►
From YouTube: Kubernetes SIG Storage 20181025
Description
Kubernetes Storage Special-Interest-Group (SIG) Meeting - 25 October 2018
Meeting Notes/Agenda:
Find out more about the Storage SIG here: https://github.com/kubernetes/community/tree/master/sig-storage
Moderator: Saad Ali (Google)
Chat Log:
A
All
right
today
is
October
25
2018.
This
is
the
meeting
of
the
kubernetes
storage
special
interest
group.
As
a
reminder,
this
meeting
is
public
recorded
and
posted
on
YouTube
today
on
the
agenda,
we're
going
to
go
over
the
planning
spreadsheet
to
get
an
update
on
the
items
that
we're
working
on
as
a
sig
for
the
1.13
release.
If
there
are
any
PRS
that
need
attention
or
design
reviews,
please
feel
free
to
add
them
to
the
agenda.
A
A
It
looks
like
what
is
not
on
the
call
so
we'll
skip
over
that
and
get
an
update
from
him
next
time,
we're
going
to
skip
over
the
CSI
drivers
and
get
an
update
at
the
end
of
the
quarter
on
those
next
up
is
moving
the
common
CSI
mount
library
or
moving
the
mount
library
to
an
external
repo.
This
is
the
mount
library
that
exists
in
the
core
of
kubernetes.
Today
we
want
to
be
able
to
reuse
it
for
CSI
drivers.
Travis
is
helping
drive
this
work,
any
updates
on
this
Travis.
Yes,.
A
B
Okay,
I
was
gonna,
say
Michelle
reached
out
to
me
and-
and
she
pointed
out
that,
not
only
is
you
know,
six
storage
interested
in
this,
but
also
sick
cloud
provider
now
so
I'm
actually
working
on
it
today
and
tomorrow.
So
it's
it's
actually,
my
my
number
one
priority
at
this
point
because
I
have
the
bandwidth
to
work
on
it.
So
awesome.
A
We
might
want
to
sort
out
if
this
library
should
live
in
the
staging
directory
of
kubernetes
kubernetes
or
this.
This
library
I
think
there
is
some
weirdness
about
if
kubernetes
kubernetes
needs
to
depend
on
external
code
that
will
also
be
consumed
as
a
library
by
other
external
code
should
be
part
of
staging.
Yes,.
A
Okay
and
then
that
gets
published
out
to
a
new
repo
and
then
the
CSI
drivers
can
pull
it
from
there
yep
all
right,
perfect.
All
right.
Thank
you
for
the
update
there.
That
looks
like
it's
on
track.
Next
up
is
the
pluggable
and
to
end
test
framework
that
patrick
has
been
driving
is
Patrick
on
the
line
by
any
chance.
I
know
he
is
Germany
time,
so
you
might
not
be
able
to
make
it
so
I.
D
A
C
A
A
And
the
deliverable
here
is
a
design,
so
that
seems
like
it's
on
track.
Next,
up
is
for
the
CSI
external
containers
to
move
leader
election
to
a
lease
object.
I
know:
Chang
made
a
change
to
the
external
provisioner
and
external
attacher
to
make
leader
election
optional,
because
it
currently
requires
additional
are
back
permissions
to
the
endpoint
object,
which
not
everybody
wants
to
hand
out.
So
this
is
something
we're
looking
forward
to
any
updates
on
SMAP.
E
E
A
A
Next
up
we
have
CSI
driver
we're
gonna
skip
over
that
then
we've
got
the
CSI
I
scuzzy
common
library
so
looks
like
the
initial
PRS
were
created
to
populate
that
repo,
and
there
were
some
discussions
about
what
the
features
should
be
and
where
this
library
should
go.
Any
updates
on
that
John
then
yeah.
F
So
I
did
a
little
update
to
that
initial
poll,
so
I
think
the
the
idea
should
be
first,
merge
it
as
a
golang
library
that
people
can
consume
and
then
from
there
we
can
figure
out
what
we're
going
to
do
as
far
as
an
actual
service.
And
so
if
we
could
you
step
one
and
then
we
can
follow
up
with
with
other
changes.
We've
tested
it
now.
F
Matt
Smith
from
de
terre
has
been
doing
a
bunch
of
work
on
this,
as
well
with
running
it
in
containers
as
ice
cozy
be
running
into
container,
as
well
as
the
initial
model
and
with
running
it
a
plugin
in
a
using
the
know,
dice
cause
you
need
both
seem
to
work.
They
each
have
their
drawbacks,
but
they
both
seem
to
work.
So
we
could
probably
even
offer
both
models.
Nice.
A
F
F
Actually,
I
put
it
on
the
agenda
to
ask
if
you
guys
are
cool
I'm,
just
gonna
merge
it
at
this
point,
because
it's
been
sitting
there,
you
know
fairly
idle
for
quite
a
while
if
Matt
has
a
chance
to
run
through
it
with
his
multipath
test
again,
that
would
be
great
and
then,
after
that
we
can,
we
can
merge
it
and
then
we
can
start
building
on
it.
Okay,.
A
A
A
The
common
library
can
be
consumed
by
other
folks
who
are
writing
ice
customized
Guzzi.
You
know
CSI
drivers,
the
generic
ice
cozy
driver
will
just
be
kind
of
replacement
for
the
entry
I
scuzzy
driver
that
we
already
have
so
excellent
work
there.
Thank
you
for
the
update
Oh.
Next
we
have
the
CSI
spec
changes
that
home
on
has
been
championing
in
around
dynamic
max
capacity
provisioning.
G
H
G
G
A
A
A
A
H
A
I
A
A
I
Yes,
so
there
were
two
items:
the
first
one
is
more
or
less
done
it's
about
spreading.
The
stateful
sets
appropriately,
as
is
done
in
the
entry
code.
So
that's
waiting
for
approval
from
yon
or
anyone
else
in
the
rumors
and
the
second
item
there
is
getting
an
external
hook
set
up
for
the
see
a
side.
Note
info
and
that'll
actually
ugly.
A
A
F
I've
actually
been
working
with
him
on
that
he's
got
a
number
of
PRS
back
in
my
repo
for
the
common
version
of
this
that
we
started
on.
He
was
actually
you
know,
we're
gonna
get
together
this
week
and
go
through
helping
about
getting
an
initial
commit
up
to
the
to
the
CSI
right
repo,
so
I'll
I'll
check
with
him.
He
is
working
on
it
he's
making
progress
so.
A
D
A
D
H
D
A
A
E
A
G
A
I
A
A
G
A
A
Okay,
it
looks
like
neither
of
them
are,
but
there
are
a
handful
of
items
that
they're
tackling
for
this
quarter,
including
adding
deletion
protection
having
a
new
reclaim
policy
similar
the
reclaim
policy
that
we
have
on
PvP
BC,
adding
topology
support
into
CSI
for
snapshots
and
adding
it
creating
a
design
for
application,
consistent
snapshots
and
so
far
all
of
those
items
seem
to
be
on
track.
If
you're
interested
in
any
of
that
work
to
be
sync
up
with
Jiang
and
Shang,
and
they
can
point
you
to
the
right
places.
A
We're
gonna
keep
this
feature
in
alpha,
we're
just
going
to
continue
to
polish
it
and
add
functionality,
so
everything
there
seems
on
track.
Next
up
is
automated
end-to-end
testing
of
CSI
sidecar
containers
before
release
both
major
and
minor,
so
Patrick
is
working
on
this
as
well.
I
think
some
of
the
work
that
Michelle
is
doing
for
the
GCPD
CSI
drivers
actually
end
up
helping
us
a
lot
here
as
well.
A
Now
for
the
CSI
items
moving
to
1.0,
this
is
on
track.
There
are
a
handful
of
issues
that
are
currently
open
for
1.0.
If
you
go
to
the
container
storage
interface,
spec
repo,
you
can
look
at
the
1.0
milestone
to
see
what
those
issues
are.
If
you
have
any
major
concerns
with
a
1.0
spot,
now
is
the
time
to
get
those
raised
either
raise
them
in
this
forum
or
at
the
CSI
community
meeting
once
the
spec
goes
to
100
making
backwards
compatibility
rating
changes
will
be
much
more
difficult,
if
not
impossible.
A
A
A
Yeah
so
I
think,
based
on
our
meetings
with
Jordan.
That
work
is
now
unblocked,
so
we
can
go
ahead
and
proceed
with
this,
but
the
challenge
is
that
there
is
a
considerable
amount,
additional
logic
that
needs
to
go
into
the
inline
CSI
support
PR
to
support
this
backwards
in
a
backwards
compatible
manner.
So
that
means
that,
while
this
is
unblocked,
we
have
more
work
on
the
inline
volume
side.
A
H
I
talked
to
Renault,
who
is
the
lead
own
on
the
on
the
PR
for
the
device
plug-in,
and
he
was
able
to
update
the
list
of
tasks
that
need
to
happen
for
for
this
to
go
to
GA.
It
looks
like
most
of
the
tasks
have
to
do
with
basically
ways
to
show
that
the
the
code
is
ready
because
they
won't
have
any
back
brake
backward
compatibility
issues
so
I'm
helping
him
with
that
and
then
I
think
there's
one
PR
where
he
actually
has
to
do
code,
so
I'll
be
following
up
with
them
on
that.
A
A
Next
up
is
the
CRD
installation
mechanism.
I
have
been
going
around
to
other
SIG's
to
try
to
get
this
picked
up.
I
went
to
API
machinery.
They
said
that
they
were
not
responsible
for
the
installation
of
objects,
and
that
was
the
responsibility
of
sig
cluster
lifecycle.
I
went
to
sig
cluster
lifecycle
and
I.
Think
Philippe
agree
to
pick
up
dis
work,
but
not
for
this
quarter
for
the
time
being,
in
order
to
unblock
cigs
storage,
what
we're
doing
is
using
the
existing
add-on
manager
as
a
mechanism
to
install
the
CR
DS.
A
Moving
the
CSI
driver
and
CSI
note
objects
from
alpha
to
beta
was
locked
on
CR
the
installation,
any
other
status
update
there.
You
know
I
haven't
even
started,
but
I
hope
this
is
going
to
be
easier
right
it
unless
we
run
into
the
same
set
of
upgrade
concrete
issues
that
Jordan
pointed
out
other
than
that
probably
should
be
fairly
simple.
Okay,.
A
H
H
A
Is
it's
only
required
for
aunt
line
to
go
to
alpha,
so
whenever
we
introduce
a,
we
need
to
ensure
that
this
additional
validation
is
in
there
right,
but
double
check.
Two
things
would
Jordan
one
is
that
and
the
second
is
if
he
would
be
willing
to
help
us
with
that
code,
since
he
understands
it
better
than
anyone
else.
H
A
F
F
A
Good
point:
okay,
I!
Think
that's,
okay,
we'll
get
a
better
understanding
of
what
we
want,
and
hopefully
the
people
concept
means
that
it'll
be
fairly
easy
to
implement
next
quarter,
then
cool
next
up
is
data
population
supporting
other
data
sources
like
cloning.
This
honestly
is
one
of
the
things
that
I'm
really
really
excited
about.
A
A
G
Yeah
so
there's
a
there's,
a
proposal
to
basically
let
you
specify
UID
so
far
we
have
only
been
using
the
first
group
to
change
the
permissions,
but
there's
a
this
change
to
allow
users
to
specify
UID.
So
the
permissions
could
be
changed
to
match
both
your
ID
and
FS
group.
So
we
need
to
like
process
that
and
make
that
change
a
homicide
on
it,
but
I'm
trying
to
do
this
for
312.
Oh
sorry,
1.13
man.
A
A
Okay,
so
I
believe
this
work
hasn't
begun.
Yet,
hopefully
you
can
get
started
on
this
soon
right
now,
there
are
two
sets
of
functionality
that
are
kind
of
stuffed
into
driver
registrar
that
are
orthogonal
and
realistically
we
need
to
split
the
driver,
registrar
into
two
separate
containers
and
two
separate
repos.
One
is
no
driver,
registrar
one
is
cluster
driver,
registrar
and
Luis
will
be
driving
this
where,
hopefully,
he
can
get
started
on
this
soon.
Next
up
is
improving
dependency
injection
for
volume,
plugins.
L
L
Is
that
looking
at
the
codebase
in
the
past,
we
have
some
issues
related
to
how
our
different
dependencies
are
injecting
into
a
volume
plugins,
for
example,
like
currently
most
of
the
dependencies
are
coming
through
the
volume
Coast
interface,
which
means
every
single
plug
plugging
it
has
to
take
advantage
of.
It
has
asked
us
to
basically
follow
in
reality.
A
C
A
A
Hard,
okay,
that's
great,
then
that's
all
that
I
have
for
status,
updates
reminders
that
feature
freeze
was
two
days
ago.
So,
if
you
don't
have
a
feature
committed
as
a
feature
issue,
then
technically
we
cannot
get
it
into
this
release
and
code
freeze
for
the
items
that
we
are
committed
to
is
November
15
and
clothe
code
slush
is
probably
about
a
week
before
that.
So
try
and
get
your
features
in
as
soon
as
possible,
we're
already
at
October
25.
So
it's
getting
very
tight.
A
A
Please
take
a
look
at
it
if
you're
interested
at
all.
If
you
have
objections
that
are
more
along
the
lines
of
feature
level,
things
that
we
should
do
instead
of
blocking
the
initial
commit,
let's
do
follow
up.
Here's
to
address
those
the
kind
of
things
that
should
block
this
PR
are
things
like
I
think
inherently
this
is
structured
incorrectly
or
is
functionally
incorrect.
Things
like
that,
so
please
follow
up
on
that
PR
and
thing
else.
You
want
to
add
on
that
John.
A
A
That
is
currently
planned
as
part
of
that
session.
At
the
end
of
the
session,
we'll
end
up
doing
a
panel
discussion,
so
some
of
you
will
probably
be
asked
to
participate
in
that
are
there
other
events
that
folks
are
aware
of?
There
is
a
containers
native
storage
day,
the
day
before
the
conference
on
Monday
December
9th
Brad.
Do
you
want
to
add
anything
else
on
that.
D
A
I
don't
know
I'm
gonna
be
able
to
make
it
out
so
I'm
still
trying
to
figure
that
out
over
the
next,
hopefully
I'll
know
by
the
next
meeting
and
I.
Don't
want
to
block
it
on
that.
If
you
guys
feel
like
it
might
be
useful
feel
free
to
go
for
it.
If
nothing
else,
we
can
do
a
small
impromptu
meeting
or
like
a
dinner
or
something
again.
A
Any
other
events
folks
are
aware
of
related
to
SIG's
storage
or
storage
in
general.
I
know
that
the
CNC
F
storage
workgroup
is
going
to
have
two
sessions,
they're
planning
on
presenting
a
white
paper
that
lays
out
the
landscape
sea
cloud
native
storage
landscape.
This
group
was
previously
a
little
bit
controversial
when
they
tried
to
do
this
last
year,
because
the
paper
was
trying
to
define
what
is
and
is
not
cloud,
storage
and
potentially
leaving
folks
out,
which
resulted
in
a
lot
of
controversy.
A
The
new
version
of
the
landscape
paper
seems
to
be
much
more,
not
opinionated
in
that
regard.
It's
more
about
here
are
the
different
types
of
storage.
Technology
is
one
of
the
explicit
non
goals
for
it
is
trying
to
define
what
is
and
is
not
cloud
native
and
doing
any
sort
of
recommendation
for
storage.
So
I
don't
anticipate
as
much
controversy
this
time
around
for
folks
that
are
interested
feel
free
to
attend
that
cube.
Con
storage
talks.
I
know
there
is
a
handful
of
talks.
A
A
All
right,
if
you
have
a
talk,
please
go
ahead
and
add
it
to
the
dock
here
and
that's
it
we'll
reconvene
in
two
weeks
approaching
code
freeze
fast,
so
please
get
your
PRS
in
early
make
time
for
code
reviews
and
let's
get
finish
off
this
year
strong.
Thank
you
very
much
and
see
you
in
two
weeks.
Thanks
up
thank.