►
From YouTube: Kubernetes SIG Windows 20180918
Description
Kubernetes SIG Windows 20180918
B
B
B
If
we
want
to,
you
know,
declare
Windows
stable
by
the
end
of
it,
because
we
need
those
results
running
there
and
then
the
couple
PRS
like
there
are
ones
for
network
metrics,
even
though
that
we
opened
them
during
12,
we
weren't
able
to
get
them
reviewed
in
time
and
then
once
they
reviewed,
and
they
said,
oh
well,
this
one's
got.
You
know
minor
changed
for
an
API,
it's
encode
freeze,
so
they
they
push
that
out
to
13,
so
that
one's
coming
coming
there
and
so
and
remain
in
terms,
are
reigning.
B
B
It
would
also
need
changes
in
the
CNI
plugins,
but
if
we
can
get
that
in
12,
that
would
be
great
because
then
the
CNI
stuff
can
come
later
without
having
to
wait
for
13,
otherwise,
I
think
if
it
moved
to
13.
That's
that's
unfortunate,
but
it's
it's
not
going
to
it's
not
going
to
be
a
huge
difference
there,
because
13
12,
where
I
could,
for
all
the
conformance
stuff.
B
No,
oh
yeah,
okay,
all
right!
Well,
we
could.
We
could
check
back
on
that
one
later,
so
that
was
just
me
a
little
bit
of
that
stuff
there
and
then
the
other
features
that
I
future
work,
that
I'd
proposed
around
implementing
runtime
classes
and
GMS
a
I
just
continued,
pushing
those
out
until
after
we
get
to
stable
for
the
rest
of
the
stuff.
So.
B
B
They
needed
it
to
be
able
to
to
do
bills
and
contribute,
and
so
yesterday,
I
went
ahead
and
started
trying
to
put
all
that
into
one
repo,
that's
right
there,
but
basically
what
I
was
trying
to
do
is
just
get
everything
that's
needed,
so
that
way,
people
know
how
to
how
to
build,
deploy
cluster
and
replace
binaries
on
there,
and
so,
if
anyone's
got.
You
know,
suggestions
on
other
things
to
add
to
this
or
ways
to
improve
it.
You
know
you'll,
please,
let
me
know,
or
course
PRS
are
welcome.
B
B
B
So
for
her
version,
thirteen
you
know
the
main
thing
I'm
going
to
be
focusing
on
is
trying
to
do
most
of
the
organization
stuff
with
other
SIG's
and
really
delving
into
things
like
the
API
Doc's
that
are
needed,
and
so
we're
I'm.
Looking
really
looking
for
more
help
is
you
know:
I've
got
some
people
at
Microsoft
that
are
I'm
helping
with
the
bug
fixes.
B
It
has
to
build
an
investigation,
but
I
think
that
there
are
are
still
more
more
bugs
to
find
that
just
aren't
yet
filed,
and
so
what
I'm
really
looking
for
is
people
to
go
ahead
and
start.
You
know
getting
this
stuff
deployed,
you
know,
run
the
tests
and
you
know
just
you
know,
sort
of
start
filing
and
fixing
bugs,
and
then,
if
we
look
at
the
what
we
already
have
over
here
in
the
the
kubernetes
repo
there's,
also
a
reasonable
backlog
that
we
need
to
start
going
through
as
well.
B
Yes,
we've
got
36
open
and
there's
already,
you
know,
hope
and
share
that,
like
you
know
that
could
be
looked
at
and
so
FEMA
wants
to
take
any
of
those.
You
know,
let
me
know,
and
we
can
you
know,
Michael
or
I
can
get
those
to
get
those
reassigned,
but
we
basically
need
to
start
going
through
this
list
and
you're
either
getting
things
closed
if
they're
not
relevant
or
start
getting
PRS
and
to
fix
the
fix,
the
remaining
ones
that
are
there.
B
B
B
You
know,
working
with
working
with
her
team
at
cloud
base,
4:30
using
cube
net
and
as
your
CNI
right
now
and
they're
working
on
ovn
next,
but
I
think
we're
basically
going
to
pick
what
the
settings
are
and
then
stabilize
on
those
for
for
upstream
testing
sometime
in
mid
October.
So
that
way
we
can
leave
that
you
know
it's
kind
of
a
constant
target
up
until
code
freeze,
which
is
probably
going
to
be
mid
mid-november
and
then
so
after
a
few
weeks
of
code,
freeze,
then
they'd
be
able
to
hopefully
release
for
cube.
B
Con
is
what
my
guess
is,
but
your
like,
as
I
mentioned.
This
is
still
a
guess
at
this
point,
because
there's
not
an
official
schedule,
but
if
they
want
to
get
something
done
by
cube
con,
that's
basically
what
we
of
what
what's,
what
sig
release
wouldn't
need
to
do
and
so
for
CRI
I've
got
a
link
down
here.
B
B
So
I
think
one
of
those
is
available
for
Windows
10
right
now,
I
need
to
double
check
and
then
hopefully
the
there
will
be
a
server
really
soon.
I
know
that
when
they
do
release
another
server
preview
bid
or
build,
it
will
be
you'll,
be
at
least
that
build,
if
not
newer,
and
so
that
should
be
able
to
work.
Okay
with
with
container
D
and
so
I'm
hoping
I
can
get
some
more
details
out
on
that
soon.
So
so
people
can
start
trying
that.
B
You
know
one
of
the
benefits
I
see
in
you
and
switching
the
container
D
model
is
that
docker
sham
is
something
that's
been.
You
know,
we've
been
maintaining
it
within
this
sig
for
the
purposes
of
Windows
support,
but
it
hasn't
really
been
getting
any
other
improvements,
and
so
my
suspicion
is
that
there
may
be
other.
You
know
Linux
specific
fixes
in
there
that
were
needed
to
pass
conformance
tests
that
Windows
just
never
got
because
the
tests
weren't
in
place
and
so
I
think
there's
also
I,
think
there's
a
chance
that
container
D
could
behave.
B
B
A
C
B
So
there's
a
PR
out
there
that
lets.
You
basically
switch
what
container
registry
you're
using
and
so
that'll
help
for
people
that
are
testing.
You
know
within
their
own
in
network
environments
as
well,
and
so
that's
basically
what
we
have
and
there's
a
PR
open
for
that.
I've
brought
this
up
with
with
the
cig,
testing
and
state
testing
infrastructure
and
there's
been
just
two
thoughts
here,
but
I
haven't
seen
any
any
action.
B
The
difficulty
that
we
have
with
a
lot
of
these
images
is
that
the
core
maintainer
are
not
supporting
Windows,
and
so
in
most
of
these
cases,
like
busybox,
is
a
really
good
example.
Someone
basically
took
a
point-in-time
snapshot
on
the
busybox
source
and
the
majority
of
the
work
that
they
did
wasn't
actually
changing
the
busybox
source.
B
It
was
creating
a
build
system
that
could
build
it,
and
so
that's
where
they
were
bringing
in
things
like
the
not
say
going,
but
the
emesis
runtime,
and
things
like
that
that
were
needed
to
be
able
to
build
that
source
code
along
with
the
you
know,
GCC
cross
compiler,
and
because
those
changes
were
so
drastic.
You
know
the
maintainer
z'
for
the
linux
stuff.
B
If
they
didn't
care
about
Windows,
they
just
aren't
taking
those
submissions
back,
and
so
one
of
the
things
I've
proposed
is
that
to
me
it
makes
sense
that
things
that
ku
Bernays
depends
on
should
probably
be
in
a
repo
where
the
kubernetes
project
can
maintain
them.
So
that
way,
let's
say
you
know
wait.
One
of
the
one
of
the
options
used
in
one
of
those
containers
was
deprecated
rather
than
us,
finding
that
with
a
broken
kubernetes
test
and
then
having
to
you
know
rewrite
it.
B
You
know
we
could
just
pick
a
version
for
that
container,
use
it
and
be
able
to
rebuild
it
from
source
if
needed,
and
so
like
I
haven't
gone
through
the
formal
proposal
on
that.
Yet
some
people
said
yeah.
Maybe
we
should
do
that
rather
than
depending
on
you
know,
publicly
available
images
that
we
don't
control,
but
I,
don't
really
know
what
a
path
to
a
decision
is
there.
B
You
know
I,
so
I
think
you
know,
one
option
would
be
focus
on
upstream
only
and
get
them
to
do.
Windows
builds
I
think
that
would
be
a
very,
very
costly
approach.
I
think
it's
possible,
but
I
think
that
I
don't
think
I
would
want
to
complete
it
with
kubernetes
conformance
because
that's
really
about
how
those
projects
are
choosing
to
to
manage
themselves
and
that
are
outside
the
purview
of
kubernetes.
B
C
No
I
I
didn't
really
have
any
thoughts
on
how
to
reconcile
the
problem.
It's
just
it
seemed
like
yeah.
It
just
seems
like
it's
still
kind
of
far
like
I
see
how
the
conformance
tests
are
working,
but
it
still
feels
like
it's
kind
of
like
based
on
a
bunch
of
like
kind
of
like
hand-rolled
images
that
you.
D
E
E
E
E
D
Plus
the
way
you
build
images
on
Linux
and
how
you
build
images
on
Windows,
especially
the
run
commands
in
a
docker
files,
can
vary
a
lot
in
a
lot
of
scenarios,
you're
mostly
using
bash
for
us
containers
where
those
things
will
not
work
properly.
For
I
mean
it's
containers,
that's
another
issue,
and
not
only
that
a
lot
of
images
are
using
visit
box.
D
D
E
So
again,
from
my
experience
going
back
into
work,
we
did
the
way
I
look
at
this
yeah.
We
have
these
images,
this
Windows
images,
but
trying
to
force
it
into
even
the
way
it
is
now
with
the
busybox
images
that
we
have
in
is
not
necessarily
proper
from
the
best
point
of
view,
we're
basically
making
them
work,
but
they're
not
really
specific
windows
tests
they're,
just
the
images,
are
just
a
way
to
make
the
best
work
on
Windows.
B
E
F
B
F
A
F
With
Flex
volumes,
but
one
of
the
other
developments
that
happened
was
like
CSI,
more
or
less
stabilized
during
112
and
113
is
a
flange
ei
for
it
mm-hmm.
The
thing
is
like
to
make
CSI
work
in
Linux
they're,
using
a
bunch
of
these
sidecar
containers
like
a
registrar
and
a
couple
of
other
ones,
to
stir
the
register
the
plugins
and
then
do
the
appropriate
handshaking
between
kubernetes
and
CSI
to
make
it
work.
F
So,
given
that
windows
will
also
G
and
113
RV
planning
to
kinda
roll
with
the
Flex
volume
model,
because
that's
completely
going
to
be
kind
of
put
on
a
frozen
mode.
So
if
there
are
bugs
that'll,
probably
not
like
you
know,
it'll
be
hard
to
get
to
push
those
peers
in
so
yeah.
Like
would
CSI
be
too
big
a
thing.
Given
all
the
other
things
we
need
to
kind
of
get
resolved
for
GA.