►
From YouTube: Data Protection WG's Bi-Weekly Meeting for 20201202
Description
Data Protection WG's Bi-Weekly Meeting for 20201202
A
A
A
Okay.
So
let's
get
started
so
the
first
one
snapshot.
So
this
is
a
going
ga
in
1.20.
We
can
look
at
what
is
remaining.
A
So
this
is
the
project
tracking
board,
so
we
are
doing
well.
I
think
a
lot
of
things
are
merged,
so
one
major
thing
that
is
pending
is
the
matrix,
adding
metric
support
to
snapshot
controller,
but
there
is
a
pr
that
has
been
reviewed.
I
think
it's
getting
close
to
be
merged
and
then
there
is
also
this
update
of
the
controller,
so
that
one
is
also
ready
for
review.
A
So
I
think
those
are
the
major
ones
that
we
need
to
get
in
before
cutting
a
release,
and
then
there
are
a
few
other
things
that
we
can
do
after
that
there
are
a
few
other
things
like
like
e3
tests
for
secrets,
those
three
we
have
the
test,
but
we
can't
merge
right
now
because
it's
frozen
for
1.20,
so
we
can
get
that
merged
later,
but
the
test
itself
is
ready.
A
Yeah,
so
this
one
I'm
working
on
creating
basically
when
you're,
creating
a
pvc
visual
data
source
you're
using
an
entry
plug-in
that
should
fail.
So
that's
something
that
I'm
working
on,
but
that's
something
that
we
can
do
after
the
wenda
chanti
and.
A
Yeah,
I
think
the
we
just
got
the
the
third
rotation
for
webhook
appear
that
it's
just
merged
and
then
use
driver
name
in
pv,
that's
also
merged,
so
yeah
so
think
of
from
from
this
project
tracking
board.
It
looks
good
and
and
also
shanti,
and
I
are
working
on
a
release
block
so
yeah.
So
that's
the
full
volume
snapshot.
A
Looks
like
benny's,
not
okay,
yeah,
so
I
we
have
weekly
design
meetings
for
that.
I
think
ben
is
working
on
a
poc
that
is
in
progress.
He
has
some
pr
submitted.
A
A
And
and
then
cozy,
I
think
cozy
also
is
doing
fine.
There's
a
lot
of
progress.
The
team
is
busy
working
on
poc,
so
there
are
weekly
meetings.
B
A
Not
cluster
api,
so
this
is
a
so
right
now.
Basically,
we
have
the
cap
merge!
That's
so
right
now
it's
provisional,
just
provisional
inventor
20.,
so
1.21.
We
will
try
to
bring
that
to
alpha
so
in
this
release,
just
trying
to
get
some
poc
ready,
making
the
project
in
good
progress,
setting
up
the
ci
and
all
of
that
so
right
now
it's
all
the
crds
is
a
lot
of
tree.
I
think
in
the
future
there's
a
plan
maybe
to
try
to
bring
some
pieces
into
it,
but
that's
not
right.
A
Now
it's
it's
a!
We
have
a
separate
repost
like
we
have.
I
think
four
repos
under
krypton
86
for
the
cozy
spec
for
the
api
controller,
adapter
yeah.
A
Thank
you,
yeah
no
problem
and
so
content
notifier.
I
so,
I
think
shenzhen.
We
still
need
to
sync
up
on
this
one
and
and
then
we
will
need
to
get
a
few
issues
resolved.
Basically,
those
are
like
api
design
issues
how
we
are
going
to
do
a
status
update
things
like
that,
so
we
will.
We
will
work
on
that
and
what
in
group
also,
I
think
I
need
to
update
the
the
the
cap
again.
A
A
Progress,
okay,
so
yeah.
So
today
we
don't
have
like
a
particular
topic
here.
I
think
we
are.
We
had
some
discussions
on
cbt,
but
I
think
we
need
some
time
to
sort
out
a
few
things
before
we
bring
it
back
to
haven't,
have
a
discussion
in
this
in
this
meeting
so
far,
I
think,
maybe
in
the
next
meeting
we
should
be
ready
to
do
an
update
on
that
in
this,
in,
like
in
two
weeks.
C
We
can
we
can
drop
some
updates,
but
I
think
we
still
have
going
on
have
some
meeting
going
on
with
the
api
with
the
cbt
right,
yeah
yeah.
We
haven't
got
up
to
any
material
that
we
can
present
yet.
C
A
In
two
weeks,
yeah,
okay,
yeah,
so
this
is
the
general
document
that
has
what
other
things
that
we
are
working
at.
So
we
already
have
so
for
the
first
section
why
we
need
to
do
data
protection
communities.
We
already
have
a
draft
doc
use
case
section.
We
also
have
that
so
we
reviewed
those
already.
A
So
what
is
currently
available,
I
think
yeah.
So
we
we
already
have
in
information.
We
just
need
to
put
put
this
together,
so
sean
chan
and
I
need
to
fill
the
section
and
then
and
then
what
are
the
missing
building
blocks
so
so
for
warning,
backups
right
now.
We
I
think
we
do
have
the
stock,
which
was
produced
by
andrew
some
time
ago.
I'm
not
sure
if
shenzhen
is
this
like
elliot
or
someone
might
be
working
on
this
or
is
it.
A
Right,
yeah,
okay
and
then
backup
repositories.
We
also
had
a
couple
meetings.
We
do
have
a
working
progress
doc.
I
think
we
are
waiting
actually
waiting
for
some
conclusion
from
the
cpt
side
before
we
going
back
to
this
so
right,
a
proshanto,
I
think
that's
which
kind
of
pending.
A
I
was
saying
that
for
backup
repositories,
I
think
we
are
waiting
for
some
conclusion
from
the
cpt
side
and
then
we
can
go
back
to
this.
I
thought
that
was
the
last
time
when
we
talked
about
this
or
some
question.
A
With
to
cbt
last
meeting,
there
was
some
there
were
some
questions
coming
out
when
we
were
going
through
this
and
then
so
we
said:
okay,
let's
finish
that
away
and
come
back
to
this
one.
Do
you
think
there
was
a
that's
how
I.
E
Of
me
I'll
check
it
again,
I
think
from
content
content
perspective,
we
had
everything
and
I
think
you
know
the
broader
apis
or
the,
and
you
know
the
good
to
haves
or
the
enhancements
that
we
thought
we
spoke
about.
I
think
we
just
decided
to
park
it
for
now
and
then
we'll
come
back
to
it
once
you
know
once
we
can
finish
off
the
white
paper,
so
I
think
for
the
white
paper
from
the
white
paper
angle.
We
should
be
good
for
now.
A
Okay,
all
right
thanks,
yeah,
okay,
so
the
next
one
did
a
popular
thing.
We
already
talked
about
this
data
populator
and
the
choice
and
unquestions.
This
is
the
content
notifier.
So
we
do.
We
do
have
a
doc
for
this
one
as
well.
We
also
review
that
and
then
the
cap
is
review,
it's
being
reviewed
and
then
cbt
yeah.
So
we
just
talked
about
that.
So
we
we
still
need
to
sort
out
a
few
things,
especially
the
api.
A
I
need
to
figure
out
the
common
apis
before
we
come
back
and
discuss
more
in
this
meeting
and
then
one
in
group
and
group
snapchat.
So
I
think
we
just
mentioned
that
and
there's
a
cap.
I
need
to
update
that
for
our
application,
beckham
restore
craft.
We
also
have
a
doc.
I
think
we
did
review
that
last
time
and
then
there
was
also
a
cab
that
was
that
was
submitted
some
time
a
while
back
yeah.
A
So
I
think
those
are
all
the
things
that
we
are
working
on
for
for
this
white
paper.
Okay,
so
there
anything
else.
F
Beijing
for
the
application
workflow:
what's
that
what
is
left
there
do
we
do?
We
need
to
take
a
pass
to
polish
it
to
make
it
to
format
it
in
the
white
paper
style.
A
So
yeah,
so
I
think
so
I'm
going
to
because
we
already
have
other
information
kind
of
yeah,
just
yeah,
so
basically
just
need
to
put
them
together
so
yeah.
So
so
I
will
be
starting
doing
that
and
then
let
you
know
if
need
any
help
on
that,
okay
yeah.
So
I
think
we
actually
have
a
lot
of
materials.
We
just
need
to
consolidate
them
and
then
fill
in
the
missing
pieces.
A
Yeah
thanks,
so
I
see
ben
is
there,
so
I
you
want
to
give
a
quick
update
on
the
the
populator
we
were
talking
about
that
earlier.
G
Hey
shane
yeah,
sorry
I
I
joined
the
meeting
like
10
minutes
late.
G
What
I
missed
at
the
beginning,
what
about
the
data
populators,
do
you
want
to
know?
Oh.
G
G
Yeah
yeah,
so
we
have.
We
have
a.
We
basically
agreed
on
an
approach
to
have
a
crd
that
will
called
like
a
volume
populator,
which
will
be
the
what
populators
will
have
to
instantiate
to
basically
tell
the
system
that
there
is
a
another
crd
that
can
be
the
source
of
a
of
a
pvc.
So,
like
you
know
today,
the
only
two
things
that
can
be
the
sources
of
a
pvc
are
another
pvc
or
a
volume
snapshot
and
they're
sort
of
implicitly
allowed
as
we
go
forward.
G
The
idea
is
anything
else
that
can
be
the
data
source
of
a
pvc
will
need
to
have
an
instance
of
this
volume,
populator
crd
and
then,
and
then
there
will
be
a
controller
that
will
look
at
new
pvcs
getting
created
and
if
they
have
a
data
source
that
doesn't
match
any
of
the
known
valid
data
sources,
you'll
get
an
event.
That'll
tell
you
hey
this.
G
This
data
source
looks
invalid,
probably
nothing's,
going
to
happen
with
the
pvc
you
just
created
and
then
the
other
half
of
the
work
and
that
that
part
of
the
work
is.
I
have
a
working
prototype,
there's
a
pr.
G
The
other
half
of
the
work
is
we're
trying
to
produce
a
prototype
data
populator
that
contains
like
a
reusable
library
portion
that
all
populators
will
hopefully
want
to
leverage,
because
we've
identified
a
lot
of
complicated
sort
of
mechanics
that
have
to
be
followed
when
you're,
when
you're
doing
this
data
population
workflow
and
then
what
we
want
to
do
is
have
a
sample
that
you
know
says:
here's
a
library
that
anyone
can
use
here's
a
sample
of
a
very
simple
populator
that
uses
the
library
and
find
a
home
for
that
and
the
combination
of
those
two
things.
G
I
think
we
can
move
the
feature
to
beta.
That's
not
going
to
happen
in
120
121
is
what
we're
looking
at
is
the
best
case
beta
time
frame.
A
Okay
thanks
so
yeah,
so
I
don't
see
any
other.
H
Can
I
can
I
ask
a
question
about
the
data
populator
yeah,
of
course
go
ahead,
hi,
hi
ben,
so
there's
a
question
for
you.
I
I
went
through
the.
I
think.
It's
the
pr
or
the
proposal
document
yesterday
and
one
of
the
use
cases
mentioned.
There
is
obviously
backup
in
the
store.
So
I'm
wondering
if
you
can
talk
a
little
bit
about
the
restaurant
scenario.
I
think
that's
where
data
populators
might.
G
Yeah
yeah,
so
the
origin
of
the
whole
data
populator's
idea
for
those
who
haven't
heard
me
talk
about
this
before
was
in
a
world
where
you're
implementing
some
form
of
a
backup
and
restore
solution
on
top
of
kubernetes.
You
can
actually
do
almost
all
of
the
work
out
out
of
tree.
G
You
know
with
your
own
crds
and
your
own
controllers
and
your
own,
whatever
that
the
one
piece
of
the
workflow
that
unavoidably
touches
the
kubernetes
core
api
is
when
you
want
to
restore
a
backup
to
a
pvc,
you
know
if,
assuming
you
want
to
do
it
through
the
way
that
we
would
restore
a
snapshot
to
a
pvc
by
just
specifying
a
pvc
with
the
data
source,
which
is
the
logical,
it's
not
the
only
way
to
do
it,
but
it's.
This
is
the
most
obvious
and
natural
way
to
do
it.
G
In
my
opinion,
you
you
have
to
change
this
little
piece
of
the
of
the
core
kubernetes
api.
I
mean
it
turns
out
that
there's
a
lot
of
other
use
cases
outside
of
backup
and
restore
that
need
a
similar
ability
to
basically
create
a
new
pvc
that
is
not
empty,
but
contains
some
data,
and
you
know
where
that
data
is
defined
by
something.
G
And
then
you
know
today,
your
only
three
options
are
mtpvc
pvc,
that's
a
clone
of
another
pvc
or
pvc.
That's
a
clone
of
a
snapshot!
So
this
this
is
supposed
to
open
up
the
doors.
To
you
know,
to
a
whole
world
of
different
possibilities
for
pvc
data
sources.
You
know
we
could
have
gone
down
the
road
of
just
adding
backup
as
as
a
fourth
option
or
adding
you
know
other
ones
as
specific
enumerated
options
in
the
kubernetes
api,
but
that's
very
slow
and
cumbersome,
and
we
wanted
a
generic
way
to
do
this.
G
So
that's
what
we're
trying
to
do
here
with
the
data
populators
and
and-
and
I
do
anticipate
it
being
used
by
the
restore
workflow
of
some
backup
solution
that
we
ultimately
agree.
H
H
I
guess
the
the
volume
is
supposed
to
be
ready
for
immediate
use,
but
in
case
of
restore,
let's
say
the
data
source
says
some
kind
of
a
data
populator.
My
guess
is
that
the
data
is
going
to
be
restored
and
will
probably
take
some
time.
Is
that
the
understanding.
G
So
so
here's
the
key.
So
when
you
create
a
pvc
that
that
is
backed
or
that
is
the
clone
of
a
snapshot,
you
don't
know
how
long
it's
going
to
take
for
that
pvc
to
get
created,
I
mean,
even
when
you
create
an
empty
pvc,
it's
not
instantaneous.
It
takes
a
few
milliseconds
or
you
know,
however
long
it
takes
to
to
to
create
that
that
empty
pvc.
G
The
point
is,
is
you
can
create
your
pvc
and
you
can
create
your
pod
and
you
can
sort
of
tell
kubernetes
your
whole
workflow
and
then
kubernetes
will
hold
off
on
creating
the
pod
until
your
pvc
is
bound,
and
so
you
as
a
user,
then
you
no
longer
care
how
long
it
takes
for
the
pvc
to
get
created.
Maybe
the
snapshot
can
restore
in
one.
G
H
H
G
There's
a
several
different
pieces
that
all
come
together
to
basically
make
it
possible
to
implement
a
workflow
similar
to
like
the
snapshot,
restore
workflow,
but
for
any
crd
you
can
imagine
it.
If
we
had
done
this
work
before
we
had
done
snapshots,
we
would
have
implemented
snapshots
on
top
of
it,
but
we
had
it.
We
didn't
so
snapshots
sort
of
have
a
lot
of
custom
code
here
and
there
that
are
handled
by
you
know.
H
G
I
mean
I
want
to
be
careful
here
because,
like
backup
and
restore
is
one
kind
of
workflow
among
many
and
there's
other
population
workflows
that
have
nothing
to
do
with
backup
and
restore,
but
in
the
context
of
backup
and
restore
yes,
there's
a
question
of
like:
are
we
going
to
have
multiple,
different
kinds
of
backup
crds
with
multiple
different
implementations?
Maybe
like?
That,
would
be
kind
of
bad?
G
I
think,
but
it
may
be
what
happens
anyways
but
like
if
we
do
agree
on
like
there
is
one
backup
crd
that
defines
a
backup
and
there's
multiple
implementations
of
that.
Then
that
goes
yet
a
layer
deeper
in
terms
of
how
we,
how
we
handle
drivers
for
that
crd,
because
the
populator
will
just
say
oh
this
is
this
is
a
backup.
It
is
a
valid
thing
to
restore
data
from
so
I'm
going
to
say
it's
okay
and
then
there
will
be
a
some
sidecar
or
some
some
controller
somewhere.
G
D
G
G
H
H
A
It's
for
restore,
so
you
can
actually
use
a
combination
of
use,
one
snapshot
api
to
take
snapshot,
and
then
you
can
use
the
data
popular
to
do
the
restore
exactly
yeah
yeah.
If
you
just
use
crit
volume
from
snapshot,
then
if
that's
like
a
local
snapshot,
you
can't
really
that's
not
really
a
backup
right.
So
all.
H
A
But
I
mean
so
typically,
your
snapshot
is
probably
uploaded
somewhere
right,
so
unless
using
so
depending
on
your
driver,
I
mean,
depending
on
your
csr
driver,
if
some
css
drivers
upload
that
one
already
to
s3
right.
So
if
you
use
aws,
cps
or
gcpd,
that's
uploaded
already,
but
if
you
use
other
drivers
most
other
drivers,
it's
just
like
a
local
snapshot.
So
you'll
have
to
upload
that
your
backup
software
usually
will
upload
that
to
another
place
right
so
yeah.
A
A
To
yeah
right
yeah
to
to
provision
the
the
pvc
with
that
data
source
yeah,
so
so
the
backup
yeah
building,
backup
there's
another.
That's
that's!
We
were
talking
about
earlier
and
there's
this.
We
don't
really
have
a
cap
banner
there's.
This
talk
talked
about
voting
backups,
so
you
can
take
a
look
of
this
one.
A
I
think
elliot
and
alexis
are
working
on
this,
but
this
in
the
google
doc
has
some
ideas
on
my
work.
Yeah.
H
A
G
A
H
Yeah,
just
in
this
respect,
right
there
were
some
discussions.
I
mean
again,
as
I
said,
I
didn't
attend
for
a
few
weeks,
so
the
freeze
and
unfreeze
requires
an
unquiet.
A
A
So
yeah,
so
this
is
for
the
yeah.
So
that's
the
the
content
notifier
proposal,
so
so
that's
for
application
yeah
for
application
consistency
if
container
notifier
is
what
used
to
be
called
the
execution
hook.
Yes,
okay,
that's
from
yeah.
Initially
we
have
that
execution
hook
right.
So
that's
like
a
crd,
but
then
we
have
some
review
comments
saying
that
we
should
not
be
doing
that
so
so
now
we
have
this.
It's
like
a
now.
It's
like
a
in-line
port
definition.
A
So
it's
more
having
kubelet
doing
those
runners
command
instead
of
having
an
another
controller,
doing
it
so
yeah
so
that
that's
still
being
reviewed
so
we'll
try
to
make
some
progress
in
the
in
1.21.
We'll
try
to
make
some
progress
on
that.
H
A
Yeah,
so
that
one
has
yeah
so
yeah,
that's
this
one
that
this
cup
actually
has
not
had
some
progress
for
a
while
yeah.
So
there
is,
there
is
a
cap
about
the
application
crd
as
well,
so
there
is
already
a
application
crd.
Actually,
this
is
like
a
application,
backup,
crd
so
right.
So
this
is
based
on
the
existing
application.
Crd
that
is
already
there
do.
We
already
have
application
and
application
there.
H
H
A
Okay
yeah,
so
if
you
go
check,
I
think
it's
under
kubernetes
six
news
applications.
So
that's
already
that's
already
beta.
I
think
russian
chain.
I
think
that's
beta
application.
A
So
yeah,
so
this
cap
is
based
on
that
basically
use
that
one
to
arrange
just
to
decide.
You
know
what
are
things
belongs
to
your
application,
use
that
one
as
the
base
and
then
now
come
up
with
some
different
crds
for
backup
and
restore.
H
A
Already
using
that,
so
I
think
shantae.
I
think
you
look
more
into
this.
It's
not
a
lot
right.
There
are
some
references,
I'm
not
sure
how
many
like
how
many
people
are
really
using
the
application.
Crd.
D
Well,
there
are
some
users
of
application
crd,
I
cannot
name
exactly
which
ones,
but
the
application
crds
serves
the
main
purpose
of
defining
application.
Crd,
it's
slightly
different
than
what
we
want
to
achieve
from
the
data
protection
perspective
right,
so
that
they
have
a
lot
of
ui
ux
considerations
in
this
crd.
D
If
you
can
take
a
quick
look
down
the
definition
you
will
understand
they
put
like
image,
versions,
sources,
etc,
etc
into
that
not
only
just
the
group,
all
these
components,
so
I
I
I'm
not
sure
the
answer
is
that
who
is
exactly
using
this
in
the
production.
F
There
is
one
use
case,
which
is:
if
you
want
to
deploy
to
google
marketplace
for
gke,
you
have
to
define
the
your
own
application
cr
that
will
let
you
let
gke
interact
with
the
life
cycle
of
your
application.
I
think
that's
kind
of
the
fundamental
thing
here
right.
How
do
you
develop
tools
that
will
let
you
interact
with
the
life
cycle
of
an
application?
And
and
what
do
you
define
that
spec
as-
and
I
think
you
know,
there's
other
definitions
as
well,
so,
for
example,
helm
charts?
F
You
know
people
in
addition
to
just
not
using
them
at
all
right,
not
having
a
centralized
definition
that
tools
can
interact
with.
So
I
think
it's
it's
kind
of
challenging
to
use
this
as
a
definition
and
as
a
basis
for,
for
you
know
another
backup
and
restore
cr.
Having
said
that,
I
think
you
know,
maybe
you
can
have
something
general
purpose
that
has
like
a
layer
of
interaction
that
would
interact
with
that
as
well.
A
Okay,
yeah,
so
that's
yeah!
We
we
we
actually
discussed
about
this
one
some
time
ago.
I
think
shantian
did
the
review
this
one
in
this
meeting
a
while
back,
but
we
haven't
got
back
to
that.
Yet.
F
A
Oh
diagrams
right
so
yeah
we
probably
but
yeah.
So
let
me
just
just
start
putting
those
together
I
mean
whatever
doc
we
already
have
and
then
then
we
can
decide
which
section
we
need
some
diagram
yeah
right
now
I
mean
it's
a
little
premature
to
think
about
diagrams
right.
So
but
yeah
we
will
see.
We
will
need
some
diagrams.
A
Okay,
yeah,
we
can
yeah.
Maybe
we
can,
let's
see
how
that
looks
like
when
we
put
those
text
together
and
then
then
see
what
what
diagrams
we
need.
Yeah,
we
can
yeah.
We
can
sync
up
on
that
later
sounds.
A
All
right
yeah,
so
in
the
next
meeting
we'll
see
if
we
can
talk
more
about
the
cvt,
and
I
think
I
think
that's
it
for
today,
then
thanks
everyone
bye.
Thank
you.