►
From YouTube: Velero Community Meeting - March 15, 2022
Description
No description was provided for this meeting.
If this is YOUR meeting, an easy way to fix this is to add a description to your video, wherever mtngs.io found it (probably YouTube).
A
Okay,
hello,
everyone,
my
name
is
urban
vasilev
and
I'm
the
new
community
manager
for
valero
today
is
march
15th
and
jonas
asked
me
to
take
over
this
session
because
he's
not
feeling
well
today.
So
please
bear
with
me.
This
is
my
first
valero
community
meeting
with
that
said,
please
follow
the
pull
of
conduct,
just
in
other
words
behave
and
be
nice
to
the
other
people
without
I've
shared
the
link
to
the
hacking
d
for
the
community
meetings.
Notes
for
today.
A
Please
fill
yourself
in
if
you're
feel
free
to
add
all
the
status
updates,
announcements
and
discussion
topics
that
you
want
to
have
for
today.
With
that,
I'm
on
my
side
since
I'm
new
in
the
into
the
product.
I
don't
have
much
to
share
right
now,
so
I'll
give
the
mic
to.
A
I
can
see
discussion
topic.
So
I'm
sorry
if
I
mispronounced
the
name.
B
Yeah,
I
just
you,
can
call
me
fong
if
you
really
want
to
go
by
the
vietnamese
pronunciation,
as
you
can
see,
it's
get
full,
but
either
way
it's
fine
with
me.
I've
been
yeah.
I've
been
a
long-term
member
of
this
team
for
quite
a
few
years
anyway.
B
I
just
want
to
give
a
quick
update
for
the
the
issues
number
nine
393,
which
is
the
priority
class,
is
not
included
in
a
backup
if
the
part
is
referent
to
it,
and
this
become
an
issue
for
us
when
we
restore
to
a
new
cluster
that
doesn't
have
that
same
priority
class.
B
So
I
create
a
fix
for
that
and
I
also
put
the
pr
in
the
meeting
note,
even
though
this
dpr
still
have
still
need
some
work
on
it,
because
I
forgot
to
put
the
chain
log,
I
forgot
to
sign
off
and
so
on
so
forth.
I
have
to
work
on
it
a
little
bit,
but
if
you
you,
you
guys
can
take
a
look
at
it
to
to
see.
What's
going
on,
you
guys
can
look
at
it.
The
chain
is
very
simple,
I
simply
add
the
priority
class
name.
B
If,
if
the
priority
class
name
is
specified,
then
I
just
simply
add
that
object
into
the
additional
items
array
during
the
part
action
either
in
backup
or
restore.
So
it's
very
simple,
straightforward,
very
easy
to
review.
I
think
I
might
add
a
few
more
unit
tests
into
into
this
case,
so
to
kind
of
make
it
make
sure
that
it
got
tested
properly.
B
I
also
already
did
some
integration
tests
with
my
setup.
That
means
when
I,
when
I
run
the
backup
with
the
high
with
the
priority
class,
specified
in
the
port
name,
it
was
able.
I
can
see
that
in
the
backup
image
the
priority
class
was
backup.
B
Then
I
go
ahead
and
delete
the
higher
the
priority
class
for
my
cluster
and
I
restore
the
restore
the
the
namespace
and
I
saw
that
the
priority
class
in
the
cluster
scope
also
got
recovered,
so
it
integration
test
is
done,
so
you
guys
can
take
a
look
at
it.
That's
it
for
me.
A
C
D
E
Yeah
I
just
wanted
to
mention
that
I've
got
the
item,
snapshot
or
comments
in
the
update
progress
comment:
the
update
progress
pr.
I
addressed
all
the
review
comments,
so
it's
on
daniel
now
to
kick
it
in.
C
C
Great
and
as
you
can
tell
a
kind
of
a
common
theme,
partly
from
dave's
work
with
item
snapshotter
and
just
we're
doing
a
lot
of
planning
and
kind
of
pulling
things
together,
really
designing
and
such
for
for
one
night
and
in
the
future.
So
three
things
one
for
folks
who
were
on
the
call
last
week,
you
know
that
some
of
the
valerite
members
have
been
doing
researching
copia
as
a
potential
replacement
for
rustic.
That
document
that's
linked.
C
Has
that
comparison
next
week's
meeting
when
we
have
the
beijing
team
members
we're
going
to
be
really
discussing,
and
I'm
hoping
that
we
can.
I
mean
our
inclination
is
to
replace
copia,
but
of
course
we
want
to
replace
rustic
with
copia.
But
of
course
we
want
the
bigger
communities,
the
wider
communities
feedback.
So
please
please
come
next
week.
If
you
have
feedback
on
that
thoughts
on
that
and
obviously
the
longer
the
more
you
can
look
at
this
doc
ahead
of
time,
ask
questions
as
needed.
C
The
the
better
discussion
it'll
be
so
that's
the
first
thing,
copia
and
rustic,
and
if
you
don't
have
access
to
the
doc.
If
you
have
questions
pound,
valero
dev
is
the
great
place
to
ask.
C
Secondly,
our
second
kind
of
design
thing
we're
working
on
is,
as
you've
heard
by
now.
I
think
we
are
hoping
to
put
a
data
mover
feature
in
valero.
I'm
working
to
kind
of
create
a
prd,
because
we're
hearing
red
hat
has
has
suggested
some
requirements,
which
I
think
will
work.
We
know
that
the
vsphere
team
may
have
requirements,
so
we're
going
to
be
chatting
with
them
anyone
else
who
might
have
requirements
or
opinions
on
well,
if
you
have
requirements
for
the
data
movement
feature
for
sure,
be
in
contact
with
me.
C
If
you
have
thoughts
on
the
architecture
of
the
implementation
of
that
data
mover
feature.
Definitely
please
kind
of
keep
watching
we're
designing
this
now.
The
community
meeting
next
week
would
be
a
good
time.
A
few
specific
thoughts
on
that
and
last
announcement
multi-tenancy
might
actually
be
designed
during
one
nine.
We've
got
some
folks
who
might
be
interested
in
doing
that.
So
just
this
is
again
a
heads
up,
and
so
we
are
looking
into
that
a
bit
and
so
again
a
few
thoughts
on
that
we
can
potentially
discuss
it.
Now.
C
I
see
some
of
the
folks
here
who
have
been
thinking
about
multi-tenancy,
but
certainly
next
week
also
is
a
great
time
in
the
community
meeting.
So.
F
Hey
eleanor,
I
wanted
to
just
mention
from
our
end,
we've
actually
been
like
prototyping
something
about
data
mover
for
a
few
months
at
this
point.
Well,
it's
something:
we've
actually
been
thinking
about.
For
almost
a
year
and
a
half,
we
have
a
like
a
very
bare-bones
design
that
we've
tried
to
pull
from
our
prototype,
to
be
something
that
could
be
generic
for
valero
and
we're
actually
hoping
like
submit
that
this
week
to
start
getting
people's
eyes
on
it.
F
I
don't
know
that
doesn't
necessarily
mean
it's
like
that
remotely
at
all
the
final
solution,
but
we're
hoping
to
get
that
submitted,
and
then
we
can
kind
of
go
back
with
what
our
product
requirements
were
and
what
that
design
entails.
But
I
just
want
to
mention
that
if
folks
will
be
in
the
lookout
for
that
or
at
least
gonna
submit
it
up
there,
it's
obviously
needs
a
lot
of
feedback
and
a
lot
of
iteration.
But
I
think
we're
only
like
a
couple
days
away
from
getting
that
submitted.
So.
F
I
have
some
surface
level
knowledge
of
it.
There
might
be
there's
another
folk
on
my
team.
Actually
I
don't
see
him
on
here
that
looked
into
that
just
to
get
a
better
understanding
as
to
what
I
think
the
original
goal
right
was
correct
me
if
I'm
wrong,
but
was
it
potentially
taking
the
vsphere
data
mover
out
of
the
plug-in
and
putting
it
into
astrolabe,
was
that
what
they're
called.
E
Well,
moving
it
up
into
valero
and
taking
the
astrolabe
layer,
and
you
know
getting
more
use
out
of
it.
Yeah.
F
And
could
you
just
give
me
a
little
color
around?
What
exactly
would
the
vsphere
data
mover
support
in
terms
of?
Was
it
going
to
be
like
a
generic
all
storage
providers,
or
is
it
only
targeting
these
sphere
volumes.
E
So
right
now
the
the
data
mover,
the
vsphere
data
mover,
sits
on
top
of
asteroids.
So,
as
we
add
more
astrolabe
support,
the
data
mover
would
pick
it
up
pretty
much
automatically.
I
mean
there's
some
tweaks
in
there
because
there's
some
vspherisms
that
crept
in,
but
those
should
be
things
we
could
take
out.
So
it
was
designed
as
like
a
scale-out
solution
and
there's
a
poc
for
copia
repo.
So
there's
a
s3
repo,
that's
built
in
to
astralis
currently,
which
is
just
a
really
stupid
one.
E
So
it
puts
things
into
very
simple
objects:
there's
a
poc
that
I
did
for
replacing
that
repository
with
one
based
on
copia
and
then
it's
pretty
much
a
drop
in
replacement
for
the
existing
one
and
we
can
also
drop
in
other
repositories.
E
So
you
know
potentially
veeam
wants
to
drop
in
like
the
vvr
back
end
or
you
know
we
could
drop
in
the
dell
emc
backend
so
kind
of
like
how
we
currently
have
like
object,
store
plug-ins
in
in
valero.
F
That's
that's
cool,
interestingly,
so
there's
engineers
that
they're
not
associated
with
our
team
that
developed
like
a
totally
upstream
project
called
volsync,
which
is
really
about
replicating
volumes
across
clusters.
F
Prototype
a
couple
data
mover
solutions,
one
of
them
actually
being
rustic
as
the
solution
for
it.
So
in
the
same
vein,
it's
just
using
an
s3
repo
as
this
backup
storage
target
and
essentially
replicating
a
snapshot
into
object
storage.
So
that
was
our
first
like
proof
of
concept
with
this
how
this
works
and
they
also
have
rsync
data
mover.
F
I
could
see
copio
data
mover
also
being
something
they're,
probably
exploring,
but
the
main
goal
of
ours
being
that
we
don't
want
to
lock
in
a
specific
data
mover
solution
like
there's
third-party
backup
vendors
that
obviously
already
have
their
own
data
mover.
They
may
want
to
bring
their
own
into
an
existing
api
that
valero
offers.
So
that
was
how
we
kind
of
reworked
what
our
prototype
was,
which
was
very
volston
centric
to
be.
How
do
we
just
support
any
data
mover
right?
How
does
someone
bring
their
own
data
mover
into
this
solution?
F
So
it
sounds
like
you've
got.
You
know
some
experience
here,
we'd
love
to
hear
your
feedback
on
kind
of
what
we've
come
up
with.
F
Cool
okay,
yeah:
I
need
to
learn
more
about
that
myself
personally
so,
but
shoe
bomb
on
our
team
did
dig
into
this
a
little
bit.
I
don't
think
he's
on
right
now,
but
I'll
have
to
stick
up
with
him
and
see
what
he
learned
is.
H
There,
I
think,
a
question
that
I
would
have
is
is
there
sounds
like
to
me
like
we're,
adding
an
extra
layer
with
astrolabe.
H
E
E
So
the
goal
with
astrolabe
was
to
handle
things
other
than
volumes,
so
the
existing
implementation
handles
volumes,
but
you
can
plug
in
other
things
there,
and
you
know,
like
the
the
constant
example
has
been
like
the
postgres
database.
So
say
you
have
a
postgres
database
and
you
want
to
use
pg
dump
instead
of
doing
a
volume
snapshot
that
plugs
right
into
ash
delay,
but
it
plugs
right
into
the
existing
repo
and
the
existing
data
mover.
E
So
if
you
have
something
that
can
provide
a
data
source,
we
can
move
it
from
place
to
place
on
top
of
the
astrolabe
apis.
So
that
was
the
reason
for
having
the
astrolabe
apis
are
a
foundation
for
the
data
mover,
so
data
movement
is,
you
know,
managing
it,
making
sure
you're
doing
it
in
the
right
order.
You
know
load
balancing,
whereas
you
know
there's
apis
for
simply
extracting
data
and
snapshotting
data.
H
So
if
I
can
restate
it
dave,
just
make
sure
I'm
rocking
you're
saying
that
astrolabe's
intention
was
to
be
like
a
container
notifier
type
of
system
where
you
could
say
like
do
these
actions,
and
now
you
have
the
snapshot,
not
necessarily
just
a
data
movement,
but
then,
after
you
did
those
things
you
could
then
move
data
with
actually.
H
Okay,
do
don't
now.
This
might
be
a
really
silly
question
and
I
apologize
doesn't
valero
already
have
the
ability,
with
some
of
those
hooks
and
stuff
to
to
do
that
today,.
E
So
what
valero
was
missing?
Was
we
didn't,
and
this
is
this
is
pretty
much
missing
throughout
the
stack,
because
we
don't
really
have
data
path
apis,
so
valero
has
the
ability
to,
for
example,
call
up
execution
hook
and
then
do
some
stuff
and
then
call
another
execution
hook
versus
you
know
packaging.
It
such
that
there
is
an
api
that
you
say:
hey
do
the
right
thing
and
what
you're
going
to
do
is
kind
of
defined
by
you.
You
know,
so
it's
a
little
more
of
an
object-oriented
thing.
So
volumes,
for
example,
are
good
examples.
E
We
say
to
a
volume,
go
snapshot
yourself
and
it
goes.
You
know
the
storage
system
goes.
It
does
does
a
bunch
of
work
for
us.
We
don't
do
that
work.
We
don't.
You
know,
reach
down
into
the
storage
system
and
say
hey.
These
blocks
are
gonna
get
copied
over
here.
Those
blocks
could
be
over
there.
It's
like!
No,
that's
that's
not
what
what
valero
needs
to
do.
The
api
is,
is
the
contract
there,
and
so
the
idea
is
that
we
can
add
other
things
in
there
as
well.
E
H
Okay,
I
see
so
then
that
means
that
the
literal
user-
a
who
has
something
but
doesn't
have
an
asteroid
implementation
of
its
apis,
for
that
thing-
means
that
that
user
won't
be
able
to
use
the
data
movement.
Is
that
a
fair
understanding
or
not.
E
H
Well:
okay,.
H
E
E
E
H
E
E
Those
tend
to
do
things
like
push
the
data
into
their
own
format,
someplace.
So
once
once
we've
done
that,
so
we
can
handle
it
one
of
two
ways,
so
you
can
handle
it
like
rds,
where
everything
is
done
under
the
covers.
That's
where
we're
currently
at
with
volume
snapshots.
So
we
say
to
the
ebs.
For
example,
we
say
snapshot
of
volume,
ubs
says
okay
and
it
comes
back.
It
gives
us
a
snapshot
handler,
but
we
don't
have
any
paths
right
now
for
saying:
okay,
we
want
to
move
that
snapshot
from
region
a
to
region
b.
E
H
E
H
E
E
That's
all
gets
pushed
down
into
the
into
the
asteroid,
protected
energy,
so
that
that
provides
you
with
an
api
that
says.
Yes,
I
can
read
your
data.
I
can
copy
that
data
someplace
else.
H
Maybe
yeah,
maybe
I'm
just
being
dense.
I
apologize.
E
E
F
Well,
how
I've
been
thinking
about
this
problem
and
I
apologize-
I
don't
actually
know
as
much
about
astrology
as
I
wish
I
did
so.
You
may
just
be
explaining
something
that
I'm
kind
of
re-stating
here.
I've
been
to
purchase
problems
like
the
csi
level,
which
is
how
do
you
perform?
How
do
you
take
a
generic
csi
snapshot
and
then
allow
it
to
be
like
portable,
and
I
wasn't
necessarily
thinking
about
it
from
the
perspective
of
like
you
know,
apis
that
a
different
type
of
application
may
support.
F
E
Just
one
level
up
from
that,
so
we
take
csi
as
abstracting
out
volumes.
It
actually
doesn't
provide
data
paths
at
the
moment,
which
is
one
of
the
problems
so
like
we
wind
up
having
to
do
like
snap
and
clone
and
then
attach
in
order
to
get
things.
So
we
need
a
data
path
now,
if
we've
gone
to
the
trouble.
So
what's
the
difference
between
a
pvc
and
any
other
object.
E
So
astrolabe
is
just
saying:
look
if
you
provide
an
api
that
lets
me
say,
snapshot
you
and
let
me
extract
that
serialized
snapshot.
Then
we
can
go
ahead
and
put
that
into
s3.
We
can
copy
it
point
to
point.
We
can
do
whatever
we
want
to
it,
and
the
format
of
the
data
doesn't
really
matter
to
us,
because
we
don't
want
to
look
inside
the
data.
We
don't
look
inside
the
volume
data
as
it
is.
E
H
H
G
E
H
Well,
why
so,
why
why?
Why
can't
the
data
remover
api
be
something
different?
Why
can't
we
separate
these
two
things
out?
It
seems
so.
H
E
H
E
H
F
Yeah
yeah,
I
think
that's
what
dave
has
he's
saying
that,
like
essentially
the
data
mover
from
their
perspective
of
what
they're
gonna
do
with
the
vsphere
plug-in
correct
me
if
I'm
wrong
david,
but
it's
essentially
afterlife's
providing
that
first
section
which
is
getting
the
bits
and
then
the
data
movement
is
still
its
own
api.
On
top
of
it
right,
that's
right.
F
How
we
were
approaching
the
problem
too?
Sean
am,
I
misunderstanding,
something.
H
Okay,
that
makes
sense
my
maybe
I
that's
where
I
was
saying
I
must
I
might
be
just
earlier,
so
I
apologize,
but
what
I
thought
we
were
saying
is
like
these
two
things
are
intro
are
linked
together
and
they
can't
be
broken
apart.
E
F
H
That's
where
I'm
getting
it
sounds
like
to
me.
What
we
need
to
do
is
to
find
that
data
mover
api
in
a
way
that
can
accommodate
multiple
things
and
that's
what
I
was
suggesting
when
we
say
when
I
said
break
it
apart
sounds
like
in
one
hand
you're
saying
it
can
be
broken
apart,
but
then
you're
also
saying
well.
Astrolabe
gives
us
all
of
these
things
that
we
need
well.
H
Divide
hold
on
a
second,
and
we
probably
need
to
divide
to
design
the
data
mover
api
in
a
way
that
can
be
used
more
extensively
right
so
that
other
people
can
use
that
data
movement,
not
just
astrolabe
users.
Right
is
my
point
that
I'm
trying
to
make
you're
saying
actually
gives
us
this,
and
I
don't
disagree
with
you.
It
does
and
that's
really
really
helpful.
H
D
H
E
H
Well
well,
well,
I
love.
F
E
Just
it
that's
why
I
want
to.
I
want
to
start
moving
us
up
a
level
so
so
so
think
about,
say
cassandra,
okay,
right!
When
you
back
up
cassandra,
do
you
want
to
snap
all
the
disks
you
can,
but
that's
not
really
the
right
way
to
do
it.
Sure
what
you'd
like
to
have
is
you'd
like
to
say:
hey,
I'm
going
to
do.
I
think
it's
medusa,
I'm
going
to
do
a
medusa
backup,
but
now
how
do
I
get
that
medusa
back
up
from
a
to
b.
F
That's
actually
an
interesting
point
because
especially
distributed
database
is
actually
pretty
good.
That's
a
pretty
good
use
case
dave
of
like
where
this
this
needs
to
be
thought
of
slightly
because
you're
right
to
sean's
point.
Yes,
you
could
do
this
thing
where
you
dump
it
all
to
one
individual
volume
you
just
do
medusa
back
up,
you
dump
it
to
a
volume
you
take.
A
snapshot
of
that
volume.
Is
that
the
correct
way
to
do
it
in
the
most
efficient
way?
Probably
not.
E
And
we
wind
up
with
a
lot
of
extra
steps.
We
have
the
same
problems
right
now.
We
were
talking
about,
for
example,
being
able
to
use
valero
to
migrate
data
from
point
a
to
point
b.
We
currently
can't
do
that.
It's
going
to
have
to
take
a
a
step
through
the
object
store,
so
you
do
like
a
rustic
backup
into
s3,
then
a
rest
of
restore,
so
one
of
the
things
we
were
doing
with
astrolabe
is,
you
can
simply
say
I'm
going
to
copy
this
disk
from
here
to
there.
E
H
H
I
think
that
we
should
potentially
think
about
this
as
a
larger
solution
in
the
context
of
like
kubernetes
and
what's
coming
and
like
think
about
it,
and
my
reasoning
for
breaking
it
up
is
that
I
think
this
is
an
important
thing
that
needs
to
exist.
I'm
not
convinced
yet
that
all
the
other
pieces
that
the
larger
solution
that's
being
pitched
was
all
tied
together
is
necessarily
should
be
part
of
that
solution.
H
E
So
we
have
the
data
pads.
Astrolabe
has
the
data
paths
already?
We've
got
an
implementation
for
fees.
Here,
we've
got
an
implementation
for
eps,
so
we
can
do
things
like
direct
volume,
access,
direct
snapshot,
access,
which
is
something
we
don't
get
yet
with
csi.
We
can
build
on
top
of
csi.
We
have
to
build
things
alongside
it.
You
know,
fong
is
doing
work
on
the
cvt
stuff,
that's
not
going
to
get
in
for
a
couple
years.
We've
got
to
implement
it
and
prototype
it
somewhere,
and
then
we
start
pushing
it
down
the
stack
so.
F
F
Right
yeah-
and
I
saw
that
was
closed,
so
I
just
don't
know
what
the
status
of
that
is
on
your
guys's
end,
but
I
guess
the
vmware
end
if
that's
who's,
continuing
that
along
but
again
like
we
were
coming
at
this
from
the
angle
of
like
here's,
a
csi
snapshot
that
has
to
get
moved
somewhere
from
openshift
storage
perspective,
that's
desperately
needed,
and
it's
actually
a
generalized
problem
in
kubernetes.
We
designed
a
solution
around
that.
F
I
think
cassandra
is
actually
a
really
interesting
idea
that
didn't
really
hit
me
until
now
that
that's
like
a
really
good
use
case
of
where
this
does
fall
short,
I
think,
there's
ways
to
work
around
it,
but
the
whole
point
being
like.
I
would
just
love
to
get
your
input
on
kind
of
what
we
did
come
up
with
to
solve
it
from
like
the
pure
csi
snapshot
level,
and
I'm
sure
we
can
have
a
good
conversation
on
github
around
this
again.
F
If
there's
someone
else
driving
a
design
elsewhere
like
we
would
love
to
work
together
and
see
it,
it's
just
it
just
seemed
like
the
one
that
bridget
before
was
going
stale.
I
don't
know
if
there's
anyone
still
continuing
it,
but
we
kind
of
it's
like
a
really
high
priority
for
us.
So
we've
gone
on
the
path
of
just
making
something
work
and
now
we're
trying
to
generalize
it,
get
it
more
like
for
what
valero
could
consume,
but
clearly
there's
a
lot
of
other
people
that
have
stake
in
this.
F
So
yeah
we're
just
looking
for
feedback.
C
And
I'll
just
note
that
I
think
there's
a
good
discussion,
but
we're
still
missing
at
least
two
key
stakeholders,
one
of
which
are
the
beijing
team
who
are
and
daniel
who
you
know
who's
driving
a
lot
of
this
and
then
also
shane
from
the
plug-in
team.
Although
dave
is,
of
course
representing
somewhat
her
views.
D
C
I
C
That
to
say
is,
I
think,
there's
a
really
good
discussion,
actually
just
paying
the
beijing
folk
to
make
sure
that
they
watch
this
recording
so
that
they're
caught
up
on
the
discussion
thus
far-
and
maybe
it's
next
commune
next
week
for
the
community
meeting
or
possibly
dylan.
I
know
that
this
is
urgent
for
you
and
frankly,
we
want
to
make
sure
the
conversation
continues.
It's
possible.
We
should
schedule
a
separate
session
if
it's
not
if
on
slack,
we're,
not
kind
of
or
on
if,
if
asynchronously
we're
not
getting
the
discussion
going.
F
Yeah,
that
might
be
a
good
idea
only
because
it's
like
this
is
like
the
topic
but
you'd
schedule
our
call
and
ends
up
going
for
two
hours,
because
of
just
how
many
things
this
deals
with,
but
I
think
from
our
end,
all
we
would
like
to
do
is
like
we're
trying
to
just
detail
what
we
did
prototype
in
a
generalized
design
we'll
get
that
submitted
before
we
do
meet
and
talk
about
this
in
a
larger
context.
F
That
way
you
just
get
an
idea
of
where
we
were
coming
from
clearly
there's
a
lot
of
things
that
dave
brought
up,
that
I
hadn't
really
considered,
and
I
would
love
to
kind
of
get
that
perspective
too
so
yeah.
I
would
love
to
have
a
much
longer
conversation
on
this.
It's
something
we've
been
thinking
about
for
a
while,
so
well,.
C
You
know
what
I
think
I'll
do
then.
I
think
I
will
try
to
organize
a
it'll
be
an
evening
north
america,
so
the
dave
and
the
beijing
folk
can
attend
so
we'll
try
to
I'll
reach
out
to
you.
We
can
find
a
time
we'll
communicate
it
to
the
larger
community
for
anyone
else
who
wants
to
come
to
it,
but
it'll
be
data
mover
specific.
I
think
we
need
that
chat.
C
I
did
have
one
other
kind
of
question
it
might
be
for
dave
because
it's
be
sphere
specific,
and
so,
if
it's
long,
we
don't
have
to
do
it.
If
it's
not
the
question
I
had,
or
maybe
it's
two
and
again
maybe
you
all
already
asked
this,
and
I
didn't
answer
this
and
I
didn't
realize
one
is.
E
I'm
not
sure
what
what
shing
said.
So
I
okay
so
right
now
the
the
vadp
stuff
is
all
encapsulated
inside
a
protected
entity,
so
so
that
you
know
in
my
thinking.
You
know
when
I
was
telling
vmware
it
was
that
that
thing
would
continue
to
be
maintained
by
the
vsphere
team,
because
it's
not
really
a
tanzoo
thing.
C
And
am
I
right
to
think
then
that,
okay,
this
vadp
specific
information,
if
a
more
generic
data
mover,
did
call
astr,
I
mean
ignoring
all
the
other
stuff.
You
said-
and
maybe
this
was
all
said-
and
I
just
missed
this-
if
the
if
the
generic
data,
if
a
valeria
data
mover
called
that
asteroid
protected
entity,
it
gets
access
to
the
vadp
specific
stuff
to
make
a
really
efficient
data
transfer.
Is
that
yes,.
C
C
F
We
are
coming
at
it
from
like
a
what's
the
generic
data
mover
thing
that
gets
you
the
basic
requirement
here
and
then
there's
a
lot
of
you
know
world-class
vendors
that
have
specialized
specifically
in
these
problems
to
get
you,
you
know
the
best
dedu
the
most
efficient
copying
like.
I
don't
know
that
we're
going
to
be
able
to
replicate
that
from
like
a
valero
perspective,
I
don't
know
that's
even.
E
So
when
we
originally
were
talking
about
data
mover
we're
talking,
for
example,
with
veritas,
who,
I
don't
think
is
represented
today-
and
one
of
the
goals
was
to
have
the
valero
data
mover,
be
something
that
was
swappable
okay
and
that
you
could,
you
know,
swap
it
out
for
certain
types
of
things,
but
we
also
wanted
to
get
kind
of
the
concept
so
like,
for
example,
like
the
the
disc
stuff
is
pretty
well.
People
have
done
a
lot
of
stuff,
there's
different
levels
of
done,
but
there's
a
lot
of
stuff.
E
That's
that's
available,
but
the
the
other
stuff
like.
If
we
we
want
to
be
able
to
say,
snapshot
cassandra
or
whatever
and
start
moving
that
around.
That's
not
done,
but
we
can
make
that
available.
You
know
as
a
set
of
plug-ins,
so
like
that's
another
good
reason
for
splitting
astrolabe
from
the
data
mover
per
se.
So
there's
a
valero
data
mover.
That's
the
generic
one,
there's
a
set
of
asteroid
protected
entities
that
also
exist
independently.
E
If
you
are
say
casting
or
veritas,
you
may
decide
to
implement
your
own
data
mover
and
do
your
own
version
of
the
disk
stuff,
because
you
have
this
stuff
that
works
really.
Well.
You
have
a
backend
repository
that
works
really
well,
but
then
we
can
take
advantage
of
the
apis
for
say
cassandra
being
generic
and
use
them
in
our
data
mover
without
having
to
write
cassandra
specific
stuff,
which
is
code
that
we
do
not
have.
C
F
C
This
is
again
going
to
show
my
lack
of
knowledge.
This
is
very
specific,
but
my
understanding,
I
dare
I
say
for
dell
and
I'll,
ask
you
and
anyone
else
here:
lots
of
people
care
about
v
sphere.
Besides,
just
us
tanzu
right,
I
mean
my
understanding
is
that
there
are
non-tanzu
vsphere
users
out
there
for
valero.
So
is
it
okay?
If
we
talk
about
vsphere
in
general
on
this
call
well
here
I'll,
ask
I'll
ask
my
question:
if
it's
too
vmware
specific,
I'm
happy
to
take
it
offline,
let's
put
it
that
way.
C
I
have
heard
from
in
this
case
it's
a
tanzy
customer,
but
I
think
it
might
be
more
general
asking
saying
that
they
don't
always
have
access
to.
I
guess
it's
an
s3
backend.
I
need
to
look
at
my
email
and
that
they
are
asking
about
whether
valera
might
support
nfs
shares,
and
I
don't.
C
F
Well,
this
is
exactly:
we've
actually
had
the
same.
Ask
from
the
red
outside.
So
when
we
did
a
migration
tool
and
valero
was
the
thing
underneath
it.
It
was
very
common
that
folks
said
I
don't
have
object.
Storage
I
have
an
nfs
share.
Can
I
just
shove
the
data
there
right
now,
I
believe
nolan?
Maybe
it
wasn't
nolan
or
someone
else.
Didn't
someone
prototype
a
backup
storage
plug-in
for
nfs?
Am
I
remembering
this
correctly.
E
F
Yeah,
I
believe
I
can
dig
up
a
repo
around
it.
I
don't
think
it
was
very
well
maintained,
but
it
wasn't
super
complicated.
I
think
it'd
be
really
cool.
If
someone
continued
that
work
and
brought
it
back
into
valero,
absolutely
interesting.
C
If
you
could
dig
it
up,
that
would
be
good,
but
actually
can
I
ask
a
question
first.
So
what
first
of
all,
why
would
people
say
that
they
don't
have
an
object
store?
I
mean
I
can
try
asking
this
person,
but
can
anyone
help
here
help
me
understand.
I
thought
it
was
really
easy
to
set
up
men.
I
o
or
whatever.
E
E
It's
going
to
be
a
shared
file
system,
so
it
acts
like
a
file
system
but
yeah.
It's
like
you
know.
You
use
windows,
file,
sharing,
right
or
well
kind
of.
C
A
mac,
but
the
is
the
point
that
with
v
on
v-
and
this
is
v,
series-
is
nfs,
share
a
vsphere
specific
thing
or
not.
E
Yes,
and
no
so
I
think
the
way
they're
trying
to
use
it.
It's
not
v
series
v
sphere,
specific.
C
Okay,
actually,
you
know
what
the
email
says
nothing
about
vsphere.
So
maybe
I
got
that
part
wrong.
Okay,
and
so
your
point
is
that
most
people
have
an
nfs
share
that
they
could
use
instead
of
an
object,
storage.
Some
people
have
that
and
that's
what
they
want
to
use.
Instead,.
E
Some
people
have
that.
I
think
we
were
getting
the
the
request
from
the
vcenter
team
that
they
wanted
to
use
ftp.
Also,
some
customers
are
using
ftp
and
it's
like.
Oh
no.
E
C
D
I
mean
you
can
already
use
nfs
as
a
pvc.
You
know
as
a
pv.
So
so,
if
your
volume
data
is
on
nfs,
we
support
that
already
through
and
I'm
assuming
copio
would
as
well
but
yeah.
This
is
different,
because
this
is
this
is
a
backup,
storage
location
which
is
specific
to
right
now,
s3
or
azure,
or
you
know
that
kind
of
thing.
So.
C
To
be
clear
too,
they
must
be
wanted
to
do
it.
They
you
know,
people
who
would
want
to
do
this
would
want
to
use
it
for
the
object
store
either
for
saving
rustic
volumes
or
for
the
kubernetes
metadata,
because
it's
the
object
store.
We're
not
talk,
okay
right,
so
I
guess
the
point
then
so.
Thank
you.
Now
I'm
caught
up
knowledge-wise.
So
what
I'm
hearing
then
I
mean,
I
guess
the
question
was:
is
it
something
we
thought
about?
C
Dylan's
point
is
it
sounds
like
possibly
nolan
did
something
like
this
in
the
past,
dylan
will
dig
it
up.
I.
F
E
But
that's
only
for
the
object
store
plug-in,
so
that's
going
to
put
the
valero
data
in
there,
but
not
anything
else.
Yeah
and
rustic
too
right
should
should
rustic,
doesn't
sit
on
top
of
the
object
store
plug-in.
So
I'm
not
sure
if
we
have
to
be
able
to
configure
rustic
to
do
that.
Interesting.
C
C
F
E
D
C
I'm
just
going
to
resave
this
because
I
really
want
to
make
okay,
so
the
actual
backup,
storage
location
which
points
to
the
object
store,
that's
shared
between
rustic
and
the
metadata
kubernetes
metadata,
but
the
object
store,
plug-in
is
used
for
the
kubernetes
metadata.
Rustic
has
its
own
ways.
That's
right!
I'm.
E
C
Thank
you,
okay,
so
cool
I
can.
I
can
send
this
to
the
person
who's
asking
and
I
guess
more
generally,
it's
a
more
general
question.
I
I
had
not
been
aware
of
these
asks
before.
Is
this
some,
so
I'm
hearing
that
for
the
valero,
it
might
be
a
nice
feature
to
add.
Is
there
still
actual
immediate
need
from
any
button
here
on
the
call,
or
is
it
just
something
something
nice
but
depends
on
user
need.
F
So
something
nice
with
mineo
available
and,
like
you
know,
I'm
sure
other
people
supply
different
s3
compatible
solutions
like
that's
all
we've
ended
up
doing
is
just
like
bundling
one
of
red
hat's,
other
s3
things
and
giving
it
to
them
and
mini,
is
so
easy
to
set
up,
and
maybe
it's
just
making
that
documentation
even
simpler.
But
it's
not
that
complicated
to
me.
But
today's
point
is
another
thing
to
maintain
so.
D
E
Yeah,
that's
possible
too.
I
think
we've
been
resisting
adding
nfs
as
a
target,
because
it's
just
yet
another
thing,
and
you
know
it
it's
like
you
know
more
work
plus,
you
know,
expands
the
testing
matrix
and
it's
gonna
be
more
support,
so
that
was
that
was
kind
of
why
we
were
resisting
doing
it.
C
A
G
Yeah
hi,
can
you
can
you
help
me.
G
Yeah
hi,
I'm
rachel.
I
work
as
an
engineer
at
cern,
so
we've
been,
I
don't
have
any
slides
to
mention
a
use
case.
I
can
briefly
tell
about
what
we
use,
how
we
use
valero.
So
I
work
in
a
team
that
handles
the
websites
etc,
both
public
facing
and
the
private
websites.
We
have
more
than
1500
2000
plus
websites
and
we
run
everything
through
operators.
So
we
have
customers
as
definitions
and
and
at
cern
we
everything
is
self-hosted,
so
we
have
our
own
data
center.
G
We
use
openstack
instances
and
then
we
use
the
okd,
the
open
source
version
of
openshift,
and
then
we
host
all
this
many
thousands
of
websites
and
and
valero.
We
used
different
cases
over
here.
One
is
it's
our
cluster
state
backup
by
default
in
most
of
our
clusters
that
only
stores
the
metadata
and
it's
specific
to
public,
pressing
websites
like
like
the
home.com,
the
home
page.
So
for
these
websites.
These
are
content
management
like
same
as
websites,
so
they
have
also
like
data
persistent
volumes.
G
So
here
we
use
another
instance
of
valero
to
take
rustic,
backups,
pre-ops
and
post
hooks,
so
we've
been
using
for
a
while.
Now
our
project
is
kind
of
stable
at
this
point
and
then
we
wanted
to
see
if
we
can
contribute
back
in
our
free
time.
So
yeah
happy
to
be
here.
G
C
To
have
you
here,
thank
you
so
much
for
joining,
and
you
got
a
very
good
community
me
to
join.
You
saw
like
a
robust
argue.
You
know
a
good
good
discussion
and
then
some
other
things.
So
I'm
really
glad
glad
to
have
you.
C
I'll
just
merely
finish,
I
I
had
cut
myself
off
and
then
so
yeah.
So
please
feel
free
to
keep
coming
to
the
community
meetings
with
julia
and,
as
you
said,
if
you
all
want
to
contribute,
I
mean
this
is
the
place
to
start
joining
and
certainly
I'm
I'm
the
product
manager.
C
But
the
engineers
can
also
help
talk
about
where
it
might
make
sense
to
jump
in
so
and
just
so
you
know
a
good
number
of
our
maintainers
are
in
beijing,
so
they're
not
here
today,
but
they'll
watch
this
video
and
see
your
introduction.