►
From YouTube: Kubernetes SIG Windows 20180821
Description
Kubernetes SIG Windows 20180821
A
B
Ok,
so
I
got
up
this
morning
and
started
doing
sort
of
an
update
on
on
what
the
progress
had
been
over
the
last
few
weeks
in
terms
of
getting
the
release
plan
together
to
promote
Windows
to
GA,
and
so
I
can
go
through
that
in
more
detail,
I've
got
one
fix,
I
want
to
call
out.
That's
got
a
back
port
to
111
that
we
should
talk
about,
and
then
it
looks
like
Peter
just
added
some
stuff
as
well.
B
C
B
So
in
terms
of
open
bugs,
we
need
to
have
those
the
ones
that
are
going
to
be
important
for
this
release
either
have
PRS
open
or
make
sure
that
we've
got
issues
marked
as
as
critical,
where
they're
still
fixed
coming
before.
We
can
get
things
in
for
code
freeze
and
then
also
dock
PRS,
where
placeholder
PRS
were
supposed
to
be
open.
Yesterday,.
B
B
So,
moving
on
down
the
list
over
the
last
two
weeks,
I've
basically
taken
the
initial
release,
criteria
at
four
different
SIG's,
signo,
networking,
storage
and
architecture,
and
and
got
a
lot
of
really
good
feedback
on.
You
know
what
they'd
like
to
see
before
we
get
windows
to
GA,
but
unfortunately
for
a
timeline
that
sort
of
puts
us
up
against
this.
This
freeze
that's
coming
up
really
kind
of
too
tight.
B
A
I
mean,
from
my
perspective,
if
the
stability
is
not
there
at
a
hundred
percent,
we'll
be
doing
a
disservice
to
our
to
our
customers.
If
we
were
to
release
it,
so
I
completely
agree
with
some
of
the
consensus
you're
sharing
that
without
notes
going
down
or
some
other
bugs
and
are
even
found
or
not
found
yet,
and
the
fact
that
we
don't
have
a
super
high
passing
rate
I
would
say
we
need
to
push
this.
It
needs
to
be
done
right.
B
C
Okay,
agree
with
Michael,
it
might
be.
A
good
idea
to
just
I
was
just
going
to
some
of
the
tests,
and
I
was
maybe
focusing
on
some
of
the
six-story
des
cuz.
That's
one
of
the
areas
where
I
have
some
background,
and
you
had
some
questions
about
that.
Like
seemed
like
a
lot
of
them
are
skipped
the
ones
that
are
on
the
cloud,
so
wasn't
quite
sure,
like
house,
and
all
the
scenarios
are
final
workout
but
yeah.
C
F
F
The
only
way
it
will
get
back
up
again
and
with
a
reboot,
so
I
tried.
We
hit
this
problem
at
the
beginning,
but
then
they
seem
to
just
get
back
within
some
time.
So
let's
say:
if
they
don't
crash
now,
it
would
say
around
10-15
minutes
and
the
Kuban
will
be
up
again,
but
now,
no
matter
how
much
time
the
node
is
down.
It'll
stay
there
forever.
There's
no
problem
with
the
node
itself.
The
Machine
we
still
have
connectivity
into
it.
I
can
still
RDP
into
it,
but
the.
E
B
E
B
F
D
B
B
B
Okay,
so
just
a
quick
touch
on
the
follow
ups
that
were
there
so
the
first
one
was
they
city
architect.
Cig
architecture
had
asked
again
for
a
way
to
track
which
api's
did
not
work
on
Windows
and
so
I'll
go
ahead
and
open
it.
The
tracking
is
sorry,
they
called
I
think
an
umbrella
PR
here,
but
they
basically
provide
an
example
of
one
that
was
listing
a
bunch
of
known
issues
with
the
API
and.
B
B
And
then
the
other
thing
that
would
that
was
suggested
by
sig
node
was
that
you
know
for
all
the
stuff
that
that
we
are
supporting
on
Windows.
They
recommend
making
sure
that
our
Doc's
have
enough
examples
of
the
use
cases.
So
that
way,
you
can
see
all
the
most
common
features
that
people
are
looking
for
as
a
full
example.
So
those
are
kind
of
the
two
things
that
were
asked
for
from
a
usability
standpoint
and
then
from
sig
storage,
which
was
the
most
recent
meeting
we
had.
B
They
asked
us
to
look
over
the
CSI
spec,
because
they're
set
to
close
it
in
2018,
that's
a
typo
there,
and
so
they
asked
deep
to
be
the
primary
reviewer
from
the
sig
there,
along
with
us
and
so
I
wanted
to
know.
Who
else
would
like
to
to
join
in?
On
that
discussion,
because
I
want
to
get
a
separate,
probably
one-hour
schedule
sometime
this
week,
let's
have
one
that
you
wanted
to
join
Peter.
B
B
B
B
So
that
was
the
storage
asks
that
were
there,
the
other
one
is
they're,
taking
a
lot
of
the
common
CSI
functionality.
That's
needed
to
do
things
like
enumerate
and
mount
and
format
volumes
and
putting
that
into
a
common
library,
and
they
didn't
have
any
recommendation
whether
that
should
be
cross-platform,
but
they
defer
to
us
to
say
you
know:
do
we
want
to
put
the
windows
code
in
there
or
not.
C
Actually
still
developed,
it's
probably
in
the
private
but
idea
there
is,
you
know,
let's
say
for
Windows:
we
would
have
something
there,
just
like
there'll,
be
a
library
for
NFS,
we'll
just
add
something
for
us
and
be
there
to
make
sure
our
windows
BOTS
can
access
it.
Okay,
also
going
to
be
a
common
library
for
I
scuzzy
as
well,
and
that's
one
of
the
ones
we
have
to
make
sure
that
it
works
on
Windows.
B
B
C
B
So
in
terms
of
getting
things
merged,
Adaline
has
got.
Pr
is
open
for
cube
test
the
test,
grid
jobs
and
then
getting
to
the
daily
test
scheduled
through
prowl,
so
I'll
be
working
with
her
on
the
account
side
of
that
to
make
sure
that
we
hand
off
the
service
account
that's
needed
for
those
to
run
from
the
central
prowl
instance,
and
that
won't
work
to
get
those
merge.
B
The
other
thing
I
want
to
mention
is
that
we
now
have
part
of
the
fix
ready
for
review
for
the
dnh
DNS
search
suffixes,
so
Peng
PHA
got
a
PR
open.
That's
going
to
use
the
CNI
capability
arguments
to
be
able
to
add
the
windows
specific
ones
there
and
that
will
need
to
do
corresponding
updates
for
each
of
the
CNI
plugins.
So
that's
available
for
review.
B
If
you're
working
on
C
and
I
plug-in,
you
know
please
review
and
comment
on
that,
because
I'd
like
to
get
that
merged
for
412
for
sure,
even
if
the
rest
of
the
C&I
providers
aren't
fixed
because
since
it's
an
addition
of
a
new
flag,
it's
not
going
to
break
backwards
compatibility
within
the
existing
ones.
They
just
won't
have
the
complete
DNS
search.
Suffix
list
set
correctly.
B
So
it's
going
to
happen
afterwards
and
then
we
still
need
to
get
the
hyper-v
isolation
refactored
to
use
the
new
alpha,
a
runtime
class.
That's
there
in
1.12,
so
they've
formalized
how
they
want
that
to
work,
and
so
we
can
deprecated
our
experimental,
annotation
and
replace
it
with
that
and
I'll
sync
with
ping
Fei
on
that.
D
B
B
D
D
G
G
G
G
There
might
be
something
useful
for
anyone.
Yeah
in
that
page
about
Windows
nodes
might
be
helpful
for
anyone
anywhere
I
running,
even
if
you're
not
running
on
Google,
it
uses
wincy
and
I
and
our
Google
Cloud
routing,
so
it
doesn't
use
oviya
and
it
doesn't
use
flannel
it's
building
on
the
scripts
that
Microsoft
provided
that
I
know.
Madam
and
another
folks
at
Microsoft's
have
shared
eventually
I
want
to
point
it
to
build
on
each
CS
and
go
library,
but
I
haven't
had
a
chance
to
today.
Mm-Hmm.
G
G
Using
Wincy
and
I,
so
just
the
binary
that
that
microsoft
is
shared,
yeah,
there's
no
I
haven't
made
any
I,
don't
actually
know
exactly
how
when
C
and
I
works,
but
it
works
fine
for
a
basic
allocating
on
GCE
and
I
haven't
had
to
implement
anything
GC
specific
in
this.
You
know
in
binary.
B
E
C
F
F
F
But
I
guess
you
are
so
basically
some
tests
are
marked
as
specific
for
a
provider
correct.
So
therefore,
as
it
says
that
g
c
gk
e
and
q
mark,
but
not
local
comments
is
basically
an
artifact
the
girl
left
over.
So
basically
when
we
compile
the
list
initially
we
manually,
so
we
basically
we
created
the
fossa
manually
and
then
again
that
manual
launch
using
cube
tests
with.
F
E
E
D
D
B
So
so
submit
discussion
actually
happening
over
in
cig
architecture
and
so
I'm
planning
to
continue
staying
there
for
those
for
those
discussions,
I
haven't
seen
a
specific
list
of
test
cases.
They
want
to
add
yet,
but
they
generally
agree
that
they
want
to
make
sure
that
the
conformance
test
is
covering
all
the
all
of
what
people
actually
use
for
in
a
scenario
standpoint
without
being
cloud
provider
or
operating
system
specific,
and
so
they
you
know,
is
it
the
generally
agreed.
You
know
with
that
stuff
in
principle,
but
there's
not
a
specific
list
of
test
cases.
D
E
B
Yeah,
so
we
haven't
had
any
specific
discussions
on
those,
yet
I
think
probably
probably
would
be
most
helpful.
There
would
be
to
take
a
couple
of
those
and
I'd
say:
let's
just
see
if
we
can
get
a
PR
through
through
cig
testing
or
at
least
get
it
up
there
for
discussion
just
because
people
have
you
agree
that
that's
sort
of
a
problem
but
nobody's
been
willing
to
take
any
concrete
steps
there.
So
I
think
we
need
to
just
sort
of
be
the
first
mover
there.