►
From YouTube: Kubernetes SIG-Windows 20230131
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
Hello,
everybody
and
welcome
to
the
January
31st
2023
iteration
of
the
kubernetes
Sig
Windows
community
meeting.
As
always,
these
meetings
are
recorded
and
uploaded
to
YouTube
so
be
sure
to
adhere
to
the
cncf
code
of
conduct
for
anybody
who's
new
here.
That
mostly
just
boils
down
to
be
nice
to
everybody.
A
Let's
take
a
look
at
the
agenda
today.
It's
the.
A
A
Next
Thursday
I
think
there
were
two
enhancements
that
we
were
maybe
tracking
for
Sig
Windows
one
is
the
node
service
log
viewer,
one
with
our
event,
which
I
think
we'll
probably
end
up
discussing
a
little
bit
more
today,
and
then
there
was
another
one
for
potentially
the
setting
the
size
of
that
route.
Storage.
Space
for
the
windows
containers
I,
haven't
had
a
chance
to.
A
I
think
I
have
like
half
of
it
kept
written
up,
but
I
haven't
had
a
chance
to
do
that,
so
I'm
that
one
may
slip.
If
there's
somebody
who
wants
to
help
pick
that
up,
let
me
know
if
not
I'll,
try
and
have
that
cap
done
this
release
so
that
they
can
get
reviews
early
next
release
and
go
in,
but
I
think
that
one
was
mostly
we
have.
The
workaround
is
to
use
empty
doors
it's,
but
it
would
just
be
nice
to
have
for
consistency.
A
A
Have
any
other
announcements
or
anything
any
other
account
updates
that
I
might
have
missed.
A
Nope
all
right,
so
next
we'll
give
some
space.
If
there's
any
new
contributors
on
the
call
I
want
to
say,
hi
I
see
at
least
one
familiar
face
hasn't.
B
A
Oh
all,
right
going
in,
let's
start
with
the
agenda
item.
The
first
agenda
item
is
any
updates
on
the
Windows
container.
Unconference
I
think
our
event
asked
that
I
spoke
with
Brennan
Smith.
We
spoke
with
him
yesterday
and
he
is
actually
interested
in
putting
something
like
that
together.
A
He
messaged
me
this
morning
saying
that
he
had
a
fever
and
a
sore
throat
wouldn't
be
able
to
be
here,
but
he
is
going
to
start
looking
at
that
some
more
if
anybody
is
interested
I
think
later,
like
after
this
meeting,
let
me
start
us
after
I
didn't
seem
going
to
slack
to
see
if
it
to
see.
If
we
can
gauge
some
interest
and
I'll
also
put
a
I
did
find
a
link
for
with
like
an
agenda
from
the
previous
unconference.
A
If
anybody
is
interested
in
what
that
is,
so
we
can
continue
that
discussion.
Do
an
interest
check
on
slack
with
that.
C
I
assume
this
is
the
conference
at
kubecon.
A
I
don't
know
if
this
I
don't
know
what
Brendan
was
thinking
exactly
in
the
past.
This
was
hosted
in
Redmond,
so
there's
on
the
Microsoft
campus,
there's
like
a
little
space
for
having
conferences
like
that
and
people
were
invited,
but
it
very
well
could
be
that,
like
Court
time,
that
coordinated
is
one
of
the
upcoming
coupons,
I
will
defer
to
Brendan.
D
C
We'll
have
someone
live
stream,
the
unconference
on
Zoom
yeah.
A
A
Yeah
so
I'll
start
a
threatens
like
after
this,
and
then
we
can
get
some
more
you
can.
You
can
ask
all
of
these
questions
for
Brendan,
I
guess
Byron
sits
here
here
and
unmuted
did
you
want
to
are?
Is
the
plan
still
to
go
to
Signal
today
to
discuss
that
I'm
able
to
attend
the
pulsate
node
meeting,
so
I
will
also
be
there
and
then
well.
E
Yeah
I
spoke
to
Bernal
Ryan
Phillips
Derek.
All
about
that
I've
put
this
on
the
agenda
I'm
going
to
show
up
and
yeah
the
general
consensus
is
if
signode
is
also
open
to
accepting
the
patch
with
it
disabled.
By
default
behind
an
option
we
should
move
forward
and
you
know
I
was
thinking
more
about.
E
E
A
E
E
E
E
A
Me,
okay,
cool
I
mean.
A
Okay,
I
can
give
a
couple
of
updates
on
what
we've
been
doing
with
the
windows
service
proxy
So
for
anybody
who's
not
going
to
wear
this.
The
windows
service
practices,
a
new
repository
that
J
help
set
up.
That's
going
to
eventually
be
the
new
home
for
an
out
of
tree
Windows,
Q
proxy.
That's
based
on
kpng,
one
of
the
big
updates
that
we
have
now
is.
A
We
actually
have
an
image,
that's
getting
built
and
published
to
the
Pates
staging
prod
for
that
that
has
the
windows
kpg
binary
to
be
used,
I'll,
put
a
link
in
I.
Don't
think
that
there's
a
link
in
the
GitHub
repository
yet,
but
these
are
all
coming
through
the
post-
submit
jobs
against
Masters,
that's
kind
of
a
big
step
forward
for
allowing
us
to
set
up
more
CI
and
and
validation
for
this.
A
I
did
I,
don't
have
a
cluster
set
up
right
now,
but
I
I
was
I,
do
have
a
pull
request.
Did
that
merge?
A
Okay,
except
that
merch
I
had
a
pull
request
open
that
has
some
deployment
files
that
to
get
this
running
in
a
cap,
Z
cluster
right
now,
there's
still
a
little
bit
of
things.
A
We
need
to
figure
out
that
there's
an
issue,
that's
discussed,
one
is
I,
think
James
put
together
a
proposal
a
while
ago
about
what
we
would
need
to
do
to
have
it
so
that
there
was
less
Reliance
on
the
some
cmi-specific
behaviors,
specifically
like
what
the
network
name
is
and
because
each
right
now
each
cni
creates
a
new
hns
network
with
like
a
unique
name,
and
it
would
be
good
to
standardize
that,
and
also
things
like
what
the,
if
there's,
a
that
Source
vape
and
a
special
endpoint,
that's
needed
for
that
or
the
the
port.
A
B
A
Exactly
what
James
was
suggesting
here
as
one
of
the
proposed
Solutions,
and
then
there
was
a
couple
of
there's
some
comments
about
like
we
could
potentially
like.
Have
it
default
to
looking
for
that,
but
take
in
an
environment
variable
that
if
it
does
need
to
get
overwritten
so
yeah
just
so
those
were.
Those
were
some
of
the
exact.
You
know,
proposals
here.
D
A
Least
what
what
I
had
to
do
for
this?
The
the
Calico
deployment,
is,
do
a
couple
like
we
wait
for
I
think
we
probably
don't
need
to
wait
for
the
networks
now
the
HMS
Network
to
come
up,
but
we
we
need
to
query
some
state
to
get
the
IP
address,
that's
associated
with
that
endpoint
that
Calico
set
up.
A
So
we
have
to
do
that
in
a
Powershell
script
and
then
before
we
can
start
kpng
and
pass
these
environment
variables
in
and
we
I
mean
we
could
potentially
generalize
the
Powershell
script-
that's
put
in
there,
but
it
since
we
have
apng
itself
right
there.
I
think
it
makes
sense
to
just
put
all
that
information
right
into
the
KPMG
itself.
Okay,.
B
Yeah
go
ahead,
James
I
think
my
goal
is
to
have
this
so
that
it's
just
a
binary,
so
we
don't
have
to
have
some
Powershell
and
the
main
reason
for
that
is.
If
we
have
to
have
Powershell,
then
we
also
have
to
have
that
Powershell
signed,
and
so
so
it
just
makes
it
a
whole
other
step
to
our
whole
build
process
and
there's
no
real
reason
to
have
the
Powershell
script
for
two
proxy,
because
now
because
we
should
be
able
to
fit
all
those
settings.
A
D
So
I
I
I
I,
my
update
is
I'm
working
with
Damon
on
getting
him
ramped
up,
so
he
can
sort
of
help
us
out
with
this
repo
and
I've
been
pairing
with
him
and
he's
get
pretty
much
up
to
speed
on
things,
I'm
expecting
him
to
be
able
to
start
doing
stuff
within
a
week
or
two
so
Mark.
If
you
have,
he
looks
like
you've,
got
eight
issues.
If
you
have
other
issues,
you
want
people
to
look
at.
D
Let
me
know,
and
then
hopefully,
I
can
start
testing
it
again
after
after
that
phase
is
over
he's
still
ramping
up
on
how
to
do
stuff
with
privileged
pods
and
our
host
process,
containers
and
stuff,
but
he
he
knows
kpng
inside
out.
So
it's
just
a
matter
of
once.
He
knows
Windows
he'll
be
able
to
do
all
this
stuff.
A
Okay,
yeah
and
I
can
definitely
help
out
with
those
process.
Stuff
I
think
the
next
steps
for
me,
which
I
have
done
soon,
is
to
get
a
periodic
job
set
up.
That's
pulling
that's
running
all
of
the
Ed
or
probably
like
a
subset
of
the
networking
tests
in
in
the
capacity
clusters
that
we
have
that
are
using
kpng
so
that
we
can
have
just
like
an
maybe
comparison
between
the
intricute
proxy
and
the
and
kpng
I
did
notice
and
I'll
like
in
order
to
get
the
load
balancer
programming
working
in
capacity.
A
G
A
A
A
This
was
I'm
guessing
the
code
from
kpng
that
was
commented
out,
and
then
this
is
what
it
turned
into,
and
so
just
the
structure
changed
a
little
bit.
There
was
there's
a
good
example.
A
B
D
A
The
point
where
it
was
able
to
create
the
the
load
balancers
that
were
this
way
but
I
think
that
there's
definitely
some
issues
in
the
delete
paths,
probably
the
same
thing,
but
we
can
I
think
after
there's
a
test
signal
kind
of
iterate,
Okay.
D
Cool,
so
for
us
we
set
up
I,
I,
didn't
know
you
were
so
far
on
the
CI
side,
because
so
what
I
did
is
I
was
like
well,
we
set
up
a
shared
kind
of
development
environment
and
we
were
going
to
partially,
like
maybe
add
that
as
a
GitHub
action
that
was
like
constantly
running
and
re
reloading
and.
D
But
it
sounds
like
you're
you're,
pretty
you're,
making
Fast
progress
on
the
CI
side,
so
you
think
you're
going
to
have
us
you're
going
to
be
able
to
finish
having
a
CI
there's
a
lot
to
do
for
a
test
in
for
a
CI
job
right
like
or
is
it
that
you
you
just
have
to
change?
You
just
are
you're
just
going
to
add
that
host
process
container
and
you
think
the
rest
of
it'll
just
work.
Yeah.
A
So
what
I
started
last
week
is
I
added
another
template
into
the
capacity
repository
that
makes
it
so
that
we
don't
deploy
Q
proxy
as
a
customer
resource
set
and
then
then
we'll
pretty
much
just
install
the
the
automatic
file
like
the
one
file
on
this
deployment
set
as
request
resource
set,
if
you're,
using
that
specific
template
and
that'll
be
pretty
much
the
same
configuration
as
the
other
Upstream
media
tests
that
we
have
in
kemc.
Okay,
cool
I
need
to
just
finish.
D
That
up,
oh
great,
okay,
all
right
cool!
If
that,
if
you
get
busy-
and
you
feel
like
you
want
somebody
else
to
like
maintain
that
or
whatever
and
you
just
write
up
instructions,
maybe
that
that
might
be
something
we
can
give
to
someone
else.
It
sounds
like
it
sounds
like
that's
one
of
those
things
where,
if
you
know
the
six
different
files
that
you
have
to
edit
and
the
who
to
Ping
On
The
GitHub
issue,
then
it
it's
not
a
hard
thing
to
do
right.
So
yeah,
yeah,
cool.
A
A
A
couple
different
ways
to
kind
of
bootstrap
this
I
think
I
might
just
talked
about
it
in
slack
too
yeah,
but
basically
like
we
need
to
when
kpng
starts
up.
It
needs
to
talk
to
the
API
server,
to
get
all
of
the
routes
and
everything
that
it
needs
to
program.
So
I,
one
issue
that
like
I,
was
running
into
was
there's
a
service
IP
address
for
the
API
server,
but
but
that
is
not
reachable
by
the
windows
nodes
until
you
start
running
kpg
to
do
that
programming.
A
So
what
there's
a
couple
options
of
what
we
could
do
and
what
we
were
doing
for
like
the
entry
queue
proxy
is
just
passing
it
a
cube
proxy
or
a
cube,
config
file
with
a
public
eye
or
with
an
external
you
know,
IP
address
or
DNS
name
in
there.
I
got
it
working
by
overwriting,
the,
but
these
environment
variables.
These
are
the
environment
variables
that
in
cluster
config
uses
when
it
tries
to
build
the
rest
client.
A
So
these
automatically
get
populated
with
the
Pod,
at
least
at
least
no.
These
ones
are
what
I'm
talking
about.
So
these
are
these
get
populated
in
the
environment
of
all
the
pods
or
containers
and
pods,
so
that
they
could
talk
to
the
API
server.
A
B
A
A
Think,
eventually
and
I
think
James
Jeff
Walmart,
probably
back
around
in
this,
but
I
think.
Eventually
we
want
cubeadm
to
create
that
file
for
us.
I,
don't
know
exactly
how
this
works
on
the
Linux
nodes,
but
in
the
cluster
API
departments
that
we
have
there's
usually
an
IT
container
that
builds
that
from
something
or
you
have
to
put
it
in
as
a
config
mapper
a
secret
in
the
cluster.
It's
a
big
and
that's
definitely
different
than
how
it
gets
bootstrapped
on
the
Linux
side
of
things.
B
Yeah
so
qbdm
creates
a
config,
our
back
roll
and
then
wires
up
cubeconfig
that
points
to
the
our
back
roll
for
Q
proxy,
and
so
we
could
just
reuse
that
probably
yeah
but
yeah
it
would
be.
But
Cube
ADM
currently
deploys
the
Linux
version
of
qproxy
and
it
would
be
nice
as
a
usability
perspective
for
customers
to
get
that
same
experience
in
Windows,
without
having
to
do
that
extra
step.
B
B
Yeah
so
I
mean
we
could
essentially
if
we
could
assume
that
there's
like
the
Cuban
am
our
back
provider
and
we
could
even
create
the
config.
So
so
the
way
they
do.
B
Know
well
well,
there's
a
they
create
an
R
back
for
qproxy
or
somebody
does
I,
don't
know
exactly
who
does,
but
that's
who
they
that's
what
they
use
under
the
hood
is
that
account
the
service
account
that
is
Q
proxy,
and
so
we
could
either
assume
it's
there
or
create
one,
that's
very
similar
to
ourselves
and
then
the
cue
config
that
they
use
just
redirects
towards
the
in
cluster
in
cluster
config.
B
No,
no,
no
thing.
D
B
D
B
B
Yeah
I
just
add
this
as
a
note
just
to
get
folks
as
a
heads
up,
but
this
Community
member
has
been
working
on
this
for
a
while
and
oh
I'll,
just
we,
we
merged
this
last
week
into
image,
Builder.
So
there's
another
provider
out
there
with
image
with
Windows
support.
Oh.
A
A
I
do
need
to
drop
in
a
minute
I'll
give
either
James
or
Jay,
though
the
host,
so
that
I
can
go
to
node
to
discuss
that
other.
H
Okay,
so
we
noticed
a
puff
impact
on
the
past
containers
and
windows,
and
this
is
because
we
were
overriding.
The
entry
point
to
open.exe,
a
material
of
what
container
image
was
being
used
and
the
proposed
tricks
that
we
have
for
this
one
is
to
basically
override
the
entry
point
with
ping.exe.
Only
if
the
base
image
is
nano
server
and
server
code
and
it
comes
with
the
default
entry
point,
which
is
C
Windows
system,
32
and
command.exe.
H
So
I
think
this
is
just
to
help
folks,
who
are
still
relying
on
Nano
server
server
code
for
the
pause
containers
which
I
highly
doubt
they
are
and
there's
just
an
attempt
to
not
break
those
folks.
But
if
someone
tries
to
build
a
pause
container,
a
pause
image
from
either
Nano
server
and
server
core
and
then
rewrite
the
entry
point
to
maybe
just
say
cmd.exe,
then
we
will
not
be
overriding
with
paint
and
they
could
break
so
just
wanted
to
see.
H
If
any
of
you
had
any
concerns
or
any
thoughts
on
this.
H
H
H
B
Cool
yeah,
so
so
essentially,
this
will
enable
actually
running
the
pause
binary
that
we
build
in
kubernetes,
whereas
is
being
overridden
previously.
C
So
basically,
this
checks.
If
the
first
argument
of
the
process
itself
is
the
full
path
for
cmd.x.
H
So
it's
907
Circle
images
are
used,
as
is
without
any
like
changes
to
it.
That's
the
default
entry
point
that
comes
with
the
image.
H
B
Yeah,
it
was
in
HCS
shim
when
a
container
was
started.
Okay,
so
it
was
I
think
it
was
causing.
Network
instability
is
that
right
could
change.
B
D
Pramita
and
David
joined.
There
are
two
new
folks
that
are
getting
involved
and
I'm
kind
of
working
a
little
in
the
background
with
them.
They're
doing
David's
got
I.
Think
pretty
good
David.
Do
you
have
any
updates
on
those
on
the.
F
Yeah
so
I
was
doing
the
deployment
with
the
sick,
Windows
Dev
tools,
and
it
was
failing
out
reached
out
to
a
meme.
He
was
able
to
help
guide
me
through
the
errors
stating
that
it
was
a
container
D
issue,
so
I
was
able
to
get
a
pull
request
in
and
updated
Maine
with
updating
the
container
D,
so
change
the
logic
on
the
install
there.
So
now
it
installs
good.
So
the
next
step
is
automating.
F
The
host
process
for
the
container
D
through
the
Calico
stuff
there
and
basically
I've,
got
everything
up,
except
for
one
bit
of
it
where
the
windows
proxies
The
Who
am
I
that
are
automatically
deployed
at
the
beginning,
aren't
able
to
go
into
a
ready
state,
but
everything
else
has
so
I'm
stepping
through
that
and
trying
to
code
that
up
that's
kind
of
where
I'm
at
now
so
but
yeah
got
two
pull
requests
up
so
far.
D
Thanks
David
all
right,
great
I'm,
working
with
Damon
in
the
mornings
he's
getting
up
and
running
with
this
stuff
too,
but
so
that's
good
and
then
pramita
you
had
some
qemu
stuff
right
like
he
was
working
almost
so.
G
So,
basically,
the
image
that
I
had
for
arm-based
architecture
that
is
working
with
RC
rsync,
that's
something
which
you
suggested
I'm
trying
to
use
the
boxes
that
we
already
have
like
the
Ubuntu
2004,
which
is
there
in
the
code
itself.
So
that
is
failing
for
me:
I'm
not
able
to
do
SSH
to
the
Box
existing
boxes.
G
So
that
is
something
I'm
debugging
but
yeah.
Our
sync
worked
with
with
you
know
my
version
of
box
that
I
had
like
the
image
that
is
Arm
based
that
is
working,
but.
G
Didn't
care
about
arm
yeah,
but
those
I'm
not
able
to
do
SSH
to
those
like
it's
it's
when
you
you'll
do
vagrant
status
it
it
is.
It
is
up
like
it's
there,
but
SSH
is
not
working.
So
is.
G
D
G
Something
UI,
no,
no,
this
so
there's
a
tool,
UTM
dot
qme.
So
basically
the
base
of
it
is
qm
qem
only,
but
it's
a
GUI
base
with
UTM,
which
actually
gives
you
a
GUI
base
for
for
creating
the
VMS
and
everything
there.
I
was
able
to
do
yeah,
SSH
and
everything,
but
with
the
commands
like
the
vagrant,
are
python
I've
been
provider
in.
D
D
But
maybe
qemu
vagrant
doesn't
support
private
Network
okay,
so
maybe,
if
you
get
rid
of
the
privacy
of
that,
maybe
if
you
get
rid
of
the
private
Network
you'll
at
least
be
able
to
get
into
the
VMS
and
then
we'll
have
a
new
problem,
which
is
that
now
we
don't
know
what
to
hard
code,
all
the
static
IPS
do.
Okay,
so.
D
All
right,
I'm
gonna
drop
thanks,
David
and
pramita
for
coming
and
helping
us
out
with
this
stuff.
Hey.
I
There's
an
issue
by
the
way
in
in
vagrant
libbert
with
some
keemu
stuff.
Here,
that's
talking
about
SSH
into
these
boxes
and
they
claim
they
got
it.
Working
I'll,
just
I,
don't
I'm
on
my
phone
right
now,
but
I'm
gonna
post
it
in
Sig
windows.
I
Thank
you,
I
didn't
notice,
you
also,
but
there
you
go
that
might
help
I'm,
not
sure.