►
From YouTube: Kubernetes SIG Windows 20220308
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
I'll
start
with
some
announcements.
First
announcement
is
code.
Freeze
for
124
is
march
9th.
I
think
it
straddles
the
time
border
difference
like
a
so
it's
it's
march,
30th
utc
and
it's
march
29th
for
pst
based
folks.
So
that's
a
little
that's
a
little
bit!
That's
yeah
exactly
three
weeks
too.
So
just
be
aware
of
that,
if
you're
working
on
enhancements
or
changes
that
you
want
to
get
into
124
release,
next
announcement
is
the
leads,
and
myself
have
completed
our
the
annual
report
for
each
sig
that's
available
here.
A
These
annual
reports
are
a
requirement
for
all
sigs
and
working
groups
in
the
cncf
and
projects,
but
they're
a
good
chance
for
the
sig
leaderships
to
highlight
things
that
the
sig
has
been
has
worked
on
in
the
past
year.
Areas
that
like
initiatives,
are
areas
that
need
more
contributors,
also
share
recordings
and,
and
everything
like
that,
so
if
you're
interested
please
take
a
look,
these
are
they're
just
markdown
files.
A
If
anybody
has
anything
they'd
like
to
add
or
anything,
please
go
ahead
and
open
up
a
pr
for
that
and
if
not
just
take
a
look.
Oh
it's
also
a
good
opportunity
to
make
sure
that
things
like
each
sig's,
readme
and
contributing
guidelines
are
up
to
date
and
we
did
update
the
contributing
guide
this
year.
It
was
quite
stale,
so
it's
a
nice
reminder
to
just
go
ahead
and
make
sure
those
resources
are
up
to
date.
A
Okay,
I
don't
see
anything
else
on
the
announcements
or
much
on
the
agenda.
I
added
I
can
do
a
demo
if
you
want,
I
did
want
to
come
into.
I've
noticed
a
little
bit
of
a
pattern
of
there
being
not
a
lot
of
agenda
items
on
the
agenda
going
into
the
meeting,
but
then
it
seems
like
each
week
we
fill
up
most
or
all
of
the
the
time.
A
So
I
was
wondering
if
people
how
how
people
wanted,
if
people
had
any
thoughts
or
concerns
about
that.
I
want
to
be
aware
of
everybody's
time,
so
I'm
not
sure
if
people
would
prefer
to
keep
doing
this
and
we
can
just
have
kind
of
ad
hoc
discussions
or
if
anybody
would
rather
us,
cancel
the
community
meeting.
If
there's
not
much
on
the
agenda,
let's
just
take
a
couple
minutes
and
give
thought
or
people
can
comment
about
that.
A
I
personally
find
these
meetings
valuable,
even
when
there's
not
really
a
structured
agenda
and
they're
half
an
hour
once
a
week,
so
I'm
I'm
fine
keeping
them
standing.
But
does
anybody
have
any
thoughts?
I
don't
know
james
claudio,
you
guys
are
brendan
alango.
You
guys
are
pretty
consistent
attendees.
So
if
you
have
any
thoughts.
B
B
Useful
actually,
so
anything
we
issues
or
something
which
we
run
into
or
the
idea
of
the
information,
what
we
collect
where
we
are
sharing
or
we
are
getting
some
feedback
or
information
from
other
folks.
It
is
really
useful
and
valuable
to
us.
Actually,
I
would
like
to
continue.
C
I
agree
with
with
him
I
think
it's
very
useful
for
especially
for
newer
fork
to
just
drop
in
and
get
to
know
the
community.
First.
C
If
we
don't
have
a
lot
of
on
the
agenda,
we
can
always
discuss
future
plans
or
new
things
that
we're
working
on
and
so
on
and
so
forth,
and
there
were
a
couple
of
cases
in
the
past
when
people
from
the
community
were
actually
coming
coming
up
with
questions
for
us
during
the
meetings,
even
if
they
weren't
specifically
set
in
the
agenda
itself.
A
D
Yeah
and
just
so
everybody's
aware,
I
think
if
you
join
the
sig
windows.
Mailing
group,
then,
which
is
at
the
top
of
this
document
here
you'll
get
access
to
be
able
to
edit
this
document.
D
And
so
you
can,
you
can
add
any
agenda
items
you
don't
have
to
necessarily
wait
to
the
day
of
to
show
up,
and
that
also
helps
my
stealth
and
mark
and
other
folks
to
just
be
prepared
for
any
kind
of
questions
that
come
in
if
they,
if
they
do
or
maybe
do
a
code
review
of
something
that
we
want
to
discuss
before
we
before
we
meet
so.
A
A
Okay,
does
anybody
have
anything
they
want
to
discuss?
I
have
something
I
can
demo
if
people
want
or
we
could
discuss
other
things
or
in
the
meeting.
It's
really
up
to
you
guys.
B
I
I
have
one
question
regarding
2022
support
david
on
this
call,
so
we
ran
into
some
issue
with
the
traffic
east-west
traffic
and
the
private,
build
what
he
provided
to
us.
Indeed,
it
fixed
the
problem,
but
the
workaround
muhammad,
like
the
moment,
is
also
called
the
one
of
the
work
around
what
is
suggested
for
overlay.
It
did
not
work
for
us,
but
again
the
os
fix.
What
is
the
dll
files,
the
fix,
what
he
gave
it?
It
fixed
our
problem.
Well,
we
have
two
questions.
B
One
is
that
about
the
workaround.
The
second
one
is
that
the
eta
approximate
each
year
is
like
april.
19Th
is
what
he
has
communicated.
I'm
not
sure
whether
it
is
officially
the
date
which
is
communicated
by
microsoft,
or
I
want
to
know
about
the
date
so
that
if
our
customers
ask,
we
will
be
able
to
communicate
to
the
customers
as
well.
Right.
Second,
one
is
that
about
the
workaround.
A
First
is
this:
is
this
the
there's
a
thread
in
that's
being
discussed
in
slack?
Is
that.
A
That's
what
I
was
looking
for
yeah,
I
just
added
it:
okay,
perfect,
okay,
yeah!
So.
E
I
think
david
is
on
the
call.
He
cannot
sleep
to
the
date
for
the
the
date
that
is
tentative,
so
it's
not
a
hundred
percent
guaranteed
and
all
the
dates
that,
even
though
the
csd
team
directly
the
servicing
division
at
microsoft,
give
out
it's
always
tentative
and
pending
additional
testing
and
validation.
E
B
Okay,
so
that
is
1081
thanks.
E
E
And
regarding
the
second
issue
of
the
workarounds,
there
were
two
issues:
one
was
that
the
load
balancing
would
not
work
for.
If
you
try
to
connect
from
a
pod
to
cluster
vip
like
the
service
flip
and
then
the
other
issue
was
that
the
pod
to
pod
traffic
might
fail,
and
so
for
the
pod
to
service
traffic.
Failing
that
one,
you
could.
E
E
I
think
l2
bridge
only
where
it
was
pod
to
service
traffic
would
fail
and
that
one
you
could
enable
the
local
traffic
policy.
B
Okay,
so
ours
is
not
a
l2,
that's
why
it
didn't
work,
but
ours
is
not.
I
think
ours
is
over.
B
Now
we
are
clear:
okay,
I
just
got
confused
because
the
thread
when
we
discuss
about
the
one
more
thing:
okay,
now
it's
clear,
so
micro,
the.
B
E
B
A
A
A
Well,
any
other
questions
about
networking
on
windows,
server,
2022.
B
One
more
thing:
what
we
are
trying
to
do
is
muhammad.
What
is
trying
to
do
is
that
we're
going
to
have
a
golden
image
with
whatever
suggestion
the
private
image
is
given
by
david,
and
we
are
going
to
run
our
ci
to
make
sure
that
what
you
call
whether
we
are
going
to
run
into
any
other
issue-
or
this
is
the
only
issue.
So
probably
we
will
do
an
update
on
that.
A
Okay,
I
think
make
sure
that
those
private
those
images
are
not
accessible
to
the
public.
I
think
right,
david.
D
B
This
is
only
for
our
testing
just
wanted
to
make
sure,
because
david
was
asking
how
about
other
traffic
and
everything
else
are
working,
so
we
want
to
make
sure
that
everything
works.
Apart
from
this
issue
right,
then
we
can
kind
of
like
kind
of
say
when
it
is
ready,
then
we
know
we
covered
everything
we
don't.
We
will
not
have
any
surprise
at
the
last
minute.
E
B
A
A
F
F
F
Yeah
yeah,
so
he
was
so
obama
was
there,
and
so
I
think
we
were
kind
of
in
this
place
where
we
are
thinking
to
redesign
the
kernel
space
proxy
now
that
it's
working
in
kpng,
because
we
can
do
that
and
since
kpng
is
kind
of
a
total
rewrite
of
the
coupe
proxy.
It's
kind
of
like
the
perfect
place
to
do
it
and
it
has
all
the
apis
that
you
would
want
that
would
allow
you
to,
for
example,
not
have
to
run
a
watch
from
a
windows,
node
and
and
so
on
and
so
forth.
F
So
I
just
it's
just
more
of
an
announcement
than
anything
else.
I
mean
you
know
if
anybody
wants
to
be
involved
in
that
they
can
be,
but
it's
a
very,
very
detail-oriented
and
time-intensive
thing
to
be
involved
in
it.
So,
but
if
anybody
wants
to
just
like
reach
out
to
me
on
the
sig
window,
slack
channel
and
and
make
sure
you
have
time
makes
you
ever
at
least
a
couple
of
weeks
to
really
dig
into
the
code
and
start
learning
it.
F
F
But
I'm
happy
to
help.
Also.
F
D
E
F
F
Yeah,
I
mean-
maybe
maybe
as
you
do,
that
maybe
we
could
sort
of
talk
through
it
and
you
could
sort
of
maybe
demo
that
at
sig
windows
how
you're
doing
it
so
that
that
sort
of
knowledge
is
there
and
we
can
sort
of
transfer
that
into
the
kpng
folks.
So
we
get
that
that
plumbed
in
there
or
if
you
want,
there's
a
pull
request.
F
I
have
into
kpng
that
ports,
the
existing
coupe
proxy
over
and-
and
maybe
you
could
sort
of-
also
add
that
there,
if
you
want-
maybe
maybe
just
reach
out
and
just
keep
me
abreast
of
what's
going
on,
if,
if
possible,.
E
Yeah
sure
yeah
yeah
surround
it's
pretty
closely
involved
in
both
the
developments
so
cool
all
right.
A
Do
you
want
to
link
that
the
poll
requests
you
mentioned
in
the
meeting
notes,
jay.
E
F
So
while
I'm
doing
this,
just
so
folks
know
like
the
way
that
this
has
been
done,
is
it's
literally
like
I've
taken
the
existing
proxy
and
sort
of
like
literally
like
sort
of
cut
it
up
into
little
pieces
and
made
it
so
that
it
can
be
triggered
from
the
kpng
api?
So
it's
not
a
very
it's
a
pretty
nasty
piece
of
piece
of
code
right
and
so
you
can.
F
You
can
make
fun
of
me
if
you
want,
but
there's
a
lot
of
context
and
really
that
this
implementation
of
kp
g
doesn't
look
like
a
major
improvement,
because
it's
not
actually
leveraging
kpng
in
the
purely
idiomatic
way,
but
the
next
one
we
can
do
that
and
if
some
folks
want
to
see
the
what
we
really
want
to
do
with
kpng
is
check
out
the
nft
or
the
ipvs
implementations
and
and
there's
no
caching
there
right.
So
so
it's
super
easy
to
understand.
F
F
A
Okay,
I
can
do
a
quick
demo
of
getting
some
of
the
more
end-to-end
workflows
for
a
csi
driver
working
as
host
process
containers.
If
anybody's
interested
that's
what
I
was
working
on
for
a
little
bit
before
or
last
week
cool,
I
have
a
pull
request
for
to
for
deployment
files
for
the
azure
disk
csi
driver
up
here.
Let
me
share
my
screen
and
this
is
not
a
very
structured
demo,
so
if
it
doesn't
work,
I
apologize.
A
So
this
is
right
now,
all
of
this,
the
csi
stuff.
First,
for
some
context,
all
of
the
windows,
csi
interactions
were
developed
before
host
process.
Containers
were
kind
of
a
thing,
and
that's
why
we
have
csi
proxy
today
is
csi
proxies
how
the
csi
proxy
is
a
component
that
the
csi
drivers
talk
to
to
manipulate
the
host
where's
my
window,
so
whereas
on
linux,
usually
those
just
run
as
privileged
containers.
A
So
today,
and
then
today
we
the
we
have
a
deployment
file
for
to
run
the
csi
proxy
as
a
host
process
container,
and
then
you
can
still
run
the
regular
containers
as
or
the
the
csi
driver
containers
as
regular
containers.
But
the
deployments
are
a
little
bit
convoluted.
There's
a
lot
of
different
value,
mounting
going
on
to
make
sure
the
right
parts
of
the
folders
that
the
drivers
need
to
get
access
to
on
the
host
have
get
mapped
into
the
containers.
We're
just
running
in
this
host
process
container
greatly
simplifies
the
deployment.
A
A
Yep
just
running
as
a
host
process,
container
and
part
of
the
csi
or
driver
installation
process,
there's
some
helm
charts
that
are
provided.
Let
me
run
through
that
and
show
the
templates.
A
A
I
think
that's
these
guys
yeah,
so
we'll
run
as
host
process
containers
and
there's
three
containers:
there's
still
there's
a
no
driver,
registrar,
a
liveness
probe
and
a
csi
plug-in.
A
Let's
wait
for
those
so
now
we
can
see
a
bunch
more
containers
running
there's
a
azure
disk
controller.
This
runs
on
the
linux
nodes.
This
is
what
actually
talks
to
azure
to
provision
the
nodes.
Why
are
we
mirroring
out
here.
A
A
A
This
okay,
I
think
I
know
what's
going
on
here.
This
is
let
me
so
with
this
yeah
I
was
kind
of
going
and
rebasing
around
some
other
changes
that
were
coming.
This
is
just
an
unsupported
flag,
getting
passed
in
which
I
thought
I
already
addressed,
but
let
me
I
can
open
up
the.
A
F
A
The
the
e
to
e
tests
are
passing
so
for
this
pull
request,
which
is
new
we,
so
it
is
working.
It's
just
a
configuration,
sorry
that
I'm
not
a
little
bit
more
prepared
for
this,
which
it
was
set.
Let
me
just
share
my
whole
screen.
F
A
A
A
So,
while
we're
at
it,
I
can
highlight
the
differences
between
the
deployments.
So
here's
the
deployment
here,
let
me
zoom
in
a
little
bit.
Let
me
first
put
them
side
by
side
right.
A
Oh
well,
so
here
you
can
see,
there's
a
lot
of
different
volume
mounts
that
are
needed.
A
There's
a
a
lot
of
named
pipe
volume
mounts
and
a
couple
of
different
folders
on
the
that
map
to
different
hard-coded
directories
in
the
cubelet
or
around
the
cubelet,
and
that's
for
registering
cubelet
plugins
and
things
like
that
and
a
lot
of
endpoints
that
get
mapped
into
folders
that
are
mounted
in
those
new
volumes.
A
A
Running
so
then,
to
prove
it
works
on
the
there's,
some
demos
there
should
be
csi
driver.
Let's
go
there,
there's
deploy
example
windows,
so
first
we're
going
to
deploy
the
storage
class
from
this
file.
A
There's
no
persistent
volume
claims
this
yaml
file
here,
which
I
can
zoom
into,
contains
some
will
deploy
this
and
then
look
at
this.
So
this
is
a
storage
class,
and
this
is
a
pod
that
uses
a
persistent
volume
claim
that's
backed
by
a
storage
type
of
disk.csi.azure.
A
So
where
is
that
yeah?
So.
A
A
Okay,
so
yeah
now
we're
seeing
the
disk
getting
attached.
So
that's
the
azure
disk
plug-in
is
attaching
the
disk
to
the
node.
This
should
be
done.
Oh
right,.
A
And
I'm
running
this
all
on
windows,
server
2022,
so
the
deployment
is
this
deployment
has
windows
server
2019
in
there,
but
we
can
edit
that.
A
Oh
yep,
and
so
we
started
the
container
so
after
we
fixed
that
up,
we
pulled
the
right
image
and
it's
running
on
the
container
now,
so
that's
using
that
persistent
volume
claim.
So
this
is
just
another
example
of
uses
of
post-process
containers
and
things
kind
of
just
working
with
the
design
that
we
have.