►
From YouTube: Kubernetes SIG-Windows 20220809
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
august
8
2022
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
and
that
could
be
summarized
to
just
be
a
nice
person.
So
let's
get
started.
First
is
announcements.
I
had
a
couple
announcements:
if
anybody
has
anything
else
feel
free
to
add
it
announcements
I
had
were
the
125
test
freezes
today
around
6
pm
pst.
A
A
Next
announcement
is
the
kubecon
north
america
schedule
is
now
available.
I
thought
I
added
a
link
to
that,
but
I
didn't
I'll
get
that
added.
A
There's
four
windows
related
talks
this
time,
so
we've
got
the
maintainer
series
talk
talk
that
james
submitted
about
all
about
host
process,
containers
got
accepted
and
chinchi
and
memes
talk
about
the
operational
readiness
got
accepted
too.
So
that's
great,
and
there
is
a
customer
submitted,
talk
from
relativity
about
their
experience,
running
windows,
containers
that
also
got
accepted.
So
that's
exciting.
It's
a
lot
of
talks
for
for
windows
so
and
if
anybody
else
like,
I
think
I
may
be
attending
in
person.
A
If
anybody
else
is
going
to
be
attending
in
person,
please
let
us
let
me
know.
Hopefully
we
can
meet
up
at
some
point.
A
One
other
kind
of
minor
announcement
is
windows.
Ser
semi
annual
channel
release,
20
h2,
is
now
out
of
support.
That's
the
last
semi-annual
channel
release
that
I
think,
we've
tried
to
say,
was
supported
in
kubernetes,
so
moving
forward.
We
will
only
be
supporting
ltsc
releases
of
windows
server.
Here
I
opened
up
a
couple.
Pr's
I
see
james
celebrating
yeah.
I
opened
up
some
prs
to
get
those
removed
from
all
of
our
proud
jobs
and
everything,
and
I
think
we
can
follow
up
with
some
more
prs
to
remove
them
from
the
container
images.
B
Can
I
get
someone
to
put
some
thumbs
up
on
an
aws
issue
that
they
are
out
there
trying
to
work
on
2022
amis
for
eks?
They
haven't
completed
them
yet
and
with
said
announcements
of
20
h2,
going
away
wouldn't
mind.
Having
those
can,
I
only
get
in
chat
and
have
some
people
throw
some
plus
ones
on
it.
For
me,.
A
A
B
It's
possible:
I
should
reach
out
to
him
anyway,
I'm
linking
it
here
in
the
chat
just
throw
some
thumbs
up
on
it,
so
they
see
that
it's
something
that
people
care
about.
A
All
right
sounds
good
next,
if
there's
any
new
contributors
here,
we
can
give
some
space
to
welcome
them.
If
so,
if
anybody's
new
here
wants
to
just
say,
hi
say
what
you're
working
on
introduce
yourself
feel
free
to
do
so,
no
pressure,
but
you
can
just
unmute
yourself
and
go.
C
A
Yeah,
I
know
us
we
may
talk
about
some
of
that
in
a
little
bit
and
if
not
I'm
happy
to
go
over
some
of
that
with
you
offline,
you
can
reach
out
and
you
can
sync
up.
D
It's
interesting,
though,
are
you
that,
were
you
looking
at
the
sig
windows,
dashboard
or
just
in
general,
just
looking
at
other
dashboards
and
then.
D
C
A
Yeah,
those
were
usually
just
if
anybody
had
any
topics
or
wanted
to
pair
on
anything
folks
who
were
available
to
join
stay,
cool,
well
welcome
and
yeah.
Hopefully
we
can
answer
some
questions
and,
if
not
feel
free
to
ask
whatever
questions,
you
have
okay,
anybody
else.
If
not,
we
can
go
to
the
agenda
pause.
D
This
is
me,
that's
me,
so
I
I
and
actually
nawaz
it
looks
like
you
came
on
a
good
day,
because
a
meme
has
is
is
he's
so
proactive
that
he
went
and
he
wrote
in
the
notes
that
he
wanted
to
make
a
new,
proud
job
for
operational
readiness.
D
That
okay
yeah,
that
would
be
great
yeah
because
we're
we're
building
that
new
operation
yeah
anyways.
So
maybe
we
can
go
after
and
look
at
that
together,
but
yeah
so
cappy
support
is
for
me.
I
saw
this
doc
that
james
you
made
about
cappy
support
and
I'm
kind
of
like
more
and
more
feeling
like
long-term.
D
Mute,
you
know
mutable,
I
know
you're
not
allowed
to
say
mutable
vms,
but
I'm
just
going
to
say
it
like.
I
just
feel
like
for
windows.
I
just
the
last.
I
just
feel
like
moving
towards
something
mutable
in
the
long
term
feels
like
something
we
should
talk
about,
and
then
I
know
that
we
have
this
windows
machine
operator
that
arvin
works
on
and
internally
you
know
vmware.
D
We
talk
about
what's
the
ideal
way
to
install
windows
and
on
kubernetes,
and
we
do
a
lot
with
image
builder
right
now:
it's
not
bad.
It
works
real.
Well,
I'm
just
wondering,
like
are
people
thinking
about
a
mutable
agent
based
approach
to
bootstrapping
windows
or
ssh,
based
approach
or
machine
operator
type
approach,
and
what's
the
status
of
that
support
proposal
also,
but
even
though
it's
not
directly
related,
you
know.
E
Yeah
I
mean
so.
I
just
give
a
little
background
on
this.
A
couple
years
ago
there
was
no
window
support,
so
I
kind
of
went
and
looked
at
various
ways
to
bootstrap
bootstrap
everything
and
get
it
working,
and
I
think
I
think
I
listed
a
few
different
options.
E
We
went
with,
at
least
for
capsi
and
general
windows
approach
was
to
use
cloud-based
init
so
that
we
could
reuse
a
lot
of
the
infrastructure
that
was
already
out
there
in
the
cluster
api
ecosystem.
So
we
wouldn't
have
to
rewrite
how
to
gather
certs
and
manage
certs,
because
cube
adm
provider
does
all
that
for
us,
and
so
that
that
was
the
initial
approach
and
I
think
the
machine
operator
was
fairly
young.
At
that
point,
I
I
don't
remember
exactly
coming
across
it
when
I
was
doing
all
the
research,
but
I
don't
know
if.
E
That
that's
kind
of
like
where
we,
where
we
chose
that
just
we
didn't
want
to
be
a
snowflake
windows,
is
just
because
there's
so
few
of
us.
If
we
had
to
maintain
our
own,
you
know
cuba,
dm
type,
bootstrap
provider,
they're,
just
there's.
You
know
just
not
enough
folks
to
do
that
type
of
thing,
and
so
that
that's
why
we
went
that
way.
But
I
mean
now:
we've
got
this
machine,
not
config
operator.
E
I
don't
know
anything
about
it,
but
it's
a
potential
option
from
a
from
like
a
cappy
perspective,
I
think
you
know
how
you
bootstrap
your
node
is
left
up
to
the
provider
to
choose
that
option,
and
so
I
I
think
you
know,
I
think
it
sounds
like
openshift
uses
the
machine
configure
operator
and
I
don't
think,
there's
anything
stopping
any
other
provider
from
using
something
else
as
well.
F
Yeah
and
I'm
happy
to
talk
about
what
we
do
with
wmco
the.
A
All
right,
let
me
stop
sharing
and
then
I
think
anybody
can
share.
E
Like
what
what
what
things
are
you
running
into
that?
You
want
to
go?
Look
at
something
else
versus
like.
D
E
D
I'm
just
wondering
whether
we
ever
thought
about
what
we
should
do
for
like
the
long
term
like
say
four
years
from
now,
are
we
still
doing
immutable
machines,
or
do
we
want
to
consider
looking
at
it
at
a
mutable
approach,
because
you
know
there's
these
stories
about
folks
kind
of
like
starting
out
with
things
immutable
and
then
eventually
moving
towards
a
mutable
model
once
things
go
more,
you
know
like
netflix
did
that
recently
and
like
there's
just
you
know,
I
I'm
okay
with
us
going
with
this
immutable
image
builder
model
forever.
D
E
F
Yes,
sir,
I
guess
I'll
just
go
over
the
quick
like
a
workflow,
that
our
users
go
through
right
and
we
have
like
machine
objects
or
machine
set
objects
that
they
create.
That
ends
up
turning
into
like
windows
that
ends
up
creating
windows,
vms
wmco,
which
acts
upon
those
windows,
vms
and
so.
E
Are
those
those
like
a
specific
type
that
openshift
exposes
or
is
that
a
cappy
object.
F
C
G
Yeah,
so
cluster
api
is
different
from
the
machine
api
that
we
use
at
red
hat,
so
we
have
sort
of
diverged
and
it
is
because
of
the
immutable
department
to
some
extent
it
is
because
of
the
immutability
part
that
jay
is
bringing
up.
We
do
it
differently
in
case
of
windows.
I
do
not
know
if
that
has
changed
I'll.
F
Yeah
yeah
and
so
I'll,
just
I'll
try
to
basically
describe
this
from
the
point
of
just
inter
just
acting
on
a
vm,
and
so
because
wmco
kind
of
just
takes
these
machine
objects
and
then
gets
the
ip
address
of
the
vm
and
users
can
also
bring
in.
We
call
it
bring
your
own
host
bioh
and
you
can
give
the
ip
and
a
username
for
us
to
log
into
with
an
sshq
that
you
provide,
and
then
we
also
act
upon
that.
F
F
F
Contains
data
about
the
windows
services
that
you
want
running
on
the
vm,
so
we
have
like
the
windows
exporter
to
collect
prometheus
metrics
running
on
the
vm,
we're
giving
the
arguments
that
it
needs
to
run
with
and
wicd
is
just
a
kubernetes
controller
and
right
now
it's
running
as
a
windows
service,
but
in
the
future
we
hope
to
run
as
a
host
cross
container.
As
that
develops.
E
F
Yeah,
so
you
know
we
also
have
another
service
and
with
with
wicked,
as
a
functionality
of
variable
substitution,
so
you
can
plug
in
node
name
and
then
give
like
then
specify
how
to
get
that
variable.
So
we
say
that's
the
node
object,
json
path
and
then
it's
a
metadata.name,
because
certain
windows
services,
their
arguments,
can
be
based
upon
variables
from
the
nodes.
We
can
also
do
that
through
powershell
scripts,
so
run
a
powershell
script
as
we're
trying
to
figure
out
what
arguments
need
to
run
for
the
service
and
then
based
on
the
output.
D
F
Almost
because
wicked,
actually
is
a
demon
or
it's
windows
service
itself.
So
it's
going
to
be
running
forever
and
so,
let's
say
someone's
changing
the
specification
of
the
service
config
map
or
they
want
to
upgrade
to
a
different
one.
F
Wiki
will
it's
a
it's
plugged
into
the
api?
It's
a
kubernetes
controller.
So
if
the
service
object,
if
the
service
config
map
or
sorry
if
the
node
object,
annotation
changes
and
the
annotations,
how
we
tell
liquid
what
config
map
to
look
at
it's
going
to
go
and
change
all
the
service
specs
according
to
the
new
config
map
that
it
should
be
using
and
to
make
that
kind
of
simpler
to
understand,
there's
a
maybe
it's
not
simpler,
because
this
is
all
ascii
art
I
did
this
morning.
F
So
wmco
copies
over
the
required
files
and
first
it
starts
the
bootstrap
commands
for
wicked
and
wiki's
running
as
a
windows
service
right
now,
but
in
the
future
could
be
a
host
cross
container.
F
D
D
D
C
F
C
F
Once
that
boot
show
command
runs
and
then
cubelet
and
container
d
are
running
on
the
vm
and
then
you
know
the
csr
is
generated
and
wmco
approves
it.
Then
we
can
start
running
wiki
as
a
daemon.
So.
E
Does
does
the
wmco
thing
do
all
of
the
like
configuration
like
cert
generation
and
all
that
to
join
the
node.
F
Wmco
is
copying
over
the
the
bootstrap
like
cubeconfig
and
the
desert
associated
with
that
and
that's
used
to
initially
bootstrap
the
the
vm
or
sorry.
I
think
you
essentially
start
cubelet.
E
F
Yeah,
so
wmco
is
going
to
stop
kublet
copy
over
the
new
cubelet
binary
and
then
sorry
so
wicd
is
going
to
be
so.
Wfco
is
going
to
be
copping
over
the
new
one
after
wsd
starts
everything
stops
everything
that
new
cubelet
is
now
going
to
be
on
the
node
and
then
wicd
will
restart
it
so
yeah.
So
it's
going
to
be
stopped
copied.
The
new
version
can
be
copied
over
and
restarted,
and
so
yeah
everything
is
mutable
for
windows.
A
C
D
That's
okay,
sebastian.
A
F
Last
two
months,
so
it's
still
very
we're
in
a
very
transitionary
phase,
I'll
link
it
in
sig
windows.
F
Once
I
hopefully
have
a
little
more
information
up
like
later
today,
so
people
can
go
through
it
and
see.
If
it's
you
know
what
people
looking
for.
A
A
A
D
A
Oh,
I
guess
one
other
thing
that
we
probably
could
have
added
to
the
announcements
was
that
there
is
now
a
cooper
96
repo
for
the
windows,
operational
readiness
work
that
got
transferred
ownership
from,
I
think,
repo
that
jay
set
up
an
org
that
j
set
up.
So
this
is
now
an
official
you
know,
kubernetes
repository
hooked
up
to
prow
and
everything.
So
if
folks
are
interested,
please
take
a
look
yeah.
A
I
think
in
slack
slack
I
mean
was
asking
what
would
it
take
to
start
running
like
the
operational
readiness
stuff
in
a
ci
job?
So
I
wanted
to
give
a
kind
of
a
minute
to
show
what
we
could
do
with
with
with
capsie
and
there's
probably
a
lot
of
different
ways
to
do
that,
but
I'm
just
the
most
familiar
with
how
to
run
extra
stuff
with
with
kappa
z.
So
I
think
now,
as
this
is
what
you
were
interested
in,
and
I
mean
this
could
be
some
future
work
in.
A
So
this
is
the
test
infrarepo
I'll
link
this
here,
and
this
is
kind
of
an
example
job
of
how
we
run
some
other
tests
that
are
live
in
another
repository
into
our
on
a
capsie
cluster
in
so
this
is
kind
of
what
the
structure
of
the
proud
jobs
look
like,
and
I
don't
think
we
have
enough
time
to
go
into
a
lot
of
detail
here,
but
we
can
go
follow
up
in
slack
in
the
pro
jobs
you
can
specify
one
or
more
repositories
to
sync
to
the
machine
when
it
runs,
and
in
this
case
what
we
do
is
we
sync
the
cluster
like
the
capsi
repository,
and
we
can
sync
other
repositories
in
this
case
we're
running
the
azure
disk
csi
driver
jet
test
passes
with
it
and
in
the
cluster
api
repository,
so
we'll
set
it.
A
We
set
that
as
a
workdir.
So
that's
where,
like
this
will
come
from.
We
have
this
script
ci
entry
point
and
what
this
does
is.
This
uses
environment
variables
that
are
all
defined
in
these
presets,
which
are
defined
at
the
top
of
the
file.
A
Yeah
kind
of
gonna
be
going
over
this
pretty
quickly,
and
then
you
guys
can
ask
some
more
information
later.
This
runs
this
ci
entry
point
runs,
and
what
this
does
is
this
sets
up
a
kypc
cluster
and
eventually
sets
the
cube
config
to
the
context
of
that.
Like
sets,
yeah
writes
a
cube
config
for
the
context
of
that
cluster
and
then
hands
over
execution
to
whatever
script
you
whatever
else
you
specify
in
these
extra
args
for
running
it.
A
So
in
this
case,
I
think
I
believe
that,
like
this,
one
is
set
up
to
run
oops.
Where
is
that
just
like
a
two
machine
cluster
and.
A
And
then,
after
that,
you
can
use.
Basically
you
can
run.
You
can
execute
like
commands
from
the
root
of
any
other
repository
here.
So
in
this
case,
this
is
the
azure
disk
repository
which
gets
cloned
to
this
path.
Based
on
on
this
and
then
what
this
does.
Is
it
just
changes
the
directory
to
the
to
that
repository
and
runs,
make
e
to
e
test,
and
then
so.
A
These
tests
are
assuming
that
there's
already
like
the
the
cluster
context
for
the
remote
cluster
is
already
you
know,
set
up
and
then
just
runs
those
tests
against
that
cluster.
So
it
should
be
possible
to
set
up
something
like
this
in
this
new
windows,
operational
readiness
repository
to
have
some
scripts
or
like
make
file
that
has
a
whole
bunch
of
make
targets
to
execute
the
tests
against
a
cluster.
That's
already
set
up.
H
Yeah,
I
think
it
makes
sense
like
what
we
are
doing
is
to
run
everything
on
sonoboy,
but
I
I
think
we
can
run
without
sonoboy
and
and
like
raw
each
wee
test
binary.
H
A
Yeah,
it
could
be
possible
to
have
like
a
make
target
to
do
run
it
the
raw
eating
stuff,
but
also
to
run
the
sono
buoy
like
to
bootstrap
sonobuoy
and
run
sono
buoy
to
I
think
that
would
make
sense
to
live
in
the
windows.
D
Operational
christmas
I'd
be
surprised
if
we
didn't
already,
because
how
does
it
run?
Maybe
it
runs
conformance
directly
yeah,
it
probably
runs
the
conformance
tags
directly.
Maybe
it
doesn't
use
certified
conformance.
C
A
Yeah
and
if,
if
anybody's
interested,
usually
what
we
do
to
set
up
these
test,
jobs
is
set
up
a
pull
request,
job
that
we
can
trigger
from
different
repositories,
and
you
could
specify
not
in
here
there's
some
other
files
windows.
No,
that's
the
unit
pre-submits.
A
I
need
to
drop
first
ignored
in
a
minute,
so
I'm
going
to
rush
and
then
we
can
follow
up
in
in
slack,
but
we
can
set
up
a
pull
request,
job
that
you
can
trigger
via
the
github
like
comments
in
the
windows,
operational
readiness
repository
and
then
so
you
can
make
a
dummy
pr
into
that
repository
and
trigger
the
unit
of
the
pr
job
until
everything
is
set
up
and
running
and
then,
after
that's
set
up.
H
A
Yeah
yeah,
I
think,
moving
up
the
windows.
Operational
readiness
stuff
into
the
kubernetes
repository
can
will
just
unlock
a
lot
of
this
potential
here,
because
we
can
now,
you
know,
just
have
set
up
some
test
jobs
and
have
prow
manage
those
and
then
use
those
to
to
iterate
on
the
all
of
the
the
actual
shape
of
the
test.
Job.
E
The
other
thing
that
might
be
cool
is,
if
you
have,
if
you're
running
these,
against
your
vmware
clusters
or
the
tanzu
clusters,
whatever
you're
running
against,
you
can
upload
the
results
into
prow
and
we
can
potentially
get
it
on
to
the
like
sick
windows
dashboard
as
well
as
like
another,
instead
of
just
caps
e
being
the
the
folks
that
are
publishing
results,
you
could
you
could
publish
the
results
too,
you
don't
necessarily
have
to
run
it
inside
the
pro
infra.
E
D
Yeah
I
filed
an
issue
in
in
in
test
infra
for
this
just
to
track
it.
Thanks
for
bringing
this
up
mark.
A
C
I
C
I
C
I
E
Gonna
say
I
think,
the
flannel
pod
that
we
post
there
has
very
specific
requirements
and
like
where
everything
laid
out
and
if
you
follow
the
guide
exactly
then
there's
a
little
bit
of
difference.
I
do
have
a
pr
that
I
like
got
it
all
working
together.
I
can
post
it
to
you
on
in
slack,
so
maybe
it'll
help.
I
C
D
I
have
an
update,
I'm
in
the
middle
of
a
big
rebase,
with
the
windows
kernel
space
for
kpng
and
as
soon
as
the
rebase
is
done,
we're
going
to
try
to
merge
it
and
we're
filing
some
follow-on
issues,
and
I
think
the
follow-on
issues
will
be
more
approachable
for
new
people
that
are
interested
in
some
of
the
windows
coop
proxy
stuff,
because
the
all
the
track
will
be
laid
out
as
far
as
integrating
with
kpng.
So
david
shot
was
on
earlier.
But
I
think
he
dropped
but
I'll
reach
out
to
the
microsoft
networking.