►
Description
Kubernetes Storage Special-Interest-Group (SIG) Volume Populator Design Meeting - 11 May 2021
Meeting Notes/Agenda: -
Find out more about the Storage SIG here: https://github.com/kubernetes/community/tree/master/sig-storage
B
B
B
We
are
two
days
from
the
enhancements
freeze
and
the
good
news
is
tim
has
reviewed
the
the
split
kep
again
and
offered
his
comments
and
he
still
seems
happy
with
them,
but
he
still
has
this
nervousness
about
the
backwards
compatibility,
breaking
part
of
the
change,
and
so
he
he
pulled
in.
I
want
to
say,
jordan.
B
Let
me
open
up
this
kept,
so
we
can
look
at
it.
B
Yeah
it's
this
case,
so
I
I
don't
know
if
jordan
is
gonna
respond
or
if
and
if
not
what
we
can
do.
B
I,
I
guess
I
guess
I
wanted
to
make
sure
that
we're
all
on
the
same
page
about
this.
This
case
you
in
particular
the
the
backwards
compatibility
yeah.
It's
a
compatible
change
with
the
data
source.
Have
you
read
tim's
concern
here
and
do
you
share
his
anxiety.
C
I
do
share
his
anxiety
in
terms
of
anytime.
You
make
a
change
in
terms
of
behavior,
that's
been
around
for
a
while
there's,
probably
someone
whether
they
should
be
dependent
on
it
or
not.
That
is
dependent
on
it
yeah.
We
found
that
out
when
we
did
the
read-only
secret
change,
and
sometimes
you
just
have
to
make
the
change.
I
think
it
would
be
interesting
to
get
jordan's
feedback
on
it
as
well.
B
Well,
yeah,
I'm
wondering
is
your
stance
that
that
we
should
just
do
this
because
the
risk
is
small
enough
or
or
should
we
be
saying,
no,
let's
be
more
cautious
and
try
to
do
something
other
than
this.
C
I
I
think
the
risk
is
there,
but
the
cost
of
fixing
it
is
pretty
low,
and
so
I'm,
okay
with
it.
B
Breaking
changes,
unfortunately,
so
yeah
I
I
don't
know
like
if
we
should
just
be
pushing
to
say
no,
the
risk
is
small,
no
there's
no
way
anyone
could
be
depending
on
this.
Let's
just
do
it
or
if
we
should
take
a
step
back
and
say
well,
you
know
you
know
we
really
want
to
be
extra
conservative
here
and
therefore
we'll
I
don't
know,
put
a
feature
gate
around
this
or
you
know
some
he
had
said.
Was
it
one
of
jordan's?
B
I'm
sorry,
one
of
tim's
comments
was
about
adding
a
like
an
example:
cube
cuddle
command,
or
was
it?
Why
can't?
I
find
it
now?
Oh
here
it
is.
If,
when
this
goes
in,
we
should
publish
a
cube
cuddle,
get
command
that
prints.
Any
instances
of
I
guess
the
this,
the
you
know
a
list
of
all
the
pvcs
that
would
tickle
this
problem.
B
B
Go
look
at
existing
clusters
and
see
who's
depending
on
its
functionalities,
because
if
anyone
ever
depended
on
it,
the
dependency
got
erased.
The
moment
they
created
the
pvc
so
yeah
we
seem
to
be
stuck
in
this
sense
of
like
we
think
the
risk
is
small,
but
we
don't
know
if
it's
close
enough
to
zero,
that
we
can
just
go
with
it.
B
No,
that's
not
that
one,
it's
the
other
one!
If
we
don't
implement
this,
then,
instead
of
just
removing
these
313
lines,
we
have
to
go
in
and
carefully
change
them
to
behave
differently,
so
that
for
core
data
types
it
continues
to
erase
it
and
allow
the
pvc,
but
for
non-core
ones.
It
would
start
going
through
the
normal
process
and
then
you
and
then
whether
it
passed
or
not,
would
be
gated
on
the
feature.
Gate
of
any
volume
data
source.
B
B
So
we
we
have
to
stop
erasing
those
and
so
there's
already
a
small
chance
that
if
we
make
this
change
and
someone's
using
a
non-core
data
type
and
the
feature
gate
is
well,
I
don't
know,
I
don't
know
how
it
would
interact
if
the
feature
gate
was
set
to
alpha
and
we
changed
the
check.
But
but
yeah
there's
this
weird
interaction
where
you
know
the
the
feature
gate
doesn't
allow
you
to
do
this,
but
the
the
the
utility
code
is
is
not
dropping
the
not
dropping
the
field.
B
Other
thing
we
discovered
was
that
I'm
I'm
trying
to
find
the
code
again.
No,
that's
not
this
one,
not
this
one,
this
one
in
the
in
the
code
that
implements
this
change
or
the
the
drop
disabled
fields.
B
I
think
that
we
weren't
able
to
perform
feature
gate
checks
here
for
some
reason
like
there
was
not
enough
context
in
the
in
this
utility
code
to
do
the
feature
gate
checks-
I
I
actually
I'm
not
100
sure
of
that.
So
don't
take
my
word
for
it,
but
there
were.
B
D
There
is
there
a
is
there
a
chance
that
this
causes,
like
a
user
like
the
user
confusion
around
core
data
type
versus
non
core
data
type,
the
difference
in
behavior
is
a
risk,
and
also
that
should
be
added
to
the
conversation
like
the
difference
in
behavior.
Is
that
not
also
a
risk
that
that,
like
should
be
weighted
against
the
breaking
backwards?
Compatibility
of
change.
B
Yeah,
so
we
talked
about
you
know:
should
we
should
we
just
allow
core
data
types,
the
same
as
everything
else
and
the
issue
with
that
was
we
know
we
will
never
allow
a
data
populator
for
one
of
the
core
data
types
because
like
for
such
a
data,
populator
would
have
to
be
sort
of
owned
by
the
community
and
all
of
the
all
of
the
implementations
of
the
data
populators
are
supposed
to
be
out
of
tree.
B
You
could
invent
a
new
crd
that
pointed
to
that
core
data
type
and
achieve
that,
so
we're
not
preventing
anyone
from
like
making
a
pod
populator
we're
just
saying
if
you,
if
that's
what
you
want,
you
have
to
wrap
the
pod
in
another
crd
before
you
create
the
the
pvc
that
points
to
that
data
source,
so
that
somebody
else
could
make
a
different
one.
That
also
was
a
pod
related
data
source,
but
did
something
different
right,
so
it
has
to
be
there
needs.
B
B
Everyone
agrees
that
just
dropping
it
was
always
the
wrong
thing
to
do,
and
now
we
have
to.
We
have
to
stop
dropping
it
for
everything
that
we
intend
to
allow,
which
is
everything,
that's
not
core,
and
it's
just
a
question
of
of
the
the
core
ones,
and
so
we
we're
yeah
we're
stuck
and
tim
doesn't
feel
like
he's
got.
B
My
reading
of
his
comments
is
that
he
doesn't
feel
like
unilaterally
approving
this
and
so
he's
looking
for
someone
to
either
support
or
or
argue
with
him,
and
I
thought
I
thought
saad
if
you
wanted
to
throw
your
two
cents
in
here
with
regard
to
this
specific
issue,
or
if
you
have
a
way
of
pulling
in
jordan,
I
I
don't
yeah.
C
I
just
posted
a
comment
and
to
throw
in
my
support
and
for
jordan
yeah.
I
can
ping
him
offline
or
if
you
want
to
send
an
email
that
I
think
should
be
sufficient
as
well.
Okay,.
C
Okay,
I'll
start
a
thread
with
you
and
him
on
there.
B
Okay,
yeah
yeah,
because
it
we
just
need
to
get
this
one,
this
one
approved
and
the
other
you
know
the
actual
pr's
for
for
the
implementation
are
ready
to
go
and-
and
that's
the
deadline
we're
facing
this
week.
So
that's
what
I'm
prioritizing
so
yeah
thanks
side
and
if
anyone
else
wants
to
put
in
comments
you're
welcome
to.
I
really
would
like
to
get
to
you
know
and
approve
from
from
tim
on
this.
B
This
cap-
I
I
wanted
to
mention
there
is
one
other
tiny
change
that
they
got
requested
on
my
other
cab
open
link,
this
one
I
just
had
because
we
merged
the
the
volume
populators
kept
last
cycle,
and
we
didn't
do
the
implementation.
It
has
the
wrong
version
on
it.
So
this
this
pr
bumps
the
version
from
121
to
122..
B
B
A
B
Okay,
yeah
yeah,
so
sorry
I'm
losing
here
it
is
yeah
david
david
approved.
It.
A
B
Here
I
I
added,
I
added
the
whole
pr
section.
I
filled
it
all
out.
Most
of
the
answers
to
the
questions
are
not
applicable
or
no,
but
I
did
answer
all
of
the
questions
in
the
prr
little
template.
So
david
was
happy
with
that.
He
approved
it.
So
that's
good
to
go
yeah.
We
just
got
to
get
these
two
things
merged.
I
have
links
here
to
both
of
the
kubernetes
related
prs.
I
have
rebased
them
and
updated
them
for
122
just
today,
so
they're
they're
well,
assuming
that
the
keps
merged.
B
We
could
merge
these
immediately
if
we
wanted
so
but
they're
here
for
review
and
if
there's
no
questions
or
discussion
about
those
I
want
to
go
on
to
the
other,
the
the
auditory
stuff.
That's
the
lower.
B
Priority
everyone's
cool
with
the
entry
stuff
in
the
cap
yep
once
okay,
so
so
yeah
I've
been
reading
through
the
docs
about
how
to
create
proud
jobs
and
it's
seems
pretty
involved.
I
I
think
that
the
the
best
thing
would
be
to
have
a
sample
to
follow.
B
So
what
I
was
going
to
ask
is
shing,
if
you
I
don't
know,
if
you
did
any
of
the
work
related
to
the
snapshot
or
setting
your
proud
jobs
for
it
or
if
you
had
help
or
if
someone
else
here
knows
who
did
the
work
for
like
other
csi
side
cars,
but
I'd
like
to
pick
your
relatively
simple
one
and
just
go
back
in
history
and
see
how
how
how
the
proud
jobs
were
added.
So
I
have
a
sample.
I
can
copy.
B
A
B
A
B
A
I
know
I
know
I
think
cozy
also
has
a
few
if
you
go
proud,
but
where.
C
Yeah
you
have
to
go
into
that
repo.
You
have
to
create
your
own
prs,
no
one's
going
to
do
it
for
you,
and
so,
if
you
look
at
one
of
the
existing
projects,
just
kind
of
follow
their
their
model
and
it
should
work
and
if
you
need
help,
I
think
I'd
recommend
reaching
out
to
michelle
she's
done
this
a
lot
for
a
lot
of
the
csi
repos.
B
Maybe
I'm
just
gonna
have
to
do
it
by
trial
and
error,
but
yeah
I
was,
I
was
going
through
the
proud
jobs
and
realizing
like
stuff
seems
to
be
spread
out
in
multiple
directories,
or
at
least
that's
what
it
looked
like
to
me.
Maybe
that's
not
what's
going
on
okay,
so
so
you
think
michelle
knows
more
about
this
than
anyone
else,
at
least
within.
B
All
right,
well
I'll
I'll,
look
for
pr's
that
she's
pushed
and
talk
to
her
when
I
get
stuck
and
that'll
be
helpful
because
yeah,
that
is,
you
know
as
soon
as
as
soon
as
we
get
the
the
caps
done.
I
don't
think
there's
I
mean
other
than
responding
to
review
feedback
on
the
the
kubernetes
related
prs.
It's
just
the
out
of
tree
stuff.
B
That
needs
a
lot
of
work
to
add
tests,
to
add,
to
add
the
proud
jobs
to
get
the
releases
out,
and
I'm
also
remembering
that
that
we,
we
did
say
we're
going
to
add
some
metrics
and
that
code
is
not
implemented
yet
so
the
metrics
remain
were
mentioned
in
the
keps,
so
that
we
need
these
new
controllers
to
be
generating
metrics.
So
that's
that's
extra
code.
I
have
to
write
as
well,
but
that's
that's
lower
on
my
priority
list
after
getting
the
alpha
release
out
so
okay.
B
Here
anything
else
for
today.
B
Any
other
questions
comments,
concerns
I'll
look
for
that
thread
with
jordan
assad.
Thank
you
for
starting
it
and
I'll
I'll
state
my
case.
Yet
again,
I
feel
like
I
keep
making
the
same
argument
and
we
keep
losing
context,
but
hopefully
we'll
get
over
the
finish
line.
This
time.
B
A
I
just
said:
there's
this
link.
I
think
it's.
Actually
michelle
submitted
this
pr.
That
added
drops
for
a
few
new
repos.
Maybe
you
can
start
with
this
one,
then
you
can
ask
her
questions.