►
From YouTube: Kubernetes SIG Windows 20170822
Description
Kubernetes SIG Windows 20170822
A
Hi
everybody
thank
you
for
attending.
Sig
windows
haven't
been
on
any
of
these
meetings
for,
for
the
last
four
weeks,
so
I've
been
getting
my
updates
through
email,
as
well
as
select
channel
readings,
so
I'll
provide
a
little
bit
of
an
update
on
what
appender
has
been
working
on
for
the
last
couple
of
weeks
and
then
we'll
open
up
the
floor
for
other
folks
to
provide
updates.
I
also
have
a
small
update
from
David
Toder,
but
I
don't
see
him
online.
A
So
if
he's
not,
he
doesn't
join
us
by
then
then
I'll
provide
the
update
on
his
behalf.
So
from
the
aprender
perspective,
who
had
a
couple
of
engineers
working
on
the
storage
and
volumes,
that's
one
of
the
prerequisites
for
us
to
get
to
beta
and
essentially
the
the
work
that
we're
doing
there
is
to
provide
support
for
things
like
empty
host
paths.
Persistent
volumes
secrets,
as
well
as
local
and
for
the
most
part,
we've
gotten
90%
of
these
working
local
is
one
of
those
things
that
is
not
working
yet,
but
that's
also
an
alpha
feature.
A
So
we
need
to
figure
out
that
something
that
we
need
to
support
or
not.
Well,
my
directive
was
to
the
team
was
if
it
ends
up
being
a
time.
Sync
we'll
skip
it
until
it
gets
to
a
beta
or
or
or
general
availability
level
feature
so
other
than
that
I've
been
doing
a
lot
of
helping
of
folks
to
install
the
sip
windows
work
using
the
ouvea
know,
vs
work
from
cloud
base,
as
well
as
the
new
installers
and
the
new
capability
to
put
into
this.
That's
that's
it.
A
Obviously
we
just
haven't
been
there,
so
my
goal
is:
if
you
can
get
some
of
the
networking
to
be
completely
robust,
especially
the
help
from
the
lean
and
a
lean
and
cloud
base
and
then
get
the
storage
on
volumes
there,
including
kinetic
mob
secrets
and
all
these
things
and
a
single
be
in
good
shape
to
declare
beta
what
we
are
still
missing.
Significantly
is
regression
runs
and
automated
testing
right
so
after
beta
that's
going
to
be.
The
first
focus
is
to
make
sure
that
we
get
all
of
our
testing
infrastructure
in
place.
A
To
be
able
to.
You
know,
catch
regressions
on
this
work.
I
wish
you
could
have
done
that
earlier,
but
you
know
we
just
didn't
have
resources.
The
other
thing
that
may
make
it
into
beta
and
that's
since
David
is
not
online
I'll
provide
some
of
his
update
is
the
work
that
David
is
doing
around
CRI
and
getting
the
container
stats
for
for
sip
windows
for
kubernetes.
A
When
you
look
at
that
work,
it's
kind
of
split
into
three
efforts
and
there's
the
node
level
and
the
pod
level
statistics
that
exist
and
we're
not
really
doing
anything
for
that
today,
kubernetes
even
for
dr.
kelso
stats
from
c
advisor,
so
they
have
imported
those
to
CRI
yet
so
whenever
to
figure
out
what
we're
going
to
do
about
that,
the
last
part
is
the
container
level
metrics
which
are
coming
from
CRI.
A
David
has
had
some
difficulty,
and
actually
he
just
joined
us
right
now.
Getting
his
PR
most
signaled
is
giving
him
some
pushback,
so
so
what
we'll
do
is
we're
going
to
attend
the
signal
meeting,
meet
David
nan
Tony
within
the
next
half
an
hour
and
have
a
discussion
with
them
and
kind
of
try
to
address
some
of
that
concerns
figure
out
what
the
issues
are
and
and
if
we
need
to
make
some
code
changes
will
obviously
evaluate
that
as
well.
A
So
David
I
gave
your
update
I,
don't
know
if
there's
anything
else,
I'll
give
you
another
there's
anything
else
you
want
to
provide,
but
that
you
can
definitely
chime
in.
So
that's
it
for
me.
So
I'll.
Let
David
speak
about
some
of
the
work
he's
doing
if
there's
additional
updates
from
what
I
mentioned
and
then
we'll
go
around
the
room.
B
B
D
B
B
D
B
Actually,
I've
been
working
offline,
a
little
bit
to
be
to
update
my
PR
to
get
to
work
with
with
the
CRI
offline,
so
that
should
one
of
them
in
concern.
The
other
thing
that
we
still
need
to
figure
out
with
them
is
regarding
the
node
metrics.
One
of
the
things
they
asked
was
that
maybe
make
sense
for
node
metrics
to
live
outside
of
the
kubernetes
codebase
and
actually
inside
a
different
repo,
and
we
need.
B
How
I
see
it
it's
kind
of
old,
a
painful,
because
there's
really
not
that
much
code
for
the
node
metrics
that
make
sense,
I
think
for
a
Bolivian
couplet.
It
lives
outside
the
cou
blue.
We
need
to
figure
out
who
will
own
that
repos,
it's
kind
of
a
more
of
a
problem.
So
that's
one
thing:
I
kind
of
want
to
address
so.
A
That's
kind
of
the
main
thing
yeah
we'll
discuss
it
with
them
later
on.
Just
given
the
fact
that
the
node
and
port
level
metrics
are
coming
from
the
advisor
and
the
fact
that
they
want
us
to
have
some
flagging
to
provide
those
in
Windows
yeah
might
not
be
a
bad
idea.
I
just
want
to
hear
you
young
kind
of
explain
it
and
see
what
their
rationale
is,
why
they
want
us
to
go
down
the
route
versus
like
an
encapsulation
or
CC
advisor
yeah,
so
sort
of.
C
A
D
I
think
it
would
be
good
to
differentiate
to
if
this
is
just
you
just
opinion
or
if
it's
something
that's
a
widely
held
belief
by
say,
dawn
or
other
members
of
the
sig.
So
because
sometimes
people's
personal
opinions
can
be
very
fixed
and
not
necessarily
representative
of
the
grander
scheme
and
also
if
this
kids
could
be
a
major
problem.
It
can
also
be
arbitrated
at
the
architecture
level.
A
C
My
opinion
is
that
these
things
should
be
in
the
sea
steer
tooth
Putin
is
actually
one
is
that
sig
Windows
should
take
a
strong
look
at
implementing
all
the
container
stuff
via
the
CRI.
That
would
involve
forking
the
sort
of
the
existing
docker
stuff,
but
it
will
be
able
then,
would
be
able
to
do
it
in
a
very
particular
Windows
way
and
not
have
to
worry
about
sort
of
the
Linux
docker
stuff
breaking
I
know.
C
Don
has
mentioned
in
the
past
that,
like
they're
always
concerned
about
doing
things
that
break
the
window
size
because
they're
not
testing
it,
and
they
don't
want
to
do
that
segundos.
The
other
thing
is:
is
that
why
nodes?
That's?
Why
shouldn't
that
be
part
of
the
CRI
like
just
like
containers,
that's
we're
getting
from,
and
that
would
make
I
think
quite
easier
for
signal.
I
mean
sick
windows
as
well.
Just
my
two
cents.
A
B
B
E
A
C
Saying
that
the
well
is
those
things
implemented,
Lexi
advisor
was
how
container
sets
were
being
retrieved
before
and
promote
so
personal
experience
like
things
like
horizontal
pod,
auto-scaling
Elise
were
not
done
with
pod
level
stats.
They
were
done
with
container
level
stats,
because
I
tried
implementing
pod
level
stats
and
ignoring
container
level
steps
and
horizontal
pod.
A
A
E
Fortunately,
I
don't
have
any
updates
regarding
those
in
obvious
with
CNI
we've
been
testing
the
metric
new
species.
That
might
represent
it
a
couple
of
weeks
ago.
I
think
and
the
good
news
is
that
they
we
might
use
the
same
CNI
code
that
they
want
to
push
to
the
kubernetes.
So
that's
that
code
would
be
used
for
both
of
us
other
than
that
nothing
new.
There.
A
A
A
D
A
A
Right
cool,
so
I,
don't
know
what
happened.
There
assume
just
crashed
on
me,
but
it
seems
it's
still
recording
so
I
don't
know
what's
going
on
there,
but
what
I
was
going
to
say
is
that
a
lot
of
our
work
can
ship
completely
out
of
band.
So
we
don't
need
to
essentially
worry
about
the
1.8
with
this
train,
so
our
goal
was
to
get
to
beta
by
end
of
September
of
2017.
A
So
a
link
that
gives
us
maybe
about
five
weeks
are
the
last
couple
of
weeks
for
stabilizations
is
about
three
more
weeks
before
we
say
who
can
code
fleet
now
and
start
stabilizing
this.
So
once
you
talked
to
Alan
and
Alessandra
and
see
you
know
how
that
fits
with
the
rest
of
your
timelines.
I
probably
should
start
a
thread
with
Alessandra
as
well,
given
that
me
and
him
are
presenting
to
the
open
source
cloud
forum
in
October
in
Europe,
so
we
want
to
have
beta
ready
by
then.
A
So,
if
it's
not
September,
definitely
by
mid-october,
we
need
to
be
done
and
get
bed
out.
Obviously,
you
know
there's
a
lot
of
mitigating
factors
where
we
are
where
we
are
today
right.
So
lack
of
resources
is
one
of
them
and
you
know.
Obviously
we
have
lots
to
have
more
contributors
working
on
this,
so
especially
thankful
to
the
folks
there
on
this
SIG
that
are
contributing
and
I
wish.
We
had
more.