►
From YouTube: Kubernetes SIG-windows 20211019
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
All
right,
hello,
everybody
and
welcome
to
the
october
19th
2021
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.
Okay,
let's
jump
into
some
announcements.
First
announcement
on
the
list
is
that
container
d
1.6
beta
was
released
last
week.
I
believe
it's
their
beta0.
A
This
has
a
couple
of
different
up.
It
has
some
updates
to
the
hcs
shim
and
which
fixes
a
couple
of
different
issues
that
we're
seeing
james
was
just
talking
about
that,
but
it
also
is
the
first
container
d
release
that
will
have
proper
host
process
container
support.
So
if
people
are
interested
in,
if
people
are
planning
on
testing
out
container
d
support
with
kubernetes
on
windows,
please
give
that
a
try
and
see
if
you're
running
into
any
issues
or
have
any
feedback
with
that.
A
That's
kind
of
related
to
the
next
announcement
is
that
docker
shim
is
still
slated
for
removal
for
124.
I
think
the
current
plan
is,
and
dims
and
sergey
have
sent
out
some
emails
to
the
google
group
saying
docker
like
that
code
path.
A
That's
slated
to
be
one
of
the
first
things
to
go
into
124.,
so
if
I
wanted
to
again
kind
of
just
open
it
up
and
see,
if
anybody
has
any
concerns
about
this
for
for
windows,
I
know
that
in
sig,
node
sergey's
been
doing
a
good
job
of
reaching
out
to
a
number
of
like
the
different
cloud
providers
and
seeing
where
they
are
with
the
migration.
A
And
if
there's
any
concerns,
I
think
everybody
so
far
seems
kind
of
in
a
similar
situation,
where
there's
not
as
much
adoption
as
people
would
like
to
see.
But
the
thing,
but
the
adoption
that
they
do
have
is
at
least
on
par
with
kind
of
performance
and
reliability
with
docker.
A
Does
anybody
have
any
kind
of
comments
or
concerns
with
docker
shim
removal
kind
of
specific
to
windows
workloads
at
this
point.
A
If
not
or
if
you
do
feel
free
to
erase
it
to
just
different
various
channels
in
slack
either
sig
windows,
sig,
node
or,
and
we
can
get
that
routed.
B
Mark,
I
don't
have
any
particular
concern,
but
I
think
in
the
same
thing
it
was
mentioned
that
datadog
or
something
I
don't
know
how
accurate
the
stats
is.
But
six,
you
know
only
six
percent
of
the
all
the
clusters
are
using
continuity.
The
rest
of
them
are
all
using
docker.
B
That
seems
like
a
pretty
big
percentage
and
I
think
there
was
concerns
of
like
we
should
document
something
in
hand-holding
folks
who
are
about
to
move,
because
you
know
if,
if
94
of
the
of
the
clusters
in
the
world
are
running
docker-
and
this
is
much
more
than
we
thought
so
I
don't
know
if
we
should
do
from
a
windows
perspective
like
some
documentation
upstream
like
is
that
needed
or
something
else.
A
B
A
Documented
on
the
kubernetes
website,
how
to
install
docker
with
on
windows
nodes
and
how
to
configure
the
cubelet
with
that,
and
my
hopes
are
that
after
that's
done,
it's
seamless,
but
if,
if
anybody
feels
like
that's
not
sufficient
or
needs
more
information
yeah,
we
can
definitely
look
into
that.
C
C
A
Yeah,
that's
that's
also
kind
of
called
that
in
the
docs
and
it's
pretty
much
bring
up
a
new
node,
that's
running
that's
configured
with
docker
they're.
I
think
signode
and
a
couple
of
other
folks
have
kind
of
explicitly
said
that
they're
not
support
or
like
they're,
not
documenting
or
advocating
for
in
place
upgrade
or
in
place
changing
of
the
cri
on
nodes.
A
C
C
Know,
especially
in
the
bvio
space,
where
they're
bringing
their
own
windows
instance
right.
If
you
ask
them
to
go
away,
reconfigure
everything
and
bring
it
back,
I'm
not
so
sure.
If
they're
going
to
be
super
happy,
but
we're
still
at
the
beginning
of
the
journey,
it
might
be
a
no-go
like
that
there
could
be
lots
of
issues
trying
to
do
it
in
place.
A
C
Okay,
I
can
talk
to
the
folks
that
must
ignore
also-
and
you
know,
ping
them
and
see
what
goes
on,
but
that's
good
to
know.
The
other
thing
I
was
going
to
call
out
is,
you
know
it's
very
easy
for
customers
on
on
windows
to
install
docker.
C
It's
not
that
easy
to
install
container
d.
I
don't
know
if
that's
going
to
be
another
barrier
for
some
customers.
I
know
it's
all
documented,
but
still
like
it's,
it's
not
like
you,
you
could
go
and
I
I
think,
enable
what
is
that
enable
something
and
just
say,
add:
docker
and
docker
would
just
get
installed
right.
It
was.
C
A
Some
powershell
commands
around
it
that
you
could
run.
You
could
say
you
know,
yeah,
just
install
docker
and
then
it'll
tell
you
you're
ready
to
reboot,
and
then
it
comes
up
and
docker's
running
yeah
yeah.
E
C
A
Yeah
the
steps
for
getting.
Let
me.
D
I
think
with
I
don't
know
if
it's
in
20
2019,
but
in
2020
is,
is
winged
available
on
the
server
now.
Is
that
something
that
like
would
be
make
it
easier
for
folks?
I
don't
know,
I
don't
know
much
about
that,
but
I
just
was
thinking
like
the
way
you
do
it
on.
Linux
is
be
an
apt-get
install
kind
of
thing,
and
so
I
wonder
if
we
had
a
win
get
type
of
thing
for
windows,
if
it
would
make
it
easier.
A
That
might
be
something
that
yeah
our
sig
should
try
and
and
publish
or
see.
If
we
can,
if
anybody
from
the
continuity
community
wants
to
work
on
that
or
not
the
yeah.
B
A
The
the
steps
I
I
post
a
link
to
the
steps
for
installing
container
d
on
on
windows
and
they're,
not
that
much
more
complicated
arvind.
I
think
the
there
are
a
lot
more
knobs
and
you
know
settings
that
you
could
tweak
that
may
not
be
they're
not
necessary,
but
you
you
could
tweak
those.
I
think.
That's
where
the
differences
are.
I
think
docker
is
a
lot
less
configurable,
but
also
just
a
lot
more
kind
of
purpose
built
for
for
doing
what
it
does
and
running
containers
there.
But
the
basic
steps
are
you
download?
A
The
package
extract,
the
contents
of
the
package
use
container
d
to
spit
out
a
default
config
file
and
then
edit
the
config
file
and
then
run
the
service,
and
also
we
have
found
that
if
you
are
running
container
d,
it
is
very
important
to
add
a
defender
exclusion
to
point
to
the
executable,
wherever
that
is
on
your
system.
That
will
help
solve
a
lot
of
performance
issues
that
folks
have
been
seeing
and
reporting.
A
A
C
Yeah
and
the
only
other
thing
that
I
have
noticed,
which
might
be
an
issue
for
some
some
folks
is
so
in
our
configuring.
In
our
cubelet
configuration,
we
increase
the
image,
pull
timeout
to
like
30
minutes,
because
all
customers
were
refusing
to
people
images
they
were
like.
We
don't
want
to
people
images.
C
That
option
is
not
available
with
continuity,
so
that
might
be
another.
I
don't
know
if
everybody
is
using
the
increased
people
time
for
the
cubelet.
That
is
not
supported
by
continuity.
You
can
set
it
in
the
cubelet,
but
it
doesn't
take
effect.
C
C
A
Cardio,
do
you
know
if
that's
the
case,
if
not,
we
can
come
back
to
this
later,
but
yeah.
That
is
a
good
thing
to
call
out,
especially
with
cervical
images
being
so
large.
A
So
arvind
was
saying
that
in
with
docker
shim
there's
a
cubelet
flag
for
setting
the
image
pull
timeouts,
that
they
set
very
for
a
long
time
out
in
their
clusters,
and
I'm
we're
wondering
if
there's
a
similar
way
to,
but
that
setting
only
would
take
effect.
If
you're
running
with
docker
shim.
We
were
wondering
if
there's
a
similar
setting
or
way
to
configure
that
behavior
for
container
d.
G
Oh
I'll
have
to
check
as
well,
but
from
what
I
saw
there
is,
but
I
may
be
mistaken,
I'll,
take
a
look
right
now.
D
C
A
C
A
Jay
all
right
and
let's
see,
if
anybody
has
any
other
announcements,
free
people,
please
feel
free
to
speak
up,
if
not
I'm
gonna
hand
it
over
to
moz
to
do
a
quick
recap
of
his
kubecon
experience.
D
One
thing
I
wanted
to
mention
about
the
container
d
1.6
beta
with
the
host
process:
support
is,
if
you
are
using
annotations
prior
to
prior
to
this
you'll,
want
to
remove
them
from
your
container
d
config.
I
was
talking
with
perry
and
I
he
didn't
realize
that
we
don't
need
those
anymore.
D
The
post
process
field
is
getting
passed
through
the
cry
layer
now,
which
means
you
can
set
like
some
sort
of
policy
saying
you
can't
set
this
for
these
namespaces
and
if
you
pass
the
annotations
through
on
the
container
deconfig,
then
somebody
could
potentially
bypass
that
type
of
policy,
so
just
wanted
to
quickly
mention
that
if
anybody
was
playing
around
with
using
annotations
earlier,
I
cannot
so
the
annotations
don't
work
or
they
do
they.
They
do.
But
I
don't
mostly
most
cases
you
don't
need
it
anymore.
A
Yeah
in
122,
when
the
feature
was
in
alpha
the
cube.
We,
we
authored
the
cubelet
to
send
both
the
annotations
on
the
cri
calls
and
to
populate
this.
The
structures
in
the
cri
call
and
but
container
d
initially
only
understood
the
annotations.
But
after
that
implementation
went
into
container
d,
it
was
no
longer
necessary
to
pass
in
the
annotations
across
as
as
well
and
just
recently,
we've
removed
the
code
to
set
the
annotations
on
the
cubelet.
F
A
There's
fields
on
the
cri
on
the
cri
objects
that
get
passed
over
on
like
container,
create
that
say
that
this
is
a
host
process
container.
Oh.
A
That's
that
the
exact
fields
are
in
the
cup.
A
All
right,
yeah
muz,
do
you
want
to
share
a
little
bit?
I
know
muzz
actually
was
able
to
attend
kubecon
in
person
and
I
think,
attended
the
save,
meet
and
greets
and
met
with
some
users
and
customers
and
was
curious.
How
everything
went.
B
Sure
yeah
I'll
try
to
quickly
some
recap
the
experience,
so
it
was
pretty
great
to
be
back
at
cubecon.
I
think
people
were
very
excited
who
who
were
comfortable
to
be
there,
and
it
was
a
lot
less
audience
like
you
know.
Normally
we
get
like
fifteen
thousand
twenty
thousand
people,
but
with
like
four
thousand
people.
So
so
you
know
the
the
crowd.
B
Wasn't
that
much
so
in
terms
of
from
a
windows
perspective,
specifically
so
on
on
the
session
mark
uj-
and
you
know
the
others
did,
I
heard
quite
a
feedback
after
the
session
like
outside,
where
people
liked
the
the
dev
aspect,
especially
what
jay
has
been
working
on
how
to
set
up
dev.
B
They
really,
I
mean
a
couple
of
people
said:
I
love
that
this
session
was
so
like
dev
oriented
I've
been
to
previous
sessions,
and
it
wasn't
that
much
dev
focus
or
developer
focus,
so
that
kind
of
resonated
with
with
folks
the
host
process
containers
people
loved
it.
They
said
literally,
like
there
were
two
folks
there
who
said
literally
that's
the
thing
that
was
stopping
me
from
using
windows
containers.
So
so
I
think
that
that
was
kind
of
resonating.
B
The
other
thing
I
was
kind
of
surprised
too,
but
it
was
kind
of
a
team
across
the
board.
Wherever
I
went
to
whether
it's
you
know
sessions
or
anything
related
to
windows,
people
were
very
interested
in
windows,
server,
2022,
which
was
a
little
bit
surprising
to
me,
because
I
mean
it's
pretty
new,
you
know
you
could
have
a
lot
of
questions
about
windows
which
is
not
windows,
server
202,
but
it
seems
like
there
is
a
lot
of
interest
and
wait
for
windows
server,
202
being
supported
in
kubernetes.
B
I
was
asked
many
times
when
it's
going
to
be
supported.
I
told
them.
You
know
the
the
tests
are
being
done.
There
are
multiple.
You
know
folks,
trying
to
to
do
that
when
I
dug
into
the
windows
server
202,
why?
It's
being
so
popular
one
is
a
lot
of
folks
starting.
They
just
want
to
start
directly
with
windows
server
2282..
B
The
other
surprising
thing
was
the
argon
compatibility,
the
backward
compatibility.
People
are
just
afraid
that
you
know
even
the
patches,
which
is
not
true
by
the
way,
like
the
the
patch,
the
patch
mismatch,
even
in
windows,
server
2019.
B
So
if
your
host
is
on
a
different
patch
and
your
your
container
is
on
a
different
patch,
it
should
totally
work
as
long
as
you
know,
the
major
version
is
the
same,
and
I
think
it's
pretty
documented,
but
folks
are
very
afraid
and
they
want
to
use
this
windows
server
2022
when
it's
like
the
argon
backward
compat
compatibility.
We
announced
they're
very
interested
that
because
they
just
feel
safer
to
use
windows,
server,
202.
argon.
B
For
folks
who
might
not
know
that
yeah
sorry
yeah,
that's
how
we
internally
say
microsoft,
so
yeah,
so
the
the
the
kernel
backward
compatibility
that
brandon
announced
moving
forward
for
windows,
server
22.
that
resonated
a
lot
with
folks.
The
last
thing
was
the
service
mesh
with
windows.
Server
2
actually
was
announced.
I
don't
know
if
you
guys
saw
it,
though,
on
the
osm
demo,
people
loved
it.
Actually,
I
heard
a
couple
of
folks
talking
about
it
outside,
so
that
was
you
know
again
a
little
bit
surprising
one.
B
Other
thing
I
think
I
noticed
which
is
you
know,
which
speaks
pretty
highly
of
this
sig,
so
I've
been
to
kubecon
like
since
probably
2017
and
every
kubecon,
like
majority
of
the
folk,
would
be
hey.
Windows
container
is
a
thing,
and
this
is
probably
the
first
time
where
most
of
the
folks
I
met
they're
already
either
using
windows
containers
on
kubernetes
or
you
know,
planning
to
use
it.
You
know
heavily.
B
I
think
there
were
like
very
few
folks
who
were
still
surprised
by
windows
containers,
so
that
was
a
little
bit
different.
The
last
thing
I
wanted
to
bring
up
here-
and
I
think
luther
is
on
the
call
he
brought
it
up
at
sigmeat,
and
then
I
met
a
couple
of
folks
afterward
as
well,
and
it
kind
of
resonated,
which
is
basically
there
are
folks
in
the
community
in
general
general
who
want
to
contribute
to
windows.
B
Containers
like
you
know,
bringing
a
parity
with
with
linux,
but
they're
not
they're,
not
able
to,
because
you
know
the
getting
started
or
getting.
You
know
vms
that
can
run
kubernetes.
You
know
windows,
so
basically
getting
the
windows
licenses
and
then
getting
those
vms
kind
of
becomes
a
hurdle
for
them,
and
I
think
jay
and
and
then
who
else
like
there
are
great
people
working
here
on
that
ross
and
all
the
other
folks
james.
B
But
I
think
there
is
a
little
bit
lack
of
awareness,
probably
on
that,
and
then
also
I
don't
know
how
the
licensing
works
and
stuff,
but
there
was
definitely
a
feedback.
So
I
don't
know
luther
if
you're
here,
if
you
want
to
say
something
on
that,
I
also
heard
the
same
feedback
from
argo
folks
who
were
working.
They
said:
hey
we're
happy
to
do
it
for
windows,
but
don't
know
how
to
get
started,
and
there
were
other
folks
like
that
as
well.
You
wanna
yeah.
E
Man,
this
is
100
a
problem
that
we've
seen
like
we
go
into.
Wm
exporter
was
one
that
was
a
big
one.
The
prometheus
guys
have
that
exporting
tool
right
and
they
didn't
care
to
even
try,
building
on
windows
and
see
what's
going
on,
and
so
we
just
went
in
and
started
doing
it
for
them,
so
I'm
gonna
add
so
we
were
talking
about
getting.
E
Resources
is
what
moz,
and
I
were
talking
about-
was
I'm
going
to
add
flannel
to
use
github
actions
and
try
to
use
some
of
the
2019
2022,
the
2022
beta.
That's
there
to
see.
If
we
can't
build
some
images,
we
don't
use
them
yet
right,
because
it's
not
they
can't
do
images
can't
run
the
cni
yet,
but
I
100
want
to
start
building
them
and
see
how
it
works.
E
Because
then
I
can
transport
that
to
the
wmi,
kids
and
some
other
stuff
there's
we
have
people
who
make
helm
charts
that
have
a
windows
component
to
it
are
monitoring
and
logging
and
just
getting
those
projects
to
update
their
stuff
has
been
a
pain.
Drone
has
been
a
pain-
I
don't
know
if
you
guys
have
ever
had
to
get
your
ci
running
on
multiple
versions
of
windows,
but
it's
a
nightmare.
E
I
thought
techdon
just
got
their
stuff
released
right,
but,
like
drone
had
the
different
sacs
and
it
was
very
hard
to
get
stuff
updated,
they
still
haven't
updated
it
right.
So
I
don't
know
a
bit
of
pain
and
I
think
anything
that
helps
people
generate
images
is
it's
just
where
you
need
to
start
just
get
people
trained
on
it.
H
E
And
go
through
how
it
all
works.
We
can
hang
out
for
a
couple
hours,
that's
always
fun,
but
the
the
problem
we
have
is
that
there
are
perfectly
good,
simple
network
apps,
not
even
complicated,
that
could
be
running
on
windows
containers.
I
could
be
running
like
just
an
api
or
something
just
sitting
there,
but
nobody
has
any
motivation
to
build
an
image
and
it's
usually
resources.
E
A
I
was
going
to
say
so,
like
claudio
did
a
ton
of
work
and
like
we
had
the
same
problem
in
the
kubernetes
project,
for
the
pause
image
and
for
all
the
test,
infrastructure
images
and
what,
with
docker,
build
kit,
we're
actually
able
to
build
simple
windows,
container
images
on
linux
nodes
and
that's
version
like
windows,
os
version
agnostic.
Now
there
are
a
lot
of
caveats
from
that
like
which
the
the
set
of
commands
you
can
use
in
the
docker
file
are
limited.
A
I
think
they're
limited
to
things
like
from
copy,
add
and
like
end
start
and
a
couple
things
like
that,
but
if
for
very
simple,
like
apps
that
where
it's
like,
I
have
a
binary
and
some
config,
and
I
want
to
just
stick
it
in
an
image
and
get
it
to
run
on
windows.
We
figured
out
how
to
do
that
from
linux
builds.
A
So
maybe
another
thing
is
to
evangelize
that
a
little
bit,
because
if
I
can
understand
that
people
don't
want
to
mess
with
you
know,
windows
hosts
just
to
build
a
like
a
smaller
part
of
their
project.
But
if,
if
we
could
say
you
know,
here's
a
way,
here's
a
docker
file,
here's
how
to
author
it.
So
you
can
build
it
on
the
same
hosts
that
you're
building
your
linux
container
images
on.
I
wonder
if
that
would
help
increase
adoption.
E
G
So,
as
mark
mentioned,
you
can
actually
build
a
lot
of
things
using
docker
build
decks,
it's
not
just
those
commands.
The
trick
is
to
use
multi-stage
docker
files,
which
basically
means
you
have
a
linux
stage.
You
can
do
literally
anything
you
want
there
of
even
for
pause,
images
and
stuff
like
that.
We
do
use
cross
compilation.
G
E
Is
exactly
what
we
do
in
one
of
our
projects
too,
and
that
works
perfect,
but
it
was
having
that
base
server
core
image
that
we
had
to
start.
Everything
from
my
my
convo,
with
with
muzz,
was
more
around
the
lines
of.
If
you
have
open
source
projects
out
there,
who
could
easily
have
a
windows
opportunity
like
just
build
it,
I
mean
there's
things
that
could
just
have
an
image
and
it
would
run
if
they
would
just
if
they
would
just
add
a
manifest
for
it,
make
it
simple.
E
It
was
like
get
the
doctor
or
sorry
the
github
action
stuff
working
and
then
just
like
start
checking
in
a
commit
and
say
look,
here's
an
action.
It'll
do
exactly
what
you
need
just
add:
a
docker
hub,
login
and
password
and
you'll
build
your
images
like
make
it
simple
that
was
kind
of
my
thing,
like
everything
you
guys
are
talking
about,
is
way
advanced
for
most
normal
people
who
just
want
to
build
some
goofy
network
app
with
an
api.
B
That's
accurate,
so
I
think
mark
the
feedback
is
and
we
can
get
into
it.
Thank
you
for
bringing
it
up
because
number
number
one
is.
There
are
a
lot
of
projects
now
in
cncf
there
are
an
incubation
state
or
different
states
that
if
there
is
demand
for
windows
and
I'm
happy
to
be
in
sigmund
or
anybody
wants
to
partner
up
with
me
to
go
to
those
things-
and
you
know,
as
luther
is
saying,
tell
them
hey-
you
want
to
build
this
for
windows
as
well
right.
B
So
that's
one
thing
number
two:
is
this
awareness
or
like
documentation,
or
you
know
whatever
jay
and
others
like
you
know,
everybody's
doing
here
for
the
dev
tools?
Is
there
something
else
needed
right
when
we
are
interacting
with
these
other
folks,
like
drone
flannel,
argo
like
these
are
all
wanna.
Do
it
on
windows,
but
apparently
it's
not
easy.
That's
the
feedback,
we're
getting
right.
So
could
we
do
something
else?
Maybe.
H
We
could
win
a
pet
store.
Maybe
what
we
need
is
like
you
know
like
we
need
that,
so
we
have
an
image
builder
that
we
put
in
the
depth.
Maybe
we
need,
like
those
three
things
right,
drone
flannel,
argo
right
like
maybe
we
should
literally
have
a
thing
in
there
like
okay,
you're,
a
developer.
You
want
an
environment
to
test
and
here's
how
you
can
do
it
you're
a
developer.
You
want
to
publish
things,
here's
what
here's
a
way
to
do
it.
It's
like
we
could.
H
We
could
integrate
that
with
the
rest
of
the
process
when
we,
when
we
spin
up
that
dev
thing
so
that,
like
at
least
next
time,
somebody
somebody
wanted
to
publish
a
windows
image.
They
had
a
single
place
to
go
where
they
could
spin
up
the
environment
to
test
it
on,
but
they
also
could
literally
follow
directions
in
there
to
build
the
images
according
to
claudio's
and
and
mark's
sort
of
some
of
the
blueprints
that
they
just
mentioned.
H
But
then
again
I
see
what
you
mean
luther
that's
still
kind
of
advanced,
but
it's
less
advanced
than
than
the
current
situation.
We
have
where
you
have
to
like
go
on
the
internet
and,
like
I
don't
know,
look
at
how
the
pause
image
is
built
right.
That's
it's
like
it's
like
an
incremental
step
right.
I
think.
A
And
I
I
also
wonder
if,
like
a
couple
times,
people
said
they
were
happy
that
sac
releases
are
going
away.
I
wonder
if
that's
going
to
simplify
things
too,
because
I
am
not
positive,
but
I'm
pretty
sure
that
there
are
github
actions
where
you
can
request
the
machine
like
a
windows
machine,
but
it
only
supported
windows,
server
2019
at
the
time.
So
I'm
wondering
if
a
lot
of
people
are
like.
A
Oh
no
there's
a
whole
bunch
of
versions
of
windows
and
I
only
see
one
of
them
supported
and
it
doesn't
look
like
the
latest
version.
So
I
don't
want
to
figure
this
all
out.
But
now,
if
there's
going
to
be
like
windows,
server,
2019
windows,
server
2022,
when
you
can
have
a
github
action
to
build
your
container
images
that
might
simplify
things
like.
E
That
has
made
my
life
so
much
easier.
I
have
to
say
I
appreciate
it.
It's
100
a
better
solution,
because
not
only
that,
but
you'd
have
people
who
would
never
upgrade
so
we'd
spend
all
this
effort
and
time
into
2004
20
h2
and
everyone
was
like.
Well,
it's
not
an
ltsc,
we're
sticking
on
2019
we're
like
well.
We
just
wasted
our
time
thanks,
nothing,
you
could
do
about
it.
H
E
There
are
so
there
are
so
many
tools
that
would
work
perfectly
fine
on
windows
nodes
perfectly
fine,
but
we
have
to
we
always
say
we
have
to
have
a
linux
worker
to
be
able
to
run
a
kubernetes
cluster
and
I'm
not
convinced
that's
true.
There's
things
coordinated
that
you
can
just
compile
and
get
it
running.
They
just
don't.
Have
it
right
like
you
could
get
to
a
point.
That's
that's
my
theory.
B
C
B
C
Second,
summary
is,
I
think,
we've
been
discussing
this
topic
for
a
while,
and
we
finally
decided
we
need
to
document
the
current
state
of
projected
volumes
that,
I
think
is
a
given.
I
can
add,
I
can
open
a
pr,
a
doc
pr
to
add
it
to
maybe
the
projector
volume
page.
I
think
the
topic
for
discussion.
I
think
james
and
I
were
briefly
discussing
this-
is
in
openshift.
C
We
have
a
little
bit
of
a
hack
if
some
user
is
on
a
cluster
where
there's
an
admission
controller
that
actually
adds
security
context,
and
you
end
up
with
run
as
user
and
run
as
username
on
a
pod
on
a
windows
pod
that
windows
pod
will
not
come
up
and
to
overcome
that
issue,
we
sort
of
don't
try
and
do
the
ch
own
for
for
the
windows
case,
james
and
I
were
discussing
whether
we
should
just
document
this
as
in
hey,
don't
do
this
or
should
we
add
this
as
an
up
as
a
as
a
workaround
in
upstream.
B
Also
so
aravind
brandon
is
from
the
windows
side
who
are
like
you
know,
expert
in
this
area
right
over
the
dev
like
maine
he's
following
up
on
that,
so
I
would
say,
add
it
as
an
option
whatever,
like
you,
feel
like
it's
the
right
thing
to
do.
We
can
get
it
reviewed
and
then
go
from
there
we'll
put
the
weight
you
know
behind
it.
I
Yes,
sorry
I've
been,
this
has
been
on
my
list
for
a
while.
I
just
a
hit
with
a
bunch
of
different
things,
so
I
I'm
definitely
let's
let's
chat
about
this.
I
I
I
want
to
get
this
documented
and
on
the
road.
C
C
B
I
Yeah
yeah,
that's
yeah,.
D
D
Get
that
I
think
it's
a.
A
I
was
gonna
re-add
it
to
the
list.
I
think
that
we'll
want
to
like
if
we
decide
that
we
do
want
to
try
and
change
it
upstream.
We'll
probably
need
to
make
a
case
to
do
that
in
sig
node
as
well,
and
they
might
have
some
feedback.
So
maybe
we
should
maybe
next
week
I'll
add
it
to
the
sig
windows
agenda
and
I
can
also
add
it
to
the
sig
note
agenda
and
you
can
talk
about
it.
I,
like
sig
node
and,
if
folks
can't
join,
I
can
report
back,
sounds
good.
So
that's
good.
A
F
Bye
starting
the
pairing
now,
if
folks
want
to
hang
out
all.