►
From YouTube: Kubernetes SIG Windows 20171003
Description
Kubernetes SIG Windows 20171003
A
Hi
everybody
and
welcome
to
another
sip
windows
meetup
how
everybody's
doing
well
today.
I'll
kick
this
off
by
giving
a
few
updates
on
the
overall
Sagan
the
work
that
that's
being
done
across
the
board
between
the
cloud
base,
aprender,
Microsoft
and
and
then
we'll
go
to
individual
contributors
that
are
actually
doing
the
work
and
give
us
a
little
bit
more
of
a
detailed
update.
So
logger
mention
our
last
meeting,
who
have
a
couple
of
different
fronts
were
making
investments.
So
we
have
a
coupie
DM.
A
They
were
working
on
to
basically
produce
an
installer
that
will
allow
us
to
install
Windows
node
to
acclimatise
cluster.
We
have
Microsoft
that
has
merged
the
CRI
stats
and
and
and
the
C
advisor
replacement
for
Windows.
So
I
was
great
news
that
that
the
pr
went
through
yesterday,
the
last
PR
and
then
we'll
also
have
microsoft.
A
That's
working
on
the
cni
and
and
the
and
the
improvements
to
take
advantage
of
the
rs3
build
of
Windows,
Server,
2016
and
Jason
will
give
us
an
update
in
a
little
bit
and
then
the
cloud
base
and
the
ouvea
novias
community
are
working
on
their
own
CNI
solutions.
On
top
of
on
top
of
Paris
tree,
as
well
as
a
separate
CNI,
without
rs3
being
required
to
create
an
overlay
network
on
kubernetes,
so
let's,
let's
dive
into
into
the
update
so
Bob
you
wanna,
get
us
started.
B
Sure
I've
been
working
on
qtm
getting
it
so
that
you
can
add
a
Windows
node
into
the
toaster.
I
did
get
a
new
window
snow
to
be
added
to
the
cluster,
using
cube,
ATM
fight
and
bypassing
the
acceleration
sacks,
the
pre-flight
checks.
So
what
I'm
working
on
now?
It's
just
getting
window
specific
validation
checks
in
there
preflight
checks,
and
you
know
some
of
the
ones
that
would
be
common
for
for
both
just
making
sure
that
we
do
have
validation
checks
and
go.
A
That's
XLS
crate,
brought
progress
and,
by
the
way
to
everyone
else,
Cuba
DMUs
is
basically
going
to
also
become
the
the
foundation
for
us
to
start,
creating
some
automation
around
sick
windows.
So
once
Bob
is
done
with
this
work,
not
only
our
customers
gonna
be
have
an
easier
path
to
adding
Windows
notes
to
their
cluster,
while
leveraging
all
the
other
work
that
we're
doing
around
networking
and
compute,
but
also
as
we
are
getting
an
automated
environment
up
and
running
cube
ADM
is
how
can
I
get
things
started
fairly
quickly?
A
C
So
we
have
Windows
Server
version
1709
internally,
it's
called
rs3
that
is
just
about
to
be
released
publicly.
There
have
been
several
insider
preview
builds
already
and
the
new
platform
features
we
introduced
were
multiple
containers
per
pod
support
from
a
network
named
these
perspectives.
So
when
you're
running
Windows
Server
containers
without
hyper-v
isolation,
you
can
add
multiple
containers
into
a
single
compartment
for
namespaces.
We
call
it
that's
one
feature
the
other
one
is
we've
done
work
to
reduce
the
number
of
endpoints
that
are
actually
required
on
the
windows,
containers
compact
cluster
itself
and
slow.
C
We've
done
some
additional
work,
also
for
load
balancing
and
so
just
like
we
have
in
the
Linux
world
in
queue
products
you
can
use
load,
balancing,
IP
table
or
user
space.
We
basically
have
something:
that's
one
that
uses
our
virtual
filtering
platform,
VFP
extension
to
do
that
load,
balancing
and
curl,
and
so
those
are
the
things
we
support
today
on
our
site.
Internally,
we've
been
working
to
bring
up
clusters
mostly
manually,
sometimes
we'll
open
shift.
C
We'd
love
to
do
it
with
cube
ATM
as
well
as
mini
cube,
potentially
I,
don't
know
if
anybody
has
tried
that
yet
but
that'd
be
something
we
would
be
interested
in
well
in
contributing
to,
but
most
of
our
process
right
now
is
manual
with
spritz
just
lines
of
paper
guidelines.
We
do
have
a
scene
I
plug
in
that
Michael
mentioned
before,
and
that's
targeting
our
host
networking
services
all
to
bridge
mode,
so
this
is
actually
doing
it,
so
hopefully
l3
routed
sub
desk.
So
no
overlay,
like
the
ovn
solution,
does,
is
just.
A
A
A
A
We
are
gonna,
have
the
cube,
Indian
work
that
they're
working
on
and
then
as
well
as
the
the
cloud
base
CNI
for
Aviano
vs,
the
two
solutions
that
they
have
and
I
think
not
only
are
we
gonna
be
able
to
go
to
beta
with
1.9
or
even
earlier,
we're
probably
gonna,
be
able
to
host
production
level
workloads
on
on
kubernetes
with
windows
think
that
we
are.
We
are
at
that
spot
to
able
to
to
support
that.
A
Like
you
know,
after
that,
things
are
gonna
come
in
are
gonna,
be
you
know,
kind
of
bookkeeping
staff
like
automation
on
and
maybe
supporting
more
types
of
apps,
but
at
the
high
level
will
be
able
to
support
to
enterprise
apps
on
these
Jason
I
don't
know.
Can
we
talk
a
little
bit
about
the
performance?
Improvements
of
Windows
Server
containers
on
artistry
is
some
of
the
public
knowledge.
C
Not
really
I
think
we're
doing
we're
still
doing
kind
of
analysis
on
the
performance.
At
this
point,
we
do
have
plans
more
baselines
of
metrics
as
far
as
through
latency
is
concerned,
and
then
also
to
improve
density,
so
that
we
can
single
up
Windows
Server
containers
with
constants
part
of
time,
basically
up
to
a
hundred.
So
those
are
kind
of
goals
that
we
have,
at
least
for
the
yep.
A
A
It's
gonna
be
a
lot
faster
to
start
Windows
containers,
they're
gonna,
have
as
much
smaller
footprint
on
the
server
OS
and
things
are
getting
better
and
better
I
know
we
actually
did
have
one
customer-
and
this
is
you
know
this-
is
that
came
through
Gardner
that
complained
that
Windows
Server
containers
on
kubernetes
were
consuming
a
lot
of
resources
on
the
machine.
I
couldn't
get
any
more
details
about
that
Gardner
protected
identity
of
the
customer,
and
you
know,
since
he
chose
not
to
contact
us
directly.
We
have
no
visibility
into
that.
A
We
have
yet
aprender
have
not
personally
seen
that
and
I've
asked
a
couple
of
others
in
the
community
and
they
haven't
seen
any
similar
issue,
but
you
guys
be
on
the
lookout,
the
vision
issue
or
Windows
containers
on
kubernetes
or
thing
up
to
many
resources
where
there
are
like
networking,
resources,
compute
or
storage.
Let
us
know
we'll
have
to
investigate
that
and
figure
out.
If
there's
a
fix
to
be
made.
I
know
for
sure.
There's
three
would
be
a
lot
better
from
a
performance
standpoint,
but
not
everybody
will
great
to
that.
E
A
G
A
So
I
think
the
kubernetes
model
is
I.
Do
have
to
go
from
alpha
to
beta
I,
don't
know
if
they're
gonna
allow
us
to
skip
the
GAE
to
be
honest
with
you.
If
they
do,
we
would
have
a
serious
thought
about
considering
making
EGA
if
they
don't
allow
us,
then
we'll,
because
the
beta
and
then
with
GA
immediately
after
that
I
think
things
are
left
that
we're
not
gonna
do
for
beta.
A
Just
to
give
you
everyone
an
update,
if
you
guys
go
to
our
channel,
there's
also
a
paint
milestone
schedule
that
include
some
of
these
things.
So
so
there's
gonna
be
some
work
around
the
Installer
right
everything
you
know
the
I
don't
know
if
we're
gonna
have
a
strip
line
solution
around
CNI,
qvd
em
and
all
these
other
things.
So
that
might
be
some
work
that
we're
gonna
need
to
do.
Then
upgrades
is
something
that
we
haven't
touched
at
all
so
far
since
been
concentrated
with
our
initial
experience,
haven't
tried
to
see.
A
What's
gonna
happen
during
an
upgrade,
whether
that's
an
application
upgrade
that's
running
on
Windows
Server
containers
or
an
upgrade
of
kubernetes
itself
right,
the
cluster.
So
we
haven't
done
any
testing
on
that.
It
might
all
work
you
might
not.
Basically
what
I'm
asserting
is
that
do
haven't
touched
that
so
we
don't
know
we
want,
and
we
won't
know
until
until
we
get
to
that
point
and
then
basically
adding
all
of
the
automation
I
mentioned
earlier,
doing,
having
regression
runs
around
with
kubernetes
or
if
signals,
for
example,
changes
something
that
breaks
us.
A
We
gonna
know
about
it
before
we
release
right
and
then
the
last
part
that
we're
not
gonna
do,
and
this
community
thing
is
validate
that
all
of
our
work
and
installer
steps
work
across
every
provider
right.
Well,
that's
a
public
cloud
provider
like
as
your
AWS
Google
cloud
as
well
as
on-premise
installations.
So
so
maybe
you
will
have
the
documentation
and
have
it
ready
by
by
December
November
December
timeframe,
and
maybe
not
we
at
aprender.
Definitely
don't
have
the
resources
for
that.
A
So
we're
gonna
be
looking
at
folks
like
Anthony
how
and
his
team
to
do
that
for
a
double
for
ACS
and
annasher.
Then
you
know
we
have
a
few
Google
folks
on
a
community
that
maybe
about
before
go
cloud.
I,
don't
know
if
anybody's
with
AWS,
who
have
that
it's
not
initially,
but
in
general,
these
are
the
things
that
kind
of
produce
a
fully
available
available
in
solution,
right,
testing,
automation,
good
documentation
across
the
board
and
and
and
and
our
ability
to
catch
regressions.
G
Probably
be
in
beta
for
a
little
bit,
that
was
the
other
thing
that
I
was
gonna.
Ask
Mike
is:
have
we
been
able
to
spend
any
time
on
on
the
test,
automation
so
that
ensure
that
there's
no
regressions
going
forward
or
I
know
there?
We
were
starting
a
little
bit
on
that,
but
I,
don't
know
how
we
were
able
to
no.
A
Absolutely
not
time
on
that
so
far,
okay,
yeah!
So
that's
something
that
we're
gonna
have
to
look
at.
You
know,
after
we're
done
with
our
in
with
our
current
work
like
when
you
look
at
our
timeline
right.
So
so
we
did
all
of
the
work
around
secrets
from
vic
mobs.
We've
made
sure
that
when
one
statistic
comes
out
to
be
able
to
test
multiple
containers
in
a
part
and
the
cni
work
that
basically
completes
all
of
the
networking
and
compute,
then
the
monitoring
solution
with
the
CRI.
A
We
haven't
test
that
now
that
it's
in
we'll
probably
have
to
validate
it
that
end-to-end
and
that
basically
the
next
thing
that's
left
is
automation,
automation
and
documentation.
So
we
all
have
a
lot
of
resources,
so
I
can't
really
commit
to
a
date,
but
you
know
we'll
keep
working
at
it
and
see
how
far
along
we
get
by
by
November
December
timeframe.
A
H
A
D
A
D
C
A
I
Have
one
question:
my
name
is
teri
I'm
coming
from
far
away,
so
we
are
trying
to
build
the
kinetics
cluster
so
support
the
windows
container
in
a
de
blasio,
a
cloud
so
I
have
two
general
questions.
The
first
is
if
we
do
it
on
the
ADA.
Yes,
we
know
that
we
can
use
the
ODN
obs,
but
beside
the
OB
n
es
today.
What
kind
of
other
network
solutions
can
we
use?
And
the
second
question
is
so
that's
a
general
talk.
I
That's
general
windows
container
questions,
so
we
know
the
Windows
Server
2016,
starting
from
the
2016
start
to
support
the
container,
but
can
I
but
many
of
our
customers,
a
their
Windows
application,
runs
on
the
older
version
like
2008
can't
build
a
container
image
that
the
base
which,
with
the
base
image,
is
the
Windows
Server
2008?
Is
that
a
skull
so.
A
I'll
try
to
answer
all
your
questions.
I'll
start
from
the
end,
which
is
ageless.
The
operating
system
of
the
Windows
Server
container
has
to
be
either
16
or
later.
There
is
no
support
for
earlier
OSS,
and
the
two
flavors
of
Windows
are
going
to
be
nano
server
in
the
server
core.
Now,
even
though
a
customer
has
an
application
that
used
to
be
running
on
an
on.
A
I
I
C
C
I
A
Wanna
had
a
muted
Cindy
said
about
that
so
with
today,
if
you
use
alpha
they're
only
networking
solution,
new
hobby,
so
Vienna.
Yes,
when
you
go
to
bed-
or
you
know
in
in
a
few
weeks,
essentially
to
you
be
able
to
to
start
testing
the
solution.
There
will
be
three
separate
networking
options
for
you.
A
The
first
one
would
be
if
you
don't
want
to
use
all
Vienna,
vs
and
you're,
using
the
rs3,
build
of
Windows
Server
16
you'll
be
able
to
use
the
building,
CNI
plug
in
the
Microsoft
will
shape
and
that's
basically,
the
layer
to
bridge
bridge
the
ni
network.
The
second
option
would
be
cloud
base
is
working
on
to
see
ni
plugins.
One
of
them
will
require
rs3
and
will
give
you
an
overlay
network
that
will
include
multi
multiple
as
per
pod.
A
So
so
it's
rather
networks
and
flat
layer,
two
for
Microsoft
or
overlay
networks,
with
or
without
rs3,
depending
how
many
containers
per
pod
you
have,
we
will
occupant
all
of
these.
That's
you
know.
A
big
part
of
what
I
was
saying
earlier
is
that,
in
order
for
us
to
be
able
to
give
this
to
customers,
you
have
to
document
all
the
options.
Different
installation
methods
cube
ADM.
All
of
these
things
before
we
can
move
forward
and
people
can
start
using
it.
I
A
C
D
So
I
basically
have
some
internal
documentation
that
I
shared,
but
I
listed
in
the
Sigma
does
think
so.
We
have
come
up
with
a
documentation
of
how
to
bring
up
on-prem
l2
bridge
solution.
The
two
things
that
I
want
to
call
out
as
one
is
both
Azure
Amazon
and
Google
Cloud
needs
to
release
our
s3
building
as
a
option
to
create
a
PM
in
the
cloud.
I.
Don't
know
when
that's
gonna
happen.
So
once
that
happens,
you
can
use
AWS
or
Google
or
as
you
to
create
this
l2
bridge
in
a
cloud
and
Raman.
D
So
until
then,
I
have
shared
a
document
which
gives
you
step
by
step
instruction
on
how
to
bring
up
the
l2
bridge
solution.
But,
as
Michael
said,
it's
not
it's
more
manual
steps
and
it's
not
integrated
with
qadian
or
anything.
It's
just
more
of
a
manual
way
of
how
we
are
bringing
in
and
testing
the
changes
appear.
Thank.
I
D
I
C
D
I
have
not
tried
an
ADA
blazer
GC,
but
I've
done
that
and
the
enrollment
so
I
say
again
I'm
trying
to
use
as
your
existing
beams
to
bring
in
some
tested.
I,
don't
know
if
the
support
statement
around
it-
I
don't
know.
So
you
have
to
talk
to
somebody
from
Azure
or
Google
or
Amazon.
Okay,
I've
done
that
manually
and
it
works.
D
A
Can't
offer
you
the
finiti
of
answer
but
they're
licensing,
the
supply
licensing
that
you're
paying
on
all
the
cloud
providers
where
there's
ws
as
your
Google
is
dependent
on
the
OS
version
you
have
so
once
you
actually
go
to
an
OS
version
that
they
no
longer
support
your
per
hour.
Licensing
that
you're
paying
for
supply
is
no
longer
valid,
so
I
would
see
that
a
public
cloud
provider
like
one
of
those
three
would
plot
you
in
that
case,
yeah
I,
guess
No.
A
So
the
lies,
because
you're
not
bringing
your
own
license
on
a
cloud
provider
right
when
you're
getting
Windows
Server
16
you're
being
charged
on
this
fly
licensing
fee
for
Windows
Server.
So
as
soon
as
you
go
upgrade
your
own
they're
gonna
figure
that
out
and
and
there
they
are-
they
are
gonna,
be
in
direct
violation.
Microsoft
at
that
point,
from
a
licensing
perspective,
so
I
could
see
why
they
would
want
to
stop
you
immediately
from
doing
that.
So
so
so.
H
A
C
Well,
you
can't
I'll
mention
one
other
thing.
What
you
can
do
is
if
you
do
have
a
VHD
image,
that's
how
you
read
it
that
has
Windows
Server
version,
7
or
9.
You
can't
upload
that
to
about
subscription
that
you
create
a
custom
yeah.
So
that's
what
I've
done
in
the
past
as
well.
So
that's
an
option.
Until
the
publicly
available
images
are
out
there.
J
A
So
there's
a
lot
of
background
noise.
If
someone
is
not
speaking,
can
you
guys
please
mute?
So
you
know
different
applications
have
different
requirements.
I,
don't
think
there
is
a
single
solution
for
everything,
but
if
you
have
an
app
that's
on
a
previous
version,
of.net
framework
there's
a
good
chance
was
the
framework
is
backwards
compatible
that
you
can
bring
that
application
to
a
newer
operating
system
within
your
version,
of.net
and
still
have
it
supported.
A
You
know,
I
can
advise
you
guys
more
than
that.
There
are
companies
that
do
this,
for
a
living
that
could
evaluate
workloads
and
on
and
basically
help
you
modernize
them.
All
I
can
say
is
that
you
know,
because
both
OS,
as
well
as
don't
have
frameworks,
are
backwards,
compatible.
There's
a
good
chance,
you'll
be
able
to
modernize
the
application
into
Windows
server
16
with
dotnet,
4.6
or
4.7.
A
J
A
Could
be
a
lift
and
shift
in
in
some
cases
they
might
have
to
rewrite
part
of
it
like
without
evaluating
the
application.
There
is
really
no
way
to
know
if
you
want
I
can
give
you
some
names
of
companies
that
do
this,
where
they
have
consulting
arms
that
basically
help
customers,
modernize
applications.
I,
don't
know
what
the
fee
structure
is,
but
you
could
start
a
conversation
with
them.
Okay,.