►
From YouTube: Kubernetes SIG Windows 20210525
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
welcome
everybody.
This
is
sig
windows,
regular
weekly
meeting.
We
are
part
of
the
cncf
and
we
will
be
following
the
cncf
code
of
conduct.
A
If
you
have
any
questions
or
concerns,
you
can
reach
out
to
myself
or
any
other
leads
in
the
sig
windows
community
and
today's
may
25th
so
welcome
and
before
we
get
into
the
agenda
here,
we
typically
try
to
open
it
up
for
anyone
that
is
new
to
the
community
and
wants
to
introduce
themselves
and
say
hello,
and
maybe
tell
us
a
little
bit
about
what
you're
here
for
why
you're
interested
in
sig
windows,
anything
that
you
know
is
relevant
to
any
of
that.
A
Okay
and
we'll
go
ahead
and
get
started
on
the
announcements.
If
you
have
anything
you
want
to
discuss
for
the
regular
meeting,
you
can
just
add
it
to
the
end
of
the
list.
Here.
Add
it
to
the
agenda
and
we'll
make
sure
we
have
time
to
to
get
to
it
to
get
started.
I
wanted
to
for
announcements.
We
have
the
host
process,
pr
merged,
which
is
huge-
that's
been
in
work
since
I
think
february,
ish
and
even
longer,
with
the
kept
and
doing
all
the
planning
for
that.
A
So
congratulations
to
everybody
that
was
involved
in
that
that
this
means
it'll
be
in
alpha
4
122,
when
it
releases
and
as
of
right
now
any
container
d
component
should
sorry.
You
can
use
any
container
d
version
because
we
passed
the
annotations
down
through,
but
you
do
need
to
use
this
particular
pr.
That's
linked
here
for
hcs
shim
and
you
have
to
build
the
container
d
shim
from
that
pr.
A
Currently,
we're
get
working
to
get
that
merged
once
that's
merged,
we'll
bump
the
version
in
container
d,
and
we
should
be
able
to
build.
You
should
be
able
to
use
the
released
versions
of
container
d
with
this
there.
Any
questions
or
comments.
B
I
guess
we
need
to
probably
test
that
out,
so
I
might
be
looking
for
volunteers
at
some
point
once
I've
once
I've
tested
it
and
rebased
it
and
done
everything
myself,
but
it
basically
just
means,
hopefully
with
that
pr
means
that
we
should
be
able
to
spin
up
cni
pods,
because
at
the
moment
I
don't
think
it
does.
So
it
seemed
to
be
okay,
but
I
I
didn't
manage
to
get
it
fully
tested
with
with
the
annotations
after
they
changed.
B
So
we
just
need
to
revisit
that
to
make
sure
I
think
danny's
already
been
through
it,
but
yeah
there's
a
bit
more
to
do.
A
Cool
yeah,
so
that's
specifically
for
being
able
to
run
pods
as
a
host
network
and
and
start
them
up
before.
The
network's
set
up
right.
B
A
Yeah,
so
I've
started
the
work
to
get
a
the
test
running
in
cluster
apa
for
azure,
and
so
I'll
probably
be
testing
that
out
this
week,
oh
cool.
A
Cool
kubecon
videos
they
got
posted.
I
think
there's
one
other
windows
one,
but
I
wasn't
able
to
find
it.
I
don't
know
if
anybody
has
a
link
to
it,
but
this
one
was
the
one
done
by
kalya
and
david
and
mark
and
jay
particularly,
has
some
really
interesting.
Information
around
networking
for
windows
really
great
overview
of
how
how
all
that
works.
So
I
recommend
checking
it
out,
but
there's
also
a
bunch
of
other
cubecon
videos
out
there
as
well.
D
Yeah,
just
a
quick
shout
out
a
shout
out
to
kalia
and
david
and
mark
and
jay,
I
guess.
Last
yesterday
I
watched
the
video
this
video
on
during
the
lunch
break.
I
think
it's
really
insightful.
We
get
a
lot
of
questions
about
networking
and
they
probably
answered
most
of
them,
especially
the
architecture
diagrams
they
presented
was
very
helpful
for
showing
the
difference
between
linux
and
windows.
Networking
stack,
so
I
would
highly
encourage
folks
to
watch
it
or
like
if
people
have
questions
about
networking,
we
should
share
this
as
well.
A
Awesome
yeah
and
then
so
there
was
a
issue
found
recently
for
q
proxy.
Essentially
it
crashed
q
proxy
if
you
pass
the
wrong
load
balancers
and
this
cause.
This
can
cause
all
sorts
of
behavior,
but
you
might
see
resets
on
some
services
for
any
open
connections,
and
so
we've
backported
that
to
119,
120
and
121.,
make
sure
you
go
ahead
and
pick
that
up
it's
going
to
help
with
stability
of
your
services
across
the
port.
A
E
Not
from
the
networking
team,
but
I
can
give
you
I
I'm
the
one
from
the
bug
actually
so
great.
E
So
what
happens
is
if
you
have
a
a
load
balancer
and
you
if
you
have
pods
that
are
distributed
across
different
nodes
when
you
try
to
hit
that
endpoint
you'll
end
up
with
a
q
proxy
crash,
so
it
had
something
to
do
with
the
ref
counting
not
being
applied
appropriately,
and
I
think
stravan
from
microsoft
is
the
one
who
had
that
fix
and
it's
interesting.
E
B
Is
that
relevant
to
user
space
as
well?
As
I
can't
remember
the
other
name.
E
B
Okay,
I
think
jay
probably
knows
a
bit
more
about
than
I
do
so.
I
might,
I
might
point
him
at
this
direction
to
make
sure
that
he
knows
about
it
if
he
hasn't
picked
up
already.
A
All
right
and
so
moz,
I
think
you
were
talking
about
this
one
last
week
with
dr
shim.
D
Yeah,
I
added
this
to
agenda
quickly
as
a
reminder,
so
I
think
jing
and
I
from
microsoft
we
create
because
we've
heard
from
customers
and
different
folks
struggling
with
moving
to
to
continuity,
especially
even
if
they
move.
They
want
some
time
for
stabilization.
D
So
that's
what's
concerning
so
we
created
an
issue.
I
could
link
the
issue
here
and
we
posted
it
on
sick
windows
and
signaled
and
tagged
ems.
D
He
came
back
and
he
said
that
we
have
updated
the
cap
and
it
is,
it
will
be
removed
in
124.,
so
he
confirmed
that
you
know
it
won't
be
removed
before
124..
You
will
get
a
warning,
but
the
docker
sim
is
still
there
until
120.
124
is
the
first
release
where
docker
shim
won't
be
there
the
expected
date.
D
For
I
mean
if
we
could
look
at
the
history,
the
expected
date
for
124
is
somewhere
early
next
year,
so,
like
you
know,
february
or
march
2022,
so
that
gives
us
almost
one
year
for
for
stabilization
and
getting
confidence
before
they're
like
forced
to
stop
using
docker
and
continuity.
So
I
just
wanted
to
share
this
with
the
with
the
group,
and
I
hope
this
is
good
news
for
everyone.
A
D
Me
as
well
so
aravind
and
jake
tagged
me,
I
think.
Initially
there
was
a
question
posted
in
sick
windows.
People
wanted
to
contribute
to
sick
windows.
The
problem
is
becoming
windows,
server
license
how
you
know
they
probably
won't
want
to
buy
licenses,
and
we
agree.
So
I
have
started
internal
conversation
and
the
feedback
we
got
is
we
are.
We
will
try
to
work
on
the
developer
program
in
future
to
give
licenses
out
for
free,
but
until
then
the
there
are
a
couple
of
options.
D
I
think
I
presented
it's
in
the
notes
and
one
of
them
is
like
using
evaluation
windows
server.
It's
completely
fine
you'll,
get
it
for
180
days,
but
r1
brought
up
a
point
that
the
evaluation
version
for
windows
server
is
not
showing
for
sac
releases
like
2004
or
20h1
and
20h2.
D
So
I'm
going
back
just
confirming
that
you
know
I
looked
at
the
link.
Yes,
it's
not
showing
so
I'll,
go
back
to
the
licensing
team
and
see.
What's
the
detail
I'll
give
you
an
update
next
week.
E
And
a
little
bit
more
to
add
to
that
most
is
that
we
did
some
more
digging
and
we
found
that
link
that
I
sent.
Apparently
you
apparently
there's
no
eval,
because
it's
some
you
can
only
get
it
through.
What
is
the
verbiage
that
I've
found?
It's
yeah?
You
can
only
partner
yeah,
it's
available
to
volume,
licensed
customers
with
software
assurance
and
loyalty.
E
E
We're
not
able
to
find
it,
but
for
some
reason
it
was
not
listed
on
the
operating
system.
I
should
ask
the
person
who
found
out
where
exactly
they
discovered
it
yeah.
D
That
would
be
good
to
know
and
if
you
can
post
it,
I
could
go
back
to
the
msdmd
or
visual
studio
team
and
say
that
you
could
they
could
fix
the
tagging
or
something
okay,
because
I
didn't
even
know
I
think
from
msdn.
It
should
never
be
a
problem.
I
was
even
thinking
if
somebody
doesn't
have
msdn,
then
how
they
will
get
it.
That's
the
question.
I'm
asking
the
licensing
team:
okay,
yeah.
G
D
I
think
one
last
update
arab
wonder
if
you
want
to
talk
about
the
cap
of
node
logs,
getting
merged,
I'm
very
interested
in
that,
and
thank
you
for
driving
that
if
you
want
to
give
a
quick
update
there
are
you
planning
to
like
do
the
the
pr
for
122
yeah,
okay,.
E
I
think
I
had
mentioned
it
in
the
last
sigmundus
call
also
so
me,
and
christian
from
red
hat
are
currently
working
on
it.
Hopefully
we'll
we'll
get
it
in
at
the
moment,
I'm
working
on
adding
the
feature
flag
on
the
on
the
cubelet
side,
so
that
you
know
only
with
the
feature
flag.
Will
this
feature
be
enabled.
So
that's
exactly
what
I'm
working
on
in
my
spare
cycles.
D
Okay
sounds
good,
and
then
once
you
have
this,
I
we
are
like.
I
am
at
least
interested
in
in
one
of
the
cases
which
is
getting
gmsa
logs
from
the
from
the
events.
Log
windows
for
this,
so
we'll
probably
get
some
devs
we'll
try
to
get
some
devs
from
microsoft
site
or
from
the
gmsa
team
to
try
it
out
once
the
pr
is
in.
E
E
Pr
once
I
push
my
changes
and
it's
ready
for
at
least
trying
out
some
basic
stuff,
everything
might
not
work,
but
you
might
be
able
to
try
out
some
basic
stuff
because
I
don't
know
how
many
people
follow,
but
that
no
log
cap
has
become
more
involved.
Now
I
have
to
do
stuff
for
even
pod
logs.
E
This
heuristic
approach
that
you
know
tim
hawkins
wanted.
So
it's
a
little
bit
more
involved
so
to
get
everything
done
it's
going
to
take
a
while,
but
as
and
when,
like
basic
features,
start
working,
maybe
I'll
tag
you
more
once
and
you
can
try
it
out.
It'll
be
a
little
bit
more
involved
to
try
it
out,
though
you
have
to
compile
the
cubelet
and
keep
control.
A
Awesome
great
to
see
that
moving
forward
all
right
is
there
anything
else
that
anybody
wants
to
bring
up
ad
hoc
here.
B
I
did
wonder
sorry
go
ahead.
Perry
like
I
was
just
wondering
if
anyone
else
is
starting
to
see
this
there's
been
an
ongoing
sort
of
flight
on
and
off
error
message
that
keeps
coming
up
in
various
issues
about
the
access
denied
log
message
and
it
came
up
again
while
I
was
doing
testing
this
week
and
one
of
the
things
I
think
you
put
on
that
that
particular
issue
james,
was
about
cpu,
spiking
and
then
causing
it,
but
I
was,
it
seems
to
be,
it
seems
to
just
be
runtime.
B
Runtime
agnostic
seems
to
happen
on
docker
and
on
container
d.
I
think,
and
I
starting
to
wonder
whether
there's
something
in
case
that's
causing
it
in
terms
of
logging
or
something
like
that.
But
I
wondered
if
anyone
had
any
ideas
of
things
where
we
could
kind
of
start
trying
to
chase
it
down,
because
it
seems
to
be,
it
seems
to
just
be
ever
present
if
you've
got
a
lot
of
pods
starting
and
stopping
and
the
high
cpu
load
that
this
access
denied
log
error
happens
like
frequently
with
I
don't
know.
A
Yeah,
so
jeremy
was
able
to
reproduce
this,
and
I
was
trying
to
get
to
the
issue,
but
it's
gonna
make
me
do
my
two-factor
authentication
and
so
in,
but
I
think
we
have
a
pretty
consistent
repo
in
the
microsoft
windows
containers
repository
and
I
was
able
to.
He
was
able
to
reproduce
it
on
gce.
I
was
able
to
reproduce
it
on
azure,
but
I
haven't
had
a
chance
to
get
back
to
trying
it.
So
maybe
we
can
link
to
that
here
and
have
somebody
take
a
look
at
it.
B
I
did
have
a
theory,
but
it's
a
bit
of
a
wild
out.
There
theory
there's
a
mutex
lock
on
log
files
in
that
happens
regularly
where
it
basically
cycles
through,
gets
all
the
pod,
like
all
the
pods
inside
the
cubelet,
and
then
it
like
locks
the
log
file,
while
it's
truncating
it,
and
I
wondered
whether
that
was
doing
something
where
it
was
locking
the
file
out
and
then
not
unlocking
it.
If
there's
some
high
cpu
that
causes
some
failure
and
then
never
kind
of
recovers.
B
D
Barry
this
this
scenario
that
you
tried
I'll,
take
a
look
as
well,
but
if
it's
in
the
the
windows
container
repo,
but
was
this
like
a
new
thing
you
discovered
like
was
this
scenario
working
before
of
exiting
log
files?
We
did
something
changed
like
what's
the
change
here.
B
B
We
were
running
the
network,
so
jay
and
I
were
working
on
network
policy
tests
for
windows
and
it
spawns
a
lot
of
agn
host
pods
up
and
down.
Quite
quite
quite
a
fast
pace
right,
so
it
comes
up
and
like
tests
and
then
tears
them
down
seems
to
increase
cpu
load
quite
a
lot,
and
we
see
this
and
it
basically
just
crashes
out
the
test
pretty
reliably
within
a
few
minutes
of
starting
it.
B
It
sort
of
crashes
out,
but
it
seems
to
only
be
when
you're
putting
the
node
under
quite
a
lot
of
load
and
creating
and
tearing
down
pods
quite
frequently,
so
it
might
be
an
edge
case,
but
it's
still
something
that
seems
a
bit
strange
that
it's
just
locking
out
and
it's
for
us
it's
container
d,
but
I
think
I've
seen
the
message
on
a
lot
of
people
are
seeing
the
same
kind
of
message
on
docker
as
well
is
this:
is
this
the
error
message?
B
I
can't
see
your
screen.
All
I
can
see
is
sorry.
A
B
Yes
similar,
I
think
obviously
now
it'd
be
a
different
path
for
the
log
file,
but
it
is
access
denied.
We
don't
see
terminating
though
what
we
it
kind
of
it
kind
of
disappears.
Out
of
case
we
can
with
our
use
case,
so
you
case
will
forget
about
it.
But
then,
when
you
go
in
the
cubelet
logs,
it's
just
constantly
cycling
saying
this
pods
like
not
deleting.
C
And
then
this
this
here
is
the
the
repo
that
jeremy.
A
B
B
Yeah
they're
still
running,
and
you
can't
open
the
log
file
yourself
because
it
tells
you
access
is
denied,
but
the
pods
are
still
there.
In
the
background
you
say
if
you
do
see,
cricut
cuddle
or
ctr,
the
pods
are
still
there,
but
kubernetes
for
us
at
least
just
basically
just
forgot
about
them,
but
the
the
cubelet
log
shows
a
very,
very
different
story.
G
I
wonder
why
it
tries
to
remove
them
before
it
stops
down.
I
mean
why
so
there
might
be
something
missing
somewhere.
B
I
mean
if
you
want
a
reliable
reproducer,
just
run
those
netpol
tests
on
windows
that
we've,
I
think
I
put
on
a
branch-
and
I
put
mentioned
it
on
there,
but
it
will
quite
quickly
lock
you
like
the
test,
will
stop
and
then,
in
the
background
in
the
cubit
logs
you'll
just
see
this
container
log
manager,
failure
constantly
moaning
about
access
denied
and
the
pods
are
still
there,
but
kubrick
has
lost
them.
B
This
was
1.44,
but
it's
been
happening
on
like
1.43
as
well.
So
one
point
four
three,
four
four:
I
can
try
it
on
one
five,
I'm
you
know
at
this
point
in
time.
My
guess
is
that
it's
it's
run
time
agnostic,
because
I
don't
think.
F
A
So
I'm
not
finding
access
denied
in
this
particular.
A
Container
log
failed
to
create
and
open
the
log
which
is
different
than
the
issue
over
here
anyways.
We
can
maybe
follow
up
off
on
offline
and
see
if
you're,
seeing
either
one
of
these.
B
I
can
quite
quickly
reproduce
so
I'll
just
grab
it
when
we
and
then
update
that
within
the
latest
vlog
message
that
I
got,
but
it
always
seems
the
same
same
use
case.
We
just
get
this.
These
pods
left
behind
and
then
cube
lit
cycling.
F
D
F
H
A
A
All
right:
well,
it's
10
o'clock,
so
we'll
call
it
good
for
today
and
we'll
see
y'all
next
week,
we'll
see
y'all
in
slack.
If
you
have
any
questions,
I
think
we
also
do
sig
windows
after
hours
now
and
jay
will
leave
that
so
I'll
hand
it
over
to
him
I'll.
Stop
the
recording
here
and
we'll
hand
it
over
to
jay
to
lead
that
with
us.
So
if
you're
interested
in
doing
some
hands-on
going
through
some
pr's
and
other
work
like
that,
stick
around
thanks
see
y'all.