►
From YouTube: Kubernetes SIG Windows 20190129
Description
Kubernetes SIG Windows 20190129
A
A
No
basketball,
injuries
so
I'll
be
fine
in
a
few
a
couple
weeks,
all
right.
So
a
couple
of
quick
updates,
so
we've
been
there's
been
a
tremendous
number
of
threads
going
back
and
forth
between
us
when
I
say
us,
I
mean
me
Craig
and
Patrick,
Ben
and
Leia,
and
deep
and
on
across
the
two
caps
that
you
have
right
now
in
our
pipeline.
One
of
it
is
the
GMs
a
cap
and
the
other
one
is
the
windows,
not
cap.
A
It
looks
like
we
have
the
final
pr4
for
the
cap
and
Patrick
put
it
in
the
notes
as
well.
So
if
you
go
to
our
meeting
notes,
you
see
it
it's
the
pr
7:43.
We
are
trying
to
make
that
the
final
PR
before
setting
up
the
windows,
not
implementable.
Our
goal
is
today
and
we've
had
the
discussion
with
Brian
and
it
looks
like
we're
heading
in
the
right
direction.
From
the
perspective
of
being
able
to
approve
this
today
or,
if
not
today,
maybe
tomorrow
it
looks
like
they're
willing
to
grant
us
an
exception.
A
If
we
don't
meet
the
deadline
today,
I
think
the
onus
of
responsibility
on
them
to
review
it
and
the
feedback
we're
getting
is
getting
smaller
and
smaller.
With
every
review
cycle
we
are,
they
are
being
asked.
They
are
asking
us
for
a
pretty
high
level
of
scrutiny
across
both
the
documentation,
the
implementation
and
the
things
we
are
committing.
That
will
work
before
we
go
GA
and
that's
actually
a
very
important
point.
A
I
know
that's
a
lot
of
you
guys
that
don't
necessarily
contribute
code
and
that's
I'm,
one
of
them
as
well,
but
you
know
testing
things
and
making
sure
that
the
specific
scenario
works
or
the
specific
capability
works.
We
should
all
be
able
to
pitch
in
so
you
know:
go
look
at
our
board
and
pull
it
up
on
our
on
our
share
screen
a
little
bit,
but
go
look
at
our
board,
find
something
that
you
feel
that
you
can
actually
test
yourself
commit
to
it
and,
let's
get
it
done.
A
A
Some
of
these
things
at
the
base
level,
we
soon
need
to
verify
if
they
work
or
not
manually,
if
you
have
bandwidth
for
some
automation.
Obviously
that's
that's
better,
but
you
know:
I
want
to
hold
ourselves
to
a
standard
I
at
least
works
once
that
at
least
works,
and
then,
if
because
there
might
be
bugs
that
we
need
to
fix
in
the
product.
A
C
A
There
is
one
other
thing
that
hurts
a
little
bit
right
there,
some
of
the
node
tests,
I,
don't
know
windows
yet
and
I.
Think
some
of
those
tests
do
have
a
lot
of
coverage
for
some
of
these
areas.
I,
don't
know
when
we'll
be
able
to
render
some
windows
are
actually
don't
have
any
visibility?
No,
no,
but
you
know
maybe
we'll
introduce
duplication
or
maybe
we
need
to
actually
go
find
the
equivalent
test
and
signo
test
and
and
figure
out.
B
Yeah
I
mean
I
think
if
I
think
that
makes
sense
if
there's
existing
node
tests
that
are
there,
you
know
when
I
was
failing
issues.
It
was
I
was
looking
at
test
grid
because
I
wanted
the
reviewers
to
be
able
to
look
at
what
was
on
test
grid,
look
at
the
list
of
open
issues
and
have
a
complete
list
from
there.
If
we
can
resolve
one
of
those
saying
hey,
we
just
enabled
on
lower
test
in
two
passes.
That's
fine.
A
B
And
so
I
guess
my
ask
was
if
you
and
Craig
could
take
a
look
at
that
and
just
figure
out,
if
there's
anything
any
content
from
there,
that
you
want
to
pull
back
in,
because
I
had
a
few
more
samples
in
there
and
then
Leah
submitted
a
sample
showing
the
scent
the
secret
stuff,
and
so
I
think
that
the
PR
I
linked
to
is
probably
stuff
that
will
help
as
Doc's
are
being
written.
But
I,
don't
know
whether
or
not
you
want
to
keep
that
PR
long-term
yeah.
A
A
B
I've
been
discussing
this
with
with
with
Michael
and
Craig,
but
basically
they
let
kubernetes
org
members
create
projects
now
I'm
here
in
github,
and
so
it's
got
this
nice
little
feature
where
you
could
tie
it
to
a
particular
repo
and
then,
when
you
do
a
query
there,
this
one
is
just
going
to
pull
everything
from
the
kubernetes
kubernetes
repo.
We
can
easily
see
what
cards
are
not
on
the
board.
Yet.
B
Yesterday,
I
dragged
all
the
stuff,
those
tags
sig
windows
over
here
into
the
backlog,
but
we
haven't,
you
know
gone
through
and
triage
that
yeah,
but
the
process
that
we
had
discussed
was
that
for
each
for
the
current
release,
we
could
go
ahead
and
you
know
track
what's
on
the
backlog
and
required
for
that
release.
You
know
what
things
are
currently
in
progress
and
review,
and
then,
whenever
things
are
merged
or
closed,
they
automatically
move
over
here
into
the
done.
B
B
Notes
are
written,
so
you
know
we'll
continue,
adding
more
columns
to
the
end
over
here
for
past
releases,
but
it
gives
us
a
chance
to
at
least
go
through
and
you
know
mark
what
we
want
to
work
on
in
the
current
release
and
hopefully
have
one
place
to
to
track
that
list
of
what
of
work.
That
needs
to
be
done
for
the
further
release.
B
Some
because
I
know
the
release
team
was
asking
for
that
last
year
or
that
like
well
yeah
I
was
last
year
for
version
13,
so
the
things
that
I
saw
were
marked
as
milestone.
114
I
tried
to
put
all
of
them
in
this
list
here,
as
we
could
see.
Most
of
are
not
assigned
to
real
people
yet,
and
there
may
be
some
other
stuff
here
in
this
backlog
that
needs
to
go
into
14
as
well.
A
A
So
let's
say:
let's
take
one,
for
example
right:
they
need
headless
service
test
on
Windows,
so
click
on
that
Patrick,
so
so
that
one
a
headless
services
the
way
the
work
is
that
basically
they
don't
expose
an
IP
on
the
load,
balancer
so
also
you're,
a
service
that
basically
needs
to
be
discovered
through
some
other
means.
So
our
goal
is
the
first
part
is:
can
someone
write
an
app
and
publishers,
endpoint
somewhere
else
and
figure
out
if
we
can
actually
find
it
up
from
another
app
or
you
know,
does
that
work
probably?
A
Is
it
as
headless?
Services
are
defined
in
kubernetes
and
if
that
all
works,
that
the
part
one
which
is,
does
it
work
on
Windows?
We
can
check
that
off
and
can
actually
call
it.
The
second
part
is:
can
we
actually
create
a
test
for
it?
Maybe
that
person
that
did
the
initial
validation
doesn't
have
the
capability
to
write
the
test.
I,
certainly
don't
have
that
I've
never
written
a
test
in
kubernetes,
so
maybe
for
that
the
second
part
is
they
can
go
to
signal
and
ask
them
hey.
Do
you
have
any
tests
around
headless
services?
A
Can
I
see
them
and
then
we
can
look
at
them.
Is
it
possible
that
you
can
adopt
that
test
to
for
our
case
by
just
changing
the
type
of
image
that
you
that
you
use
or
some
others,
more
minor
change?
And
if
that's
not
the
case
either,
then
you
know
put
that
for
that
issue,
bug
in
the
queue
and
say:
hey
a
validate
works
for
Windows
there's
a
similar
test
in
Linux.
That's
it
has
XYZ
and
we
just
need
a
developer
to
come
and
adopt
it
whenever
we
can
and
then
we'll
make
a
decision.
B
B
B
So
that's
basically
split
out
what
I
had
on
that
project
board.
I
think
something
would
be
helpful
is
if
you
see
if
you
can
actually
get
to
the
board,
and
if
you
can't
you
know,
then
let
us
know
I
believe
everybody
should
have
read
access
to
this,
but
I
think
there's
a
chance
that
that
actually
moving
things
from
one
column
to
another
may
require
some
additional
github
group
members.
Memberships
I,
know
that
Michael
and
I
can
already
do
it
as
severe,
but
I.
Don't
know
what
I
guess,
I
guess.
B
C
A
A
B
Then
then,
up
within
that
group,
I
think
we
could
also
talk
about
what
the
schedule
should
be,
because
I
do
think.
It
might
be
helpful
to
consider
doing
something
that
overlaps
with
the
EU
time
zone
as
well
and
potentially
potentially
Asia,
because
we
also
have
Peng,
PHA
I,
think
that
could
help
with
some
of
them.
So
maybe
do
something
like
you
know
morning,
one
day
to
overlap
with
EU
and
and
afternoon
in
the
US
another
day
to
overlap
with
Asia.
Something
like
that.
A
Alright,
so
I
guess
you
know
from
from
next
steps
perspective
the
moment
of
truth
with
the
with
the
cab
is
probably
gonna
be
today
there
is
a
Sikh
architecture,
call
on
Thursday
patrica.
Does
sick
architecture,
Cole,
I'm,
gonna,
call
sick
class,
the
lifecycle
again
I,
remember
the
discussion
we
had
about
a
month
and
a
half
ago
with
them
where
they
said
they
don't
want
to
support
hybrid,
no
hybrid
heterogeneous
clusters.
So
I
want
to
resurface
that
back
saying,
hey
with
the
unaudited
reducing
the
cap
of
outline,
how
we
think
this
should
work.
A
B
B
B
B
A
A
E
F
So
basically,
the
CAF
PRS
are
really
hard
to
check.
I.
Think
I
love
some
comments
in
previous
videos
for
those
got
merged
and
then
I
didn't
go
back
to
them,
and
then
there
were
some
new
ones.
So
I
just
need
to
like
read
again
make
sure
the
CRI
part
in
the
cubed
dark
ocean
part
is
what
I
understood
as
time
or
actually
changes
or
not
so
yeah.
How
try
to
do
that
today
by.
A
Rating,
you
really
appreciate
all
your
help,
who
know
that
you
were
looped
into
this
in
the
last
minute
by
Dawn,
so
both
on
the
windows
node
cap,
as
well
as
the
group
managed
service
account
cap.
You
know
you,
you
came
in
last
minute,
so
thank
you
for
putting
the
time
to
review
it
on
the
GMS.
A
I
know
you
haven't
reviewed
it
yet
and
that's
what
Dib
is
waiting
for,
because
we
want
to
set
that
incremental
today
possible,
but
on
the
windows
node,
one
who
addressed
all
your
comments.
A
E
E
So
I
had
a
question
around
one
of
the
comments
I
made
around
this
effort.
That's
going
on
with
CSI
Magnussen
right.
So
basically
the
idea
there
was
I
was
just
speaking
to
Peter
as
well
on
slack
and
it
sounds
like
like
the
GCPD
works
today
with
Windows,
not
bad.
You
should
be
able
to
increment
ation
of
Windows
workload
is
able
to
use
gcpd
persistent
volumes
in
men
operating
under
the
GC
cloud
provider.
So
I
guess
my
question
was
once
we
moved
to
energy.
A
is
the
vacation
that
a
feature
like
CSI
migration.
E
Just
works
for
videos
cuz.
We
know
it
won't
write.
Cal,
CSI
grabbers
wanted
to
be
supported
on
windows.
So
do
we
have
any
thoughts
on
like
how
we
gonna
like
put
a
statement
or
on
that
kind
of
just
say
that
maybe
for
Alpha
CSI
migration?
It
won't
work,
but
hopefully,
when
it's
beta
are
getting
close
to
gr,
stable,
the
windows.
Support
comes
in
to
support
CSI
or
we
explore
another
path
and
CSI
where
the
node
plugins
are
not
containerized
so
that
it
works
in
Windows.
E
So
that
makes
sense
for,
like
you,
know,
CSI
overall,
but
then
there's
also
this
parallel
effort
going
on
called
CSI
migration,
where
the
work
is
basically
like.
Take
an
entry,
PV
provider
like
let's
say
you
know,
I
hear
disc
or
DC
PD
or
Amazon.
Yes,
and
basically
the
idea
is,
can
be
move
out
the
entry
code
and
basically
have
like
a
shim
layer
that
goes
out
to
the
CSI
plug
it
right.
So
you
have
everything
working
the
same
way.
A
E
The
code
that
acts
on
it
is
the
CSI
code
and
the
idea
behind
all
this
or
the
main
motivation
is
that,
then
you
can,
you
know,
start
removing
the
entry
code
at
defined
point
and
also
get
rid
of
all
the
vending
or
some
of
the
rendering
that
comes
with
it.
So
that's
like
the
overall
goal
and
the
idea
is
if
something
works
today,
with
entry,
for
example,
say
GCPD
and.
C
E
F
B
B
If
it's
written
down
someplace,
we
should
probably
just
respond
to
it
with
what
the
windows
plan
is
there
and
and
like
you
know,
and
in
this
case
I
think
that
as
it
stands
today
or
what
we
would
probably
say
is
that
the
windows
CSI
plugins
could
be
built
from
the
same
source
tree
a
lot
of
cases,
they're
just
going
to
be
things
that
could
be
crossed
built,
but
in
cloud
provider
setups,
you
know
like
like
GCE
or
orin
orin
Azure.
We
would
probably
just
pre
copy
those
windows
exe
files
into
the
VM.
B
At
the
same
time,
you
set
the
cubelet
and
other
things,
and
so
I
think
we
could
probably
just
outline
something
that
says
you
know:
here's
how
we
would
call
these
on
the
host.
It
doesn't
address
the
it
doesn't
address,
tying
them
to
a
container
lifecycle,
but
I
think
it
could
at
least
less
migrate
to
that
API
and
a
shim
that.