►
From YouTube: Kubernetes SIG Storage 20170105
Description
Kubernetes Storage Special-Interest-Group (SIG) Meeting - 05 January 2017
Meeting Notes/Agenda: https://docs.google.com/document/d/1-8KEG8AjAgKznS9NFm3qWqkGyCHmvU6HVl0sk5hwoAE/edit#heading=h.6e6i9ioqbcq7
Find out more about the Storage SIG here: https://github.com/kubernetes/community/tree/master/sig-storage
Moderator: Saad Ali (Google)
Chat Log:
09:16:06 From hekumar : 1
09:16:18 From gopalsharma : 1
09:16:21 From aditya : 1
09:26:11 From erinboyd : storage locality
09:26:24 From erinboyd : call it that? ^^
09:26:38 From erinboyd : scheduable storage locality
09:26:47 From erinboyd : just a thought
09:36:43 From Tom G. : Storage Inventory?
09:42:48 From Tom G. : Add Clones to Snapshots
09:45:24 From steve watt : can we go on mute if we’re not talking, the typing is drowning out the speaker
09:52:40 From Luis Pabón : I agree, i think this is outside the scope of kube
09:54:03 From Luis Pabón : i think the example would be specific to one storage implementation
09:54:43 From Tom G. : How does the scheduler handle the hand off of the keys?
09:55:19 From Luis Pabón : the problem is that there are too many iterations of an encryption setup
09:58:23 From hekumar : we have made lot of fixes in AWS
09:58:54 From hekumar : there are some papercuts problems though
09:59:26 From steve watt : justinsb is like the nicest guy in kubernetes
10:06:21 From Luis Pabón : i'll start reviewing also
10:08:49 From Saad Ali : “justinsb is like the nicest guy in kubernetes“ +1
A
Happy
new
year,
everyone-
this
is
the
bi-weekly
meeting
of
the
storage
special
interest
group.
Today
is
january
five
2017
and
today
on
the
agenda,
we've
got
a
few
items
for
primarily
going
to
focus
on
2017
planning.
We've
got
a
I'm
going
to
start,
adding
a
couple
of
line
items
regularly
to
the
agenda,
their
design,
reviews
and
PRS
to
discuss
that
need
attention,
we're
going
to
leave
those
as
reoccurring
line
items
and
anybody
that
has
any
design
reviews
that
they
want
to
go
over
our
PRS
that
need
attention.
A
Please
add
them
there
and
the
idea
here
is
I.
Think
a
lot
of
these
meetings
have
ended
up
being
a
lot
about
what
the
current
status
updates
are
on.
What
folks
are
working
on
and
potentially
things
that
we
want
to
work
on
in
the
future,
but
we
haven't
really
used
these
meetings
as
an
opportunity
to
do
design,
reviews
or
discuss
some
of
the
PRS
that
are
in
review
a
lot
of
that
happens,
offline
or
at
the
face-to-face
meetings
and
I.
Think
you
should
try
to
use
it
use
these
meetings
to
do
more
of
that.
A
B
Okay,
sorry,
my
connections,
a
little
screwy
right
now,
not
sure.
What's
going
on
yeah,
so
we've
got
the
spreadsheet
I,
don't
know
how
many
of
you
had
a
chance
to
look
but
I
see
a
few
people
are
jumping
in
right
now.
I
would
like
to
go
through
and
go
through
these
one
by
one
and
feel
free
to
add
some
more
while
we're
talking
about
it
and
kind
of
prioritize.
B
What
we
think
needs
to
be
done
as
a
group
and
then
kind
of
put
down
how
long
we
think
it's
going
to
take
to
do
and
if
anybody
has
wants
it
to
take
ownership
of
it.
That
would
be
great.
We
can
also
put
their
names
down
as
well
and
if
we
see
something
that's
could
be
broken
down
into
multiple
deliverables
over
the
course
of
you
know,
16
17,
I'm
frame.
That
would
be
great.
We
can
split
those
up
as
well,
so
I
haven't
feeling
something
like
you
know.
Out
of
tree
volume.
B
A
C
A
D
B
E
C
Cool
I
on
the
schedule,
you
know
if
we
have
things
that
are
more
than
I
one
release
at
you
know,
I
think
it'd
behoove
us
to
look
at
stepping
it
across
multiple
releases
right.
So
what
we's?
If
we
have
a
feature,
that's
some.
You
know
a
snapshot
into
something
we
probably
could
do
over
three
releases
if
we
wanted
to,
but
I
think
it'd
be
better
to
have
like
you
know,
something
that
we
could
have.
You
know
made
each
one
of
those
three
releases
to
get
us
up
to
the
final
product.
So.
A
G
I'm,
not
a
comment
suit.
That's
all
some
big
features
like
snapshot.
I,
guess
you
really
eat
erase
what
I've
said,
but
it
just
one
Lee
clarified.
Also
snapshot
has
a
multiple
use
cases.
Maybe
some
people
want
to
have
a
consistent
snapshots?
Are
people
want
to
have
a
consistent
groups?
So
if
we
can
just
make
a
targets
which
one
we
can
deliver
I
speech
releases,
so
you
can
make
a
judgment
rather
than
see
the
face
ball
and
what
is
our
something
is
sufficient
for
products
release,
let's
be
clear,
as
we
have
the
grades
so.
G
B
How
would
I
would
kind
of
I
would
suggest,
maybe
just
if
it
snapshotting
is
something
that
covers
several,
has
several
different
features
associated.
I
might
just
call
those
out
as
separate
features,
and
that
way
we
can
prioritize
them
separately
from
unless
it's
something
that
dependencies
between
them.
That's
something
we
need
to
deal
with,
but
I
feel
like
if
there's
different
types
of
snapshots
we
want
to
support.
Maybe
those
are
just
different
features
and
just
have
different
rows
for
those
all
right.
This
is
slime.
E
To
look
super,
like
a
you
know,
a
release,
planning
thing
where
you
know
what
what
these
are
sort
of
more
than
agile
process
are.
These
are
epics
that
are
broken
out
into
multiple
user
stories.
That's
like
the
first
thing,
like
you
know,
I'm,
really
asking
questions
like
what
does
this
mean,
and
it's
showing
that
fact
that
you
know
we
probably
need
a
link
to
a
proposal
or
something
like
that
in.
Is
it
right.
H
I
think
we
need
to
take
this
in
phases
where
we
first
proposed.
What
the
feature
set
is
some
of
them
to
be
implemented
before
others
then
defining
the
interface,
because
on
many
of
these
like
snapshots,
I
would
suggest
there's
going
to
be
multiple
implementations.
I
mean,
ultimately
it
isn't
the
main
screen.
Kuber
Nettie's
code.
That's
going
to
be
doing
this!
It's
going
to
be
delegated
down
to
the
backend
storage
providers
that
in
many
or
most
cases,
are
the
ones
that
actually
own
then
initiate
the
snapshots.
H
H
E
H
A
And
then
those
owners
or
teams
can
go
out
and
do
exactly
what
you
guys
are
proposing,
which
is
come
up
with
the
proposals
for
that
particular
feature.
Come
up
with
the
various
possible
implementations,
bring
it
back
to
the
sig
report,
back
the
proposals,
select
a
final
design
and
then
move
forward
with
implementation.
So
that
means
today,
instead
of
focusing
on
the
specifics
of
any
individual
area.
B
B
Okay,
so
hopefully
everyone
can
see
this.
Let
me
know
if
you
want
me
to
zoom
in
I,
think
it's
not
seeming
a
little
bit
okay,
so
the
first
one
I've
got
here
is
the
out
of
tree
volume
plugins
as
a
feature.
We
discussed
this
a
lot
during
the
November
face-to-face
meeting,
so
I
have
a
feeling
it's
pretty
high
priority,
but
just
to
get
a
sense
do
we
think
this
is
a
priority
one
between
one
and
ten?
What
do
we
think
so?
I
know
this
is
hard
to
solicit
from
about
30
people.
B
C
I
E
B
J
A
A
D
A
H
D
A
E
K
K
B
A
Every
feature
that
we
propose
needs
to
have
an
alpha
beta,
jee,
a
proposal,
so
one
thing
that
we,
the
larger
Cooper
Nettie's
organization,
now,
has
a
features
repository
where
we
track
all
the
features
that
are
shipping
for
a
particular
release.
There's
a
template
for
each
feature,
and
one
of
the
fields
that
we
define
is
whether
it's
alpha
beta
or
GA,
and
that
ends
up
dictating
what
it
looks
like
in
the
release
notes
and
what
the
documentation
for
it
looks
like,
etc,
etc.
A
B
A
A
K
B
Sounds
good
yeah.
It
makes
sense,
especially
if
CBI
is
going
to
kind
of
be
the
ultimate
replacement
for
two,
so
schedule
will
sorich.
This
is
something
that
we've
been
discussing.
I
think
internally,
I'm,
not
sure
how
much
the
sig
has
talked
about
this,
but
the
idea
is
that
we
would
only
scheduled
pods
at
high
level
the
ideas
that
we'd
only
scheduled
pods
unknowns.
It's
supported,
given
volume
type
so
has
to
have
the
right
drivers
after
it
has
or
a
binaries.
B
K
Also
I
think
a
big
deal
for
one
of
the
things
that's
coming
up
quite
a
bit
across
Cooper
netted
fragmentation,
concerns
and
I
think
this
is
something
that
we
can
do
to
help
make
sure
that
what's
in
the
source
code
base
is
going
to
work,
and
we
have
more
assurances
will
work
across
from
the
less
images.
And
you
know
people
can
have
a
better
handling
of
what
happens
in
those
sorts
of
different
configurations.
I
kind
of
feel
that
this
is
a
high
to
maybe
a
low
one.
L
G
L
K
We
should
talk
maybe
about
that
offline.
There's
a
I'm.
Sorry,
there
was
a
there's
gonna,
be
more
motion
and
local
storage
that
I
want
to
talk
about.
When
we
got
to
it
or
I.
Don't
know
people
wanted
to
talk
about
it
now,
so
you
tell
me
actually
I'll
talk
about
it
now
I've
been
working
with
the
node
team.
K
Also
and
I-
don't
know
if
you
guys
are
familiar
with
who
wishes,
but
this
is
going
to
be
working
and
focusing
quite
a
bit
on
disk
management
and
disk
resources,
and
he
is
also
going
to
be
working
with
the
Google
storage
team
on
local
storage
and
I.
Think
we
feel
that
this
is
an
area
that
it's
kind
of
past.
Due
at
this
point-
and
you
know,
I
feel
like
it's
a
definitely
item.
10
is
a
priority.
K
One
item
I,
don't
know
if
we're
going
to
get
it
in
one,
not
six
I
think
we're
going
to
have
to
make
a
lot
of
progress
on
it.
This
quarter
and
we're
going
to
have
at
least
two
people
our
side.
You
know
thinking
about
this
quite
a
bit.
This
quarter,
I
kind
of
think
it's
deciding.
It's
slightly
related
to
you
item
four,
but
I,
don't
know
exactly
I'm
hoping
we
can
keep
them
decoupled
and
not
dependent
on
each
other.
L
Go
ahead,
cut
out,
I'm,
basically
saying
local
storage
support
either
you
do
it
by
hand
or
otherwise.
Basically,
the
very
first
thing,
which
you
said
spirit:
google
storage,
only
scheduled
partner
nodes
right.
So
there
has
to
be
basically
a
notion
of
where
this
volume,
where
this
local,
then
what
is
it
called,
so
you
need
some
sort
of
it
from
the
scheduler.
I
think.
K
F
K
With
storage
requirements
yeah,
so
let's
let
let's
take
a
step
back
for
number
ten.
What's
going
on
to
the
node
team?
Is
there
going
to
be
looking
at
making
the
disk
a
first
class
schedule,
ogle
resource
like
CPU
and
memory
and
I?
Think
that
may
be
something
that
you
know.
So
you
can
schedule
your
jobs
according
to
disk
capacity
and
eventually
disk
I
ops,
they're
going
to
be
working
with.
You
know
the
storage
cig
to
do
this.
I
was
talking
them
about
this.
K
We're
going
to
be
spitting
up
a
team
of
people
to
be
working
on
this
and
you
know
q1
that
it
to
me
is
the.
I
think
what
you're
saying
about
scheduling
is
based
on
the
you
know,
storage
of
the
first
class
sets
and
resource
and
we're
doing
that,
partly
in
the
node
team,
because
the
node
team
is
pretty
focused
on
being
able
to
export
the
nodes
resource
management
pieces
and
we
want
to
integrate
with
the
system.
That's
already
present
and
leverage
that
well,
of
course,
still
working
closely
with
the
cigs
into
the
storage
team.
K
The
item
number,
for
maybe
we
should
come
up
with
a
different
name,
and
the
idea
here
is
that
when
you
are
asking
to
provision
a
cod
on
a
node
that
node
is
going-
and
you
know,
new
provision
of
volume
or
storage
associated
with
that
pod.
That
pod
will
be
able
to
actually
a
node.
The
pot
is
scheduled
on
will
be
able
to
support
those
red
requirements.
F
L
K
H
D
Maybe
a
third
dimension.
If
I'm,
I
know
that
was
neither
what
I
was
speaking
of
just
to
get
I
guess
more
I,
just
one
one
extra
data
point
that
we
probably
do
one
the
item
number
for
schedule,
storage,
simply
to
prevent
discarding
crashing
right.
You
need
to
prevent.
We
need
guarantee
isolation
between
processes
and
storage
right
now,
one
of
the
biggest
I
solution
escapes
that
we
experience
number
ten
to
me
about
providing
a
persistent
volume,
abstraction
that
is
backed
by
local
media,
which
is
required.
D
K
Along
with
Stallone
I,
think,
there's
two
older
coming
out
of
this
or
stuff
like
I
agree
with
10
mins
dimensions.
Make
sense
to
me.
One
I
think
if
we
put
the
name
schedule
inside
the
name
of
you
know
all
these
items.
It's
not
that
useful,
because
a
lot
of
these
things
are
going
to
its
scheduling
system
in
Coober
Nettie's.
A
lot
of
what
we're
going
to
do
is
going
to
interact
with
the
scheduler
I.
K
I,
don't
know
like
I,
think
centering
things
on
the
scheduler
to
me
seems
it's
it's
like
saying.
You
know
you're
working
on
the
colonel
everything
that
deals
with
memory.
We
should
you
know,
break
it
down
and
how
we're
using
the
memory
I
mean
it's.
It's
such
a
core
piece.
You
know
everything
kind
of
does
I
guess
to
me
for
seems
like
a
very
distinct
body
of
work
and
10
also
seems
like
a
distinct
body
of
work.
K
K
K
D
K
Its
most
basic
I,
I
thought
four
I
thought
the
whole
point
of
four
was
that
as
I
understand
it,
it
would
have
more
to
do
with
klug
in
support.
So
if
you
are
NFS
or
if
you
were
cephalopod
or
if
you
were
rook
or
if
you
were
any
of
these,
like
woah
bikes
or
all
of
these
different
things
that
are
going
on
that
when
you
try
to
schedule
your
pot
on
the
node,
you
knew
that
you
know
we
try
to
schedule
your
job.
K
K
Knowing
that
your
cluster
has
the
capabilities
to
meet
the
storage
plug-in
technology
that
you
wanted
to
have,
and
if
it
doesn't,
it
will
tell
you
clearly
why
it
doesn't,
whether
it's
a
kernel
lacking
or
a
distribution
gap,
or
you
know,
something's
broken
where's
today
you
just
get
is
a
big
jam
out
and
all
right,
I
think
we
have
too
much
dependence
in
the
u.s.
image
for.
K
K
This
one,
no
absolutely
so
local
storage
could
be
one
of
those.
You
know
technologies
that
you
want
to
hit,
but
there's
another
piece
with
local
storage
support
that
needs
to
happen
that
I
think
is
distinct,
where
we
almost
have
to
do
like
plug
in
support
for
local
storage,
and
this
comes
in
several
dimensions-
one.
We
have
to
start
teaching
Cooper
Nettie's
to
treat
storage,
the
first
class
citizen,
so
that
you
know
and
of
itself
we
can
schedule
able-
and
maybe
this
is
a
separate
item.
K
Stories
I
agree,
and
that's
that
that's
just
part
of
it.
So
what
did
it
should
be?
You
know,
let's
call
it
storage
as
a
first
class
schedule
resource
and
then
the
first
class
resource
and
which
will
be
shared
by
the
node
team
and
then
the
other
one
is
just
local
storage
in
general
and
I.
Think
there's
there's
more
work
in
local
storage,
too
I!
K
B
A
Poor
I
think
actually
going
back
to
Aaron's
idea
of
breaking
item.
Number
four
into
sub
items
may
make
sense.
Based
on
what
Tim
and
Michael
said,
my
understanding
was
closer
to
what
Michael
was
mentioning,
which
is
capabilities,
but
what
Tim
mentioned
was
also
important,
which
is
what
are
the
resources
that
are
actually
available.
So
I
think
capabilities
and
resources
are
two
distinct.
Things
does
that
this
is
inviting
team.
K
Of
another
ask
that
he's
asking
for
which
was
not
within
the
ask
of
horrible
I'm
thinking,
right,
she'd
be
listed
and
also
I.
Don't
know
it
good.
You
know,
I,
don't
know
when
it's
gonna
land
yeah,
let's,
let's
add
it
to
the
list.
Okay
and
we'll
call
that
like
resource
availability,
maybe
yeah,
and
just
also
to
let
you
know,
though
disqus
resource
management
in
the
nose
which
in
us
is
or
Disqus
resource
management
period,
in
something
a
lot
of
attention
coming
this
year.
Also,
okay,.
A
B
K
What
do
you
think
I
would
say
already
too
just
for
now,
I
mean
I,
I,
guess
I.
Look
at
Tim
had
a
really
good
description
of
priorities
recently,
which
was
sort
of
like
when
you
have
a
tower
defense
game.
You
know
which
enemies
are
closest
to
you
and
the
ones
that
are
closest
the
ones
you
put
a
lot
of
attention
on
Harper.
This
is
sig,
you
know,
I,
think,
storage,
capable
scheduling
is
something
where
if
we
get
a
doc
out
in
next
quarter
will
be
happy.
Okay,.
K
A
Haven't
heard
that
much
asks
for
snapshots,
I
think
I've
heard
far
more
for
for
some
of
the
other
features
things
like
local
storage,
I.
Think
people
are
really
really
craving
over
snapshots.
I
think,
is
a
nice
to
have,
but
nobody
going,
I'm
not
going
to
use
uber
Nettie's
because
doesn't
have
stop
shouting.
So
here's.
K
My
thinking
and
by
the
way
I
agree
with
you,
said,
damn
it
you're
right.
I
think
snapshot
should
be
it
to
the
reason
I'm
pushing
it
is
as
the
work
that's
going
on
in
Federation
and
especially
as
right
now,
it's
trailing
on
its
storage
functionality,
but
I'm
expecting
by
the
second
half
a
20-17,
we're
going
to
have
to
start
doing
things
like
replicating
data
may
be
a
cross
cluster
in
my
then
I
believe
having
snap
chat,
functionality
Gordon
be
very
important.
So
let's
say
two
that's
already
there.
That's
eight.
A
A
You
know
I've
been
thinking
about
that
a
replication
and
I
was
wondering
how
much
of
it
would
be.
The
responsibility
of
Coober
Nettie's
on
where
I
seek
uber
Nettie's
is
exposing
a
common
API
that
a
bunch
of
different
storage
providers
provide
and
then
providing
a
consistent
interface
against.
It
agree
that
replication
I'm
not
sure
how
cou
burnetii
barely
fits
into
that,
because,
generally
the
storage
system
tends
to
be
aware
of
where
it
is
and
figure
out
where
it
needs
to
replicate
the
data.
What
are
your
thoughts
on
that?
A
N
A
K
K
I,
admit,
I'm,
a
little
snapshot,
biased
based
on
my
experience
in
network
appliance
I,
would
love
to
schedule
some
time
at
the
end
of
January,
like
a
one
hour,
just
out
of
a
band
meeting
to
discuss
data
replication
and
what
this
is
gonna,
you
know
be
alike
and
how
we
may
want
to
attack
it.
I,
don't
know
how
urgent
that
is
for
other
people,
but
I
also
think
that
Leslie
stay
on
five
I
believe
that
Jane
really
wants
to
keep
working
on
snapshots
this
quarter.
E
Three
I
think
the
data
locality,
especially
in
the
big
data
use
cases
on
premise,
is
really
important
right
and
so
local
disk,
right
and
I
think
there's
some
interesting
teachers
that
this
is
a
community's
as
a
cloud
native
platform
and
public
baths
have
very
little
available,
whereas
on
premise
they
tend
to
have
a
lot
and
so
we'd
be
quite
public
cloud
focus,
but
not
quite
so
private
cloud
around
or
sorry
on-premise.
You
know
clouds
where
they
do
have
access
to
lots
of
local
disk
36
terabytes
on
average
could
note
you
know.
H
E
K
K
K
B
C
B
A
Think
jion's
been
doing
a
lot
of
work
here
on.
He
recently
sent
out
of
PR
that
Tim
commented
on
buying
a
complete
consensus
on
this
I
think
he
was
making
any
trying
to
rectify
the
difference
between
selectors
and
storage
classes.
That
was
a
big,
open
design
question
from
the
last
time
we
had
a
face
discussion
on
this,
so
that
design
still
need
to
be
finalized
and
once
the
design
is
finalized,
then
we
need
to
implement
it.
It
seems
plausible
that
we
can
do
this
in
1.6
time
frame.
P
A
K
No,
no
I
think
that's
good,
because
you
don't
work
your
tail
off
on
it
due
to
means
that
you
know
we
are
not
kind
of
you
know.
P
ones
are
a
very,
very
strong
commitment.
We're
gonna
try
to
do
at
this
border.
P.
2
means
that
were
active,
you
know
doesn't
mean
we're
not
actively
trying
to
kick
tail
on
it.
You
know
to
me
I
kind
of
like
having
a
few
key
ones
and
a
lot
of
peaches
and
we
have
freedoms
of
student
loans.
Yeah
also
I,
don't
see
test
here.
K
B
C
C
L
C
K
C
It
is
beautiful,
nasaka
sigh,
we
split
up
some
of
the
shared
controller
libraries
that
were
still
debating
over
and
we're
not
sure
that
should
try
to
go
and
they
cadiz
trunk
I.
It
really
shouldn't
go
indeed
in
our
incubator,
NFS
provisioner
project
that
doesn't
make
a
whole
lot
of
sense.
So
we're
really
trying
to
figure
out
where
to
put
the
shared
libraries
now
and
then
after
that,
so.
P
M
A
C
F
C
Right
so
not
every
external
provisioner
will
be
a
flex
plugin,
but
every
extra
every
provisioner
for
flex
will
be
an
external
revision.
So
if
you
want
to
write
an
extra
provisioner
for
a
flex
plugin,
that's
the
only
way
to
do
it,
but
you
could
write
an
external
provisioner
for
NFS
AWS
CBS,
something
else
intruder
I.
A
A
C
Have
I
the
Flex
provisioner
exclusively
call
out
and
with
those
API
calls
in
there,
we've
talked
about
doing
it
that
way
as
well.
It's
just
two
different
paths
and
they're.
Bringing
the
external
provisioner
path
is
probably
the
better
path
to
follow
and
the
design
that
we
want
to
push
more
versus
having
a
call.
A
J
C
Yes,
so
exactly
we
it
would
it
be
extra
work
to
have
an
explicit
call
out
to
create
delete
from
the
Flex
plug
in
this
entry.
When
you
could
accomplish
the
same
work
by
the
using
extra
provisioner
I
sure
we
could
say
that
we
do
both
good
I
think
just
for
the
sake
of
time
and
resources.
We
were
going
to
be
one
or
the
other
for
the
next
iteration
of
flex.
Okay,.
C
I
would
say
we
leave
the
second
1i,
the
explicit
call
out
from
the
entry
flex
provisioner
for
the
I,
the
next
iteration
of
fries
or
whatever
their
auditory
volume
plugins
discussion,
because
I
don't
know
that
we
necessarily
eat
it.
There's
going
to
be
a
way
to
achieve
that
functionality
without
putting
those
calls
in
mm-hmm,
so
I
don't
know
that
we're
going
to
actually
want
it,
but
maybe
there's
some
debate
there
that
we
should
have
both
just.
A
A
Alright,
my
understanding
so
far
has
been
that
encryption
is
owned
by
the
storage
system
itself
and
usually,
when
provision
a
piece
of
storage,
you
specify
whether
you
want
encryption
or
not
in
our
storage
class
interface
is
flexible
enough
to
allow
whatever
set
of
parameters
you
want
to
pass
if
you
wanna,
if
that
happens,
to
be
in
a
flat,
tentative
christian
storage
classes
kind
already
enable
that
so
my
question
here,
for
whoever
put
that
item
is:
is
there
anything
for
Cooper
Nettie's
to
do
at
this
point?
So
is
this
in
inclusion
addressed
good
question?
K
K
You
know,
as
the
experts
in
the
sig
storage
team
we
can
say,
encryption
is
that's
already
taken
care
of,
but
how
do
consumers
you
know
either
enterprise
or
non
learn
how
to
do
it,
and
is
it
easy
for
them
to
do
it?
Is
it
worth
having
someone
go
off
and
just
figure
that
out
and
write
that
up,
and
you
know
we
don't
go
possibly
even
put
a
you
know
the
Cooper
Nettie's
blog
post
to
describe
it.
A
K
Think
like
having
it
a
walkthrough
one
way
to
do
it,
given
her
that
sort
of
responsibility
would
be
invaluable,
so
maybe
having
an
example
is
everyone
who
could
sign
up
to
doing
that?
Example:
I,
don't
think.
Actually
at
Google
we
have
anyone,
our
storage
team,
who
is
qualified
to
do
this,
otherwise
I
volunteer
for
it
I
think,
would
be
really
cool.
E
K
You
know,
is
it
more
of
a
local
storage
thing?
I
don't
know
like
the
fact
that
you
just
said
that
ended
of
itself
is
something
that
I
would
argue
that
civilians,
you
know,
I'm
gonna,
save
civilians
as
people
who
are
not
you
know,
as
aware
of
Cooper
Nettie's
were
storage,
would
be
aware
of.
You
know
to
you,
it's
obvious
and
simple
to
other
people.
They
may
not
even
know
I'm
going
to
Google
gonna
origin
encryption
right
now
and
then
ask
how
that
works.
K
N
B
K
B
K
Design
design
an
owner,
I,
know
sod,
and
this
is
a
use
to
you.
I
know.
Fish
will
be
involved
in
this
and
I
think
fish
and
sod
for
now
and
we'll
put
great
other
people
in
the
community
are
interested.
You
know
please,
we
I
really
like
this
to
not
be
a
google
only
one
so
that
it
gets
the
best
solution.
Possible.
I
was.
A
K
B
C
K
A
K
A
K
A
K
G
K
Require
an
architectural
change
to
how
we
do
things
in
Coober
natives
to
support
them,
but
then
Justin
kind
of
came
back
and
said
he
might
be
wrong
on
that.
You
know
we
were
going
back
and
forth,
so
this
might
not
be
a
big
effort.
It
might
just
be
fixing
bugs
tamal
rugs
I'm
guys
we
leave.
This
is
kind
of
a
question
mark
and
get
back
to
Justin
sfp,
okay,.
C
K
K
So
I've
been
I've
been
kind
of
having
offline
conversations
with
Justin
about
this
stuff
anyway,
and
we
were
going
through
them
and
I
basically
said
look.
You
know,
I
made
a
suggestion
of
making
this
an
effort
for
this
quarter
and
he
said.
Let
me
get
back
to
you
before
we.
You
know
open
up
that
can
and
we
didn't
have
a
list
of
what's
wrong
with
her
woods.
Constable.
K
K
M
B
B
K
Okay,
as
you
know,
I've
been
kind
of
died,
just
passed
off
the
mantle
of
TL
ship,
which
I
had
for
three
whole
weeks
of
Federation,
one
of
the
tails
left
and
a
new
one
just
showed
up,
and
I
think
that
there's
a
lot
of
stability
work,
they
need
to
do
and
they
have
some
interesting
ways
we're
going
to
be
approaching
state.
So
I
don't
think
we
have
much
to
do
this
quarter.
Besides,
maybe
think
about
our
ideas
on
items.
Five
and
six.
E
K
About
I
provide
a
an
email
to
the
storage
cig,
with
some
of
the
things
and
conversations
have
been
going
on
in
the
Federation
and
like
that,
lead
to
sleep,
I,
just
an
email
if
I,
sound,
okay,
yeah
sounds
great,
wonderful
or
actually
I'll
get,
maybe
Christian
Bell
who's,
a
one
of
at
least
the
Google
TL.
For
that
now
you
know:
I
asked
him
to
do
it.
Okay,.
B
All
right
so
how
people
started
dropping
off
an
hour
just
over
time,
but
let's
try
to
blow
through
the
last
couple
right
here,
so
the
resource
available
scheduling.
This
is
closely
related
to
what
we
talked
about
earlier.
I
would
assume
this
is
also
a
p2
I'm
going
to
stick
my
neck
out
and
say
that
it's
probably
a
design
type
settings.
The
docs
and
sod.
Do
you
mind
being
the
owner
for
that
for
now
or
does
someone
else
want
to
take
not
responsibility.
K
B
K
A
A
In
a
better
state
than
we
were
before,
what
we
did
for
the
last
quarter
was
go
through
a
bunch
of
different
bugs
and
highlight
areas
where
we
had
regressions
in
the
past
and
try
to
write
an
10
test
to
patch
those
or
units
has
to
touch
those.
So
a
bunch
of
tests
were
written
doc
said
we
still
have
a
huge
gap
so
that
a
lot
of
that
testing
was
for
the
cork
uber
Nettie's
functionality.
A
Things
like
attached
detached
mount
unmount.
What
happens
when
qubit
restarts
things
like
that?
We
still
have
work
to
do
on
those
areas
themselves,
because
so
far
we
just
added
regression
tests.
We
didn't
go
through
them
methodically
and
say
we
have
full
coverage.
Then
we,
the
the
storage
volume
plugins
themselves,
are
completely
exposed
at
the
moment,
meaning,
for
example,
if
we,
if
there's
a
big
change
in
Seth,
that
or
something
break
stuff,
we
don't
notice
it.
C
Q
O
K
O
I
think
we
got
some
good
attention
to
get
some
pieces
completed
before
the
holiday.
I
want
to
continue
that
momentum
as
we
move
forward,
though,
to
make
it
a
priority.
It's
also
becoming
a
priority
for
this.
You
know
see
I
type
environment
within
open
shift,
so
I
think
that's
going
to
facilitate,
feel
measurement
more
of
a
habit
on
our
side
as
well,
but
I
think
we
need
to
not.
O
You
know,
focus
on
so
many
features
and
less
test,
and
you
have
maybe
is
part
of
our
reviews
that
we
do
have
testing
and
coverage
so
that
we're
not
doing
this
in
hindsight,
because
that's
really
what
sudden
I've
done
is
it's
been,
though,
we're
trying
it's
still
a
rate
act.
You
know
a
reactive
testing
design,
instead
of
in
line
with
the
actual
right.
K
So
the
side
news
is
hired.
Now
we
have
four
people
and
the
Google
storage
team
for
Cooper
Nettie
was
dedicated
hundred
percent.
It's
kind
of
weird
in
that
we've
only
really
kind
of
had
one
for
like
six
to
nine
months,
because
either
a
we
had
a
new
person
or
be
saddles
release
our
and
now
we've
got
four
so
I'm,
hoping
that
this
will
both
help
with
review
bandwidth
and
also
help
with
the
testing
workload.
B
K
K
That
would
be
great
to
also.
I
want
to
make
sure
that
this
used,
I
I
think
of
people
I,
don't
want
it
just
to
be
Googlers,
as
always
the
reviewers
I
trust
yon,
very,
very
much
I.
Think
woman
also
has
like
commit
access
and
should
be
reviewing
stuff
in
because
teams
into
so
you
know
I
don't
know
like
Steve,
wat
or
Brad.
We
should
also
be
talking
and
making
sure
that,
as
we
think
people
are,
you
know,
I
don't
know
aching
and
ready,
which
we
should
talk
more
about
that.
That's
all.
K
A
K
Yeah
I
think
yeah
I
think
why
don't
we
give
that
in
the
featured
repo
right
you
can
put
it
yet,
oh
and
then
assign
the
milestones
and
so
yeah
can
somebody
like
I
know
Matt.
Do
you
have
the
bandwidth
to
go
to
the
repo
and
for
everything
that
doesn't
have
a
miles
done
associated
with
it,
send
that
to
the
sixth
Red
List,
and
then
we
find
owners
for
each
one
yeah.
A
K
K
You
know,
I
was
actually
really
happy
when
I
think
Jeff
fans
on
me
in
the
face
to
face
incident
a
little
bin
I'm.
Writing
all
this
code
and
no
one's
reviewing
it.
What
the
hell
that
was
really
good
feedback
and
he
was
actually
a
lot
more
polite
than
I
was
I'm.
Just
not,
but
you
know
that's.
You
know
how
we
address
these
issues
and
Brad
also
cornered
me
and
was
like
telling
me
they
poked
tunnel
holes
into
the
my
technical
ideas
and
I.
A
A
Q
Just
yeah
I
just
need
someone
to
take
a
look
at
that
pure,
so
it
can
go
through
the
start
going
to
the
testing
process.
This
is
the
lip
storage
PR
right.
No,
this
is
a
new
one.
This
is
specifically
for
I'm
one
of
our
dell
emc
storage
product
called
scaleo,
and
this
particular
volume
plug-in
as
a
standalone
plugin
for
scale
arrival
got.
C
Q
From
last
time,
when
I
think
when
we
had
a
before
Christmas
on
floor
on
one
of
those
meetings,
that
was
a
question
that
I
brought
up
specifically
because
I
know
this
was
coming
and
your
consensus
was
yes
at
this
point.
We're
still
in
the
position
where
the
community
is
still
accepting
plugins,
so
I
hope.
C
A
Still
going
to
be
calling
flex
alpha
and
if
you
want
a
volume
plugin
that
is
going
to
be
quote,
unquote:
production
ready,
you
don't
want
to
rely
on
the
alta
api
and
right
now,
the
only
GA
api
that
we
have
is
the
internal
api.
So
I
think
it's
ok
to
add
volume
plugins
at
this
point.
If
they
need
to
be
converted
in
the
future.
That's
ok!
Our
ownership
for
volume
plugins
rests
on
the
ownership
of
the
person
who
writes
the
volume
plug
in
their
into
a
state
of
disrepair.
It's
on
the
person
who
wrote
it.