►
From YouTube: SIG - Storage 2022-12-05
Description
Meeting Notes:
https://docs.google.com/document/d/1mqJMjzT1biCpImEvi76DCMZxv-DwxGYLiPRLcR6CWpE/edit#
A
A
A
Foreign
I
would
point
you
guys
to
the
agenda
doc,
which
is
in
the
meeting.
Invite
and
I
would
encourage
anyone
to
add
agenda
items
for
this
or
future
meetings
in
order
to
ensure
that
they
can
be
discussed
here.
So
please
feel
free
to
do
that
and
I
see.
Some
of
you
are
adding
your
names
to
the
attendees
list.
So
that's
great.
A
And
I
wonder
I
see
Andrew
is
here
I'd
like
to
thank
him
for
setting
this
up,
as
this
is
our
first
kind
of
Sig
storage
meeting
I,
don't
know
if
there's
anything
that
you
wanted
to
mention
Andrew
at
the
beginning,
as
we
get
organized.
A
Yes,
yes,
it's
for
this
basically
storage
topics
related
to
for
cute
bird,
yes,
foreign.
A
And
I'm
trying
to
see
who
I'm
not
exactly
sure
who
has
each
event
individual
agenda
item,
let's
see
so
we
have
the
VM
definition.
Contents
on
VM
export,
maybe
I
should
share
my
screen.
Let's
see
if
I
can
do
that.
A
All
right,
let's
see
I,
think
maybe
Alexander
Wells-
may
have
added
the
first
item
and
I
don't
see
him
here
yet.
A
All
right,
let's
see
foreign
who
has
the
next
topic,
bzvm
snapshot
and
wait
for
first
consumer
yeah.
D
Yeah,
so
the
problem
is
when
you
have
a
storage
that
is
Snapshot
capable,
but
at
the
same
time
it
can
be
paid
first
consumer
and
we
create
a
snapshot
and
create
a
VM
restore,
and
this
VMware
object
doesn't
get
to
complete
State,
because
the
restore
DV
stays
the
meters
consumer
and
the
PPC
stays
pending,
and
it
eventually
starts
importing
when,
when
we
try
to
start
the
restored
VM
but
I'm,
not
sure
if,
like
as
a
user,
if
I
create
the
restore
object
and
it's
not
completed
I'm,
not
sure
that
I'm
gonna
try
to
start
the
game.
A
Yeah
so
it's
kind
of
leaving
us
in
a
situation
where
the
restore
appears
like
it's
not
successful,
but
everything
actually
is
okay
right,
so
I
see
that
you've
proposed
a
couple
of
solutions
here.
A
So
I
guess
yeah
for
for
a
non-vm
workload,
you
would
re
you're,
basically
just
restoring
the
the
PVC
from
snapshot
and
in
that
case,
you're
probably
running
your
pod.
Along
with
that
so
I
think
in
that
case,
yeah
I
would
just
it
would
stay,
wait
for
first
consumer
until
you
until
you
run
the
pod.
So
we
have
something
for
the
import
use
case
where
we
create
like
a
doppelganger
pod.
A
I
guess
is
the
best
way
to
describe
it
that
that
causes
like
forces
things
to
kind
of
get
moving,
so
we're
not
in
a
stuck
situation
and
then
CDI
is
able
to
do
the
import.
So
yeah
we've
changed
the
behavior
kind
of
for
this
particular
case
for
the
restore
it
might
be.
Okay,
I
think
to
have
it
be
marked
complete
the
what
I
worry
about
option
two
with
The
annotation
is
like
I'm
thinking
about
this
in
terms
of
what
is
the
behavior
I?
A
Guess
we
we
probably
don't
really
care,
let's
say
if
a
if
a
VM
hold
on
a
second
I'm,
actually
I,
always
confuse
myself
with
wait
for
first
consumer
versus
local
storage,
but
you
can
actually
have
non-local
storage.
That's
weighed
for
first
consumer,
so
that's
never
mind.
My
point
is
not
relevant
here.
The
bind
immediate
though
it
can
can
cause
unwanted
side
effects
if
we're
binding
it,
where
we're
not
going
to
actually
run
the
VM.
A
Okay
and
then
Okay,
the
third
option
is
let
the
controller
start
and
stop
the
restored
VM.
So
this
would
be
essentially
the
doppelganger
approach.
A
And
I
think
in
the
populators
case,
we're
going
to
be
tolerating
the.
How
actually,
how
is
that.
F
G
I
think
we
can
I,
don't
know
when
exactly
the
restore
is
marked
as
complete,
but
it
could
be
potentially
be
complete
what
once
all
the
resources
are
created
and
updated?
A
It's
because
the
right,
the
PVC
is
impending
until
the
VM
actually
gets
started.
So
everything
is
good
to
go,
but
because
we're
waiting
for
the
PVC
to
be
bound
yeah.
G
I
mean
I'm
surprised
that
we're
waiting
on
that
but
yeah
we
shouldn't
I,
think
once
everything
has
been
created
and
the
restore
should
be
considered
complete.
A
A
And
we
can
start
to
go
on
to
the
next
topic.
I,
don't
know
if
we
wanted
to
bounce
up
to
the
first
item,
the
VM
definition
Alexander.
That
might
be
your
item
right.
Yes,
all.
F
Right,
okay,
I
just
wanted
to
sort
of
explain,
what's
going
on
right
now
and
and
see
if
anybody
has
any
thoughts
or
suggestions.
So
basically,
we
recently
added
a
VM
export
functionality
and
essentially
it
starts
a
pod
and
it
allows
you
to
download
disk
images
in
in
either
raw
or
compressed
format,
and
the
second
phase,
where
we're
in
now
is
where
we've
added
or
where
we're
adding
a
new
rest
endpoint
from
the
same
plot
and
that
rest
endpoint
will
contain.
F
If,
if
the
source
of
the
export
was
a
virtual
machine
or
a
virtual
machine
snapshot,
the
endpoint
will
contain
the
definition
of
the
virtual
machine.
So
you
can
download
the
definition
and
apply
it
to
a
different
cluster
or
something
like
that
and
sort
of
there's
a
few
open
or
open
questions
on
on
what
we
want
to
do.
A
Alexander
could
I
just
ask:
do
you
happen
to
have
an
a
draft
PR
or
an
open
PR
for
this?
Not
yet.
F
I
I
have
a
local
version.
I
haven't
pushed
the
address
PR
up
yet,
okay
thanks,
but
basically
what
adjustment
generating?
Is
this
VM
definition?
It
takes
the
The
Source
VM
definition
runs
it
through
the
instance
types
and
preference
types
and
expands
it
into
a
fully
expanded
virtual
machine
so
there's
no
instance
types
or
anything
in
the
actual
output,
and
then
it
looks
for
a
data
volume
templates
and
replaces
the
source
of
the
data
volume
templates
with
the
URL
of
the
export
endpoint.
F
If
there's
no
data
volume
templates,
but
there's
a
PVC
or
a
data
volume
Source
in
the
VM
itself,
it
will
generate
data
volumes.
This
makes
the
assumption
that
CDI
will
be
available
wherever
we're
trying
to
create
this
new
virtual
machine
so
I'm.
My
question
sort
of.
Is
that
a
reasonable
assumption,
or
you
know
people
might
not
use
CBI
and
then
I,
don't
know
how
useful
this
whole
rest
endpoint
is
without
CVI.
You
know
it's
just
the
VM
definition
and
you
can't
do
much
with
it
at
that
point.
F
So
does
anybody
have
any
issues
or
questions
about
that.
A
The
one
one
use
one
usage
pattern:
first
of
all,
I
think
that
the
CDI
you
know
the
usage
pattern
of
importing
it
into
a
different
cluster
is
a
good
and
interesting
one
which
sounds
like
this
will
solve
really
well.
One
other
use
case
that
I
wonder
about
is
so
in
the
VM
export.
There
exists
the
URLs
to
just
download
the
the
disks
locally.
A
So
if
somebody
was
trying
to
basically
pull
all
the
pieces
down
to
like
a
local,
workstation
and
Cobble
things
together,
in
that
case,
the
the
XML
with
the
source,
references
wouldn't
make
sense
in
the
CDI
anymore,
but
again,
I,
don't
know
exactly
what
people
would
do
with
that.
They
still
would
have
to
rewrite
the
XML
slightly
like,
for
example,
if
they
were
doing
an
upload
or
whatever
they're
doing
so
as
long
as
it's
pretty
obvious
how
you
would
replace
those
which
I
guess
it
is
because
if
you
understand
the
format.
F
Right,
it's
it's
just
replacing
the
The
Source
in
the
data
volume
to
be
you
know
whatever
it
is.
You
want
to
be
if
you
have
an
alternate
Source
I
I've
seen
somebody
that
sort
of
wanted
to
take
the
exported
disk
images
storm
in
an
S3
bucket
and
then
use
the
S3
bucket
as
a
source
for,
for
you
know,
creating
copies
of
this
virtual
machine
sort
of
like
an
offline
storage
and
then
create
new
VMS
in
other
clusters.
So.
A
Okay,
yeah
and
I
think
it's
important
for
us
to
kind
of
mention
too.
That,
with
this
feature
like
I,
think
we're
trying
to
do
things
in
an
incremental
approach.
So
we're
not
trying
to
create
a
canonical
Cube
vert
VM
format
like
a
you
know
like
a
self-contained
format
or
anything
like
that.
We're
just
trying
to
kind
of
have
a
raw
export
capability
see
how
people
use
that,
and
then
we
can
figure
out
what
the
next
steps
are
in
terms
of
how
it
might
be
useful.
C
So
well,
I
I,
don't
know
if
you
can
hear
me
because
yes
yeah,
okay,
so
with
image,
I
o
we
we're
looking
at
actually
both
uploading
and
downloading
disk
images.
This
is
for
Rev
and
one
idea
that
we
had,
we
didn't
implement
it
in
the
end,
but
we
think
it
would
be
much
more
efficient
would
would
creating
an
MBD
endpoint.
C
C
So
the
advantage
of
this
is
that
you,
the
downloading
tools
we
have
for
MBD
are
much
more
efficient
than
just
downloading
everything.
Over
http
I
mean
obviously,
if
you're
already
compressing
stuff,
then
that's
probably
not
true,
but
if
you're,
if
you're
just
sort
of
downloading
raw
disk
blocks,
then
the
MBD
one
would
skip
over
all
the
blank
space
and
stuff.
So
so.
C
C
I
mean
we
also
have
like
you
know,
MBD
Uris,
which,
which
means
you
can
actually
Canon,
actually
refer
to
a
specific
remote
location
with
a
defined
format
and
everything.
So
if
we've
sort
of
solved
a
bunch
of
those
kind
of
problems
but
yeah
up
to
you.
A
So
since
we've
oh
go
ahead,
Alexander
I
didn't
say
anything:
okay,
sorry
I
thought
someone
was
jumping
in
here.
I
had
one
comment,
since
we
got
Richard
here
too
and
I
would
wonder,
would
would
any
of
the
V
to
excuse
me
V
to
V
or
guest
Fest
type
of
utilities
want
to
use
an
export
of
this
nature,
or
would
they
be
more
likely
to
to
build
things
essentially
build
the
export
environment
by
itself?
I
guess
would
be
the
question.
C
I'm
not
really
sure
I
mean
I'm,
not
really
sure
what
the
use
case
of
this
is
actually
in
general,
I
mean
if
we
were
talking
about,
for
example,
modifying
disk
images,
then
having
a
an
actual
modifiable,
which
is
not
what
you're
talking
about,
because
it's
just
downloading,
but
if
you,
if
you
could
also
modify,
then
that
would
be
interesting,
although
kind
of
exciting,
in
the
way
that
you'd
have
to
ensure
that
the
VM
was
was
switched
off.
A
F
The
use
case
is,
is
going
from
one
cluster
to
another
and
having
the
target
cluster
connect
to
the
source
cluster
directly,
there's
a
few
more
things
that
we
need
to
generate
so
that
CDI
can
properly
connect
to
the
source
cluster,
in
particular
the
Source
CA
cert,
so
that
it
it
knows
it's
talking
to
the
right,
endpoint
and
then
there's
also
a
you
know,
secret
token,
to
authenticate
that
you're
actually
allowed
to
talk
to
this
endpoint.
F
It's
essentially
the
the
ca
service
is
just
you
know.
If
you
talk
to
any
ATP
endpoint,
you
need
to
know
that
you're
talking
to
the
actual,
correct,
endpoint
and
that's
what
this
shirt
is
for
too.
So
you
can
so
you
can
determine
that
the
endpoint
you're
talking
just
the
right
one.
It's
not
like
a
man
in
the
middle
okay,
no.
B
Can
I
talk
something
yeah,
please
we
are
doing
these
in
large
scale
for
you'll
understand.
We
have
a
vgi
services
on
top
of
convert
and
we
have
100
clusters
with
1250
nodes
each
cluster
and
to
do
that.
Also,
for
you
know,
we
update
on
weekly
basis
the
golden
plot.
We
call
Golden
image
of
Windows,
Linux
and
Mac
that
we
offer
as
part
of
our
services
completely
automated
via
cicg.
B
For
you
understand,
we
didn't
find
another
way
that
you
are
mentioning
about
certificate
and
things
for
us.
This
doesn't
fit
our
needs
because,
when
I
during
the
Clone
downloading
first
the
files-
this
is
take
a
lot
of
time
and
for
us
we
keep
the
latest
version,
always
in
the
local
storage
of
each
node.
B
For
you
understand
of
the
cluster
to
be
able
to
to
be
clone
it.
This
is
what
we
have
done
so
far
in
automated
way
way:
over
cicg
automation,
copies
for
small
deployments,
okay,.
A
Yeah
so
if
I
understand
you're
talking
about
the
like
the
VM
provisioning
flow
and
sort
of
the
maintenance
or
the
yeah,
the
creation
and
update
of
the
golden
images
that
are
used
as
sources
there
sounds
like
you've
got
a
pretty
good
approach
there.
A
Just
wanting
to
kind
of
make
sure
that
I
understand
you
right,
because
the
the
feature
that
we're
discussing
here
is
more
about
exporting.
So
this
is
wouldn't
be
about
provisioning,
but
more
like
if
you
have
a
VM
that
needs
to
move
to
a
different
cluster
or
to
be
backed
up
somewhere.
B
Now
I
understand
it's
like
the
server
that
I
have
running
here.
I
need
to
run
in
another
cluster
and
we
would
like
to
have
that.
Also.
Let
me
explain
you
why,
if
it's
possible
to
later
on,
we
are
working
in
another,
a
solution
to
do
that
to
do
live
migration
across
clusters.
B
Okay,
okay,
we
are
interested
to
to
interact
with
the
the
team
that
are
doing
that,
because
what
we
are
doing,
it's
already
available
to
lose,
live
migration
when
we
have
a
simple
CPU
attach
it,
but
we
are
doing
live
migration
with
GPU,
attach
it
on
the
VM
okay,
and
this
is
what
we
are
working
on
to
make
it
available,
because
this
is
not
available.
A
F
This
particular
export
the
VM
has
to
be
down.
The
export
actually
will
not
become
ready
if
the
VM
is
running.
A
A
F
I
I
was
just
you
know.
My
questions
were
just
around
I'm
I'm
thinking
about
putting
you
know
this
on
the
rest.
Endpoint
does
that
make
any
sense,
and
you
know,
is
there
anything
I'm
missing
or
is
there
anything
I'm
doing
that
I
shouldn't
be
doing
and
forward
the
use
case
where
I
want
to
basically
just
copy
a
VM
from
one
cluster
to
another
or
download
it
to
your
local
machine
or
some
other
storage
for
backup
purposes?
That's
that's
the
the
use.
A
Case
do
we
have
Alexander?
Do
we
have
a
design
dock
in
the
cube,
vert
Community
repo
for
VM
export?
Yes,
okay,
it
would
be
real
cool
if
perhaps,
if
you
could
link
it,
maybe
like
right
right
here
or
something
just
so
that
folks,
who
are
who
are
curious,
can
get
some
more
context.
Yes,
let
me
go.
G
The
the
doc
that
exists
is
mostly
about
just
the
raw
disk
export.
We
should
probably
expand
it
to
include
the
the
new
spec,
a
new
design
for
the
manifests
and
stuff,
but
definitely
maybe
we
can
add
it
to
that
doc,
but
as
far
as
live
migrating
between
clusters,
I
think
that's
really
cool
and
interesting,
but
I
think
you
know
the
whole
Lab
migration
requiring
shared
storage
and
and
all
that
we
may
have
to
start
investigating.
G
Basically,
the
other
mode
of
lab
migration,
where
you
copy
storage,
so
I
think
that
that
you
know
migrating
between
clusters.
Having
shared
storage
may
be
prohibitive
in
a
lot
of
cases.
A
Also,
another
tangential
Point
here
that
that
I'm
starting
to
think
about-
and
some
of
us
are-
is
disaster
recovery
and
doing
replication
to
a
cluster,
and
you
can
have
you
know,
synchronous
or
asynchronous
replication,
depending
on
what
your
use
cases
and
what
the
environment
or
the
distance
between
the
Clusters.
A
It's
not
exactly
live
migration,
but
it's
kind
of
like
another
flavor
of
this
multi-cluster
view
of
the
world.
So
a
lot
of
this
stuff
is,
you
know,
I,
think
becoming
topical
and
we're
definitely
interested
in
hearing
like
I'm.
This
is
one
of
the
great
things
about
I.
Think
about
this
call
is
to
get
to
just
get
a
view
on
how
different
people
are
are
using
using
Cube
vert
and
the
storage
parts.
A
B
This
is
me
again
Andre
from
the
desk
I
I.
We
are
doing
a
large
deployment
here.
For
you
understand,
the
size
is
125
000
notes
with
one
million
VMS.
B
Okay
with
that,
we
cannot
use
something
that
across
the
cluster
we
was
using
on
the
beginning,
Seth
and
Seth
they
do.
The
duplication
part
is
on
Alpha
stage.
We
cannot
use
it,
and
then
we
find
a
very
good
solution
that
is
also
open
source
call.
It
lean
store
that
a
few
other
gaps
missing
on
the
the
self
I
don't
know
why
converts
you
recommends
to
Yousef
until
it
has
production
solution
with
the
duplication.
This
is
no
go
with
Seth
for
me,
okay
and
also
the
second
item.
B
With
that
solution,
I'm
able
to
mimic
what
is
called
VMware
instant
clone
with
cloning,
with
copy
on
right
I
can
instantly
have
the
the
clone
of
the
master
VM.
B
A
Are
you,
are
you
using
the
smart
clone
function
at
well?
It
sounds
like
you're
not
using
stuff.
B
Anymore
from
linstore
itself,
okay,
they
have
the
amounts
Solutions
completely
different
from
from
convert,
and
that's
why
I
would
like
to
raise
a
rent
here.
Let's
move
away
from
Seth,
because
it's
not
feature
complete,
doesn't
have
lots
of
features
that
must
be
in
place
to
to
do
a
large
scale
deployment.
Okay,
yes,.
A
I
just
have
to
I
have
two
comments.
I
wanted
to
add.
So,
first
of
all,
with
with
smart
clone
with
stuff,
you
can
achieve
a
pretty
I,
don't
know
if
it's
I
don't
know
Michael.
What
would
you
say
like
it's
a
few
seconds
that
we
can
achieve
on
a
large
clone
and
it
doesn't
matter
how
large
the
volume
is,
because
we
we
cloned
from
snapshot.
A
So
it's
a
it's
an
efficient
clone
from
Seth's
point
of
view,
so
I
would
mention
that
and
secondly,
what
I
would
say
is
one
of
the
big
challenges
that
I
think
we
have.
Is
we
really
want
to
be
storage,
agnostic
and
essentially,
you
should
be
able
to
run
VMS
on
any
storage,
that's
supported
by
kubernetes.
We
really
tried
to
do
that.
A
B
It
has
also
another
feature
that
is
I
did
at
least
here
we
use
Ram
as
the
actually
disk
of
the
user,
where
the
the
PVC
are
are
actually
running
on
and
with
this
Ram
disk
I'm
not
able
to
do
this
from
Seth
and
RAM
disk
is
available
from
Lim
store.
I
can
use
actually
Ram
as
a
disk.
Then
I
have
one
million
times
faster
than
ssgs.
For
you
understand,
and
with
the
duplication
together,
we
can
have
like
a
100
VMS
with
300
gigs,
and
each
VM
has
100
gigs.
A
B
G
G
D
G
Wondering
when
you
were
when
your
phoning
was
very
slow
for
you,
I
wonder
if
perhaps
volume
snapshots,
weren't
configured
correctly
or
or
something.
B
A
Yeah,
it
would
be
interesting
to
hear
a
little
bit
more
about
your
architecture
and
your
platform.
So
if
you
guys
were
ever
interested
in,
you
know
sharing
more
about
that
or
doing
a
doing
a
talk
sometime,
you
know
at
Cube,
vert
Summit
or
something
about
how
you
have
your
stuff
set
up.
I
think
that
could
be
informative.
A
I
know,
it's
probably
also
your
secret
sauce
to
some
extent,
but.
G
B
F
We
could
look
at
maybe
adding
a
new
Lane
that
uses
Lin
store
for
testing
I'm,
assuming
you
would
like
that
so
that
you
have
a
slight
better
confidence
that
everything
still
works
when
we're
running
a
CI
Lane.
F
Yet
we
can
look
into
that.
A
Yeah
we
should
I
I,
haven't
really
been
exposed
to
this
provisioner
before
today.
So
it's
definitely
something
I'm
interested
in
in
seeing.
B
F
The
if
you
follow
the
link
for
their
GitHub,
they
actually
have
a
specific
Cube
word
to
go
to
examples.
Yeah
then,
there's
a
cube
for
under
there
and
there's
a
exactly.
A
A
Oh
looks
like
I'm
not
making
some
assumptions
here
based
on
names
which
Maybe
misguided
but
I
see
lvm
thin,
so
maybe
it's
using
lvm
behind
the
scenes,
cool
yeah
I'll,
definitely
be
taking
a
look
at
this
all
right.
Anything
else
on
this
topic
before
we
move
to
Richard's
questions,
I'm,
fine,
all
right.
C
Sounds
good
so
this
this
question
is
almost
stupidly
simple.
It's
like
the
opposite
of
the
previous
question
in
a
way
on
my
kcli
Cuba
instance,
which
has
got
one
node
I'm,
currently
creating
storage
by
basically
sshing
into
the
node,
creating
Loop
devices
and
then
creating
I
think
local
persistent
volumes
from
those.
Is
that
like
how
it's
supposed
to
be
done,
because
it
seems
like
really
weird
and
hacky
to
do
that?
Or
is
this
just
something
I'm
missing
about
how
to
do
this
properly?.
A
Yeah,
so
for
a
block,
sorry,
the
little
Echo
there
so
for
Block
storage
I
mean
that's,
definitely
a
a
bit
of
a
clunky
way
to
go
about
it.
There
is
a
local
storage,
provisioner,
that's
capable
of
discovering
block
devices
and
then
you
know
automatically
making
them
available.
A
I,
don't
remember
if
that
handles
recycling.
I,
don't
know
if
somebody
else
knows
about
that,
so
that
you
can
actually
use
them
and
give
them
back.
But
can
you
talk
about
the
local.
A
So
yeah,
so
lso
is
more
about.
Usually
it's.
It's
used,
for
example,
by
staff,
where
the
local
storage
operator
will
discover
the
devices
on
the
nodes,
make
them
available
as
PVS
in
a
storage
class,
and
then
a
storage
system
will
sit
on
top
of
that
and
consume
those.
This
is
how
stuff
works
to
actually
get
its
storage,
but
yeah
just
for
testing
the
block
devices
like
you're
doing
I
mean
that's
for
a
testing
environment,
probably
about
as
good
as
you're
gonna
get
I
would
say.
E
I
just
want
to
ask
Richard:
if
do
you
have
already
local
disk.
E
And
do
you
want
to
see
a
side
driver
or
is
will
be
enough
to
have
a
reference
to
the
PVC,
because
in
that
case
you
can
just
create
manually
a
PV
and
PVC
bound
together
and
then
your
VM
can
start
using
that
device.
I
can
send
you
an
example
if
you,
if
you
need
yeah.
E
C
A
I
think
you
know,
block
storage
is
a
little
bit
more
complex
and
if
you,
if
you're
okay
with
you,
know,
file
file
based
storage,
something
like
HPP
that
we
have
you
know
just
for
a
development
environment
might
be
simpler,
so
it
really
depends
on
what
you're
trying
to
do,
though,.
C
A
Yeah
so
HPP
I've,
we
found
the
beak
it's.
We
noticed
that
it
seems
to
be
used
by
people
who
are
doing
pocs
or
just
kind
of
getting
started.
It's
a
real
great
way
to
to
solve
the
problem
of
dynamic
provisioning
with
local
storage.
A
C
A
Right
so
I
have.
Are
there
any
other
topics?
We
want
to
discuss.
There's
a
couple
of
GitHub
issues,
one
of
the
things
that
I
thought
it
would
be
useful
to
do
in
in
this
call
is
to
periodically
go
over
some
of
the
issues
that
we
have
just
to
make
sure
that
those
are
getting
addressed.
So
I
thought
we
would
open
up
a
couple
of
these
and
I'll
do
that
now,
foreign.
B
I
would
like
to
ask
one
thing
before
that
Adam
sure
yep
is
there
anyone
I
know
that
is
a
repository
on
quietio
or
Ubuntu
Fedora
like
the
latest
image,
be
ready
to
be
copied,
but
no
one
is
doing
for
Windows.
A
Yeah
I
think
that
just
has
to
do
with
like
licensing
and
redistribution
rules
from
Microsoft,
so
they
have
I
mean
I,
think
they
have
some
cloud
images
that
are
behind
like
a
authentication
or
something
as
far
as
I
know.
Does
anybody
else
have
any
suggestions
or
thoughts
on
it?.
F
I
tried
looking
for
them,
I
couldn't
find
them
so
I
I
had
to
download
the
iso
and
install
it
manually.
Thank.
A
You
yeah
it's
definitely
something
that'd
be
nice
to
see
them
Distributing
more
regularly
in
this
modern
environment,
but.
B
B
They
built
with
that
script,
and
there
is
an
API
also
to
be
able
to
reach
it
that
behind
the
scenes
grab
the
files
direct
from
Microsoft.
That
is
no
problem
with
that.
Okay,
because
it's
public
information,
how
you
update
the
the
Mahal
Microsoft
is
updating
the
windows
itself.
B
B
I'm
gonna
add
who
is
responsible
on
to
to
do
the
quality
of
your
golden
images.
A
I
think
it
really
depends
on
the
distro.
You
know
if
certain
distributions
have
it
as
part
of
their
release
process.
A
Okay,
so
I
put
it
as
a
note
in
the
in
the
meeting
for
folks
to
they
can
check
that
out
thanks
for
sharing
that
all
right.
So
let's
go
to
the.
We
just
have
a
few
minutes
here
left.
So
the
first
issue
that
we
have
is
about
resizing
a
volume
after
the
Clone
and
so
right
now
we
we
require
the
size
to
be
the
same
as
the
source,
and
then
it
would
be
a
separate
action.
A
A
This
has
been
around
for
a
little
while
it's
getting
the
yeah
the
life
cycle,
rotten
and
stuff.
So
I
guess
the
question
is
well
I'm,
just
raising
awareness
here.
Is
there
any
discussion
about
it.
F
A
So
right
so
yeah,
obviously
a
resize
has
well,
maybe
not
so
obviously
has
two
parts
which
is
the
making
the
storage
container
itself
larger
like
and
then
how
the
how
the
the
virtual
machine
actually
uses
the
free
space
and
I
think
in
the
past.
Cdi
has
taken
a
hands-off
approach
to
that
second
piece,
because
we
can't
possibly
understand
how
a
user
is
going
to
want
to
make
use
of
the
additional
space.
A
It
does
seem
that
we
could
resize
the
PVC
though,
but
that's
less
useful
than
because
yeah
I
mean
you'll
have
a
bigger
disc,
but
the
image
will
have
to
be
sort
of
taught
how
to
use
that
extra
space.
So.
F
F
A
Okay,
I'm
not
seeing
that
here
it's
solanki.
F
F
Initial
I'm,
assuming
that
they're
doing
something
specific
for
their
for
dash
system,
but
you
know
I'd
like
to
see
it
because
I
I
do
think
it.
It
would
be
nice
to
have
something
that
actually
also,
you
know,
resizes
the
guest.
It's
just
making
sure
we
do
that.
For
you
know
all
the
major
operating
systems
Etc
be
nice.
A
Yeah
I
mean
I
just
think
it's
impossible
to
know
how
to
do
that
automatically
without
understanding.
What's
in
the
image,
that's
being
cloned,
so
there's
lots
of
different
ways
that
you
could
construct
that
image
and
we
don't
have
a
way
of
making
sure
we
do
that
that
right
for
everything.
G
F
F
G
F
G
D
G
One
point
in
time:
you
could
only
clone
to
the
same
size,
Target
and
then
you
have
to
resize
the
PVC,
but
I.
Don't
know,
I'm,
not
sure
what
what
exactly
is
being
asked
for,
because
you
you,
as
far
as
I'm
aware
you
should
be
able
to
clone
into
a
larger
PVC
and
then
yes,
the
disc
image.
The
the
image
file
will
then
still
be
the.
A
And
it
seems
that.
A
Will
handle
any
resizing.
A
And
do
this
and
then
see
if
we
get
any
comments,
yeah
I
think
the.
So
there
are
things
you
can
do
for
guest
resizing
like
partitions
and
stuff.
You
can
use
tecton
pipelines
or
other
kind
of
automation
that
takes
care
of
that
if
you
want
to-
and
that
gives
you
the
opportunity
to
have
something-
that's
a
little
bit
more
custom
to
the
environment
that
you
have
okay,
this
next
issue,
which
we
I'm
just
about
zero
time
to
discuss.
So,
let's
see
is
data
volume,
progress,
Improvement,
metrics
collection,.
F
Yeah,
so,
basically,
right
now,
the
controller
actually
connects
to
the
Importer
pod
directly.
To
get
the
you
know,
progress
updates-
and
this
looks
like
it's
instead
of
hitting
the
product
P
directly,
they
want
us
to
create
a
service,
and
then
we
can
hit
the
service
instead
of
the
pod.
F
Okay,
this
you
know
it's
apparently,
for
if
you
have
a
specific
Network
policies
that
don't
allow
you
to
directly
connect
to
pods.
A
Okay,
all
right
I'm,
sad
to
say
that
we
don't
have
time
to
dig
into
this
one
any
further.
I
wanted
to
take
just
one
second,
to
thank
everybody
for
joining
this
first
Sig
storage
meeting.
It
was
great
to
get
everybody
together
in
a
forum
where
we
can
discuss
storage
issues.
A
We
will
be
doing
this
on
a
bi-weekly
basis.
So
in
two
weeks
we'll
be
back
here
again.
I
would
love
to
see
you
guys
back
and
yeah.
Let's
definitely
put
those
agenda
items
onto
the
agenda
and
and
have
some
good
discussions
here.
So
thanks
again
and
have
a
great
week.