►
From YouTube: Kubernetes SIG Windows 20210330
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
march
35th
2021
iteration
of
the
kubernetes
sig
windows
community
meeting.
As
always,
these
meetings
are
recorded
and
uploaded
to
youtube.
So
please
be
sure
to
adhere
to
the
cncf
code
of
conduct
and
let
me
share
my
screen
and
then
we
can
kind
of
get
started.
A
All
right,
I
see
a
couple.
People
are
still
adding
some
things
to
the
notes,
but
we
can
get
started
here.
Let
me
pull
up
the
chat
too,
so
I
don't
miss
anything
all
right.
First,
we
can
see
if
there's
any
new
folks
on
the
call
who
would
like
to
kind
of
just
give
an
introduction
or
anybody
who
wants
to
just
kind
of
say,
hello,
and
I
see
a
couple
of
people
who
look
new
on
the
call
here.
A
But
if
not,
we
can
kind
of
we
can
get
started.
First
thing
is
announcements.
I
don't
see
any
new
announcements.
I
know
we're
kind
of
rapidly
approaching
the
end
of
the
121
release.
I
believe
tomorrow
is
the
deadline
for
having
reviewed
and
open
doc
prs
set
up.
Those
don't
necessarily
need
to
merge
tomorrow,
because
the
docs
review
team
likes
to
take
a
look
at
them,
but
they
do
ask
that
they're
that
all
of
the
doc
peers
are
reviewed
by
the
subject
matter,
experts.
A
A
Yeah
so
jay
mentioned
mike
k
and
mauricio
might
be
new.
I
believe
so.
A
But
does.
B
C
I
don't
I
don't
really
have
any
I'm
just
kind
of
clearing
my
cash
from
some
other
stuff
so
but
as
always,
we're
going
to
be
pairing
after
this
today
and
I
think
we
might
maybe
go
into
the
kublet
and
some
other
sort
of
interesting
stuff.
So,
as
always,
everybody
feel
free
to
stay
on
afterwards
and
do
do
the
sig
windows
after
dark.
A
Sounds
good,
let's
get
into
some
of
these
issues.
I
believe
jing
added
the
first
issue
about
this
container
tear
down
issue.
I
think
that
a
couple
of
other
folks
have
been
looking
at
it.
Unless
this
is
a
different
issue.
Jing
are
you
on?
I
actually
don't
see
jing
on
the
call.
A
C
C
C
I
am
because
so
I'm
I've
talked
some
to
james
about
this,
but
we
are
going
to
try
to
really
redo
the
entire
code
proxy,
with
this
new
tool
called
kaping,
and
we
have
essentially
a
coupe
proxy
working
group
that
meets
up
every
friday
just
fyi,
and
I
think
it's
gonna
be
really
good
for
the
for
the
windows
side
of
things,
because
it's
it's
gonna,
be
a
lot
more
easy
to
understand
and
it's
gonna
completely
decouple
the
code
bases
from
different
group
proxy
implementations.
C
So
I
I'd
like
to
take
a
pulse
after
we
talk
about
this
about
whether
anybody
cares
about
us
being
on
the
bleeding
edge
of
that
anyway.
So
this
is
something
I've
been
playing
around
with
I've
been
talking
to
perry
about,
but
building
images
is
real
important
for
windows,
since
you
can't
distribute
it,
and
so
I
feel
like
it's
kind
of
a
windows
related
thing.
C
It's
too
bad
james
isn't
here
because
he
might
have
some
strong
opinions
on
this,
but
I
we
were
just
kind
of
talking
about
like
what
would
be
the
best
way
to
really
serve
up
coup,
proxy
and
kubla
and
all
the
other
artifacts
downstream.
We
have
this
concept
of
like
this.
C
We
call
it
the
burrito
right
this
idea
that
you
have
like
all
this
stuff
and
you
need
to
put
it
into
an
image,
and
you
know
we're
kind
of
solution,
solutioneering,
stuff,
downstream
and
kind
of
internally
bike
shedding
a
lot
about
like
what's
the
right
way
to
do
it
because
there's
a
thousand
ways
to
do
this
and
we
haven't
really
come
up
with
a
standard
way
of
doing
it.
Yet
so
I've
kind
of
tried
to
upstream
this
question
and
see
whether
anybody
might
be
interested
in
image
builder,
possibly
being
aware
of
things.
E
C
E
You
know,
if
you
don't
have
an
internet
connection,
if
you
don't
have
access
to
the
web,
how
do
you
build
a
windows
image
if
you
are
a
person
that
you
know
has
got
to
build
a
windows
image
because
you're
not
allowed
to
be
supplied
that
by
a
another
vendor
or
whatever?
And
it's
trying
to
understand
them?
E
Whether
we
need
to
find
a
mechanism
to
be
able
to
pre-load
that
or
whether
we
need
to
build
and
copy
in
or
whether
we
need
to
put
something
into
image
builder
that
allows
us
to
load
the
artifacts
kind
of
by
by
the
nature
of
just
basically
re-rooting
the
h
like
the
url
or
the
ansible
uses
to
pick
it
up
from
localhost
or
whatever,
but
yeah
it's
trying
to
find
a
way
around.
The
lack
of
web
is
quite
tricky.
A
Yeah
I
know
for
for
aks,
we
do
do
kind
of
air
gap
scenarios
too,
and
we,
while
we're
not
using
cluster
api,
we
do
have
kind
of
a
golden
image
concept
and
where
we
do
ensure
that
for
different
releases,
all
of
the
artifacts
that
would
be
needed,
at
least
for
the
standard
cluster
bring
ups,
are
kind
of
included
in
that
image.
I
see
a
couple
people
on
the
call
who
are
pretty
active
in
sync
cluster
lifecycle
or
cluster
api
too,
like
david
and
lubamir
and
others.
A
I
think
that
this
is
probably
going
to
overlap
quite
a
bit
with
kind
of
those
projects
as
well.
So
I
don't
know
if
anybody
wants
to
comment
or
if
we
should
just
kind
of
think
about
this
and
wait
for
james
to
come
back
everything
but
yeah.
I
do
see
this
kind
of
being
an
important
issue
in
the
future.
E
A
C
Yeah
I
mean
so
that's
the
thing:
we've
been
we've
been
using
the
tool
internally
called
image
package,
which
kind
of
tries
to
put
a
little
bit
of
a
facade
around
the
idea
that
you
have
you
know,
maybe
20
or
30
different
docker
images
or
files
or
something,
and
then
it
gives
you
a
way
to
kind
of
package,
those
in
a
container
with
a
unique
sort
of
aggregate,
signature
and
stuff.
So
it's
it
solves
some
of
the
solves
some
of
the
easy
problems,
but
it
doesn't.
C
You
know
it
doesn't
give
an
end-to-end
solution
to
us
for
the
problem
of
a
actual,
like
some
kind
of
way
to
just
spin
a
whole
step
that
serves
the
artifacts,
and
so
you
get
into
this
weird
intermediate
state
where
you're
like
well,
I
could
give
people
a
shell
script
or
a
python
script.
That
did
this
in
like
four
lines
of
code,
and
then
you
have
this
weird
bike,
shedding
problem
that
happens
afterwards,
because
everybody
has
their
favorite
web
server,
and
so
it's
so
so
you
it
winds
up.
It's
not
a
tech
problem.
C
D
C
C
A
Which
may
be
okay
yeah,
but
yeah
yeah?
Let's,
let's
kind
of
revisit
this
in
a
little
bit,
I
do
see
that
jing
is
on
the
call
now
jing
do
you
want
to
talk
about
the
issue?
I
saw
you
adding
earlier
about
the
container
teardown
issue.
B
We
have
seen
the
custom
issues
when
the
container
during
the
container
tear
down
right.
Basically,
we
delete
the
container
and
it
failed.
Has
this
error
message
saying
failed
to
remove
that
system?
The
arrow
would
be
like
create
file
some
darker
file
like
a
path
and
accesses
the
night
arrow.
B
So
this
kind
of
arrow
is
almost
the
same
as
the
one.
Let's
say
I
discovered
for
let's
say
when
you
run
os
that
if
it
is
a
singling
to
a
directory,
it
will
give
you
error
message
showing
create
file.
Create
file
is
very
misleading.
A
Is
there
an
issue
or
anything
that
we
can
add
more
details?
I
can
add
a
little
bit
of
maybe
flavor
to
this.
I
know
that
in
windows,
one
of
the
most
common
ways
to
get
a
file
handle
is
to
just
issue
a
create
file
call
and
with
whatever
permissions
that
you
what
you
want
on
it.
So
that's
probably
why
so
many
things
like
the
os
stat
call
and
the
and
potentially
this
are
having
errors
with
the
actual,
create
file.
F
I
I
created
a
issue
under
the
windows
container
hi,
I'm
julie,
I'm
from
google.
G
F
Hi
mark,
let
me
explain
a
little
bit.
It
seems
that
the
the
file
seems
to
be
the
docker's
container
log
file.
It
seems
that
the
docker
try
to
remove
the
container,
and
you
know
it
comes
out
this
of
access
denial
issue
and
it
fails
to
remove
the
container
so
the
port
start
in
the
termination-
and
I
saw
the
similar
issue-
was
reported
in
aks
before
and
it
seems
that
they
submitted
some
pr
and
the
pr
was
fixed.
F
It
was
available
in
117
12,
but
the
actually
the
customer
saw
this
issue
repro
in
117.15.
So
I'm
wondering
if
you
know,
I'm
also
wondering
if
there's
some
besides
the
kubernetes
pr
side
change.
If
there
is
some
also
some
change
applied
to
the
os
layer
or
the
aks
one.
I
I
also
attached
the
link
of
the
ats
one
in
the
bug
I
opened.
F
I
can
I
edit
the
beam.
F
B
So
someone
on
the
chat
mentioned
a
dw
share
mode
problem.
I
I
don't.
I
don't
know
what
is
dw
share
mode.
H
So
basically,
if
somebody's
walked
the
file
for
right
access,
you
have
to
be
careful.
What
how
how
you
execute
your
current,
create
call
create
file
call,
and
I
can
link
you
to
the
docs
for
that,
but
you
basically
have
to
debug
whether
something
has
already
obtained
access.
A
A
Yeah,
let
me
see
if
anybody
on
the
call
would
be
able
to
take
a
look
at
that
this.
Let
me
try
and
follow
up
with
some
people
inside
of
microsoft,
who
are
closer
to
the
the
docker
kind
of
layer
or
the
hcs
layer
and
see,
if
there's
anything,
that
we
can
kind
of
do
to
help
with
this
investigation.
H
Also,
what
it
was
access
is
denied
error.
Did
you
obtain
that,
with
the
get
last
error
function.
F
It
is
full
on
the
node
log
because
we
we've
reported
in
the
gke
windows,
so
it's
like.
We
read
it
from
the
cloud
logging,
so
it's
the
it
is
some.
I
think
it
is
produced
by
the
docker
daemon
and
it
is
ported
to
our
login
yeah.
H
I
F
B
Also
in
the
kubernetes,
so
kubernetes
will
will
carry
this
ironman
because
it
failed
to
delete
the
container
and
so
from
dr
daemon
error
message.
It's
also
logged
in
like
no
kubernetes
side.
A
Would
it
be
possible
to
to
attach
a
larger
section
of
the
cuba
log
to
this
issue
to
see
what
other
calls
may
be
happening
around
the
same
time
or
beforehand
leading
up
to
the
situation
that
might
help
figure
out?
If
maybe
something
I
know,
we've
seen
issues
in
the
past,
where,
when
containers
start
and
then
immediately
fail
like
there
were
some
issues
either.
F
Yeah
I
see
if
customers
delete
this
this
log,
you
can
see
the
error
message.
It
mentions
some
log,
it
is
the
docker
log
I
think
docker
container
lock.
If
you
manually,
delete
that
lock,
the
the
then
the
clean,
the
container
can
be
removed
and
the
product
can
be
terminated.
F
A
C
Phone
question:
this
is
just
a
stupid
question.
You
know
hayden's
here.
He
may
have
some
thoughts
on
this,
so
I
actually
couldn't
mount
a
windows
iso
today
in
in
a
vmware
workstation,
and
I
couldn't
figure
out
why-
and
I
was
just
wondering:
if
has
anybody
tried
to
develop
against
windows,
server
2019
in
vmware
workstation
and
run
into
an
issue
with
packer?
C
E
C
Well,
it's
the
generic
question
of
like
how
the
hell
is
it
even
possible
that
the
windows
installer
is
booted?
Is
it
is
there
something
that
I
don't
know
about
how
windows
is
embedded
into
the
bios
of
of
of
hypervisors
or
something
like
this,
because
how
is
it
that
you
could
even
have
a
thing
that
had
a
windows
dialogue
that
was
trying
to
install
the
operating
system?
If
the
iso
wasn't
visible.
A
A
That
is
wim,
is
kind
of
an
internal
windows
image
format,
and
that
includes
all
of
kind
of
the
bits
that
get
laid
out
to
disk.
When
you
do
a
windows
install
that
file
may,
when
I
was
checking
before
was
pretty,
I
was
pretty
close
to
approaching
the
maximum
file
size
that
the
fat32
file
system
could
read.
A
Oh,
it
may
be
possible
that
that
file,
due
to
you
know,
ongoing,
continued
servicing
has
gotten
so
large
that
so
windows
kind
of
operates
by
it
boots
into
this
windows,
kind
of
pre-install
environment
or
win
pe
and
that
they're.
I
know
that
there's
been
issues
kind
of
you
know
accessing
files
larger
than
the
four
gigabyte
limit
that
is
set
for
the
file
system.
So
there
may
be
an
issue
expanding
that
install
whim
file
on
there.
J
E
E
I
put
them
into
the
unattended
and
then
loaded
them
in
the
unattended
in
the
pre-boot
I
fit
in
the
in
the
specialized
or
something
I've
tried.
It's
been
about
six
months
since
I
did
this
so
really.
J
It
and
it
depends
on
your
virtualization,
I'm
more
experienced
with
kvm
and
others,
but
in
the
past
I've
actually
had
to
rebuild
the
iso
with
drivers
to
get
past
windows
pe
into
you
know
to
access
drives
and
stuff.
I.
A
J
A
There
may
be
some
kind
of
extra
packages
that
you
can
install
too.
I
think
virtualbox
has
this,
where
you
can
sell
these
extra
packages
and
it'll
detect
if
it's
trying
to
boot
up
windows
and
kind
of
also
set
inject
those
drivers,
any
drivers
that
are
needed
into
that.
E
Yeah,
I
can
see
the
para
and
j's
screen
there.
I
can
see
the
pvc
drivers
are
loaded
as
part
of
the
floppy
drive.
I
I
can
look
at
it
tomorrow
with
you.
Jay
there
might
be
a
it
might
be.
Something
between
fusion
and
workstation
is
slightly
different
at
the
way
that
it
loads
these
drivers
and
or
the
default
drive
that
it
uses
or
something.
C
Yeah
I
mean
it
just
getting
this
working
on
ubuntu
was
tricky
because
I
actually
actually
this
is
ubuntu
focal
faucet
2020
and
I
actually
had
to
install
the
vm
monitor
and
another
kernel
module
myself
manually
just
to
get
the
boot
up.
So
I
suspect
it's
something
happening
between
the
hypervisor
and
the
operating
system
and
because
I've
already
had
to
do
one
weird
thing
so
so
to
uplevel
it
perry.
Like
you
know.
The
reason
I
was
doing
this
was,
I
kind
of
was
trying
to
go
through
the.
C
What
if
I
didn't
have
vsphere
what,
if
I
didn't
have
forget
about
these
here,
like
what,
if
I
didn't,
have
a
external
hypervisor
to
run
image
builder
against,
and
I
just
wanted
to
build
the
ova
offline.
What's
the
experience
going
to
be
like,
and
so
that's
what
I
ran
into
so.
Has
anybody
ever
done
this
on
windows
or
mac
or
os
10?
Has
anybody
ever
run
image
builder
on
this
environment,
yeah.
A
E
G
D
G
Question
when
you're
talking
about
image
builder,
are
you
talking
about
a
container
image
builder
or
virtual
machine
image
builder?
We
need.
C
J
C
K
E
C
I
A
I'm
gonna
kind
of
close
the
meeting
out
at
that
and
hand
it
over
to
jay
and
start
on
the
the
pairing
sessions
I'm
going
to
stay
since
the
segway
windows.
Our
signature
meeting
is
closed
too,
but
let
me
okay.
Stop
that
hand
it
over
thanks.
Everybody
for
attending
feel
free
to
stay
and
pair
on
us
with
image
builder
on
windows.
C
K
Hey
jay,
I
just
had
some
questions
about
the
q
proxy
next
generation
cool
cool.
I
was
just
kind
of
curious
like
what
kind
of
timeline
you
see
that
happening
in.
C
K
C
Know
like
the
biggest
thing
now
is
getting
people
to
dig
in
I
mean
if
we
got
one
other
person
to
really
dig
in
and
help
own
a
piece
of
it
ricardo's
owning
ipvs
and
I
hope
to
get
involved
with
the
windows
side.
C
C
That's
there
as
far
as
how
long
it
takes
to
be
dot
all
the
t's
and
cross
all
the
t's
and
dot
all
the
eyes,
and
all
that
stuff
I
mean
that's
up
to
the
cncf
gods
of
you,
know
keps
and
all
that
stuff,
but
I
mean
it's
out
of
tree,
so
it's
like
you
know
we
get
it
going
and
get
it
working
and
people
are
going
to
be
able
to
start
to
use
it
and
you're
going
to
be
confident
using
it
in
releases.
C
I
think
if
we
just
get
one
other
person
to
really
dig
into
it,
so
I
know
that's
kind
of
a
roundabout
question,
but
I
think,
probably
two
months,
if
we
didn't
it
would
probably
be
one
to
two
months
depending
on
how
many
people
we
really
get
digging
into
it.
Now
that
said,
there's
a
lot
of
people
that
came
to
that
first
meeting,
I'm
assuming
some
small
percentage
of
those
people
are
gonna,
do
work
so
like
if,
as
long
as
there's
a
few
people
that
are
really
gonna
dig
in,
I
think
we'll
be
fine.
K
Cool
yeah
just
trying
to
get
an
idea
of
like
how
far
ahead
it
was
no
seems
good,
I'll,
look
more
into
the
project.
C
Yeah,
so
we
have
so
laurie
laurie
apple
and
she
she
helped
me
to
make
a
doc
that
wasn't
so
dysfunctional
about
this
and
that
actually
kind
of
so
let
me
put
a
link
to
that.
D
D
C
Okay
cool,
so
I
put
a
link
in
both,
but
yes,
I'm
asking
if
you
feel
like
you
know,
you
want
to
get
involved
and
maybe
help
own
a
tiny
piece
of
it
or
have
a
friend
that
wants
to
or
whatever
or
if
nothing
else,
just
come
and
hang
out
with
us
on
fridays
I
mean
having
other
people
showing
up
makes.
It
helps
us
to
keep
the
momentum
going
right.
It
gives
us
a
little.
D
A
little
community
fair
fanfare,
and
that
goes
a
long
way.
So
even
if
you
could
just
show
up
every
once
in
a
while
that'll.
C
Yeah,
so
the
coup
proxy
has
been
a
thorn
in
our
side
on
the
windows
side,
for
a
long
time
and
and
everywhere
else,
because
it's
very
hard
to
change
it,
because
if
you
change
something
for
example
and
point
slices,
you
add
those
to
ipvs,
then
you
have
to
add
them
to
ip
tables.
But
then
what
you
didn't
add
it
to
windows.
C
Now
you
have
to
add
it
to
windows
too.
So
so
we
have
a
problem
in
windows
where
we
are
not
really
tracking
the
changes
that
are
happening
in
the
cool
proxy
entry
and
it's
it's
diverging
right.
We
also
have
a
general
understandability
problem
because
there's
almost
no
real,
clear
decoupling
between
the
generic
logic
of
the
coupe
proxy
and
the
way
that
the
individual
coup
proxy
providers
are
implemented.
C
So
the
coup
proxy's
job
is
to
write
routing
rules
between
services
and
pods,
and
it's
a
lot
more
comp
for
something
that
does
something
so
simple.
There's
a
very
there's,
not
a
very
there's,
just
not
a
deep
level
of
sort
of
modularity
of
object
orientation
in
the
code
base,
as
is
right.
So
a
lot
of
code
drift
going
on
it's
making.
It
hard
for
us
to
really
know
what
the
difference
between
windows
and
linux
is
from
a
coupe
proxy
support
perspective.
The
tests
aren't
the
best
I
mean
you
can
pass
conform.
A
Then
yeah
just
one
more
question
just
so:
I'm
kind
of
on
the
same
page
is,
and
it's
fine,
if
this
is
still
kind
of
being
discussed
or
planned,
is
the
plan
to
move
q
proxy,
at
least
for
windows
out
of
tree.
Then
I
I
heard
somebody
say
out
of
tree
or
have
a
second
cue
proxy
that
is
out
of
tree
or
move
the
entirety
of
q
proxy
out
of
tree.
C
A
K
Sounds
like
a
great
idea
moving
it
out
of
case,
slash,
k,
yeah
repo's
too
big.
A
C
Yeah
yeah
and
that's
fine,
you
know
we'll
have
a
crappy
and
good
proxy
entry
and
then
we'll
all
do
all
of
our
work
and
everybody
who's.
Shipping
windows
or
supporting
it
is
gonna.
Have
a
really
good
alternative
and
they're
gonna
know
how
it
works
and
I
think
that's
great
for
the
vendors
that
are
being
gonna
get
involved
in
right.
So
I'm
fine
with
the
crappy
dead
code
entry
as
long
as
it
doesn't
prevent
our
ability
to
build
a
community
around
something
better.