►
From YouTube: Kubernetes SIG Storage 20170202
Description
Kubernetes Storage Special-Interest-Group (SIG) Meeting - 2 February 2017
Meeting Notes/Agenda: https://docs.google.com/document/d/1-8KEG8AjAgKznS9NFm3qWqkGyCHmvU6HVl0sk5hwoAE/edit#heading=h.rlll9qe7mmj6
Find out more about the Storage SIG here: https://github.com/kubernetes/community/tree/master/sig-storage
Moderator: Saad Ali (Google)
Chat Log:
09:14:55 From rkouj : pls mute if you are not speaking
09:34:03 From Christopher M Luciano : I will be there
09:34:07 From Luis Pabón : I probably be there
09:34:22 From Luis Pabón : I probably [will] be there
A
So
this
is
the
bi-weekly
meeting
of
the
Ku
burnetii
storage.
Special
interest
group
today
is
februari
to
2017.
As
a
reminder,
this
meeting
is
look
and
recorded.
We
have
a
lot
of
items
on
the
agenda
today,
starting
off
with
status
updates,
1.53
and
1.49,
or
plan
to
go
out
at
the
end
the
next
week,
februari
10.
If
you
have
anything
that
needs
to
go
into
either
those
releases,
make
sure
to
get
them
cherry-picked
as
soon
as
possible
for
1.53
assign
them
to
me
for
1.49
assignment
to
jess
any
other
status
updates
from
anyone.
B
So
this
is
jeff.
I
just
wanted
to
say
that
John
and
I
have
been
making
progress
on
and
end
testing
focused
on
persistent
volumes
mostly,
but
we're
also
starting
some
disruptive
steps
that
involve
volume
and
making
sure
the
people
it
gets,
amounts
unmounted
and
cleans
up
stuff
like
that
and
I'd
like
to
just
say
that
I
know
there's
been
work
to
improve
the
pr
of
workflow
and
beat
up,
and
in
my
experience
I
know
the
scene
of
improvement.
So
I
just
wanted
to
let
you
know
that
it's
appreciated
and
I
do.
A
D
C
C
How
to
manage
local
storage?
This
is
for
both
ephemeral
and
persistent
use
cases.
The
document
basically
covers
the
high
level
use
cases
that
we're
trying
to
solve,
with
the
overall
workflows
of
how
you
would
use
features
to
solve
these
use
cases,
and
what
we
would
basically
like
for
everyone
to
do
is
to
read
through
the
document
and
make
sure
that
all
first
we
have
covered
all
of
the
use
cases
and
then
second,
do
our
workflows
aim
to
you
know,
be
able
to
cover
the
youth
pieces
in
a
good
way.
C
We
had
a
few
questions
open
questions
about
some
of
the
use
cases,
namely
so
the
at
least
for
persistent
storage.
The
main
use
cases
that
we've
been
targeting
are
distributed,
file,
systems
and
databases
using
stateful
sets.
So
we
would
like
to
hear
from
anyone
if
they
have
any
use
cases
that
don't
use
full
sets.
The
second
use
case
we
weren't
sure
about
is
for
ephemeral
storage.
Is
there
a
need
for
a
femoral
storage
to
be
able
to
have
dedicated
SSD
disks
and
then
the
third
use
case
we're
looking
for
something
back
on
our
level
storage?
C
E
One
quick
question:
John:
well:
albion:
the
supportive
in
the
position,
chemical
storage,
right.
C
Now
our
proposal
does
not
include
lvm,
mainly
because
cases
that
we've
been
targeting
and
the
people
that
we've
talked
to
that
what
they
really
want
is
dedicated
disk
and
I.
Guess
one
of
the
things
with
LV
on
is
that
it
will
be
difficult.
If
her,
you
know
dynamically
partitioning,
the
storage,
then
it'll
be
difficult
to
guarantee
the
consumers.
Will
people
get
dedicated,
mistress
and
get
the
dedicated
I
ops,
I.
E
Think
about
well
on
some
promotion
also
have
cell
signal
this
case
or
the
crust
I
think
he
has
a
issue.
The
issue,
a
device
back
available.
I
can
just
link
that
to
your
on
pure
okay,
look
coupled
up
yeah.
C
Great
yeah,
if
you
yeah
just
you,
know,
comment
on
the
PR
with
any
of
the
links
to
issues
you
may
have.
If
you
have
more
use
cases
that
we
haven't
covered
here,
then
we
want
to
know
about
it
as
well.
Right.
E
F
So
to
be
discussed
us
like
there's
a
need
of
supporting
mount
options
in
volume,
plugins
and
exposing
that
as
an
API
at
API
level,
so
that
users
can
configure
options
they
want.
I
have
opened
a
proposal,
design
proposal.
There
are
few.
There
is
at
least
one
problem
that
is
kind
of
a
are
still
exploring
an
option
how
to
solve
that
properly,
which
is
like
how
to
disallow
options
that
are
not
supported
or
are
problematic.
So
we
are
considering
several
options
like
blacklisting.
F
F
A
F
Oh
I
open
the
proposal
for
like
we
have
had
problems
that
we
had
like
a
double
oscoda
being
blown
up
and,
and
users
are
noticing
it
quite
later,
and
at
that
point
they
might
already
get
locked
out
of
their
own
clusters,
so
we
want
to
basically
provide
matrix
so
that
people
can
see
how
company
is
using
the
cloud
provider
ETA.
What
is
the
latency
uncertainly
take
calls
like
wonderful
evening
for
thing
is
taking
time
or
so
yeah.
There's
a
proposal
out
for
those
matrix
people
have
already
a
lot
of
developers
have
commented
on
it.
F
It's
it's
a
it's
in
good
shape,
though
I
think
the
and
I
have
already
written
it.
Do
you
see
that
kind
of
works,
but
yeah
anybody
has
comment
currently
I'm.
Only
covering
storage
calls.
It
could
be
extended
to
cover
other
API
calls
as
well
and
I'm.
Only
considering
GC
and
AWS,
so
in
I
think
I
can
open.
A
F
A
B
G
Just
went
out
a
few
more
points
on
the
local
storage
design
that
mission
already
vent,
or
basically
there
are
a
few
open
discussion
or
like
questions
of
the
questions
that
we
have
for
Stella
PR,
and
we
need
feedback
on
that.
G
A
A
H
A
H
I
I
H
I
So
this
project
is
something
that
we
started
chorus
and
we
want
to
get
your
opinion
because
we
want
to
see
if
this
is
something
of
a
continuance,
something
that
this
sake
thinks
that
it's
a
good
idea
and
something
that
the
SEC
will
also
like
to
participate
in,
because
we
definitely
need
the
community's
help
around
us.
So
what
quartermaster
is
is
really
a
a
pod,
a
container
that
it's
on
communities
and
it
listens
for
banks
and
some
of
the
eventual
going
to
a
little
bit.
I
But
what
it
does
is
it
main
goal
is
to
deploy
upgrades
week
and
do
everything
that
needs
to
be
done
on
a
storage
system
that
is
normally
done
by
an
administrator
and,
for
example,
if
you
know
opentext
wrist
or
if
you
have
Cooper
nation
at
the
moment,
and
you
install
a
an
external
storage
system,
no
one
with
kinetics
of
one
with
OpenStack.
You
have
to
set
it
up
yourself.
You
have
to
go
through
all
the
installations
that
I'll
be
on
to
decide
the
disk
and
so
on.
I
Well
we're
trying
to
do
with
Google
Dennis's
change
that
model
and
be
able
to
deploy
inside
communities
using
those
technologies
that
communities
provides
and
with
quartermaster
we're
trying
to
do
is
leverage
that,
specifically
for
deploying
storage
inside
communities
and
be
able
to
provide
status
and
and
also
create
a
simple
way
to
get
information
from
the
storage
systems
and
deploy
them.
So
right
now,
when
we
deploy
storage
system
onto
communities,
we
have
to
first
start
set
it
up.
We
have
to
find
the
disks
we
want
to
use.
I
We
have
to
deploy
them,
the
storage
we
have
to
initialize.
It
make
sure
it
has
all
the
information
needed
to
be
able
to
satisfy
provisioning
and
then,
and
we
have
to
them
provision
it
whenever
request
comes
in
and
attach
it
to
an
application.
So
the
last
two
are
again
provided
by
google
Nettie's,
but
the
first
two
are
the
ones
I've
provided
by
quartermaster.
So
how
does
it
do
this?
I
If
the
way
does
it
is
that
a
storage
cluster,
a
third-party
resource
is
registered
with
google
Nettie's,
and
what
happens
is
that
the
administrator
define
the
cluster
I
want?
You
know,
whatever
type
of
stories
that
I
want,
it
could
be
set,
it
could
be
glossy,
it
could
be
whatever
container
historicism
they
want
to
deploy
inside
communities,
they
define
the
entire
cluster
and
they
submit
that
to
communities.
Once
that
happens,
quartermaster
then
reach
that
information
looks
at
the
description
and
it
figures
out
that
each
one
of
the
cluster
is
made
of
a
certain
number
of
nodes.
I
So
then
what
quartermaster
does
is
that
it
created
new
object.
New
third
party
resource
object
called
storage
node,
and
it
provides
that
back
to
kinetic
and
then
once
it
does
that
it
takes
that
information
and
actually
calls
the
drivers.
They
have
quartermaster
specific
drivers
to
deal
with
each
one
of
those
systems
to
be
able
to
initialize
the
node
and
the
storage
according
to
the
requested
storage
type.
I
So
the
architecture
is
kind
of
simple
it
has
to
third-party
resource.
Operators
are
called
or
a
way
to
get
a
bench,
and
once
they
get
to
mention
those
and
wants
to
take
action,
it
deals
with
everything
that
Cooper
Nerys
all
the
cool
davis
communication
for
you
for
the
driver
developer,
and
then
it
calls
the
drivers
only
to
special
to
exercise
those
things
that
I
needed
for
that
for
that
storage
system.
I
One
of
the
things
that
also
does
is
that,
because
it
it
leverages
Cooper,
Nettie's
and
its
ways
of
defining
status
for
its
objects,
is
kind
of
you
follow
the
same
line
and
providing
status
for
the
storage,
node
or
storage
cluster.
So
now
an
API
as
an
API.
It
provides
that
information
externally
so
that
clients
can
link
ensuing
it
doesn't
matter
what
now
storage
system
you
have
deployed
in
Tsukuba
net,
one
of
some
other
things
in
the
future
that
that
we
would
like
to
do
with
this.
I
Also
is
the
ability
to
operate
one
of
the
things
that
happens.
A
lot
is
we
operate
companies?
We
all
create
all
these
containers
as
great,
but
then
the
storage
systems
I
have
a
friend
way
of
being
up
created,
and
what
we
would
like
to
do
is
that
we
would
like
the
the
administrator
of
the
client
to
be
able
to
change
the
revision
of
the
image
change
the
image
and
then
that
costs
an
upgrade
and
have
the
drivers
understand
how
to
do
an
upgrade
correctly
for
that
storage
system.
I
Lastly,
is
adding
or
removing
notes,
for
example,
if
you
want
more
storage,
everyone,
less
storage
of
one
one
of
your
notes
died.
How
do
you
do
that?
How
do
you
go
through
that
everything
again
handled
by
quartermaster,
but
we
just
need
to
be
able
to
edit
the
third-party
research
q
coda,
but
I'm
sure
this
other
ways,
so
we
can
go
around
that,
but
goal
of
quartermaster
again
is
to
be
able
to
deploy,
manage,
get
status
and
of
your
storage
system.
I
Again
from
a
simple
point
of
view,
we
don't
want
to
manage
every
specific
thing,
but
from
dusty
I
want
more
storage,
I
want
less
storage.
One
of
my
stories
dies.
What
I
do
now,
and
things
like
that
so
I
have
a
patrick
document
that
I
put
together
and
a
demo
also
that
I
ice
also
recorded
that
you
can
see
and
please
provide
feedback
of
what
you
think.
Please.
Let
me
know
my
emails
at
the
top
I
think
it's
definitely
on
my
architecture
document
and
let
us
know
what
you
think.
I
A
Sorry,
the
definitions
that
you
provide
to
set
up
the
storage
cluster
is
you
they
define
exactly
which
disks
on
the
node
to
use
yeah.
So
that's
a
great
question.
Let.
I
C
I
There's
towards
the
bottom
bottom
there's
a
definition
of
the
storage
cluster
and
here's
an
example
of
a
room
of
one
right
here.
So
bless
it.
Now
again
it
has
to
do
with
what
your
what
the
storage
driver
requires,
for
example,
for
Gloucester.
It
required
raw
disk
which
of
the
raw
disk
on
which,
on
that
note
can
I
take
over
it
could
be
either
you
deploy
a
storage
system
that
requires
file
systems
may
be
on
top
or
require
something
else.
So
it's
up
to
the
storage
system
when
it
gets.
I
A
I
Know
so
that
well
other
things
that
we
going
to
do
too,
if
we,
if
we
see
that
you
know
people
like
this,
and
what
we're
going
to
do
is
we're
going,
you
know,
I
have
it
as
a
POC
right
now
we're
going
to
open
source
it,
and
we
want
like
to
work
with.
You
know:
vendors
or
those
out
there.
That
container
is
their
storage
products
to
be
able
to
easily
deploy
them
into
communities.
Their
usual.
I
A
A
B
B
Well,
so
I
just
know
that
at
some
point
you
know
I
know
and
and
yon
feel
free
to
jump
in
if
you're
still
on
the
call
I
know,
there's
been
tiny
windows
and
issues
with
the
recycler
and
that
there's
been
a
desire
to
deprecate.
It
I
believe
Brad
created
an
issue
or
a
PR
to
do
that,
but
I
think
it's
sold
out.
B
I
know:
Aaron
talked
to
some
open
red
hat
openshift
project
managers
to
see
if
we
had
customers
using
NFS,
relying
on
NFS
recycling
in
the
online
version
of
open
shift
or
in
the
enterprise
version,
and
she
came
back
with
the
answer
that
no,
which
is
a
good
answer.
If
we're
trying
to
deprecate
it
and
I
know
John
on
our
team
has
restructured
some
persistent
volume
test
and
Men
test.
B
So
as
the
recycling
is
all
consolidated
into
a
single
test
suite,
so
it
would
be
easy
to
remove
it
once
recycling
is
deprecated,
but
I
don't
have
any
status
on
if
there's
forward
progress
and
doing
that,
if
there's
been
an
official
announcement
so
that
the
clock
can
start
again
on
the
deprecation
saw.
Do
you
know
anything
about
that
part
of
it
I'm.
A
A
A
J
You
/
vsphere
VI,
should
talk
provider
for
vcr,
be
added
a
bunch
of
new
test
and
we
have
opened
a
PR
for
them.
So
I
would
like
to
fix.
Take
a
look.
Some
up
does
coach
right
now
is
written
only
for
vsphere,
but
the
genetic
tests
that
we
have
identified
should
be
applicable
for
the
horse,
emeka
wider,
and
maybe
a
lot
of
the
code
can
be
reused
so.
A
A
B
H
J
I
elected
as
the
reviewer
for
storage,
plus
G
and
I
was
struggling
to
find
what
exactly
I
need
to
do
now.
I
searched
for
the
issue
that
Oprah
get
storage
classes
are
a
bunch
of
issues,
opened
again,
storage
classes
and
I,
don't
know
if
the
desire
is
to
GA
storage
class,
as
is
today,
and
only
change
the
API
definitions,
or
is
there
some
blocking
bugs
or
some
implementation
that
way
to
like,
for
example,
dashboard
trudat
doesn't
have
any
visibility
into
storage
trouble.
C
H
A
H
C
So
yeah
I
was
keeping
track
of
that
issue.
Currently,
someone
from
the
add-on
team
is
making
some
changes.
A
I
A
K
Have
one
really
good
quick
question
just
for
fun?
We
anyone
else
going
to
keep
con
Europe.
They
were
designed
a
separate
of
the
meeting.
When
I
came
in
late.
Oh,
we
should
know
that
that's
a
good
question
I
know
that
sod
will
be
there
and
I
think
I'll
be
there.
So
I
don't
think
we
need
like
a
big
formal
face-to-face
because
it
might
be
just
hard
for
when
you
get
there,
but
it
might
not
be
a
bad
idea,
so
hook
up,
say
hi.