►
From YouTube: Kubernetes SIG Node 20181106
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).
B
A
B
A
All
right,
excellent,
yes,
so
I'm
patrick
laying
on
one
of
the
co-chairs
for
sig
windows.
A
couple
quick
things
I
wanted
to
mention
today
were
we've
been
continuing
to
to
work
on
on
the
windows.
Effort
and
I've
got
a
release
criteria
doc
linked
here.
This
is
the
one
that
we
started
back
in
the
1.12
timeline
and
then
we
brought
it
forward
to
1.13.
A
The
main
things
have
changed
between
then
and
now
is
that
a
lot
of
the
PRS
we
needed
to
get
things
like
dns
working
in
Windows?
We
were
just
merged
thanks
for
everyone
that
helped
with
that
over
the
last
couple
days,
I'm
getting
that
in
before
code
freeze,
but
we're
now
at
a
point
where
the
test
results
are
much
more
consistent
and
if
you
look
down
later
in
the
doc,
we've
got
a
list
of
all
the
test
cases
I've
imported
and
are
running
on
Windows.
A
The
main
thing
that
we're
focusing
on
now
is
getting
more
of
the
organizational
work
done.
I'm
going
to
be
sending
this
document
back
out
to
say,
architecture,
windows,
node,
storage
and
networking
see
if
there's
any.
You
know
final
feedback
on
that
around
the
set
of
features
and
that
the
scope
of
the
release
and
then
the
other
thing
I'm
doing
on
right
now
is
I'm
working
with
Michael
Michael
and
some
others
on
getting
all
of
the
docs
done
that
have
windows,
use
cases
and
samples
in
them.
A
So
I've
got
a
work
in
progress
PR
there
that
will
be
finished
before
the
19th,
which
is
the
cutoff
where
they're
going
to
start
reviewing,
and
the
other
thing
I'm
doing
in
parallel
to
that
is
I've
got
a
tracking
issue
here
for
updating
the
API
reference
Docs
to
clarify
what
which
options
and
the
API
are
are
covered
in
Windows
and
which
ones
are
not
covered
because
they're
at
their
linux
specific.
So
I've
got
a
rough
list
there,
but
that's
going
to
be
replaced
with
pr's.
A
So
that
way
you
know,
developers
and
and
and
admins
that
are
digging
through
the
API
side
of
the
reference
will
be
able
to
find
the
info.
They
need
there,
whereas
your
users
they're
getting
started
with
all
their
first,
we'll
be
able
to
go
through
this
one
here
and
I've
got
like
a
proposed
outline
of
what
of
what
the
main
topics
are
that
we're
going
to
cover
so
that
way,
people
will
have
what
they
need.
A
B
A
So,
starting
from
the
Windows
version
support,
initially
we
were
planning
to
support
Windows
Server
2000
19,
that
was
when
it
was
announced.
It
was
supposed
to
be
out
in
October
that
hasn't
happened
yet
so
what
we've
been
doing
is
just
focusing
on
running
Windows,
Server
version
1803
and
it
works.
It
works
well
with
docker.
The
only
problem
is
that
Microsoft
is
going
to
be
dropping
support
for
that
version
in
the
winter
of
2019.
A
So
for
the
next
version
for
kubernetes
1.14.
That's
what
we're
going
to
focus
on
making
sure
that
Windows,
Server
2000
19
is
supported
because
I
think
that's
what
most
customers
will
want
to
be
on
long
term,
there's
also
a
reasonable
chance
that
2019
may
just
work
with
docker,
but
we
don't
have
the
test
pass
in
place
for
that,
yet
so
I
guess
in
alternative
plans.
A
If
we
wanted
to
bring
the
this
docker
support
with
Windows
Server
2000
19
in
maybe
that
can
make
something
like
a
13.1
release,
because
there's
not
any
substantial
changes
to
the
functionality.
That's
there
in
docker
that
works
with
Windows
Server
2000
19.
The
thing
that's
really
important
about
2019
is
that
we
need
it
to
move
to
container
D
and
bringing
up
container
D
is
going
to
require
version
14
anyway,
I
don't
see
us,
you
know
realistically
putting
a
new
CRI
into
version
14
and
then
bringing
that
back
to
13.
A
We're
just
simply
not
not
testing
those
at
this
point
well
because
they
don't
work
with
things
like
secrets,
config
maps
and
shared
volumes,
and
so
I
don't
think
that
they
would
really
be
be
useful
for
most
people.
You
know
my
workout
reduced
functionality,
but
that's
not
what
we're
going
to
be
publishing
on
tests
great.
A
So,
in
terms
of
you
has
a
setup
experience
all
that
we're
running
on
on
Windows
is
just
the
binaries
that
are
there
on
the
on
the
node.
So
the
you
know
your
your
scheduler
API
server
controller
manager,
all
that's
so
going
to
be
run
on
on
the
NEX
the
same
way
as
it
is
today
we're
just
going
to
be
using
the
cube,
let
and
cube
proxy,
which
are
already
there
and
the
kubernetes
builds.
Today.
A
The
main
one
I've
been
focusing
on
is
the
azure
sea
and
I
plugin
that's
what
we're
going
to
be
using
for
the
results
on
test
grid
that
I
run
on
Azure,
and
then
we've
got
another
plug-in
that
Peters
looking
at
that
should
be
able
to
work
on
GCE
and
it's
basically
more
of
a
I
think
they're,
calling
it
a
meta
plugin.
But
that's
not
part
of
the
kubernetes
repo.
That's
like
all
these
CNI
plugins
are
part
of
separate
repos,
but
this
one.
A
You
can
basically
use
to
just
program
the
nut
windows
network
interfaces
for
the
containers,
and
you
can
still
use
something
else
for
ipam,
so
things
like
flannel
or
GK,
CNI
plugins
could
just
layer
on
top
of
that.
So
we're
getting
good
good
feedback
on
on
that
as
well.
The
same
thing
with
ovn
kubernetes
son,
which
works
with
the
open
virtual
switch
project,
and
so
we'll
have
links
to
that
over
in
the
doc
set
as
well,
because
they've
got
ansible
job
for
people.
A
Also
update
that
real,
quick
and
then
and
then
we
have
to
flex
volume
plugins
that
are
published
over
there
on
github
for
SMB
and
I
scuzzy
and
then
entry.
We
also
have
the
the
azure
providers
that
are
therefore
as
your
disc
and
as
your
file
I
don't
know
if
anybody
is
started.
Work
on
stuff
for
GCE
are
those
entry
or
are
those
out
of
tree
flex
volume
right
now.
B
A
Okay,
sync
with
Peter
and
update
that.
A
A
A
You
know
the
secrets
and
config
maps.
Those
are
working
working,
fine
on
the
current
releases,
although
one
thing
I
should
call
out
on
storage,
is
that
subpath
mounts
are
not
implemented
on
Windows,
because
we
can't
map
individual
files
that
are
there.
So
you
have
to
basically
map
you
just
map
the
whole
volume.
A
So
that's
something
I'll
make
sure
is
updated
in
the
API
documentation
being
able
to
do,
subpath
Mouse
is
going
to
require
Windows,
Server,
2000,
19
or
possibly
even
a
later
version.
Just
because
the
file
system
driver
isn't
there
for
being
able
to
restrict
those
down.
So
that's
something
that
we
will
be
looked
after
later.
A
The
metric
stuff
came
in
mostly
during
during
12.
There
is
still
one
open,
PR
around
Network
stats,
where,
if
we
get
enough
reviews
they'll
be
in
13,
but
that's
the
like
per
network
adapter
stats.
A
The
rest
of
this
stuff
is
all
is
all
there
for
the
pods
and
then
resource
some
limits
implemented.
Those
came
in
in
12
HPA
should
work
on
top
of
those,
that's
something
that
we'll
be
including
in
the
docs
with
examples,
as
we
get
that
going
and
then
networking
I
mentioned
the
plugins
that
are
already
there
one
you're,
the
one
main
Delta
with
with
the
like
support,
is
that
we
don't
support
the
host
networking
mode.
A
A
So
it's
there
today
is
still
that
experimental
and
annotation
which
lets
you
run.
You
know
one
container
in
a
pod
you're
using
the
hyper-v
hypervisor
boundary.
Once
we
move
to
2019
a
container
d,
that's
where
I
will
be
able
to
get
support
for
multiple
containers
per
pod
and
actually
implement
runtime
class.
The
way
the
way
we
had
discussed
it
before
and
then
in
terms
of
you
know
the
the
other
testing
we
listed
all
the
test
cases
down
there.
A
A
Otherwise,
for
all
those
cases,
and
so
that's
still
something
that's
you
know
under
review
there,
and
so,
if
that
doesn't
merge,
then
we'll
need
to
refer
back
to
this.
This
list
of
I've
known
failures
down
there
and
we
can
just
use
it
Jenko
test
exclusion
list.
Instead,
if
we
want
to
focus
that
down
on
test
great,
don't
only
what's
what's
known
to
be
working
on
Windows
long-term,
the
way
siga
architecture
wants
to
handle
this
is
to
create
what
our
we're
going
to
be
called
conformance
profiles.
A
But
that's
not
something
that
they
had
started
working
on
or
is
in
scope
for
version
13
and
so
I've
been
working
with
with
Brendan
burns
on
that
and
we're
basically
just
going
to
keep
keep
pushing
that
direction
and
say
architecture,
because
they've
got
another
case
where
they
want
profiles
as
well,
and
so
that's
kind
of
been
identified
as
the
direction
that
the
cig
architecture
would
like
to
go.
And
so
you
know
what
we're
proposing
to
do
is
take
this
list
of
cases
that
we
have
here
and
you
know
push
them
in
that
in
that
direction.
B
A
B
A
Okay,
so
so
using
the
hyper-v
isolation,
that's
there
today
you
can
use
containers
built
on
Windows,
Server,
1709
or
1803,
and
then
likewise,
when
customers
move
forward
to
2019
they'll
be
able
to
run
1803
based
containers
as
well.
So
today,
if
you
want
to
backwards
compatibility,
you
would
use
the
experimental
hyper-v
isolation
and
then
that's
something
that
you
know
will
update
into
the
runtime
class
when
we
go
into
version
14
with
container
D.
But
that's
that's
the
other
main
use
case
for
hyper-v
isolation.
So
it
gets
either.
B
A
A
Yes,
so
the
first
link
there,
the
github
link
shows
that
the
PRS
I'm
tracking,
which
is
down
to
just
the
remaining
two
tests
and
doc
PRS
and
then
the
second
link
that
I
have
there
in
the
signal
meeting
notes
is
the
release
criteria,
doc,
which
also
includes
the
test
case
list.
That's
the
one
that
I
had
on
the
screen
share.
Okay,.
B
B
C
Well,
this
question
up
in
Cuban
at
this
dam
we
could
two
weeks
back
or
so,
and
there
was
some
discussion
about
that
and
they
were
basically
two
suggestions
or
two
alternative
suggestions.
But
what
to
use
for
the
node
node
labels
and
I
was
I
was
going
to
ask
ask
advice
from
the
signal
meeting
what
how
to
how
to
proceed
with
that?
Can
we
make
a
decision
here
or
should
I
take
the
question?
What
in
in
the
signal,
maybe
mailing
list,
for
example?
C
So
basically,
there
were
like
two
alternatives
suggested
in
the
Cuban
and
disturbed
mailing
list.
First,
one
for
the
for
the
namespace
was
featured
note
that
kubernetes
io
to
use
for
the
NFD
labels
and
the
other
one
was
like
end
of
day
but
note
Cuban
at
least
attire
and
color,
the
former
one
Pollock
sort
of
bitter
or
better
describes
what
the
label
is
about
and
the
other
one
is
like
project
or
anybody
specific.
C
C
C
A
B
Actually,
I
I
I'm,
okay
lose
the
MFT
I
personal,
that's
better!
That's
just
I.
Don't
have
good
taste
one
company,
the
animal
who
is
the
second
a
name
and
Lefty.
C
C
C
B
B
C
B
C
B
C
A
B
You
Remus
to
the
signal
to
come
the
meeting
so
anyway
do
we
have
any
topic
or
any
issue.
You
want
brought
up
any
concern.
You
need
to
the
r1
that
certainly
nice
and
the
project
to
progress
all
the
things.