►
From YouTube: Kubernetes SIG Windows 20220719
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
A
Let's
get
started
with
some
announcements
first
announcement
for
today
is
that
there's
a
pr
open
to
remove
the
hns
v1
code
pass
from
q
proxy.
A
I
believe
that
we
had
discussed
previously
that
all
the
cni's
are
likely
on
the
h
and
sv2
apis,
so
this
should
be
safe
to
do
if
anybody
has
any
concerns
about
that.
Please
comment
on
the
pull
request:
james
jiffany,
any
anything
else
with
that,
or
is
that
pretty
much
it.
B
Yeah,
I
know
that's
pretty
much
it.
I
think
we
actually
broke
support
for
hms
v1
already
because
of
the
pr
that
david
had
opened
up.
That
doesn't
like
the
extra
caching
and
speeds
up
q
proxies,
and
so
this
is
just
making
it
official.
A
Sounds
good
next
announcement
is
code.
Freeze
is
august,
2nd,
that's
two
weeks
away.
It's
coming
up
pretty
soon.
So
if
there's
any,
you
know
pr's
that
we
want
to
get
in
for
any
enhancements.
I
think
I
know
ravi's
still
tracking
some
for
the
os
fields
start
letting
us
know
if
we
need
to
prioritize
reviews
finding
our
viewers
from
other
sigs
and
there's
other
peers.
You
want
to
get
in.
Let's
just
try
and
do
that.
Does
anybody
else
have
anything
to
announce.
A
Yeah
I've
been,
I
saw
some
activity
on
that
pr
recently
I
know
tim
hawkin
was
asking
if
there's
anything
else
from
him,
and
I
saw
arvin-
responded,
we'll
try
and
get
I'll
take
a
look
at
that
today
and
then
we
can
maybe
deep
dive
next
week
to
see
where
we
are
with
all
the
enhancements
that
are
tracking.
A
Yeah,
okay,
next
is:
we
can
welcome
some
new
contributors.
If
there's
anybody
on
the
call
who
is
new
and
wants
to
introduce
themselves
or
ask
some
questions,
feel
free
to
do
so
and
yeah
no
applications,
but
if
you're,
if
you'd
like
to
do
that,
feel
free.
C
Hi
everyone,
my
name,
is.
B
A
Okay,
I
think
we
can
what
what
I
heard
was
that
was
that
was
funny
he's
a
new
pm
at
microsoft
and
he
was
joining
if
he
audio
issues
get
resolved.
We
can
hand
it
back
over
to
him
too,
but
now
he
can
get
started
on
the
agenda.
James,
do
you
want
to
demo
your
cube
ctl
plugin.
C
Sure
screen
sharing
all
right.
Let
me
see
if
I
can
do
this
pretty
quick.
Let's
look
at
a
few
other
things
on
the
agenda
too.
B
A
B
Cool
awesome,
so,
with
with
the
host
process
container
feature,
I
wanted
to
make
it
easier
for
us
to
be
able
to
get
onto
a
windows
node
for
debugging,
I'm
often
sshing
through
the
master
node
into
a
windows
node,
which
requires
some
fancy,
ssh
stuff
in
particular
like
on
aks,
since
we
don't
have
the
control
plane
nodes
available
to
us,
I
often
have
to
do
a
bunch
of
port
forwarding
fanciness
to
kind
of
eventually
get
to
the
node
and
so
with
host
process
containers,
since
it
gives
this
access
to
the
node,
I
wrote
a
little
script
that
that
did
this
for
me
and
then
I
turned
it
into
a
crew
plug-in.
B
So
this
is
the
the
website,
so
you
can
go
see
the
code
and
everything
that
runs,
but
I
wanted
to
just
give
you
a
quick
demo
of
what
this
actually
looks
like
so
on.
On
the
right
hand,
side,
can
everybody
see
this?
This
is
too
small,
or
should
I.
B
B
Looks
okay
for
me,
okay,
so
here
I
have
a
windows
node,
this
capsi
comp
and
then
over
here
I've
got
crew
installed.
So
if
you
don't
aren't
familiar
with
crew,
crew
is
kind
of
the
cube,
ctl
plug-in
location.
Where
you
can
go.
B
Let's
see,
I
think
I
have
a
link
here.
It's
a
place
where
you
can
get
all
the
all
these
different
plugins
that
plug
into
cubectl,
and
so
I've
got
the
windows
one
that
I've
added
down
here
and
hopefully,
oh
it
just
brings
us
to
the
url
I
gave
before,
but
there's
a
whole
bunch
out
there,
they're
really
cool,
there's
a
few
that
I
use
pretty
regularly.
B
B
Install
typically,
you
just
say
like
windows
debugging,
it
will
install,
but
I
found
it
bug,
and
so
I'm
just
fixing
it,
and
so
I
and
it's
being
published
to
the
manifest
right
now.
So
it's
not
quite
there
yet.
So
I'm
just
going
to
install
the
latest
one.
So
I
can
show
you
a
couple
features
once
that's
done.
It's
it
gets
installed
locally.
B
B
B
So
this
is
my
container
root
folder.
This
will
get
cleaned
up
in
the
new
version
in
1.7
that
mark
and
folks
are
working
on.
But
in
here
I've
created
a
folder
for
apps
and
I've
got
vim
in
here,
so
I
can
go
cd
into
there
and
I've
added
them
to
the
path
and
in
the
current
version
continuity.
This
doesn't
quite
work
right,
but
it
will
be
on
my
path
and
so
then
now
I
can
drop
in
and.
B
Kind
of
right
for
my
windows
thing,
so
I
like
I've,
added
some
extra
tools
to
this,
to
make
it
interesting
and
usable
since
vim
isn't
something
I'm
not
gonna,
necessarily
probably
install
on
my
host
machine.
I
bring
it
along
with
the
container
and
all
of
a
sudden.
I
can
edit
all
the
files
kind
of
quickly
and
tweak
tweak
configurations
and
that
type
of
thing
so
yeah.
You
can
also.
B
Also
specify
images-
and
so
I
can
specify
mark-
has
this
the
scratch
image
that
makes
it
doesn't
have
anything
in
it,
so
it
boots
really
fast.
So
that's
one
option
I
have.
I
can
also.
B
B
This
I
have
another
image
that
I
have
some
other
tools
installed
on
it
like
sysinternals,
and
so
now
I
can
come
in
here
and
I've
got
all
the
assistant
journals
running
and
I
can
just
you
know,
do
whatever
I
need
to
do
here.
I
think
it's
except.
B
Oh,
it's
going
to
hang
because
I
didn't
accept
it,
but
I've
heard
the
exact
syntax,
but
I
can
actually
accept
the
euler
here
and
and
use
like
assist
internals
pretty
quickly
on
the
on
the
machine.
So
that's
that's
everything
I
wanted
to
just
quickly
show
if,
if
you're
interested
it's
up
here,
you
can
give
it
a
try.
Hopefully
pretty
soon.
This
latest
release
will
be
in
the
crew
index
and
you
can
just
pull
it
and
not
have
to
use
the.
B
Yeah
aks
has,
and
in
123
we
have
host
process
support,
so
you
can
use
it
right
on
acas
or
any
cluster
that
has
host
process
enabled
so
123,
plus
and
122.
If
you
enable
the
feature
gate,
you
also
have
to
be
running
container
de
1.6,
plus.
C
A
question
since
I
actually
needed
something
like
this
today:
is
there
any
way
to
also
copy
files
from
you
know,
broccoli
to
windows,
notice,.
B
B
C
A
A
C
Like
to,
I
guess,
go
for
the
process
to
upload
this
as
a
crew
plug-in.
Was
that
hard,
like.
B
Yeah,
I'm
not
doing
that.
Yes,
can
you
see
my
screen
again
with
the
debug,
so
I'm
using
this
crew
release
spot
component
to
do
that,
and
so
I've
got
a
couple
workflows
that
enable
this.
B
The
my
plugin
is
really
simple:
it's
not
even
a
go
plugin
right
now,
it's
not
even
a
go
tool,
it's
just
a
bash
script,
and
so
I
just
tar
up
the
packages
and
then
release
it
to
github
actions,
and
then
I
call
this
here
and
it
and
this
actually
automates
pushing
this
out
to
the
crew
plug-in.
For
me,
it
took
me
a
little
bit
figuring
out
like
how
this
works.
B
I'm
actually
going
to
do
a
cncf
workshop
on
on
this
plugin
in
a
few
months,
and
I'm
going
to
talk
through
how
to
do
the
whole
crew
thing,
but
so
this
is
the
plug-in.
This
is
the
actual
pr
that
I
was
mentioning
earlier.
Oh,
oh,
hey
he
merged
it,
so
I
didn't
even
have
to
do
it.
It's
supposed
to
automate
the
whole
merging
thing.
This
I'm
not
sure
why
I
didn't
automate
it,
but
he
gave
me
the
approval,
so
you
can
actually
go
use.
B
The
version
that
I
just
showed
today
by
just
doing
cubes
k,
crew
install
but
yeah
once
you
have
it
automated.
It's
pretty
easy
to
the
the
rest
of
the
automation.
All
works,
pretty
simple,
so.
B
A
So
yeah
this
is
something
that
we've
been
that
I
was
experimenting
with.
I
was
seeing
if
we
could
make
a
scratch
image
for
host
process
containers
that
didn't
have
any
that
didn't
really
have
any
files
in
it,
and
the
answer
is
yes,
we
can
it
works
for
host
process
containers.
I
think
we'll
have
more
information
on
that
in
the
in
the
future.
What's
the
size
of
the
image
just
is
it
literally
the
size
of
the
scratch
image.
C
A
Maybe
two
or
three
kilobytes:
oh,
that's,
awesome
all
right
yeah!
It
only
works
with
those
process
containers
and
especially
when
brendan
gets
back.
We'll
have
more
to
share
about
that.
That's.
C
B
B
Okay,
so
we'll
move
on
to
chi
stats,
so
christ,
hats
is
a
kept,
that's
in
alpha
right
now
and
it
was
implemented
for
linux
and
basically,
what
they're
trying
to
do
is
move
everything
from
they're
trying
to
move,
remove
all
the
c
advisor
information
from
cubit
and
there's
a
bunch
of
other
collection
stuff
in
cubelet
and
push
that
down
to
the
runtimes,
and
so
this
is
being
implemented
in
container
d.
I
think
cryo
also
implements
it
and
we
we're
missing
windows
implementation.
So
we
to
do
this.
B
B
B
We
don't
actually
use
a
bunch
of
those
things
for
windows,
and
so
I
went
through
and
specified
everything
that
we
do
and
do
not
use
and
what's
the
val
equivalent
and
then
also
we're
any,
is
working
on
hyper-v
support,
and
so
we
wanted
to
make
sure
that
these
pod
stats
would
work
for
both
hyper-v
and
process
isolated
in
continuity,
and
so
he
so
I
went
through
and
said
like
which
ones
we're
using
and
which
ones
are
we
aren't,
and
so
I
think
we
had
a
little
bit
of
a
conversation
on
here
around
like
how
to
like.
B
B
We
do
have
this
opportunity
to
kind
of
design
this
and
have
a
windows
first
experience
here,
instead
of
just
going
to
the
linux,
and
so
I
think
we've
had
a
little
bit
of
conversation
back
and
forth,
and
I
think
the
last
few
things
is
just:
how
do
we
do
we?
I
think
we
decided
that,
essentially,
we
would
like
to
make
windows
specific
fields.
Do
we
reuse
some
of
the
ones
that
are
not
windows,
specific,
that
we
can
fill
out
completely?
B
How
window-specific
it
is?
Do
we
change
all
the
container
stack
calls
to
use
all
these
new
windows
fields
that
type
of
thing,
so
please
take
a
look
at
this
and
give
some
feedback
I'm
going
to
bring
it
to
sig
node,
actually
right
after
this
meeting
and
get
some
of
their
feedback.
Just
I
want
to
get
some
someone,
someone
from
their
side.
A
Yeah-
and
I
think
I
had
a
comment-
I
think,
where
we're
settling
then
some
folks
from
node
are
have
commented
on.
This
too,
is
like
these
cri
based
stats
are
not
meant
to
be
like
completely
exhaustive
they're,
not
meant
to
be
everything
that
you
would
need
to
monitor
your
system,
but
they
are
enough
to
kind
of
provide
the
equivalent
of
what
the
cubelet
is
exposing
through
c
advisor
today,
just
not
using
the
c
advisor
endpoints.
A
So
I
think
the
idea
is
to
to
still
expose
those
to
the
cubelet,
but
then,
if
you
want
really
in-depth
monitoring
stats
and
potentially
things
like
knowing
that
there
are
different
stats
for
different,
you
know
hyper
or
like
container
modes
on
windows
with
container
d.
Those
are
probably
best
to
be
exposed
through
the
container
runtime
itself
and
then
have
something
like
node
export
or
like
something
scan
scan
those.
So
this
is
not
meant
to
be
completely
exhaustive
and
completely
transparent
to
the.
A
B
A
I'll
share
my
screen
again.
I
think
I
had
the
next
update,
so
I
wanted
to
give
a
quick
update
on
host
process
containers
and
where
we
are
a
few
weeks
ago,
we
mentioned
that
we
are
going
to
be
delaying
the
ga
of
this
until
hopefully,
well
probably
we're
delaying
it.
So
we're
not
getting
the
feature
in
125,
hopefully
we'll
be
able
to
ga
the
feature
in
126.
A
And
now
we
think
that
we
have.
We
have
that
the
current
there
were
a
number
of
changes
that
went
into
hcshim
recently
that
will
be
getting
vendored
into
the
new
or
container
d
1.7
for
release
that
will
should
bridge
that
gap,
and
we
can
demo
that
at
another
time,
maybe
after
code
freeze
since
some
of
that's
more
urgent,
but
for
of
several
at
least
several
releases
of
continuity
and
kubernetes.
A
What
we'll
do
is
we'll
mirror
the
volume
we'll
mirror
the
any
volume
mounts
that
go
into
the
container,
both
in
the
new,
more
desirable
location,
which
can
be
accessed
just
through
absolute
paths.
Instead
of
needing
to
use
this
environment
that
that
container,
sandbox
one
point
environment
variable
and
we'll
mirror
them
into
the
old
location
and
build
up
the
same
directory
structure
that
you'll
see
where
you'll
yeah,
where,
where
and
where
you'll
be
able
to
use
that
container
stamp
excellent
point
to
read
them
off
of
that.
A
I'm
currently
in
the
process
of
ensuring
that
all
of
the
existing
ed
tests
that
we
have
pass
with
the
newer
behaviors
there's
a
couple
of
differences
that
I'm
working
through
with
with
danny
about
that.
But
I
think
overall
we're
in
pretty
good
shape
too.
Both
support
the
new
volume
out
strategies,
the
way
we
wanted
it
to
even
on
windows,
server,
2019
and
also
be
able
to
have
a
smooth
transition
for
folks
who
are
upgrading
their
clusters
or
updating
their
container
runtime
without
needing
to
rebuild
and
all
of
their
their
workloads.
D
Yeah
sure,
let
me
share
my.
Let
me
stop.
C
A
C
Okay,
cool,
okay,
so
otherwise.
C
D
D
Oh
here
we
have
10
test
cases
here
and
we
also
in
our
operators.
We
match
each
of
the
test
cases
with
the
upstream
e2e
test.
So
if
you
use
this
binary,
if
you
download
this
binary
and
run
the
find
and
build
it
right
like
this,
so
you
define
the
categories,
the
core
network
and
specify
your
cluster.
You
will
start
all
the
networking
tests
of
the
windows
cluster,
something
like
this.
It
will
like
take
20
minutes
to
finish
the
test.
I
won't
let
oh,
we
don't
need
to
wait
for
it
right
now.
D
D
So
when
you
start
it
will
like
run,
there's
ten
test
cases,
and
currently
we
are
running
the
first
one,
and
here
is
the
case
we're
running
and
you
see
if
we
looking
for
this
one,
so
each
of
the
test
cases
will
be
run
one
by
one
here
and
it
it.
Actually,
you
can
see
it
passed
here.
So
if
I,
if
we
go
to
the
second
one,
so
if
you
go
to
each
of
there,
you
can
find
that
it
most
of
the
mars
success.
D
There's
only
one
test
case
about
the
staphon
set,
which
is
not
success
at
this
time,
but
I'll
fix
that
later,
but
you
can
see
that
most
of
them
are
success.
So,
with
this
banner,
you
can
run
the
networking
test
of
this
in
just
one
command,
and
that
is
what
I
get
here,
oh
by
the
way.
D
One
more
thing
is
that
there's
a
test
case
about
the
the
this
one
restarting
the
service,
so
I
I
have
a
pr
open
for
this
and
it
haven't
been
merged
yet
so
the
result
you
see
here
the
e2e
test
binary,
I'm
using,
is
beautiful.
My
own
branch,
so
after
the
task
is
merged,
you
can
use
the
upstream
test
upstream
into.
It,
has
binary
to
run
this
to
run
this
apparently
and
yeah.
C
Yeah,
it
would
be
cool
if
we
could
eventually
be
able
to
upload
these
someday
to
kate's
conformance,
but
I
don't
know
if
they'll
they'll,
let
us,
but
I
filed
an
issue
there.
Nobody
responded
on
it
yet,
but
for
broader
context
like
there's,
always
for
us
a
lot
of
confusion
about
what
a
windows
cluster
is
from
a
conformance
standpoint,
because
the
sonoboe
plug-in
has
a
declarative
way
of
specifying
linux
conformance,
but
not
windows,
conformance
and
so
we're
helping
to
once
shinji
finishes.
C
This
tooling
actually
be
able
to
have
a
sona
buoy
plug-in
that,
like
then
triggers
this,
so
that
we
can
have
some
kind
of
a
declarative
one-liner,
for
you
know
any
customers
or
whatever.
When
they
have
problems,
they
can
just
run.
That
and
they'll
have
a
very
clear
readout
of
exactly
what
does
and
doesn't
work
in
their
in
their
windows.
Environments.
D
Yeah,
actually,
this
one,
this
operational,
redness
tool
can
trigger
the
assembly
test.
I
just.