►
From YouTube: Kubernetes SIG Windows 20210928
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
september
28
2021
instance
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
and
conduct
standards.
A
Let's
start
with
some
announcements,
the
first
announcement
is
again
we're
congratulating
claudio
for
his
promotion
to
tech
lead,
and
we
wanted
to
thank
deep
again
for
all
of
his
work
before
this
was
announced
back
in
either
end
of
april
or
may,
and
I
didn't
we
didn't,
have
anybody
kind
of
have
any
like
we
didn't
have
any
reason
to
not
move
forward
with
the
with
this
promotion,
except
that
I
went
on
parental
leave
very
shortly
after
and
this
kind
of
just
got
dropped.
A
So
we're
finishing
this
up
and
I
believe
all
the
prs
to
swap
deep
for
claudio
are
kind
of
in
place.
So
congratulations
again
claudio
and
I
don't
see
deep
on
the
call
deep's
been
unable
to
join
sig
windows
meetings
for
a
while.
But
I
do
want
to
thank
him
again
for
all
of
his
work.
He
did
oh
incredible
amount
of
work
kind
of
early
on
and
with
the
windows,
bring
up
and
then
continued
with
a
lot
of
windows.
A
Storage,
related
work
with
driving,
csi
proxy
to
stable
for
windows,
and
all
of
that,
so
thank
you
claudio
congratulations.
The
next
announcement
is
docker.
2010-8
has
some
fixes
for
windows
containers
that
people
are
probably
interested
in
consuming.
There
was
an
issue
where
pot
or
container
deletion
would
sometimes
fail
if
the
hcs
shim
returned
in
a
kind
of
an
error
code
that
wasn't
handled
and
containers
would
be
left
around
and
stuck
on
the
machine.
A
There's
more
details
in
the
debug
there,
but
if
anybody's
using
docker
or
mobi
highly
recommend
updating
to
version
to
the
version
a
newer
version,
I
don't
think
that
docker
ee
is
out
yet,
but
I
think
we're
expecting
that
package
anytime
from
mirantis.
A
All
right
first
agenda
item
I
wanted
to
ask:
is
anybody
planning
on
attending
in
person
for
kubecon
north
america.
A
I
don't
think
any
of
the
leads
are.
I
was
just
curious
if
there's
going
to
be
any
representation
from
sing
windows
or
not,
there
are
a
couple
of
in-person
events
like
a
sig
meet
and
greet
that
are
scheduled.
Brenda
did
you
raise
your
hand.
A
Okay,
if
you
guys
are
interested,
we
can
maybe
have
some
of
you
attend
on
like
some
of
the
sig
meet
and
greets
on
behalf
of
sig
windows.
I
will
follow
up
with
you
guys
offline
and
if
anybody
else
is
interested
in
in
any
of
that,
please
just
reach
out
to
myself
james
claudio
j.
Any
of
the
leads
here
and
also,
I
will
probably
probably
be
canceling
the
community
meeting
on
during
that
week
to
give
folks
a
chance
to
attend
and
any
talks
or
kind
of
virtual
events
or
personal
events
that
they'd
like.
C
We
also
have
at
least
one
recording
that
will
be
played
during
the
week
kind
of
showing
off
post-process
containers,
as
well
as
the
dev
environments
that
everybody's
been
working
on.
So
I
think
there
might
be
one
other
one
if
you
yeah.
A
Let
us
know,
I
think
there
was
this
one
related
to
service
mesh.
I
will
we'll
make
sure
to
get
those
on
the
agenda
next
week
and
share
the
times
for
anybody
who's
interested.
A
Yes,
thank
you
for
highlighting
that
james
all
right.
Jay
next
agenda
item
is
yours.
Do
you
want
to
cover
jay
on
the
call.
A
Yeah,
okay,
next
one
is
one
that
I
thought
was
interesting.
A
A
So
there's
proposed
changes
to
how
I
think
claudio
found
a
library.
This
library
go
acl
that
allows
you
to
set
file
permissions
for
windows
in
a
manner
similar
to
how
they're
kind
of
specified
and
expressed
in
linux
and
in
a
lot
of
the
golang
libraries.
A
So
there's
some
discussion
here
and
I
think
that
we
determined
that
the
library
was
healthy
in
terms
of
kind
of
an
open
source
project
and
being
able
to
take
a
dependency
on
it.
The
next
question
was:
is
this
something
that
we
should
should
kind
of
move
forward
with,
and
it
doesn't
seem
like
there's
any
kind
of
red
flags
here
or
reasons
why
we
shouldn't
cardio?
Do
you
want
to
comment
on
this.
D
Yeah,
I've
known
this
library
so
for
some
time
ago,
I
tried
to
do
something
similar,
but
ultimately
I
found
this
library
instead,
which
basically
does
the
same
thing.
D
I
had
the
idea
of
mapping
the
linux
style
permissions
to
windows
for
some
time,
and
apparently
we
can
do
it
to
some
extent
due
to
windows,
windows,
api
and
that's.
What's
actually
what
this
library
is
doing.
It's
setting
those
generic
read,
write
and
execute
permissions
for
the
file
owner
file
owner
group,
and
I
think
others
are
something
like
that
which
pretty
much
correlates
to
how
the
links
permissions
are
set
up
to
be.
D
This
will
also
allow
us
to
enable
a
couple
of
features
for
windows
as
well,
and
if
you
check
the
pr
we're
actually
enabling
a
couple
of
e3
tests
for
windows
as
well,
which
should
relates
to
file
permissions.
D
There
are
still
a
couple
of
them
which
are
not
yet
enabled,
but
that's
mostly
because
it
requires
to
run
as
non-root,
but
that's
something
that
can
also
be
addressed
by
running
us.
You
know
as
a
container
user
instead
of
container
administrator
for
windows,
so
the
list
of
tests
that
we
support
from
conformance
extents.
D
There
are
still
a
couple
of
things
that
I
was
still
having
issues
with
and
I
kind
of
forgot
exactly
which
they
were,
but
they
were.
Regarding
that
the
which
one
was
it
there
were
a
couple
of
empty
deer
scenarios
which
were
a
bit
more
interesting
and
are
still
failing
for
some
reason.
You
still
have
to
look
into
that.
D
Yeah,
but
there
are
two
different
tests
which
are
still
failing.
I
don't
think
I've
noted
them
here.
I
don't
think
so.
I
should
have
noted
them
yeah.
I
should
have
included
the
list.
A
No
problem,
yes,
so
go
ahead.
James.
C
D
All
I
haven't,
I
was
hoping
that
it
would
help
him,
but
I'm
not
sure
how
much
it
does.
A
This
is
kind
of
discussed
here.
It's
our
events
on
the
call
and
then
lumiere,
I
see
you're
raising
your
hand.
E
D
The
the
first
link
in
the
pr
there
links
to
the
well-known
sids.
E
Yes,
I
know
that
they're
stable,
but
I'm
just
basically
trying
to
poke
at
the
problem
in
a
way
that,
like
what
can
go
wrong
like
with
this
binding
in
the
library
and
maybe
nothing,
maybe
everything
is
going
to
be
fine.
A
Yeah
brandon
is
there
somebody
that
do
you
think
I
think
we
could
probably
take
this
offline
and
take
a
look
at
this.
This
library.
B
A
Okay,
arvin
did
you
have
anything
to
yeah.
G
Cool,
so
to
answer
james's
question.
This
would
help
me
in
the
sense
that,
when
I
was
working
on
that
pr,
I
actually
ended
up
writing
some
of
the
code
instead
of
just
using
that
library.
So
it'll
help
me
in
that
way
that
if
we
are
okay,
because
at
the
time
I
think
james
you
had,
you
were
worried
that
the
library's-
you
know
you're
not
100
sure,
but
if
you,
if
we
have
overcome
that
we
could
the
way
would
help
me
is.
G
G
So
you
know
it
really
won't
help
there,
but
we
could
use
this
library.
There
is
what
I'm
thinking.
Okay,
thanks
yeah
and
I
have
not
done
any
work
in
that
in
that
area.
Yet
I'm
just
waiting
to
hear
back
from
you
folks
from
from
microsoft,
about
how
that
the
user
permissions
work
up,
and
it's
been
on
my
back
burner
to
write
up
this
talk.
I've
sort
of
written
up
a
bit
of
a
doc
to
document
the
current
state
of
projected
volumes,
and
I've
been
going
back
and
forth
with
david
eads
of
that.
A
Okay
yeah,
so
I
guess
the
next
steps
are
somebody
from
microsoft
to
kind
of
just
give
that
a
a
review
and
make
sure
that
there's
nothing
that
could
be
subverted
or
anything
there.
A
So,
okay,
jay,
do
you
want
to
we
skipped
this,
but
since
you're
here
do
you
want
to
talk
about
this?
This
agenda
item
for
a
little
bit.
H
H
So
I
think
now
we
know
what
needs
to
be
done,
and
now
it's
just
a
matter
of
kind
of
doing
it,
which
is
going
through
the
code
and
kind
of
figuring
out
how
it
works.
I'll,
probably
start
on
it
if
no
one
else
gets
involved,
but
I'm
sort
of
tied
up
in
other
things
right
now
too.
So
I
just
wanted
to
bring
that
up
again
for
folks.
H
If
does
anybody
want
to
look
into
that
or
if
folks
want,
we
can
look
into
it
during
the
pairing
session
or
whatever,
but
I
think
it's
a
relatively
well-bounded
project
now
that
we
know
we're
doing
so.
It
could
be
a
good
getting
started
issue.
A
Yeah
for
anybody
who
isn't
who
wasn't
here
last
week,
some
folks
kind
of
were
following
up
on
a
conversation
that
started
with
this
issue
where
nodes
would
get
into
an
unhealthy
state
if
pods
were
admitted
too
frequently
into
a
cluster,
and
then
that
led
to
a
discussion
about
what?
What
can
we
do
and
who
is
interested
in
working
on
kind
of
more
official,
perf
and
scale
tests
for
windows?
And
this
is
following
up
with
that.
Following
up
on
that.
F
But
my
time
is
highly
limited,
as
I'm
doing
the
exact
same
thing
for
container
d,
so.
F
If,
along
the
line,
somebody
else
wants
to
I'll
gladly
help
you
get
set
up
and
anything
else,
anything
else
that's
needed,
but
yeah
I
I
can
start
looking
into
it.
A
You
mentioned
that
you're
doing
the
same
thing
with
container
d.
Is
that
something
that
you
might
be
able
to
share
in
the
next?
Well,
some
of
the
next
coming
upcoming
community
meetings
about
what
any
findings
that
you
have
there.
F
Yeah
definitely
I
have
some
some
numbers:
I'm
still
trying
to
get
some
scenarios
working
properly
and
yeah.
A
Yeah
thanks
so
yeah
if
anybody
else
is
interested
reach
out
to
any
of
these
folks
here-
and
I
think
this
is
something
that
we
could
like
sig
windows
could
really
benefit
from.
A
I
don't
know
if
I
have
time
to
look
into
that
this
release
either
kind
of
in
the
same
boat.
It's
a
lot
of
people
all
right.
Next
agenda
item
is
host
process
pr
and
kept
z.
This
is
this.
Is
you
james.
C
Yeah,
I
just
wanted
to
point
it
out.
I
was
able
to
get
the
capstic
and
end
test
passing
and
I
also
ran
conformance
for
the
windows
end-to-end
test
that
one
kubernetes
and
tests
against
it.
So
it's
it's
ready
for
folks
to
take
a
look
at,
and
you
know,
give
a
good
look
over
and
make
sure
things
are
going
the
right
direction
here.
So
I
just
wanted
to
call
attention
to
it
because
it's
ready
for
review.
C
So
if
anybody
from
sig
windows
is
interested,
but
it's
using
host
process,
containers
and
then
and
calico
to
to
do
this
and
I've
built
the
containers
and
sig
windows
tools
as
kind
of
examples
to
to
get
started.
I
also
started
to
do
a
little
bit
of
work
upstream
in
calico
to
push
this
up
there,
but
with
the
changes
that
we
have
coming,
potentially
with
the
way
that
the
volumes
get
mounted
into
these
host
process
containers.
C
That
kind
of
is
stalled
because
it
required
some
pretty
not
pretty
changes
to
to
map
all
the
environment
variables
right.
So
currently
I
just
have
like
a
powershell
script
that
does
some
find
and
replace
and
pushes
all
the
binaries
in
the
right
place,
but
yeah
go
ahead
and
take
a
look,
I'm
looking
for
some
feedback
yeah.
I
think
it's
it's
pretty
close
to
being
ready
here.
I
C
Yeah
so
I
ran
the
300
or
so
tests
and
I
think
four
failed
two
of
them
I
knew
about
because
we
had
some
some
issues
with
the
images
last
week
and
then
there
was
one
that
I
didn't
quite
know
what
happened,
but
I
haven't
rerun
them
yet,
but
I'm
working
on
making
that
an
automated
process
so
that
we
can
run
them
continuously
and
figure
out.
If
there's
any
flakes
or
something
like
that,.
E
H
So
I
guess
james,
the
the
one
question
I
would
have
is
this
stuff
is
also
bleeding
edge
and
advanced.
We're
all
really
excited
about
it.
But
how
could
we,
unless
your
name
is
danny
or
perry
like
like
how
can
like
us
mortals
help
you
with
this,
I
guess
or
is
there
anything
specifically
looking
for
I
mean
it's
really
exciting
right,
like
it's
gonna
change,
how
all
of
us
install
cni.
So
I'm
definitely
looking
forward
to
stealing
this.
The
question
is:
is
there
anything
we
can
do
in
the
interim
to
help.
C
Yeah,
I
think
that
this
one
here
is
this
is
pretty
specific
to
capsi
and
just
sets
up
the
configs
properly,
but
that
other
pr
that
I
have
in
sig
tools
has
q,
proxy
and
calico
images
built,
and
you
should
be
able
to
take
those
and
reuse
them
in
your
environment.
You
could
build
a
docker
image
there
and
give
it
a
go,
and
I
think,
even
for,
like
sig
windows,
dev
tools,
we
could
potentially
try
using
this
over
installing
it
directly
on
the
node.
C
Yeah,
so
you
have
to
give
an
image,
but
the
image
isn't
actually
used.
Well,
the
some
of
the
binaries
from
the
image
are
used,
but
the
like
the
booting
of
the
image,
the
processes
that
just
run
right
on
the
host.
So
there
was
discussions
during
the
the
kep
whether
or
not
we
might
potentially
have
a
like
a
scratch
image
that
you
could
use.
There's
like
three
or
four
components
that
are
needed
to
actually
move
the
image,
but
I
think
it
got
scrapped
temporarily.
C
So
if,
if
there
is
a
desire
for
that,
maybe
we
can
bring
that
conversation
back
up
brandon.
I
don't
know
if
you
have
more
to
add
there.
A
Yeah
we,
I
think
that
there
was
during
the
discussions.
There
was
still
a
need
for
some
sort
of
a
windows
based
image,
similar
to
like
nanoserver
or
server
core,
that
you
build
your
images.
On
top
of,
I
think
that
we
like
to
discuss.
It
said
that
it
could
be
feasible
to
make
a
super
slim
image,
similar
to
a
scratch
image
that
people
could
base
it
off
of,
but
you
do
need
to
base
it
off
of
a
windows
based
image,
yeah
brandon
danny.
I
think
you
guys
were
the
most
involved
in
that
conversation.
B
Yeah,
I
think,
for
the
most
part,
it's
just
a.
If
there's
demand
for
it,
then
you
know
we're
happy
to
take
a
look
into
it,
but
it
seems
like
most,
people
are
happy
with
the
regular
base
image.
J
J
A
J
E
A
There's
a
feature
flag
or
feature
gate
that
needs
to
be
specified
in
the
cube
api
server
and
the
cubelet
for
all
of
the
nodes,
and
you
do
need
to
be
running
with
a
kind
of
a
pre-release
version
of
container
d,
which
is
called
out
here,
because
the
most
recently
released
version
of
container
d,
which
is
154,
I
believe,
does
not
have
full
support
for
host
process
containers.
But
as
long
as
you
have
those
two
things,
you
can
where's
an
example.
A
Does
that's
a
requirement
for
the
host
process
feature
right
now?
There
we
had
plans
to
eventually
have
another
enhancement
come
in
where
you
can
mix
so
right
now,
there's
a
requirement
that
all
the
containers
in
the
pod
need
to
be
host
process
containers
and
need
to
request
host
network
permissions.
A
C
A
All
right
does
anybody
have
anything
else
to
add
we're
nearing
the
end
of
the
30
minute
session
and,
if
not,
we
can
hand
it
over
to
jay.
Are
you
going
to
do
the
pairing
session
this
week.
H
Yeah
yeah,
I
have
an
interesting
hts
gym
issue
that
might
be
worth
looking
into
I'll
post
the
link
to
it
here
just
in
case,
since
we
have
three
minutes
this
and
then
we
can,
we
can
pair
on
it
after
the
three
minutes.
But
I
have
a
general
question.
That's
a
two-minute
question,
which
is
does
anxious
seem
like
nest
errors
inside
of
itself
like?
Is
that
a
thing
like
because
I'm
looking
at
this
error
message
that
I
just
posted
in
zoom
and
I'm
like
it
looks
to
me
this
came
from
an
internal
engineer.
H
A
It
wrong
I've
done
some
a
little
bit
of
changes
and
they
do
tend
to
wrap
error
messages.
I
think
like
and
danny
or
brandon
can
correct
me
if
I'm
wrong,
but
a
lot
of
times.
The
error
codes
returned
from
like
the
actual
operating
system
calls,
don't
have
any
context
that
would
help
in
actually
debugging
it's
just
like
here's.
The
error
code,
so
hcs
gym,
tends
to
wrap
theirs.
A
That
got
returned
from
some
of
the
system,
calls
to
add
context
on
them,
and
I
believe
that
sometimes
the
context
that
gets
added
to
the
errors
may
not
look
like
it
necessarily
matches
the
errors
that
are
returning,
but
in
general
those
are
added
for
debuggability
purposes
and
actually
are
relevant
to
the
error.
But,
okay.
H
We
can
dive
more
into
it
right
now.
Folks
want.
A
H
So
anybody
got
anything
else
they
want
to
look
at
before.
I
dig
into
this
other
thing
that
I
posted.
J
Jay
me
and
frederick
has
simply
dates
in
the
sea.
Dev
seeing
windows
have
tools,
maybe
would
be
cool
to
share
something.
H
J
So
basically,
we
have
been
struggling
the
last
days
to
integrate
the
image
builder
inside
our
environment,
see
windows,
dev
tools,
and
we
have
some
successful
approaches
in
this
process,
and
this
is
good
news,
because
we
now
have
a
centralized
way
to
bring
the
image
builder
cap
user
in
production
for
for
every
everywhere,
and
then
we
can
bring
this
in
our
environment
instead
of
building
our
own
form.
This
will
reduce
the
time
as
well,
so.
J
We
have
created,
like
these
experiments,
image
builder
folder
here,
where
you
can
keep
all
these
things.
We
have
been
documenting
this
information
here,
all
these
tabs
and
what
the
image
builder
is
doing
so
spend
a
lot
of
time
working
through
the
image
builder
process,
to
understand
how
we
can
build
the
same
team
for
virtualbox.
J
So
now
we
have
one
way
to
do
this
with
virtualbox,
as
well
as
a
hypervisor,
create
this
simple
script
here,
call
it
image
builder
as
such,
and
this
is
only
a
wrapper
for
the
packer
and
and
make
build
node
box
local.
So.
H
Is
this
all
merged
in
the
code
I
mean,
or
is
it
still
under
review,
it
is
merged.
J
Yeah
cool,
so
what
we
are
doing
right
now
is
like
overriding.
Can
we
creating
this
explain
very
fast
this?
What
the
approach
we
took
here.
So
basically,
we
have
this
building
virtual
box
that
creates
this
pr
is
open
there
on
image
builder,
but
on
our
side,
sig
dev
windows
box.
We
have
two
steps
here.
The
first
one
is
building
configuration
where
you
can
override
the
packer
configuration
with
the
information
you
want,
so
you
can
pass
like
easel
the
runtime
and
custom
roles
for
ransomware
to
run
on
top
of
the
existing
packer.
J
And
after
that
we
have
some
default
settings
that
we
keep
on
oversight.
We
can
improve
these
like
doing
some
jk
fans,
stuff
or,
but
we
have
xml
as
well.
So
it's
a
little
bit
tricky
to
to
get
a
more
generic
roof
or
measuring
for
merging
these
these
files
here,
but
basically
we
create
another
overlay
and
copy
one
by
one,
the
files
you
want
to
override
in
the
in
the
custom,
virtualbox
folder.
J
I'm
not
building
this,
but
it's
simple
as
running.
It
will
take
a
little
bit
time
to
build
this,
but
we
can
like
pass
a
few
variables
like
I'm
sending
my
repository
and
my
own
branch
and
the
ease
of
image
I
am
using,
so
we
can
run
like
image
builder.
In
the
end,
the
default
will
be
the
upstream
that
comes
through
overall.
So,
basically,
all
you
need
to
pass.
Is
the
iso
I'm
image
that
you
want
from
msdn?
J
So
if
you
create
a
dot
box
then,
and
in
the
example
here,
I'm
importing
inside
our
vagrant
file,
so
our
vagrant
file
got
very
smaller
because
we
don't
need
to
install
all
the
hyper-v
and
ssh
and
all
the
other
ps
that
we
had
before
now.
Everything
is
all
right
inside
the
box
I
need
to
do
is
like
join
the
the
node
in
the
closer
and
install
calico.
J
This
process
is,
is
the
the
cni
is
still
needed
to
to
do
in
the
end,
because
we
need
the
node
to
be
joined.
So
if
you
have
one
way,
this
is
like
some
to
do
something.
So
if
I
have
one
way
to
join
the
node
in
the
middle
of
the
view,
we
could
install
the
cni,
but
this
is
some
other
other
improvements
that
I
can
do
so
we
keep.
We
are
keeping
this
the
same.
Cine
installation
and
join
them
basically
have.
J
If
you
see
the
I
have
everything
set
up
here,
so
we
have
the
control
plane
and
the
windows
node,
so
you
can
ssh
in
the
control
plane.
J
Yeah,
okay
cool,
so
I
have
this
all
running
here:
I'm
just
using
the
nano
server
without
parameters
and
these
stupid
coup
cto
logs
issue,
great
yeah.
So
this
is
running
the
windows
box.
So
but
but
I
I
didn't
had
any
comments
just
getting
the
power
shotgun,
but
yeah
it's
it's
kind
of
running,
so
the
join
is
running
the
the
notes
running,
join
it
in
the
cluster
we
have
like.
J
H
L
It's
actually
a
pretty
convenient,
so
I
was
trying
to
avoid
ansible
for
so
long.
But
then
I
learned
that
there's
actually
a
really
nice
plugin
for
choco
in
ansible
and
you
just
if
you
want
to
add
more
packages
from
choco.
It's
super
easy
and
oh
my
god,.
H
D
If
you
try
it
on
a
linux,
you'll
see
something
very
nice
hollywood
you
might
have
to
install
it
first.
A
C
Yeah
so
I
tried
running
this
last
night.
I
posted
in
the
issue
that
I
got
in
the
chat
there,
but
I
was
chatting
on
like
my
personal
machine,
which
doesn't
have
a
whole
lot
of
cpu
and
memory
so
to
be
fair,
but
I
saw
that
there
was
like
a
light
variable,
so
I
copied
those
over
to
your
variable
file
and
then
I
got
the
error
that
I
pasted
in
there.
So
I
don't
know.
C
I
just
installed
virtualbox
and
vagrant
and
then
just
ran
this
on
linux.
No,
it's
on
a
windows,
machine
windows,
10.
H
L
H
H
L
But
I
always
used
hyper-v
on
windows.
L
My
favorite,
so
I
so
have
we
worked
on
my
machine.
The
problem
is:
what
was
it
because
my
machine
is
a
private
machine
is
not
part
of
a
company
network.
That's
not
a
problem,
but
I
think
the
shared
folders
do
not
work
in
in
company
networks,
because
the
way
hyper-v
uses
some
kind
of
network
device
virtual
device
to
do
the
synchronized
folders
and
there
was
always
a
problem
with
people.
I
never
could
repla
reproduce
the
problems.
H
A
problem
on
hyper
v2,
so
he
switched
to
virtualbox
and
that
worked
better
for
him.
So
I
think
we,
the
final
decision
that
we
made
was
virtualbox,
was
going
to
be
the
way,
the
thing
that
we
tested
and
supported,
because
we
just
can't
support
everything
so
yeah.
So
what
we
need
to
do
is
basically
figure
out
yeah.
We
need
to
get
this
okay,
we
need
to
think
of
whether
or
not
we
expect
it.
H
So
the
fact
that
you
got
it
to
work
on
hyper-v
in
windows
means
that
it
probably
can
work
on
virtual
boxing
windows
if
we
just
fix
this
mount
issue-
and
I
assume
this
is
file
system-
is
made
available
by
the
virtualbox
guest
additions.
L
By
a
kernel
module
but
isn't
guest
editions,
something
virtual
box
specific.
C
D
D
H
L
H
Yeah
I
mean
I'm
just
wondering
like
whether
this
is
something
that
can
be
fixed
with
a
one-liner,
and
then
we
automate
it
or
is
it
an
easy
thing
to
re-test
for
you,
james
yeah,
I
have
to
go
grab
that
machine.
Okay,
like
I
think
I
think
all
you
need
to
do
I'll.
Just
update
I'll
update
the
pairing
dock
with
this.
With
this
with
this
one
liner.
H
So
I
think,
there's
a
possibility
that
you
might
be
able
to
literally
run
that
one
thing
on
windows
and
then
maybe
that'll
work
and
then
you're
right
friedrich.
We
need
to
update
the
docs
and
everything
I
mean.
I
think
the
bigger
issue
is.
We
need
to
start
testing
on
windows.
H
So
with
virtualbox
I
have
a
you
know
I
could
repurpose
this.
I
have
a
machine.
I
think
I
could
repurpose
it
as
a
windows
machine.
L
H
K
K
C
L
H
J
L
J
L
C
By
the
way,
seeing
the
image
builder
get
used,
this
way
is
awesome
because
it
was.
It
was
always
something
I
wanted
to
try,
but
it's
cool
to
see
it
used
locally.
Do
you
have
an.
I
C
O
J
You
can
you
can
get
the
box
that
frederick
just
based
in
there.
I
use
a
2022.
H
J
E
J
Yep,
that's
what
frederick
just
pasted
the
end
there!
He
built
the
one
and
just
below
that
for
testing
already.
L
J
P
J
Because
you
need
to
to
use
my
branch
so
paste
the
command
I
have.
You
need
to
set
two
two
variables:
the
image
builder
repo,
then
image
builder
branch.
There
are
two
other
environment
variables
that
need
to
send.
This
is
not
an
upstream
image
builder,
yet.
H
J
Yeah,
maybe
that's
because
you
already
cloned
the
repository,
try
to
do
where
I'm
minus
effort
delete
the
image
builder
in
the
same
folder.
K
K
C
H
H
K
H
K
H
C
J
Yeah,
I
have
a
branch
with
the
cni
fix,
so
I
clean
up
the
the
calico.
I
need
to
get
like
the
it's
a
in
a
branch.
C
C
C
H
I
think
one
of
us
is
gonna
have
to
just
dig
into
the
windows
stuff
and
see
why
the
how
to
get
those
things
directories
working
without
hyper
v.
So
I
have
a
windows
machine.
Maybe
I
can
do
it
unless
friedrich
beats
me
to
it,
but
does
the
image
builder
work
on
windows,
james.
L
C
L
So
vagrant
downloads
boxes
right
and
we
just
upload
a
nice
ready,
provisioned
box
to
vagrant,
and
then
people
just
download
this.
H
L
J
K
L
I
don't
know
I
don't
I
think
it
came
on
linux.
It
came
with
the
necessary
plugin
vagrant
plug-in
list.
L
H
Yeah,
I
mean
you
know
we
just
need
to.
We
need
to
we
like
this.
Basically,
this
is
a
new
thing
that
we
just
haven't
done
yet,
so
we
just
got
to
get
it
working,
so
all
right
we'll
get
on
it.
James,
cool,
so
yeah,
like
I'm
glad
you
kind
of
found
this.
So
oh!
No.
What
did
I
do
I
just
hit
x
on
my
terminal?
Does
it
mean
that
killed
the
process.
K
H
H
I
feel
like
this
is
a
good
stopping
point
I'll
just
let
this
thing
run
in
the
background
and
let
you
know
this
is
great
work
friedrich
and
I
mean
just
it's
just
really:
cool
man.
H
Yeah,
we're
gonna
we're
gonna,
be
able
to
do
a
lot
with
this,
so
we're
gonna
need
to
like
show
people
how
to
use
it,
and
I
think
the
docs
look
really
good
to
start,
and
I
think
we
can
do
a.
We
can
do
a
tgik
on
this
in
a
week
or
two.
If
you
all
want,
maybe
the
two
of
you
all
run
it
and
I'll
just
I
can
show
a
meme
internally
how
to
host
a
show.
H
H
K
L
K
L
H
Kind
could
be
the
next
thing
right
like
we
could
potentially
have
kind
for
the
linux
mode
and
then
windows
as
the
I
think,
we've
talked
about
that
we
just
never
did
it.
Do
you
think?
Oh
windows
no
run
on
kind.
No
linux,
you
use
kind
for
the
linux
node
and
then
we
just
run
kubernetes
join
print
join
token
from
inside
the
kind
cluster
right
and
then
whatever
that
url.
As
long
as
the
windows
node
can
route
to
it,
then
you
can
have
a
kind
cluster
with
an
added
on
windows.
H
J
C
Have
because
they
have
the
cap
d
for
for
linux,
but
we
can't
run
any
of
the
cappy
tests
against
that
because
we
don't
have
windows
nodes.
So
if
we
had
like
a
cap,
vb
yeah,
I
could
actually
run
the
cappy
tests
against
windows.
H
H
J
H
I
mean
I
think,
as
is
this
thing,
you
know
we
could
do
all
that,
but,
like
looking
just
taking
a
step
back,
I'm
like
well,
we've
been
like
kind
of
gunning
it
to
like
make
this
thing
better
and
better
and
better
for
for
a
while
now
and
I
feel
like
we
should
probably
do
the
basic
stuff
like
it
works,
you
just
need
vagrant
and
virtualbox.
We
should
just.
H
H
Actually
that's
actually
going
to
make
the
demos
more
useful
to
people
right,
because
you
know
if
we
can
just
how
to
use
csi
proxy
with
this
and
how
to
use
different
c
I
providers
and
how
to
test
host
process
containers
and
how
to
use
sonaboo
to
run
all
the
end-to-end
tests
like.
I
think
people
are
going
to
get
a
lot
more
value
out
of
that
than
having
the
cluster
api.
I
mean
it's
just
you
know,
I
think
we
should
yeah.
H
No,
I
mean
it
actually
is
a
it
actually
is
a
really
cool
thing
to
do,
but
I'm
just
thinking
about
it.
I'm
like
I'm,
like,
I
think,
csi
and
c
getting
the
entry
and
all
that
stuff
is
probably
easier.
It's
hard.
That
stuff
is
also
going
to
be
hard.
Don't
get
me
wrong,
but,
like
it's
probably
going
to
be
more
for
us,
it's
going
to
be
more
valuable
because
I
think
we're
going
to
get
a
huge
audience
with
that.
C
Yeah
I
mean
to
me
like
if
we
could
point
if
someone's
like,
because
I've
had
several
people
come
to
me
and
say
hey.
I
want
to
work
on
this
end-to-end
test
for
windows
in
kubernetes
kubernetes
and
I
have
to
point
them
at
aks
engine
and
then
they
need
azure
and
then
they
or
they
need
google
and
yada
yada
yada.
If
I
could
just
say,
go
here,
run
up
yeah
and
run
the
test
against
it.
Then
they're
up
and
running
right.
It's
like
that.
H
Yeah
we
need
these
parts
and
by
the
way
you
can
do
that,
while
he's
doing
it
so
wendley's
doing
all
of
her
stuff
on
these
clusters.
So
you
can
run
the
engine
test.
We
just
haven't
proven
that
to
anybody
yet
really
in
the
public.
So
I
think
that's
what
we
need
to
do.
We
just
need
to
basically
like,
like
you
did
with
those
experiments.
You
know
we
need
to
basically
really
flesh
out
several
readmes
on
different
things.
You
can
do
with
this
right.
L
L
A
sneaky
t-shirt
just
arrived
yesterday
from
reamer
and
I'm
not
not
allowed
to
look
at
it
because
it
is
a
secret
t-shirt.
Really
I
guess
you,
I
guess
you
guys
know
about
this.
It's
it's
hidden
in
a
drawer.
I
can't,
I
can't
see
it
the
the
hidden
re-ember
t-shirt.
Yeah,
I
didn't
get
it,
I
don't.
I
don't.
O
L
K
J
J
Yeah,
I
know
what
it
is
about
the
z2
test.
This
is
the
same,
the
gap
that
we
are
doing,
the
conformance
ones
like
how
I'm
very
curious
how
we
can
degrade
this
cap
with
this
work
we
are
doing
and
how
these
things
will
fit
if
they
will.
H
I
think
I
don't
think
it
fits
into
the
cap.
I
think
I
mean
I
think
that
yeah
I
think
it's
just.
I
think
it's
supportive
of
the
cap,
because
now
anybody
can
run
these
tests
on
it,
but
it's
not
like,
I
think
the
cap.
We
just
need
to
finish,
defining
the
ginkgo
tags
for
those
tests
and
we
need
to
get
consensus
from
other
folks
and
just
be
like
okay.
What's
left,
maybe
maybe
I
gave
a
review
last
week,
okay,
okay,
so
we
have
some
review
comments
there.
H
So
yeah
we
just
got
to
finish
the
cap,
but
I
don't
think
there's
there's
like
there's
not
a
lot
of
code
to
write
for
that
cap,
it's
more
like
getting
consensus
on
it,
so
we
can.
H
All
right,
let's
I
gotta
go
too,
let's,
let's
just
follow
up
on
this
later.
This
is
awesome,
though
friedrich.
This
is
really
cool.
H
Oh,
this
is
a
big
big
win,
so
this
is
huge
for
big
for
cluster
api
and
it's
big
for
it's
going
to
get
vmware
really
behind
this.
I
think
you
know.
That's,
that's!
That's
part
of
it
because
now
the
vmware
and
the
cluster
api
community
is
going
to
be
able
to
point
to
this
as
an
example
of
how
their
tool
is
valuable
and
it'll
help
boost
us
in
that
sense
right,
as
opposed
to
us
just
building
all
of
our
own
tools.
H
H
H
Cool
all
right,
we'll
catch
up
later,
see
you
guys.