►
From YouTube: Kubernetes SIG Windows 20181030
Description
Kubernetes SIG Windows 20181030
A
Hello,
everybody
and
welcome
to
another
say
Windows
meetup
I.
Thank
you
all
for
attending.
We
actually
have
a
pretty
good
attendance
today.
Lots
of
folks
are
here
just
a
couple
of
bit
of
news,
so
I
just
started
at
VMware,
so
you
know
you
guys
have
noticed.
Prenda
is
no
longer
a
company,
so
I'm
now
with
VMware,
and
one
of
the
things
that
I'll
be
overseeing
is
also
the
windows
server
containers
on
pivotal
and
PKS.
A
So,
while
my
goal
is
to
start
introducing
a
couple
more
open-source
developers
from
VMware
and
people
into
the
project
that
can
help
us
go
to
GA,
since
our
goal
is
to
GA
in
pretty
much
months
from
now.
So
me
and
Patrick
are
going
to
talk
about
resources
of
line,
as
was
the
tail
end
of
this
week
and
maybe
update
you
guys,
the
next
Meetup,
that's
it
for
me.
Now
I
have
resources
again
for
from
my
own
company,
so
we
will
have
us
on
this
progress
and
I'm
excited.
A
Absolutely
I
guess
someone
complained
that
it
was
merged
automatically
after
I
approved
it
I,
don't
know
what
happened
with
that,
but
essentially
there's
a
sick
conformance
that
basically
was
formed
to
kind
of
start
hopping,
define
what
the
goals
in
charge
era
are
for
each
one
of
the
different
things
being
part
of
communities
and
our
Charter
is
kind
of
very,
very
cut
and
dry.
Our
goal
is
to
make
sure
that
we
can
support
containers
on
the
windows
ecosystem.
A
We
make
sure
that
all
the
new
features,
functionalities
and
bells
and
whistles
are
added
on
kubernetes
in
a
work
on
Windows
or
we
call
it
out
when
they
don't
work
and
essentially
make
sure
that
we
test
and
we
take
the
conformance
test
and
everything
else
are
happening
around
the
test
fit
for
kubernetes
and
make
them
work
for
Windows
with
him
within
our
platform
in
hybrid
environment
of
Windows
limits.
So,
given
that
charter,
that's
kind
of
our
our
guiding
principle
as
well,
going
through
GA
with
1.13.
B
Yeah
so
a
couple
things
so
basically
that
the
Charter
that
that's
on
there
basically
is.
You
is
saying
you
know,
none
of
the
governance
is
Divi
deviating
from
you
know
the
regular
defaults
for
the
community,
and
then
it
just
calls
out
the
two
bullet
points
of
Windows
specific
code
in
all
parts
of
the
database
is
being
or
so
at
the
code
base
as
being
a
responsibility
as
well
as
testing
of
Windows
specific
features
and
cluster
clusters.
So
I
think
it's
pretty
clear
to
me.
A
C
B
All
right
so
I've
got
links
in
there.
I
guess
'sir.
The
one
one
question
I
want
to
be
thinking
about
is
I
know
that
the
sig
was
initially
formed
to
handle
the
Windows
specific
stuff.
But
you
know
we
were
basically
formed
out
of
sig,
node
and
I.
Don't
know
if
this
is
something
that
would
go
into
the
Charter
or
if
this
is
a
separate
discussion,
but
you
know
is
there
some
point
at
which
we
should
consider
just
becoming
part
of
sig?
Note
again.
Is
that
something
that
that
you've,
given
a
thought
to.
A
I'll
tell
you
my
personal
opinion:
I
think
the
signals
meetings
are
fairly
polluted,
with
a
lot
of
very
specific
things
that
signaled
cares
about,
and
they
basically
take
a
hundred
percent
that
time
so
I
think
that
would
be
a
fly
on
the
wall.
If
we
were
part
of
those
meetings-
and
maybe
we
are
not
able
to
get
enough
face
time
to
raise
our
agenda
and
talk
about
our
things,
noting
that,
given
that
we
have
our
own
charter
and
maybe
our
own
set
of
responsibilities,
it
makes
sense
for
us
to
still
stay
out
of
signal.
A
We
also
span.
You
know
between
signal
signal
works,
a
cluster
Rob's
who
have
a
lot
of
other
SIG's,
that
we
span
responsibility
on
and
make
sure
we
need
to
interface
with
that
it
might
make
sense
to
keep
our
own
our
own
agenda,
but
mobile.
If
everybody
wants
to
move
to
signal,
you
can
do
that.
It's
just
I
feel
like
different
if
I'd
imagined
that
different
different
objectives
within
that
sig,
okay.
B
C
B
B
So
I
think
that
as
there's
new
people
coming
onto
the
project,
that
would
be
a
great
place
for
them
to
help
sort
of
look
at.
What's
there
and
said
you
know
it
does
this
make
sense,
because
here
for
those
of
us
I've
been
looking
at
this
for
a
long
time,
you
know
there's
probably
steps
that
I
just
do
automatically
and
without
even
thinking
about
it,
so
getting
fresh
eyes
on
that
would
be
helpful.
B
C
Cluster,
a
hybrid
cluster
on
GCE,
so
anyone
who's
out
there
running
on
Google.
If
you
have
a
Google
cloud
project,
then
you
should
be
able
to
follow
these
steps
and
bring
up
a
cluster
just
by
running
a
few
commands.
It's
based
on
the
cube
up
script,
which
is
under
the
cluster
director
cluster
subdirectory
and
the
kubernetes
repository.
C
So
we
are
using
this
as
the
basis
for
some
automated
testing
now
on
GCE,
which
you
Joe
and
I,
are
working
on.
Presently
I've
have
the
test
running
with
the
exclusion
list
that
Adelina
published
and
they
take
about
three
hours
right
now,
I
haven't
really
looked
at.
What's
passing
and
failing
yet,
but
we're
making
progress
here
and
what
more
to
share
on
actual
test
results
soon:
okay,.
B
C
B
C
B
C
B
C
A
B
Yeah,
so
we
could
probably
just
basically
keep
a
list
of,
like
you
know,
supported
just
supported
deployment
or
sorry
would
be
like
the
supported
deployment
tools
or
distributions,
or
something
like
that
and
then
once
we've
got
a
preview
from
from
docker.
We
can
put
that
on
the
list
same
with
openshift,
so,
okay.
D
Yeah,
so
basically,
the
main
this
question
for
well
going
for
this
from
the
discussion
last
week
is
what
approach
we're
gonna
take
to
the
conformance
test.
So
just
a
quick
recap
and
correct
me
if
I'm
wrong.
So
basically,
we
cannot
exclude
the
conformance
test
if
we
want
to
go
to
GA
and
basically
say
that
this
Windows
is
conformers
conformant
according
to
the
company's
community.
D
D
Issue
more
than
a
meter,
they
do
understand
that
some
features
now
work
on
Windows.
They
can
never
work,
but
we
eat
me.
So
the
approach
there
was
discussed
last
week
was
running
business
on
a
hybrid
class
there.
So
basically
forcing
the
tests
to
run
on
Windows
and
if,
for
some
reason
they
cannot
run
and
Windows
they
they
will
be
ran
on
the
Linux
node
in
that
hybrid
class.
D
D
D
Now
the
problem
with
this
approach,
obviously,
is
that
the
way
that
tests
are
written
basically,
there's
no
uniform
way
in
which
the
pods
back
is
defined,
for
some
test
is
directly
into
the
test
for
some
tests,
there's
a
helper
method,
helper
function
and
so
on
and
so
forth.
So
there's
no,
let's
say
framework
function.
That
gives
you
a
positive
basis,
some
parameters.
D
So
thus
we
have
to
go
with
each
test
hand
by
hand
by
hand.
Now
we
in
that,
if
you
look
at
that
PR,
which
is
not
to
change
all
the
tests,
so
that
PR
is
just
the
work
in
progress.
It's
marked
as
working,
because
once
you
discuss
the
exact
mechanism,
if
the
testing
the
the
folks
from
testing
and
Sagami
that
you
are
okay
with
the
mechanism
will
go
forward
and
change
will
be
first,
but
there's
no
point
in
changing
200-plus
tests.
D
D
D
D
D
Organization
under
which
there
with
under
which
today
they
are
so
all
the
images
there
are
manifest
lists.
If
you,
if
you
pull
one
of
them
and
look
into
them,
you
will
see
that
there
are
manifest,
is
not
just
an
image.
One
of
the
issues
that
we
might
encounter
here
is
that
which
one
equation
once
you
create
the
manifest
list,
you
cannot
reference
images
from
different
repositories,
so
registry.
So
basically
we
have
the
images
in
a
registry
and
the
window
switches
in
docker
hub.
D
D
The
amd64
is
the
Linux
one,
but
it
is
included
inside
the
mount
is
one
that
is
a
manifest
list.
So
that's
just
a
little
bit
of
under
the
hood
stuff.
Now,
if
you're,
if
you're
interesting
interested
to
test
on
hybrid
class,
the
class
or
with
this,
these
know
that
you
have
to
use
this
repository
have
to
update
some
Doc's
to
mention
this
as
well.
B
D
This
rarely
built
I
mean
they're
just
built
once
they
change
something,
but
that
versions
or
whatever
I
think
I
need
to.
They
have
automation
in
place,
I
think
to
check
for
changes
in
the
specific
image.
Folder
I
need
to
double
check
this
I
I,
don't
remember.
First
time,
I,
look
but
I.
Think
the
actual
layer
see
this
thing
so
so
building
our
image
is
in
in
kubernetes
now
being
great
and
awesome,
because
that
way
we
will
just
take
them
into
their
manifest
lists.
D
D
B
D
Build
the
Windows
version
of
the
globe,
I,
Nerys
and
so
on
and
so
forth.
So
it's
not
it's
not
really
easy
task,
Plus
that
the
the
way
they
build
images
I
mean
I,
haven't
investigated
this,
but
I,
don't
think
that
they
have
some
some
tool
that
we
can
use
and
just
point
them
so
say
we
provide
a
machine,
a
Windows
machine
where,
where
these
feel
so
I,
don't
think
it's
it's
realistic.
For
this.
D
This
time,
what
I
do
wanna
test
out
is
so
the
error
have
been
having
when
I
try
to
make
manifest
lists
from
images
into
different
in
the
ruler
container.
If
she
and
docker
hub
was
that,
obviously
they
cannot,
they
can
a
bit
different,
but
I'm
curious.
If
the
repository
is
their
problem
or
exactly
the
registries,
so
I
will
push
some
images
in
the
Google
today.
D
B
I
think
the
first
thing
that's
most
important
is
that
since
there's
a
since
there
are
some
unit
tests
that
are
run
before
then
Dan
tests
are
on
I
think
that
that
would
be
a
higher
priority
to
get
that
going
first
and
then
once
that
and
that
sort
of
requires
getting
a
Windows
build
machine
there.
So
I
think
this
right
scope
could
be.
You
know,
run
the
unit
tests
on
things
like
the
cubelet
and
cube
proxy
and
then
also
see
about
building
the
windows.
Specific
image
is
needed
for
ete.
B
B
Okay,
all
right
sounds
good,
well
make
sure
those
are
on
the
stick,
testing
and
Civic
architecture
agenda.
So
we
can
keep
that
discussion
moving
forward,
so
cool
all
right
else,
Michael
or
Peter
I'm,
not
nothing
from
me.
Thank
you.
Okay,
all
right!
Well,
thanks!
Everyone
have
it
have
a
great
day,
I'm
gonna
head
over
to
the
signal
meeting.