►
From YouTube: Data Protection WG Bi-Weekly Meeting for 20200506
Description
Data Protection WG Bi-Weekly Meeting for 20200506
A
A
Today's
agenda
so
today,
we'll
all
be
doing
a
design
review
of
the
warning
group
cap.
There
are
several
people
asking
about
it,
so
let's
take
a
look
and
discuss
body
today
and
then
we
will
give
an
update
on
the
container
notify
proposal
which
we
discussed
in
previous
meetings,
and
after
that
we
can
have
some
opening
issues.
If
there
are
people
have
any
issues
to
discuss.
A
A
The
reasoner
is
the
cap,
so
this
is
the
actually
the
application
snapshot
cup
that
introduces
API
to
do
application,
snapshot
and
back
up,
but
there
are
also
other
cases
that
are
not
covered
by
that
cap.
The
first
use
case
is
that
we
can
have
a
boring
group
that
allows
users
to
manage
mobile
learning
together
and
those
volumes
could
be
ill
onto
the
same
application.
A
For
example,
they
can
use
this
to
group
all
the
volumes
in
the
same
stiphu
set,
and
then
the
second
use
case
is
that
for
us,
I
think
for
some
storage
systems,
maybe
not
a
lot,
but
for
some
Street
systems
they
always
manage
the
volumes
in
the
same
group
even
for
them
to
create
one
single
volume.
They
have
to
create
a
cook
first
and
then
create
creative
onboarding
in
that
group.
A
So
this
would
be
convenient
and
then
the
third
use
case
is
that
we
can
also
use
this
construct
as
a
source
to
take
a
snapshot
for
all
the
volumes
in
the
same
one
group.
And
then,
if
the
system
can
provide
consistent
group
snapshot
support,
then
they
can
use
this
to
do
that,
and
the
fourth
use
case
is
that
this
is,
of
course,
is
not
something
that
we
are
continuing,
our
trust
immediately
by
the
destination.
That's
the
potential
future
use
case.
A
You
can
also
use
this
constructor
to
do
a
group
replication
if
a
storage
system
supports
it
use
case.
Five
is
this
is
for
the
the
volume
placements
use
case,
so
we
could
use
this
one
to
say
that
we
want
to
spread
volumes
across
the
street
pools
or
stack
them
on
the
same
soda
pool,
but
this
one
since
there
are
actually
there
is
already
a
cap
that
is
out
so
G
pool.
A
So
I
think
this
is
something
that
I'd
like
to
figure
out
whether
we
can
make
a
decision
on
this
one,
whether
we
should
just
focusing
on
this
other
cap
won't
in
place,
then
they're,
not
using
group
concept
for
that.
So
this
is
a
chibi
determined,
and
the
six
use
case
is
that
this
can
also
be
used
together
with
that
application
snapshot
cap.
A
So
this
would
be
one
resource
managed
by
that
application.
Snapshot.
Er
D,
second,
seventh,
is
that
if
some
applications
don't
use
the
workload
API
such
as
deficit,
then
they
probably
can
just
use
this
boring
group
and
they
can
still
manage
those
volumes
together.
So
those
are
the
use
cases
that
I
have
here.
There
are
any
questions
regarding
the
use
cases.
A
A
Okay,
so
it
goes
so
goes
says
to
providing
API
team
and
motive
audience
to
gathering
a
group
and
provide
an
API
to
take
snapshot
of
a
group
of
volumes
and
this
the
third
ago.
So
that's
we
can
also
use
this
one
to
do.
You
want
emplacement,
but
we
want
to
decide
whether
we
want
to
do
that
or
not,
and
the
fourth
go
is
that
I
think
this
group
API
should
be
extensible
so
that
if
we
want
to
support
other
features,
then
we
should
be
able
to
use
it.
B
B
C
A
A
B
Like
some
of
the
goals
and
some
of
the
use
cases,
they
don't
necessarily
they
don't
necessarily
fit
into
each
other's
picture.
For
example,
some
of
the
storage
systems
does
require
volumes,
get
provisioned
the
time
when
you
specify
when
you
create
a
group,
because
for
them
to
able
to
be
able
to
support
topology,
they
require,
let's
say
consistency.
A
consistent,
wouldn't
group
could
be
physically
allocated
in
same
black
right
yeah.
B
Doesn't
apply
to
existing
Warrens
in
the
system
already
in
occupy.
The
second
thing
is
that
group
snapshot
doesn't
really
make
sense
for
especially
for
some
of
the
use
cases
either
start
there,
for
example,
if
you
want
to
simple
it,
just
manage
the
lifecycle
of
group
volumes,
a
using
one
group
to
achieve.
That
seems
to
be
a
little
bit
too
much
from
my
perspective,
because
you
know
effectively,
you
cannot
really
do
consistency.
Group
consistent
snapshots
were
those
randomly
grouped
together
forums
and
a.
B
Secondly,
even
though
underlying
storage
system
that
support
group
consistence,
it's
not
shock,
we
ran
into
the
risk
where,
if
the
worm
are
not
allocated
on
the
same
rack
or
something
that
which
the
underlying
storage
system
does
require,
however,
we
still
expose
this
worm
group
API
to
the
users.
It
makes
this
feature
not
usable.
A
A
Do
we
want
to
be
able
to
start
with
something
small
and
then
is
that
something
that
we
can
make
progress
on
yeah?
Definitely
there
are
a
lot
of
concern.
That's
why
this
was
actually
we
started
discussing
such
a
long
time
ago
and
then
I
think
kind
of
getting
stuck,
because
there
are
a
lot
of
concerns.
I
guess.
B
A
That's
why
I
want
you
bring
this
opera
is
so
it's
especially
I
think
the
one
group
group
snapshot
and
what
emplacement
those
are
very
two
very
different
goals.
Right,
I
think
he
usually,
we
started
with
a
group
snapshot
and
then
we'll
in
placement.
I
think
this
is
an
something
that
it
was
added
later
so
I
do
want
to
make
a
decision
and
then
see
what
we
want
to
I
was.
C
Groups,
yeah
volumes
that
you
can
put
into
a
consistency
group
and
they
are
consistency
across
all
the
volumes
in
the
group.
Okay
and
I
understand
kind
of
the
more
use
cases
you
coupled
together,
the
more
difficult
it
is
to
get
this
across
the
line,
but
we're
talking
about
core
kubernetes
api
s,
and
we
want
to
kind
of
make
them
as
generic
and
reusable
as
possible.
So
I
would
shoot
up
try
to
at
least
look
at
all
the
use
cases
and
see.
Would
it
fit?
Does
it
make
sense?
Can
we
tackle
this
use
case?
A
A
So
those
newly
it's
either
they
support
consistent
group
snapshot
or
consistent,
good
replication,
so
I
some
some
stresses
and
they
kind
of
support
those
two
as
two
different
features.
Some
can
support
both
so
here
I
have
this
consistent,
consistent
group
snapshot
here,
replication
I,
think
I
think
you
mentioned
here
actually
so
here
I
did
mention
consistent,
good
replication
here,
so
it
is
covered
naturally
by
it's
like
a
sub
case
of
three
and
fork,
and
maybe
I
can
add
a
supple
it
to
make
it
more
obvious.
A
Okay,
sure
sure,
okay,
so
I
yeah
I,
don't
think
this
part
is
important
that
you
know
if
we,
if
we
want
introduce
this,
so
we
want
to
make
this
one
generic
and
extensible
okay
and
then
now
goes
okay,
so
I
just
mention
here
that
the
we
are
not
really,
although
we
mentioned
this
use
case,
but
this
may
be
something
in
the
future.
We're
not
really
focusing
on
talking
about
replication
group
here.
So
that's
the
non
go.
A
Over
to
the
proposal
section
so
here
we
are
proposing
to
introduce
three
new
CR
DS
voting
group
on
a
good
class
and
group
snapshot
to
create
a
new
voting
group.
So
here
we
have
two
ways:
one
way
is
you
create
a
group
with
existing
volumes
and
either
you
provide
a
list
of
PVCs
or
just
use
a
selector,
and
the
second
way
is
create
a
empty
group.
First
then
create
an
individual
volume
with
that
group
ID
and
then
basically
add
one
one
new
volume
to
the
group
at
a
time
and
then
for
snapshot.
A
Group
snapshot
can
be
created
from
a
source,
rolling
group
and
the
restore
so
restore.
There
are
also
we
could
support
it
in
two
ways.
Maybe
we
can
start
cinco
and
but
then
so
there
are
two
possible
ways
that
well
is
actually
the
second
one.
We
can
just
create
one
wall
in
front
snapshot
at
a
time
in
a
creative.
All
you
create
a
group
with
all
of
those
volumes.
A
A
D
A
F
A
A
G
A
Yes,
I
think
we
could
simplify
things.
I
was
thinking
even
like,
for
example,
this
I
can
think
of
like
let's
start
with
just
this
option,
then
you
know,
then
we
can
decide
why
that
you
add
this
one
day
right.
So
this
is
one
one
thing
and
then
for
this
great
group
of
course,
we
also
have
like
too
few
options.
We
can
decide
whether
we
want
to
start
with
something
simple
so
but
but
then
we
also
have
this
other
place
of
what
in
place
in
the
use
case
right.
A
So
then,
if
this
is
already
existing
volumes
and
that
doesn't
make
sense
anymore,
his
wording
is
already
placed
right.
So
for
that,
then
we
will
need
this
for
volume
placement.
Then
we
need
to
say,
create
a
new
volume
right
otherwise
for
existing
volume,
then
that
that
one
really
does
not
have
a
place
here.
So
yes,
we
need
to
make
some
decisions
and
they
may
be
narrow
down
to
say
well,
should
we
start
with
something
simple,
you
know
so,
but
I
think
with
you.
E
A
E
E
A
Here
we
are
I
think
here
we
are
only
talking
about
like
a
new
create
a
new
PVC,
so
I
probably
should
make
this
more
clear
yeah.
This
is
more
like
consistent
with
what
do
we
have
for
the
existing
snapshot
feature
the
exist
existing,
create
warning
from
snapshot
feature,
so
you
need
to
create
a
new
PVC
Friday,
that's
what
we
hired
currently.
So
that's
what
I'm
having
mine-
oh
yeah,
okay,
because
otherwise
I
think
I
need
something
else
that
we
need
to
design
something
I
was
just
support
in
other
cases,
but
yeah.
A
Okay,
so
for
volume
placement
support,
then
we
didn't
see
many
results.
There
are
already
straight
post
exists
and
then,
in
the
warning
group
class
there
is
a
allowed
topology
field
for
force
to
decide
at
the
accessibility
of
the
volumes
to
the
pause
and
the
notes.
I
just
I
mentioned
the
signs.
It
does
not
have
a
question
just
saying
it's.
What
the
that's
the
other
cab
the
street
before
has
the
capacitive
group
does
not
have
a
capacity
is
really
so
ok.
A
So
let
me
go
over
the
definitions,
volume
group
class,
so
he'll
here
right,
so
we
have
driver
of
your.
We
have
the
parameters.
Then
we
have
the
sphere
to
specify
whether
we
support
whether
this
class
is
force.
Group
snapshot.
Not
this
is
a
boolean
and
then
this
is
allowed
topologies
so
like,
for
example,
this
API
I
was
thinking.
You
know,
I,
don't
know
if
we
really
need
it.
So
that's
something
we
can
also
discuss.
If
you
want
to
make
it
as
simplified,
maybe
we
don't
really
need
to
us.
A
A
Otherwise
we
will
create
that
with
some
existing
PVCs,
so
the
source.
So
actually
I.
Should
me
just
look
at
this
right
here.
So
this
is
a
source
field,
so
it
can
have
a
PvE
PVC
list
or
a
selector.
So
that
means
we
are
select.
We
are
just
adding
existing
PVCs
to
this
group.
We
are
choosing
one
of
this
too.
A
Otherwise,
if
we
are
choosing
this
one,
then
that's
for
restore.
Of
course,
this
isn't.
When
you
discuss
it,
we
want
you.
Do
we
want
to
support
this
one
right?
What
do
we
want
to
just
L
restore
time?
We
can
actually
just
create
a
one
or
in
front
snapshot
at
a
time
instead
of
introducing
this,
but
this
is
something
that
we
can
discuss
and
then,
if,
if
the
sauce,
if
this
source
is
now,
then
then
that's
an
empty
or
a
group.
A
Status
field,
so
we
will
have
the
PVC
list,
so
we
know
what
volumes
are
added
party
B
C's
are
added
to
this
group,
and
then
we
have
this
indicating
whether
this
one
is
created
successfully
or
not,
see
how
it
was
ready
and
then
also
having
an
arrow
field,
and
the
full
group
snapshot
also
have
a
stack
and
status
for
the
group
snapshots
back.
So
here
we
have
this
source,
so
we
maybe
we
don't
really
need.
A
C
A
A
A
One
good
yeah
so
so
warning
cooks
back
has
this
voter
group
class
names
filled
right,
okay,
yeah
and
then,
and
then
the
source
could
be
if
we
are
creating.
If
we
are
just
creating
that
with
existing
volumes,
then
is
going
to
be
either
a
PVC
list
or
based
lecture
that
select
the
PVCs
that
we
want
now.
This
one
is
really
for
the
third
one
is
ready
for
restore.
So
if
we
don't
want
it-
and
you
know,
we
can't
start
without
this
I
think
so.
This
is
like
yeah.
A
H
A
A
I,
so
I
need
to
double
check
and
I.
Think
I,
remember,
I,
look
at
the
imputations
in
cinder
seems
to
be
they
usually
they
just
creating
the
video
for
waterings
from
snapshots
and
then
put
them
into
groups,
but
what
I
was
trying
to
say
they
is
that
I,
don't
know
if
any
serious
matter
you
require
I
mean
you
can
only
give
this
whole
group
snapshot
as
a
source.
I
think.
Usually
they
just
create
one
in
Tampa
and
of
course
they
put
them
in
this
group.
A
A
So,
okay,
so
okay,
let's,
let's
start
to
go
back
to
this,
think
I
have
that
somewhere,
maybe
in
the
beginning,
so
I
understand
the
case
that
you're
talking
about
I
so
that
can
actually
be
covered
by
yes,
so
create
a
created
empty
group
first
and
then
create
each
individual
bottom.
So
those
warring
will
be
like
a
volume
created
from
a
snapshot.
So
you
can
just
you
can
actually
do
a
create,
create
volume
from
France
snapshot
and
then
add
that
to
the
scoop
at
the
same
time,
that
should
work
for
those
systems.
Wray.
G
A
G
A
I
think
this
look
flow
with
the
work
which
I
think
this
is
not
yeah.
I
think
this
one
is
actually
not
required.
I
haven't
seen
evidence
that
this
is
required.
I'm
just
saying
this
is
like
a
communist,
but
so
for
those
systems
they
need.
You
have
a
idea
right,
so
they
can
create
an
empty
group
first
and
then
it
can
create.
A
A
G
A
Understand
I
because
I
you
know
what
done
this
before
so
I
think
this
case
would
work
for
everyone.
I
haven't
seen
any
that
this
does
not
work
the
case
that
does
not
work
is
you
can
for
some
store
systems.
You
cannot
add
one
individual.
What
in
one
existing,
even
volume
to
existing
group
that
you
cannot
do,
but
if
it's
new,
then
that
would
work
for
everybody,
so
I.
A
I
mean
because
I
so
since
I
leave
the
scene,
sender
I
said
I
know
what
is
for
is
Ives,
so
we
can
list
clearly
list
us
out
here
like
what
are
the
things
that
might
be
a
problem,
yeah,
I,
understand
or
definitely
I
think
that
was
widely.
We
study
this
a
lot
time
ago
and
kind
of
put
it
aside
for
a
while.
So
there
were
definitely
issues
but
yeah
I
think
we
just
need
to
list
all
of
those.
F
I
I
A
A
I
I
G
A
In
in
cinder
we,
when
we
implement
this
one,
we
did
implement
this
one
for
a
suti
system
that
does
not
have
like
consistent
group
snapshot
right.
So
if
they
don't
really
have
this
concept
this
to
a
system,
but
we
allow
them
to
kind
of
group
things
together.
So
it's
mostly
the
information
would
be
stored.
The
cinder
layer
would
have
something
it's
just
a
and
yeah
I
think
that,
should
you
should
work,
because
that
incriminated
is
in
kubernetes
database
right.
I
A
A
C
G
G
They
don't
have
to
be
group
to
restore
time,
but
if,
if
you
want
to
end
up
with
a
a
new
group
of
things
that
you
can
again
take
a
consistent
snapshot
of,
then
you
might
be
required
to
tell
the
system
that
at
the
time
you
do
the
restore
otherwise
the
result
of
the
restore
won't
be
something
you
can
turn
back
into
a
group.
So.
A
G
G
D
A
So
you
know
you
wouldn't
know
at
this
time
right.
So
you
have
your
snapshot.
Have
a
group
snapshot
names?
You
know
what
a
snapshot
it
you
need.
Then
you,
basically
just
you
create
a
voting
group
first,
then
you
create
a
volume
from
a
snapshot.
Add
that
to
the
group.
So
you
do
this
one
at
a
time
until
you
got
to
all
of
them
all
of
them,
whistled
all
of
them,
but.
C
A
A
A
A
A
G
So
so
I
expect
I
guess
I
expect
there
will
be
implementations
of
this,
where
the
the
group
snapshot
will
be
implemented
by
basically
co-locating
all
the
data
for
all
the
volumes
on
one
thing
and
taking
one
giant
snapshot
of
it
in
such
a
way
that
it
can't
be
restored
piece
by
piece.
But
of
course
the
driver
can
pretend
to
restore
things
piece
by
piece
by
you
know
when
the
first
restore
is
requested,
restore
the
whole
group
and
then,
as
other
restores
requests
that
you
could
just
provide
pointers
to.
A
G
A
A
G
G
G
C
Is
the
hardest
challenge?
Is
that
we're
gonna
have
to
support
a
bunch
of
different
storage
systems?
Some
storage
systems
may
allow
you
to
create
an
empty
volume
group
and
then
add
in
volumes
at
a
time.
Others
may
only
allow
you
to
add
empty
volumes,
and
there
may
be
some
storage
systems
that
don't
allow
you
to
create
an
empty
volume
group.
They
say
you
must
specify
you
know
five
volumes
at
creation,
time
together
and
that's
it.
So
we
have
to
think
through
those
use
cases
and
then
on
the
snapshot
on
the
on
the
snapshot
restore
side.
C
We
have
the
same
problem
of
you
may
have.
Storage
systems
like
Dave
was
mentioning
where
they
don't
really
have
a
concept
of
restoring
by
group
ID.
They
just
create
a
bunch
of
individual
volumes
and
they
would
have
to
figure
out
how
to
synthesize
a
group
ID
or
the
ones
that
you're
mentioning
been
where
they
don't
have
individual
snapshot
IDs.
They
only
have
a
group
ID,
and
so
we
need
to
decide
whether
we're
going
to
support
all
of
these
use
cases.
B
A
So
yeah
yeah,
that's
a
good
point,
so
we
could.
So
those
are
the
things
that
we
probably
choose.
The
certain
thing
we
pollinates
you
like
say:
how
do
we
prevent
those?
If
you
know
they,
we
don't
want
them
to
delete
one
snapshot
from
a
group
snapshot
and
you
know,
then
you
could
cost
the
whole
course
I'm
sure
to
be
invalid
right.
So
we
need
to
think
about
how
to
disable
that
not
sure
it's
a
we
need
a
web
cook.
The
whole
thing
mm-hmm
so
yeah.
So
things
like
that
at
me.
We
I.
B
Mean
if,
if
direction
is
that
women
creative
woman
group
snapshot
where
some
newly
Agassi
as
I
call
I,
would
say,
including
one
snapshot
as
in
in
the
media,
where
he
doesn't
make
sense,
because,
because
eventually,
the
won't
group
snatcher
I
should
be
managing
all
this
and
exposing
those
two
and
user
seems
to
be
a
lipid
too
much.
If
is,
on
the
other
hand,
we
dedicate
individual
snapshot
creation,
then
yeah.
It
makes
sense.
We
just.
A
That's
how
actually
your
well,
so
if
you
look
at
the
this
case
that
we're
actually
creating
one
at
a
time
is
not
like
either.
So,
if
you
look
at
this
we're
actually
either
we
are
adding,
because
this
symbol,
PVCs
or
we
are
I-
owe
ya
doing
a
snapshot
not
talking
about
the
wall
in
the
group
right
it
took
about
on
you.
The
snapshot
groups.
Nash
up
are
not
here.
A
Part:
okay,
yes,
so
I
think
that
also
is
tied
to
we
need
to
decide.
How
are
we
going
to
do
restore
if
we
can
only
restore
by
the
time,
then
we
do
need
to
have
those
right.
Otherwise,
maybe
we
can
look
at
you
know
if
we
only
need
just
for
this
one
type
of
restore
just
you
know
we
have
just
restore
from
this
group
snapshot.
We
don't
really
need
to
worry
about
the
individual
pieces,
then
maybe
we
don't.
We
can
hide
those,
but
also
I,
think
for
maybe
for
other
drivers.
They
couldn't
really
support
that.
A
B
A
Mean
yes,
I
think
that
same
thing
we
need
to
decide.
I
know
those
yeah,
that's
actually
pretty
yeah
yeah.
Those
are
the
that
that
are
either
weird
right.
So
you
really
suppose
you
use
an
either
screwed
snapshot,
but
then
you
also
have
those
you
need
to
snapshots
that
are
showing
as
resources
and
how
are
you
going
to
protect
them
from
being
deleted
by
accident
right?
So
those
are
the
things
that
we
knew
needed
is
we
need
to
decide
how
to
what
I
do.
A
A
Okay,
so,
and
with
so
for
group
snapshot
that
we
have
status,
that
we
have
this
ready
to
use
flag
and
also
we
have
a
list
of
bonus
snapshots,
so
in
volume
snapshots
that
we
have
the
group
snapshot,
name
and
then
the
here
so
I
have
this
slice,
because
if
we
want
to
support
both,
you
want
to
do
is
this
just
for
both
the
useless.
You
take
a
good
snapshot
and
placement.
Then
then
this
could
be
in
you
know,
different
groups,
that's
why
this
is
a
string.
This
is
a
like
array
means
that
just
one
name.
A
So
so
this
is
the
llamó
example.
So
if
we
create
a
so
this
is
the
group
class.
So
it's
a
group
snapshot.
So
this
is
supports
group
snapshot.
So
we
have
the
statute
set
to
true,
and
then
this
is
the
warning
group.
You
lost
you
this
class,
so
this
basic
says
that
okay
want
to
add
all
of
those
volumes
that
matches
the
label
in
to
this
warning
group
and
that's
that
plans
to
this
one,
a
group
class
one
in
a
good
snapshot.
We
just
specify
the
source.
A
Is
this
warning
group
and
and
a
PVC
belongs
to
this
rolling
group?
We,
so
we
all
have
this
group
name
here
and
then,
if
we
also
do
that
for
group
for
the
policemen's,
then
in
the
in
this
Placement
class,
that
would
have
this
topology
is
and
then
in
the
group
we
specify
the
placement
group
class
and
then
the
PVC,
then
you
know
it's
possible
for
it
to
belongs
to
both
the
motor
groups
for
group
snapshot
and
also
placement.
Then
in
our
case
it
belongs
to
two
different
groups.
A
A
A
C
C
Types
of
storage
systems,
so
what
I
did
was
add
a
item
to
tomorrow's
SIG's
storage
called
to
talk
to
vendors
about
this?
Okay,
we
can
say:
hey,
you
know,
we've
got
this
kept
in
order
for
this
to
work
for
everyone.
We
need
to
kind
of
know
what
works
for
you
and
we
can
start
getting
their
input.
Yeah.
A
H
H
H
C
H
D
H
H
K
H
To
end
picture
of
a
particularly
use
case-
mm-hmm,
okay,
and
that
way
it
also
gives
us
the
ability
to
be
like
you
know,
for
the
first
phase,
we're
gonna
tackle
this
this
one
or
two
use
cases
for
the
second
phase.
Here's
like
ideas
we
have
for
how
to
address
the
other
use
cases
but
like
we
don't
have
to
you,
know,
really
get
iron
out.
The
API
and
you
know,
get
the
details
for
these
use
cases.
Yet
mm-hmm.
D
K
Tag
the
volumes
dynamically
with
the
storage
group
do
they
need
to
be
aware
of
which
drivers
support,
which
method
or
is
the
expectation
that
all
drivers
will
support
both
methods,
which
we've
we've
heard
some
kind
of
fit
with
one
or
the
other,
so
I'm
at
Dell,
and
we
have
like
five
different
systems
and
they
do
things
slightly
differently.
My
concern
is:
what
is
the
user
experience
when
he
tries
to
provision
against
yes.
A
Yes,
I
think
that's
something
that
we
need
to
discuss
right.
So,
if
your
drivers,
some
drivers,
can
only
support
one
way,
the
other
driver
or
can
support
here
in
a
way.
Is
that
what
you're
trying
to
say
or
can
they
do
things
the
same
way,
I
think
all
your
drivers
support
the
same
kind
of
weight?
You
create
a
what
a
group
well.
A
A
I
also
think
that
you
know,
of
course,
then
decides
that
you're,
the
the
PVCs
that
is
selected
has
to
be
correct.
Right,
if
you
say
it's
a
consistent
group,
but
if
they
are
not
on
the
same
tour,
maybe
your
system
cannot
support
because
it's
a
good
place.
So
maybe
that's
a
problem
so
so
I
think
this
one,
the
first
one
would
be
harder
it
could
you
know
they?
Could
it
get
me
wrong
depending
on
what
your
support?
But
if
you
just
want
to
group
them
together,
not
really
saying
I'm
guaranteeing
consistency,
then
that's
part.
A
G
Also,
the
concern
that,
no
matter
what
we
design
like
the
people
who
consume
this
API,
were
the
applications
that
consume
this
API
may
prefer
one
or
the
other
mechanism
to
the
inclusion
of
the
other
one
and
that's
gonna
drive
what
people
actually
implement
right.
If
we
design
this
to
give
you
two
options,
but
in
practice
everyone
uses
just
one
of
them,
then
vendors
are
going
to
ignore
the
other
one,
because
yeah
so.
A
L
D
A
K
D
L
A
I
A
L
Had
one
last
question
before
share
with
us,
so
we
are
also
adding
a
support
to
add
any
previously
in
their
group
dynamically
right
depending
on
the
leader,
because
just
to
start
with,
you
might
have
created
that
volume
group,
but
later
on,
you
added
more
volumes
through
the
particular
application.
So,
yes.
A
D
A
L
A
L
Example,
I
can
give
I
suppose
you
have
one
application
for
that.
You
are
using
three-volume
the
street
or
they
start
doing,
and
you
have
some
replication
policy
which
just
creates
a
group
with
all
the
PVCs
being
part
of
that
group,
and
you
take
a
snapshot
consistent
set
of
that
group,
and
then
you
replicate
to
the
DSi
now.