►
From YouTube: Kubernetes SIG Windows 20220412
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
Okay,
yep.
We
were
just
discussing
before
the
recording
started,
not
really
anything
on
the
agenda.
I
believe
that
21.24,
zero,
beta
or
rc.0
was
released
this
week.
A
Yes,
now
is
a
good
time
to
start
doing
testing
if
you're
planning
on
you
know
adopting
124
release
when
it
gets
released,
shouldn't
be
that
many
changes
or
any
changes
between
now
and
the
the
final
build.
A
That's
all.
I
think
I
have
four
announcements
today.
Let
me
check
the
contributor
or
the
the
participants
list
yeah.
As
always,
if
there's
anybody
who's
new
to
contributing
and
wants
to
introduce
yourselves
or
say,
hi
get
some
get
some
help
now's
the
time
to
do
that,
but
I
I
think
I've
seen
everybody
on
the
participants
list
multiple
times
here,
so
we
can
skip
that
and
then
we
can
go
into
agenda
items.
I
don't
have
much
to
talk
about.
A
I
think
one
thing
that
we'll
note
is,
I
think,
last
time
we
talked
about
a
pr
to
improve
the
performance
on
q,
proxy,
oh
policy
or
yeah
policy
syncing.
We
were
hoping
to
get
that
in
124.
That
was,
I
actually
filed
an
exception
for
that
that
got
denied.
So
we
will
work
on
merging
that
for
124.1
and
then
back
for
it
as
appropriate.
A
Oh
I'll,
add
it
to
the
agenda.
After
the
fact,
david
was
the
one
who
was
working
on
that,
so
that
that
that's
a
pretty
significant
time
savings
for
network
to
become
ready
after
a
new
node
comes
online.
If
there's
a
large
number
of
services
in
the
cluster
arvind,
do
you
want
to
give
us
a
quick
status
update
on
that.
B
Yeah,
so
the
no
log
viewer
will
not
make
124
and
we're
targeting
it
now
for
125..
The
main
issue
was
with
api
review.
It
looks
like
they
want
the
shape
of
the
api
itself
to
change
tim
hawkin
was
not
really
happy
with
specifying
a
path
and
a
service.
B
He
does
not
want
the
user
to
even
care
about
stuff
like
that,
but
rather
have
a
generate
query
input
and
you
can
give
the
name
of
a
service
as
a
query
or
a
file
name
as
a
query,
and
the
server
needs
to
be
smart
enough
to
figure
that
out
and
and
make
that
right.
You
know
sort
of
call
as
to
whether
to
look
at
a
journal
or
just
you
know,
return
a
file.
B
B
We
need
to
do
this
encoding
and
decoding
into
the
api
struct
itself.
So
that's
the
other
thing
that
came
about.
I
think
I
got
a
review
from
jordan
leggett
about
that
yesterday.
So,
given
these
kind
of
changes
that
we
need
to
make,
this
doesn't
make
much
sense
for
us
to
try
and
get
an
exception
on
this.
Given
there's,
I
think,
a
little
bit
more
work
that
needs
to
go
in.
So
that's
the
update
on
that.
Let
me
know
if
anybody
has
any
questions.
A
I
it
seems
like
they
keep
moving
the
goal
post
a
little
bit
here,
unfortunately,
but
I
think
we're
in
should
be
in
pretty
good
shape
to
merge
this
very
early
in
125..
Hopefully,.
A
And
yet
yeah,
given
that
I
think
this
the
exception
for
this
q
proxy
fixes
was
rejected
with
like
out
of
concern
of
stability,
and
this
was
testing
only
windows
code
paths.
I
I
think
that
the
chances
of
having
an
an
enhancement
that
touches
the
api,
surface,
cubelet
and
cubesee
tail
will
probably
have
very
very
little
if
no
chance,
if
any
chance
of
being
admitted
here
too
so
the
best
we
can
do
is
yep
merge
it
as
early
as
we
can
once
the
range
opens
for
new
check-ins
cool.
A
And
then
yeah
yeah
okay,
so
I
found
the
pr
here
anybody
who's
interested
there's
a
lot
of
details
on
on
that
on
some
of
the
time
savings
that
we're
able
to
have
with
these
changes.
And
then
there
are
some
additional
time
savings
with
some
additional
os
updates,
which
will
be
coming
out
in
the
next
couple
of
months.
A
And
the
link
to
pr
jamie
was
this:
you
who
added
an
update
on
the
gmsa
charts,
yeah.
C
So
james
did
a
little
bit
of
help,
probably
a
week
and
a
half
two
weeks
ago,
and
I
finally
been
able
to
pull
down
his
changes.
I
rebased
off
the
latest
main
changes
and
all
the
unit
tests,
but
the
last
one
all
the
integration
tests,
but
the
actual
chart
deployment
test
passed
and
if
the
chart
deployment
test
fails
because
of
a
go
mod
that
works
on
the
other
eight
integration
tests
that
do
the
same
go
mod
command.
C
So
once
that
gets
resolved,
I
think
we
may
be
close
to
being
able
to
get
that
merged
in.
I
did
also
update
to
use
the
0.4
image
0.4.0,
the
newest
one.
So
all
those
updates
are
in
place,
so
it's
getting
close
and
I'm
also
getting
a
few
pings
and
like
reports
from
people
actually
using
it
already.
So
that's
been
some
good
feedback
to
get
incorporated
in.
A
Okay,
great
yeah,
I
can
maybe
take
a
look
at
some
of
the
go
mod
issues.
It's.
C
The
it's
the
same
exact
command,
it's
just
failing
to
pull
on
the
last
one.
I
have
no
clue
it
worked
seven
or
eight
times
before
so
it's
it's
just
have
to
run
it
locally
and
try
to
debug
it.
A
A
Yeah
thanks:
does
anybody
have
any
questions
for
jamie
about
that
or.
A
All
right,
if
not
jay,
see
you
have
some
more
some
questions.
D
Hey
so
right
now
for
the
kernel
space
proxy,
you
have
to
create
the
hns
network
and
the
net
firewall
rule
outside
of
coup
proxy
before
it
starts
on
windows.
So
so
I've
just
have
my
own
automation.
That
does
this,
but
like
normally,
I
think
what
people
do
is
they
make
these
things
on
their
own?
D
D
Like
you
tell
it
whether
you
want
to
do
an
overlay,
whether
it's
what
what
type
of
hns
network
and
whether
it's
an
overlay
or
not,
so
I'm
failing
to
see
whether
there's
a
good
reason
why
we
couldn't
just
sort
of
create
a
network
firewall
rule
in
the
coup
proxy
when
it
runs
and
an
hns
network.
And
then
you
could
run
it
like.
You
wouldn't
need
your
cni
provider
to
do
that.
For
you.
D
C
D
C
It
makes
the
aws
metadata
call
and,
of
course,
that's
kind
of
going
to
fail
until
you
re-add
the
route
back
and
that's
because
it's
going
through
like
a
proxy's,
not
the
right
way
to
describe
it,
but
it's
basically
proxying
it
through
its
network
to
call
out
through
the
local
host.
But
then,
when
you
get
your
calico
network
up,
you
can
it'll
start
actually
getting
the
it'll.
Add
the
rule
back
in
to
be
able
to
call
the
metadata
api
in
whatever
cloud
provider
you're
using.
C
D
C
D
D
Yeah,
so
I
don't
yeah,
I
so
what
I
do.
It's
two
calls
and
I
just
still
literally
stole
the
calls
from
calico,
and
I
I
literally
do
the
two
things
they
did
and
it's
really
funny,
because
if
you
look
at
the
calico
bootstrap,
actually
the
port
4789
is
hard-coded,
and
so
is
the
address
prefix
and
the
gateway
for
the
hns
network.
So
I
don't
even
know
how
it
works.
D
It
is
network
thing,
so
I
feel
like
there's
like
a
if
if
we
use
ipvs
as
a
archetype,
then
the
idea
would
be.
We
should
have
our
own
default
logic,
for
if
nothing
is
provided,
we
create
a
firewall
rule
and
an
hns
network
that
starts
the
coup
proxy
in
kernel
space
that
that's
kind
of
where
it's
going
so
jamie
sounds
like
he's
not
against
the
idea
and
yeah.
Okay.
E
One
quick
question:
do
we
have
all
the
information
available
to
create
the
hns
network
like
we
need
to
know
the
pod,
twitter
and
the
subnet
range
to
use
for
the
hns
network
and
then
in
the
l2
bridge
case
after
we
created
the
hns
network?
We
also
have
to
create
a
bridge
endpoint
and
other
things.
So
would
we
also
do
all
of
that
as
well
in
q,
proxy.
D
Yeah
I
mean
I
didn't
talk
about
the
l2
bridge
case
because
that's
the
other
one
right,
there's
overlay
and
then
there's
l2
bridge
and
those
are
separate
right
right.
So
there's
a
whole
conversation
about
like
yeah
like
if
the
coop
proxy
is
going
to
do
this.
We
need
to
plumb
a
few
parameters
into
it.
Like
I
mentioned
like
right
now,
when
I
do
it
it's
hard,
I
hard
code,
these
things
and
I
don't
even
know
how
why
we
do
192..
C
D
C
C
C
D
C
Calico
does
their
thing
we
have
psyllium
coming
or
calico
with
ebpf
like
it
would
be
nice
if
there
was
a
standard
way
for
for
them
to
be
able
to
to
like
not
have
to
boot,
like
you
could
bootstrap
a
cluster
without
seeing
eye
being
active,
but
then
it
can
have
the
hooks
like
what
you're
proposing
here.
I
think
just
me
throwing
it.
D
Here's
what
I'll
do
I'll
file
an
issue
upstream
and
then
I'll
tag
david
and
jamie
on
it
and
I'll
write
up
what
we
sort
of
talked
about
here,
and
we
can
talk
about
adding
this
in,
at
least
in
kp
g.
D
C
C
C
So
that
way,
because
it
doesn't
know
the
state
of
h
s
networks,
so
I'm
probably
getting
way
too
often
the
weeds
here.
But
I
was
just
thinking
if
q
proxy
is
going
to
handle
some
of
that.
Then
we
want
to
maintain
hms
networks
and
not
blow
them
all
away
right,
so
yeah
yeah.
A
C
A
C
C
That's
interesting
that
you're
saying
that
does
it
I
mean
it's
a
common,
it's
common
enough
that
people
do
that.
I
don't
know
if
flannel
recommends
doing
that
or
not,
but
I'm
sure
it
probably
does
it's
just
it's
it's
because
the
unknown
state
right,
which
I
would
I'd
be
concerned.
If
you
implemented
an
hns
network
for
kaping
and
then
you
booted
calico,
it
would
delete
it
right.
So
I'd
have
to
check
and
recreate
it
like.