►
Description
Meeting of Kubernetes Storage Special-Interest-Group (SIG) Workgroup for Container-Storage-Interface (CSI) Implementation - 09 November 2020
Meeting Notes/Agenda: -
Find out more about the Storage SIG here: https://github.com/kubernetes/community/tree/master/sig-storage
Moderator: Michelle Au (Google)
A
All
right
today
is
monday
november
9th
2020..
This
is
the
kubernetes
csi
implementation
meeting.
I
think
the
main
topic
for
today
is
going
to
be
120
code.
Freeze
is
this
thursday,
so
let's
go
ahead
and
run
through
the
items
that
we
need
to
make
by
this
thursday.
So
let's
skip
some
of
these
other
items
that
are
not
in
the
kubernetes
core.
A
Let's
see
this
first
one
makes
csi
drivers
spec
pod
info
on
mount
mutable.
I
think
this
is.
This
is
more
of
a
nice
to
have
capability.
I
think
nothing.
A
Nothing
is
urgently
needing
this,
so
I
will
probably
put
this
on
the
post
code
freeze,
review
cycle.
It
probably
won't
make
120..
I
think
this
performance
e2e
test
is
also
not
critical
for
120..
Let's
skip
over
that
fixing
this
race
condition
between
node
and
stage
and
delete
volume.
I
think
this
is.
This
is
also
a
nice
to
have
to
fix
in
120,
but
I
don't
think
it's
going
to
make
it
in
time.
B
A
All
right
next
up
is
checking
if
removing
the
check
for
mount
points
in
cleanup
orphan
pod
directories.
A
A
C
A
Okay,
I
think
that's
fine,
if
you've
already
manually
tested
the
operation
generator
stuff.
C
A
All
right
cool,
as
long
as
like
we
have
some
some
sort
of
testing,
I
think
that'd
be
good
all
right,
but
so
other
than
that.
I
think
this
is
looking
in
good
shape,
so
we
should
be
able
to
make
this
next
up.
Is
fs
group.
B
A
All
right
awesome,
thank
you
and
then
slightcot
has
a
e2epr
out
for
the
change
policy.
I've
been
reviewing
it.
I
think
this
one
should
be
close
to
being
able
to
merge.
It
looks
like
he
pushed
something.
A
The
also
the
feature
flip
to
beta
is
also
as
part
of
this
pr.
I
think
hamad
were
you
gonna.
Do
you
have
another
api
pr
with
the
feature
flip.
D
Yeah
I
opened
it
and
yeah
there's
another
pier.
I
have
for
I'll
link
it
in
this.
This
thing
yeah.
I
think
jan,
took
a
look
in
the
morning
and
the
main
thing
there
is
what
discussion
is
like.
I
dropped
the
wording
about
the
default
because
that's
what
the
con,
like
the
tim,
was
suggesting.
Actually
so
we're
not
explicitly
calling
out
what
is
the
default,
but
but
yeah,
but
basically
it
says
that
if
not
specified
it,
it
always
applies
a
permission
change.
So,
okay.
D
Jordan
has
filed
that
issue
under
a
long
term
thing,
so
I
don't
think
anybody
is
like
right
now
with
this
is
going
to
investigate
it
on
that
one.
So
I
can
ping
him
again
today
and
but
I
can
ask
him
to
take
a
look
at
this
one
if
this
seems
reasonable.
A
Okay
sounds
good
and
then
have
you
had
a
chance
to
look
at
this
pr
for
the
e2e
test.
A
Okay,
yeah,
I
think,
basically
he's
adding
two
test
cases,
one
with
the
always
mode
and
one
with
the
on
change
mode,
yeah,
yeah
and
he's
enabling
it
for
csi
drivers
too.
There's
a
test
capability
for
for
fs
group
already.
A
All
right
and
let's
see
other
than
that,
did
the
metrics
change
go
in
yet.
A
Okay,
it
did
merge
okay,
so
we're
good
on
the
metrics
cool
all
right.
I
think
that's
it
and
other.
So
I
think
we
still
have
what
three
pr's
still
remaining:
the
csi
driver
change
the
api
wording
and
then
the
e2e
tests.
I
think
all
of
them
seem
to
be
nothing
controversial,
so
we
should
be
on
track
for
being
able
to
merge
all
these
countries.
E
Yeah,
so
I
tried
so
I
just
changed
this
a
little
bit
just
trying
to
break
out
like
two
buckets.
One
is
things
that
we
must
merge
before
the
thursday
and
others
are
since
we
must
get
it
done
before
central
ga,
but
not
necessarily
by
this
thursday.
E
So
for
this
thursday,
so
we
have,
you
know
the
stress
test,
so
that's
the
scalability
one,
that's
merged
finalizer,
test
merged
and
then
we've
done
some
menu
upgrade
tests.
So
myself
and
a
few
others
testers,
the
the
hostpath
is
a
driver,
so
that
works
and
then
and
then
manu
test
also
with
the
gcpd
driver
that
also
works.
So
what
I
remind
remaining
the
I
think
one
thc
is
the
matrix
work
that
the
grant
is
working
on.
I
think
his
pr
is
almost
I
mean
I'm
almost
ready,
I
think
they're,
just
some
small
things.
E
Sean
chan
is
going
to
take
another
look.
He
was
out
for
a
few
days
last
week.
So
now
one
question
we
have
is:
let's
say:
if
we
get
this
one
merged,
he
has
a
e3
test,
but
then
that
one
I
mean,
doesn't
really
have
any
official
image
that
we
can
use.
E
So
how
are
we
going
to
do
that?
I
mean
it's
a
chicken
egg
problem
again
which
e
to
e
test
oh
matrix,
matrix
so
matrix.
This
is
like
yeah,
so
he
submit
so
grant
submitted
a
work
in
progress
e3
test.
But
then
we
don't
really
have
an
image
that
oh,
this
isn't
kubernetes
kubernetes
yeah
yeah,
that's
the
and
then
we
can
really
and
then
for
this
one.
This
matrix
is
like
a
new
feature
right.
We
can't
really
cut
a
patch,
that's
not
right!
So
yeah
I
mean.
Can
we
just
do
something.
E
E
E
Is
you
know
the
right
now?
It's
not
that
easy.
I
mean
I
haven't
figured
out.
I
asked
grant
to
take
a
look
also.
Maybe
we
can
ping
you
and
you
can
help
us
figure
out
how,
because
I
was
trying
to
change
that
in
the
okay
image.
You,
though,
maybe
I
didn't
change
the
right
place.
Maybe
we'll
continue
yeah.
A
E
E
A
E
E
Okay.
Okay,
so
that's
that's
a
match.
Some
that's
metrics
yeah!
So
that's
the
top
priority
for
today
and
then
and
then
after
that.
Basically
there
are
a
few
prs.
I
think
they
are
in
good
shape.
It's
just,
I
think,
you're
waiting
for
this
like
a
matrix
work
before
we
converge
or
we
can
merge
now.
Do
you
think
what
do
you
think
the
basically
this
moving
snapshot
crd
to
b1?
That's
in
external
snapshotter,
uh-huh.
A
Yeah,
I
think
we
should.
I
think
we
should
merge
that
I
was
deferring
to
saad
to
do
the
api
review,
for
it.
E
Yeah
but
saad
is
so
busy
because
I
so
you
have
already
taken
a
look
right.
You
have
already
taken
a
look
of
the
earlier
earlier
prs
and
and
shantae
also
took
a
look
of
it.
It's
there's.
No,
we
don't
have
any
schema
changes.
So
what
else
do
you
want
to
look
we're
not
reading
any
schema
changes.
A
Yeah,
the
the
main
thing
is,
I'm
currently
trying
to
review
all
the
the
fs
group
related
stuff.
So
I
I
that's
why
I
defer
to
sod
to
handle
the
snapshot
stuff
but
I'll
I'll,
first
ping
sod
today
and
see
if
he
has
time
to
take
a
look.
E
Okay,
can
we
defer
that
to?
Can
we
also
ask
like,
if
does
not
have
time?
Can
we
ask
shan
chan,
take
a
look,
because
it's
really
just
making
sure
that
the
the
two
api
versions
are,
you
know
one
is
like
both
are
served.
One
is
the
store,
the
other
one
is
not
it's
mainly
because
of
that.
E
Just
that
right
and
there's
nothing
else.
That's
yeah
yeah
the
schema
change,
and
then
we
also
tested,
at
least
if
you
look
at
the
pr's
that
are
submitted
the
past
tests
except
for
unit
tests.
That's
because
we
you
know
manually
do
something
so
unit
test
is
not
going
to
pass,
but
all
the
other
tests
passed
right.
So
I
mean
what
else
are
there
that
we
actually
really
need.
A
E
A
I
think
the
most
important
is
going
to
be
the
merging
the
one
that
adds
the
new
v1
api,
but
doesn't
change
any
of
the
colors.
I
think
that's
the
first
step.
E
Right
and
then
there
are
also
controller,
and
then
there
is
a
the
main
thing
that
actually
has
to
get
merged
by
thursday.
Is
this
entry
one
right,
the
one
that
I
highlighted,
remove
the
feature
gate
yeah.
B
E
That's
the
only
thing
that
we
have
to
get
merged,
because
everything
else
is
arbitrary
and
there's
really
no
relationship
between
them.
Actually
I
mean:
can
you
merge
that
one
I
mean
tim
has
already
given
a
he
has
already
approved
it.
A
E
A
You
know
flip
the
la
it's.
The
last
thing
we
flip
the
the
main
thing
is,
I
guess
so,
maybe
once
we
merge
this
change,
that
adds
the
v1
apis.
A
A
Keep
the
controller,
keep
the
controller
with
the
older
version,
so
keep
the
controller
on
the
3.0
and
that
should
still
function
yeah.
So.
A
E
We,
oh,
can
we
do
that?
Oh
yes,
are
you
saying
just
changing
just
any
images,
only
change
the.
E
A
A
E
E
E
Okay,
yeah:
I
can
update
that
one
yeah.
Maybe
I
will
submit
another
one
just
for
that.
Okay,
so
so
that's
this
is
those
are
the
things
that
I
think
must
merge
by
by
thursday.
E
Okay,
do
you
agree?
Those
are
the.
A
E
Are
there
are
other
things
that
I
below
this
one
add
other
things
that
we
we
need
to
do
before
ga,
but
not
really
necessarily
by
thursday.
So
those
are
like.
We
need
to
update
release
tools.
We
need
you
know
all
of
those
external
to
use
a
new
image
which
I
already
have
a
pr.
It's
just.
The
images
are
not
not
the
real
one
and
then,
and
also
oh,
there's
this
new
one,
the
reload
third
thing
for
web
hook,
so
yeah,
so
I
yeah.
E
E
E
Yeah
and
then
okay
yeah,
I
think
the
rest
of
them
are
or
like
we
can
do
like
after
thursday
yeah,
maybe
it's
the
thursday,
so
maybe
yeah.
Maybe
the
first
thing
is
to
just
to
change
the
api
entry
crds.
Maybe
that
would
be
the
first
one
yeah.
That
would
be
good
good
to
try
that
out.
Without
all
the
images.
A
Right
I'll
I'll
go
ahead
and
ping
start
again
to
take
a
look
at
that
api
pr,
so
that
we
can
get
moving
on.
A
E
E
E
E
Okay,
so
then,
maybe
just
ask
me
just
to
avoid
too
many
confusion
too
many
pr's,
maybe
just
to
show
him
the
one
that
we
without
controller
one.
Maybe
that's
just
you
know,
hopefully
that's
okay
forward
and
then
so
basically
just
merge
that
one
and
then
the
other
one
is
the
the
intricate
one
right.
So
those
two
those
are
like
a
must-have
for
thursday
yeah.
Okay,
all
right.
A
All
right,
I
guess
all
the
other
items
we're
not
targeting
120,
but
do
folks
want
to
give
an
update
on
them.
D
I
don't
have
any
new
update
for
expansion,
I
think
kk
on
on
slack
and
he
was
saying
that
he
hasn't
been
able
to
work
on
anything
related
to
kubernetes
for
last
month
or
so
so
he
and
he
took
the
task
of
like
the
pvc
being
deleted
while
spending.
So
I
not
sure
if
we'll
be
able
to
get
this,
and
so
we'll
have
to
try
next
quarter.
D
A
F
Just
a
short
reminder
for
the
generic
fml
inline
volumes
and
csi
fml
inline
volumes.
You
will
have
that
api
meeting
in
half
an
hour.
F
F
F
Yeah,
it
sounds
good.
It
might
just
be
that
it's
too
early
anyway,
but
then
I
need
to
start
working
on
on
my
plan
b.
A
Yeah
sounds
good,
probably
at
least
for
the
next
few
days.
I
think
most
of
us
are
gonna,
be
busy
with
reviewing
the
120
the
code
freeze
items,
but
I
think
the
later
half
this
week
should
be
more
freed
up.
Okay,
all
right!
Thank
you
very
much.
F
Just
a
question:
you
wanted
to
ask
six
scalability
about
running
more
scale
tests
about
volume
provisioning.
Have
you
started
that
or
shall
I
do
that
on
this
first
year's
meeting?
I've
put
basically
some
topics
into
the
regulatic
storage
meeting
agenda
and,
and
one
of
them
is
doing
more
skill,
testing.
A
I
pinged
six
scalability
on
slack.
They
did
not
really
respond.
I
think
some
one
person
on
test
infra
responded
and
said
they
don't
have.
A
They
don't
yet
have
capacity
planning
for
large
clusters,
at
least
from
a
test
infra
point
of
view,
but
I
think
it's
the
current.
I
think
I'm
I'm
a
little
confused
because
they're
talking
about
moving
scalabilities
tests
over
to
the
kubernetes
infra,
I
guess
currently
the
current
scalability
tests
are
not
running
on
top
of
the
the
the
kubernetes
testing
the
front
thing
which
I
guess
is
confusing,
but
I
think
it
it
might
be
worth
actually
going
to
like
the
sixth
scalability
meeting
and
bringing
it
up
with
them.
A
F
F
And
and
proposals
basically
and
I'll
try
to
get
their
opinion.