►
From YouTube: Kubernetes SIG Storage - Bi-Weekly Meeting 2022-02-24
Description
Kubernetes Storage Special-Interest-Group (SIG) Bi-Weekly Meeting - 24 February 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
Hello,
everyone
today
is
february
24th
2022.
This
is
the
kubernetes
storage
sig
meeting.
So
today
we
will
go
through
the
1.24
planning,
spreadsheet
and
and
then
I
think
we
do
have
an
issue
here.
We
will
talk
about
it
at
the
end
and
also
there
is
this.
Opt-In
1024
feature
block
tracking
spreadsheet.
A
If
you
are
working
on
a
feature,
you
want
to
write
a
blog
and
that's
not
added
already,
please
ping.
One
of
the
six
load
leads
to
get
added
there,
and
the
next
deadline
is
so
okay,
so
march
24th.
This
is
the
feature
block
opt-in
phrase
and
then
march
30th.
That's
the
code.
Phrase
deadline.
A
B
I'm
still
working
on
it.
I
have
made
a
lot
of
code
changes,
but
I'm
basically
I
had
this
problem
with
read:
write
menu
that
I'm
trying
to
fix
and
working
on
it.
Basically,
I
should
have
appear
by
hopefully
like
yeah,
maybe
tomorrow
or
next
week's
it's
pretty
close,
but
oh
geez.
A
Okay
thanks
this
next
one
is
another
update,
or
is
this
together
with
the
first,
the
recovery
from
resize
failure.
A
And
the
next
one
is
an
issue
related
to
assuming
volumes
are
mount
points.
This
is
a
bug
fix,
let's
see
is
it
is
chain
here.
A
A
Oh
yeah
yeah,
so
is
there
this
this
one?
I
think
you
have
been
reviewing
right,
determine
my
points
without
relying
on
proc
procoma
this.
I
remember
that
if
you
remember.
A
C
A
C
I'm
not
sure
if
there
is
a
pr
or
not.
I
wasn't
tagged
on
one
yesterday
there
is
one
today
I
missed
it.
A
And
oh
next
one
is
the
storage
capacity
tracking.
So
I
I
wonder
if
I
it's
a
isn't
it
michelle.
A
Wonder
you
want
to
talk
about
I'm
here?
Oh
okay,
do
you
want
to
talk
about
this
now
or
do
you
want
to
talk
about
this
later,
so
the
story
comparison
is
checking
me
and
probably
talk
about
it
now,
since
you
joined
this
right,
so
patrick
added
this
issue.
Basically,
we
did
not
track
this
in
the
beginning,
because
we
thought
we're
not
going
to
work
on
this
in
this
release.
A
But
then
this
feature
was
already
data
for
one
year,
so
automatically
there's
a
deprecation
warning
added
to
the
api
itself,
so
so
jordan
notices
about
this
one.
So
basically
this
one
we
either
have
to
go
to
ga
v1
or
we
have
to
move
to
weapon
beta.
We
want
beta
2.,
so
patrick
has
this.
I
think
he
sent
out
email
to
the
minis
as
well-
and
this
is
also
here
talk
about
this.
There
are
a
few
issues:
yeah
patrick,
why
don't
you
go
ahead?
Talk.
D
D
Okay,
so
I
disabled
that
particular
assumed
intelligence
thing
now
now
now
is
it
okay?
Now,
yes,
yeah?
Okay,
so
I
wanted
to
clarify
this
thing
without
a
scalar
first
and
that
took
longer
than
expected,
because
mostly
the
autoscaler
folks
were
not
necessarily
unresponsive,
but
we
did
when
we
did
go
through
a
proposal
that
I
had
and
the
status
was
that
it's,
we
kind
of
agreed
that
it
works,
but
it
is
complex,
but
that's
mostly
because
the
problem
itself
is
complex
of
the
different
players
here,
cdci
csi
driver,
auto
scaler
cluster
provider.
D
D
My
proposal
is
that
the
cluster
administrator
needs
to
provide
that
information
by
creating
some
manual
and
and
owned
by
the
cluster
administrator
csi
storage
capacity
objects,
and
that
approach
works.
That's
basically
what
I've
implemented.
In
my
pr,
I
also
implemented
the
necessary
code
for
auto
scaler,
but
we
really
got
stuck
on
this
question
of.
D
Can
it
be
done
in
any
simpler
way,
and
I
don't
think
it
can
be,
can
be
made
simpler,
but
that's
where
we
are
so
now,
given
that
jordan
wanted
to
do,
I
wanted
us
to
do
something
and
we
have
to
do
something
in
this
cycle.
I
went
back
to
the
cluster
autoscaler
folks
and
asked
about
this
pr
again
and
the
first
thing
that
I
that
I
was
told
is
that
we
shouldn't
wait
for
them.
If
we
think
that
the
feature
is
used
and
useful
without
autoscaler,
we
should
we
could.
D
D
D
That's
where
this
slack
thread
comes
in,
but
I
started
last
week
and
then
also
the
mailing,
the
list
that
I
sent
to
the
six
storage
google
group
trying
to
figure
out
whether
this
feature
is
used.
I
think
it
is
used.
I
got
a
few
bug
reports
over
the
last
few
months,
the
latest
actually
this
week
from
someone
from
redhead
who
has
tried
it
with
his
csi
driver.
D
It
worked
for
the
developer
and
then
the
qa
folks
ran
it
with
an
unusual
note,
set
up
a
cluster
set
up
where
the
node
name
were
longer
than
expected.
So
that
is
one
open
issue
against
the
external
provision.
D
Now
the
fix
would
be
easy
it,
but
it
does
show
that
we
that
the
feature
is
used
and
this
person
also
replied
on
on
the
six
storage
group
saying
that
yeah
it
worked
for
him
and
it
seems
to
be
okay,
so
yeah,
it's
it's
not
as
much
feedback
as
we
would
have
hoped
to
get
from
people,
but
that's
partly
also
because
people
have
been
not
reacting
to
pings.
I
I'm
fairly
sure
that
ldm
developers,
for
example,
are
using
it,
but
they
didn't
reply
on
slack,
although
I
pinged
one
of
their
developers
explicitly.
D
I
know
pmm
csi,
but
physios.
I
drive
that
I
work
on
uses
it
with
without
running
into
any
issues.
So
personally,
I
think
the
feature
is
ready
for
ga.
I
don't
expect
any
breaking
changes
or
any
any
api
changes
in
fact,
intrigue
in
in
in
kubernetes
itself,
and
also
the
the
auto
skill
I
think
will
get
resolved.
Somehow
we
we
will
have
the
attention
of
the
autoscaler
refu,
who
looked
at
it
half
a
year
ago,
or
so
last
time
at
least
again
next
week,
he's
currently
on
vacation
and
any
kind
of
improvement.
D
D
D
Because
if
we,
we
basically
need
a
decision
relatively
soon,
because
we,
regardless
what
we
do,
we
need
to
ask
for
an
exception
to
make
changes
in
125
or
124
and,
and
that
is
either
a
question
to
do
another
beta
or
go
to
to
ga
my
concern.
I
perhaps
I
should
add
that
is
that
whatever
we
do,
it
means
we
need
to
change
external
provisioner.
C
So
maybe
a
stupid
question:
if
we
wait
another
release,
will
we
get
more
feedback.
D
No,
I
don't
think
well,
it
looks
like
it
looks
like
there
are
now
some
csi
drivers
like
the
one
that
red
hat
it
was
for
cube,
weird
something
that
are
now
in
development
are
currently
getting
tested.
Those
might
get
more
actual
user
feedback.
D
D
Yes,
yeah,
but
that's
that's
what
I
did
as
part
of
the
development,
because
we
were
unsure
how
well
the
what
the
cluster
load
would
be.
That
was
all
done
as
part
of
the
the
alpha
or
beta
development.
It
might
have
even
been
alpha.
I
don't
remember
when
it
was,
but
I
did
did
scale
testing
with
six
scalability
bringing
up
a
cluster
using
a
fake
host
path,
driver
that
didn't
any
didn't
do
anything
locally,
except
simulating
capacity
and
then
creating
volumes
pots
with
volumes
and
and
that
worked.
E
D
E
Mean
I'm
I'm
personally
also
okay
with
a
v1
api.
I
you
know,
I
guess
in
the
worst
case,
if
we
find
out,
we
need
to
change
the
api
in
the
future.
We
can
just
add
new
fields
or
something
to
make
it
work.
It.
D
I
try
to
ask
jordan
whether
there
is
an
alternative,
well
it
for
doing
a
new
api
version.
That
is
exactly
the
same
as
the
old
one.
Just
doesn't
make
sense
to
me
from
a
technical
perspective.
The
only
reason
really
is
to
create
artificial
churn
in
the
ecosystem
to
remind
people
that
v1
beta1
is
not
what
they
should
be
using.
D
A
A
Unfortunately,
we
cannot
just
stay
as
is,
but
on
the
other
hand,
we
actually
still
need
to
get
an
exception
right
either
way.
We
need
to
raise
an
exception
request
from
the
release
team
as
well.
A
As
soon
as
possible,
that's
why
I'm
hoping
we
can
make
a
decision
today
and
then
you
can
start
the
thread
with
them.
Then
they
they
need
to
also
evaluate,
even
if
we
say
hey,
we
want
to
go
j.
They
may
not
agree
right
because
they
they
also
look
at
this
and
see
what
are
the
risks
it's
getting
so
late.
D
A
But
but
they
will
look,
but
they
I
would
say
okay,
so
it's
not.
I
mean
we.
Of
course
we
have
to
make
a
decision
first
and
then
they
will
decide
they
even
like.
Sometimes
even
when
the
the
sick
decides
it's
okay,
they
may
still
say
no,
so
I
have
seen
that
happen.
So
I'm
not
sure
what's
going
to
happen
with
this
one,
but
I
think
the
sooner
the
better
so
michelle
do
we
can
we
make
a
decision
today.
E
Yeah,
I
don't,
I
don't
think
more
time
will
really
get
us
more
information.
I
would
probably
say:
let's
start
the
exception,
to
go
ga
and
we
will
need
as
part
of
that,
we
also
have
to
update
the
prr
and
get
a
pr
review
for
ga
right
and
we
can
let's
go
with
that
and
if
we
get
pushback
about
going
to
ga,
then
we
can.
You
know,
drop
down
to
beta
at
that
time.
But
let's
try
for
ga.
A
D
D
D
A
No,
you
can
start
on
the
slack
channel
just
to
get
some
just
so
they
can
maybe
ask
you
questions
so
I
mean,
and
then
you
can
send
that
the
formal
one
through
email,
so
you
can
start
with
at
least
so.
They
can
start
thinking
about
this
right.
I
think
they
they
need
to
read
through
all
the
information
and
understand
you
know.
What's
the.
A
Approve
anything
that
means
we
will
have
this
deprecation.
A
A
Okay,
thank
you
all
right,
so
all
right,
so
basically
yeah
we
need
to
in
the
future.
I
think
we
need
to
think
more
about
it
when
we
go
into
beta.
I
think,
but
I
think
this
is
the
only
entry
beta
api
right
now
for
us
other
than
that.
We
have
crds
that
center
in
beta
or
just
like
a
field,
and
I
use
this
ng
api
so
so
exception.
A
Moving
on
to
the
next
csi
ethmoid
volume,
so
this
okay,
we
are
going
ga
jonathan.
F
A
A
A
Next
one
is
oh
samba,
sift
csi
drawer2ga.
E
There
was
some
julie
came
in
to
try
to
fix
the
build
so
that
we
could
actually
push
the
image
to
our
our
registry.
I'm
not.
I
think
there
was
some
failure,
though,
I'm
not
sure
if
it
got
resolved.
E
A
A
So
I
think
masaki
said
he
will
be
working
on
this
with
mastaba,
but
I
have
not
heard
about
an
update
next
one
is
csi
warning
house
additional
metrics.
So
this
one-
and
there
is
a
pr
out
review-
is
in
progress.
So
there
are
some
review
comments.
I
just
replied
just
just
so
trying
to
go
getting
this
through
next
one
is
csf,
warning
house,
programmatic
response,
just
pin
the
nic
and
see
if
he
can
start
working
on
this.
Just
waiting
for
his
response.
G
Yeah,
the
I'm
going
to
to
push
the
the
change
to
kubernetes
core
repair
today
to
change
the
the
the
feature
gate
because
that'll
be
a
one-line
change
and
I'm
going
to
also
be
pushing
another
pr
to
move
the
api,
the
out-of-tree
api
to
beta,
so
that
we
can
get
the
beta
release
out
and
then
the
metrics
changes
is
also
in
progress.
A
Okay,
maybe
okay
I'll
ping,
him
open
him
and
see
if
we
can
join
okay,
we'll
go
to
the
next
change
block
tracking,
well,
basically,
the
the
cozy.
I
think
I
think
we
still
need
reviews
right
or
I
think
michelle
you
reviewed
it
and
then
sid
also
updated,
so
probably
just
need
another
review.
H
H
They're,
like
there's
a
new
contributor
pradeep,
who
has
been
working
on
just
like
performance
improvements
overall,
mainly
switching
some
of
the
partial
calls
to
use
windows,
eye
octals.
A
Next,
one
is
a
node
expansion
secret,
so
humble
is
not
on
the
call.
I
think
I
need
to
check
with
him.
A
C
So
yeah
I
am
working
on
it.
The
scenics
part
looks
working.
However,
I
got
stuck
with
volume
reconstruction
because
right
now
we
reconstruct
only
volumes
after
qubit
is
restarted.
That
needs
to
be
unmounted,
and
I
need
to
change
it
to
reconstruct
all
volumes
to
see
how
they
are
mounted
and
that's
actually
a
blocking
point.
So
I'm
recording
reworking
the
reconstruction
and
but
the
rest
looks
kind
of
okay.
A
You
can
put
it
put
a
link
somewhere
in
maybe
in
the
dark,
so
people
want
to
review
it.
Okay,.
A
Okay,
so
next
two
are
done
and
the
csm
migration
core
bugs
and
issues
do
we
do.
We
have
any
update
here.
Michelle.
E
Jaway
has
one
bug
fix.
That's
I
think,
waiting
on
me
to
review
related
to
the
attach
limit
and
yeah
it's
on
my
plate
to
review.
A
E
I
think
the
andy
has
pr's
open
for
both
of
them.
I
had
I
merged
the
azure
file,
one
I
think,
for
azure
disk.
I
still
need
to
go
and
double
check
that,
like
we're
running
all
of
the
tests
that
we
documented,
that
is
needed
for
ga.
A
Okay
and
and
what
about
this
next,
one.
E
Update,
probably
yeah
just
to
document
the
feature
gate.
A
G
A
Right:
okay!
Next,
thanks
for
the
update-
and
next
one
is
a
gc-
oh
this
one
I
think
we
delayed,
so
I
think
yeah
delayed.
So
I
think
we
probably
don't
need
an
update
on
this,
since
this
is
delayed
and
the
next
one
is
also
delayed.
A
And
okay,
seth,
rbd
and
cfs
those
are
or
humbles.
Does
anyone
know
any
update
on
this?
What
was
in
the
pain.
A
So
so
ronald
is
also
working
on
it.
There
is
already
a
working
progress
pr
out.
A
And
next
two,
I
think
the
secret
production
and
eu's
production-
those
are
somasaki,
I
think,
he's
not
on
the
call.
I
believe
he
hasn't
made
progress.
So
it's
supposed
to
stay
in
design.
A
And
the
next
one
is
use
id
ownership
in
config
maps
and
secrets,
so
so
this
one
that
we're
not
sure
okay
yeah.
This
is
the
one
that
we
have
a
question
mark
last
time.
We
don't
know
if
the
author
is
still
working
on
it.
So
how
much
is
this
still
the
same?
Because
we
are
we're
not
sure
just
need
someone
to
own
it.
A
A
Next
one
is
the
now
chris
for
no
shutdown,
so
yeah,
so
ashtosh
is
working
on
the
pr
is
testing
it,
so
it
should
be
out
soon.
A
A
So
and
next
one
is
board
expansion
for
staple
set.
This
shalini
on
call
penny
chance.
A
So
I
think
he
I
think
she
is
busy,
so
I
have
not
got
an
update
from
him
yet
not
her.
Yet.
B
It
may
be
worth
verifying
if
she'll
have
time
to
work
on
this,
because
there's
a
lot
of
demand
for
this
video.
A
Next,
one
is
content
notifier
for
application
snapshot,
so
this
one
is
still.
We
still
need
to
get
the
issue
resolved
with
signaled
and
they
also
learned
some
comments
from
the
api
reviewer.
So
it's
still
not
resolved
yet
so
stay
in
design.
For
this
release,
that's
all
we
have
on
the
spreadsheet
is
a
is
sid
back
here.
A
G
A
Okay,
so
yeah,
so
we
already
talked
about
this
issue
so
basically.
A
A
For
g,
so
we'll
start
that
and
see
how
that
goes
and
other
than
that.
I
think
we
don't
have
any
other
issues.
So
are
there
any
and
there
anything
else
so
that
folks
want
to
talk
about
here.
So
this
one?
I
just
want
to
show
you
this
right.
So
this
is
a
this
brush
that
has
all
the
feature
blocks
that
we
opt
in.
So
we
have
a
few
quite
a
few
from
the
sixth
story
side.
A
So
if
you
you,
if
you
are
working
on
a
feature,
you
want
to
write
a
blog,
but
it's
not
here.
You
can
pin
one
of
the
seeks
or
delete
to
get
added
here
yeah.
So
I
think
we
don't
have
anything
I
mean
other
than
we
want
to
wait
for
for
sid.
So
maybe,
if
you,
if
you
you
know,
if
you
are
interested
in
hearing
an
update
from
cozy
and
stay
here
and
michelle,
can
you
please
stay
just
for
a
couple
more
minutes
for
for
si
to.
H
Join,
I
can
quickly
point
out
something
that
I
found
was
interesting.
Okay,
sure
so,
there's
there's
this
sig
instrumentation
work.
That's
happening
to
instrument
the
cubelet
for
tracing
basically
tracing.
A
H
And
it
seems
right
now,
it's
mainly
targeting
the
the
tracing
around
like
cubelet
talking
to
the
cr,
the
cri
runtime,
and
so
in
the
cap.
I
pointed
out
that
you
know
that
that's
the
same
framework
can
potentially
be
extended
to
csi
plugins
as
well.
H
Is
like
2831,
it's
not.
H
A
So,
okay,
so
so
potentially
we
can
also
do
that
for
the
for
the
volumes
that
are
used
to
buy.
A
H
A
H
Heads
up
that's
coming
in
and
I
don't
think
they're
tracking
csi
right
now,
but
something
we
can
potentially
look
at
doing
this.
H
Yeah,
I
think
it's
just
basically
tracing
like
a
context
and
like
what
is
the
actual
call
that
was
made
right
now.
That's
the.
B
Focus
we
already
trace
like
not
trace
exactly,
but
we
do
log
the
calls
like
the
meet
requirement
trick
to
the
operations
that
we
we
call.
First,
no,
no
csr
driver
but
trace,
may
be
useful
in
the
sense
that
when
a
action
was
triggered
and
exactly
and
that's
what
I
think
yeah.
A
This
is
already
merged,
so
this
is
going
to
be
in
window
24.
Then.
A
A
Thanks
for
putting
this
out,
okay,
so.
A
H
H
A
E
E
A
Okay,
yeah
sounds
good
yeah,
so
we
can.
We
can
let
him
know
yeah.
So
I
think,
let's
see
there
anything.
Okay,
yeah.
I
think
then
that's
that's
it
for
today.
I
don't
want
to
keep
everyone
here.
I'm
not
sure.
Well
we'll
see
this,
but
I
I
think
he's
you
know.
We
have
already
passed
his
message.
Basically,
he
just
I'd
like
to
get
his
get
the
cab
merged.