►
From YouTube: Kubernetes SIG Storage Meeting 2022-12-01
Description
Kubernetes Storage Special-Interest-Group (SIG) Meeting - 01 December 2022
Meeting Notes/Agenda: https://docs.google.com/document/d/1-8KEG8AjAgKznS9NFm3qWqkGyCHmvU6HVl0sk5hwoAE/edit#heading=h.a4oc0xnjnezo
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
December
1
2022.
This
is
the
meeting
of
the
kubernetes
storage
special
interest
group.
As
a
reminder,
this
meeting
is
public
recorded
and
posted
on
YouTube.
A
So
we're
gonna
go
through
the
agenda
today.
First
up
we're
gonna,
go
through
the
126
planning
spreadsheet
and
get
a
final
status
update
on
the
items
that
are
going
into
the
126
release.
The
upcoming
deadline
is
the
release
date,
which
is
December
6th
next
week,
so
coming
up
very
quickly,
so
we
just
want
to
get
a
end
of
cycle
update
for
all
the
features
that
folks
have
been
working
on
then
we'll
go
over
PRS
to
discuss.
A
If
you
have
any
PRS
or
designs
that
need
review,
please
feel
free
to
add
them
or,
if
there's
anything
else
that
you
want
to
add,
feel
free
to
add
to
the
agenda
and
we'll
get
to
that
afterwards.
So
with
that
we'll
go
ahead
and
switch
over
to
the
planning,
spreadsheet
and
start
getting
status
updates.
A
So
first
up
we
have
delegate
FS
group
2
CSI
driver
instead
of
cubelet.
This
includes
CSI
updates.
So
let
me
make
a
new
column
here.
A
Okay,
anyone
have
updates
on
this
is
the
line.
B
B
C
A
B
I
opened
a
PR
that,
like
I'm
working
on
test,
I
open
a
PR
that
refactored
some
of
the
mock
tests
and
then
I
have
a
PR,
not
I
have
I
don't
have
a
period,
but
I'll
have
a
peers
soon
in
order
to
that,
actually,
as
the
the
new
tests
I'm
working
on
them.
So
as
soon
as
the
Master
branch
merge
opens,
we
should
have
this.
These
two
things
done
like
22.,.
D
A
It
are
we
still
targeting
126
for
this
release
or
for
this
task.
A
Staying
in
I
guess
we
could
leave
it
as
is,
and
then
just
mark
it
as
incomplete
and
copy
it
over.
Okay.
A
Okay,
next
up
is
issues
related
to
assuming
volumes
are
Mount
points.
Do
we
have
Jing
looks
like
we
haven't,
had
an
update
on
this
for
a
while?
Does
anyone
know
what
the
status
of
this
one
is.
B
I
think
there's
still
a
bunch
of
issues
related
to
mount
points.
I
think
I,
I
Michelle
knows
some
and
then
there's
a
issues
during
Reconstruction
when
we
assume
why,
when
we
assume
the
volumes
are
Mount
points
and
we
don't
properly
reconstruct,
sometimes
so
yeah,
we
still
have
to
fix
those
things.
As
far
as
I
know,
I
think.
If
we
look
at
the
last
like
update,
there
will
be
some
updates
on
this.
This
one.
A
We'll
keep
that
open
seems
like
this
is
probably
going
to
get
punted
to
the
next
release.
Next,
two
items
already
complete
so
I'll
skip
over
those.
Then
we
have
volume
group
API
anything
new
on.
E
This
one,
no
so
I
I
pinned
the
gems,
have
not
got
a
reply,
so
I
probably
should
have
sent
him
an
email
and
see.
If
can
I
I
just
need
to
set
a
meeting
with
him
to
go
over
his
concerns.
I
have
already
addressed
his
comments
on
that
PR.
E
A
A
Anyone
have
an
update
on
this.
One
looks
like
last
update.
Api
code
was
merged,
docs
and
review
PR
and
external
provisioner,
and
review
and
volume
populator
pending.
F
Sorry
this
is
the
cross
namespace
stuff,
yeah.
A
Cool,
do
you
know
what
the
latest
status
is
and
has
any
of
this
changed.
F
F
E
E
So
yeah
run
actually
submitted
a
working
progress
PR
for
the
ete
test.
A
Okay:
next
up,
we
have
new
rwo
access
mode,
add
n10
tests.
A
So
last
status
was
testress,
merge,
scheduler
improvements,
question
mark
anyone
know
the
latest
on
this
one.
C
D
F
A
F
A
G
D
A
Then
we
have
CSI
proxy
for
Windows
transition
to
privileged
containers.
It's
the
last
status
PR's
out
anything
new
here.
H
I,
don't
know
if
Mauricio
or
Alex
is
here,
but
there's
been
a
continuing
work
in
figuring
out
how
to
launch
it
correctly
because,
there's
you
know
incompatibilities
and
such.
A
Cool
thanks,
Matt.
H
F
Yeah
like
so
this,
this
item
basically
came
up
I
think
because
someone
was
doing
performance,
experiments
comparing
Linux
to
Windows
and
saw
that
Windows
had
lower
performance
and.
D
F
H
Gotcha:
okay;
okay,
thanks
for
the
contact.
A
So
we'll
we'll
keep
this
open
and
then
see
if
someone
can
confirm
if
this
is
still
an
issue
or
not
node
expansion
secrets
anybody
here
from
Humble
on
this
one.
E
I
think
this
one
is
Michelle.
This
is
still
someone
some
new
contributor
saying
yeah.
A
A
And
feel
free
to
reach
out
offline,
we
can
figure
it
out.
Next
up
is
SC
Linux
relabeling,
using
Mount
options,
end-to-end
tests,
volume,
reconstruction.
C
Yeah
everything
is
merged.
I
still
have
some
tests
of
working
progress.
We
can
do
it
in
the
next
High
Court
everything
I
wanted
is
merged.
You
got
it.
A
Okay,
that
sounds
good.
Next
skip
this
one:
vsphere
Windows,
CSI,
migration,
move
to
ga
codes,
merge,
dot,
PR
and
review
anything
new
here.
E
We
also
added
more
notes
in
the
release
notes
we
added
more
like
the
windows
not
GA
yet,
but
we
waited
for
the
next
season.
Driver
release,
which
is
for
1.24,
got.
A
Then
we've
got
Azure
disk
and
Azure
file.
Anyone
have
updates
on
this
one.
So
I
guess
Azure
disk
is
moving
to
127.
Azure
file
code,
merge
doc,
PRN
review
anything
new
here.
D
A
D
E
So
there's
a
talk,
that's
getting
very
close!
I
think
combo
just
replaced
it
today.
So
hopefully
that
will
get
merged
soon.
Cool.
A
It
sounds
good
next
up,
we
have
always
honor
reclaim
policy.
C
C
Yes,
so
the
fixes
and
lip
external
provisioner
has
been
merged,
so
we
need
to
do
a
release
of
the
battery
and
then
merge
it
into
the
external
provisioner.
D
E
A
Matter
it's
external
provisioner
stuff.
Then
you
have
control
volume,
mode,
conversion
between
source
and
Target.
E
Yeah
this
one
so
so
Roanoke
is
working
on
adding
this
E3
test
in
external
psychos
repo.
So
it's
still,
he
has
a
few
PR's
Patrick
is
helping
reviewing
it
so
once
those
are
in
then
I
think
Deepak
can
add
a
test
using
the
same
basic
same
same
method.
A
Got
it
cool?
Thank
you
Shane.
Next
we
have
items
that
are
co-owned
with
other
cigs,
starting
with
Sig
node
node
non-graceful,
node
shutdown.
E
Yeah
this
one
basic
gesture
has
the
the
blog
pending.
It's
been
it's
just
stream
reviews.
It's
already
submitted.
E
Can
you
help
me
review
the
blog
since
you
reviewed
the
the
other
version?
Okay,
yeah
sure
I
will
send
you
a
link.
Okay,.
E
A
Yeah,
that's
awesome.
Next,
up
enable
username
space
in
cubelet,
so
uids
get
shifted,
rootless
mode.
B
Actually
this
is,
there
are
two
different
stories.
This
is
a
bit
of
confusion.
We
should
track
them
differently,
one
is
to
run
route,
run
cubelet
without
root
and
that
that
is
a
different
cap
and
there's
a
second
cap
is
the
one
that
we
were
talking
in
the
last
meeting,
which
is
like
enable
username
space
inside
Cube
and
yeah.
B
So
the
last
time
we
had
folks
from
signal
presenting
the
enabling
username
space
in
cubelet
and
I
tested
I
tried
to
use
that
feature
and
if
it's
it's
release,
Alpha
in
the
last
release,
but
I
found
that
it
doesn't
actually
work
with
any
configuration.
Any
combination
of
container
runtime
and
analytics
kernel
right
now,
either
there
are
bugs
in
container
runtime
or
bugs
in
Linux
kernel.
B
So
it's
just
like
to
provide
more
feedback
to
the
Sig
note
folks,
who
came
to
our
meeting
last
last
meeting
I
think
we
need
a
little
bit
more
time
like
cryo
had
some
bugs
which
prevents
the
the
username
space
to
work
properly
when
integrated
with
kubernetes.
B
Let
me
find
the
use
I
think
the
the
cap
that
we
have
linked
is
for
Cube,
rootless,
Cube,
rootless,
cubelet,
yeah,
okay,.
B
A
A
Okay,
so
I
think
we
should
have
two
different
tasks
here.
We
can
track
them
separately.
Moving
forward,
I
think
you
added
a
comment
here:
yeah.
Let
me
go
ahead
and
promote
that.
E
Oh
before
we
move
to
the
next
one
question
for
Ben:
do
you
have
any
plan
to
continue
to
work
on
your
PR
in
that
CSS
back
Force
detach
thing.
C
I
Would
like
to
bring
that
back?
Yes,
if
there's
openness
to
actually
moving
forward
like
I,
can
outline
the
design
in
more
detail.
If
that
would
help
people
understand
it
better.
I
mean
I
still
think
that
the
the
pr
as
is,
is
okay
to
merge
and
it's
what's
what's
missing,
is
sort
of
an
understanding
of
the
rest
of
the
the
rest
of
the
design
so
yeah.
If
people
want
to
see
that
merge,
I'll
work
on
it.
E
I
I
can
possibly
make
time
if,
if
the
outcome
is
likely
to
be
successful,
I.
E
Think
I
think
that
chance
he
still
has
those
concerns
right
so
probably
need
to
talk
to
him.
Yeah.
E
A
H
D
E
E
To
take
a
loose
notice,
someone
who
actually
has
a
PR
there,
a
cap
there
right
come
on,
you
want
to
chairman.
B
It
was
there's
a
there's,
no
enhancement
it
was.
There
was
a
some
POC
I'm
I'll
have
to
take
a
look.
E
G
A
Okay,
let's
follow
up
on
that
offline
sounds
like
possibly.
There
was
a
cap
or
POC
by
somebody
unclear
if
we
have
a
solid
owner
for
this
or
not.
B
D
A
Thank
you
both
and
last
we
have
better
default.
Storage
class
code
was
merged
doc
PR's
in
review
anything
new
here.
A
Cool
thank
you
John
and
with
that
we'll
switch
back
to
our
agenda
and
first
up,
we
have
a
PR
from
deep
I
think
this
is
the
one
that
you
mentioned
p
deep,
around
runtime
assisted
mounts,
so
that
needs
to
be
reviewed.
Please
help.
Take
a
review.
Take
a
look
at
this
anything
else.
You
want
to
add
deep.
G
A
Yeah
we
got
the
CSI
spec
meeting
coming
up
next
Wednesday,
so
you
can
discuss
this
there
as
well
yep
thanks.
Next,
we
have
from
Beyond
Dynamic
provisioning
configuration
in
PVC
issue
or
pull
pull
request
808
here
in
the
external
provisioner.
C
So
it
comes
every
now
and
then
that
somebody
comes
and
wants
to
configure
the
dynamic
for
the
provisioning
parameters,
not
in
storage
class,
but
in
PVCs.
C
So
far
we
have
been
resisting
because
that
makes
the
PVCs
not
portable,
but
it
doesn't
stop
the
people
from
requiring
death.
So
I
don't
know,
has
anything
changed
in
this
area
or
is
it
still
the
same
I
linked
the
previous
discussion
on
Disco
two
years
ago
and
also
like
five
years
of
old
tip
that
was
almost
merged,
but
then,
in
the
end
nobody
really
wanted
to
implement
it.
E
C
That
what
you
mean
yes,
I,
think
if
I
remember
correctly,
that's
the
last
status
it
introduced.
What
was
the
conflict
map
and
user's
name
space
and
then
a
PVC
would
start
a
route
that
points
to
a
storage
class
and
to
the
conflict
map
so
and
somehow
managing
the
parameters
from
both.
But
that's
it's
complicated
from
Atomic
cluster
admin
perspective,
because
some
storage
class
parameters
could
be
dangerous
and
they
need
to.
The
question
means
they.
C
F
I
guess
how
much
so
we're
having
discussions
on
qos
right
now
right,
but
I
think
that
some
folks
are
wanting
to
Leverage
things
to
do
more
than
qos
like
how
much
do
we
think
that
discussion
can
help
with
this.
C
H
Kind
of
qos
discussion
is
sort
of
talking
about
Plumbing
PVC
parameters
down
to
the
driver,
which
is
actually
I
yeah,
which
which
has
to
be
used
a
provisioning
time
for
TOS.
So
it
would
have
would
help
help
here
too,.
D
C
I,
don't
throw
the
quality
of
service
discussion
closely?
Is
it
still
like
some
predefined
I,
don't
know
classes
like
fast
and
soap.
H
Yeah
I
I,
it's
I
mean
it's,
it's
it
it.
We
haven't,
decided
anything.
So
there's
still,
you
know
plenty
of
opportunity
to
change,
but
the
touch
of
the
direction
is
it's
going.
Is
they'll
we'll
be
driver,
specific
parameters
like
number
of
iops,
so
yeah
instead
of
like
classes,
it
should
be
kind
of
General
parameters.
H
Right
yeah,
just
like
right
now,
the
motivating
cujs
are,
is
quality
of
service
I
mean
I,
think
there
are
a
couple
options.
There's
also
some
discussion
about
using
the
new
new
resource
claim
cap,
which
might
not
be
General
and
help
this
case.
F
I
guess
also
in
terms
of
the
question
of
portability
and
how
important
that
is
I
would
say
it
is
still
very
important,
especially
to
service
providers
who
are
writing
applications
to
work
across
multiple
environments.
F
They
don't
want
to
have
to
have
special
logic
to
deal
with
every
different
platform
or
infrastructure
that
they're
using
I
was
actually
watching
or
not
watching.
I
was
I
was
looking
at
some
slides
of
talk
and
I.
Think
it
was
someone
from
it
was
like
a
maintainer
of
the
kubernetes
Cassandra
operator,
I
believe,
and
they
were
saying
that
you
know
portability
is
key
to
them
and
you
know
they
actually
specifically
called
out
like
PVC
being
portable,
as
kind
of
you
know
really
kind
of
key
to
to
letting
them.
F
H
Yes,
if
you
scroll
down
that
link
there,
please
note
that
I
corrected
yeah.
Okay,
let
me
link
in
these
suggestion,
but
yeah
I'm,
trying
to
consolidate
all
discussion
to
this
talk
and
we're
also
planning
on
meeting
Tuesday
next.
H
Yeah,
yes,
cool
great
okay,
thanks.
A
Yeah
and
I'll
I'll
say
big
plus
one
to
Michelle.
I
know
portability
is
hard.
I
know
it
makes
it
difficult
to
not
be
able
to
add
arbitrary
things,
but,
like
Michelle
said
very
eloquently,
you
know
it
is
what
makes
kubernetes
attractive.
You
write
your
application
deployment
once
and
kind
of,
don't
have
to
worry
about
it.
So
let's
try
and
hold
up
that
principle.
If
we
can
okay
sure
with
that,
let's
move
on
to
the
next
item:
copying
PVC
annotations.
C
So
again,
people
come
and
ask
for
a
way
how
to
transfer
the
PVC
labels
or
something
doesn't
need
to
be
usable
into
a
label
in
the
cloud.
So
they
can.
They
can
see
what
is
what
and
in
similar
fashion
transfer
labels
from
BBC
to
PVS.
So
they
can
filter,
PVS,
better
and
again,
I'm,
resisting
because,
like
peace,
labors
and
peace
be
available
as
they
are
different,
they
are,
they
have
different
purposes,
but
I
think
there
should
be
some
way
how
to
transfer
some
labels.
Some
selected
name
labels.
F
Today
we
are
ready
to
pass
in
like
the
namespace,
the
namespace
and
the
name
of
the
PVC.
C
C
C
B
B
But
like
it
fetches
the
PVC
and
it
syncs
it
to
the
to
the
vcenter
back
end.
So
I
don't
know
like
in
that
case.
Obviously
it's
a
optional,
not
so
optional
component
of
the
driver,
but
I
don't
want
to
derail
the
discussion
too
much,
but
is
that
something
that
that
could
work
for
other
driver
authors.
A
A
C
A
Hey
cool
thanks
John.
Finally,
we've
got
volume
quests
from
Matt
carry
Matt.
H
Yeah
so
I
actually
figured
these
previous
two
topics
were
going
to
fill
up
the
time,
so
it
would
be
better
to
make
a
meeting
for
next
week
to
discuss
the
cause.
Stuff
I
mean
I.
Think
as
mentioned,
it's
possible
that
this
is
actually
just
the
general
question.
We've
just
discussed
about
passing
on
my
driver,
specific
parameters
or
around.
H
H
H
A
H
A
Okay,
anyone
else
have
any
topics
they
want
to
bring
up
before
we
go.
Thank
you.