►
From YouTube: WG Data Protection Bi-Weekly Meeting
Description
WG Data Protection Bi-Weekly Meeting
A
A
Okay,
so
we
want
to
make
high
level
discussions
on
what
it
means
to
do.
You
take
protection
and
how
to
achieve
that,
and
also
we
want
to
have
some
more
detailed
design
discussions,
specific
specific
things
that
we
want
to
do
to
fill
those
gaps
in
the
protection
support
and
we
want
to
create
documents,
make
them
available.
So.
A
A
Okay,
I'll
just
go:
go
over
the
the
topics,
potential
topics
right
so
here
illicitly
there
are
potential
topics,
meaning
they
could
change.
There
could
be
more,
they
could
be
less
right
based
on
what
we
think
are
needed.
So
the
first
way
is
three
data
for
an
snapshot
without
creating
a
new
volume.
So,
right
now
we
do
have
snapshot
functionality.
We
can
create
a
new
volume
in
front
of
snapshot,
but
here
what
it
says,
okay,
is
it
possible
just
to
read
from
the
data
type
from
the
snapshot
here?
A
A
So
would
you
have
what
in
snapshots,
but
that
you'll
a
mean
you
take
a
snapshot
and
your
snapshot
is
under
local
storage,
together
with
your
volume
for
some
cloud
providers
that
may
be
different,
but
in
general
that's
the
case.
So
what
in
backups,
meaning
that
you
back
it
up
to
a
different
device
which
is
separate
from
your
primary
storage
and
data
populate
er?
So
this
is
something
that
has
been
discussed
in
six
storage,
miss
lis.
B
A
A
Yeah
this
is
someone
for
all
this
Apple
I
think
we
don't
really
have
any
design
for
this
one.
So,
basically
so
currently
you
could
create
a
new
volume
for
us.
So
that's
like
a
whole
volume
right
and
then
this
is
the
wrist
data
from
snapshot,
meaning
that
you
can
actually
just
do
actually
like.
You
can
mount
a
snapshot
and
the
read
from
it
it
directly.
So
you
don't
have
to
you,
go
create
another
volume
from
it
make
it
faster.
The.
A
Backup
this
is
for
backup.
Yes,
of
this
is
whole
everything
is
related
to
backup
right.
So,
if
you
do
so
right
now
to
do
a
back,
you
have
to
you
have
to
create
a
foot
action,
restore
a
backup
and
restore.
She
say
when
you
do
a
way
to
do
a
restore,
you
will
have
to
create
a
you.
You
will
have
to
create
in
your
volume
when
you
do
a
backup.
You
also
have
to
create
new
water
right
now.
I
think.
B
A
A
B
A
C
C
A
Yeah,
so
this
first
one,
okay,
so
probably
should
expand
this
one
yeah
so
yeah
we
could
we
well
potentially
there's
no
designer,
but
potentially
could
you
know
monk,
that's
not
shot
right
and
then
make
that
available.
Then
we
could
you
either
read
your
at
the
firm
for
the
whole
thing
or
part.
You
know,
guess
that's
the
this
case.
A
A
Next,
one
is
consistency
following
groups
the
group
snapshot,
so
this
also
is
something
that
we
have
been
working
in
six
or
so
I'm,
actually
working
on
a
cap,
this
one
and
allow
you
to
a
group
volumes
together,
mention
them
together
and
take
a
snapshot
of
this
multiple
volumes
instead
of
just
one
and
the
next
one.
Application
snapshot
back
on
recovery,
so
icd-9
is
also
on
a
car.
He
has
a
cap
for,
for
this,
allow
you
to
snapshot
a
whole
application
and
recover
it
and
I
think
it
was
also
back
up
back
apart.
A
A
A
The
next
one
data
protection
policy,
so
we
don't
really
have
any
design
for
this
one,
but
thanks
somebody
brought
this
up
and
also
at
some
time
ago.
Jing
actually
did
some
POC
of
this
using
the
metadata
controller
to
do
periodic
snapshots.
So
the
securing
you
know,
periodic
backups.
You
can
schedule
for
that
or
set
retention
policy
to
automatically
clean
up
the
old
backups
topology
to
set
about
you
specify
backup,
location
or
it
could.
You
could
include
other
things.
A
F
A
So,
nor
is
a
I,
can
I
can
add
a
link,
Oh,
actually
I.
Don't
think
we
have
anything
that
that
is
really
really
clearly
written
yet,
but
I
think
there
is
a
Google
Docs
somewhere.
Basically,
the
idea
is
to
allow
allow
you
to
use
some
external
data
source.
So
currently
you
can
use
either
a
volume
snapshot
or
a
PVC
as
a
data
source
to
create
a
to
create
a
new
PVC.
This
process,
that's
also
okay,
now
to
allow
and
allow
for
other
sources
like.
If
you
have
a
s3,
you
have
an
s3
bucket.
A
C
I
guess
also
a
key
thing
here
is:
the
existing
mechanism
requires
a
change
to
the
core
kubernetes
data
type.
Every
time
you
would
have
a
new
data
source.
The
idea
here,
I
think,
is
that
you
introduce
an
intermediate
level
of
a
thing
which
is
sort
of
a
data
populate
err,
and
then
you
could,
in
direct
the
data
populate
err
to
have
lots
of
different
sources
for
the
data,
so
it
allows
for
an
evolution
path.
It
doesn't
require
changing
the
PBC
definition.
Every
time
you
want
to
add
a
new
data
source
yeah.
A
Yeah
I
think
this
led
yeah
I
think
the
challenge
is
mainly
that
I
think
the
the
API
reviewer
I'm,
not
sure
at
least
that
was
you
in
the
beginning,
when
we
added
the
snapshot
supported
by
we
added
that
up
first
source.
So
we
asked
you
make
it
super
super
strict,
but
you
can
only
allow
this
one
source
and
then
next
time
we're
talking
about
for
cloning,
which
is
TVC
source,
so
it
so
we'll
see
what
were
they
saying
so.
B
B
Yeah,
so
so
the
the
required
changes.
You
stop
validating
that
field,
and
just
say
you
know
what
put
whatever
you
in
there
and
and
if
it's
a
volume
or
a
snapshot,
the
appropriate
CSI
sidecar
will
handle
it
and
if
it's
something
else,
the
appropriate
something
else
will
handle
it.
And
if
it's
something
you
know
what
he's
ever
heard
of
it'll
just
sit
there.
Nothing
will
happen.
A
G
A
B
A
A
A
Basically,
there
are
things
that,
because
we
are
like
cross,
sick
abs
and
on
the
sixth
orage,
but
there
are
topics
that
are
very
specific
in
the
authority
that
we
may
not
be
interested.
Like
am
I,
agree,
Michael
entry
drivers.
She
sees
a
driver's,
that's
gonna,
be
very
specific
topic
for
data
protection
or
like
see
gaps.
They
are
doing
some
design
on
api's.
Then
you
know,
then
we
we
may
need
to
make
some
changes
there,
but
the
main
design
will
be
owned
by
mystic
apps.
A
So
things
like
that
and
also
I
this
one
I
want
to
make
sure
everyone
understand
that
the
working
group
does
not
own
code
a
mistake
on
code,
so
we
will
have
discussions.
We
will
do
design
discussions
under
cat.
We
can
can
discuss
caps
and
review
caps,
but
the
cab
approval
and
the
code
implantation
that's
still
owned
by
sick,
absent,
sick
story.
G.
A
We
can
talk
about
that,
but
it's
only
by
the
six
and
then
the
the
third
one
is
that
this
is
actually
something
came
up
during
the
the
review,
which
are
we
trying
to
form
this
working
group.
There
appears
to
be
some
confusion
on
the
definition
of
data
protection,
because,
if
you
search
in
Google
Google,
then
then
you
will,
you
will
see
that
it
talks
about
not
only
backup
and
recover
aspect
of
it,
but
it
also
talked
about
this
security
and
privacy
part.
A
There
are
all
stakeholders,
also
under
the
disband
the
criteria
that
we
are
required
to
put
in.
They
are
basically
saying:
okay,
we
need
to
working
on
all
the
things
that
are
we
listed
in
the
in
scope
section
and
we
need
to
produce
some
documents
and
then,
if
we
completed
everything-
and
there
is
nothing
more
to
discuss
and
investigate
them,
then
at
that
time
we
polish
it
just
disband.
H
A
A
C
C
D
D
The
set
up
of
this
group
because
we
had
there
were
definitely
things
that
didn't
work
for
VMware
and
I
want
to
make
sure
we
had
as
many
different
inputs
in
here
as
possible,
and
so
that's
why
we
brought
in,
for
example,
all
the
data
protection,
vendors,
a
Veritas
team,
etc
del
EMC
to
start
having
some
input
on
this
as
well.
Yeah.
C
I
guess
I
guess
the
only
the
reason
why
you
hear
that
hesitation
in
my
voice
is
at
this
point,
I
mean
the
beta
API.
Now
has
that
has
the
the
beta
association
right.
So
we
can't
like
throw
it
out
and
do
something
new.
Are
you?
Are
you
talking
like
we're,
gonna,
throw
it
out
it
gonna
do
something
new,
but
if.
A
A
A
That's
okay,
yeah,
so
that
all
we
could.
We
could
also
talk
about
that
time.
So
I
have
some
ideas
because
there
are
we
got
feedbacks
right,
but
I
don't
think
we're
going
to
do
like
a
dramatic
change,
but
maybe
fix
some
of
the
of
the
box
or
some
of
those
are
designed
that
way.
So
we'll
just
look
at
those
and
see
if
there
are
any
solutions,
but
not
that
not
necessarily
have
to
be
the
focus,
because
there
are
a
lot
of
topics
here
right.
D
G
D
B
A
So,
let's
just
go
over
those
topics
because
I
know
some
of
those
already
have
owners
right.
Others
have
not
right.
So,
let's
just
I'm
going
to
go
over
just
a
I.
Just
look
at
this
that
I
know
or
I
have
owner
so
I'm,
currently
working
on
warning
groups
and
then
I
know:
inna
is
working
on
application,
snapshot,
backup
and
recovery.
So.
A
B
C
Know
yeah
I
would
love
to
because
one
of
the
things
that
strikes
me
is
that
we
have
two
kinds
of
things
here:
things
where
multiple
folks
will
have
to
write
to
a
standard
and
things
where
there
is
a
separate
functionality
that
can
be
driven
kind
of
by
one
and
the
volume
backup.
One
strikes
me,
as
kind
of
like
volume
snapshots
it's
going
to
need
lots
of
sort
of
working
out
of
how
this
might
work
without
necessarily
kind
of
the
storage
level
implementation
that
makes
sense,
and
so
that's
one
where
I
think
we
can.
C
A
A
A
C
We
can
start
with
some
principle
statements
and
yet
make
sure
we
all
agreed
on
on
some
basic
principles,
and
then
we
can.
If
somebody
then
has
api's
that
meet
all
of
those
principles.
We
can
put
those
forward
and
discuss
them.
We're
gonna
argue
on
some
of
the
principles
and
rather
than
call
away
in
the
API,
we
can
argue
on
the
sort
of
some
of
the
more
fundamental
things.
Yeah.
B
C
So
I'm
certainly
willing
to
draw
up
a
little
dock
with
my
our
view
of
sort
of
some
of
those
principles,
and
maybe
we
can
as
a
discussion
kind
of
go
through
what
several
folks
provide
for
that,
because
I'm
assuming
other
folks
would
be
able
to
do
the
same
for
like
the
next
meeting.
Maybe
that
sounds
good.
B
Yes,
I
only
mentioned
the
data
populated
stuff,
because
I
am
you
know,
trying
to
propose
a
cap
for
the
kubernetes
1:18
time
frame
with
with
my
ideas
and
I.
Don't
know
how
far
we'll
go
but
I
don't
want
to
wait
and
while
the
perfect
to
be
the
enemy
of
the
good,
so
I'm
trying
to
pursue
that
right
now
before
the
1:18
deadline.
B
C
C
B
B
A
B
A
H
Sir
sorry,
that
the
idea
Ben
is
that
this
whole
big
working
group
is
going
to
meet
bi-weekly
where,
as
a
specific
domain
area,
for
example,
William,
backup
or
Ted,
a
popular
or
Warren
group
feel
free
to
schedule
as
many
minutes
as
he
wanted
right.
We
just
don't
want
to
call
this
like
30
plus
engineer
or
people
or
PMS
meeting.
You
know,
won't
be
home
and
discussing
everything
right
and.
B
D
A
So
normally
the
one
group
yeah
it's
like
this
right,
so
you
have
something
you
have
to
work
off,
not
like
everything
will
be
discussed
any
in
this
bi-weekly.
But
if,
if
even
there's
the
deadline
you
you
have
to
have
additional
meetings,
just
just
sort
out
with
those
issues
right-
and
it
happens
like
this-
is
the
first
meeting
we
have
so
many
things
listed
here
but
doesn't
mean
like
everything
we
can
do
everything
in
a
month.
That's
not
possible
right,
so
we.
F
C
H
A
Yeah
yeah
definitely
I
also
I,
think
in
this.
In
this
meeting
we
can
also
have
time
for
design
discussions
if
we
have
time
right.
So
that's
normally
I
see
in
other
I
think
every
single
work
differently
so
like
here.
We
can
also
have
a
section
for
design
discussions
like
in
six
storage
in
the
written
section
for
that
by
usually
report.
I
don't
have
much
time
because
in
that
meeting
went
over
status,
but
here
we
we
can.
A
A
D
H
A
That's
that's
this!
That's
like
this!
Well,
okay,
yeah,
it's
not
so
that
would
be
be
under
voting
backups
if
you
talk
about
incremental
backups.
So
when
you
do
backups
and
full
backup
and
was
incremental
backup,
so
it
should
be
a
sub
topic
underwater
backups.
So
here
we
have
this
a
diffs
between
two
snapshot
like
that
you
can
probably.
C
G
Moving
this
okay,
let
me
just
explain
that
push
in
a
little
bit
further.
So
when
I
look
at
this
list,
I
see
some
of
the
items
that
may
not
require
API
level
changes.
For
example,
the
volume
backups
is
such
a
generic
domain
that
even
today,
people
are
doing
volume,
backups
right,
they're,
just
wanting
the
filesystem
and
copying
the
file,
so
you
have
volume
backups
in
some
sense,
you
don't
really
need
any
API
level
support
strictly
speaking.
G
C
Definition
of
volume,
backups
and
mine
are
different
because,
yes,
by
volume,
backups
certainly
I
did
not.
I
did
not
intend
to
mean
copying
the
full
copy
of
the
of
the
data,
as
seen
from
the
from
the
client
OS
right,
it's
more
like
the
like
a
snapshot
but
stored
outside
of
the
storage
pool,
so
inherently
incremental,
inherently
done
ideally
more
efficiently.
You
know
with
with
some
sort
of
diff
calculation.
You
know
all
of
the
sorts
of
things.
D
I
mean
we've:
we've
been
looking
at
this
quite
a
bit
on
the
VMware
side
and
with
our
backup
partners,
because
our
but
our
traditional
backup
workflows,
have
been
take
a
snapshot
and
then,
but
that's
strictly
a
local
snapshot
that
doesn't
go
anywhere
except
local
storage
and
then
the
backup
application
goes
ahead
and
extracts
the
data
and
puts
it
in
its.
You
know
it's
storage
system
and
then
deletes
the
snapshot.
D
G
Okay
I
mean
even
even
accepting
this
right
that
the
principles
would
clarify
this
volume
backups
and
things
like
that.
I
still
believe
there
is
a
clear
distinction
between
the
API
level
work
as
opposed
to
more
MORE.
Let's
call
just
you
know,
probably
backup
provider
level
work.
Consistency
groups
definitely
require
support
from
CSI,
so
just
to
be
very,
very
narrowly
focused.
We
can
ask
what
are
all
the
things
that
CSI
spec
needs
to
support,
and
you
know
start
with
those,
for
example
the
consistency
groups
and
things
like
that.
Yeah.
C
D
C
D
Kind
of
what
we
meant
by
reviewing
the
existing
api's
was
to
go
through
and
look
at
them
in
terms
of
you
know
as
many
different
eyes
as
many
different
workflows
and
use
cases
to
say
this
is
what
I
and
things
like
consistency
groups.
You
know
we've
been
talking
about
them,
but
it's
like
what
scale
of
consistency
group
do
you
need?
You
know,
building
out
a
large
scale.
You
know
multi
machine,
multi,
storage
system.
Consistency
group
is
hard
and
it's
cool,
but
will
one
actually
solve
a
problem
and
that's
something
we
need
to
understand.
B
D
B
Sympathize
with
with
the
person
who
said
we
should
focus
on
real
CSI
API
is,
but
it's
it's
not
like.
We
haven't
thought
of
this
down
at
the
CSI
working
group
before
that.
The
problem
is
that
we
couldn't
figure
out
what
it
meant
or
how
to
do
it,
and
so
I
think,
maybe
when
the
outputs
of
this
group
could
be
a
clearer
goal
for
for
the
CSI
working
group
to
actually
go
go
do
because
we
know
that
it's
it's
a
desirable
thing.
We
just
don't
know
how
to
do
it,
so
we.
A
Need
to
know
the
whole
workflow
right,
so
it's
we
can't,
we
can't
do
bottom-up,
we
need
to
know
so.
How
do
we
get
this
to
work?
What
are
the
API
needed
in
qualities,
and
what
are
we
needed
in
series
that
we
need
to
look
at
this
whole
things
so,
like
forwarding
group,
so
we
need
to
post,
we
will
need
the
communities,
API
changes
and
also
CSI
changes.
A
So
yeah,
so
that's
definitely
very
important.
That's
like
the
high
level
thing.
Definitely
that
that
is
very
important
right.
So
that's
actually
why
it's!
If
you
look
at
the
in
scope
right,
so
that's
the
first
right.
So
it's
like
understanding
the
the
high
level
how
to
do
things
and
then
those
are
or
individual
topics,
but
of
course
each
of
them
could
be.
There
are
big
guys,
Wow
right.
H
Yeah
I
just
want
to
add,
on
top
of
what
Angela's
saying
when
we
say
a
woman
back
up
yet
traditional
morning,
back
up
me
as
what
they
was
describing.
I'm
sorry
idea,
the
percent
on
your
name,
sorry,
but
we
also
need
to
think
about
how
do
we
effectively
can
restore
the
backup
into
a
warning
right?
It's
not
a
one-way
thing
right.
You
can
take
as
many
backup
as
you
wanted
about.
How
do
you
even
cope?
All
this
stuff
into
a
kubernetes
native
construct
is
and
very,
very
important
thing.
H
A
Commenting
a
chat,
I
mean
a
list
or
workflows
and
fun,
so
so
that
yeah,
so
there
are
different
levels
of
backups
of
their
protection
right
here,
so
that
will
be
like
different
levels
of
workflows.
Right,
so
backup
by
individual
worrying
will
backup
application
back
up
at
cluster
level
is
all
very
different.
D
One
comment
I'd
have
at
this
point:
I'm
at
the
group
level
is:
do
we
have
the
right
set
of
participants,
it
looks
like
we've
got
storage
vendors
and
some
DP
vendors
and
I
was
hoping
or
thinking
that
it
would
be
good
to
get
application
or
database
vendors.
You
know
the
Cassandra
type
folks
involved
as
well
cuz
a
lot
of
these
things.
You
know
we
have
to
ask
them.
How
are
we
going
to
effectively
backup
your
application?
D
E
I
did
some
research
mostly,
for
example,
for
a
moment,
a
B.
How
do
you
actually
like
inflict
on
the
way
you
want
to
backup
a
MongoDB
cluster?
So
first
thing
you
you
have
to
do
is
to
pick
the
right
replicas
to
snapshot
or
or
backup.
So
there's
a
lot
of
application-specific
logic
seems
that
these
kind
of
scenes,
so
we
have
to
figure
out,
and
so
we
don't
have
currently
have
a
standard,
API
or
controllers
for
doing
these
kind
of
things.
E
C
The
interesting
question
is
whether
we
go
the
approach
that
we
have
pursued
so
far,
which
is
to
say
these
products
all
have
commands
for
doing
backup
or
for
preparing
for
a
backup,
and
we
all
basically
provide
hooks
for
tickling
those
limits.
I
think
their
third
approach
might
be
actually
saying.
Are
there
some
primitives
that
we
can
expose
that
are
new,
that
these
guys
might
sign
up
to
using
I''m
and
I'm
reaching
here
but
I'm
thinking
of
sort
of
like?
What's
the
windows
know
nobody
at
this.
D
E
There
there
are
two
paths
here:
one
is
the
basic
features
rely
on
file
system
level,
backup
and
but
you
have
to
know
how
to
actor
pick
the
right
replica
to
backup
in
the
first
place.
The
second
is:
how
directions
before
support
using
application,
specific
tools
like
a
know,
for
example,
dump
or
or
much
my
sequel
dome.
So
these
are
two
different
tracks
and
I.
Think
yeah.
D
And
I
think
there's
very
much
the
topic
of
is
the
application
involved
in
a
backup
or
other
cases
where
the
application
is
not.
You
know
we
right,
we've
done
crash,
consistent
backup
for
a
while.
So
what
you
know
is
that
valid
is
a
dollar
that
scale.
Is
it
something
that
you
know
our
current
tools
can
support?
Because
right
now
you
know
we
don't
have
a
consistency
group
that
goes.
You
know,
multi,
pod
or.
D
G
D
A
C
J
C
G
A
A
A
A
A
Yeah
so
yeah
so
so
start
with
your
design,
doc
in
a
Google,
Doc
Google
Doc,
and
then
you
can
send
that
through
the
slack
and
also
an
email,
then
people
can
start
adding
adding
those
comments
on
the
on
the
dock
itself.
So
that
will
be
at
least
that's
saved
or
an
email.
Sometimes
it's
hard
to
find
them
and
then
then
we
can.
A
C
A
A
C
Like
that,
your
proposal
is
basically
what
you
described,
which
is
to
allow
basically
unvalidated
CRD
references
in
the
data
source.
Yes,
have
you
talked
with
any
of
the
folks
that
are
going
to
be
doing
the
approval
on
that
kept
ahead
of
time,
because
that
that
is
likely
to
either
be
a
dead
in
the
water
or
a
reasonable
approach.
It
doesn't
seem.
K
C
B
Yes,
so
those
are
the
two
outcomes
that
I
am
anticipating
and
no
I
haven't.
If
you
have
the
names
of
the
people,
that
I
could
maybe
fill
them
out
or
give
them
a
preview,
but
I
think
that
maybe
any
reticence
that
we've
had
to
that
approach
in
the
past
has
just
been
around
not
knowing
what
the
implications
of
that
would
be.
So
I
want
to
have
a
POC
that
can
demonstrate.
This
is
not
as
scary
as
it
might
seem.
It
actually
works
perfectly
well.
Try
to
get
people
more
comfortable
with
the
idea.
Yeah.
A
C
A
J
A
A
No,
no
I,
don't
think
this
one
well
getting
to
you.
Why
not
I
think
I,
don't
don't
believe
so
it's
going
to
take
a
long
time,
but
I
didn't
just
saying
that
I
I
need
you
put
together
cap,
but
we
do
have
with
some
Google
Doc.
It's
been
there
for
a
long
time,
so
it's
probably
doesn't
help
you
just
to
go
through
that
again.
Maybe
just
let
me
just
put
together
what
happened
and
we
can
look
at
it.
Yeah.
G
A
Your
next
meeting,
and
so
we're
going
to
list
everything
that
we
want
to
go
through,
but
then
we're
running
out
of
time.
Then
we're
not
going
to
cover
them
right
so
and
also
depending
on
the
priority
deadlines.
If
you
want
to
get
something
right
now,
18
them
part
of
us
should
review
that.
First,
okay,.
A
Yeah
and
the
application
snapshot
that's
announcement.
Design,
of
course,
is
there
interesting
one
there's
already
a
cap
yeah
I'm,
not
saying
that
we
need
to
review
all
of
them.
You
in
my
meetings,
just
impossible
right
so
usually
like
we,
we
schedule
like
a
one-hour
review
meeting
for
one
cap,
and
it
couldn't
even
finish
like
I
was
doing
this
avoiding
on
the
house.
The
other
day
took
about
an
hour.
We're
halfway
through,
so
yeah
definitely
need
a
lot
of
time
to
go
through
those
things.
A
A
H
A
A
Should
be
so,
I
need
to
figure
that
out,
because
right
now
that
I
think
the
link
it
does
not
work
but
I
recorded
it.
So
I
need
to
ask
people
to
figure
out
how
to
get
this
one
post
you
somewhere
yeah,
but
we
should
never
figure
this
out.
Those
are
all
are
just
six
ones
for
me
now.
A
new
group
so
well
know
that
one.