►
From YouTube: Kubernetes SIG Storage - Bi-Weekly Meeting 2022-03-24
Description
Kubernetes Storage Special-Interest-Group (SIG) Bi-Weekly Meeting - 24 March 2022
Meeting Notes/Agenda: -
Find out more about the Storage SIG here: https://github.com/kubernetes/community/tree/master/sig-storage
Moderator: Xing Yang (VMware)
A
Today
we
will
go
over
our
1.24
planning
spreadsheet
and
then
we
will
just
to
go
back
to
review
this
archive
sub
projects
again
just
to
see
if
we
can
move
forward
with
it
all
right.
So,
let's
start
with
the
spreadsheet.
B
Yeah
I'm
here
I
have
a
peer
out
that
is
being
reviewed
for
what
was
the
for
read,
write,
many
expansion
and
then,
and
then
I
opened
a
pr
for
removing
the
expansion
feature
here.
I
will
link
it
in
chat.
B
It
just
toggles
the
feature
gate
to
ga,
but
because
previously
tim
recommended
us
to
remove
the
feature
gate
checks
from
the
code
if
you're
moving
the
pc
gate,
so
it
became
kind
of
big
because
you
know
like
we
had
to
delete
a
lot
of
code.
So
it's
mostly
deletion
of
code,
but
at
the
same
time
we
we
are
deleting
a
lot
of
code.
So
it
touches
a
lot
of
packages,
so
it
will
require
approval
from
like
more
than
just
pkg
feature
owners
and
yeah.
B
So
that's
it
from
my
site
which,
if
I
can
get
approval
for
the
for
the
like
the
code
deletion
that
I'm
doing
in
this
pr.
We
should
be
good
to
go.
But
jan
was
telling
me
that
the
api
review,
pier
is
full,
so
I
don't
know
but
yeah
so.
B
Yeah,
that's
what
right
man,
it's
a
new,
not
a
new
api
change.
It's
just
I'm
deleting
the
code,
but
still.
A
Okay,
what
about
the
documents
like
place
for
the
pr
for
the
dog
and
for
the
blog?
Do
you
have
those
as
well.
B
Blog
was
pushed
out.
I
have,
I
will
open
the
dark
pier
while
I
was
working
on
it
not
yet.
A
A
A
B
Yeah,
I
think
so
well
to
reset
back
the
the
size
all
the
way
to
original
side.
It
requires
like
not
big
change,
but
still
requires
a
little
bit
change
and
I
haven't
had
time
to
think
through
the
the
the
updated
design
yet
so
I
need
to
I'll
try
it
in
125
to
move
it
to
beta.
I
should
be
okay
to
keep
in
alpha
for
one
release.
I
guess.
D
B
And
we
did
add
actually
one
more
state
like
from
this
thing,
like
we
have
a
research
status,
we
have
allocated
resources,
so
we
have
quite
a
few
states,
so
it's
it
could
be
possible.
It's
just
that.
I
haven't
thought
through
the
design.
I
haven't
thought
through
the
implications
of
the
whole
thing
yet.
C
B
E
B
It's
a
nice
to
have
like,
I
think,
like
I
don't
know,
michelle
is
on
the
call,
but
to
recover
too
but
yeah
I
don't
know
yet.
I
I
can't
answer
but
okay,
if
we
need
we,
we
do
have
us
the
allocated
resources
and
pvc
status
does
track
towards
like
the
resize
towards
which
the
resizer
is
working.
So
do
we
do
have
a
field
and
to
track
that
so
it's
possible
to
like
make
it
work
with
existing
fields
that
are
there
just
that.
I
need
to
think
through
every
every
aspect
of
it.
F
I'm
still
working
on
that,
so
I
was
trying
to
document
like
where
the
mount
point
tracking
were
used
and
also
trying
to
go
through
a
list
of
pr
issues
related
to
this.
It's
quite
long
list.
F
So
I
think,
by
doing
this
I
can
better
understand
the
other
cases
and
avoid
you
know
issues
we
experienced
before.
If
we
make
any
changes
here
so
yeah
it's
to
make
any
change,
we
have
to
be
careful
yeah
around
the
mount
point.
A
F
I
think
manu
right.
He
opened
a
pr
trying
to
update
the.
F
Mob
sys
version
first,
okay
and
the
after
that
is
merged.
We
can
make
the
changes
to
to
use
the
faster
like
a
way
of
tracking
mount.
That's
pr
need
approval,
so
I
don't
have
permission
to
approve
it.
A
A
Okay,
what
is
the
number
again.
A
F
To
update
the
movie
says
mount
package
version.
Okay,
I
see
I
can
send
the
link,
I
think
the
owner
right
now
I
see
it's
like
a
clinton
team,
then
dong.
Oh,
it's
a
node,
okay
yeah,
mostly
like
a
knowledge
team.
A
A
A
A
So
there
is
a
pr
already,
but
I
think
we'll
have
to
wait
for
the
test
to
be
stable
and
then
promote
it
to
conformance
and
also
he
also
has
all
the
like
dark
peers
open.
I
think.
A
Did
you
go
to
the
enhancement
issue
and.
A
A
Next
one
is
a
volume
group
api,
so
I
just
updated
the
cap
because.
A
A
Okay,
yeah,
he
didn't
add
a
status,
I'm
not
sure.
Okay,
so
we'll
just
say,
arizona
update.
H
A
Thank
you,
and
next
one
is
warning
health
additional
metrics.
So
this
one
so
the
pr
got
merged
but
then
reverting
there's
a
bug.
The
bug
got
fixed,
so
the
pr
is
submitted
again
and
being
reviewed
so
trying
to
get
that
merged
by
the
deadline
and
the
next
one
is.
Since
I
want
to
health,
pragmatic
response
so
that
I
need
to
pin
nick
again
because
he
said
he
will
start
the
design
in
a
few
weeks.
That's
the
last
update,
so
I
will
check
and
see
if
they
started
or
not.
C
So
shin
you
know
most
of
the
status
but
for
everyone
else's
benefit
all
the
prs
for
this
are
out
and
ready
for
review.
Most
of
them
are
merged.
The
ones
that
remain
are,
I
think,
the
ede
test,
the
metrics
patch
for
the
library
and
the
entry
change
to
move
the
feature
gate
to
beta.
C
Once
the
metrics
patch
merges
to
the
library
I
can
do
releases
and
that
should
unblock
the
ede
test
but
yeah.
So
I
guess
the
main
priority
is
getting
the
metrics
patch
in
and
getting
the
entry
feature.
Gate
patch-
and
I
heard
earlier
in
the
meeting
someone
mentioned-
there's
trouble
getting
api
reviewers
to
review
things
like,
obviously
the
feature
gate
alpha
beta
will
require
an
api
reviewer,
but
it's
like
a
one-line
change
is
that
is
there
gonna
be
a
problem?
Getting
that
review
done.
C
A
Nice,
this
one
is
so
small,
it's
probably
maybe
you
can
try
since
the
team
reviews,
some
people
you
can.
This
is
just
like
one
line
right
so.
B
B
A
B
A
C
A
A
A
And
the
and
then
expo,
and
also
you
have
the
other
dock,
the
you
know
all
the
placeholder
pr's
for
dog
and
blog.
You
also
have
all
of
those
submitted
right.
A
Next
one
is
the
ultra
storage
api,
cozy,
okay,
so
sid
is
probably
not
here.
He's
traveling,
I
think
it's.
I
think
I
actually
did
not
check
michelle.
Did
you
check
the
the
cup
recently?
Is
it
I.
E
Haven't
had
a
chance
to
check
in
the
last
week,
but
we
have
been
going
back
and
forth
a
little
bit
and
I
think
mainly
just
I
had
a
couple
outstanding
questions.
I
haven't
checked.
If
he's
responded
to
them
yet.
A
J
Yeah,
this
is
still
in
progress,
basically
going
through
how
we
can
plumb
the
story
apis
between
the
cri
runtime
and
something
like
kara.
J
J
Deep,
so
this
is
pretty
much
in
progress,
no
major
updates,
but
I
think,
like
we
pretty
much,
have
a
direction
on
where
we
want
to
take
it
and
that
that's
basically
like
we
are
going
through
some
of
the
finer
details
of
exactly
how
the
the
the
layering
will
happen,
model
and
the
new
model.
J
But
overall
we
want
to
get
out
of
the
the
grpc
based
service
that
used
to
have
to
kind
of
like
a
direct
api
invocation
between
the
windows
based
csi,
note,
client
and
the
implementing
api.
A
Would
you
be
still
waiting
for
api
reviewer,
which
one
is
this:
this
is
the
node
expand
node,
it's
my
secret.
A
So
and
also,
I
think,
the
believe,
the
dog
pian,
those
things
are
already
out.
A
And
next
one
secure
linux,
labeling,
so
yeah
you
want
to
give
an
update.
D
Yeah,
I
have
done
a
bunch
of
work,
but
still
autofocus
remaining
including
api
review,
so
this
is
going
to
slip
from
124
and
I
will
try
again
in
125..
A
A
E
I
think
drawing
has
a
bug
fix
out,
I
think
it
merged.
I
haven't
doubled,
I
haven't
double
checked,
but
that
was
the
only
outstanding
csi
migration
bug,
video
about.
E
It
was
when
you
have
a
when
you
have
in-line
volume,
pods
plus
csi
native
pods.
A
This
is
migration
vsphere,
so
this
one
we
have
some
discussions
on
slack,
so
we
actually
need
to
deprecate
67u3
because
it
is
actually
well
the
end
of
support
in
october.
That
was
announced
like
two
years
ago.
So
we
are
missing
a
deprecation
message
in
kubernetes,
so
deviant
has
submedia
pr
with
the
you
know,
checking
the
versions
with
the
deprecation
message.
So
please
take
a
look,
and
also
that
means
we
cannot
turn
this
on
by
default.
E124.
E
A
A
That's
a
little
weird
right,
but
but
if
we
I
mean
in
the
future,
why
is
this
move
to
csi?
Then
you
know,
then
they
would
just
follow
that
documentation
or
say
what
is
supported.
Why
did
not
support
it
even
like
in
our
system,
our
documentation?
A
We,
I
don't
think
we
say
something
like
this.
Normally
it's
a
following
the
you
know,
because
vmware
has
its
own.
You
know
life
cycle
for
those
products
right
and
then
there
are
blocks
talking
about
those.
So
I
don't
think
we
normally
just
say
vsphere
application
in
our
vcu
css
driver
dock.
A
A
A
One
is
a
specific
one
right,
so
we
initially
have
the
message
we
crafted
messy
around
that,
but.
B
B
I
mean,
from
our
perspective
the
main
reason
of
deprecating
it.
Removing
support
for
67
u3
is
because
it
cannot
support
migration
like
it
doesn't
support
migration
for
all
cases
like
nfs
before
people
are
using
it,
then
it
cannot
support
migration,
although
so
entry
plugins
customers
who
are
using
entry
pvs,
they
will
break
simply
if
they're
running
on
6.7
u3
and
they
might
they.
They
have
migration
level
they'll
break.
So
we
need
to
I'm
not
too
much
concern
about
when
we
vmware
ul's
the
67
u3,
but
basically
we
cannot
support
migration
on
6
7,
u3.
E
E
D
A
E
A
E
Okay,
maybe
I'm
confusing
things
okay
hold
on,
let's
see
if
I
can
find
okay
I'll.
Let's
come
back
to
this
okay.
A
I'll,
just
I'll
just
say
what
I'll
check.
A
A
A
E
A
A
K
Yeah,
so
the
main
pr
is
actually
merged
the
one
which
actually
turns
on
the
changes
into
peta
yan
actually
suggested
a
bunch
of
venture
and
tests.
I
haven't
actually
gotten
to
start
on
it
yet.
A
I
actually
did
not
know
that
you
will
be
joining
today.
If
you,
I
thought,
you'll
be
out
for
a
few
days,
so
I
have
pinned
somebody
else
to
start
the
working
on
the
et
test,
so
I
can
sync
up
with
you:
there
is
a.
E
A
A
Okay,
yeah
thanks
and
also
I
see
that
you
submitted
a
a
pr
for
the
blog
right,
placeholder.
A
So
nurse
I
know
that
previously
was
saying:
probably
we
don't
need
a
a
dog
because
it's
kind
of
confusing,
but
there
are
some
people
from
the
dog
team
there.
They
say
that
we
should
have
a
doc.
Can
you
also
work
on
that
so
right
now
we
only
have
a
like
after
after
the
alpha
release.
We
only
have
that
in
the
blog,
but
we
don't
really
have
that.
In
the
you
know,
the
actual
document
area
right.
A
No
okay,
so
I
think
ronald's
been
doing
good
with
this
one.
It's
in
good
progress.
There
is
a
couple
prs
already
merged,
he
submitted
another
pr
and
you
can.
I
need
to
review
it.
So
this
is
on
the
snapshotter
external
snapshot
side.
Once
that's
merged,
we
can
cut
a
rc
build
and
then
he
can
update
external
provision.
Based
on
that,
so
let's
say
we
are.
A
A
Oh
okay,
which
one
are
you
updating
the.
I
I
A
A
A
All
right
next
one
is
a
non-grease
for
no
shutdown,
so
the
pr
is
out
is
being
reviewed.
I
think
we're
getting
close.
A
So
I
see
how
long
do
you
review
that
you're
you're
good
with
this
as
well
ching,
also
reviewed
it.
I
think
we
just
need
to
pin
the
signal
approver
again.
I
think
he
already
reviewed
it.
It's
pretty
much
okay!
So,
let's,
let's
see
if
we
can
get
the
submerged
and
the
the
block
yeah
the
placeholder.
A
And
one
expansion
for
stiffer
set
so
charlene
said
she
has
a
conflict
this
morning,
but
she
gave
me
update.
She
did
update
the
cap,
address
the
some
comments,
but
she
still
have
quite
a
few
questions,
so
I
think
she's
going
she's
going
to
be
in
harmony
just
to
get
those
questions
sorted
out
before
updating
it
again.
A
L
A
And
and
then
the
so
the
merge
that
code
merge.
That
line
is
coming.
It
is
it's
29th
right,
it's
less
than
a
week.
Actually
it's
a
code,
merge
deadline
and
okay,
so
about
this
archiving
some
sub
projects.
We
talked
about
this
in
last
meeting
and
I
also
have
not
heard
anything
else
about
those
projects
so
other
than
this.
You
know
cluster.
The
two
cluster
projects
humble
said:
we
should
still
keep
them
so
we'll
keep
them.
A
We
will
proceed
with
the
archiving
for
those
remaining
projects
that
we
mentioned
in
the
email
yeah.
So
if
you
have
any
concerns,
please
speak
out.
Otherwise
we're
going
to
start
the
archiving
very
soon.
J
Want
to
talk
about
this
yeah
hey,
so
this
is
a
recent
issue
that
we
ran
into
ebo
who's
in
the
column.
I,
basically,
what
we
found
is
that,
when
using
local
pvs
after
a
node
restarts
the
expectation
is,
there
is
typically
a
path
backing
that
local
pv
in
the
host
right,
and
I
think
it's
up
to
the
overall.
You
know
cluster
admin
to
figure
out
like
how
to
make
sure
their
local
disks
in
the
machine
is
getting
mounted
into
the
path
associated
with
the
local
pv.
J
So
what
we
found
is
that
they're,
currently
in
the
entry
co
at
least
like
no
specific
checks
around
whether
that
mount
has
happened
or
not.
One
thing
I
was
curious
about
is:
would
it
make
sense
to
kind
of
add
a
protection
against
that,
and
are
we
even
kind
of
like
adding
stuff
to
the
local
volume
plug-in,
or
is
that
kind
of
considered.
J
It
seemed
like
in
the
code
we
do
not
like.
We
look
for
the
presence
of
the
path
that's
specified
in
the
local
volume
spec
as
the
source,
and
if
that
path
is
not
present,
we
we
report
an
error
right
and
that
gets
reported
up
through
setup
at
as
as
a
failure,
and
that
gets
retried.
But
if
that
path
exists
but
is
not
backed
by
a
mount
point,
then
what's
what
ends
up
happening
is
the
pod
that
is
mounting
that
local
pv
gets
access
to
the
underlying
host
file
system,
which
is
almost
a
security
issue.
E
J
E
B
So
we
added
a
change
in
cubelet
that
it
it
will
try
to
mount
previously.
We
did
have
this
ability
to
mount
the
volume
in
the
cubelet,
but
we
added
a
change
in
cubelet
sometime
like
later
on
after
the
local
feature
was
released,
that
it
will
try
to
mount
the
volume
if
you
are
trying
to
use
it
as
a
file
system
volume
mode,
and
you
will
try
to
mount
the
volume
and
if
the
path
that
you
have
specified
the
device
path
in
the
pv
source,
it
doesn't
it's
not
that
the
one
you
want
then.
E
B
Can
see
this
case,
what
deepak
is
reporting
will
happen
where
it
will
try
to
mount
something
from
host
path.
Basically,
because.
J
Exactly
yeah,
so
basically,
your
bot
ends
up
getting
access
to
the
node
like
root
os
almost
and
it's
kind
of
exactly
the
host
path.
Type
scenario
where,
where
a
customer
might
end
up,
writing
like
lots
of
data
and
like
us
other
parts
out
of
the
node.
M
Yeah
yeah
in.
J
In
the
in
the
in
the
local
volume
api
object
that
we
have.
E
So
I
think
the
I
think
hamad
mentioned,
so
we
have
a
mode
where,
if
the
path
is
a
device
path
instead
of
a
file
like
a
directory,
then
it
will
try
to
mount
it.
For
you.
B
N
I
just
want
to
be
clear,
so
this
is
if
the
lpv
gets
bored,
local
volume
gets
provisioned
as
a
like
as
a
block
like
it's
a
block
but
not
mounted,
and
the
cubelet
will
mount
that
for
you
but
like
if
the
source
is
already
like
formatted
with
the
file
system.
But
it's
not
mounted.
There
is
no
such
check
there
for
cubelet
to
actually
mount
for
you
right
or
am
I
not
understanding
correct.
N
So
I
think
in
that,
in
in
our
case
that
we
discover
is
the
path,
is
a
file
system,
but
it
wasn't
mounted
so
then
what
happened
is
basically
like
hublet
did
not
do
that
extra
check,
and
then
you
ended
up
essentially
giving
out
the
file
the
root
file
system
passed
to
the
pod.
J
B
Right:
it's
it's
there's
a
there's,
a
valid
use
case,
though,
to
mount
a
host
to
use
a
host
directory
as
a
local
volume
like
there
are
people
using
it.
Yes,.
J
B
Okay,
I
would
if,
if
I
have
to
say
like
that,
I
would
just
for
all
these
use
cases.
Why
didn't
you
use
the
device
directly.
J
Mainly
because
we
do
some
more
stuff
on
the
device
like
configure
some
encryption
and
like
mount
the
file
system
in
a
specific
way
kind
of
things,.
N
There
are
some
like
large
disks
and
then
like
doing
a
make.
File
system
could
take
a
while
so
like
we
do
that
ahead
of
time
to
not
like
stall
the
like
cubelet.
So
we
we
we
don't
like
cubelet
took
a
while
just
to
do
the.
B
J
Right,
like
we
specifically
configured
like
a
device,
mapper
thing,
the
other
point
is
around
this-
is
you
know,
at
least
in
our
internal
stuff,
like
we
both
have
a
like
using
the
entry
and
the
local
volume.
So
basically
I
mean
like
a
local
volume,
csi
plugin
as
well,
but
the
csi
path
is
easy
right
and
in
in
the
note
publish
for
the
csi,
we
can
make
our
own
checks
and
easily
reject
it.
J
So
I
I
can
put
it
in
a
issue,
but
are
there
any
concerns
around
this
having
a
potential
security
aspect
and
should
not
be
put
in
an
open
issue.
E
A
I
thought
someone
raised
an
issue
on
slack
and
seeing
racist
issue
in
today's
meeting,
but
no,
I
don't
know
if
this
wasn't
joined
this
call.
A
Okay,
I
think
it's
there's
some
requests
to
add
the
secrets
to
get
capacity.
Csi
call
well
anyway,
I
I
don't
see
the
person
who
raised
this
meeting
so.
B
I
think
there
has
been
a
need
to
like.
There
has
been
a
pr
or
request
to
report.
If
a
metal
volume
used
size
wire
matrix
for
a
while
and
we
have
not,
we
don't.
I
think
we
don't
report
if
a
mirror
volume,
how
much
space
that
that
used
by
a
matrix.
So
I
don't
know.
A
That's
a
that's
a
for
a
different
issue.
This
one
is
more
like
I
just
like
the
note
explains
secrets,
it's
more
like
adding
adding
secrets
in
the
get
capacity
csi
course
more
like
that.
Oh.
A
Just
but
I
don't
see
the
the
person
here,
it
was
asking
this
question.
I
was
like.
A
All
right,
so
are
there
any
other
issues
that
anyone
want
to
talk
here?