►
From YouTube: Kubernetes SIG Windows 20190226
Description
Kubernetes SIG Windows 20190226
A
Hello,
everybody
and
welcome
to
another
sig
windows.
Meetup
were
actually
in
the
last
two
weeks
of
the
release.
Actually
a
little
less
than
two
weeks
code.
Freeze
is
coming
up
very
fast
towards
the
tail
end
of
next
week
and
as
well
as
were
marching
towards
a
ga
release
or
a
stable
release
of
Windows.
There's
a
lot
of
work
and
a
lot
of
a
lot
of
churn
going
on
in
our
PRS.
A
So
I
really
appreciate
both
Patrick
and
I
really
appreciate
all
the
work
that
everybody
is
putting
in
both
on
and
off
work
hours
to
make
forward
progress,
and
you
know
your
your
critical
parts
of
the
success
of
the
seep
called
little
bit
of
congrats
to
deep.
The
the
first
PR
for
GM
has
emerged
that
one
fairly
well,
so
that
was
that
was
awesome
to
see
that
go
through
Jeremy
as
well.
It's
as
other
folks
that
worked
on
that
PR
on
CD
born,
but
I
did
pass
him
the
congrats,
the
PR
merge
last
night.
A
So
let's
go
through
our
agenda.
The
first
item
and
I:
don't
have
it
on
the
list.
Here
was
the
GC
art
that
IO
upload
of
images
and
does
appear
that
Claudio
had
created
I
kind
of
wanted.
You
know
what
the
latest
status
on
that
I
know.
The
Patrick
and
I
talked
about
that
late
last
night
around
the
distribution
based
on
licensing
of
Windows
images.
A
B
C
B
A
A
D
B
D
C
C
D
A
Maybe
the
easiest
way
to
just
get
moving
forward,
so
we're
not
stuck
on
this
or
we
have
let
a
lil
bit
over
a
week
to
finalize
it.
Can
you
just
clone
the
test
case
ban
and
just
have
a
point
to
whatever
you
upload,
whatever
that
image
upload
works
and
and
so
can
make
forward
progress?
I
know
it's
not
ideal.
We
have
two
test
cases
doing
the
same
thing,
but
just
phone
into
different,
repos
and
or
at
least
different
versions
of
the
image,
and
then
we
can
fix
that
later
sure
I
mean
eventually.
A
A
Okay,
all
right
so
so
Ben
you're,
you're,
unblocked
and
Claudio.
On
your
end,
can
you
message
ting
Hawking
on
slack
or
do
you
want
me
and
Craig
to
ping
him
and
see
if
we
can
actually
get
him
like?
If
you
can
get
me
on
slack
or
a
quick
message,
what
it
is
that
you
want
him
to
do
exactly
give
me
the
manifest
list
that
you
need
him
to
upload.
I'll
start
a
conversation
with
him
on
on
slack
and
see
if
we
can
actually
get
him
to
do
the
work
today.
B
A
A
All
right
next
item
so
on
to
28th
we're
meeting
with
Brian
grant
to
discuss
our
efforts
towards
GA
and
kind
of
kicking
the
heads
up
on
where
we
are,
and
some
things
are
working.
Some
things
are
not
working
under
the
premise
that
he's
not
surprised,
come
code
freeze
and
we
kind
of
get
his
blessing
that
yes,
who
are
on
track
for
GA,
and
you
know
all
the
other
peripherals.
A
big
part
of
that
obviously
is
gonna.
C
C
E
E
So
we
have
two
two
problems:
we're
facing
the
major
one
and
the
reason
for
the
blocks
of
red
is
the
basically
cube
pellets
re
restarts
itself.
It
gets
a
pending
code,
division
by
zero
error,
it
panics
and
it
restarts
itself
after
it
restarts
itself
the
cni
agency,
and
I
will
be
basically
blocked.
It
will
not
respond
to
or
any
in
AD
command
or
any
anything
like
this.
So
basically,
any
part
of
Lu
will
try
to
schedule
afterwards
will
just
not
start,
thus
the
whole
failed
cases.
E
A
E
E
Every
time
is
the
same
division
by
zero
problem.
However,
sometimes
they
just
and
I
stopped
working
before
we
saw
the
first
division
by
zero,
but
for
the
agency
and
I
I
think
we
can
narrow
it
down
to
the
cubelet
restarting
so
I'm
convinced
that
that
is
the
problem.
This
division
by
zero
is
just
something
that
induces
it,
but
we
do
have
to
figure
it
out.
So
I
have
to
look
closely
into
it.
E
A
F
A
C
F
G
C
G
E
G
C
Okay,
so
we're
looking
at
that
issue,
one
of
the
possible
mitigations
that
we're
looking
at,
which
is
why
I
want
to
actually
talk
to
Peter
on
the
on
the
Google
results,
is
that
we
also
have
a
plugin
that
is
more
like
cube
net.
That
uses
something
called
wind
bridge
and
I.
Believe
that's
what
Peters
using
over
in
these
results?
C
E
Now
these
this
is
affecting
a
couple
of
tests.
It's
definitely
an
issue,
for
you
know
real-world
scenarios,
but
for
the
test
we
can
work
around
it.
In
the
images
mean
we
can
pull
and
wait
for
it
to
have
Network
before
we
do
any
network
related,
related
operation
or
just
start
the
entry
point.
So
that's
a
little
less
severe
in
that
we
have
a
workaround,
but
it
it's
still
not
ideal
that
it
happens.
Okay,.
C
That
also
reminds
me
that
this
test
case
here
that
says
DNS
should
provide
DNS
for
the
cluster.
We
have
a
PR
to
fix
that
and
that
was
actually
just
merged,
and
so
that
should
be
in
the
next
test
last
year
and
basically
that
test
case
was
covering
some
Linux
specific
functionality
around
Etsy
resolve
comp,
and
so
that
was
separated
out
into
a
separate
test
case,
and
so
the
one
that's
there
remaining
should
should
turn
green
on
Windows
and
you'll
see
that
same
failure
over
on
the
GCE
results.
H
C
C
H
A
E
E
A
F
So
we
actually
have
one
problem:
we
haven't
actually
had
Cuban
resource,
but
we
had
no
crashing
so
we
have
block
of
res
because
of
that
and
we
got
a
memory
dump
file,
but
we
haven't
actually
figured
out
what
exactly
it
was
causing
that
so
sometimes
a
node
will
just
reboot
and
that
would
fail
like
about
four
or
five
tests
and
we've
been
trying
to
disable
Windows
Defender
and
see
if
that
would
actually
help.
So
it
was
still
trying
to
figure.
I
was
going
out
with
that.
Okay.
I
H
J
I
C
F
So
we
just
tried
to
say
going
this
last
night,
so
haven't
had
a
chance
to
look
at
the
new
results
and
see
the
same
problem,
and
we've
also
had
some
flaky
creating
passing
box
issues
and
we'll
be
meaning
to
try
to
get
down
to
the
bottom
of
that.
But
it's
still
like
flaky.
Sometimes
when
you
try
to
create
a
pot
and
just
fail.
Okay,.
A
B
B
A
B
A
D
A
A
J
J
A
C
Yeah
so
I
think
the
main
feedback
I
have
from
a
Windows
standpoint
on
this
is
I,
mean
I
agree
with
the
sentiment
that
removing
dr.
shim
from
in
putting
it
into
a
separate
CRI
repro
is
the
right
thing
to
do,
but
I
think
that
once
we
get
container
D
working
since
it's
using
the
newer
windows,
api's
I
think
we
would
rather,
but
basically
I,
think
we
should
accelerate.
C
C
Okay,
excellent,
yes,
I
think
we
I
think
we
straight
to
continue
working
with
working
with
him
on
that
stuff.
But
I'll
need
to
talk
with
him
about
the
organizational
side
of
that
in
terms
of
getting
getting
releases
and
sync
with
the
team,
that's
working
with
container
D
and
see
if
that's
something
you
can
do
or
if
we
need
to
get
someone
else
on
that.