►
From YouTube: Kubernetes SIG Windows 20210223
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
february
23rd
2021
instance
of
the
sig
windows
community.
B
A
For
kubernetes,
as
always,
these
videos
are
recorded
and
uploaded
so
just
make
sure
you
and
here
to
all
the
cncf
code
of
conduct
and
also
just
be
nice
in
general,
all
right,
let's
get
started
so
we've
got
some
more
updates
on
the
caps.
A
This
is
kind
of
ongoing,
but
I
think
kind
of
wrapping
this
up
at
least
for
121.
For
now,
so
the
current
status
is-
and
I'm
not.
A
I
know
this
is
discussed
in
slack,
I'm
not
sure
if
the
tracking
spreadsheets
have
been
updated
or
not,
but
I
need
to
keep
working
with
the
release
lead
to
get
those
updated
the
statuses,
so
the
privileged
containers
kept,
I
believe,
on
wednesday
we
were
able
to
get
consensus
with
all
of
the
different
sigs
and
that
got
merged,
and
on
thursday
the
release
lead
for
121
granted
an
exception,
granted
the
exception
requested
and
are
going
to
allow
that
into
the
121
release
for
an
alpha
status
for
the
node
keplog
even
going
until
last
week.
A
There
was
a
lot
of
discussion
about
if
this
should
be
made
more
general
and
then
more
questions
about
the
security
aspect
of
this
which
we
believe
were
already
answered
and
there's
just
a
lot
of
discussions
still
going
on
with
this.
So
I
think
the
decision
right
now
is
to
use
the
rest
of
121
to
get
this
kept
into
a
state
that
everybody
agrees
on,
and
hopefully,
work
towards.
Implementing
this
in
an
alpha
in
122.
A
arvind
has
been
doing
a
great
job
for
this.
I
think
he
has
an
ask,
and
that
is
for,
if
anybody
is
kind
of
would
if,
if
folks
could
comment
on
the
kept
pr
about
how
this
could
be
used
for
windows,
I
think
that
would
help
or
like
and
about
how
they
plan
on
using
it.
I
hope
that
would-
or
I
think
that
would
help
kind
of
clarify
some
use
cases
that
folks
not
familiar
with
windows,
might
have
and
help
might
help
drive
this
along
arvin.
C
I
think
mark
you
captured
most
of
it.
The
key
the
key
bit
of
contention
is
like
clayton
wants
a
bunch
of
features
added
even
on
the
on
the
pod
log
side,
and
he
thinks
that,
given
that
it'll
cause
a
mismatch
between
pods
and
nodes,
it
needs
to
be
done
as
part
of
this
cap.
So
I'll
call
something
out
there.
The
thing
that
I'm
finding
it
hard
to
normalize
is
tim
hawkin
wants.
C
He
wants
that
to
be
sort
of
transparent,
and
you
know
which
makes
it
really
hard
right,
because
it's
almost
like
saying
that
you
need
to
know
information
about
the
cubelet
service
and
the
fact
that
it's
logging
to
warlog-
I
don't
know
cubelet.log,
which
could
be
anywhere
right.
You
could
set
up
your
cubelet
service
to
log
pretty
much
anywhere
anywhere
that
you
want.
So
that's.
D
D
From
from
a
support
standpoint,
I
thought
I
think,
that's
not
a
great
plan
anyway,
because
that
implies
that
people
are
going
to
be
using
this
to
get
around
knowing
anything
about
where
the
logs
have
been
generated
and
fight.
If
you're,
making
it
completely
invisible.
You've
got
to
have
a
way
of
being
able
to
make
it
like
make
it
visible
so
that
you
can
pick
out
where
the
issues
are,
whereas
if
it
was
completely
invisible,
you
could
find
an
mention
in
the
log
and
not
know
where
it
came
from.
C
A
Yeah,
so
I
think
the
kep
right
now
you
you
would
say
like
I
want
to
pull
this
log
file
from
you
know
var,
like
that
lives
in
this
nested
under
our
log,
or
you
can
point
it
to
a
service
name
and
for
windows
that
would
fire
off
the
etw
query.
For
that
kind
of
service
log
name
and
on
linux
it
would
query
journal
d.
A
So
I
think
the
question
is:
how
do
we
make
it
so
that
the
users
don't
need
to
know
if
things
are
getting
logged
to
just
a
generic
text,
file
or
a
service?
And
I
think
that's
a
very
difficult
question,
especially
because
there
could
be
so
many
different
configurations
like
node
configurations
like
who
knows
exactly
what's
going
to
be
running
on
the
service
or
like
on
the
nodes,
every
cluster
shape's
gonna
be
different.
D
It
sounds
similar
to
the
question
I
got
in
slack
the
other
day
about
about
aks
and
it
was
like
where's
the
start,
cubelet
ps1
file
for
that-
and
I
was
like
I
don't
know
it-
turns
out.
It's
in
c
drive
program
files,
eks
bootstrap,
dot
ps1.
So
it's
the
same
kind
of
question
right.
A
Yeah
anywhere,
I
actually
got
that
question
first
and
I
directed
them
to
sig
windows
and
maybe
you
but
yeah.
I
was
like
I'm
familiar
with
how
like
how
azure
sets
it
up,
but
I
don't
know
how
the
services
are
set
up
like
every
single
cluster
could
potentially
be
different
here.
So
yeah
I
don't
know
if
anybody
has
any
thoughts
or
extra
comments,
that'd
be
good
yeah.
I
I
don't
know,
is
there
more
that
we
should
can
discuss
here
or
not.
C
A
E
I
have
a
quick
general
thought
that
I've
been
having
a
lot
lately
because
in
sig
network
we've
started
moving
on
the
policy
side.
You
know
we
have
a
lot
of
stuff
coming
down
the
pike
and
we
go
back
and
forth
between
the
idea
of
at
what
point.
Do
we
just
make
a
kubernetes
dash
cigs
like
set
of
extended
tools
that
you
know
we
can
use
and
like
with
the
windows
stuff?
Sometimes
I
feel
like
between
the
conformance
tests
that
we've
been
talking
about,
trying
to
figure
out
a
solution
to
and
then
between.
E
Now
we've
got
logging
here
and
then
we've
got
some
of
the
security
context
stuff
I
just
like,
and
we've
got
the
sig
windows
tools,
repo.
I
wonder
at
some
point
if
we
think
outside
the
box
about
building
our
own.
Like
thing,
you
know
what
I
mean
like
that,
like
building
our
own
tools
that
are
really
really
really
good
at
the
thing
that
we
care
about.
E
You
know-
and
I
know,
there's
tons
of
consequences
to
that,
but
I
just
all
this
stuff
I
keep
thinking
of
it
like
we
could
build
something
so
useful
if
we
just
if
we
could
combine
some
of
these
concepts
into
like
our
own
swiss
army
knife.
You
know
what
I
mean
of
of
tooling,
but
I
don't
know
how
we
would
do
that,
but
just
a
general.
A
This
case,
if
like
what
like,
if
it
was
potentially
just
cube,
cdl
updates,
we
could
maybe
make
a
plug-in
for
that
and
host
it
in
there,
but
this
really
needs
features
enabled
and
code
enabled
in
the
cubelet
too.
So
I
I
don't
know
how
to
reconcile
that
either
with
how
do
we
put
together
a
useful
solution
in
synwindest
tools,
without
actually
being
able
to
do
some
of
the
updates
needed
to
enable
those.
A
But
yeah
we
do
have
sig
windows
tools
and
I
think
that
that
might
be
a
little
bit
underutilized.
So
if
anybody
has
any
ideas
on
how
to
like
for
projects,
that
would
be
a
good
candidate
for
that.
I
think
well,.
E
You
know
where
they
don't
have
to
memorize
like
what
secret
tool
they
have
to
use.
You
know
what
I
mean,
but
this
is
all
just
yeah,
but
yeah
I
mean
I
I
I
see
that
it's
probably
not
possible
for
this
one,
but.
A
Good
good
idea,
though,
all
right
next
is
I'm
not
sure
who
added
this
agenda
item
it.
F
F
Yeah
so
we
cut
over
to
the
promoter
image
repository
the
gc
gcr
promoter
image
repository
last
week
that
caused
quite
a
bit
of
failures
in
our
tests.
The
two
notable
ones
are
the
httpd
which
we
have
a
fix
for,
and
we
actually
even
have
a
general
idea
of
what's
going
on
there.
The
other
one
was
the
nginx
image
is
failing
and
so,
and
I
I'm
just
looking
into
it
a
little
bit
and
the
image
copies
directly
from
your
nginx
image.
G
G
It's
failing
interesting,
so
there
was
a
reason
for
that.
Nginx
image.
G
I
didn't
find
any
64-bit
engine
x
available
on
the
entire
internet.
So
I
built
one-
and
this
was
already
known-
I've
talked
about
this
during
a
previous
sig
windows
meeting
and
said:
hey.
If
you
want
to
use
nginx
on
a
nanoserver
based
image
there
you
go
so
yeah,
it
was
supposed
to
be
working
and
yeah.
F
It
looks
like
that
image
on
docker
hub
got
updated
14
days
ago,
and
and
since
then,
if
you
pull,
if
you
just
pull
the
image
even
but
not
through
kubernetes,
it
just
doesn't
move.
F
F
G
G
Those
are
just
normal
tags
there
and
basically,
it's
just
two
different
nginx
versions.
H
G
I'd
say:
no:
when
I
build
stuff,
I
build
on
my
own
registry
for
obvious
reasons
and
then,
when
I
told
sick
window
say
you
can
use
that
nginx
image.
I
put
it
on
the
e3
team,
because
maybe
people
don't
want
to
use
an
image
called
claudia
value,
slash
nginx.
You
know.
H
Yeah,
but
the
weird
thing
is
that
when
we
use
the
nginx
image
from
ede
team,
it
works
and
then
if
we
use
the
one
from
the
promoter
that
test
case
just
keeps
failing
so
yeah
yeah.
We
can.
F
A
All
right,
yeah,
let's
continue
to
this
one
offline.
It
seems
like
there's
just
a
couple
of
folks
that
really
can
take
action
here,
but
yeah.
Let's
maybe
try
and
move
this
all
over
to
the
ede
team.
Repository,
I
think,
is
that
are
all
of
the
are
all
these
issues,
especially
the
ones
you
just
linked
james.
These
are
all
related
right.
F
Yeah
they're
all
related
to
switching
over
to
the
promoter
registry
and
at
the
time
we
didn't
have
an
answer
on
the
htb
d
permissions
stuff.
But
we
got
an
answer
last
night,
so
we
don't
have
to
talk
about
it
here.
A
Okay
sounds
good.
The
next
item
was
which
is
kind
of
related.
Is
we
merged
the
pr
that
james
did
to
reorgan
or
to
add
a
new
board
on
test
grid
for
our
ci
signal?
So
right
now,
there's
some
there's
some
issues.
This
is
what
we
were
discussing
on
the
15
minutes
prior
to
the
community
meeting,
but
this
is
a
good
place
to
look.
A
These
are
the
dashboard
or
these
are
all
the
test
passes
that
the
release
team
will
look
at
and
help
us
kind
of
paying
us
if
there's
any
issues,
so
I'm
glad
that
this
merged
this
makes
it
easier
to
kind
of
focus
on
what
the
release
team
will
look
at.
So
thank
you
very
much
james.
A
Does
anybody
have
any
questions
about
this?
I
think
our
goal
going
forward
is
going
to
be
to
keep
these
test
passes
green.
Obviously
we
want
to
keep
all
of
them
green,
but
we'll
prioritize
these
over
all
of
the
rest
of
the
windows
ones
first
and
then
we'll
after
the
signal
goes
green,
we'll
look
at
the
rest
of
the
test
passes.
A
All
right
next
is
perry
and
jay.
I
think
you
had
a
call
out
here.
E
Yeah
we
had
a
couple
of
stragglers.
What
do
we
got
here?
Yeah,
so
quick
ping?
I
I
mentioned
this
up
in
upstream
channel
the
policy
stuff.
I've
got
it
like
working
we've
got
it,
we've
got
it.
We've
got
the
tests
running
and
they're
pretty
stable
for
us
for
the
most
part,
so
that
means
I'm
gonna,
be
removing
the
linux,
only
tags
from
them,
and
so
I
had
two
questions.
One
is
for
the
folks
in
the
room
that
host
cni's
and
stuff.
E
Do
you
all
have
road
map
for
when
you
all
are
going
to
support
network
policies
on
windows
for
with
either
calico
or
andrea
or
using
azure
cni
or
any
other
like?
I
just
wanted
to
get
a
quick
pulse
of
how
many
people
are
actually
going
to
support
that
as
a
feature
like
in
production
or
if,
if
nobody,
that's
that's
fine,
I'm
just.
C
Yeah
there
has
been
some
loose
requests
on
the
openshift
side.
I
mean
in
general
right
there's
always
the
question
about
when
it
comes
to
windows.
They'll
always
ask
us:
oh,
can
you
not
do
this
feature
that
linux
support,
so
network
policy
comes
under
that
we
haven't
prioritized
all
of
that,
but
stuff
like
this.
I
think
our
intent
should
be
to
always
try
and
you
know,
capture
sort
of
feature
parity
with
with
linux.
Okay,
if
not
immediately,
but
eventually.
E
E
So
that
would
be
yeah
I'd
love
to
know
more
if
people
are
running
in
production,
but
if
not,
I
guess
this
is
just
a
warning
that
my
plan
was
to
remove
the
linux,
only
tag
from
those
tests
and
hopefully
find
a
way
to
start
running
them.
Well
start
running
them.
I
assume
I'm
gonna
have
to
put
skip
tags
in
places
per
james's
thing,
but
so
am
I
gonna
break
stuff.
When
I,
when
I
remove
that
linux
only
tag,
am
I
gonna
break
ci,
stuff,
ravi
or
or
claudio,
or
does
anybody
know
james.
G
Pretty
much
yeah
could,
if
you
remove
the
linux
only
tag,
could
you
possibly
add
skip
some
type
of
other
tag
which
we
can
then
use
in
the
skip
projects
for
the
other
jobs
which
do
not
support
network
policies?
Yes,
sir,
absolutely.
E
So
is
it
not
already
a
net
poll
skip
there's,
there's
not
because
it's
usually
relying
on
linux.
Only
so
I
guess
in
the
future
we
probably
should
put,
should
rely
on
linux
only
as
as
little
as
possible
and
be
more
explicit
about
what
we're
skipping,
but.
H
A
E
Well,
this
goes
back
to
the
other
thing.
I
was
wondering
it
in
over
and
sig
network.
You
know
we
have
now
service
conformance
and
we're
thinking
about
a
network
policy
conformance,
and
I
I'm
wondering
whether
this
is
one
example
of
one
of
those
things
where
we
do
a
kubernetes,
sigs
windows
conformance
and
we
just
have
our
own
tests.
That
we
know
are
really
awesome.
I
mean
that's
what
I
would
like
to
have
the
most
right,
rather
than
jumping
around
the
ginkgo
stuff,
to
find
bugs
and
images
and
other
things
that
I
don't
care
about.
E
Did
that
with
the
network
policy
stuff,
and
we
just
made
our
own
stuff
and
people
find
it
so
useful?
You
know
because
it's
a
tool
that
actually
solves
the
problem
of
knowing
whether
policies
work
or
not,
as
opposed
to
a
tool
that,
as
opposed
to
spending
all
of
our
time,
trying
to
figure
out
what
dance
we
need
to
do
to
get
something
merged
into
e2e.
You
know
I
mean
we
did
get
it
into
e2e,
but
we
were
very
close
to
just
going
anyways.
B
A
I'll
should
I
add
that,
to
a
future
a
future
agenda
item
either
going
through
and
cleaning
up
the
like
the
test
tags
or
the
windows,
networking
conformance
right
with
rlc
conformance.
I
think
that.
E
I
mean,
I
think,
it's
a
broader
if
you
feel
free
to.
I
feel
like
it's
a
broader
question
of
windows
performance
to
me.
Just
let's
have
some
awesome
tests
that
always
work
that
only
use
one
image
or
two
images
that
we
all
agree
on.
You
know
that's
the
way,
but
that
actually
print
out
useful
stuff
you
know
like,
but
I
don't
know,
there's
a
lot.
E
A
I
think
it
could
go
either
way.
I
think
that,
especially
if
there's
a
kind
of
an
effort
to
get
more,
that
the
test
results
from
more
cloud
providers
onto
test
grid
for
windows
that
could
be
a
big
benefit
there
or
or
more.
I
should
not
even
say
cloud
providers,
but
more
infrastructure
providers,
I'll.
B
D
Yeah
it
was
99360
was
that
the
was
that
the
one
I
raised
this
morning
about
the
in
the
port
yeah
the
pause,
build,
there's
a
there's,
a
public
docker
image
in
the
pause,
build
for
windows,
the
docker
cross
or
doc
cross
or
whatever
that
image
pulls
from
docker
hub,
and
when
I
tried
to
do
a
build
on
a
build
server,
it
died
so
yeah.
D
G
Yeah,
honestly,
I'm
a
bit
surprised
you
fit
you've
hit
the
rate
limiting
from
docker
hub
I've
been
building,
probably
hundreds
of
images.
D
G
Yeah,
that
makes
a
lot
more
sense,
so
we
actually
mirrored
quite
a
few
docker
images
to
the
kids
registry,
for
example
the
httpd
image,
the
nginx
image
power,
image,
basebox
image
and
a
few
others
which
were
commonly
being
used
in
into
intestine
e3
tests.
G
But
for
this
we
haven't
typically
for
image
builders
and
so
on
so
forth,
which
run
infrequent.
G
If
you
actually
find
a
cube,
close
equipment
on
gcr
io
definitely
send
the
sender
up
request
for
that.
But
in
this
case
I
think
you
can
just
solve
your
issue
by
just
doing
a
docker
login,
because
you,
you
basically
have
docker,
offers
a
rate
limit
for
anonymous
users
and
a
different
rate
limit
for
authenticated
users.
E
G
G
E
D
Which
show
you
know
which
I've
done,
I've
I've
I've
done
a
I've,
basically
just
downloaded
and
done
a
cache,
but
I'm
just
thinking
in
future.
I
would
I'd
rather
not
have
to
put
in
a
step
that
downloads
one
in
a
proxy
cache
and
then
uses
that
I'd
rather
have
just
this
sort
of
mimic
windows.
Linux.
Sorry.
E
G
G
E
D
I
I
It
with
the
servicing
team
with
the
wsd,
it
is
part
of
it.
The
reason
there
is
no
release
notes.
If
you
look
at
it,
there
won't
be
any
release
notice.
It's
a
little
bit
confusing
for
general
users
right
like
it's
a
very
specific
scenario
about
smb
shares
using
ccg,
so
it
was
better
not
to
release
it.
You
know
with
the
notes,
but
it
has
definitely
made
it,
so
it
is
2c.
So
the
b
patches
are
all
only
security.
The
non-security
updates
are
usually
c.
A
So
yeah,
so
that
means
it'll
be
included
in
the
3b
update
correctly,
which
should
get
automatically
applied
or
like
which
you
can
download
and
will
be
kind
of
a
recommended
update.
But
if
you
want
it
sooner,
you
can
manually
install
the
two
c
updates,
which
will
not
be
kind
of
showed
up
or
showing
up
as
a
manual
or
as
a
recommended
update.
D
I
D
A
You
can
do
wenver
and
look
at
the
the
this
is.
This
will
be
the
the
minor
version.
If
your
windows
version
at
least
someone
to
serve
2019,
is.
A
10.0.17763.1790
or
greater
you'll
have
this
fix.
The
those
the
servicing
patches
are
always
increasing,
so
the
3b1,
which
will
come
out
next
week,
will
be
higher
in
a
couple
weeks,
will
be
higher,
we'll
have
a
higher
patch
version.
E
A
A
like
no
so
a
while
look
like
there
used
to
be
individual,
like
the
these
updates
used
to
be
cumulative
or
the
these
updates
used
to
be
roll-up
packages
which
contained
individual
kbs
that
you
could
check
to
see.
Hey,
there's
one
kb
and
the
package
would
include
many
of
them,
but
I
I
actually
have
no
idea
of
the
reason
why,
but
now
there's
just
single
kb
packages
that
are
installed
and
there's
no
way
like
you
kind
of
just
have
to
know.
What's
in
it,
there's
no
way
to
check
if
you
have
specific
things
and.
E
H
D
I
I
must
admit
I
was
expecting
a
kb
number.
That
just
says
this
batch
is
the
one.
So
that's
a
bad
day
as
well.
Yeah,
that's
fine!
Is
it
worth
then,
once
the
the
next
update
comes
out
in
2c,
should
I
once
we've
got
kb,
I'm
quite
happy
to
update
the
docs
with
the
you
know,
for
if
you
want
gsma
sport
on
on
windows
on
container
d,
you
need
to
you
know
you
need
to
apply
this
kb
number,
but
I
guess
right
now:
it's
probably
not
worth
it.
If
it's
not
a
main
update.
A
D
A
All
right
anything
else
for
this
I've
got
one
thing
to
update
and
then
I'm
going
to
drop
for
the
sig
node
meeting.
This
is
more
of
just
an
fyi,
but
we
discovered
an
issue
with
container
d
and
any
storage
plugins
that
mount
disks.
So
this
is
azure
disk
and
probably
the
google
pd
driver
with
container
d
and
windows
there's
more
details
in
there.
A
J
So
yeah,
I
I
discovered,
I
think
same
or
similar
issue
in
step
pass,
so
I
fixed
there.
I'm
not
sure
this
can
be
fixed
similar
way.
Basically,
I
replace
evaluating
link
call
with
some
powershell
commands
like
get
pass
like
item
yeah.
A
B
A
If
we
can
fix
this
in
any
kubernetes
components
or
not,
which
is
why
I
created
a
separate
issue,
we'll
need
to
keep
working
on
this,
to
figure
out
the
appropriate
place
and
way
to
resolve
this.
So.
A
If
you
are
certain
dr
certain
plugins
are
like
azure
disk
is
because
azure
disk
will
mount
a
disk
to
a
vm
provision,
the
disk
and
then
create
a
volume
on
the
disk
and
then
link
to
that
volume.
But
azure
file
which
uses
an
smb
share
is
not
so.
It
really
depends
on
kind
of
the
underlying
up
or
like
operations
of
the
driver.
A
G
Okay,
speaking
about
volumes
and
mappings
and
sub
paths
and
symbols
and
so
forth,
I
also
had
an
issue
which
I
discovered
on
container
d
itself
regarding
siblings
and
so
on
so
forth,
which
may
cause
for
some
people
to
have
some
failures,
so
I've
linked
that
in
chat
and
I've
also
sent
a
fix
for
something
like
this
in
kubernetes
itself,
which
is
this
one
and
there's
also
an
issue
on
commodities.
That's
linked
on
the
request
itself.
G
Apparently
there
are
certain
scenarios
in
which,
if
you
have
sim
links
in
the
sub
path,
you
are
mounting
as
a
volume
in
a
pod.
Basically,
the
point
will
not
be
able
to
start.
The
container
will
not
be
able
to
stop
because
of
that,
because
of
how
the
symmetric
is
evaluated
in
h,
h,
h,
c
s,
shame,
but
that
has
been
fixed
in
his
his
shim
as
far
as
I
know,
but
that
has
to
be
ported
to
container
d
as
well.
A
G
A
G
J
So
I
think
before
I
raised
the
issue
for
windows
about
evaluating
function
and
they
had
a
huge
discussion,
but
it's
still
kind
of
someone
proposed
something
that
is
not
approved,
so
it
seems
like
they
are
just
trying
to
document.
Every
single
link
will
not
work
for
windows.
J
There
are
some
details
in
that
discussion.
I
can
also
put
some
the
link
somewhere.
F
E
Topics
nope
nothing
for
me
for
anybody
that
wants
to
dive
into
the
container
d
stuff.
We
are
having
a
call
with
the
microsoft
folks
in
a
little
bit
today,
just
fyi
around
cni
stuff.
I
I
didn't
get
a
chance
to
add
every
single
person
here,
but
for
folks
that
really
want
to
dive
into
it.
Just
ping
me
in
upstream
slack,
sig
windows,
I'll
post,
that.
E
B
Yeah,
the
other
thing
that
I
want
to
bring
this
bring
up.
This
is
related
to
what
jay
was
talking
about
earlier,
which
is
related
to
having
confirmed
and
stress.
So
there
is
this
conformance
working
group,
which
actually
has
said
in
the
past
that
the
conformance
tests
are
specific
to
lungs.
We
do
not
have
anything
which
is
window
specific,
so
I
think
that
is
something
that
can
be
started
as
a
project.
B
I
just
want
to
second
jay's
thoughts
there,
so
they
they
have
told
it
in
the
past
that
they
are
supposed
to.
J
F
For
this
too,
so
I
don't
know
if
you're
familiar
with
that,
but
adelina
started
the
conversation
at
one
point.
So,
let's
see
if
I
can
find
the
open
issue
there.
E
But
this
is
why
I
feel
like
I
feel
like
we
should
talk
about
this
offline,
ravi
or
anyone
else
like
the
idea
of
like
a
windows,
tooling
or
testing
subgroup,
or
something
that
actually
is
actively
working
on
on
solving
this
problem
for
us
in
a
way
that,
like
makes
it
really
easy
to
like
verify
that
a
windows
cluster
is
doing
the
right
thing
and
so
that
we're
spending
less
time
jumping
through
hoops
and
and
more
time
actually
verifying
the
core
functionality.
E
That's
kind
of
how
I
feel
like
I'm
spending,
that's
I
feel
like
I'm.
You
know.
E
B
Yeah,
the
other
thing
that
I
wanted
to
ask
you
jay,
is:
what
is
the
cnc
plugin
that
we
are
using
for
the
network
policy
test?
Is
it
flannel
or
is
it
calico?
What
what
are
we
using
there.
E
E
Your
practices,
one
of
the
plugins,
did
you
use
we
any
any
cni
plugin.
You
use
you're,
going
to
find
network
policy
issues
so,
but
the
overall
test
framework
is
actually
in
upstream
kubernetes,
but
we
were
like
inches
away
from
from
going
out
of
tree
just
so
that
we
could
get
it
standardized
and
get
it
to
live
somewhere
and
initially
we
lived
in
andrea,
but
it's
not
cni
dependent
right.
It's
a
it
verifies
the
policies,
work
and
we've
built
other
tools
around
it.
E
That
will
automatically
test
psyllium
and
other
things,
and
we
have
a
working
group
for
that,
but
and
we're
working
on
publishing
results
for
that.
But
right
now
we
mostly
use
it
like.
We
use
it
with
oven,
kubernetes
to
find
issues.
We've
used
it
for
andrea
to
find
a
lot
of
issues.
We've
found
some
issues
in
calico
calico
tends
to
be
the
best,
but
but
that's
just
one
example.
I
feel
like
we've
got
that
with
windows.
E
We
just
haven't
like
I
feel
like
we're,
but
yeah
I
don't
know
we
should
talk
about
this
offline
yeah.