►
From YouTube: Kubernetes SIG-Windows 20230228
Description
No description was provided for this meeting.
If this is YOUR meeting, an easy way to fix this is to add a description to your video, wherever mtngs.io found it (probably YouTube).
A
I
will
start
with
some
of
the
announcements
and
move
on
first
announcement
is
code.
Freeze
is
March
14th,
which
is
in
exactly
two
weeks
from
today.
C
It's
become
easier
now
in
some
ways,
because
we
only
need
Cuba
changes,
because
apparently
this
is
the
thing
that
I
didn't
fully
understand
during
the
cigarch
call.
I,
don't
know
whether
you
got
that
Mark
is
like
they
don't
even
want
the
cube,
CDL
changes.
They
want
all
the
cube,
CDL
changes
into
like
a
plug-in
or
something
like
that
which
I
can
write
after
the
code
freeze,
even
so
I'm
just
focusing
now
on
the
cubelet
changes.
A
C
Yeah
I
don't
think
it's
needed
as
an
initial
knowledge,
but
I
yeah,
like
I,
came
away
from
the
cigars
call
like
not
100
sure
whether
the
cube
CTL
changes
were
needed
or
not,
but
so
I
just
threw
it
in
there
and
both
mache
and
and
Jordan
said.
No,
you
shouldn't
do
any
Cube
CTL
changes.
We
don't
want
this
exposed
in
anyways
through
official
means,
so
I
I
I've
dropped
those
changes,
but
he's
asked
me
for
a
bunch
of
other
changes
that
I
hopefully
can
make.
A
Okay,
okay
sounds
good
yeah,
let
us
know
what
you
need
help
and
then
the
another
announcement
is
so.
The
kind
of
lazy
consensus
period
ended
for
the
decision
to
promote
Arvin
to
Sig
windows
co-chair
so
today,
so
we'll
be
kind
of
merging
out
those
PRS
and
getting
that
done.
I.
A
Yeah
two
weeks
ago,
at
the
community
meeting,
so
anybody
who's
interested
in
that
didn't
get
that
pretty
much
recording
and
yeah.
Thank
you
for
all
once
again,
Urban
for
everything
and
happy
to
have
you.
A
No
okay
doesn't
seem
like
it,
so
next
is
sometimes
we
give
space
if
there's
any
new
contributors
who
just
want
to
say,
hi
or
introduce
themselves,
I
see
at
least
one
but
I
think
they're
going
to
be
giving
a
demo
later.
So
you
wanna
in
that
introduction,
then
that's
fine,
but
go
ahead
now.
A
Pretty
sure
everybody
else
is
pretty
is
already
all
right.
Let's
get
into
the
agenda
so
yeah
I
guess
do
you
want?
Let
me
stop
sharing
because
I'm
guessing
you're
going
to
be
screen
sharing
and
go.
B
Ahead
and
yes,
yes,
I
will
be
screen
sharing
Just
moment
all
right.
So
let
me
just
hit
this
working.
Hopefully
this
works.
Hopefully
this
works
cool.
Can
you
see
it?
Can
everyone
say
that
sweet
cool
so
for
some
context?
B
B
You
know
it
can
set
up
a
Linux
controller,
Linux
worker
node,
a
Windows
worker
node
and
I,
was
speaking
with
Jay
a
little
earlier
in
the
week
about
whether
or
not
this
tool
could
be
used
for
entirely
entirely
Windows,
set
up
by
running
the
Linux
bits
on
WSL
now
I'm
going
to
show
a
quick
demo.
I
have
been
fixing
bugs
in
this
for
the
last
six
hours.
B
So
whether
or
not
the
the
demo
gods
are
going
to
shine
on
us
today
is
still
to
be
seen,
but
basically
I've
just
got
a
build
of
rkm
here.
This
is
a
this
is
just
a
server,
it's
just
running
in
the
other
room,
it's
not
a
virtual
machine,
so
this
could
easily
be
someone's
laptop
or
whatever
you
want
now.
B
The
normal
way
to
run
rkm
is
it
would
install
itself
as
a
background
service,
and
it
will
just
kind
of
continuously
run
the
background
set
up,
giving
you
these
blah
blah
blah
blah.
Unfortunately,
something
I
found
out
tonight
is
that
WSL
really
doesn't
like
it.
If
you're
running
a
system,
so
I'm
going
to
have
to
run
this
as
the
local
administrator.
Instead,
it's
basically
the
same.
The
only
thing
that
doesn't
work
when
you
do
that
is
that
host
process
host
processors
can't
run
a
system
because
the
administrator
can't
Elevate
to
that.
B
So
that's
something
that
needs
to
kind
of
get
fixed
eventually.
The
other
thing
that
currently
was
working
but
isn't
working
right
now
is
the
windows.
Pods
can't
seem
to
send
TCP
traffic
to
Linux.
B
Pods
I
did
fix
the
DNS
issue
where
they
weren't
able
to
resolve
DNS,
but
now
it
seems
like
TCP
is
playing
up
so
anyway,
this
whole
thing's
very
experimental,
but
hopefully
it
gives
you
an
idea
of
where
this
could
go
in
the
future,
with
a
bit
more
work,
so
I'm
just
going
to
run
the
I'm
just
going
to
run
the
rkm
in
controller
mode
and
I'm
mostly
running
it
as
administrator,
which
is
just
what
the
service
does
Now
by
default.
B
And
so,
if
we
start
this,
it's
going
to
do
a
bunch
of
stuff,
including
create
a
virtual
switch.
So
we're
going
to
lose
connectivity
for
just
a
moment
here
and
then
it
will
once
this
RDP
comes
back.
B
It'll,
basically
be
downloading
all
of
the
assets,
setting
up
all
the
components,
starting
everything
up
and
then
hopefully,
we'll
be
able
to
get
to
group
CTL
in
a
very
short
period
of
time,
all
right,
sweet
yeah.
So
the
first
thing
it's
doing
right
now
is
it's
going
to
download
the
WSL
Ubuntu
image.
B
This
is
going
to
take
a
hot
minute,
but
it
basically
the
the
way
that
RPM
Works
to
make
it
really
easy
to
basically
throw
away
everything
and
start
from
scratch
is
that
everything
gets
put
into
a
unique
installation
directory.
Hopefully
you
can
see
this
well.
B
Everything
gets
put
into
its
own
unique
installation
directory
that
is
generated
when
there
is
no
Active
installation.
So
it
checks
for
this
active
file.
If
that
doesn't
exist,
it'll
go
cool
I'm
going
to
set
up
a
new
installation
for
you,
it's
a
unique
path,
and
this
is
really
useful
for
making
sure
that,
like,
if
you've
ever
had
to
like
tear
down
the
kubernetes
like
install
and
it's
like
containers,
are
using
files
and
there's
file
locks
all
over
the
place
right
by
having
a
unique
install
path.
B
You
can
always
just
delete
the
active
file
and
then,
when
you
run
rkm
again,
you'll
get
a
brand
new,
install
folder
you're
not
going
to
run
into
any
file
locks.
It
also
means
it's
super
easy
to
uninstall,
because
it's
literally
stopped
the
service
and
delete
this
folder
and
you
that'll
delete
everything.
It
will
delete
all
the
containers,
all
the
state,
everything
you
can
reset
to
scratch.
So
that's
pretty
nice
in
terms
of
like
repeatability
and
now
we're
just
waiting
for
stuff
to
download.
At
this
point.
D
So
when,
when
June
showed
me,
this
I
was
like
okay,
this
kind
of
turns
the
whole
Sig
Windows
Dev
tools
thing
on
its
head
because
we're
like
well,
we
were
building
VMS
for
that
for
that,
for
that
sort
of
ephemeral
thing
but,
like
this
gives
you
a
completely
local
development
environment,
that's
Windows
native,
when
you
don't
have
to
delete
anything,
I
mean
I,
just
like
it
I'm
still
making
my
head
spin
that
you
could
do
this.
It's
really
cool.
C
B
That's
pretty
amazing
I
started
trying
to
set
up
kubernetes
with
rke
too
and
then
I,
because
I'm
running
Windows,
11
notes
I
found
out
quickly.
That
containerd
does
not
do
the
version
check
property.
So
I
was
like
okay
I
want
a
patch
container
D,
but
then
trying
to
get
a
patch
container
D
into
okay,
you
two
and
then
that
was
horrible
and
then
trying
to
uninstall
our
key
too,
and
it
was
like
I'm
gonna
delete
your
C
drive
and
I'm
like.
D
B
Not
great
so
that's
also
another
reason
why
I
went
this
like
completely
isolated
directory
installation,
because
I
want
to
make
sure
that
it
was
super
easy
to
get
rid
of
stuff
very
safely.
C
B
So
this
doesn't
use
Cube
ADM
at
all
yeah.
This
is
literally
doing
it
from
scratch.
So
it
is
downloading
all
the
individual
components
like
the
cube,
API
server.
B
Pretty
much-
and
this
is
also
why
it
doesn't
need
to
use
tokens
or
anything
like
that,
because
that's
just
a
coupe
ABM
thing
I
was
saying
this
to
Jay
the
other
day.
Is
you
can
just
as
long
as
you
have
asserts
everything
just
works,
so
it
does
so
RPM.
Does
all
the
certificate
provisioning
does
all
of
the
coupe
config
Generation?
B
It
runs
container
D.
It
sets
up
the
network,
it
sets
up
it
installs
Calico.
It
does
all
of
that
work
to
get
everything
running
so
yeah.
C
C
B
B
While
that's
all
that's
coming
up
to
show
you
roughly
what
how
this
this
kind
of
structured
so
basically
internally,
it
has
a
bunch
of
components
and
the
components
are
responsible
for
running
electricity
processes
and
just
kind
of
controlling
when
things
exactly
start
up,
because
they
also
have
they
have
interdependencies.
B
But
you
can
see
stuff
like,
for
example,
this
is
the
public
component
which
will
run
on
Windows
or
Linux.
It's
fairly
simple,
it's
basically
just
extracting
a
config
file
and
filling
some
stuff
in,
and
then
it's
literally
just
running
the
Kubler
process.
So
you
can
pretty
much.
B
Imagine
like
you
could
replace
this
fairly
easily
with,
like
you
know,
go
run
for
example,
or
whatever
you
wanted
to
to
just
run
the
kublet
from
like
your
source
code
and
then
it'll
be
using
your
own
version
of
it
and
there's
a
bunch
of
other
components
in
here
like
that
handle
everything.
So,
as
I
mentioned,
like
you
know,
provisioning
certificate,
setup,
XD,
all
that
sort
of
stuff
is
handled
as
components
cool.
So
this
anyway,
this
looks
like
it's
going.
B
B
If
we
go
so
group
CTL,
we've
got
a
big
git
nodes,
oh
wide
cool,
so
we
can
see
that
we
have
both
nodes
ready.
We
have
the
windows
10
or
the
windows
11.
It
is
labeled
as
Windows
10,
but
it's
the
windows
side
of
things.
We
also
have
the
WSL
cubelet.
B
Running
as
Ubuntu,
and
we
can
see
it's
running
the
the
WSL
kernel
and
all
that
jazz,
let's
go
have
a
look
at
some
pods.
E
B
I
have
things
but
it's
cool
there.
We
go
all
right
cool.
So
now
we
can
see
the
now.
We
can
see
the
pods
as
the
core
set
of
stuff
comes
up.
You
know.
D
B
And
call
DNS,
basically
it's
just
downloading
images,
and
so
at
this
point
yeah,
so
we
at
this
point
we
have
FD
the
API
server,
the
kubelets
and
Kube
proxy
or
running
in
WSL.
Then
there's
also
the
kublet
core
DNS
group
proxy
and
a
few
other
bits
and
pieces
all
running
on
Windows.
B
B
With
regard
to,
like
you
know,
at
the
moment,
we
don't
have
TCP
working
between
the
windows
pods
and
the
Linux
pods
and
I'm
suspecting
that's
something
to
do
with
how
the
switch
is
the
hypervis,
which
is
doing
routing
and
stuff
I,
had
to
use
quite
a
few
undocumented
commands
to
get
the
WSL
to
even
be
able
to
access
the
network
through
the
external
switch
that
the
overlay
network
is
also
on.
B
But
whatever
issue
is
there
it's
specific
to
running
in
this
mode,
because
I
do
have
my
actual
Network
that
I
actually
run
or
my
actual
cluster
actually
actually
run,
has
a
separate
Linux
as
it
runs
the
Linux
controller
on
a
separate
machine,
and
it
has
a
bunch
of
Windows
machines
that
are
just
worker
nodes
and
all
the
interconnectivity
works
fine
there.
So
it's
it's
something
specific
to
do
with
running
an
overlay,
Network
and
wsl2
off
the
same
hybrid
switch.
B
Yeah
it
could
be,
it
could
be
any
number
of
things.
I
I
found
the
very
undocumented
vfp
control
command.
That
is
not
explain
anywhere
on
the
internet,
so
it's
quite
possible
that
there's
just
like
policies
or
something
that
are
missing
anyway.
If
we
look
at
okay
everything's
up
so
let's
go
and
make
some
make
some
pods,
so
Coupe
CTL.
B
It's
not
cool
so
now
we've
got
so
basically
I'm,
just
creating
nginx
and
Powershell
Nano
container.
So
we
can
see
pods
coming
up
on
both
the
windows.
B
Of
things
this
will
take
a
minute
to
basically
do
stuff,
but.
A
B
Mentioned
it
basically,
you
know
it
all
works,
just
with
the
exception
of
whatever
networking
stuff's
going
on
right
now
and.
D
B
D
D
B
B
D
C
D
D
Yeah,
you
do
your
own
control
planning.
You
just
run
rkm
to
join
to
it.
A
B
B
B
B
This
is
actually
downloading
the
the
Ubuntu
zip
and
then
basically
doing
the
setup
of
WSL
itself.
So
if
I,
just
actually
that
that
this
here
is
a
is
the
windows
pod.
But
if
I
look
in
the
distribution
list,
you
can
actually
see
it
sets
up
its
own
distribution.
B
So
it's
actually
separate
from
any
double
cell
distributions
that
the
user
has
on
their
machine
and
again
this
means
that
each
of
the
installs
is
completely
separate,
because
each
of
these
each
of
these
distributions
here
is
actually
inside
I,
believe
it's
inside
Ubuntu
Adobe.
So
yeah
like
it's
set
up
inside
here.
So
again,
if
you
delete
this
folder
you're
deleting
that
WSL
distribution
and
clearing
out
all
of
the
Linux
State
as
well,
when
you
switch
to
a
different
distribution,
so.
A
Yeah,
that's
really
awesome.
Yeah,
really,
nice.
D
Yeah
yeah
I
mean
I
just
had
I
have
no
questions,
because
this
is
the
second
time
I've
seen
this
and
I'm
just
trying
to
the
question
I
have
is,
for
us
brought
more
broader.
D
Like
we
should,
this
seems
like
it,
you
know
it
seems
like
after
a
few
months
we
could.
We
could
start
adopting.
This
I
mean
June's
open
to
donating
this
to
us
and
seems
like
if
we
had
somebody
who
want
to
do
it,
sort
of
adopt
this
for
our
workflow,
like
what
Arvin
is
saying.
This
could
be
our.
This
could
be
our
development
workflow.
D
This
could
be
Sig
Windows
devtools
too,
or
this
could
be
its
own
thing
and
we
could
deprecate
the
old,
sick,
Windows
Dev
tools,
and
we
could
say
if
you
want
to
add
a
feature
to
Windows.
Here's.
Here's
the
development
environment
go,
get
pay
a
hundred
dollars
for
Windows,
put
it
on
some,
some
shitty
old
laptop
and
you
know,
and
just
go.
You
know,
like
I,
think
you
know,
I
think
it
would
be
work
because
none
of
us
are
net
programmers
and
I.
None
of
us.
B
But
well,
actually
it
already.
It
already
runs
on
links
because
it
actually
makes
Linux.
It
makes
Linux
binaries
from
the
published
step,
and
so
I
run
it
on
Linux
already
to
do
the
Linux
control
plane
for
my
own
cluster,
because
it
will
set
up
a
Linux
control,
plane
or
a
Linux
worker.
B
C
C
D
D
You
have
a
Windows
laptop
you've
got
wsl2
you
let
this
thing
go
and
do
the
weird
wsl2
runs
directly
plugged
into
your
router
thing
that
June's
got
in
here
and
and
then
the
work
for
us
is
putting
in
the
compilation,
steps
that
we
have
in
synchronous
devtools
in
here,
which
isn't
a
lot
of
work,
but
it's
somebody
at
least
digging
in
and
figuring
out
exactly
what
we
need
to
swap
out
in
the
different.net
or
C
sharp.
D
It's
like
the
same
thing
to
me
because
I
don't
know
the
difference
like
to
find
out
those
component
files
that
June's
got
and
say
like
okay.
How
do
we
need
to
change
this?
How
what
do
we
do
for
this,
putting
some
logic
in
there
and
then
we
could
say:
okay,
you
got
a
Windows
laptop.
You
can
run
this
and
then
somebody
smart
comes
along
and
says.
Well,
you
could
also
do
nested
virtualization
and
run
it.
D
You
know-
or
maybe
we
start
with
the
nested
virtualization
workflow
I,
just
feel
like
that's
a
harder
one
to
to
document
because
be
totally
different
on
different
system
machines
and
I
still
don't
know
what
RMR
Mac
users
are
going
to
do.
You
know
what
I
mean
like
it's.
It's
like
a.
C
C
Yeah
yeah
I
I,
don't
want
to
make
that
statement
right,
go
get
a
laptop
like
Windows
instance
of
some
sort
yeah
we
can
make
it
work
with
nested,
virtualization
yeah
and
the
other
question
I
was
going
to
ask
June.
Was
it
you
know?
One
thing
like
you
were
showing
me
the
spots
where
you're
saying
this
is
where
you
go
and
get
your
Cube
cubelet
from.
Maybe
we
could
make
that
configurable
to
point
at
like
a
local
file
system
where
we
have
the
cubelet
so
so
Jay.
D
B
C
I
see
your
point,
I
see
your
point,
you're
saying
equal
element
of
like
when,
when
June
does
the
command
that
that
was
running,
it
would
also
build
instead
of
just
pulling
from
you
know,
whatever
the
Upstream
binaries
are
yeah.
D
B
A
D
D
Yeah
so
some
of
this
we
talked
about
last
week,
June,
but,
like
you
know,
our
problem
is
or
we
talked
about
this
Sunday
I
guess
June,
but
some
of
it
is
like
you
know
we
just
right
now
in
kubernetes
people,
developing
features
for
Linux
can
go
and
they
can
run
one
command.
It's
called
like
kind,
build
node
image
and
that
uses
kind,
which
is
a
thing
called
kubernetes
and
Docker.
D
You
may
have
heard
of
it
or
used
it
and
it
literally
that
thing
goes
and
it
just
spins
up
a
new
cluster
from
the
code
that
you
just
wrote.
You
can
change
the
kublic,
you
can
change
the
Kube
proxy.
You
can
change
the
API
server,
you
can
add
a
new
flag
to
the
kubernetes
controller
manager.
You
can
change
anything
in
the
kubernetes
code
base,
run
one
command
and
get
a
cluster.
So
we
built
this.
D
This
mediocre
version
of
that
called
Sig
Windows
Dev
tools
that,
like
is
a
vagrant
recipe
and
you
can
create
a
branch
and
then
you
can
do
make
all
and
it'll
spin
up.
Two
VMS
and
it'll
smash
the
changes
you
made
into
those
VMS
and
it
takes
you
know
two
hours
or
whatever
an
hour
to
get
it
up
and
running.
But
at
least
it's
just
a
click
of
a
button
right
and
then
this
is
like
you
know,
and
but
the
problem
is
that
all
our
users
can't
do
that
right,
like
you.
D
So
we're
like
we've
got
to
give
people
something
other
than
this
virtual
box.
Macintosh
Centric
experience
and
we
haven't
been
able
to
figure
it
out.
Yet
we
tried
qemu
that
seems
to
it
seems
to
be
cobbling
along
we're
getting
closer,
but
this
kind
of
just
disrupts
our
entire
thinking
to
the
point
where,
if
you
have
a
like
I'll,
say
Windows
machine,
even
though
I
for
me,
the
thing
that
makes
sense
from
a
developer
perspective
is
a
laptop.
A
E
E
Would
just
also
add
a
couple
years
ago,
I
had
a
conversation
with
some
of
the
sick
testing
folks
and
one
of
the
things
they
want
to
do
for
kubernetes
CI,
for
like
a
new
PR
and
things
is
to
build
them
and
the
reason
they
built
kind
was
to
be
able
to
build
this
thing
quickly
on
a
local
machine
and
then
be
able
to
make
testing
really
fast
in
the
cloud
for
it.
E
E
Now,
with
what
you
just
presented,
we
could
potentially
do
that
type
of
thing
on
a
Windows,
Server,
2022,
VM
and
they
would
be
able
to
anybody,
would
be
able
to
replicate
that
locally
as
long
as
they
had
some
VM
with
Windows
on
it,
and
so
that
opens
the
door
again
to
being
able
to
potentially
get
Windows
CI
long-term
as
part
of
PR
submits
pre-submit.
So
I
think
there's
quite
a
bit
of
potential
here.
We
just
need
some
folks
to
to
lead
it
and
move
it
forward.
So
awesome
work.
A
B
D
B
No
I'm
not
going
anywhere
in
this
next
six
months,
but
that
is
a
good
good
point
to
race
and
context
like
I
built
this
tool
primarily
to
solve
my
problem
internally
in
my
business
right,
which
is
I
need
to
run
a
Linux
master
and
a
bunch
of
windows,
work
I
know
it's
like
I,
don't
have
the
I
need
to
run
the
whole
thing
on
Windows
I.
B
Don't
have
that
use
case
right
so,
in
terms
of
my
daily
experience
is
not
going
to
be
using
that
workflow
I
mainly
did
it
because
you
know
we
talked
about
it
and
it
was
like.
Can
we
actually
do
this?
Yeah
I
think
it's
viable?
Let's
explore
it
and
see
if
it
will
work
so
longer
term
Sig
Windows
definitely
needs
to
have
somebody
who's
capable
of
you
know,
building
upon
it
and
developing
it,
because
it's
just
purely
from
a
business
standpoint.
B
Now
that
my
problem
or
right
now
that
my
use
case
is
solved,
I
don't
have
a
lot
of
I.
Don't
have
really
any
development
time
to
kind
of
go
in
and
be
like
right,
I'm
going
to
take
this
for
another
three
months
and
improve
my
opponent
like
I.
Just
don't
I
don't
have
that
time
capacity,
so
it
does
need
to
be
in
order
for
it
to
make
that
transition
from
where
it
is
now
to
being
a
useful
tool
for
Sig
Windows.
B
There
is
going
to
have
to
be
somebody
who
basically
takes
over
the
development
of
it
or
supports
it
or
transfers
it
or.
However,
however,
it
makes
sense
to
fit
it
in,
because
I
won't
be
there
to
do
the
actual
development
I
can
I'm
more
than
happy
to.
You
know
Point
people
in
the
right
direction
and
be
like
this.
How
this
works?
You
know
so
that
people
can
get
across
how
it
works,
but
in
terms
of
the
actual
development,
that's
not
something
I'm
going
to
be
able
to
have
time
for
moving
forward.
A
Jay
I'm
also
wondering
it
sounds
like
the
Linux,
like
the
control
plane.
Stuff.
That's
running
is
also
set
up
all
manually
I'm
wondering
if
we
want
to
take
kind
of
the
overall
approach
of
this,
where
we
have
the
Windows
machine,
spinning
up
the
WSL
instance
and
then
running
things
in
there
and
then
actually
just
run
kind
in
the
WSL
instance.
A
C
D
C
D
D
A
D
D
Yeah
and
doing
this
tune
is
like
compiled
a
lot
of
assumptions
we
all
have
about
Kube,
ADM
and
other
things
like
that:
UDP
broadcast
things,
I
love
that
you
know,
there's
no
you're,
not
passing
tokens
around,
and
you
know
stuff
like
that.
Join
tokens
like
we
all
just
kind
of
yeah.
It's
way
simpler
than
what
we're
used
to.