►
From YouTube: Kubernetes SIG Node 20200218
Description
Meeting Agenda:
https://docs.google.com/document/d/1j3vrG6BgE0hUDs2e-1ZUegKN4W4Adb1B6oJ6j-4kyPU
B
C
D
D
So
I'm
working
with
Victor
and
basically
I'm
just
a
sec
okay,
so
the
first
one
is
about
the
multiple
normal
container,
basically
I'm,
adding
a
few
new
few
end
to
end
tests
to
exercise
that
apology
manage
with
more
than
one
container
per
pod
and,
of
course,
more
than
one
per
pod
and
others
I
have
another
another
noteworthy
one
about
upgrade,
which
I
would
like
to
get
a
design
review.
Meaning.
D
Okay,
is
this
the
upgrade
flow
we're
looking
at
I'm,
not
sure
we
I
captured
the
request
about
testing
the
great
flow,
and
this
also
pulled
a
man
refactoring,
which
I
think
is
the
lower
priority.
One
also
I
would
like
to
stress
if
some
test,
if
you
people
mean
believe
some
tester
missing,
just
add
the
note
in
the
review
and
let
them
I
will
add
that
I
eventually
figured
out
the
way
to
test
to
exercise
the
policies.
D
Because,
right
now
we
focused
on
the
most
strict
one,
the
single
numa,
an
old
and
I'm,
going
to
pull
up
man
to
put
up
the
DPR
and,
like
tomorrow,
most
likely
again
review
and
that
will
complete
the
coverage
and
that
will
build
on
the
infrastructure.
We
added
with
the
PRS
already
posted.
If
there
are
any
questions,
I'd
be
happy
to
answer
so
just.
B
C
Well,
that's
there
that's
right
for
these
two
PRS
that
Francisco
was
talking
about.
There
are
some
functionality
PRS
at
the
bottom
of
the
list.
If
you
look
down
toward
the
bottom,
there's
three
that
need
attention-
and
you
know
one
of
them-
is
if
you
look
at
the
one
at
the
bottom,
where
Nolan
is
it's
more
about
guaranteed
guarantee
a
lot
of
resources
across
multiple
containers?
The
PR
right
up
here
at
the
top
is
a
test
for
that.
To
make
sure
we
can
align
resources
in
multiple
containers
in
a
pod,
but
there
are
some
functionality.
C
B
C
B
B
C
Have
refactored
on
that
reef
active
it
added
some
new
aware
tests
to
the
CPU
manager,
infant
tests,
so
you
know
this.
One
here
just
needs
an
approval,
so
we've
ran
it
locally
and
the
remote
and
on
multi
Numa
bare
metal-
and
this
is
passing
so
we
just
need
to
get
that
in
the
next
one.
Whoops
is
really
just
this
one.
C
The
next
one
on
the
list
is
eight
eight
one
one
zero,
and
this
is
some
code
refactoring
test
Francesco
mentioned
that,
and
it's
really
just
some
clean
up
stuff
for
your
organization.
It
looks
like
there's
already
Patrick,
said
he'll
check
on
this
regarding
Windows
and
give
some
feedback
there
Francesco.
You
want
to
add
anything
to
that
comment.
Four,
eight,
eight
one,
one:
zero!
No.
D
C
C
The
test
starts
and
let's
see,
cubelet
logs,
killing
the
hollow
nodes
and
after
a
bit
of
analysis,
the
theory
is
that
there's
some
strict
memory
limits
in
there
and
when
we
add
topology
manager
which
listens
to
pods
and
containers,
we're
adding
some
extra
memory
to
the
cubelet
and,
as
a
result,
the
containers
are
being
killed.
What
I
have
done
is
pushed
a
debug
PR
down
below
and
it's
actually
need
to.
Okay
to
test,
but
we
got
that
so
it's
running
the
test
now
to
see
in
what
I
did
was
updated.
C
One
parameter
so
I
labeled
the
topology
manager
feature
and
also
updated
increased
the
memory,
the
cubelet
kill
memory.
If
you
look
at
that
PR
to
see
if
that's
really
what's
going
on,
it's
been
difficult
to
pinpoint
exactly
why
cupid
is
killing
the
containers.
I,
don't
know
if
it's
the
kernel
out
of
man
but
I've
looked
at
the
current
log,
there's
not
enough
logs
in
there,
at
least
for
me,
to
figure
out
why
they
bother
hollow
nodes
is
not
going
to
ready
State
I've
reached
out
to
six
scalability
and
suggesting
for
some
help.
B
B
B
A
So
the
the
issue
that's
linked
here.
This
is
Kevin's
issue
that
he
created
Kevin
Clues.
So
there
was
three
possibly
four
PRS
already
merged
in
this
I
think
there
you
did
some
of
the
review
and
on
them,
and
so
a
lot
of
the
groundwork
has
already
done
on
this
book.
Kevin
is
actually
on
vacation
now.
So
he
asked
me
just
to
take
over
that.
There's
two
remaining
PRS
there's
one
large
one,
and
so
this
was
to
do
it
and
guaranteeing
the
hints
generated
from
multiple
containers.
A
So
it's
all
explained
in
the
issue
and
each
each
PR
is
explained
or
what
they're
doing,
but
the
last
one
is
kind
of
a
large
one.
Like
I
said:
Kevin
had
to
leave
us
and
he
asked
me
just
to
take
it
over
and
there
was
some,
maybe
I
think
maybe
I
had
a
five
or
six
commits
to
the
end
of
if
you
are
just
tie
it
up
and
so
that
that's
that's
open
for
review.
A
A
At
that
time
talked
about
it's,
it's
eight,
seven,
seven,
five,
nine!
So
that's
the
one!
That's
open
for
review
the
one
after
that,
and
so
that
has
some
of
the
commits
that
are
in
the
previous
PR,
so
that
that
will
need
to
be
rebased
once
the
one's
eight
seven.
Seven,
five
nine
is
actually
merged
and.
D
A
We
just
broke
them
up
into
two,
because
there
was
one
change
in
the
second
one
which
is
up
in
Cuba
taco.
So
we
just
we
took
that
a
nice
letter
into
its
own
PR
just
for
convenience
sake,
so
yeah
right
now,
the
eight
seven
seven
five
nine
is
kind
of
the
big
one
we
want
to
just
get
over
the
line.
Once
that's
once
that's
words,
then
we
can
rebase
the
second
one
and
yeah
after
that,
then
I
think
is
just
to
feature
Gator
and
we
should
be
okay.
Okay,.
A
I
haven't
actually
I
haven't
retested,
because
I
was
expecting,
I
might
have
to
push
some
more
cords
and
I'm,
not
sure
if
those
tests
are
just
flaking
or
not
so
yeah
I
can
I
might
just
retested
and
see.
But
I
was
expected.
I'm,
probably
a
push
more
code
because
it
hasn't
been
reviewed
as
a
notion
that
might
be
requests
come
and
Kevin.