►
From YouTube: Kubernetes SIG Storage Meeting 2021-11-04
Description
Kubernetes Storage Special-Interest-Group (SIG) Meeting - 04 November 2021
Meeting Notes/Agenda: https://docs.google.com/document/d/1-8KEG8AjAgKznS9NFm3qWqkGyCHmvU6HVl0sk5hwoAE/edit#heading=h.hjhvd5gxf91p
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
november
4
2021..
This
is
the
meeting
of
the
kubernetes
storage
special
interest
group.
As
a
reminder,
this
meeting
is
public
recorded
and
posted
on
youtube.
A
A
This
is
where
we
track
the
items
for
that
that
the
sig
is
currently
working
on
for
the
next
release,
upcoming
deadlines
to
be
aware
of
november
16th
is
the
code
freeze.
A
So
if
you
are
working
on
any
features
for
the
upcoming
1.23
release,
please
be
aware
of
that
date.
All
features
need
to
be
complete
merged
approved
by
that
date,
and
let's
go
ahead
and
switch
over
to
the
planning,
spreadsheet
and
start
getting
status.
Updates,
I'm
going
to
create
a
new
column
here
for
today's
date.
D
C
Yeah,
indeed,
thank
you
jamon
for
actually
having.
F
A
G
A
Update,
oh
okay:
let's
take
a
quick
look
at
that
update
for
generic
ephemeral
volumes.
All
pr's
are
for
ga
and
one
two
three
okay.
A
D
A
A
I
No
so,
but
we
we
do
like
also
have
some
other
issues
related
to
one
point.
So.
I
Not
exactly
the
same
as
this
and
here,
but
it's
also
relevant
to
long
points,
checking
my
plans,
and
so
we
were
discussing
in
the
csi
economics.
I
Oh,
no,
the
the
performance
issue
is,
I
think,
being
addressed
by
different.
I
So
this
one,
I
think
I
still
need
to
verify
some
behaviors
to
see
it
is
actually
already
fixed
by
cooperative
changes.
Yeah
recently.
A
I
A
Okay
sounds
good,
let
me
add
a
quick
line
here.
So
let's
say
no
related
issues.
A
All
right,
thank
you.
Jing
next
item
is
pvc
and
line
ephemeral.
We
already
talked
about
that.
Thank
you
patrick.
A
Then.
We're
gonna
skip
these
two
items
and
go
to
volume
group
api
snapshot,
consistency
group
spreading.
H
I
don't
have
a
update
for
this
one.
I
think
I
need
to.
A
A
A
A
I
don't
think
we've
got
an
update
on
this.
Is
it
worth
continuing
to
tracking
for
this
quarter.
C
I
think
so
I'll
follow
up
with
julie.
I
mean
I
think
this
actually
may
be
pretty
close
to
oh.
C
The
smb
csi
driver
is
marked
as
a
ga
got
it,
so
that
part
at
least
is
done.
C
J
I
think
one
thing
to
check
here
to
see
if
it's
actually
going
through
our
build
pipeline.
J
A
Okay
sounds
like
there's
some
traction
here,
but
we
need
to
do
some
follow-up
validation
and
get
and
also
get
an
update
on
the
sif
site.
All
right.
Thank
you
both
for
that
update
next
item
is
sending
out
a
deprecation
notice
for
flex
volume.
The
last
update
here
was
proposed
to
be
added
to
123
release.
Notes
any
follow-up
here
are
we
done.
I.
A
And
mustafa
is
back
and
we'll
look
into
it
was
the
last
status
any
anything
new
here.
K
A
G
A
G
The
the
prototype
is
there
a
link
to
it
in
the
doc.
I'm
not
looking
at
it
right
now,
but
I
was
going
to
say:
where
is
that
prototype.
K
By
the
way,
mustafa
will
continue
his
work.
If
this
idea.
A
Okay,
thank
you.
Thank
you.
Misaki,
for
that
update.
Next
is
csi
volume,
health,
additional
metrics
and
or
events
moving
to
alpha
2.
last
status
was
pr
had
been
submitted
and
reviewed.
H
Pr
is
still
there
being
reviewed
and
I
think
there
is
a
one
ci
issue
that
the
team
is
trying
to
figure
out.
H
A
You
shane
next
is
in
fact
I
should
probably
start
marking
these
csi
volume
health,
programmatic
response.
This
would
be
a
design
for
the
quarter.
H
Yeah
so
I
talked
to
nick
about
this,
so
basically
we
just
discussed
what
other
common
scenarios
that
they
encounter
with
this
one
house
issue
and
what
we
could
do
to
fix
this.
Oh
yeah,
there's!
No.
We
don't
really
have
any
design
proposals.
G
G
But
it's
going
to
require
me
finding
more
time
than
I
think
I
have
so.
I
want
to
say
it's
a
risk.
I'm
still
working
on
it,
trying
to
get
everything
right
if
it
does
somehow
miss
the
deadline.
It'll
be
ready
early
on
in
the
following
release,
but
I
can't
commit
to
having
it
in.
A
No
worries
sounds
good,
I
don't
think
anybody's
completely
blocked
on
data
source.
So
that's
fine.
G
H
G
A
Okay,
I
would
don't
want
to
do
that
good
to
call
out
the
dependency
thanks.
Ben
next
item
is
cozy
object,
storage,
api.
Anyone
want
to
give
an
update
on
this
one.
H
I
think
it's
still
having
meetings
discussing
the
point
you
are
trying
to
address
the
comments.
A
H
So
I
think
fan
is
looking
at
it.
I
think
he's
trying
to
figure
out
how
to
add
a
proxy
service,
so
I
think
yeah.
A
I
Yes,
so
kind
of
we
talk
about
this
and
the
plan
to
work
on
it.
But
it's
because
it's
not.
I
There's
no
much
progress
yet
yeah,
okay,.
A
Sounds
good
and
then
next
up
is
csi
migration.
J
I
think
I
do
need
to
follow
up
with
him
there's.
I
think
one
outstanding
bug.
A
Cool
sounds
good,
thank
you,
michelle,
and
next
up
is
core
bugs
and
issues.
A
J
Yeah,
I
still
need
to
follow
up
on
this
one
yeah.
A
Sounds
good!
Thank
you
michelle.
Next
we
have
the
different
cloud
providers
and
their
entry
plugins.
So
first
up
is
vsphere,
deviant
shane.
Anyone
have
an
update
on
this
one.
H
I
think
deviant
is
trying
to
make
this
on
debug
by
default,
but
I
think
there
are
some
concerns.
H
Yeah,
I
think
it's
not
yeah
so
the
alpha,
meaning
that
dev
has
tested
it,
but
qe
has
not,
and
you
can
see.
I
think
it's
almost
out.
So
I
don't
think
you
know
in
2.4
that's
going
to
go
ga,
so
I
think
it
has
to
be.
H
Dev
has
checked
in
the
code
and
the
test
he
did,
but
I
think
that
qe
has
has
not
yeah.
So
that's
why
we
call
it
alpha
if
we
has
not
tested
it.
D
H
It
should
I
yeah,
I
thought
it's
supposed
to
be
today,
but
probably
delayed
a
little
bit
yeah,
it's
it's
so
long.
I
don't
think
the
qe
has
resources
to
test
windows
in
2.4
release,
so.
D
H
Then
we
actually
have
a
feature
gate
in
the
driver
and
I
actually
need
to
check
this
one.
Normally,
that's!
Okay!
So
if
you
have
a
feature
it
should
be
so
we
have
a.
We
have
to
have
a
feature
gate
for
that
particular
one.
So
a
user
will
have
to
go
turn
it
down.
Basically,
there's
a
flag:
it's
a
config
map,
so
we
can
yeah.
We
can
talk
about
this
one
separately.
We
can
talk
to
deviant.
D
A
All
right
cool
sounds
like
there's
a
follow-up
discussion
that
needs
to
happen
here.
Thank
you
both
for
the
update.
Next
item
is
azure
disk.
An
azure
file,
azure
disk
has
already
been
completed.
It
should
be
beta
and
on
by
default.
This
this
next
release
and
then
azure
file
we
were
waiting
last
status
update,
was
waiting
on
fs
group
to
go
beta.
D
A
D
C
This
is
continuing
getting
closer,
we've
got
the
core
tests
all
sorted
out.
Working
now
on
proof,
tests
turns
out,
they
need
a
pd
for
prometheus
stuff
and
the
usual
things
are
still
on
cube
test,
one
which
is
causing
problems,
etc,
etc.
But
I
think
we
are
kind
of
getting
to
the
point
where
we
could
probably
switch
it
on
by
default
and
then
deal
with
the
tasks
you
know
as
post
code
code
freeze.
A
Cool,
thank
you
matt
yeah.
It
sounded
like
there
was
a
lot
of
work
here
to
get
all
the
existing
antenna
tests
off
of
gcepd.
So
thank
you
for
all
that
hard
work.
Yeah
next
up
is
aws
windows
support.
Anyone
have
an
update
on
that
one.
A
M
So
the
peers
out
and
it's
being
reviewed
by
ayan.
H
So
yeah,
so
we
talk
to
you,
seek
us
about
our
photo
and
have
to
be
concerned.
We
have
some
different
suggestions,
so
we
want
to
update
the
proposal
and
click
on
that
suggestion
and
we
actually.
H
G
G
The
point
is,
I
think,
everyone's
agreed
on
like
what
should
what
it
should
do.
The
question
is
just
how
does
the
admin
say
who
gets
to
do
it
and
that's
that's
the
and
we
have
a
bunch
of
good
ideas
and
it's
just
a
question
of
which,
which
one
is
the
least
weird,
so
I
I
feel
very
good
about
this.
H
Yeah
so
I
asked
ronald
to
investigation
and
then
maybe
write
down
these
different
proposals,
and
then
we
can
talk
about
each
of
them
and
compare
and
see
what
we
want
to
do.
A
Makes
sense?
Okay,
so
it
sounds
like
there's
a
definitely
active
discussion
going
on
here.
Different
proposals
being
considered
and
sounds
like
there
will
be
a
follow-up
at
least
to
figure
out
which
path
to
go.
So
thank
you.
H
G
Let
me
throw
out
one
one
question
to
this
wider
audience,
which
is
the
question
came
up
of.
Maybe
we
can
just
create
a
new
verb
for
using
in
the
rbac
system
that
would
be
applied
to
pvcs,
because
verbs
are
just
strings,
as
I
understand
it
and
they're
scoped
by
the
object
that
they're
attached
to
so.
My
question
is
because
pvcs
are
a
core
object.
Doesn't
adding
a
new
verb
that
the
rbc
system
supports
on
pvc's
count
as
a
change
to
a
core
object
that
sig
api
would
want
to
weigh
in
on.
G
H
Just
for
our
you
know,
maybe
they
have
the
same
reason
saying
hey
you
just
for
this
one
use
case
and
same
reason.
They
have
against
this.
What
this
voting
security
standard
we
were
thinking
about
earlier.
D
Not
you're
not
going
maybe
once
the
recovery
from
research
goes
ga
by
that
time
we
might
want
to
do
but
like
for
part
resizing,
there
was
similar
like
adding
a
verb
for
the
resize
operations
rather
than
putting
it
in
status.
So
yeah,
I'm
not
not
sure
yet,
but
there
were
some
discussions
around
and
tim
suggested
that
in
in
his
original
design
in
his.
D
H
H
A
A
All
right
cool.
Thank
you.
Thank
you
all
for
the
update
on
this
one
sounds
like
there's
a
lot
of
follow-up
to
do
here,
but
we're
making
good
traction
and
getting
the
this
designs
at
least
started.
A
Next
up.
We
have
secret
protection.
Prevent
deletion
while
in
use
depends
on
in-use
protection
kept
below,
and
the
protection
cap
is
co-owned
with
sig
api
machinery
masaki
an
update
on
either
of
these
yeah.
A
A
Okay,
thank
you
no
problem.
Thank
you
for
the
update.
Next
is
cohen,
with
segat
user
id
ownership
in
config
maps
and
secrets,
preserve
default
file
mode
bit
set
and
atomic
writer
volumes.
A
Okay
sounds
good.
Thank
you.
Next
item
is
address.
Issues
address
this
issue
with
youtube
created
by
status,
set
not
being
auto
removed.
C
Yeah
it's,
it
should
be
it's
at
the
point
of
like
just
trying
to
grab
people
and
get
them
to
approve
it.
It
was
approved
for
122
but
missed
the
deadline,
and
so
I'm
you
know
pinging
people
on
slack
and
that
kind
of
thing
a
bit
frustrating,
but
you
know
hopefully
it'll
get-
could
get
done.
A
Next
item
is
volume.
Expansion
for
stateful
sets
any
updates
on
that.
D
A
H
Yeah,
we
don't
really
have
an
update.
I
think
we
are
kind
of
stuck.
I
think
we
probably
need
to
try
to
schedule
a
meeting
with
derrick
and
just
want
to
make.
A
Cool
all
right-
that
is
all
that
we
have
for
updates
today,
let's
go
ahead
and
switch
back
to
the
primary
agenda.
So
first
item
that
we
have
here
is
from
manu
gupta
using
open
et2
to
detect
mount
points
faster
comment
here
is
that
all
comments
have
been
addressed.
Looking
for
feedback
to
see,
if
anything
additional
is
needed,
and
should
we
close
this
other
pull
request
in
favor,
open
at2,
or
should
this
be
considered
for
older
kernels.
B
Well,
I
just
wanted
to
point
out.
There
was
a
long
and
legally
discussion
in
the
pr
and
you
definitely
need
something
for
the
older
panels,
but
still
wait.
It
will
help
a
lot
to
merge
that
vr.
G
M
A
Okay,
what
are
the
blockers
here
from
getting
it
merged
right
now,.
M
M
But
like
since
I'm
not
an
expert
at
storage,
I
might
probably
need
someone
else
to
help
with
testing,
as
it
runs
to
the
core
part
of
it,
and
the
other
question
was
like.
Could
we
have
it
for
1.23
if
everything
goes
right.
A
Yeah,
I
think
if
we
can
get
it
into
123,
that
would
be
awesome.
So
I
guess
question
for
the
rest
of
the
sig.
Is
anyone
on
the
call
willing
to
help
manu
with
testing
this.
A
Okay,
thank
you
jay,
so
my
move
will
be
single,
jing
offline
and
see
if
you
can
get
some
intensive
testing
going
here.
This
is
core
core
fundamental
logic
for
pretty
much
every
every
single
volume,
and
so,
if
there's
a
regression
here,
it's
going
to
be
major,
say
it's
worth
doing,
math
testing
to
make
sure
we're
confident
in
it
and
it's
definitely
an
expert
in
this
area.
So.
A
Thank
you.
Next
item
is
from
umasaki.
We
discuss
this
a
little
bit
right
before
asaki
was
to
follow
on
volume,
snapshot
transfer
you're
either.
If
you
want
to
give
a
intro
here,.
K
Yes,
can
I
share
my
screen
sure.
K
K
A
Sure
yeah
hang
on.
Let
me
share
my
screen.
A
Okay,
can
everyone
see
that,
yes,
is
there
a
specific
comment?
You
want
me
to.
K
Scroll
to
yeah
sorry,
but
could
you
go
down
to
almost
the
end.
K
K
This
is
a
proposal
for
transfer
feature
and
we
are
currently
focused
on
only
on
volume
snapshot
and
seeing
her
advised
us
to
use
different
policy.
K
Feature
I'm
thinking
about
arrowing
cloning
to
from
the
volume
snapshot
link
if
it
is
allowed
by
the
reference
policy.
G
K
Example
we
are
trying
to
share,
so
there
is
user,
a
and
user
b
and
user
a
is
in
source
namespace,
and
the
user
b
is
in
target
namespace
and
user.
A
create
difference
policy
to
allow
user
b
to
reference
his
volume
snapshot,
who,
in
his
name
space
to
the
target
name
space.
G
G
Yeah,
I'm
sorry
sorry
you're
right,
so
some
csi
sidecar
will
do
the
heavy
lifting
and
then
the
so
so
the
way
you
get
agreement
from
both
users
is
that
the
source
namespace
has
to
create
this
reference
policy
that
says
it's:
okay
and
then
the
destination
namespace
has
to
create
the
the
link
object,
which
says
I
want
it,
and
so
you
really
do
need
like
a
handshake
from
these
two
objects
before
you
can
do
anything,
so
it
feels
pretty
good.
G
H
My
next
question
is:
what
about
the
secrets
is
that
handled
here
or
the
secret?
It
also
has
to
go.
Go
through
this
transfer.
First.
K
K
The
snapshot
so.
K
K
Yes,
but
it
is
all
behind
the
controller
that
actually
copies
it,
so
the
user
of
the
snapshot
link
doesn't
see
anything
related
to
the
secret.
G
G
F
H
I
don't
know
when,
actually,
when
we
create
a
volume
from
snapshot,
we
actually
for
the
snapshot
itself.
We
don't
use
the
secret,
it's
the
volume
side
that
that
may
have
a
secret
use
for
snapshots
only
when
you
create
or
delete
snapchat
or
list
snapshots.
There
are
secrets.
So
if
it's
create
world
in
france
now
shall
then,
whenever.
G
G
K
Actually,
volume
snapshot
link
only
allows
the
user
to
clone
from
it
not
create
or
delete.
So
in
that
sense,
I
don't
think
we
need
to
allow
access
to
the
volumes
and
absolute
secret.
G
Yeah,
I
guess,
but
what
I
want
to
say.
Is
I
really
like
this
design
for
the
the
user
interface
for
it?
I
think
it's
better
than
any
other
one
that
I've
seen,
and
so
you
know
congratulations
on
having
a
better
design,
but
we
needed
to
fully
go
through
an
example
of
like
with
a
csi
driver
that
uses
secrets.
G
You
know,
and
just
just
answer
those
questions,
because
because
I
think
that
the
the
challenge
for
most
of
us
is
because
most
csi
drivers
don't
use
the
secrets
feature
like
none
of
us
know
we're
not
familiar
with
it,
but
it's
because
some
csi
drivers
can
require
it.
This
feature
might
break
with
those
csi
drivers.
K
Yeah,
I
understand
that
we
need
to.
G
K
A
Your
discussion
and
yeah
thank
you
for
for
the
proposal.
It's
nice
to
kind
of
get
it
improved
overall
design
sounds
like
there's
still
one
sticky
area
around
secret,
so
we
need
to
think
through
that,
and
maybe
the
answer
is
that
you
just
don't
handle
secrets
and
you
don't
get
this
behavior
when
you
have
secrets,
but
regardless,
I
think
their
question
stands.
We
need
to
walk
through
that
and
make
a
decision
on
what
to
do.
H
So
I
think
there
is
still
one
use
case-
that's
not
covered
by
this,
because
there
are
use
cases
that
you
really
need
to
have
a
transfer
of
the
snapshot
itself
so
basically
be
able
to
have
a
volume
snapshot
created
in
a
different
name
space
pointing
to
the
same
snapshot-
that's
not
covered
by
this.
But
I
think
this
is
of
course
solving
some
use
cases.
H
H
G
G
G
H
I
think
there
are
cases,
actually
they
want
to
share
this
one
there's.
Actually
someone
write
a
blog
about
it
in
kubernetes
on
the
website.
Sometimes
they
have
different
teams.
They
want
to
share
the
same
data,
they
actually
have
like
a
gold
copy,
and
then
they
have
some
other
copies
of
snapshots,
pointing
to
that
same
data.
G
H
Is
not,
this
is
a,
but
this
one
actually
does
not
really
create
another
another
snapshot.
I
think
the
proposal
is
actually
creating
another
pvc.
H
F
H
H
H
A
Okay,
so
it
sounds
like
we
have
good
progress
here,
but
there's
a
couple
of
areas
that
we
still
need
to
look
into
and
we'll
continue
to
do
so.
So
thank
you.
Everyone
we're
at
the
top
of
the
hour,
so
we're
going
to
go
ahead
and
end
the
meeting
now.
Thank
you
very
much
for
your
time.