►
From YouTube: Kubernetes SIG Windows 20180807
Description
Kubernetes SIG Windows 20180807
A
A
B
A
C
A
A
Well,
I
guess:
I'll
go
ahead
and
and
give
a
quick
update
on
that
the
GA
progress
so
a
few
weeks
ago,
you
know
I
had
written
up
some
proposed
release
criteria
that
we
discussed
with
with
the
group,
and
so
those
are
linked
over
in
the
notes
as
well.
But
I
was
I
discussed
it
briefly
with
sig
note
last
week
and
they
recommended
that
we
also
needed
to
get
reviews
from
sig
storage
and
sig
networking
so
I'm
going
to
try
to
take
this
to
to
those
meetings
and
get
some
feedback
from
them
as
well.
A
A
A
So
you
know
at
this
point
this
isn't
100
percent
guaranteed
for
112,
but
at
least
we
know
what
the
process
is.
So
once
we
go
through
all
these
reviews
get
the
tests
passing
and
then
you
will
be
able
to
move
that
forward
and
from
from
a
doc
standpoint,
you're
the
main
things
that
that
signal
brought
up
was
they
want
to
make
sure
that
we
were
really
clear
and
in
terms
of
you
know
which
features
were
supported
on
Windows
at
to
kubernetes
api
level,
and
so
we'll
need
to
work
on.
A
So
we've
got
a
little
bit
of
that
today
in
that
dark
set,
but
I
think
we
need
to
sort
of
separate
this
out
a
bit
into
you
know:
here's
it
getting
started
and
here's
some
you
know
deployments
and
best
practices
then
also
have
like
an
API
reference
that
corresponds
with
that
as
well.
That
covers
more
specifics
for
people
that
need
to
know
more
at
the
API
level.
A
E
A
A
A
Couple
PRS
to
continue
fixing
the
dns
resolution,
so
I
should
have
a
better
ETA
on
those
later
today,
but
we're
working
I'm
working
with
with
two
Nash
and
Peng
Fei
today
once
pancrase
online,
but
we
basically
need
to
update
how
the
windows
CNI
plug-ins
are
setting
the
DNS
servers
search
suffix.
So
that
way
you
can
access
things
in
the
global
service
namespace,
not
just
the
pods
namespace.
A
A
E
So
one
of
my
questions
was
about
Linux
containers
on
Windows
that
was
listed
as
future
work
and
I
was
wondering
if
we
expects
kubernetes
users
to
need
that,
because
we're
gonna
need.
As
far
as
I
know,
kubernetes
clusters
are
going
to
need
a
Linux
master
for
the
foreseeable
future.
So
why
wouldn't
we
just
use
Linux
nodes
to
run
Linux
containers?
It
must
say
yeah.
A
So
we've
had
a
few
customers
that
that
have
asked
about
this,
because
we
have
the
technical
capability
to
run
Linux
containers
directly
on
Windows.
Then
we
could
potentially
run
you
know
the
the
the
controller
or
the
schedule
of
the
API
server
as
links
containers
on
Windows
and
actually
remove
the
need
for
that
Windows
know
it
all
together.
The
reason
I
call
that
out
in
the
docs,
specifically
as
future
work
is
because,
in
order
for
that
to
really
make
sense,
we
also
need
to
get
to
a.
A
We
would
need
to
figure
out
what
a
deployment
model
would
be.
Where
you
know,
one
I,
don't
I,
guess
nodes,
not
the
right
word,
but
one
Windows,
OS
installation
could
be
hosting
containers
for
both
Linux
and
Windows,
because
that
breaks
the
kubernetes
assumptions
around
a
node
from
an
API
standpoint
today.
A
So
maybe
that's
something
that
is
done
through
more
of
like
you
know
the
node,
this
paradigm.
That's
you
know
it's
a
in
a
separate
working
group,
or
maybe
you
know
the
node
is
a
Windows
node,
but
we
also
have
you
know
a
you
know,
sort
of
a
node
--less
instance
that
can
manage
the
Linux
containers
for
burst
capacity.
A
E
A
So,
like
that's,
that's
one
case
that
that
we've
have
some
people
have
expressed
interest
in
we're
still
trying
to
gauge
that
the
other.
The
other
case
is
that
if
a
customer
wants
hypervisor
based
isolation,
they
may
desire
to
get
that
from
hyper-v.
And
so
you
know,
the
capabilities
of
hyper-v
are
very
similar
to
what's
available
with
technologies
like
cada
today
and
at
this
point
in
time,
I
don't
know
well
what
the
priority
is
of
that
or
if
there's
people
that
are
looking
for.
But
it's
something
that's
definitely
definitely
technical
technically
feasible.
A
So
you
know
if
they're
already
running
hyper-v
on
their
physical
clusters
today
and
they
may
desire
hyper-v
for
Linux
isolation,
but
rather
than
putting
that,
as
part
of
you
know,
the
initial
release
I'm
just
intentionally
saying:
let's,
let's
hold
off,
let's
hold
off
on
that,
because
that's
a
different
level
of
discussion
that
we
would
need
to
probably
take
through
Sega
architecture,
which
is
separate
from
running
Windows.
Workloads
on
Windows
nodes
makes
sense.
I.
E
A
E
A
So
it
could
be
misunderstanding
this
one,
so
so
the
way
I'm
I've
been
looking
at
so
I
haven't
deployed
sto
personally,
because
I've
been
working
mostly
on
the
windows
side.
But
my
understanding
is
that
some
of
those
needed
access
to
additional
capabilities
to
do
things
like
mounts
and
we
don't
have
any
way
like
that's,
not
a
separate
privilege
that
can
be
asserted
on
Windows
and
so
I
may
have
oversimplified
that
statement
there.
A
A
All
right,
I'll,
try
to
see
I'll
try
to
find
someone
else
that
can
give
give
some
more
feedback
on
that
because
out.
That
was
my
understanding
of
those
architectures.
But
okay,
all
right,
I
guess:
I'm
ready
to
move
forward
to
the
fluency
topic
of
I
and
James
and
then
actually
did
I
get
your
name
name
right.
C
Yeah
so
I
recently
opened
a
pull
request
to
the
fluent
D
repo,
with
a
Windows
version
of
their
docker
container.
So
now
you
can
run
the
fluent
D
sidecar
alongside
I
mean
sidecar
on
Windows
nodes.
Now
I
think
the
next
thing
would
be
to
get
it
into
the
kubernetes
repo,
but
I
can
post
the
pull
requests
in
the
chat.
C
C
C
There
wasn't
really
an
easy
way
to
get
container
level
logs,
especially
with
I
think
for
us.
We
were
seeing
customers
not
being
able
to
easily
change
their
applications
to
write
to
standard
out
mm-hmm
with
Windows
applications.
So
we
wanted
a
way
to
be
able
to
pull
those
logs
from
from
files
on
disk
and
with
fluent
D.
It
made
it
really
easy
to
do
that.
C
They
just
didn't,
have
a
Windows
version
and
what
you
do
is
you
just
run
it
as
a
sidecar,
and
lo
and
D
will
basically
follow
logs
on
a
file
for
you
and
basically
put
that
or
forward
that
on
to
wherever
you
want.
So
in
the
kubernetes
clay.
The
case
and
games
probably
speak
more
to
the
elasticsearch
application,
but
then
you
would
be
able
to
use
elasticsearch
with
fluency
as
well
in
the
on
Windows
nodes.
C
C
F
So
I
I
guess
part
of
it
was
trying
to
figure
out
if
we
can
get
any
feedback.
We
saw
that
as
a
issue
in
the
backlog
for
the
Windows
sig
group
was
kind
of
giving
in
the
fluid
D
working
so,
but
this
is
kind
of
the
first
step
and
looking
to
look
at
the
kubernetes
has
an
example
of
setting
up
some
D
on
your
cluster
for
doing
log
collection,
pushing
to
elastic
stack
and
with
the
latest
changes
to
enable
symlinks.
F
You
can
actually
do
that
now
with
with
Flemington,
so
you
can
configure
to
pull
up
all
the
metadata
about
all
the
pods
and
namespaces
that
the
locks
are
coming
from
and
then
it
pushes
it
to
elastic
search-
and
you
can
you
know,
sort
of
search
on
all
that
I'm.
So
the
idea
was
to
initially
get
this
in
there
fluent
ease
of
people
can
use
it
and
be
successful,
and
also
on
top
of
that
be
able
to
have.
F
A
F
That
would
be
in
the
instance
where
you're
running
from
the
d
on
each
node
as
a
daemon
set,
so
that
you
can,
it
would
just
end.
So
then
each
pod
would
write
to
standard
out
and
the
fluid
d
kubernetes
send
links
all
of
the
container
logs
to
the
to
it.
I
think
it's
like
some
temper
logs
slashed
and
then
it
adds
all
that
metadata
and
then
the
fluid
e
payments
that
would
pick
all
that
up
and
ship
it
off
to
wherever
you
want
it
to,
but
in
the
example
that
they
have
is
the
elastic
search.
F
C
D
E
D
A
A
F
I
think
I
think
once
we
once
it
kind
of
goes
to
a
people,
gonna
start
wondering
about
hey.
How
do
I
actually
keep
logging
at
a
production
level,
and
this
is
you
know,
trying
to
look
at
what
they're
doing
on
the
Linux
side
and
replicate
it
on
the
window
side
as
much
as
we
can
I
think
it's
a
good
first
approach,
unless
anybody
has
any
other
ways
that
they
suggest
or
recommend.
G
G
F
A
H
H
H
F
Ok,
yeah,
we
haven't
completely
tried
the
whole
thing
I
wanted
to.
We
just
got
that
initial
one
working,
but
we
haven't
done
the
fluent
Diaz
on
the
host,
and
so
there
may
be
some
restrictions
there
and
if
that's
one
of
the
issues
that
comes
up,
then
we
can
address
it
in
the
in
this
meeting.
I
guess
it's
a
bug.
D
A
Just
because
I
think
a
lot
of
customers
are
typically
I'm
not
going
to
be
using
one
container
per
pod
for
their
existing
applications,
because
you
know
things
like
asp.net
already
have
their
own
hosting
environment,
so
I
think
the
most
likely
case.
Someone
you
know
is
running
multiple
pods,
you
know
is
for
the
logging
sidecar,
so
that'll
be
great.