►
From YouTube: Kubernetes SIG Windows 20181127
Description
Kubernetes SIG Windows 20181127
A
B
B
B
A
I
think
I
think
that's
the
right
approach
and
you
know
I
think
that
you
know
not
only
from
a
long-term
support
perspective,
but
also
from
a
feature
standpoint.
That
makes
sense.
Now.
There's
one
caveat
with
that,
and
obviously
this
is
we
don't
know
18:03
sorry,
19:03
should
come
out
sometime
at
March
timeframe,
correct
and
19:03
is
gonna
have
necessary
improvements
for
phenolic
for
flannel.
The
excellent
support
remember
some
of
the
changes
that
they're
doing
for.
A
We
were
told
that
it's
possible
that
David
basically
said
that
it's
possible
that
you
guys
my
back
poor
things
to
server
nineteen,
but
it's
not
guaranteed.
So
so
that's
that's
something
that
we
will
leave
the
door
open
for
that,
because
1903
might
be
a
necessary
support
matrix
for
1.14,
obviously,
pending
on
you
know,
if
it's
going
to
be
releasing
time
for
us
to
test
for
it,
but
there's
something
for
us
to
think
about.
A
D
D
E
A
A
B
Okay,
I
think
they're,
going
at
the
end
of
a
single
release
is
something
that
that
would
be
helpful
in
terms
of
in
terms
of
the
sakes
capacity,
because
you
know,
as
we
saw
in
13
now
that
the
release
bar
is
raising
on
us.
We
have
more
technical
debt
that
we
need
to
pay
and
we
don't
have
more
people
yet
and
so
I
think
we
just
need
to
be
very
careful
to
to
make
things
as
narrow
as
possible
to
succeed
in
14,
so
I
think
that's
kind
of
going
to
be
the
trade-off.
B
D
B
Okay,
can
you
see
that
that
edge
gee
doc?
Yes,
okay,
so
that
this
is
something
I
put
together
just
sort
of
lat?
Last
week
after
we
had
or
I
guess
it's
two
weeks
ago
after
we
had
the
discussions
with
with
cig
architecture,
but
what
I
wanted
to
try
doing
for
this
release
is
making
sure
that
we're
doing
a
little
bit
more
planning
upfront
and
actually
getting
people
assigned
to
to
focus
on
some
of
these
different
priorities
that
are
there.
B
Like
I,
don't
remember
exactly
what
his
words
were,
but
they
basically
said:
if
you
don't
have
enough
people
in
the
community,
you
need
to
fix
that
problem,
and
so
what
I'm,
looking
for
or
so
I
think
what
I'd
like
to
do
is
sort
of
say
you
know.
Are
these
sort
of
you
know
the
right
work
areas
that
we
could
potentially
split
up
and
then
can
we
set
volunteers
to
drive
each.
C
A
Yeah
we
need
to
do
that
for
sure.
I
think
that
you
know
put
a
lot
of
burden
for
you
in
the
last
release
and
you
know
we
need
to
get
some
volunteers.
I
will
drive
individual
items
there
from
your
big
list.
You
know
just
from
the
perspective.
You
know
we
don't
know
the
big
community
so
I'll
do
it.
You
can't.
A
We
can't
really
spread
ourselves
too
thin
either
there
were
a
couple
of
items
that
I
think,
like
maybe
the
Container
D
and
the
hyper
V
one
might
be
the
two
things
that
we
might
not
wanna
tackle
414,
but
it
can
try
to
tackle
everything
else
just
because
both
of
those
are
bigger
items
right.
So
if
we
want
to
do
the
hybrid
class
or
support
with
runtime
class,
which
I
have
as
well,
which.
C
A
Had
on
my
list
and
the
group
managed
service
accounts
to
start
running
production
workloads,
you
know
that's
like
three
features,
plus
the
stability
that
you
have
I.
Think
that's
that's
about
it
from
our
ability
to
to
deal
with
this
as
a
team
and
other
resources
from
from
my
company.
But
you
know
you
know
if
I
get
one
of
them
who
might
be
able
to
advances,
but
I,
don't
know
if
we
can
tackle
container
D
as
well
as
hyper
V
isolation,
there's
no
for
114.
B
B
B
B
B
A
A
B
B
B
B
D
D
F
F
B
A
We
can
do
that
so
maybe
put
a
little
table
there
with
you
know.
There
are
three
four
things
from
confirmed:
ones,
hybrid
classes
up
or
run
time.
Class
msas,
like
I,
said
for
my
personal
preference,
I
wouldn't
put
container
D
in
hyper-v
Beda
they're.
More
importantly,
I
know
that
I
notice
that
you
said
that
for
continuity,
we'll
discuss
in
a
future
in
the
future
sip
windows
meeting
so
right
now
we
just
don't
have
enough
data
about
that.
So
I
don't
know
if
you
can
spread
our
substitute
in
to
try
to
tackle
that
too.
B
Yeah
I
mean
I
think
in
terms
of
the
the
area
that
I'd
be
most
probably
the
best
to
work
on
would
be
your
continuing
to
work
with
with
that
Elena
and
Claudia
in
terms
of
driving.
You
know
just
the
test
effort
up
there.
I
could
probably
look
into
maybe
doing
the
unit
testing
stuff,
but
based
on
the
way
on
the
way
things
are
turning
out.
A
D
B
Yeah
yeah
well
I
think
that
for
the
container
D
I
think
that
wouldn't
necessarily
need
to
go
through
that
whole
process,
because
we
can,
you
know
if
there
were
people
that
could
get
that
work
done.
I
mean
CRI.
Implementations
are
out
of
tree
right,
so
I
think
we
could
just
say
you
know
we
support
it.
Here's
here's
a
link
to
the
code,
but
I
think
that
the
hyper-v
support,
because
it's
going
to
rely
on
runtime
classes
and
the
group
managed
service
account
support
since
it's
an
API
change.
Those
would
all
need
to
go
through
that.
D
B
A
The
bigger
goal
there,
what
they
wanted
to
do
is
run
time
class
was,
can
we
have
customers
avoid
having
to
target
their
specific
pot
specs
for
Windows
or
non
Windows,
and
maybe
use
the
runtime
class
to
make
smart
decisions
about
that?
That's
kind
of
what
we're
trying
to
solve
and
and
we're
trying
to
avoid
having
tea
stains
and
toleration
x'
to
do
that,
so
I
guess
the
biggest
thing
here
is.
A
F
One
thing
that
I'm
not
sure
is
that
I'm
not
sure
110
class
would
be
ready
in
114,
with
all
the
features
that
we
want.
So,
instead
of
just
betting
on
mint
and
cost
to
do,
oh
then,
maybe
we
should
look
into
other
methods
that
would
probably
improve
this
user
experience
and
scheduling
and
everything
yeah.
A
So
I
mean
the
other
alternative
that
we've
discussed
was
to
basically
put
a
taint
on
all
of
the
windows
nodes
and
have
that
ten
medically
applied
by
the
cubelet,
at
which
point
no
pods
will
be
scheduled
on
it
unless
they
have
the
Toleration
and
then
the
owners
will
be
on
the
windows:
application
developer
in
kubernetes
to
not
only
put
a
node
select
or
the
targets
windows,
but
also
add
the
Toleration.
I
think
that
plan
would
work
no
matter.
A
What
the
caveat
here
is
that
you
have
to
do
add
two
extra
things
to
your
pods
back,
because
that
Linux
developers,
don't
have
to
add
I,
think
that
while
we
are
runtime
class
can
satisfy
your
needs
in
the
long
term.
We
can
invest
in
that
and
in
the
meantime,
in
the
short
term,
if
we
do
require
developers
to
do
a
couple
of
extra
steps,
I
think
that's
not
a
that's
an
acceptable
solution
for,
for
a
small
period
of
time,.
A
You
know
if
you're,
if
you're
a
private
registry,
so
being
able
to
inspect
images
and
figure
out
what
OS
version
know
that
not
only
do
you
have
to
figure
out
their
West,
but
also
there
was
version
because
that's
very
dependent
on
well
I
guess
if
we're
only
supporting
2019,
that's
not
as
big
of
a
point
but
having
the
ability
to
do
that
might
be
a
much
bigger
undertaking,
I'm,
not
against
it.
If
someone
wants
to
be
a
POC
and
see,
how
is
it
that
is
but
dealing
with
private
registries,
we're
gonna
have
to
solve
that.
D
F
You
don't
really
have
to
expect
private
registry
for
this,
so
much
there's
some
hints
in
the
pass
back
that
can
indicate
that
this
is
for
Windows
and
the
mission.
What
hook
can
actually
applied
foundations
and
everything?
So
it's
just
a
way
to
help
users
use
this
feature
easier
and
I'm
just
asking,
because
this
is
listed
as
a
release
required
item
I'm,
not
sure
that
is
actually
we
can
achieve
meaningful
things
in
this
release
with
rental
cost
that.
A
So,
essentially,
the
approach
there
is
in
well,
you
have
a
demon
side
response,
Windows
and
Linux
I,
don't
even
know
how
that's
gonna
work.
Just
from
the
perspective
that
you
were
for
the
ones
that
you
only
have
one
specification,
so
single
image
will
not
be
able
to
be
scheduled
on
both
Windows
and
Linux
nodes.
A
H
H
C
There
aren't,
like
oh
I,
looked
into
the
the
new
layout
of
what
you
guys
are
talking
about
what
the
daemon
sets
and
the
classes
and
everything
else
I
spent
like
two
hours
last
week
looking
into
it,
and
it
looks
like
the
people
using
Solaris,
for
example,
or
another
operating
system.
That's
Linux,
are
you
Linux
or
UNIX,
based
they
basically
flag
the
same
way
where
we
do
Windows
beta
note
equals
Windows.
If
that
makes
sense,
and
it
doesn't
seem
like
there's
anybody
else
currently
looking
into
the
docker
support.
C
If
that
make
sounds
like
into
the
containers
and
figuring
out
from
the
metadata
of
an
inner
registry,
it
doesn't
look
like
anybody
in
the
kubernetes
Club
you
go
system
currently
is
looking
into
that
that
deep
and
so
I
couldn't
find
a
way
to
actually
go
digging
to
it.
Unless
we
built
some
type
of
API
interface
between
docker
or
whatever,
we're
calling
you
know,
docker
or
its,
can
you
see
our
container
d
or
any
of
that
right?
Yeah.
A
C
A
A
About
the
a
the
the
container
itself,
but
if
they're,
if
the
repo
or
the
registry,
does
that
work,
you
only
do
it
once
right.
So
if
they
inspect
the
image
at
the
upload
time,
for
example,
they
can
add
the
metadata
as
part
of
the
image,
rather
than
ask
having
to
write
an
admission
controller
that
actually
takes
in
that
data
and
find
something
in
at
run
time,
figures
out
what
otherwise
you
have.
A
B
A
B
A
Off
very
quick
things,
so
thank
you
to
a
few
folks.
I
started
adding
the
customer
voice
I
into
Windows.
If
you
remember,
we
talked
about
that
last
week,
please
continue
to
add
more
things.
You
know
we
have
bigger
team
meeting
today.
So
there's
there's
13
of
us.
So
if
you
have,
customers
are
interesting,
Windows
Co,
add
them
to
the
customer.
Voice
is
linked
on
the
notes
from
last
week.
That's
going
to
help
us
kind
of
build
consensus
and
build
momentum
for
Windows,
and
and
that's
it
all
right.