►
From YouTube: KubeVirt Community Meeting 2021-06-23
Description
Meeting notes: https://docs.google.com/document/d/1kyhpWlEPzZtQJSjJlAqhPcn3t0Mt_o0amhpuNPGs1Ls/edit#heading=h.u43ltfrghjfd
A
Hello,
everybody
I'm
chris
caligari
and
I'm
your
host
for
for
project
kubert,
and
this
is
our
weekly
meeting
to
discuss
any
topics
that
we'd
like
to
bring
in
front
of
the
community.
A
I'm
gonna
post
the
link
to
our
meeting
notes
to
oh,
no,
not
that.
Okay,
sorry
about
that
meeting
notes
are
posted
to
chat.
A
And
do
we
have
any
new
community
members
this
week?
That
would
like
to
say
hi.
B
Okay,
so
I
have
shared
an
actual
email
with
all
the
information
as
well.
Probably
maybe
I
should
share
it
too,
but
it's
a
link.
The
this
resize
one.
B
So
we
were
interested
in
making
data
volumes,
expand,
clone
a
bit
faster
and
just
to
check.
You
can
hear
me
right,
yep,
okay,
cool,
so
we
were
interested
in
making
it
faster
and
one
of
the
things
that's
required
is
expanding
the
disk.image
file
in
some
scenarios
and
we're
going
to
make
that
into
a
feature
where
you
can
expand
the
data
volume
offline.
But
I
was
going
to
present
all
of
that
yesterday
last
week,
but
I
realized
I
don't
have
an
explanation
for
why
it's
not
online.
B
So
it's
online
now
online
meaning
it's!
It
works
running
yeah,
it's
gonna
the
implementation,
I'm
working
on
in
keyword
code,
so
keep
coordinate.
It
says
feature
guess
it's
background,
expand
persistent
volume,
expanding
use,
persistent
volume!
You
can
expand
a
pvc
what's
missing
for
the
vm
to
take
advantage
of
it
one.
If
it's
a
distant
image
file,
we
need
to
increase
the
size
of
it
and
if
it's
and
regardless,
whether
it's
a
blog
or
a
file
system,
we
should
also
notify
the
vm
about
this.
So
I
have.
B
The
sizes
of
the
pvcs
and
the
virtual
articulator
tries
to
see
what
the
expected
size
should
be
sends
that
to
the
field
launcher
with
the
virtual
machine
options,
it
compares
it
to
the
actual
size
and
expands
if
it's
not
big
enough
and
sends
the
block
resize
command
to
the
vm.
With
the
leverage
for
a
blocker
size
it
will.
B
Let's
exchange
okay
pass
around
the
change
disks
and
send
the
block
resize
to
the
vm.
So
I
have
some
code
that
works
there.
It's
not
full
complete,
there's
no
working
progress,
pr,
there's!
No
things
might
change
a
little
bit.
I
hope
this
is
a
good
way
to
pass
information
around
with
the
virtual
machine
option
between
the
built
under
wheel
launcher
some
pitfalls.
B
I
have
only
implemented
the
to
these
dot
image
profiles,
not
to
qca
too,
because
a
few
image
image
is
not
like
running
on
a
file
that
is
being
used
by
qmu
and
a
theoretically
should
be
solvable,
and
I've
only
tried
this
with
rook
7.
B
It
does
online
resize
and
I
don't
know
how
common
or
uncommon
offline
research
expansion,
but
not
in
use,
and
maybe
I
should
address
that
one
two
weeks
like
a
status
for
the
condition
for
the
vm
bmi
that
if
you
stop
and
start
it
again,
then
it
would
resize
the
file
system.
Does
that
this
that
image?
For
you
that's
my
plan
with
some
implementation.
B
C
Yeah,
I
think
this
is
a
great
feature.
My
question
is
so
generally
like
the
so
the
okay.
So
my
guess,
my
first
question
is
about
the
api.
C
So
if
you
want,
if
a
user
wants
to
resize
or
expand
a
disk,
do
they
expand
the
pvc
directly
or
via,
like
a
data
volume,
template
or
the
data
volume.
C
Okay,
so
then
the
so
the
data
volume
templates
and
the
data
volumes
stay
unchanged,
and
so
and
then
the
vert
handler
will
be
watching
all
pvcs
and
it
will
detect
when
the
size
changes
and
update
the
virtual
machine
options
for
vert.
D
C
Okay,
yeah,
I
think
I
think
that's
fine.
I
guess
the
one
thing
to
know
in
perhaps
it
makes
sense,
it's
just
the
kind
of
division
of
labor
between
data
volumes
and
pvcs.
So
right
now
we
have
this
perhaps
interesting
behavior,
where,
if
a
pvc
is
it's
owned
by
a
data
volume
is
deleted,
it
will
get,
you
know
repopulated.
C
C
Yeah-
and
I
guess
one
thing
that
going
along
with
this
is-
and
maybe
community
has
input
on
this,
so
this
will
essentially,
you
know,
trigger
this
resize
when
a
vm
is
started
so.
C
Imagine
a
flow
so
we're
we're
really
trying
to
evangelize
this
golden
image
use
case
where
you're
cloning
from
a
pvc
and
basically
one
issue
we
have
now
is
if
you're
cloning
from
a
smaller
pvc
to
a
bigger
pvc.
You
know
the
disc.img
is
still
the
same
size.
So
there's
just
a
lot
of
unused
space
and
you
know
the
question
is:
do
we
want
to
also
have
an
expansion
phase
during
this
cdi
clone
process,
or
is
it
enough
to
just
wait
until
the
vm
has
started
and
do
it
then.
E
I
mean
I
can
weigh
in
my
opinion,
which
is
that
I
think
it's
probably
more
consistent
to
just
have
have
it
always
happen.
E
Is
that
vm
startup,
rather
than
having
multiple
flows
where
it
has
to
be
handled.
C
Okay,
that's
definitely
makes
life
easier.
G
How
do
we
know
how
big
to
make
the
file
if
we're
talking
about
the
file
system
one?
So
we
see
the
size
of
the
pvc?
How
are
we
figuring
out
what
that
corresponds
to
for
the
file
like
do
we
understand
the
overhead
and
everything
I
guess
of
the
file
system
and
all
of
that.
B
So
this
is
like
an
unfortunate
I
probably
should
have
written
down
on
the
pvc
which
overhead
it
uses.
I'm
gonna
have
some
kind
of
eq
compatibility
code
to
not
trample
over
existing
things.
That
reserve
some
overhead,
but
I
tend
to
have
an
annotation
on
the
pvc.
If
it's
preserving
some
of
the
space
for
overhead,
then
our
do
the
calculation
in
red
handler
based
on
some
annotation.
G
G
A
E
Worthwhile
to
note
it's
a
good
question,
but
I
see
that
aliche
is
on
the
call
and
she's
working
on
some
live.
Guest
fs
integrations
for
convert
and
it
will
be
possible
in
the
future
to
actually
have
a
pipeline
where.
F
E
E
You,
of
course,
can't
have
live
guest
fs
going
in
there
and
making
changes.
So
it
is
a
good
thing
that
we
should
think
about,
because
it
would
be
nice
to
be
able
to
actually
create
some
sort
of
a
pipeline
or
job
that
could
could
do
that,
because
live
guest
at
best
does
know
how
to
to
enter
a
guest
and
expand
file
systems.
G
Oh
interesting
so
for
offline
resize
is
lib
guest
fs,
always
the
thing
that's
being
used.
E
A
E
I
mean
you
could
use
that
we
don't
have
any
kind
of
integrations
for
it,
but
yeah
I
mean
like
if
without
automatic
processes
you
could,
I
guess,
expand
your
pvc
and
then
you
could
attach.
You
know
manually
construct
a
pod
that
has
lid
guest
fs
in
it
and
and
have
it
do
some
things.
I
believe
that
would
be
possible.
F
Yeah
I
mean
you
can
just
use
v3
size
to
to
do
that
and
use
don't
have
a
task
in
tectum
pipeline.
It
do
the
resize,
but
that's
of
course
possible
to.
F
I
E
Doing
a
live
resize
operation,
it
can't
have
coordination
with
the
operating
system,
but
if
you
were
doing
some
sort
of
offline
operation
you
could,
but
I
think
so.
E
I
would
wonder
yeah
if
the
vm
is
not
running
and
the
I
guess
the
cube
would
still
be
monitoring
the
pvcs
and
triggering
the
resize
operations
and
then,
like
we
discussed
earlier,
they
would
they
would
the
disk
image
file
and
these
things
would
be
updated
at
the
time
the
vm
is
starting.
E
G
Kind
of
locking.
E
Mechanisms-
and
I
guess
you
guys-
could
stay
tuned
for
some
more
info
from
alicia
about
that
issue
as
well
she's
looking
at
it,
we're
trying
to
find
a
way
to
basically
have
some
sort
of
system
to
to
indicate
that
a
certain
set
of
pvcs
are
in
use
by
you
know
either
either
a
tecton
pipeline
or
by
cubird
itself,
but
not
both.
G
Interesting
yeah,
this
is
pretty
cool.
I
am,
I
guess,
conflicted
about
directly
working
on
the
pvc
rather
than
working
on
the
data
volume.
Just
because
of
the
the
source
of
creation
would
be
the
data
volumes.
That's
the
thing
the
user
has
created,
so
I
would
kind
of
expect
that
to
be
the
interface
they
go
through,
but
I'm
not
necessarily
I
don't
know.
I
guess
it's.
The
best
thing
I'd
have
to
I'd,
want
to
see
a
little
bit
more
before
we
completely
commit
to
somebody
mutating
letting
a
user
mutate.
E
E
Object
during
provisioning
and
then
pvcs
become
the
important
object
during
consumption,
and
this
is
one
of
the
reasons
why
code
was
added
so
that.
E
To
specify
data
volume
anymore,
if
the
pvc
was
actually
provisioned
by
one,
because
we're
trying
to
make
it
less
important
to
users
of
a
vm,
that's
already
been
provisioned
for
them
to
have
to
worry
about.
The
data
volume
object
like
maybe
even
someday
that
object
could
be
detached
from
the
pvc
once
it's
been
ready,
made
ready,
and
then
you
just
use
it
like
a
regular
pvc
and
sorry.
I
just
basically
asked
you
to
contribute
maya
and
then
like
talked
anyway.
A
Yeah,
so
from
a
community
perspective,
the
only
thing
I'd
ask
is
to
follow
the
the
design
standard
that
we
recently
built
and
get
this
into
the
design
folder.
A
Get
this
documented
and
then
every
like.
We,
we
have
the
the
recording
of
everybody's
comments,
but
it
would
be
really
good
to
have
them
tracked.
The.
A
A
A
B
Okay,
sorry,
I
had
some
microphone
issues.
I
wanted
to
say
that,
while
this
is
propagated
through
the
pvc,
there
is
no,
it
doesn't
limit
the
ability
to
start
this
by
changing
the
data
volume.
It's
just
that
if
this
were
to
happen,
it
would
happen
in
cdi
in
an
independent
pr.
G
Okay
makes
sense
one
of
the
things
that's
in
the
back
of
my
head-
that
I
guess
I
haven't
even
said
yet
is
I'm
concerned
about
vert
handler
watching
all
pvcs
because
that's
could
be
potentially
a
lot
of
resources
and
then
it's
like
every
node
would
be
watching
a
lot
of
resources
as
well.
So
it's
compounded,
so
I
was
trying
to
think
of
ways
to
reduce
that
and
one
of
the
things
I
was
thinking
I
was
like.
G
Well,
we
just
watched
pvc,
our
data
volumes
instead,
but
that
won't
work
because
it's
totally
possible
to
just
use
a
pvc
without
a
data
volume
and
that's
where
my
train
of
thought
was
going.
So
that's
a
separate
train
of
thought,
but
somehow,
when
we're
doing
this
design,
I'd
like
for
us
to
think
of
a
way
of
isolating
or
figuring
out
some
way
to
narrow
how
many
objects
we're
watching
on
vert
handler
so
maybe
add
a
label
to
a
resize
pvc
that
actually
needs
to
be
worked
on
by
bird
handler.
G
So
we
don't
have
to
watch
all
of
them,
but
we
get
alerted
when
there's
one
that
needs
to
be
resized
on
the
exact
node.
That
needs
to
be
resized,
something
like
that.
I
know
that
might
sound
like
an
early
optimization,
but
we
can
quickly
get
into
some
some
trouble
if
we
like,
for
example,
have
a
100
node
cluster,
that's
watching
all
pvcs,
so
everyone
knows
watching
all
pvcs
and
there's
thousands
and
thousands
of
pvcs.
I
can
get
kind
of
messy,
so
something
else.
B
G
No,
that
could
be
a
problem
that
might
actually
I
might
go
away.
G
Let
me
all
right
well
we'll
take
we'll
yeah
catch
up
on
this
offline
yeah,
so
just
the
background
quickly
we're
doing
some
scale
and
performance
testing
and
profiling,
and
that's
probably,
if
we're
doing
anything
inefficient
there,
it's
probably
going
to
get
caught
and
addressed
fairly
soon.
So
I
don't
know
if
we
can
depend
on
that.
We'll
we'll
see
what's
actually
happening
behind
the
scenes
there.
G
A
Yeah
definitely
all
the
more
reason
to
to
get
this
into
the
proper
design
directory.
B
A
G
I
don't
have
anything
just
look
forward
to
having
more
discussion.
I
guess
out
of
band
of
this
meeting,
so
more
design
discussion.
It
looks
really
good
or
it
looks
like
a
really
cool
feature.
I'm
also
interested
in.
I
wouldn't
want
to
discourage
just
working
on
a
poc
to
prove
out
the
technology
immediately
as.
A
Well
definitely
important
discretizes,
just
meat
and
potatoes.
That's
the
sad
meaning
you
gotta
have
some
kind
of
feature
in
there.
E
Just
something
to
hit
that
hit
my
mind
and
then
let's
move
on.
If
no
one
else
has
anything,
but
we
should
consider
the
implications
of
snapshots
when
it
comes
to
resizing.
So
if
you
have
the
size
of
a
pvc
changing
between
snapshots,
it's
going
to
be
a
corner
case
that
we
addressed
in
rev
as
well
so
or
an
overt.
So
just
keep
that
in
mind
in
the
design.
A
A
Still
still
and
an
open
agenda
and
floor
does
anybody
else
have
anything
before
I
talk
about
event.
A
A
Going
once
twice:
okay!
Well,
I
have
good
news:
everybody.
We
just
got
a
notice
from
all
things,
open,
2021,
that
our
paper
has
been
accepted,
so
stu
and
I
will
be
going
in
person
given
that
nothing
drastic
happens
with
with
the
ongoing
coba
19
pandemic,
we'll
be
going
to
raleigh
and
and
doing
our
our
presentation
of
an
internet
distributed
kubernetes
cluster
with
coobert
on
raspberry
pi's.
A
Yes,
I've
just
thought
of
something
stu.
All
of
our
systems
will
be
at
our
house
while
we're
in
raleigh.
So
that's
correct.
Oh.
A
We'll
be
calling
up
our
wives
and
or
our
kids
to
say,
quick
re
reboot
that
raspberry
pi
just
pull.
A
A
So
I
I
suggest
that
we
create
a
working
meeting,
so
we
stay
on
track
with
getting
our
demo
built.
A
I
A
G
A
A
I
We
can
we
can
yeah,
we
can
figure
that
offline.
I
think,
on
slack.
A
A
I
built
this
beautiful
rack
out
in
my
garage,
took
a
serious
network
outage
on
my
on
my
home
network
and
impacted
my
kids,
remote
learning
remote
school
and
got
everything
working.
A
Okay,
pull
requests
that
need
attention.
Do
we
have
anything
that
has
been
lingering.
J
No,
I
think
we're
we're
good
there.
We
have
a
pr
on
the
way
and
we
had
a
couple
of
discussions
on
the
pr
itself.
So
it's
fine.
A
Okay,
thank
you:
how
to
keep
tcp
sockets
connections
alive
after
live
migration.
This
is
a
topic
that
came
up
on
slack
also.
A
Is
anybody
available
to
to
comment
on.
A
A
Okay,
I'll
I'll
just
ask
him
to
create
an
issue.
Please.
I
I
A
A
G
K
So
that
this
was
my
email
that
I
threw
out
here,
there
was
a
issue
that
4207,
where
someone
had
put
in
an
issue
saying
hey:
could
we
use
the
default?
Vnc
client,
that's
built
into
mac,
so
I
did
a
little
bit
of
research
on
it
and
the
the
answer
basically
comes
down
to.
K
No,
you
can't
use
it
because
the
default
client
that's
built
into
the
os
x
absolutely
requires
a
password
and
you
can't
seem
to
bypass
it
at
least
nothing
that
I
could
find
allowed
you
to
bypass
it,
and
this
has
been
seen
by
some
other
projects
where
they
had
the
same
sort
of
issue.
So
I
mean
quite
frankly,
I
think
probably
the
thing
could
probably
just
be
closed
out
is
can't
do
it,
but
I
don't
know
I
just
I.
K
I
threw
the
the
threw
it
out
there
to
the
mailing
list,
because
I
didn't
really
see
anything
on
the
issue
when
I
had
gone
through
and
done
any
research
on
it.
G
K
K
But
I
also
wasn't
quite
sure
you
know,
I'm
more
than
willing
to
kind
of
maybe
dig
into
it
a
little
bit
more
and
see,
but
I
would
need
a
little
bit
of
guidance
in
where
to
kind
of
look
at
or
what
you
know.
What
generates
that
the
actual
vnc
server?
How
we
do
that
and
whether
or
not
we
could
put
a
password
into
it.
K
I
G
Would
just
look
into
how
to
set
a
password
using
the
domain,
xml
invert
for
vnc,
and
then
we
would
just
put
in
a
secret
or
something
and
attach
it
to
the
vmi.
If
that
was
desired,.
I
K
So
I
guess
the
other
question
would
be.
Is
that
something
that
we
want
to
do
like
across
the
board
is
have
some
type
of,
because
I
this
would
only
based
on
my
research
on
it
that
this
would
only
be
fixing
a
problem,
for
you
know
os
x
and
using
the
default
client,
because
the
default
client
absolutely
requires
a
password.
So
if
you
you
know,
it
would
be
one
of
those
things
where,
if
you're
in
that
very
situation,
then
we
start
assigning
passwords
to
all
your
vms
and
then
you
can
use
it.
K
It
seems
kind
of
like
a
rough
rough
situation,
almost
seems
like
it's
better,
just
to
use
a
third-party
client
that
doesn't
require
a
password
unless
we
think
there's
a
real
need
to
start
locking
down,
or
you
know,
applying
a
password
to
all
the
vnc
connections.
Wouldn't.
G
A
K
K
You
know
adding
adding
a
password
to
the
vnc
server.
I
think
would
be
a
separate
it's
a
more
of
a
separate
issue
or
a
separate
request
than
what
this
is
here,
because
this
is,
you
know
very
specific
to
using
that
built-in
vnc
client
on
mac.
K
You
know
it
definitely
could
see
a
new
feature
request
or
a
new
issue
saying
hey.
I
would
like
to
be
able
to
put
passwords
on
my
vms
or
I'm
sorry
put
passwords
on
my.
You
know
vnc
server
for
my
vm,
but
I
think
that's
a
separate
issue.
Then
this
one
right
here.
A
I
also
noticed
that
this
this
issue
is
from
september
18
2020.,
so
it's
been
sitting
around
for
quite
a
while.
A
If
any
of
you
got
anybody
knows
him,
I.
H
K
Possible
yeah,
I
think,
like
I
say
I
think
more,
you
know
there
might
be
a
different
issue
or
a
different
feature
request
here
to
actually
be
able
to
put
passwords
on
vms
or
vmis
that
are
created,
for
you
know
more
secure
access,
although
I'm
wondering
how
that
would
work
with
the
the
the
open
shift
ui,
but
I
think
that's
a
separate
request
from
what
this
is,
which
is
really
just
you
know.
K
Can
we
support
that
built-in
client
in
the
the
mac,
and
I
think
that
the
answer
at
least
by
default
is
no,
but
that's
just
my
thought,
like
I
said
it's
more
that
this
is
almost,
I
think,
like
you
know,
issue
scrubbing
or
something
like
that,
just
closing
something
out.
If,
if
we
really
don't,
you
know
if
it's
not
doable
now
without,
you
know,
potentially
affecting
every
other
user's
interaction
with
it.
H
I
think
that's
fair,
I
I
would
add,
as
you
close
the
issue
also
give
a
hint
that
maybe
this
is
the
issue
you
ought
to
open
asking
for
a
password
feature.
K
Yeah,
I
can
certainly
I'll
I'll
go
in,
and
you
know
just
add
a
little
bit
more
about
that,
and
you
know
mentioned
that
it
might
be
possible
to
handle
it
through
that
way
and
then,
but
I
don't
think
I
have
any.
You
know
ability
to
actually
close
it
out.
It
was
more
just
a
helpful
hint
that
might
be
able
to
close
out
an
issue
at
this
point
once
you
know,
everybody
agreed
on.
K
A
A
A
You
guys
want
to
type
something
in
there,
yeah
you're
looking
for
me
to
just
type
something
in
around
yeah.
It
could
put
a
comment
in
there
about
maybe
creating
an
issue
about
the
vnc
server
and
having
an
option
to
enable
password.
K
Yeah
I
didn't
catch,
there
was
somebody
who
actually
was
just
looking
at
the
qmu
stuff.
I
don't
know
if
you've
got
a
link
that
you
could
put
in
there
for
in
that
doc.
For
you
know
what
that
feature
is
or
what
that
that
callout
was
just
as
a
reference
that
might
be
helpful,
or
I
can
look
it
up
and
find
it
okay.
I
can
have
that.
A
K
K
The
only
other
thing
you
know
I
was
just
gonna,
I
didn't
wanna.
I've
been
editing
these
things
before
where
text
is
changing
around
at
the
beginning,
so
I
was
gonna
put
something
in
there
about
the
the
the
default
client
in
mac.
Oss
only
supports
password
vnc
servers,
so
it's
not
something
that
can
be
supported
right
now,
but
and
then
leave
the
rest
of
it
in
there,
because
I
think
that
fully
addresses
the
the
issue.
A
Mark
we
can,
how
about
you
comment
on
that
issue
offline
here,
so
we
can
move
on
yeah
yeah
by
all
means
sure,
thank
you,
and
by
moving
on
I
mean
it's
7
50
a.m
and
we've
about
run
out
of
time.
A
So
I
suggest
we
skip
bug
scrub
this
week,
which
we
did
that
last
week
also
so
next
week
we
should
focus
on
a
good
bug
scrub
rather
than
get
down
in
the
get
deep
into
any
agenda.
A
A
Yep
all
right,
cheering
and
clapping,
so
we
have
it
I'll
return.
Nine
minutes
to
you.
Thank
you
for
joining
us
this
week
and
have
a
good
week
we'll
see
you
next
time.