►
From YouTube: Kubernetes SIG Storage 20170706
Description
Kubernetes Storage Special-Interest-Group (SIG) Meeting - 06 July 2017
Meeting Notes/Agenda: https://docs.google.com/document/d/1-8KEG8AjAgKznS9NFm3qWqkGyCHmvU6HVl0sk5hwoAE/edit#heading=h.e2xu3sl4x6bo
Find out more about the Storage SIG here: https://github.com/kubernetes/community/tree/master/sig-storage
Moderator: Saad Ali (Google)
Chat Log:
[none]
A
All
right
welcome.
Everyone
today
is
July
6
2017.
This
is
the
bi-weekly
meeting
of
the
storage
special
interest
group
for
kubernetes.
As
a
reminder,
this
meeting
is
public
and
recorded
today
on
the
agenda,
we're
going
to
go
over
b2b.
One
point:
eight,
two,
three
planning
the
second
round
of
this.
We
did
the
first
round
last
time.
A
If
you
have
any
PR
that
needs
attention
on
the
wider
community
based
into
the
agenda
duck
and
we're
going
to
go
over
some
design
reviews,
we've
got
two
here
with
not
very
much
information
we
can
go
over.
Those
then
brad
has
an
item
to
go
over
the
external
storage
repo
and
then
Tushar
added
an
item
about
updating
on
how
to
improve
collaboration
between
multiple
groups.
This
is
an
issue
that
I
think
our
community
is
starting
to
run
into
now
that
we've
matured
and
I
think
we've
got
more
than
50
people
showing
up
to
this
meeting.
A
So
we
need
to
make
sure
that
we
remain
transparent,
open
and
collaborative
so,
first
off.
Let's
kick
it
off
with
the
planning
spreadsheet
we
went
through
and
added
items.
If
you
think
there
is
an
item
for
1.8
that
we
should
be
working
on.
That's
not
on
this
spreadsheet.
Please
add
it
to
the
agenda.
We
could
review
those
at
the
end.
In
the
meantime,
let's
just
go
through
these
line
by
line
CSI
volume,
plugin
I
think
priority
here
is
valid
one.
It's
a
multi
release
a
multi
quarter,
a
project
for
this
quarter,
we're
targeting
and
design.
A
A
We
want
at
least
one
person
as
owner
and
reviewer,
but
obviously
these
are
going
to
be
multiple
people
working
on
it.
So
the
purpose
of
this
is
to
at
least
have
one
person
assigned
to
each
and
then
you
can
grow
from
there.
We
just
want
to
make
sure
that
there's
someone
that
we
can
talk
to
usually
I.
A
All
right,
that's
the
isolation
resource-management.
This
is
something
that
Jing
is
brought
to
alpha
in
1.7.
She
is
going
to
continue
to
drive
it
and
what
they
we
already
have
a
reviewer.
If
anybody
else
wants
to
work
on
that,
please
reach
out
to
Jing
local
persistent
storage
support.
This
is
something
Michelle
brought
to
alpha
and
1.7
we're
going
to
continue
to
drive
it
Michelle
drive
it.
A
A
Block
volumes
support
Erin
who's
driving
this
conversation
as
well
and
looks
like
we
have
a
couple
other
owners
assigned
as
well
and
reviewers
look
good.
This
is
a
p1
for
our
group
for
this
quarter.
I
think
that
makes
sense
targeting
alpha
for
this
milestone
seems
reasonable.
The
conversations
in
the
electrical
so
far
enable
containerization
of
mount
dependencies.
I
think
the
priority
is
correct
here.
One
Yan
has
a
proposal
out,
there's
already
a
feature
open
for
it.
Humble
has
agreed
to
review
it.
I
should
beneath
that
refactor.
A
A
A
A
A
A
A
A
D
A
A
I
can
hear
you,
it's
not
I
feel
like
you
might
be
overloaded
this
quarter,
but
thank
you.
A
A
Improving
the
trimmer
tack,
I
haven't
had
a
chance
to
talk
to
Jing
Xin
just
got
back
from
vacation,
so
I
need
to
sync
up
whether
there's
any
work
that
still
needs
to
be
done
here.
So
for
now,
we're
going
to
leave
this
empty
and
also
the
scent
afterwards
fix,
broken
vol
reconstruction
Chakri
is
working
on
this
for
1.8
priority
seems
reasonable
any
reviewers
for
this
work.
This
is
making
changes
to
the
quality
and
manager
code
into
blip.
A
A
A
A
A
A
And
everything
else
looks
good
here.
Next
up
we've
got
a
p3
which
is
PVC
capacity,
usage
stats,
it's
nice
to
have
user
friendly
feature,
I,
don't
think
we
anybody
signed
up
to
own
it.
Yet
does
anybody
want
to
sign
up?
If
not
it's
a
p3?
We
can
leave
it
on
a
sign
for
now
and
you
can
take
it
up
later
in
the
quarter
needed
I.
D
A
So
we're
going
to
need
a
feature
open
in
the
future
repo
for
this
I'm,
going
to
leave
that
red
for
now
and
I'll
leave
it
up
to
you
to
create
that
as
well
data
replication
Brad,
this
is
assigned
to
you.
We've
made
it
a
rate
priority,
so
I
don't
think
it's
worth
assigning
a
reviewer
at
this
point.
We
don't
really
expect
much
traction
on
this
disk
order.
But
if
something
happens,
you
can
keep
us
updated
and
update
this
accordingly.
Okay
support
attach
detach
for
rwo
volumes
such
as
I
scuzzy,
fiber
channel
in
our
BD.
F
A
I'm
going
to
assign
myself
as
reviewer
for
this
because
I
don't
quite
understand
what
attach
detach
for
ice
code,
you
might
look
like
just
a
budget
will
attach
to
watch
this.
Okay,
I
think
the
confusion
that
I
have
is
my
understanding
was
that
requires
code
to
be
executed
on
the
node
we
so
I,
guess
that's
where
the
disconnect
is,
but
we
can
follow
up
offline.
A
B
A
That's
mine,
let's
call
it
beta
and
let's
just
make
sure
that
any
changes
the
API
are
not
permanent
until
we
go
to
GA,
but
it
sounds
like
it's
very
reasonable
and
priority
on
this.
We
want
to
call
it
too
sure.
Okay
should
be
these
worthy.
It
does
need
a
feature,
repo
issue
and
that
I'll
leave
that
up
to
you
open
and
who
wants
to
be
a
reviewer
on
this
I
can
do
that.
So
this
is
Jeff
all
right,
cool
thanks,
Jeff,
all
right
cool.
A
A
A
F
A
Okay,
so
but
I'm
going
to
go
ahead
and
add
this
to
the
list
of
items
to
track
for
1.8
and
okay
sign-u
as
the
as
the
owner
for
this,
and
so
you
we're
going
to
make
sure
that
this
gets
in
for
1.8.
You
just
need
to
make
sure
that
you
follow
the
process
and
create
documentation
and
all
that
kind
of
thing.
So.
A
B
F
B
F
A
Suppose
you're
going
to
need
to
open
a
feature
bug
to
track
this.
This
is
in
the
kubernetes
flash
features
repository
you
just
go
in
there
create
an
issue
and
there's
a
template
for
the
work
that
needs
to
be
tracked
when
you've
got
that
issue
opened
up
just
link
to
it
here,
and
this
red
coloring
will
go
away.
A
A
I
This
is
steep
yeah,
so
we
did
get
a
little
feedback
from
the
community.
Thank
you
for
that,
and
and
we
do
have
a
planner
option,
so
we
are
probably
not
going
to
do.
They
use
this.
The
new
CRD
way
to
to
do
the
volume
attached
amount.
We
had
a
good
conversation,
we'll
lick
it
from
the
API
team
and
I
think
we
are
yeah.
We're
going
to
be
doing
two
other
mechanics
in
like
perfect
Maps
I'll
update
this
PR.
With
these
changes
and
and
I'll
let
a
bombing
against
you
see
the
reviewer
I
know.
A
I
So
we
wanted
to
do
the
Rivoli
plugging
in
a
to
open
in
a
slightly
different
way.
We
wanted
to
make
it
more
of
a
controller
basis.
Everything
kubernetes,
more
controllers,
great,
where
you
do
you
set
a
desired
state
and
there's
something
else,
something
there
that
watches
it
and
then
you
know
and
make
that
a
state
happen,
and
we
wanted
to
have
CPR
in
this
case
CRD
to
define
the
state.
So
basically
the
holy
plugin
move
very
same
will
be
just
the
main
purpose
of
it
would
be
just
to
create
a
CRD.
I
There
was
some
contention
about
things,
because
the
Rupali
forgave
lists
in
core
and
because
see
are
these
the
custom
resource
definition,
our
api
expression
point.
We
are
not
really
part
of
possible
and
when
you
have
something
in
core
which
is
the
volley
plugin
calling
something
that
is
extension,
it's
been
considered
as
the
layer
violation.
If
so
that
that's
that's
pretty
much
like
the
dilemma
right
now.
So,
okay.
A
I
We
are,
we
are,
we
are
where
we're
exploring
doing
a
config
Maps.
We
don't
want
to
do
a
blocky
API,
because
we
want
to
take
advantage
of
all
the
things
that
community
has
in
terms
of
discovery,
city,
security
and
authentication.
So
if
we
have
a
blocking
API,
we
need
to
specify
up
with
the
user.
We
have
to
specify
the
endpoint
user
secrets
and
all
that
stuff,
and
we
don't
want
to
kind
of
avoid
that.
So
that's
what
we
wanted
to
do.
We
try
to
do
the
CR
disk.
First.
F
I
We
do
something:
if
use
a
complete
mapping
set,
then
we
don't
have
to
worry
about.
The
user
has
whatever
in
an
end
point
exactly
what
I've
got
going
so
possible,
so
that
will
be
managed
by
kubernetes
up
well,
if
the
end
of
the
ankle
is
his
actually
to
make
the
user
experience
better,
I
guess:
okay,.
A
Yeah
so
you're
going
to
hate
me
for
this
I
think
we
pushed
back
on
previous
volume
plugins
that
try
to
write
data
into
a
config
map
scale
like
scale
IO
was
one
I.
Think
Vlad
was
the
author
of
that
yeah.
F
A
But
we
could
take
that
discussion.
Offline,
I,
think
it's
better
than
during
a
CRD,
but
we
can
take
that
offline.
Okay,.
I
A
A
A
J
To
it
sure
I've
been
getting
so
the
statuses
we're
very
briefly,
is
I've,
been
getting
good
support
from
Michelle
I
think
she
might
be
off
work
a
little
bit
this
week
or
all
this
week
you.
J
So
it's
stalled
some
but
I've
been
addressing
her
comments
as
well
as
other
reviewers,
and
it's
just
it's
just
sitting
there
waiting
for
final
approval.
Ok,.
H
Yeah
kami,
yes,
Jem
hi
everyone.
This
is
Leone
and
I
work
from
oh
hi
Jackie.
So
about
this
proposal,
my
colleague
Howard
has
given
a
simple
introduction
on
the
last
meeting
and
today
I
will
give
our
folks
more
on
the
use
case
and
some
basic
imputation
facade.
Can
you
table
open
the
file?
The
pol?
Okay,
not.
A
H
Okay,
so
about
this
proposal,
so
about
the
motivation,
so,
as
we
all
know,
kinetic
energy
pod
anywhere,
including
because
person
than
they
are
metals,
but
right
now,
as
for
story
result,
if
you
have
to
choose
specific
story
back
and
according
to
the
environment
community
to
be
deployed
here
are
some
example
and
about
the
use
case.
When
use
case,
is
we
found
less
some
story
system
the
legs
Inga,
the
guy
Oh
can
provide
volume
resource
for
permit
hood
directly,
and
so
in
this
case,
how
can
you
that
use
these
resources?
H
If
the
kinetic
cluster
is
deployed
on
this
pair
metal
and
the
another
use
case
is
that
we
got
some
testing
from
users
that
we
should
be
rich
canary
stories
ever
since
they
were
using
some
story
system
with
on
support
and
then
always
have
a
oxygen
which
is
platform
in
part.
I
think
it
would
be
a
hard
work
for
the
to
develop
every
store
ever
because,
as
we
know,
the
operational
attaching
volume
to
host
is
cloud
a
complicated,
so
I'm
wondering.
Can
we
share
the
compact
implementation
and
the
providers
simple
interface
for
their
technology
development?
H
We
can't
just
call
this
leverage,
finish,
volume,
discovery
and
the
amount
of
the
device
class
to
container,
and
we
also
won't
want
this
project
to
serve
for
flex
volume
and
the
stairs
as
again
and
even
other
any
other
storage
system
routine
go
the
here
next
are
some
proposed
design
so
because
there
are
a
lot
of
story
protocols
and
some
of
them
are
implemented
in
different
ways
according
to
different
system
types
and
operating
system
types.
So
so
it
will
be
very
complicated
if
we
directly
added
this
device
drivers
into
volume
plugin.
H
So
we
want
to
create
a
library
to
communicate
with
kernel
and
expose
or
unifying,
is
to
volume
packing,
and
here-
and
next
are
some.
If
you
have
objects
in
first
week,
we
we
will
design
some
constants,
which,
which
includes
some
platform
tabs
and
operating
system
types
and
and
the
next
are
some
device.
Travel
types
and
and
the
connector
is
a
interface
that
indicate
what
also
the
world
can
do.
H
With
this
lever
and
the
in
this
in
this
interface,
you
can
see
we
we
will
design
four
methods
in
the
first
stage
and
the
connector
process
is
used
to
tell
story
tech
and
how
to
initialize
national
volume
and
as
forth
next
information
it
can
use
you
can
use
to
connect
a
volume
and
in
this
drug
we
have
travelling
pipe
and
under
map
called
connection
data,
which
is
which
differs
from
gravel
volume
volume
travel
volume
type,
for
example,
a
scrappy
driver.
You
can
use
SS,
escapa
connection
data
here,
and
here
is
the
summary
of
the.
H
B
H
Well,
because
ye
OpenStack
and
they
have
a
project
called
Westrick
to
do
some
host,
decide
one
discovery
and,
for
example,
if
you
want
to
attach
a
block
device
to
the
pair
metal
you
have
to.
You
have
to
do
a
lot
of
things
and
such
as
such
as
dog,
in
the
car
target.
Long
and
and
then
and
the
connect
connects
the
volume
could
using
the
initiator
and
and.
B
H
C
F
C
Sir?
Can
you
please
say
it
again?
Well,
it
sounds
like
what
what's
being
proposed
is
a
volume
plug-in
portal
or
interface
from
kubernetes
to
this
external
entity.
This
host
volume
discovery
it
potentially
has
more
than
one
storage
together
attach
step
at
the
exterior.
So
the
way
this
likes
me
is
being
something
similar
to
CS.
I
know.
H
H
D
I
think
what
this
is
trying
to
do
is
like
an
adapter,
for
example,
if
you're
trying
to
use
time
to
mount
OpenStack
Center
on
a
bare
metal
cluster,
then
you
have
to
consider
flow
providers
cloud
providers
to
be
able
to
use
OpenStack
standard
volumes
either
way
it
won't
work.
So,
but
if
we
use
this
adapter,
then
we
don't
need
to
configure
like
a
cloud
provider
for
the
subscriber
cluster
and
we
can
go
through
this
adapter
in
states
like
like
that.
G
D
H
C
H
J
F
A
H
A
So
I
think
the
way
that
I
try
to
simplify
everything
in
kubernetes
storage
side,
because
I'm,
not
a
very
smart
guy
and
for
me,
volume
plugins
in
kubernetes,
are
basically
a
mechanism
for
consuming
different
types
of
storage
systems
and
so
far,
what
we've
done
with
a
volume
plugins
is
that
for
each
type
of
storage
system
you
get
one
volume
plug-in.
It
makes
it
a
very
nice,
simple
and
easy
user
experience.
A
It
sounds
like
it's
not
just
about
consumption
of
storage,
but
also
making
local
storage
available
yeah,
and
we
do
have
a
local
storage
project
that
is
currently
underway,
which
is
designed
to
make
global
storage,
expose
local
storage
to
to
kubernetes
end-users
and
I.
Wonder
if,
rather
than
this
being
a
completely
separate
project,
that
kind
of
competes
with
that,
maybe
we
could
roll
some
of
the
ideas
of
this
into
that.
H
A
What
is
what
was
your
end
goal
with
this
presentation?
Is
it
to
add
a
new
volume
plugin,
or
is
it
what?
What
is
your
intended
end
goal
I.
H
A
A
So
then,
maybe
we'll
point
Michelle
at
this
and
see
if
she
can-
or
maybe
you
can
reach
out
to
Michelle
her
alias
is
Anna,
hey
you.
She
is
right
here
somewhere
MSA,
you
42
and,
and
you
can
take
conversation
offline,
okay,.
A
A
A
All
right
we're
moving
it
last
item
on
the
agenda
is
to
char
on
the
line
he
isn't,
but
so
I
actually
added
this
on
his
behalf
last
time,
and
it
was
based
on
a
comment
that
he
mentioned,
he
was
trying
to
collaborate
on
the
snapshot
design
and
it
was
moving
very
very
quickly
and
when
you
comment
on
something
the
design
doc
would
be
out
of
date
and
he
was
acting.
What
a
better
way
to
collaborate
with
me
and
I,
hear
that
and
I
don't
want
it
to
be
brushed
under
the
table
and
ignored.
A
I
think
this
community
is
growing
very
quickly.
There
is
a
lot
of
people
that
attend
very
regularly,
but
I
think
there's
only
a
handful
of
people
who
contribute
regularly
and
I
don't
want
everyone
to
feel
like
they
are
not
part
of
this
community.
They
should
be.
Whoever
wants
to
contribute
should
be
able
to
contribute.
A
A
K
A
A
A
L
Going
into
this
world
I've
been
listening
in
the
last
few
weeks,
I'm
a
chemical
Forester,
and
we
have
a
kind
of
an
occupied
great
version
of
AFS.
If
you
promote
the
Andrew
file
system
and
we're
starting
to
architect
how
to
connect
you
into
kubernetes,
and
so
really
basic
question
is
and
is
that
what
Walters
the
open
source
open
s
world
war
a
closed
source?
You
know,
so
we're
not.
You
know
to
throw
that
out
there.
What
do
you
recommend
is
the
best
way
to
hold
our
sin.
L
A
A
B
Sorry,
my
internet
connection
been
choppy
I.
So
if
you
know
anybody
has
questions,
comments
are
or
just
trying
to
get
started
in
the
community,
and
they
don't
feel
comfortable
emailing
like
the
broader
list
with
questions
I
know
myself
and
you
know
others
are
ahead-
are
always
want
to
share
for
people
or
to
spend
time
with
them
to
you
know,
help
bring
them
up
to
speed.
They
just
want
to
reach
out
directly
and
try
to
ask
direct
questions
or
to
get
the
lay
of
the
land
we're
more
than
help
at
a
24
the
right
direction.
B
L
L
L
D
That
was
the
upstream
the
other
side.
Never
the
document
to
that
documents
actually
like
the
all
wall
and
plugins
and
data
interface,
is
what
they
look
like
not
on
to
find
the
link
I'll,
probably
post
it,
making
it
considered
by
some
someone
in
the
community.
That's
pretty
good
starting
point
to
learn
the
basics
of
like
how
does
how
the
whole
public
wall
of
Dragon
ecosystem
works
and.
A
I'll
be
great,
and
then
we
can
also
go
in
and
update
the
sig
storage
community
page
and
as
the
presentation
like
that,
I
think
I
did
that
at
some
point,
but
may
have
gotten
wiped
out
recently
I'll
go
and
take
a
look
at
that
cool,
I
post
very
much.
For
your
time.
Those
of
your
America
has
a
soup
of
July
and
have
a
good
week
we'll
reconvene
in
two
weeks.