►
From YouTube: Kubernetes SIG Storage 20161027
Description
Kubernetes Storage Special-Interest-Group (SIG) Meeting - 27 October 2016
Meeting Notes/Agenda: https://docs.google.com/document/d/1-8KEG8AjAgKznS9NFm3qWqkGyCHmvU6HVl0sk5hwoAE/edit#heading=h.8r1a3x6d838k
Find out more about the Storage SIG here: https://github.com/kubernetes/community/tree/master/sig-storage
Moderator: Saad Ali (Google)
Chat Log:
09:15:37 From steve watt : huamin chen - @rootfs on github
09:15:39 From steve watt : and slack
09:44:39 From SIG Storage : https://docs.google.com/document/d/1drdxPkZEiGA06-jnsbSqywzQ6z0bT8uMpJdnIzkRoPo/edit?ts=57ed5df2#
09:48:14 From hchiramm : #sig-storage
09:48:15 From pat : #SIG-Storage
09:48:38 From SIG Storage : https://kubernetes.slack.com/messages/sig-storage/
09:53:26 From barakmich : +1 to mixed feelings about Slack :P
09:55:37 From hchiramm : Happy to help on glusterfs
09:56:16 From eparis : http://kubernetes.slackarchive.io/sig-storage
A
Alright,
let's
get
started
so
hello,
everyone.
This
is
the
bi-weekly
meeting
of
the
storage.
Sig
today
is
Thursday
October
27.
As
a
reminder,
this
meeting
is
public
and
recorded,
and
today
on
the
agenda,
we've
got
a
lot
of
items.
First
up,
let
me
give
an
update
on
NFS
and
cluster
in
GC
I.
So,
as
a
lot
of
you
know,
Google
compute
engine
and
container
engine
are
moving
to
a
new
base.
Image
base
OS
image,
it's
GCI,
it's
based
on
chromium
instead
of
Debian
7,
which
is
being
deprecated.
The
old
image
was
called
container
vm.
A
A
So
we
were
looking
for
a
way
to
get
those
utilities
packaged
in
because
users
kind
of
were
looking
for
these
things
and
without
them
Cuba
can't
mount
NFS
or
I
scuzzy
or
cluster
FS.
We
were
hoping
to
put
them
in
the
base
image,
but
GCI
is
a
very
minimal
image
and
what
we
decided
was
that
we're
going
to
have
to
find
a
way
to
be
able
to
package
these
binaries
that
are
required
for
mounting
outside
of
the
operating
system,
and
we
decided
to
come
up
with
some
sort
of
containerization
scheme.
A
B
Basically
feels
the
same
way
so
like
atomic
is
like
the
super
low
peer
down
image,
and
they
don't
want
to
include
any
basically
mount
drivers
or
kernel
modules
that
you
know
this
thing
needs
that
the
vet,
different
storage
endpoints
need
to
do
the
mount,
and
so
you
know
we're
pretty
interested
in
solving
this
problem
too.
Yeah.
A
Is
a
design
doc
internally
will
have
to
publish
something
externally
there
are
PRS.
If
you
look
for
stuff
from
dish,
you'll
you'll
find
some
of
the
PRS
around
this
I
think
Matthew,
Wang
and
rootfs
have
been
on
some
of
these
PRS,
so
they're
kind
of
familiar
with
the
area
you
might
be
able
to
ping
them
as
well.
A
And
no
worries,
so
this
was
something
that
we
were
planning
on
doing
anyway.
It's
just
that
the
timeline
got
escalated
on
and
the
to
start
off
with
what
we'll
have
is
probably
not
something
that
we
it'll
be
pointing
in
the
right
direction.
It's
the
way
that
we
want
to
go,
it's
not
the
fully
fleshed-out
solution,
but
it's
pointing
in
the
right
direction,
all
right.
A
A
E
A
And
so
the
recommendation
here
is
to
look
at
existing
volume
plugins
and
follow
the
pattern
for
them.
I
wish
we
had
better
documentation
on
how
to
add
new
storage
volume
plugins,
but
at
the
moment
we
don't
so.
The
best
thing
you
can
do
is
look
at
other
volume,
plugins
and
ping
folks
who
have
already
worked
on
volume,
plugins
I
think
rootfs
is
a
great,
a
great
person
to
reach
out
to
quammen
chen.
He
has
written
a
lot
of
the
existing
plugins,
so
you'll
have
a
lot
of
experience
to
share.
E
E
D
E
B
Hailey-
and
this
is
steve-
and
I
think
you
know
it's
often
in
your
own
benefits,
what
happens
is
that
you
know
historic.
Team
storage
just
hasn't
had
much
regression
test
coverage
so
other
folks
in
other
six
or
even
folks
in
our
own
sick.
Well,
like
you
know,
break
stuff
periodically
and
it
won't
get
caught
because
it's
not
in
the
regression
stuff
right,
and
so
I
select
in
the
last
probably
two.
Maybe
three
releases:
we've
really
tried
to
like
increase
our
test
coverage
both
like
at
the
unit
test
level,
but
also
it
system
and
into
entities.
B
E
Okay,
I'd
looking
to
the
DVD,
does
not
run
inside
of
a
container
the
actual
kernel
module
getting
it
order.
The
container
it's
a
really
possible.
So
I
did
I
just
see
there
is
one
digit
least
wanted
in
depth
SL
that
does
use
like
a
virtual
machine
instead,
so
it
has
to
be
something
of
that
nature.
I
I
can't
just
benefit
container.
Do.
A
I
think
that'll
be
the
biggest
challenge
then,
because
I
there
is
no
existing
infrastructure
to
start
virtual
machines
either.
Your
your
your
storage
service
has
to
be
started
as
a
service,
somehow
which
it
means
it
needs
to
be
containerized
beyond
that.
I
guess:
you'll
have
to
rely
on
unit
tests
and
if
you
could
come
up
with
some
sort
of
mock
or
stub
interface
for
your
for
your
storage
service,
that
would
be
the
best
to
do
an
end-to-end
test.
B
And
just
we
for
Gloucester
and
safe,
we
ran
into
the
same
problem
to
write,
which
you
know
in
order
to
test
against
the
real
cluster
and
safe
endpoint
for
the
volume
plugin
or
the
provisioner
week.
We
have
to
actually
have
that
thing.
Containerized
running
in
the
CIA,
so
like
the
safe
container
is
you
know
we
call
them
all
in
one
container
where,
basically
it's
nothing
you
would
ever
actually
use
in
the
real
world,
but
its
interface
compliant.
So
it'll,
you
know
very
validate
all
the
differ.
Function
calls
and
responder.
You
know.
C
B
It's
it's
the
safe
aio,
all
in
one
container,
but
better,
yet
just
reach
out
to
rootfs
warm
and
chen,
and
he
is
the
author
of
that
container
image
and
a
woman
can
sort
of
work
you
through
like
how
how
he
created
it,
what
he
did,
how
he
tests
against
the
things
like
that
and
I'll
put
his
name
in
chat
for
you.
Oh.
D
Have
a
request
as
good
community
citizens,
it's
always
nice
as
you're
going
through
that
process.
If
you
can
document
what
you're
doing
based
on
what
woman
has
done,
we
can
then
create
two
more.
You
know
robust
set
of
documentation.
Zaz
more
people
come
along
to
make
this
whole
process
easier.
That's
a.
B
Yeah
and
feel
free
to
reach
out
to
all
of
us.
You
know
on
the
seek
oui
oui.
It's
not
that
often
that
we
have
somebody
come
new
person
come
along
and
want
to
write
a
plugin
plugin.
So
we
want
to,
you,
know,
make
sure
you're
successful
and
you
know
I'm
sure,
there's
a
lot.
That's
not
documented
and
so
feel
free
to
ask
questions
right.
A
All
right
so
next
up
on
the
agenda,
I
put
a
put
a
bullet
point
in
the
agenda,
doc,
asking
folks
to
leave
PRS
that
need
attention.
As
a
reminder,
we
have
less
than
two
weeks
we
have
about
a
week
and
a
half
left
until
code
freeze
for
1.5.
That
means
we
want
to
get
code
in
ASAP.
So
if
you
have
PRS
that
need
my
attention
or
need
somebody
else's
attention,
please
put
it
in
that
doc.
D
Because
John
on
I
don't
want
to
speak
for
him
if
he
wants
to
present
status
and
diligently
committing
pierre-jean.
Are
you
on
ok,
so
John
and
Jeff
have
both
been
very
busy
I
think
there
were
like
four
tests
committed
yesterday:
Thank
You
sod
for
reviewing
those
actively
getting
into
the
destructive
testing
I
think
we
provided
a
lot
more
coverage
just
to
the
basic
PvP
bc
tests.
We
haven't
started
any
of
the
storage
class
tests,
which
might
have
a
Swizzle
of
difference,
so
we
need
to
probably
go
into
those.
D
We
have
a
bi-weekly
meeting
that
anyone
is
welcome
to
join.
To
offer
up.
You
know
anything
that
they
see
they
feel
like
could
be
hardened,
but
we're
still
going
on
the
idea
that
you
know
the
first
test.
We
creator
on
bubs
we
find
so
otherwise
it's
your
movie
I
think
we're
moving
forward
at
a
really
good
pace.
Yeah.
H
H
Basically,
what
we're
learning,
there's
kind
of
a
few
things,
I
think
that
are
all
coming
together
at
the
same
time.
One
of
them
is
that
there's
a
desire
to
be
able
to
add
plugins
that
are
out
of
tree,
so
that
people
are
not
basically
bottlenecked
by
reviews
and
development
from
inside
the
repo,
and
this
is
the
conversations
that
we're
having
around
flex
and
I
think
that
these
are
going
to
be
having
probably
one
of
the
main
things
that
we
talk
about
in
just
a
couple
weeks
at
the
storage
face-to-face
sig.
H
H
On
top
of
that,
we're
also
discovering
I,
don't
know
how
many
people
are
aware
of
the
images
that
Cooper
Nettie's
uses
overall,
how
many
people
are
like
in
that
world
and
maintain
those
images
like
container
vm,
and
I
know
red
hat:
what's
the
name
of
the
redhead
image?
Is
it
Brad
or
you
know
time,
yeah
atomic?
You
know
it's
a
lot
of
work
and
keeping
all
of
these
straight
for
all
of
our
storage.
You
know
functionality
it's
something
where
either
a
we
make
it
very
distro
dependent
and
then
we're
asking
our
customers
to
say.
H
You
need
to
run
this
distro
to
get
this
sort
of
storage
functionality
and
this
distro
to
get
that
storage
functionality
and
then
qualification
becomes
difficult
and
then
suddenly
they've
got
a
matrix
if
the
customers
hire
confronted
with.
All
of
this
is
it's
complex
and
it's
probably
not
a
road
to
success,
given
all
of
the
options
that
we
want
to
offer,
customers
and
ourselves
look
Steve's,
giving
me
the
nod,
so
I
think
you're,
you're
agreeing
with
this
sort
of
complexity,
and
then,
on
top
of
that
you
know
to
be
transparent.
H
Google
has
had
a
bit
of
a
snafu
with
its
own
images.
Lately
in
you
know,
we
have
container
vm,
which
is
this
image
that
we've
been
maintaining
for
a
while
that
we
would
rather
people
move
on
to
another
stronger,
safer,
more
up-to-date,
maintainable
image,
frosties
guy
and
again,
there's
the
other
images
that
other
companies
have
to.
So
given
all
of
that,
we
have
started
working
on
some
PRS
little
faster
initially
than
we
had
expected.
H
Now
we're
slowing
it
down
a
little
as
you
realize
it's
complex
to
container
eyes,
clients
so
that
when
you
have
things
like
the
NFS
mount
clients
or
the
bluster,
Mouse
clients
or
Seth
or
our
DD
or
who
knows
what
else
is
going
on
that
we
can
come
up
with
some
system.
That'll
be
sort
of
safer
and
more
I
mean
the
word
in
both
the
English
translation
and
the
technical
use
of
containing
the
complexity.
H
So
you're
gonna
see
a
lot
of
work
coming
out
of
us
probably
end
of
this
week
early
next
week
with
how
to
do
this,
and
it's
different
than
provisioners
I.
Think
it's
funny
that
we're
all
trying
to
do
this
at
the
same
time.
How
do
we
simplify
and
maude
relies
our
ability
to
enable
different
types
of
storage
technologies
in
Coober
Nettie's,
and
it's
it's
a
little
chaotic
as
we're
all
attacking
this
from
different
directions
and
different
needs.
H
I
think
we
should
be
talking
about
this
face-to-face
igg
a
little
bit
of
what
problems
we
want
to
solve
and
what
order
we
want
to
solve
them
in
I.
Think
it's
great
actually
that
all
these
different
groups
are
doing
them
in
different
ways.
I
was
really
happy
to
see
the
NFS,
provisioner
and
I'm.
Looking
at
that
code,
I'm
also
really
happy
to
see
that
you
know
this
and
jingchou
are
looking
at
container
izing
the
clients
I.
Don't
think
we
have
the
solution.
H
So
you
know
to
me:
it
feels
like
the
storage
cig
and
these
efforts
are
kind
of
I.
Don't
want
to
overstate
this,
but
it's
like
where
we
know
we're
beginning
to
run,
as
opposed
to
beginning
to
walk
and
we're
trying
to
find
the
right
way
to
do
that
without
tripping
over
her
own
feet
because
go
makes
this
hard.
In
my
opinion,
go
does
not
have
a
clear
ABI
like
see.
Does
it
does
not
have
deal
open
or
shared
objects?
So
you
know
my
finding
ways
to
make
these
dependencies.
H
You
know
dependable,
clean,
programmable
and
and
safe
is
hard.
You
know
another
piece
that's
here
is,
and
this
might
be
what
sod
was
talking
about
in
our
rant
and
it's
not
a
rant
about
the
team
or
anything
going
on
in
the
cig.
It's
more
of
just
AI
think
this
is
hard.
How
do
we
enable
things
to
work
out
of
tree
while
still
maintaining
quality
and
I?
Think
it's
not
an
easy
problem,
because,
though
these
er
experience
to.
H
Experience,
that's
a
better
way
to
say
it.
Thank
you
very
much.
How
do
we
keep
a
strong,
consistent
user
experience
in
both?
You
know,
predictability
of
what
these
plugins
will
do
and
then
also
making
sure
that
they're
consistent,
so
that
people
who
are
not
part
of
the
cig
or
just
you
know
the
the
people
on
the
streets
of
San,
Francisco
and
Berlin.
H
Think
that
that
is
it's
hard
and
I
think
that
we're
gonna
have
to
organize
a
lot
in
order
to
figure
out
what
that
looks
like
and
I
think
that
at
some
hand
you've
got
the
the
thing
I
talked
about
last
time,
which
is
the
easiest
way
to
do.
This
is
to
keep
everything
in
tree,
but
there's
a
lot
of
drawbacks
to
doing
that,
and
you
know
the
easiest
way
to
keep
things
consistent.
User
experience
is
to
keep
everything
entry.
H
On
the
other
hand,
what
I'm
also
seeing
is
that
you
know
reviewing
bandwidth
and
the
testing
infrastructure
that
we
have
now
makes
that
really
hard
to
maintain
velocity
for
all
the
interest
that
we've
got
in
this
area,
but
I
really
wouldn't
want
to
stop.
You
know
like
the
word
bruda,
fests
and
I
think
Brendan
Byrnes
are
doing,
and
maybe
it's
not
these
people
like
an
RB
Deanna's
or
you
know,
I,
don't
want
to
slow
down
the
inclusion
of
this
work.
H
I
think
was
looking
at
the
device
open
to
PR
just
this
week
and
I
think
that's
a
good
example
of
something
that
works
for
most
of
our
plugins,
but
not
all
and
has
to
be
handled
and
architected
a
little
bit
differently
for
things
like
our
BD,
and
we
know
we
need
to
be
aware
of
that,
so
that
that's
my
spiel
and
I'm
curious
what
other
people
think
and
if
you
want
to
know
how
it
goes
to
next
actions.
We
can
talk
about
that
too,
but
that's
all
just
a
perspective.
Yeah
I
totally.
B
Agree,
I
think
you
know
we.
The
entry
model
is
Swift
as
well
so
far,
but
you
know,
as
I
totally
agree
with
all
the
points
that
you've
made
in
and
additionally,
like
you
know,
whenever
we
want
to
patch
a
bug
or
something
like
that,
you
know
it's
got
to
come
out
in
the
next
Cooper
knees,
release
and
people
have
to
wait
right,
and
so,
whereas
out
of
tree
model
storage,
vendors
that
are
contributing
volume,
plugins
and
provisioners
can
quickly
they're
not
tied
to
cuban
Eddie's.
B
They
can
patch
them
and
ship
them
at
their
own
cadence
and
ultimately
de
view
with
happier
customers
right
or
users
for
our
community.
So
I
totally
think
like.
This
is
the
right
time
for
us
to
like
new
things,
I'll
just
refer
to
figure
out
having
a
new
things
out
a
tree
figure
out
a
way
for
it
to
be
elegant,
I
think
this
inner
eyes
client
mount
driver
thing.
It
is
problem
to
solve
around
or
berate
rather
container
eyes.
Dependencies
are
the
problem
to
solve
around
user
experience
and
then
there's
the
other
problem.
B
I
agree
that
you
point
at
two
witches
there's
a
bunch
of
stuff.
That's
in
flight
for
entry
now,
so
you
know
the
the
plugins
that
you
mentioned:
fladam
ears,
lip
storage
thing,
and
how
do
we
communicate
what
the
roadmap
is
for
things
that
are
entry
now
and
when
things
without
of
true
I,
think
we've
got
to
figure
all
of
that
out.
B
H
So
this
should
be
a
big
agenda
item
that
we
talk
about.
I,
don't
think
we're
gonna
solve
this
honestly
we're
not
going
to
solve
this
I.
Don't
think
the
customers
will
see
the
solution
to
this
in
the
first
half
of
next
year,
but
I
think
that
it's
something
that
is
large,
its
architectural
and
we
need
to
start
planning.
You
know
in
a
week
or
two
and
then
start
figuring
out
what
next
steps
are
and
I've
got
some
ideas
for
next
steps,
but
I
don't
think
this
meeting
is
the
right
time
for
them.
H
We
should
we're
all
going
to
meet
together
in
person
and
I.
Think.
The
most
important
thing
honestly
is
the
work
that
just
Aaron
when
I
say
Aaron,
but
that
that
effort
to
me
is
a
requirement,
no
matter
what
we
do
and
in
addition,
you
know
we
at
least
on
our
side,
we're
hiring
more
people
in
the
storage
area
for
Cooper
Nettie's
and
one
of
their.
You
know
some
of
their
focus
is
going
to
be
on
accelerating
that
effort
to
I.
D
Think
I'm
super
excited
about
this
I
mean
I'm,
selfishly
excited
about
it,
because
my
team
does
usability
and
I.
Think
beyond
the
e
de
it'd
be
great
to
have
like
a
road
map.
How
we
achieve
consistent
usability
without
a
tree
I
do
think
it's
the
right
way
to
go,
but
I
don't
think
we
have
a
good
process
for
achieving
that.
There
are
steps
to
make
that
way.
So
I
hope
that
we
all
can
kind
of
brainstorm.
How
we
get
there.
H
Have
opinions
and
thoughts
and
everything
it's
a
problem,
but
I
think
that
you
know
figuring
out,
which
ones
makes
sense
and
don't
and
working
with
everyone
together
on
a
plan
is
the
important
part
and,
like
I,
said,
I
I
think
we
get
a
lot
done
at
the
face
two
faces
and
so
I'm
very
excited
for
the
one
coming
up
in
November
I
mean
if
you
want
I
can
give
ideas
now,
but
I.
You
know,
I
think.
The
way
things
are
going
is
the
right
first
step,
at
least
for
the
next
couple
months.
H
A
Let's
save
that
for
the
face
to
face
I
think
it's
a
larger
discussion
than
we
have
four
for
today,
but
thank
you
Michael.
That
was
that
was
great
and
I.
Think
a
lot
of
fuss
completely
agree
with
that.
It's
a
it's
been
on
the
minds
of
a
lot
of
people
in
the
sig
for
a
long
time.
So
to
have
that
captured
is
great
and
it's
a
perfect
segue
to
our
next
item.
Brad
child's
proposal
for
an
NFS
provisioner
incubator.
C
C
On
that,
yet
we
have
I
Clayton
on
our
side.
Acting
is
missing
or
in
it
responsibly
and
then
me
acting
as
the
champion,
I
I,
don't
I
need
to
figure
out
what
the
next
step
or
what
the
next
step
in
the
process
is
to
actually
get
that
into
incubator.
Wait
if
you
guys
would
go
out
and
take
a
look
at
that
and
maybe
comment
on
it
or
you
know,
leave
any
feedback.
Is
you
see
appropriate
we're
going
to
be?
C
B
F
Nothing
can
go
into
the
communities
project
directly
at
this
time.
Everything
needs
to
go
into
it,
an
incubator.
It
means
to
show
that
as
an
active
community,
it
has
users,
it
has
done
a
couple
of
releases
and
I
think
after
six
months
or
so
it
can
go
into
the
corker
benetti's
repo
I
was,
in
the
other
room,
not
really
paying
attention.
When
the
question
exactly
came
up,
was
it
what's
involved
in
incubator?
Yes,
the
process
yeah,
so
the
process
is,
is
you
have
to
have
a
cig,
a
champion
and
a
sponsor?
F
Obviously
the
cig
in
this
case
would
be
general
agreement
from
the
people
in
this
cig
that
the
project
it
makes
sense
is
something
that
kerber
Nettie's
cares
about.
A
champion
is
going
to
be.
Anyone
with
you
know
their
name
and
an
owner's
file
somewhere
inside
kuber
Nettie's
has
to
say
yes,
I
think
this
is
a
good
idea
and
I
am
willing
to
help
the
people
with
any
issues
they
have
around.
You
know
their
software
or
working
with
kerber
Nettie's
stuff
like
that
and
a
sponsor
is
going
to
be
it's
a
very.
I
F
List,
there's
like
five
sponsors
or
like
I,
mean
Brian
grants
not
even
a
sponsor
officially
according
to
the
rules
but
I.
Don't
think
any
of
us
would
tell
them
no
and
that's
just
one
of
the
people
who
try
to
pay
attention
to
the
the
project
as
a
whole
and
make
sure
that
these
things
fit
in
with
the
Cooper
Nettie's,
like
mindset,
so
I
mean.
Obviously
we
would
have
no
problems
finding
a
champion.
A
sponsor
will
also
likely
be
extremely
easy.
F
So
that's
the
idea
when
you
sit
in
incubation
I.
Think
it's
like
minimum
time
is
about
six
months
and
you
show
that
it's
active.
Then
people
use
it.
Then
it
can
graduate
if
it
goes
about
six
months
without
being
paid
for
or
not
pay
for
like
play
with
not
being
looked
at,
then
it
goes.
The
other
way
into
I,
don't
know
what
it's
called
deprecated
or
dead
or
done,
or
something
so.
B
F
The
code
to
the
kid
repository,
then
it
goes
straight
in
if
you
want
your
own
repository-
is
an
incubator.
There
is
a
rule
that
anything
that
is
currently
inside
Cooper
Nettie's
that
you
wish
to
break
into
its
own
repository
is
perfectly
fine
without
following
any
process.
So
if
we
wanted
to
take
the
GCT
plugin,
if
we
wanted
to
take
the
GCE,
plugin
and
spin
it
out
into
its
own
repo,
we
would
be
able
to
do
that
without
any
of
this
process.
B
F
B
H
H
Yeah,
it
may
not
just
be
reviewer
bandwidth.
I
know,
there's
some
incubated
projects
where
they're
still
getting
the
same
set
of
reviewers
that
are
doing
the
Cooper
Nettie's
repo.
But
you
have
the
freedom
to
say:
I'm,
not
going
to
wait
on
the
quarterly
mark
or
I'm,
not
waiting
for
a
patch
release
in
order
to
get
things
out
there
for
folks
in
the
open
source
world
and
that's
very
powerful.
H
A
A
H
We
get
billed
for
it
one
or
two
based
on
a
question.
Actually
that
may
lead
to
one
or
two:
what's
the
packaging
like,
if
you
put
things
in
an
incubator,
besides
people
downloading
it,
I
mean
a
its
awareness
and
then
downloading
and
communication
to
the
customer
right
right
and
I
think
that's
that's
not
to
be
minimized
right.
H
When,
when
I
download
linux,
you
know
and
then
I
feel
like
we're
back
in
the
days
cuz,
you
know
I'm
very
old
when
I
downloaded
the
Linux
tree
and
then
I
had
to
go
somewhere
else
to
download
the
drivers.
For
my
machine
and
I
didn't
know,
I
had
to
go
somewhere
else
to
download
that
other
piece.
You
know
learning
about
that.
You
know
communication
channel
that
I
had
to
do
that,
the
first
time
it
was
a
headache
and
on
the
second
time
it
was
easy.
H
You
know
that
the
packaging
and
then
the
communication
about
you,
know
the
coordination
between
the
two
and
then
the
last
one
I've
got
is
version.
Skew
and
version
compatibility
did
largely
unsolved
right.
So
to
me
those
are
the
the
downsides
of
taking
things
out
of
tree
is
a.
How
do
you
know
where
to
find
it
and
then
be?
How
are
you
sure
that
it
works.
C
H
F
Know
like
just
the
open
source
Travis,
you
can
use
I,
don't
know
not
all
incubator
repose
automatically
have
any
kind
of
Jenkins
like
the
main
stuff
we
use
in
Cork
you
so
I,
don't
know.
Maybe
the
testing
for
guys
have
set
it
up
for
some
and
could
be
to
repose,
but
most
of
them
just
use
a
the
free.
You
know
Travis
stuff.
A
A
I
Okay,
great
yeah
I
just
wanted
to
introduce
us
we're
new
to
the
community,
and
so
just
want
to
say:
I
am
basically
we
were
working
on
it
where
we
would
like
to
write
an
integration
with
pernetti's.
We
provide
a
block,
storage
software
products
deployed
is
container
and
we
think
it
would
be.
It
would
be
great
if
occur,
benetti's
we're
looking
at
in
the
entry
plug
in
at
the
moment.
I
We
were
very
much
like
use
of
dynamic
provisioning
as
well
we're
very
early
days
into
the
integrations
in
the
moment,
so
yeah
I'm
sure
we
coming
up
with
lots
of
questions
but
as
well
we'll
be
in
that
in
Seattle
and
complete
weeks
time
and
I'd
love
to
mean
something
else
festivus
as
possible.
That's
that's!
Basically
all
I
wanted
to
mention
if
anybody
has
any
questions
on
the
product,
but
essentially
word
one
block
storage
products,
so
we
focus
on
it
up.
Disco
use
cases,
running
databases
and
containers.
I
B
So
just
just
some
background
for
yourself
and
other
folks
on
the
call
periodically
and
we'll
do
a
face
to
face
in
person
get
together
just
because
it
gives
us
much
higher
bandwidth
for
technical,
technically
complex
discussion
and
we're
doing
one
on
the
monday
before
Paul.
So,
if
you're
coming
out
on
hi,
there
we're
going
to
be
meeting,
you
know
walking
distance
somewhere,
we
should
I
believe
there's
a
document,
Souths
logistics
and
the
proposed
agenda.
But
that
could
be
able
to
make
screaming
everybody
who
was
to
get
some
sense
of
the
road
map
exactly.
A
A
C
I
we
brought
this
up
a
few.
You
know
for
the
last
I
know
three
or
four
storage,
cigs
I,
it's
one.
Annotation
change
were
final
review.
We
want
to
merge
this
like
today
or
tomorrow.
All
right
we've
had
some
review.
Where
you
live
n
reviews,
we
had
some
comics
on
it,
just
kind
of
want
to
give
everybody
notice.
This
isn't
a
big
change
of
no
api
change,
its
one
extra
annotation,
not
ever
names
field,
but
just
heads
up.
A
A
H
J
Yeah
just
minute:
unmuting,
okay,
so
if
you
could
RSVP
by
the
31st
of
October,
so
we
can
get
a
headcount
for
meals.
That'd
be
great,
just
RSVP
by
putting
your
name
in
the
dock
that
is
linked
to
in
this
meeting
agenda
and
if
anybody
with
food
allergies
or
any
other
requests
for
logistics
for
the
meaning.
Please
email
me-
and
I
think
I
put
my
mail
in
that
dock
as
well.
J
J
A
H
One
more
thing:
I'm
still
service
I,
have
no
answer
to
this
at
all
bugs
old
issues.
It's
really
hard
for
me
to
understand
when
things
are
a
lot,
that
the
flake
system
I,
believe
that
coo
benetti's
has
in
place
is
sometimes
a
little
extra,
noisy
and
misleading,
and
so
just
as
somebody
who
you
know,
I
have
the
ability
to
listen
and
sort
of
direct
people
and
who
pernetti's
on
the
team
in
Google
to
say,
hey.
This
is
something
that's
important.
We
need
to
look
at.
It
is
anyone
else
with
me
just
having
confusion.
C
H
Something
Vladimir,
oh
no,
so
one
of
them
is
one
that
five
is
coming
and
it's
really
hard
for
me
to
know
not
bug
wise.
I
think
the
features
repo
does
a
good
job
of
knowing
what
features
people
are
expecting
to
have
in
15,
but
what
bugs
people
view
as
blockers
for
one
dot?
Five-
and
I
I
don't
understand
if
we
have
a
mechanism
to
communicate
that
and
the
other
one
is,
is
there-
and
this
is
just
my
ignorance,
because
I
kind
of
jumped
in
and
don't
know
how
things
work
all
the
time.
H
A
H
Then
I'll
start
logging
into
it
and
if
someone
can
maybe
offline
or
start
asad
can
teach
me
how
to
do
it
sure,
because
I
think
we
need
a
forum
that
says
hey.
This
is
a
big
deal
and
it's
not
it's
not
easily
a
parent
see
on
github
issues,
but
we
need
to
devote
attention
to
it
and
then
either
a
Disqus
if
it
is
or
isn't
or
be
just
jump
on
it.
H
But
right
now,
I
worry
that
you
know
we
can
have
a
group
of
people
on
a
DW
thinking,
it's
a
big
deal
and
then
another
boot
group
of
people
working
in
the
same
space.
You
don't
see
I
like
the
work
that
I
see
jingchou
and
Justin
doing
where
she
works
on
something,
and
he
goes
oh
yeah.
This
really
helps
me.
H
Let's
work
on
this
together,
or
this
is
more
important
than
you
knew
so
please
drive
it
harder
and
Justin's
been
really
really
helpful
for
us
in
directing
some
of
the
people
on
my
team's
efforts
on
what
to
focus
on.
Does
anyone
else?
You
know
later
to
talk
about
experiences
like
this
and
offered
advice,
or
you
know,
plus
ones
or
anything
is
funny.
You
mention
that
like
I
was
I
reached
out
to
Justin
yesterday,
because
so
here's.
B
My
experience
on
my
select
trying
to
run
a
storage
platform
in
Coober
natives
and
one
of
the
cluster
up
thing,
runs
on
dbm.
That's
missing,
like
d
m
fijn
pool
and
a
bunch
of
like
dependencies
for
logical
volumes
that
just
aren't
in
the
OS
and
so
given
to
your
point
around
dependencies
for
storage
not
being
in
the
Corps
image.
B
I
ran
straight
into
that
yesterday,
right
and
I
was
hosed
because,
like
cluster
up
is
incredibly
tightly
tied
to
like
it
and
only
tested
on
a
particular
OS
which
happens
to
me
d
me
and
for
amazon,
so
I
reached
out
intact
mr.
amazon,
just
in
Santa
Barbara
on
the
shoulder
in
the
community
on
slack
and
I
said:
hey
dude.
What
do
you
think
and
you
run
into
this?
B
Have
you
tried
something
on
Mike's,
a
fedora
that
includes
the
package
and
just
that
we
started
talking
and
just
since,
like
oh
yeah,
I'm
working
on
these
gnarly
problems
with
volumes
in
ews,
can't
believe
it
was
like.
We've
figured
out
this
burg
can't
believe
that
it
was
even
working
this
far
and
I'm
like
what
are
you
talking
about
you
know,
and
so
it's
like
there's
like.
What's
it
like
this
dimension
of
like
serious,
that
people
are
running
into
you
that
aren't
making
its
way
through
the
noise
to
us.
B
H
H
It's
kind
of
like
we're,
throwing
a
great
party-
and
you
know
we're
you
know,
there's
a
party
going
on
and
we're
kind
of
in
charge
of
like
the
drinks
and
the
security
and
the
house
that
everyone's
having
the
party
in,
and
we
can't
tell
when
people
either
need
more
drinks
or
someone
set
a
small
fire
in
one
side
of
the
building,
or
you
know
anything
like
that
and
its
large
fines
resolvable.
But
we
need
to
increase
our
pointed
communication
and
github
issues
right
now.
H
I,
don't
think
we're
gonna
clean
that
up
anytime
soon,
so
I'm
looking
for
alternatives,
and
so
if
anyone
has
any
I
would
love
if
one
other
thing
to
let
people
notice
is
when
I'm
talking
to
other
cigs
they're
beginning
to
look
at
this
cig
as
a
model,
both
the
work
that
Erin's
doing
the
way
that
we
have
so
many
people
contributing
I
know
that
we're
not
doing
as
good
a
job
responding
to
some
of
the
larger
efforts
that
are
coming
into
it.
But
in
many
ways
you
know
it's
pretty
well
structured
and
I.
H
Think
that
you
know
it's.
If
we
can,
you
know
unlock
this.
It
might
be
another
model
for
other
things
in
Cooper
Nettie's
to
to
figure
out.
So
you
know
like
the
cig
face
to
face.
I
think
we
pioneered
that
I
can't
say
for
sure,
because
I've
been
here
so
short
but
I
do
think
it's
the
success
that
other
people
are
looking
at
and
saying
was
useful
and
then
the
other.
So
so,
let's
start
using
the
cigs
storage
channel
more
and
then
trying
to
direct
people
to
it.
H
C
F
D
H
D
The
IRC
hack
instructions,
and
then
you
guys
can
also
just
use
the
IRC
client
to
slack.
So
you
don't
even
have
to
go
into
slack
you
just
log
in
to
that
server
through
your
normal
IRC
client,
whatever
you're
using
and
then
you
can
you'll
miss,
of
course,
like
the
graphics,
but
you'll
still
be
able
to
talk
and
be
notified
within
IRC.
Instead
of
having
to
use
I.
Don't.
H
Want
us
to
build
like
two
channels
where
people
are
lot
like
to
me
if
we
can
just
have
a
pointer
to
a
log
and
then
have
everyone
else
on
slack,
because
that's
the
lingua
franca
of
Cooper
Nettie's
for
now,
you
know.
I
worry,
then,
that
we
have
people
talking
in
the
IRC
and
I.
Don't
know
how
this
works.
In
all
honesty,
it's.
D
H
H
H
H
If
there's
one
else
is
interested
in
helping
us,
please,
like
let
us
know,
because
we're
trying
to
make
some
koopa
Nettie
stuff
work
with
the
containerized
NFS
clients
just
sort
of
a
request
for
help,
and
if
you,
if
you
are,
you
know,
feel
like
you're
an
expert
and
that
you
could
maybe
just
answer
some
of
our
questions
as
we're
trying
to
container
eyes
these
pieces
and
get
a
better
understanding
of
what
is
sort
of
table
stakes
for
NFS.
Please
reach
out
to
sod
or
Jing
Xiu.