►
From YouTube: Kubernetes SIG Storage Meeting 2021-06-03
Description
Kubernetes Storage Special-Interest-Group (SIG) Meeting - 03 June 2021
Meeting Notes/Agenda: https://docs.google.com/document/d/1-8KEG8AjAgKznS9NFm3qWqkGyCHmvU6HVl0sk5hwoAE/edit#heading=h.rirnvdqiskmp
Find out more about the Storage SIG here: https://github.com/kubernetes/community/tree/master/sig-storage
Moderator: Saad Ali (Google)
A
All
right
today
is
june
3
2021.
This
is
the
meeting
of
the
kubernetes
storage
special
interest
group.
As
a
reminder,
this
meeting
is
public
recorded
and
posted
on
youtube
on
the
agenda.
Today,
we're
going
to
go
over
the
122
planning
spreadsheet
to
see
what
the
status
of
the
various
items
are,
that
the
sig
is
committed
to
dates
to
keep
in
mind
the
next
upcoming
one
is
code
freeze
on
july
8th
we
have
a
little
more
than
a
month
left
so
plenty
of
time.
A
If
you
have
a
feature
that
you're
working
on
that
you're
interested
in
writing,
a
blog
for
please
make
sure
to
reach
out
to
one
of
the
sig
leads,
and
we
can
help
you
get
registered
for
that
and
if
you
have
any
pr's
that
need
attention
designs
that
need
reviewed
or
anything
else
that
you'd
like
to
discuss,
feel
free
to
add
to
the
agenda
doc
and
we
can
go
over
it
after
the
planning
spreadsheet
this
agenda.
Doc
link
is
in
your
invitation,
calendar
invite.
A
A
A
Month,
I
think
a
month's
on
the
line,
but
I
cannot
hear
him.
Can
you
hear
me
now
sorry?
Oh
yes,
there
we
go.
Okay,.
B
A
Cool,
so
I
think
that
should
be
okay
to
merge
soon.
The
plan
for
the
csi
spec
is
to
cut
a
release
on
monday
next
week
and
so
I'll
take
a
look
at
getting
that
merged
this
week.
A
B
So
we
dropped
that
copy.
Allow
volume,
expansion
item
from
this
quarter
pick
and
I
have
reached
out
to
api
reviewers,
like
tim
and
jordan,
about
like
how
to
deprecate
if
you're
going
to
duplicate
the
parameter
from
switch
class,
because
it
would
be
a
change
in
api
change
sort
of
thing,
so
we
need
guidance
from
them.
B
Then
yeah,
but
that's
the
only
outstanding
item
left
actually
as
far
as
issues
are
concerned.
But
when
we
talked
about
this
in
the
csr
implementation
meeting,
we
we
do
not
think
that
this
should
be
like
copying.
The
allow
volume
expansion
should
be
a
blocker
for
resizing
as
such,
going
to
ga
like
not
in
this
release,
but
next
series,
if
you
want
to.
A
And
was
there
general
consensus
on
that,
or
is
that
up
for
debate.
B
But
I
will
schedule
we'll
obviously
have
like
one
or
two
more
calls
when
we,
when
we
plan
resizing
for
ga
but.
A
B
So
kept
mismatched,
I'm
working
on
implementation.
I
have
to
start
split
the
stories
split,
the
card
so
that
I
can
work
badly.
I
haven't
done
that,
but
I
I
do
have
a
poc.
C
So
I
think
humble
replied
to
the
pr
saying
that
he
would
do
a
rebase,
I'm
not
sure
if
he
has
time
to
move
this
forward,
I
can.
I
can
sing
up
with
them
offline.
D
A
All
right
next
item
is
issues
related
to
assuming
mount
points.
Volumes
are
mount
points.
Jing.
Are
you
on
the
line.
E
Yes,
so
yeah
there's
no
further
update,
but
there's
some
p,
because
this
issue
related
to
a
race
we
discovered
but
not
yet
be
able
to.
You
know,
have
a
good
solution
or
they
relate
to
the
part
and
volume
management
on
this
condition.
E
By
clinton
so
which
has
a
like
a
big
change
related
to
like
a
pod
status
management
like
update
which
might
help
on
this
race,
so
we
are
trying
to
review
that
change
and
we
have
a
bit
more,
maybe
address
some
of
the
issues
here.
A
A
Okay,
we're
gonna
mark
that
one
as
no
update
pvc
inline
ephemeral
volumes
working
with
the
csi
driver.
Any
updates
on
that.
G
H
A
Okay,
next
up
is
csi
out
of
tree
moving
iscsi
driver
image,
building
testing
ci
cd
documentation
christian.
Do
you
have
any
updates
on
this.
J
A
A
A
J
Yeah
actually,
yesterday
I
had
a
call
with
a
christian
and
he
showed
me
already
the
stuff
so
yeah
so
everything
he
did
before
and
I
think
actually
we
should
go
with
a
more
generic
implementation,
but
yeah.
I
want
to
see
how
to
proceed.
A
So
if
there's
a
few
different
proposals,
I
think
the
best
approach
may
be
to
try
and
find
some
time
with
just
kind
of
the
wider
sig
and
do
a
30
minute
or
one
hour
meeting
and
talk
through
the
different
ideas
and
see
if
we
can
come
to
a
consensus
on
which
approach
is
the
best.
A
Cool.
Thank
you
for
helping
with
this.
H
Yeah
so
we
actually
talked
about
this
in
yesterday's
csr
meeting,
so
I
think
I
will
write
down
some
more
videos
on
various
different
directions
because
we
had
a
meeting
last
time
right.
So
we
talked
about
looking
at
how
to
use
those
two
enhanced
scheduling,
but
I
thought
about
that.
I
don't
see
how
this
can
help
with
scheduling
actually,
so
I
think
I
just
need
to,
and
then
we
also
got
some
new
requests
on
making
this
appear
as
a
matrix
yeah.
H
So
there
are
various
requests
and
there's
a
reaction.
So
I
have
a
doc
that
have
some
of
this
that
I
just
want
to
write
down
more
details
then,
and
then
we
can
have
a
meeting
to
talk
about
it.
K
This
one
is
the
development
of
the
entry
feature
is
done
and
I'm
adding
automated
tests
and
it
will
be
ready
for
reviewers,
probably
early
next
week.
I
want
to
say.
K
A
K
Yeah,
I
can
do
an
update
here,
so
I
forget
what
the
last
update
was
so.
A
K
Out
of
this,
we
have
been
spending
a
lot
of
issues,
a
lot
of
time,
arguing
about
sort
of
esoteric
api
issues
that
were
not
so
core
to
the
feature,
and
and
now
that
we've
sort
of
have
this
new
viewpoint
that
we
can
just
deliver
an
alpha
out
of
tree
and
get
the
kept
approved
later,
we're
really
focused
on
the
the
core
issues
of
how
are
people
going
to
consume
this
thing,
what
is
the
downward
api
going
to
look
like,
and
then
how
does
that
affect
things
like
bucket
access
and
some
of
the
neglected
parts
of
the
original
design?
A
A
All
right,
thank
you,
ben
for
that
update
and
for
folks
who
are
interested
in
any
of
those
discussions
immediately
after
this
meeting
will
be
the
will
be
the
cozy
meeting
so
feel
free
to
attend.
That
next
item
is
change.
Block
tracking
last
looks
like
we
haven't,
had
an
update
here
for
a
while.
H
Yeah,
I
think,
still
right
now,
there's
we
haven't
done
any
update
on
the
api
design
part.
I
think
we
are
right
now
working
on
some
documentation.
A
All
right,
thank
you.
Shane
next
item
is
new
rwo
access
modes
chris
has
been
working
on
this
last
status
was
kept,
was
updated.
Chris,
are
you
on
the
line
by
any
chance.
L
L
A
Cool.
Thank
you.
Chris
next
item
is
the
csi
proxy
for
windows
looks
like
we're
doing
new
api
changes
required
for
driver
changes,
I'm
moving
it
to
ga
and
I
think
mauricio
jing
deep
are
working
on
this
any
of
those
folks
on
the
line
for
an
update.
E
E
I
think
two
of
them
merged
and
the
rest
of
them
should
be
able
to
merge
like
very
soon
by
this
week,
and
then
the
plan
would
be
cut
rc
release
early
next
week.
So
we'll
ask
the
driver
owners
to
test
it
out,
and
then
we
can
have
a
final
like
a
v1
api
upgrade
to
v1
and
have
this
release.
A
E
A
Okay,
next
item
is
csi
migration,
so
csi
migration
has
the
core
logic
and
then
logic
for
specific
entry
drivers.
A
So
for
the
core
logic,
jawway
or
matt,
do
either
of
you
have
an
update
here.
Last
status
was
bug
on
bag.
A
M
Yes,
so
that
bug
has
already
been
fixed,
cherry
picking
is
in
progress.
M
I
think
in
general,
there's
no
big
issues.
There
are
a
couple
of
small
issues
like
up
up.
They
upgrade
the
annotation
to
ga
version,
or
something
like
that.
I
I
need
some
time
to
get
some
time
to
fix
that
other
than
that.
I
think
it's
all.
A
Good
cool,
good,
stuff,
shawway,
thank
you
and
related
next
step
is
deprecation
of
entry.
Drivers
that
will
not
be
migrated.
Remove
plugins
that
were
previously
deprecated.
A
M
I
think
yeah
lost
the
status
is
this.
I
think
this
is
still
true.
I
had
a
pr
out
for
the
website
update.
M
I
also
actually
find
I
find
out
that
on
1.17,
jc,
pd
and
aws
has
already
been
deprecated,
which
is
kind
of
unexpected.
I
haven't
updated
that
in
this
part
of
the
website.
D
M
A
Okay
and
so
for
the
rest
of
these
drivers,
let's
get
an
update
for
vsphere.
First.
H
I
think
stevia
has
a
pr
out
for
moving
the
topology
label
to
ga
other
than
that,
because
everything
else
is
the
same.
I
don't
know
if
your
jaw
will
you
get
any
more.
M
Update
there
is
a
windows
support.
M
It's
sorry,
I
forgot.
I
need
to
check
the
the
spreadsheet
we
have
for
csm
migration,
okay,.
A
No
worries
I'm
gonna,
mark
windows,
support
as
outstanding,
and
we
can
get
an
update
next
time
on
that
next
item
is
azure
disk
and
then
azure
file.
Anybody
have
updates
on
that.
M
N
Yeah,
we
are
still
trying
to
figure
out
the
sidecar.
A
F
D
Yeah,
that's
accurate,
there's
a
pr
out
in
the
evs
csi
driver
repo
for
windows
support.
That's
just
pending
some
other
stuff.
You
see
csi
pages
and
stuff.
A
And
finally,
we
have
cfs
and
seth
rbd
last
status
here
we're
still
in
discussion
whether
we're
going
to
support
it
or
not.
O
Right
right,
yeah,
yes,
so
yeah,
I
had
some
discussions
on
this
with
the
victor
audience
group
and
also
in
general
teams
at
red
hat,
carried
out
some
more
discussions
on
the
pr
car.
But
it's
not
enough
to
take
the
final
call,
because
we
shall
also
also
say
I
have
to
get
the
community
interest
on
this.
So
I
have
an
action
item
to
send
an
email
to
the
company
that
is
mailing
list
and
see
the
community
interest
before
we
take
the
final
call.
Okay,.
A
A
All
right,
thank
you,
humble
and
good
to
see
you
back
next
item
is
updated.
Topology
labels
from
beta
to
ga
for
vsphere,
and
so
it
sounds
like
we
got
an
update
for
that
already.
You
want
to
add
anything
to
that.
Shane.
A
Next
is
always
honor
reclaim
policy
last
status
here
was
we're
going
to
move
this
to
123.
Is
that
still
the
case.
H
Yeah
so
well
yeah.
He
will
get
back
to
this
right
now.
There's
no
update,
I
think
we'll
get
back
and
we'll
come.
A
Okay:
next
item
is
control
volume,
mode
conversion
between
source
and
target
pvc.
A
H
I
don't
know
if
we
have
a
masaka
here,
so
jordan
has
some
concerns.
I
think
those
are
still
outstanding.
Those
are
not
resolved
yet,
so
I
think
we
probably
missed
the
deadline
for
requesting
an
exception.
It
does
not
look
like
it
can
be
resolved
easily
and
jordan
is
also
tied
up
with
a
1.22
committed
items.
A
And
got
it
so
maybe
let's
switch
this
to
design.
I
guess
it
is
designed
for
this
quarter.
A
G
B
We
are
not
yeah,
we
is.
I
wish
I
responded
to
you
on
slack,
like
it's
just
a
design
item
and
we
haven't
heard
anything
from
him
from
the.
M
Not
really,
I
think,
I'll
pee
him
again
and
see
if
he
gets
times.
Okay.
A
A
I
think
it's
fine
okay,
so
let's
leave
it
for
now
and
then,
if
we
decide
that
it's
not
being
committed,
then
we
can
go
ahead
and
remove
it
all
right.
Thank
you,
hamad
and
jaway
for
the
update.
Next
item
is
ungraceful.
Node
shutdown
prototype
for
this
quarter
last
status
was
updated,
kept
try
to
get
it
merged
as
provisional
check
with
sig
release.
If
provisional
is
allowed.
H
Yes,
that
is
allowed.
Actually
we
could
still
do
that.
So
I
think
I
think
we
talked
about
having
the
cap
at
least
reviewed
and
because
that
has
in
the
cab
we
talked
about
how
to
how
to
use
this
change.
That
band
has
made
in
css
back
the
first
detach
and
then
so
it's
yours,
you're
planning
to
cut
it
in
a
monday
right
next
monday.
H
J
H
Should
let
you
know
on
monday
and
see
how
that
is
and
and
see
if
we
can
get
pr
merged,
so
I
got
some
review
comments
from
sean
jean,
so
I'll
be
addressing
them.
A
B
That
was
different
design.
This
is.
B
No,
actually
I
didn't
I
didn't
I,
the
cap
didn't
merge.
Last
I
looked
so
I
haven't
had
a
chance
to.
B
B
And
the
owner
is
different
person:
okay,
akihito.
B
A
A
Okay,
so
next
item
is
with
sig
apps
address.
What
is
this
address?
Issues
pvc
created
by
stateful
set
will
not
be
auto,
removed,
kk
or
matt.
Any
updates
on
this.
N
I'm
continuing
to
work
on
the
implementation.
The
api
change
is
basically
approved.
A
Awesome
sounds
like
good
progress
and
matt
I'll
go
ahead
and
mark
you,
as
the
primary
here.
A
H
Though
shalini
is
looking
at
this,
I
think
last
time
I
checked
with
her,
she
said:
she's
still
reviewing
it
trying
to
update
the
cap,
so
I
need
to
sync
up
with
him
again.
A
H
Yeah,
so
we
are
trying
to
get
signaled
to
review
and
approve.
A
A
Next
item
is
sig
architecture,
kubernetes
utils
mounts
split
it
into
a
new
repo
and
add
end-to-end
testing.
Do
the
formal
release
for
this
sereni
was
tracking.
This
last
status
was
that
the
pr
was
close
to
merge
any
updates.
Since
then,.
F
I
need
to
double
check,
but
I
I
think
it
has
merged
got
it
either
either
that
or
like
he
needs
to
rebase
or
something
like
that.
But
it's
been
lgt
and
then
approved
at
least.
A
Okay,
so
next
item
is
prioritization
on
volume
capacity
with
six
scheduling
this
will
be
staying
in
alpha
and
122.
Any
new
status
updates
here.
F
No,
I
need
to
sync
up
with
yay
chang
to
see
if
there's
anything
else
that
needs
to
be
done.
I
know
there
was
like
one
api
change
that
potentially
was
deferred.
A
A
Awesome
thanks
michelle
and
with
that
we
have
completed
the
planning
spreadsheet,
update.
Let's
move
on
to
the
remaining
items,
so
I
don't
see
any
prs
to
discuss
or
designs
to
review.
One
item
of
miscellaneous
discussion
from
michelle
is
flex
volume
deprecation
michelle.
You
want
to
talk
about
this.
F
Yes,
so
on
our
website
on
our
like
doc's
website
and
sig
pages
flex,
volume
is
marked
as
deprecated,
but
we
actually
have
not
sent
out
a
deprec
official
deprecation
notice
in
like
the
release
notes
for
it.
So
I
think
I
just
put
this
item
here
that
you
know
we
should
probably
target
doing
that
this
release.
F
I
think
over
the
last
couple
of
releases,
various
functionality
of
flex
volumes
has
been
being
removed.
I
think
from
like
the
I
think,
with
the
switch
to
distur
lists
in
the
controller
cube,
controller
manager
has
basically
made
flex
volumes,
not
workable
by
default.
You'll
have
to
like
build
your
own
custom
image
of
cube
controller
manager
in
order
to
make
that
work.
N
A
Release
sounds
good
to
me.
Are
there
folks
on
the
call
who
are
still
using
flex
volume
by
any
chance.
B
B
Not
sure
my
my
personal
feeling
says
that
we
should
just
deprecate
the
control
plane
thing,
because
we
have
very
little
like
nobody's
using
it
and
everything,
but
but
for
other
stuff
we
have
tests,
we
are
running
here,
we
know
it
works
and
there
are
people
using
it.
I'm
not
sure
if
they
will
respond
to
the
main
list,
but
there
are
people
using.
F
B
Yeah,
of
course-
and
it
has
to
have
similar
features
to
this
work-
do
we
are
applying
to,
but
we
have
no
plans
to
support
like
any
flex
volume
to
csi
migration,
the
people
who
are
using
those
drivers
they
will
have
to
pretty
much.
A
P
I
think
it's
big
windows
used
to
advertise
a
couple
of
select
volume
based
portions.
F
Yeah,
so
I
guess
it
sounds
like.
Maybe
we
do
still
have
some
active
users,
I
think
maybe
a
better
step
for
122
is
we
should
start
trying
to
advertise
the
csi
equivalents
instead
of
the
flex
equivalents
and
also
work
on
over
the
next
couple
of
releases.
Work
on
making
the
csi
equip
replacements,
more
stable
and
with
future
parity.
A
I
think
that
seems
reasonable.
Do
we
want
to
add
an
item
for
this
release
to
make
some
progress
here.
F
Yes,
so
we
can
say
I
don't
know
we
can
have
an
item
for
releasing
the.
F
F
I
guess
that's
something
we
would
need
significant
work
on.
I
think
right
now
we're
just
getting
the
iscsi
driver
just
working
on
linux
and
we
haven't
even
looked
at
windows
at
this
point.
P
P
A
proxy
iscsi
so
seems
like
he
may
be
working
on
something
for
iscsi
csi
on
windows,
and
he
was
pretty
active
in
that
area
last
couple
of
weeks,
okay,
but
yeah.
Definitely
that
api
group
is
alpha
for
sure
not
nowhere
close
you.
A
A
Okay,
so
andy
and
julie
have
done
some
work
here.
Michelle
and
matt
will
sync
up
with
them
to
see
if
they
can
commit.
Does
that
sound,
accurate.
B
A
B
B
So
should
we
like
what
should
be
the
signal
that
okay,
this
can
be
removed
safely,
for,
like
the
flex
volume.
A
I
think
flex
volume
is
probably
we
can't
remove
it.
We
just
need
to
mark
it
as
deprecated
and
kind
of
leave
it
as
a
stub
that
we
continue
to
support
to
some
extent
slowly
deteriorating
over
time,
but
trying
to
maintain
it
as
much
as
possible.
A
H
A
So
I
guess
it
sounds
like
this
cycle.
We're
gonna
work
on
some
of
the
csi
drivers.
We're
gonna.
Are
we
gonna?
Do
any
messaging
updates
this
cycle.
A
Do
we
want
to
send
out
any
preemptive
messages
to
sig
storage
or
to
kubernetes
dev,
to
say
hey
by
the
way
flex
volume
has
been
deprecated,
you
should
start
transitioning,
or
should
we
hold
off
on
such
messaging.
H
H
Until
michelle
mentioned
this.
Actually
we
have
not
really
sent
out
a
formal
message,
but
if
people
will
read
about
the
docs
and
they
will
see
that
it
says
it's
deprecated,
so
we
should
get
our
message
consistent.
A
I
think
that
sounds
okay
to
me.
I'll,
go
ahead
and
add
an
item
here,
stand
out,
deprecation
notices
for
flex.
A
All
right
anything
else
on
flex
volume,
deprecation.