►
From YouTube: Kubernetes SIG Storage 20170330
Description
Kubernetes Storage Special-Interest-Group (SIG) Meeting - 30 March 2017
Meeting Notes/Agenda: https://docs.google.com/document/d/1-8KEG8AjAgKznS9NFm3qWqkGyCHmvU6HVl0sk5hwoAE/edit#heading=h.mjv7xb5aiecb
Find out more about the Storage SIG here: https://github.com/kubernetes/community/tree/master/sig-storage
Moderator: Brad Childs (Red Hat)
Chat Log:
09:33:22 From hekumar : is the code ready for review btw?
A
A
C
On
mute
there
we
go,
can
you
hear
me
now?
Yes,
okay,
so
Matthew,
Leo
and
I
went
through
the
planning
spreadsheet,
which
is
linked
in
the
agenda
and
we
went
through
and
marked
each
of
the
items
as
done
not
started
or
started.
We
can
review
those
items
it's
folks
would
like
and
then
what
we
did
was
we
copied
items
that
were
not
parted
or
started
into
the
1.7
tab.
That's
a
starting
point.
C
You're,
let
me
run
do
the
list,
do
you
want
me
well,
I
could
I
could
do
it
I'll
go
ahead
and
share
my
screen,
then
here
we
go
all
right,
so
this
is
the
list
of
items
that
we
filtered
down.
So
NFS
wedging
issue
was
something
that
we
dropped
last
quarter.
C
It's
something
that
we
should
not
drop
this
quarter.
It
requires
a
larger
refactor,
a
larger
redesigned,
to
address
the
issue
we
have
work
around
in
place
to
prevent
it,
especially
now
with
mount
options.
You
can
do
a
soft
soft
mount
to
prevent
this
from
happening
if
it's
really
an
issue,
but
it's
something
that
we
should
tackle
at
this
point.
We
don't
have
it
for
it.
Does
anybody
want
to
step
up
and
take
ownership
for
this
freak?
You
too,
I'll.
Take
it
I've
been
seen
months,
good
man,
all
right.
C
Badging
yeah.
C
B
C
F
C
C
C
The
initial
proposal
go
out
a
couple
weeks
ago:
lots
of
comments
on
that
I'm
going
to
continue
to
drive
that
forward,
we're
probably
going
to
hold
meetings
separate
from
from
the
primary
storage
sig
meetings
to
discuss
those
because
they're
going
to
require
a
larger
design
discussions,
but
I
plan
to
continue
to
drive
that
and
I
think
there's
a
lot
of
folks
on
the
on
the
reviewer
line.
Please
feel
free
to
add
yourself
on
a
reviewer
line
if
you
want
to
be
involved
well,
not
involved.
C
If
you
want
to
be
part
of
the
approvals
I
guess,
but
the
aim
for
this
quarter
is
going
to
be
to
get
agreement
across
the
cholesterol
orchestration
systems.
On
what
the
CSI
interface
looks
like
we're,
probably
not
going
to
talk
about
implementation
or
design
on
this
quarter,
but
then,
hopefully
the
subsequent
quarter.
We
can
focus
on
that
part:
local,
assemble,
femoral,
storage
and
local,
persistent
storage
Michelle.
Do
you
want
to
talk
about
both
or
how
you
want
to
break
it
up?
Yeah.
H
Sure
so
we
have
we
had
our
high
level
design
angel
on
with
this,
with
just
the
user
interface
details
for
22.
We
want
to
start
to
implement
an
alpha
feature
based
off
of
that,
so
I've
linked
the
proposal
that
we
had
out
and
we
also
start.
H
H
That
would
be
very
helpful
right
now.
If
you
look
at
our
tracker
issue,
we
have
quite
an
aggressive
goal
for
alpha
I
think
we
only
have
a
little
more
than
a
month
to
develop
all
the
Alpha
features
in
order
to
meet
the
tories,
so
yeah
all
the
help
that
we
can
get
in
terms
of
reviewing,
and
all
of
that
would
be
great.
H
Let's
see
the
owners
are
correct,
I
guess,
I,
don't
know
about
other
reviewers
correct
I
I
would
like
to
have
more
than
one
person
read.
You
know,
mix
ups
and
some
time
to
dedicate
some
time
is
quarter
to
help
you.
We.
A
Can
help
review
all
right
which
I
want
to
notice
on
these
I
on
the
proposals?
There's
a
pretty
big
I
requirement
on
the
scheduler
around
this
and
it
like
some
of
that
expertise
of
the
outside
of
just
fig.
If
we
had
enterprise-ready
and
we
identified
anybody
else
outside
of
storage,
stick
the
help
yeah.
H
F
I
F
C
E
C
C
C
D
C
Fair
all
right,
I'll
go
ahead
and
remove
that
item
and
we'll
have
to
remember
to
readd
it
should
testing
is
always
a
high
priority.
We
are
going
to
continue
to
drive
on
this
Aaron's
been
doing
an
amazing
job.
I
think
she
can
use
all
the
help
she
can
get.
There's
been
more
and
more
folks
dialing
into
the
test
meeting.
She
holds
bi-weekly
on
Fridays
Aaron.
Do
you
wanna?
Are
you
online.
G
Again,
I
added
one
item
on
the
agenda
to
after
we
go
over
the
spreadsheet,
no
I
mean
simply
that
John
and
I
are
chugging
along
on
with
an
eye
toward
for
the
future,
where
we
get
more
of
the
feature
offers
to
also
add
an
n
test,
we're
trying
to
keep
the
trying
to
make
the
test
consistent
and
structured
well
so
that
they
can
be
so
that
any
file
you
randomly
grab
and
in
the
UAE
directory
you
know,
is
a
good
choice
to
copy
for
your
own.
G
You
test
we're
trying
to
pull
out
common
elbow
routines
and
we're
trying
to
deflate
the
test.
There's
there
are
some
issues
with
the
way
to
test
the
written
that
make
them
fail
when
they're
running
parallel
and
we're
working
on
cleaning
that
up
as
well.
Another
pull
out
a
tree
issue
that
might
that,
hopefully
sig
stored,
or
at
least
them
look
at
a
high
level
to
decide
whether
there's
any
role
in
entry
e,
2
e
and
see
I.
If
there's
any
role
between
that
and
out
of
trees,
storage,
cuz,
that's.
C
Super
valuable
on
helping
the
whole
team
be
productive.
So
thank
you
all
right.
So
next
up
cloud
provider,
storage,
metrics-
this
was
an
item
that
was
dropped.
Last
quarter
is
yon
still
to
the
right
person
to
work
on
this.
No.
C
D
C
And
so
for
everyone
who
has
a
feature,
that's
going
to
be
part
of
1.7
click
on
your
future
repo
issue
and
update
the
milestone
to
1.7.
If
it's
not
and
you
don't
have
the
permissions
to
ping
either
me
or
Matt,
the
Leo
there
and
you
haven't,
created
a
1.7.
You
all
right
ignore
that
we'll
figure
that
out
later.
C
E
C
C
C
And
both
of
these
are
going
to
GA
right,
yeah
school
scheduling
by
note
capabilities.
I
think
this
is
a
feature
that
we're
going
to
kind
of
piggyback
off
of
CSI
I'm,
not
sure
we
want
to
design
it
separately,
I'm
going
to
be
prioritize
it
for
1.4,
for
this
quarter,
I'm
going
to
double
check
with
michael
rubin
to
make
sure
he's
okay
with
it
as
well.
But
if
anybody
has
any
comments
on
it,
let
me
know.
C
C
I
A
E
C
C
C
C
C
C
C
D
B
60
miles
actually
interesting.
Can
you
link.
D
C
So
then
I'll
keep
the
the
line
item
there
and
then
we
can
follow
up
on
it
later
and
see
if
anybody
is
actually
committed
to
working
on
it
sure
I
imagine
this
work,
it's
probably
going
to
be
led
by
the
Federation
team.
So
maybe
somebody
on
the
Federation
team
decided
this
is
something
they're
going
to
pick
up.
C
C
C
A
C
C
A
C
C
C
A
C
Future
yeah
I
mean
that
was
literally
gutting,
everything
and
separating
it,
and
that
was
like
replacing
the
entire
chorus
of
speed
storage
stack
looking
at
this
it's
large,
but
it's
not
nearly
that
large
market
related
mostly
to
the
attached
to
touch
controller,
not
into
the
volume
manager
or
any
of
the
other
volume
code.
So
it's
big
but
and
English.
A
C
C
C
C
A
H
So
I
think
we're
basically
at
the
final
round
of
reviews.
So
if
there's
still
any
concerns
that
I
haven't
addressed,
please
ping
me
about
them.
I,
make
sure
to
respond
to
everything.
So
if
I
haven't
responded
to
one
of
your
questions
that
I
probably
missed
it.
So
just
pay
me
again,
but
we
would
really
like
to
get
some
get
to
some
consensus.
You
know
which
in
the
next
week
or
so
on
this
so
that
we
can
start
implementing
for
alpha.
H
A
H
Cool
yeah,
so
so,
basically,
I,
think
allow,
though
young
gave
some
suggestions
about
how
to
handle
the
block
devices.
His
idea
was
basically
have
you
can
have
all
the
discs
available
as
a
block
resource
and
then,
depending
on
how
the
user
wants
to
use
that
they
can
decide
at
the
time
they
create
the
pod
if
it
wants
to
be
a
blocked
if
they
want
to
use
as
a
block
device
or
if
they
want
to
use
it
as
a
as
a
file
system.
H
Part
yeah
so
add
that
still
needs
to
be
worked
out.
I
think
we
we
added
in
you
know
this
kind
of
high-level
proposal
for
block
devices
just
kind
of
as
to
cover
all
the
bases,
but
we
haven't
actually
looked
into
the
details
of
how
we
would
actually
you
know,
access
the
block
devices.
I
did
I,
did
a
quick
look
at
how
docker
does
it
and
docker
does
have
a
a
device
option?
H
That's
so
that's
separate
from
the
volume
mounts
they
have
a
device
option
where
you
can
pass
an
actual
block
device
there.
So
if
we
do
want
to
handle
the
block
support
in
the
future,
I
think
we
would
have
to
go
through
some
interface
like
that,
then
the
question
there
is,
you
know
if
we
have
to,
we
have
to
do
everything
through
CRI,
so
then
the
CRI.
H
A
So
I'm
stuff,
like
that
why
this
is
always
confused
me
about
the
Cooper
Nettie's
process,
like
you
know,
I,
there's
kind
of
like
this.
This
urge
to
have
designed
a
feature
but
then
kind
of
road
map
it
out.
We
want
to
do
less,
I
mean
like
I,
don't
mind
if
we
just
say
like
hey
we're
not
going
to
implement
this.
We
thought
about
it.
We
haven't,
but
if
we
specify
this,
then
we're
going
to
modify
the
PVC
to
include
block
or
files,
it
kind
of
seems
like
we're,
complying
that
it's
going
to
be
partially
implemented.
J
C
Don't
think
that's
what
the
intention
here
is
from
what
dish,
and
michelle
has
been
telling
me
I-
think
what
they're
trying
to
do
is
get
a
proof-of-concept
out
the
door
as
fast
as
possible,
because
what
they've
learned
is,
you
could
spend
just
months
and
months
on
end
trying
to
design
the
perfect
design,
but
when
you
start
actually
implementing
it,
you're
going
to
discover
a
lot
more
issues,
so
they
want
to
get
some
sort
of
prototype
working
and
then
help
that
lead
the
design
rather
than
kind
of
design
in
isolation.
If
that
makes
any
sense.
H
H
A
H
That
would
be
good.
It
would
be
good
to
know
you
know
what
potential
use
cases
we
would
have
I
think
maybe
I
guess
somewhere,
maybe
at
the
top
of
the
doc.
We
should
mention
that
this
is
kind
of
like
a
multi
release,
vision
document
more
than
you
know
what
we
are
currently
doing.
We're
currently
planning
on
doing
for
the
immediate
alpha
feature.
I
H
I
H
A
C
It
yeah
I'll
take
a
look
at
this
PR
in
general
for
the
approver
issue.
I
think
that's
kind
of
a
issue
across
the
project.
We
just
need
to
get
better
about
making
sure
that
the
correct
directories
have
the
correct
owners.
I
think
PR
is
like
this
tend
to
be
problematic
because
they
spew
across
multiple
different
directories,
I'm
looking
at
it
right
now,
you
got
stuff
in
API.
C
You've
got
stuff
all
over
the
place,
so
usually
the
Hoover
bot
won't
be
happy,
because,
even
if
you
get
approval
from
someone
who's
in
one
directory,
they
probably
are
not
going
to
have
approval
somewhere
else.
Unless
you
get
approval
from
someone
who
is
very
low
in
the
tree
or
from
a
bunch
of
people
that
have
approval
across
a
bunch
of
different
directories,
I
think
it's
just
the
nature
of
these
large
PRS.
A
J
I,
this
is
just
a
reminder
that
recently
the
location
was
entered
there,
there's
a
number
of
daily
MC
building
from
the
area
and
it's
based
on
head
out.
So
I
think
it's
currently
moved
to
great
america
parkway
vs
mission
mission.
Something
is
where
it
was
before:
I
think
they're
in
the
same
general
neighborhood,
and
then
please.
J
Again,
I
was
thinking
we
do
the
standard
zoom,
although
we
have
to
use
a
laptop
with
a
camera.
I
did
not
the
only
conference
rooms
that
we
have
meaning
Delhi
MC
with
cameras
are
set
up
to
use
the
cisco
telepresence
wing
and
you
have
to
have
cisco
at
the
other
end,
to
get
it
to
work.
So
they
don't
wire
the
conference
rooms
with
generic
webcams.
C
C
C
A
C
To
our
backlog
over
here,
stabilizing
Azure
support,
so
I
recall,
there
was
a
bunch
of
folks
that
I've
been
working
on
the
azure
plugins,
who
have
had
I,
guess
issues
with
the
core
code
and
they
wanted
to
make
changes
there.
My
belief
Seth
was
one
of
the
people
working
on
that
I
wanted
to
have
a
line
item
here
so
that
we
actually
carve
out
space
and
people
to
review
this.
Do
we
have
anybody
working
on
the
asher
plugins?
You
want
to
comment
on
this
or
is
interested
in
driving
it
well.
C
C
Awesome
going
to
college
EA,
but
it
doesn't
really
mean
anything
in
this
case
a
priority,
we're
going
to
put
this
to
and
will
be
a
single
release
again,
I'm,
not
sure
if
it's
worth
tracking
as
a
separate
feature,
we
didn't
do
that
for
you
WSS.
We
won't
do
a
treasure.
Second
item.
I
added
here
was
PVC
capacity
usage
stats.
This
is
something
that
Jiang
Xiu
on
our
side
was
requesting
on.
C
The
idea
here
is
that
when
you
lift,
when
you
list
the
tvs
or
sorry
PDF
TVs
using
a
cute
little
command,
people
get
TV,
it
should
be
able
to
give
you
what
the
usage
capacity
if
those
tvs
is
be
a
pretty
sweet
feature
to
have
doing.
I'm
not
sure
how
committed
you
wear
to
working
on
it.
This
milestone,
you
have
any
thoughts
on
that.
E
C
C
Even
if
it's
bug
level
work,
if
it's
tracked,
it's
a
lot
easier
than
if
it's
kind
of
add
hot.
So
if
anything
comes
up
in
the
quarter,
feel
free
to
ping.
This
meeting
or
ping
you
directly
and
say
here
is
something
that
we
should
probably
be
tracking
and
as
we
go
through,
the
rest
of
the
quarter
will
probably
just
look
back
at
this
dis
planning,
doc
and
use
it
as
as
the
reference
point
to
checking
on
status
for
things.
I
Hi
this
is
Lisa.
I
just
added
a
section
at
the
end,
where
I'm
just
requesting
the
ability
to
take
the
master,
which
is
a
project
that
we
talked
about
about
I,
don't
know
three
or
four
weeks
ago,
both
the
architecture
of
it
and
how
it
allows
you
to
install
solar
systems
on
to
communities
and
I
would
like
to
move
it
to
Cumbres
incubator
and
I'm
following
the
incubator
community
website
scription,
and
they
says
that
it
would
need
to
get
acceptance
by
sick
first.
So
this
is
my
first
project
moving
this
into
the
Creator.
C
A
So
when
the
external
parishioner
was
a
little
more
obvious,
I
think
everybody
was
fairly
familiar
with
it
I
you
are
hitting
the
right
process
of
Mozilla
to
have
to
have
a
signature
support
and
then
a
champion
from
well
probably
deeper
down
the
creek
or
deeper
down
this
source
tree
than
anyone
on
this
call
right
now
to
champion
it,
but
I
can
help
you
with
that
process.
I
think
that
the
main
point,
though,
is
it
needs
to
be
socialized
with
the
sink,
and
we
have
to
have
a
pretty
good
understanding
of.
A
I
G
G
I'd
like
to
understand
from
six
storage
point
of
view,
if
there
is
a
desire
to
in
some
way
integrate
test,
infra
with
and
CI
for
out
of
tree
stores,
in
other
words,
the
proven
as
community
value
being
able
to
claim
that
a
certain
release
of
cumin
at
ease
works
with
the
following
external
storage
providers
and
some
lists,
because
we've
run
some
level
of
end
and
testing
for
out
of
tree
providers.
Is
that
even
desirable
for
cuva
nettings
to
be
able
to
do
that
and
make
that
kind
of
clean.
A
G
Only
ones
today
you
know
we
have
19
plugins
or
so
and
they're
all
intrigued
and
they
all
get
tested
based
on
end-to-end
tests
and
recently
you
know,
there's
a
lot
of
new
test
added
for
vsphere
and
that's
great.
Some
provide
some
have
more,
some
have
fewer,
but
we
do
cover
it.
When
we
move
to
out
a
tree,
those
most
of
those
19
providers
will
now
be
out
of
tree
and
it
will
take
deliberate
effort
to
be
able
to
run
those
tests
in
an
automated
fashion.
And
do
we
want
to
do
that.
A
Yeah
I
think
thing
we
do
I
yeah
I
mean
I.
In
my
mind,
is
absolutely
we
want
to
do.
Testing
I,
like
you,
said,
I,
don't
think
I
understand
what
the
design
for
wouldn't
quite
look
like.
Quite
yet,
but
if
your
proposal
is
that
you'd
like
to
propose
a
design,
I
think
just
yeah
we'd
be
more
than
happy
to
review
it
and
consider
with
high
priority
one.
D
Problem
is
the
tests
that
are
added
like
19
plugins,
that
is
Ted
just
just
like
vsphere,
we
don't
run
those
tests
as
part
of
the
CI
process.
Likewise,
kasi
tests
are
not
running
policy
a
process
because
if
it
does
exist,
but
we
don't
have
the
setup
that
how
communities
has
doesn't
run
doesn't
have
this
here,
it
doesn't
have
I
discusses
it.
Those
tests
are
not
run
actually,
so
one
calendar
is
like
finding
out
the
setup
that
is
required
to
run
those
specific
volume,
plugging
set
a
test,
and
that's
the
test,
this
one
big
challenge.
D
G
G
It's
not
clear
to
John
and
I
have
been
thinking
about
this
problem.
How
we
do
that
when
a
test
is
tagged
with
a
feature
it
gets
skipped
by
CI,
but
my
understanding
can
talk
to
mystic
testing
earlier
and
peak.
Was
that
even
though
it's
it's
skipped
as
part
of
the
pr
acceptance
process,
that
featured
tests
do
get
run
at
some
point
and
and
if
they
fail
they
fail
and
then
it's
just
the
standard
process
for
dealing
with
a
failed
test.
G
A
G
D
Individual
providers
who
have
added
that
plugins
are
what
adding
the
plug
in
it
that
willing
to
donate
hardware
or
willing
to
like
have
some
sort
of
hook.
We
talked
about
it
before
with
the
tests
are
occasionally
done
on
that
particular
infrastructure
and
and
the
results
but
external,
or
whether
it's
posted
back
to
github,
so
that
I
mean
right
now.
These
fear
tests
are
not
run
at
all,
actually,
not
even
as
part
of
features
thing
or
anything.
It's
just
this
lion.