everybody.
Every
time
we
miss
couple
meetings.
We
have
a
lot
of
that
to
go
through
just
churn
and
clear
adjectives
in
bugs
I
hope
that
we
will
get
out
of
this,
maybe
after
holidays
for
sure,
and
then
we'll
get
into
real
improvements
that
we
can
make
together.