►
From YouTube: Kubernetes SIG Windows 20220222
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
february
22nd
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.
Let's
get
started
with
some
announcements,
jay
I
mean
or
shinji.
Do
you
want
to
start
off
with
this
announcement.
B
Yeah,
so
we
are
planning
to
do
a
demo
in
the
operational
readiness
in
a
few
weeks,
with
a
few
of
the
tests
that
we
have
and
the
prototype.
A
B
Yeah
we
are
having
some
discussions
on
putting
the
status
upstream
directly,
so
I
think
we
are
going
to
start
with
a
prototype
outside
kk
first
and
then
move
the
tests
when
you
have
a
good
amount
of
of
them
right.
So
we
have
like
categories
of
tests
and
starting
with
the
gmsa,
we
are
starting,
creating
a
few
of
them.
So
we
have
two
options
here.
We
go
like
with
all
we
with
one
one
batch
of
yards
and
just
merge
everything,
or
we
can
like
prototype
everything
outside
kk
and
then
move
when
we
are
ready.
A
Thanks
for
the
update
next
announcement,
I
added
for
people
who
are
using
the
gmsa
web
hook
image.
We
have
made
some
progress
with
turning
that
into
an
official
kubernetes
image.
A
The
we
are
now
publishing
images
to
the
kate's
staging
gcr,
the
official
kid
staging
gcr.
This
is
the
path
for
the
current
container
image.
You
can
think
that
we
always
publish
the
latest
image
too,
so
folks
can
go
ahead
and
test
that
the
next
steps
is
to
we're
working
on
image
promotion
to
promote
it
to
the
prod
google
container
registry
images.
A
So,
if
you're,
using
that,
I
previously
were
publishing
the
image
to
the
segment
of
schools,
docker
hub
accounts,
but
we're
slowly
trying
to
stop
publishing
images
to
that
so
start
pulling
from
the
staging
repository.
If
you
get
a
chance
last
announcement,
james
did
you
want
to
take
this
one.
C
Sure
so
I
think
two
weeks
ago
we
announced
that
we're
just
moving
over
to
capsi
from
aks
engine
for
the
windows
tests,
and
so
we
merged
that
pr
today.
So
just
everything
124
and
forward
we're
going
to
be
using
cluster
apa
for
azure.
This
lets
us
have
all
the
code
inside
the
cncf
repositories
instead
of
aks
engine
and
should
make
it
a
little
bit
more
manageable
and
easier
for
folks
to
be
able
to
contribute
fixes
because
it's
built
on
top
of
cluster
api,
I'm
thinking
maybe
next
week.
C
A
That'd
be
great,
and
for
those
of
you
who
joined
at
the
end
of
the
sig
windows,
ci
meeting
we've
discovered
that
the
aks
engine
jobs
are
no
longer
reporting
logs,
so
I
think
james
said
that
logs
are
different,
but
right
now
we're
only
getting
logs
for
the
cap
c
jobs.
So
it's
another
reason
to
move
over.
A
Okay
does
anybody?
I
don't
see
anything
on
the
agenda
today.
I
don't
really
have
anything
either.
Does
anybody
on
the
call
want
to
do
want
to
talk
about
any
of
the
124
enhancements
and
the
status
of
those?
I
think
we're
coming
up
to
just
about
four
or
five
weeks
of
coding
left.
Otherwise,
if
anybody
has
any
suggestions,
feel
free
to
add
it
to
the
gender
just.
D
A
Yeah,
I
think
you
there
are
new
api
types
added
to
facilitate
the
kubelet
cube
ctl
communication
right,
yeah
yeah.
That
would
be
great.
Do
you
want
to
just
let
us
know
on
slack
when
you're
ready
for
that
first
round.
A
An
update
all
right
and
there
was
a
demo
for
anybody-
who's
interested.
There
was
a
demo
of
that
at
one
of
the
previous
meetings.
A
I'll
we
can
I'll
add
a
link
to
the
recording.
After
this
too
then
okay
sounds
good,
that'll
be
exciting.
A
Ravi
are
you
on
the
call
we
can
quickly
go
over
the
pod
os
field
to
beta
I
I
saw
that
there
was
a
pr
open.
Jordan,
leggett
had
a
couple
of
comments.
I
think
the
big
one
was
just
wanting
to
make
sure
we
had
all
the
necessary
test
coverage
that
we
were
committing
to
in
the
cap.
E
Sorry,
I
am
an
another
call,
so
I
may
not
be
able
to
give
an
update,
but
I
think
I'll
post
it
on
slack
in
a
few
minutes.
A
Okay,
that
works
I've
I'll
link
to
the
the
issue
or
the
pr
after
after
this
to
not
on
the
call
yeah,
I
guess
we'll
just
open
it
up
to
any
other
discussions
or
we
can
go
into
the
pairing
early.
A
F
All
right
go
ahead
whenever
okay,
so
the
problem
I
kind
of
wanted
to
talk
about
is
charts
that
don't
work
with
windows
right
and
there's.
There's
really
two
concepts
here:
there's
charts
that
can
deploy
to
windows
and
then
there's
charts
that
can
deploy
in
a
cluster
with
windows
right
for
mixed
workloads.
So
start
manager
is
one
of
those
that
I
just
have
it
set
up
here
easy
for
me.
F
So
if
I
install
this,
I'm
pretty
much
going
to
guess
that
it's
going
to
schedule
because
there's
no
node
selectors
defined,
I
could
add
some,
but
by
default,
there's
not
any.
If
I
install
this,
I'm
pretty
positive,
it's
going
to
get
scheduled
on
my
windows
note
and
when
this
is
done,
you'll
notice
that
it's
not
going
to
start
up
right
and-
and
that's
mainly
because.
F
So
we'll
see
this
comes
out
here,
I'm
curious
where
it
got
scheduled,
but
we'll
start
seeing
them
when
they
start
coming
up
like
the
injector
it
might
got
scheduled
for
windows.
Yep,
see
got
scheduled
on
windows.
This
is
going
to
fail
right.
This
is
a
common
problem
you
run
into
because
I
have
a
mixed
workload.
Cluster
right.
F
I
have
both
windows
and
linux
workers,
and
so
anything
you
deploy
is
really
going
to
fail,
and
so
you
know
you
could
easily
just
add
a
node
selector
to
this,
and
then
it
would
schedule
only
on
linux
right,
but
that's
not
by
default,
and
that's
that's
really
frustrating
for
me
because
I
just
want
to
be
able
to
grab
a
chart
and
deploy
it
right.
So
there's
our
error
image.
F
F
And
so
we
spend
a
lot
of
time
like
doing
some
of
the
under
underlying
work,
but
a
good
way
to
get
involved,
at
least
in
my
opinion,
is,
if
you
come
across
these
things,
just
kind
of
fix
them,
and
so
the
the
idea
here
is.
Is
that
like
what
do
we?
How
do
we
fix
this?
The
easy
answer
would
be
to
add
a
node
selector,
but
if
you
go
to
the
project-
and
you
can
see,
there's
actually
a
ticket
open
that
says,
you
know:
hey.
F
F
So
that's
kind
of
interesting
that
they're
asking
for
that.
They
consider
this
a
priority.
They
keep
saying
it's
not
stale.
They
really
want
somebody
to
pick
up
and
do
this
work
and
so
part
of
doing
this
kind
of
work.
As
some
of
us
know-
and
maybe
some
people
newer
to
the
sig
don't
know-
is
that
not
every
not
everything
just
works
on
windows
and
there's
lots
of
comments
in
a
lot
of
these
issues
open
around
this
like
well,
it
compiles
and
it
in
fact
does
compile
for
windows.
F
F
Zoom
is
really
annoying
these
bars
there
we
go.
So
if
I
pull
up
a
terminal,
I'm
in
here,
if
I
just
if
I
just
do
a
go,
build.
F
Typically,
if
you
just
try
to
build
something,
I'm
I'm
sure
most
of
us
kind
of
know
this
in
some
ways.
But
if
you
just
try
to
build
something
on
windows,
if
it's
pulling
in
something
that's
linux
specific
from
like
the
go
standard,
library
or
a
component,
then
it
won't
compile
right
because
it'll
fail,
because
that's
not
available
for
this
architecture,
so
this
compiled
so
the
next
step
in
some
of
this
stuff
is,
is
that
these
all
compile
which
I've
tested
most
of
these?
F
But
the
one
issue
comes
into,
is
now
that
I
have
this
compiled
binary
just
sitting
here.
This
is
not
using
the
build
system
at
all
that
they
use
here,
which
is
a
different
conversation,
and
I
don't
know
if
they're
using
seago
so
which
I'm
sure
in
some
instances
I
believe
they
are
so
that
requires
mngw
to
be
installed
right,
so
it
it.
This
gets
complicated,
really
quick,
just
trying
to
support
windows.
A
F
Yes,
and
so
that
gets
even
more
interesting,
so
I
don't
know
about
the
controller
runtime.
That's
the
first
thing
that
jumped
out
to
me.
It
seems
to
compile
just
fine
for
windows,
but
that
is
used
here.
So
I've
made
kind
of
notes
about
this,
so
my
idea
here
is
kind
of
just
to
walk
through
potentially
like
solving
this
problem.
F
You
know,
because
it's
it's
a
lot
of
a
lot
of
moving
pieces,
a
lot
of
parts
for
people
to
learn
and
kind
of
tackle.
If
you
wanted
to
even
start
doing
this
and
so
like,
for
example,
this
compiles
so
now
that
that
compiles,
I
should
figure
out
how
to
build
a
windows
image
for
it
and
then
from
there
you
know,
deploy
it
and
see
if
it
works,
run,
run
some
basic
tests
against
it
make
sure
that
it
actually
works
as
as
designed.
F
So
that
has
led
me
down
the
road
of
like
preparing
for
this
looking
through
here.
So
if
anybody
knows
basil
or
basil,
I
could
probably
really
use
some
help
there,
because
it
does
not
seem
to
work
on
windows.
A
Yes,
they
did.
Is
there
a?
Can
you
check
real,
quick
in
the
hack
folder?
Is
there
like
a
hack,
update,
bazel,
a
lot
of
projects
that
I've
seen
yeah
just
rely
on
these
scripts
to
do
to
update
all
the
basal
dependencies?
I've
never
done.
I
think
that's
about
the
extent
of
my
interaction
with
bazel,
so.
G
My
my
suggestion
suggestion
would
be
to
do
cross
compilation.
Basically,
you
would
build
the
windows
binaries
on
the
linux
node.
In
that
case,
you
won't
encounter
any
sort
of
dependency
issues
like
bezel
or
whatever,
whatever
and
basically
just
copy
over
the
windows
binary
into
windows.
Whenever
you
need
it
now,
you
can
even
do
this
for
container
images.
G
For
example,
we
even
do
this
for
a
couple
of
the
kubernetes
test
images
and
I
think
also
for
the
pause
image
which,
by
the
way,
also
requires
the
min
jab
mean
gw
compiler
for
windows,
so
we
are
actually
using
that.
Basically,
it
compiles
on
linux
in
a
linux
stage
and
just
copies
over
that
compiled
binary
into
a
windows
image.
Typically
a
nanoserver
image,
and
you
have
a
working
image
for
windows
yeah.
We.
F
So
you
know
that's
I'm
used
to
that
part.
It's
just
kind
of
an
interesting
thing
to
make
sure
that
you
get
this
all
built
correctly
and
where
are
the
right
hooks
at
because
I
seem
to
can't
be
no
I'm
using
open
souza.
So
you
know,
maybe
that's
my
problem.
F
It,
the
the
basil,
scripts
and
and
some
of
the
stuff
don't
seem
to
work
correctly,
at
least
on
tumbleweed,
so
I've
I
need
to
get
a
new
ubuntu
instance
set
up
to
to
see
if
it
works
better
there,
because
it's
kind
of
written
in
that
direction.
It.
G
Seems
I
think
that
might
be
the
best
option
in
my
opinion,
because
if
they
have
automated
image
building
processes
like
some
github
actions
or
whatever
they
might
already
be
using
an
ubuntu
image
or
something
and.
F
F
This
is
an
interesting
project
and
a
really
like
deep
one
to
dive
into
so
they
have
all
these
make
steps
for
building
and
they
use
ctr,
which
is
well.
This
is
their
tag
for
here
so
they're
doing
some
interesting
stuff
with
this
cmctl,
which
is
something
that
they
compile
themselves
here.
H
F
Jet
stack
is
like
the
sponsor
of
their
home,
repo
and
stuff
like
that,
but
I
don't
know
who's
officially
running
cert
manager.
They
do
have
a
slack
channel
that
I
need
to
get
plugged
into,
but
this
one's
a.
F
G
True,
I
also
have
some
vested
interest
insert
manager
and
I
actually
tried
it
out
last
week
and
it
didn't
quite
work
for
me.
F
G
F
To
get
it
set
up
and
figuring
out
how
to
use
the
injectors,
and
I
have
that
in
the
gmsa
chart
to
use
cert.
C
G
Yeah,
although
the
way
I
was
trying
to
use
it,
apparently,
you
can
also
add
an
annotation
into
the
in
into
the
engineering
english
resources
in
kubernetes
and
that
will
basically
cause
the
search
manager
to
auto
generate
certificates.
For
you
using
the
mentioned
search
provider
in
annotation,
and
you
can
even
use
different
plugins
for
cas,
for
example,
let's
encrypt
or
whatever,
but
it
didn't
work.
For
me,
I
mean.
F
G
A
See
here
also
cncf,
it's
a
cncf
sandbox
project
right
now,
sir
manager.
I
just
checked
so
it
is
under
the
cncf
review.
Now.
F
Okay,
good,
then
I'm
going
to
get
plugged
in
for
this,
but
I
think
this
is
just
good
to
talk
about,
because
these
are
the.
These
are
types
of
challenges
that
somebody's
going
to
face
like
getting
into
this
and
wanting
to
know
how
to
contribute,
and
so
right
here
is
where
I'm
using
it
to
inject
the
ca,
and
it
knows
how
to
for
web
hooks
to
automatically
add
the
the
ca
bundle
config
stuff.
F
If,
if
you
have
it,
enabled
so
yeah
it
just
it's
magic,
it
just
works
and
it
seems
to
have
worked.
But
it's
definitely
hard
to
read
through
and
logic
about,
and
I
looked
at
a
lot
of
web
hooks
that
you
start
managing
before
I
kind
of
got
the
idea
of
what
combination.
G
G
G
F
Of
stuff
to
learn
here
that
I
hope
to
like
post
back
and
share
for
like
some
of
these
more
complex
projects
and
problems,
because
this
one's
this
one's
been
frustrating
the
little
bit
I've
I've
looked
at
it
because
I
can't
just
check
it
out
and
build
it.
So
I
can't
even
just
you
know,
switch
architecture
and
get
all
the
things
in
right
places
and
then
worry
about
making
images.
H
So
one
thing
we
should
do
like
you
know
if
you,
if,
if
you're
filing
good
first
issues
for
these
and
other
applications,
we
need
help
on.
Like
you
know,
jamie
one
thing
you
could
do
is
like,
if
you
could
give
us
like
a
set
of
high
level
instructions
for
like
a
new
contributor
that
we're
like
okay,
like
you,
see
something,
and
you
want
to
run
on
windows
first.
H
Do
this
then
do
this
then
do
this,
because
I
was
just
doing
this
with
kp
g
last
night,
like
I
was
like
okay,
I
want
to
run
this
in
a
container
first,
let
me
compile
a
windows
binary
and
just
run
that
windows
binary
on
a
windows
os
to
see
if
it
works
right.
That
was
like
what
I
was
doing.
I
was
doing
that
yesterday
and
I
use
the
dev
tools.
H
Sig
windows
dev
tools
for
that,
so
so
I
did
that
and
it
worked
and
then
I
started
asking
myself
like
how
do
I
put
this
kpng
coop
proxy
thing
into
a
container,
and
then
I
didn't
really
know
how
to
answer
that
question
other
than
pinging,
maybe
james,
because
I
know
he's.
I
think
done
this
and
I
know
some
folks
at
ranch
have
done
it
and
then
I'm
like.
F
Yeah,
I
think,
going
forward
to
me.
Co
brings
a
lot.
Ko
brings
a
lot
of
stuff
to
the
table,
so
that's
that
uses
crane,
which
we
had
the
demo
for
a
few
months
ago,
yeah,
I
think
jason
hall
came
and
demoed
that
yeah
a
co
is
made
specifically
for
building
go
projects
and
putting
them
in
windows
containers
all
from
any
platform.
So
it
doesn't
matter.
You
don't
need
a
windows
box
to
build
the
windows
container.
It
just
layers
it
in
to
the
existing
base
image
that
you
pick
now.
F
A
F
A
A
Yeah,
I'd
also
like
to
highlight,
too
that
I
think
that,
where
we
started
was
that
there
was
a
no
node
selector
on
the
on
the
issue,
one
thing
that
I've
seen
in
a
couple
places,
including
one
of
the
projects
that
I
was
converting
to
work
for
windows,
was
that
even
once
we
get
a
container
image
out
that
works
for
windows.
There's
no
guarantee
that
the
fields
that
the
pods
that
need
to
be
set
in
the
pod
specs
are
even
going
to
be
compatible
between
windows
and
linux.
A
A
Now,
if
we
do
start
using
like
helm,
helm
is
smart
enough
to
you
can
set
in
like
a
configuration
for
what
os
to
use
and
we
could
kind
of
conditionally
put
in
all
the
you
know,
specific
pod,
spec
fields
for
each
os
2.
But
I
think
that's
another
common
concern
that
people
who
are
focused
on
linux
and
the
linux
deployments
don't
really
realize
either.
So,
just
because
we
can
build
the
container
images
and
binaries
to
run
on
windows,
and
linux
doesn't
mean
that
we
can
share
the
same
deployment.
C
C
F
What
I
would
what
I
would
like
to
happen?
Is
it
just
install
without
me
thinking
about
it
from
upstream,
which
means
that.
C
C
I
think
it
would
be
maybe
a
bad
situation
if
a
customer
deploys
this
thing,
they
think
they
deployed
it
to
linux.
It
ends
up
on
a
windows
node
and
then
they
have
some
sort
of
strange
issues
because
it's
not
tested
or
maintained
by
anybody.
So
I'm
wondering
if
we
actually
go
back
to
them
and
just
say
you
know,
we've
discussed
this
and
maybe
we
need
to
approach
it
from
a
different
angle.
F
F
I
want
to
get
it
running
and
see
if
it
works
or
doesn't
work
without
any
code
changes
and
if
it
doesn't
work
without
any
code
change
changes,
then
I
think
you
know
coming
back
with
even
more
of
the
information
and
you
know
coming
back
with,
like
we
tried
this
here's
the
problem,
here's
here's
the
difficulty
there
and
this
is
what
it's
going
to
take
to
change
the
project
that
I
think
that
they'd
probably
be
more
receptive
of
doing
node
selectors
at
that
point,
so
I'm
trying
to
do
all
the
work
up
front
to
like
plus
it's
kind
of
a
fun
little
thing
to
work
on
right.
F
It's
like
how
do
you
take
something
complex
and
make
it
work
in
a
windows,
a
cluster
with
windows,
notes,
and
I
think
you
know
I
just
want
to
approach
it
as
this
is
what
they
want.
Let's
try
to
do
it
and
if
and
if
it
doesn't
work
or
we
find
the
thing
that's
going
to
prevent
it
from
from
happening,
then
we
can
just
come
back
and
have
the
conversation
about.
F
F
A
I
think
that
yeah,
I
think
that
maybe
the
first
step
in
that
guide
that
jay
is
suggesting
to
write,
I
think,
should
probably
be
do.
Does
this
even
make
sense
to
run
on
windows?
I
think
that
there's
definitely
cases
like
both
cases.
There's
a
lot
of
projects
like
you
know,
node
exporter,
that,
yes,
this
is
a
like,
should
run
as
a
demonstration
on
every
node
in
your
cluster.
So
in
those
cases
we
do
want
this
to
run
on
windows
and
linux
things
like
cert
manager.
I
don't
know
it's
like
it.
A
A
And
as
we're
talking
about
this
more,
I
do
see
that
guy
that
jay
suggested
to
write
could
be
good
to
publish-
and
you
know
point
to
like
reference
from
these
issues-
that
the
the
issues
in
these
github
requests
where
people
are
like.
You
know
what
we
want
to
support
windows,
but
we
don't
know
how
to
do
it
so
this
could
be.
I
think
this
could
add
a
lot
of
value.
F
Yeah,
I'm
making
a
note,
and
I
think
I
think
this
is
all
a
good
point,
because
this
ties
into
windows,
operational
readiness
right
like
at
a
high
level
like
does
all
your
things
going
to
work
in
a
in
a
cluster
with
windows.
You
know:
can
everything
be
deployed
without
any
errors?
Do
you
have
node
selectors?
Do
you
have
your
affinities?
Do
you
have
all
these
things
like?
F
I
think
that
kind
of
feeds
into
that
same
same
line
of
thinking,
because,
because
that's
that's
a
big
conversation
we're
having
right
now
is
you
know
we
we
have
mixed
workload:
environments
like
we
we're
that's
a
big
story,
we're
kind
of
pushing
and
if
charts
aren't
compatible
with
mixed
workloads
or,
if
charts,
don't
support
it
without
modification
at
installation
which
can
be
done
almost
all
charts
support,
adding
node,
selectors
or
node
affinities
or
whatever
you
want
to
do.
F
It
still
is
just
a
problem
that
people
need
to
know
to
do
that,
and
you
know
it's
it's
just
one
of
one
of
those
things.
I
think
that
you
know
I
added
windows
to
my
cluster
and
out
broke
everything.
H
F
Okay,
well,
that's
that's
all
pretty
much.
I
had
was
just
like
high
level
the
problem,
and
I
told
jay
I
could
do
something
like
every
other
week
or
or
whenever
I
get
some
progress
on
this
and
kind
of
show
what's
happening
and
what
was
done
to
like
move
it
forward
and
I'll
start
writing
up
this
guide
as
I
go
so
that
way,
we
can
have
something
to
point
people
to
to
help
them
like
make
the
decision
if
they
want
to
take
on
the
work
or
not,
do
it
because
yeah.
A
And
I
can
help
find
a
place
for
that
guide.
We
recently
the
some
of
the
leads
were
doing.
The
annual
report
and
part
of
that
was
doing
a
scrub
of
like
the
sig
windows
like
contributing
guide,
and
this
might
be
a
good
I'm
wondering
like
we
should
at
least
link
it
to
there,
but
I
can
help
find
the
place
for
that
and
publicize.
It.
F
F
You
know
they
just
they're
picking
they're
picking
a
choice
that
they
prefer
and
you
know
we're
gonna
have
to
we're.
Gonna
have
to
do
the
work
or
somebody's
gonna.
Do
the
work
to
either
say
this
this
this
efforts
not
work
that
are
not
gonna
work.
You
know
and
go
with
a
different
solution,
so
I
think
that's
going
to
be
that's
going
to
be
the
challenging
part
with
some
of
these
projects.
A
F
F
From
that
standpoint,
and
then
also
it's
very
much
focused
on
linux
right,
so
it's
hard
to
tell,
but
we
know
what
they
want,
and
so
I'm
gonna
I'm
gonna,
try
to
go
through
and
do
some
of
this
work
just
learning
about
other
projects
and
and
stuff
like
that.
So
I
figured
I
would
share
that
with
everybody
as
I
go
through.
A
A
F
F
A
F
A
Are
you
going
to
caveat
that
with
I'll
stop
the
recording
and
then
we
can
talk
a
little
bit?
Okay,.