►
From YouTube: KubeVirt Community Meeting 2021-04-14
Description
Meeting notes: https://groups.google.com/g/kubevirt-dev/c/ai6JHpuMxhk
A
A
There
we
go.
The
meeting
notes
document
is
posted
to
chat.
So
if
you'd
like
to
follow
along
on
your
local
computer,
you
can
otherwise.
I
have
my
screen
shared.
A
And
if,
if
you'd
like
to
add
yourself
as
an
attendee
here,
that
would
be
nice.
A
And
we
start
off
with
introducing
making
a
little
introduction
for
new
people
who
have
joined
us
this
week.
If
we
have
anybody
new
and
you'd
like
to
say
hello,
now
is
your
chance.
B
Hey
my
name
is
lewis,
I'm
a
red
hat
employee,
I'm
just
interested
in
the
project.
There's
a
couple
use
cases
that
we
might
have
around
cooper,
mostly
around
isolation,
so
yeah,
I'm
just
kind
of
watching
just
seeing
what
it's
all
about
and
I'll
just
be
in
the
background
kind
of
lurking.
A
Cool
welcome
aboard
louis
many
of
us
are
our
red
hat
employees
and
we're
trying
to
become
less
red
hat
oriented
and
more
internet
focused,
but
always
welcome
to
have
more
people
join
us.
C
A
I
don't
think
that
anyone
else
looks
strange
to
me.
Okay,
let's
move
on
to
our
agenda
and,
of
course,
as
a
community
organizer,
I
usually
hijack
the
first
bullet
point.
A
Sorry
guys
some
late
breaking
news
yesterday,
our
our
sponsor
at
red
hat
asked
me
if
we
minded
releasing
our
slot
at
kubecon
eu
and
we
we
were
going
to
have
a
very
minimal
presence
at
kubecon
eu
anyways,
just
having
a
monitoring
a
chat
board,
we
weren't
going
to
do
any
official
presentations
or
or
one-on-one
meetings
and
breakout
meetings.
A
So
I
agree
to
release
our
slot
and
let
another
community
project
take
our
place.
A
A
A
So
it's
a
it's
sad,
but
if,
if
we
think
about
it
it
it
releases
our
time
to
really
focus
on
having
a
solid
presence
at
red
hat
summit
and
and
also
coffer
papers
has
opened
up
for
all
things
open
in
raleigh,
north
carolina
and
I've
been
to
all
things
open,
probably
five
times,
and
that
conference
is
awesome.
There's
lots
and
lots
of
people
go
to
that
conference.
A
It's
got
a
wide
range
of
of
topics,
and-
and
I
spoke
to
our
our
sponsor
about
presenting
at
that
conference,
and
he
was
surprised
to
hear
about
it
because
it's
not
something
that
we
would
normally
participate
in,
so
it
it
really
could
be
a
good
opportunity
for
us
to
talk
to
a
user
base
that
would
not
we
not
normally
get
to
talk
to
so
volunteers.
Need
it.
A
It's
it's
going
to
be
a
a
hybrid
event
for
sure
and
possibly
a
virtual
only
event
if,
if
things
go
bad
with
with
the
koba
19
spreading,
so
anybody
that's
interested
in
and
participating
and
developing
that
presentation
or
or
actually
doing
the
work.
A
D
Of
that
event,
I
haven't
been
before
I'm
curious,
how
we
would
I
guess
within
it
like
what
what
would
they
respond
to
well
there
as
far
as
presentations
go
with
keeper.
A
The
good
thing
about
all
things
open
is:
we
can
use
demos
that
we
already
have
in
inventory,
introduction
to
to
cover
and
like
gpu
usage
and
and
virtual
machines.
So
we
could.
We
could
really
start
from
from
square
one,
whereas
if
we
go
to
kubecon,
everybody
already
knows
that
we
can
run
virtual
machines.
So
we
really
have
to
do
some.
Some
pizzazz.
F
Let's
see
so
dave
to
answer
your
question,
having
looked
at
the
things
that
they're
interested
in
in
terms
of
potential
topics
cloud
is
on,
there,
kubernetes
is
on
there.
Containers
is
on
there.
So
as
kristen
said,
anything,
that's
like
a
direct
intro
to
keyboard
even
would
be
fair
game
because
it's
not
within
their
wheelhouse,
and
it
is
something
that
they
have
listed,
that
they're
interested
in.
A
Okay,
yeah
yeah,
that's
pretty
simple
and
I've
gone.
I've
gone
to
ben
suter
presents
at
all
things
open
every
year
and
if
you've
ever
seen,
him
speak,
you
walk
out
of
his
presentations
with
like
the
hairs
on
your
neck,
rising
like
they're.
His
presentations
are
so
awesome,
and
I
saw
is
the
the
big
black
blockchain
or
the
big
bitcoin
guy
he's
a
a
greek
fellow.
I
can't
I
can't
remember
his
name
now.
I've
got
his
book
and
same
thing
like
just
this
presentation
after
presentation.
That
is
so
amazing.
A
So
for
us
to
be
able
to
if
we
can
get
in
there,
we
got
big
shoes
to
fill.
A
F
About
two
weeks,
chris,
a
suggestion
on
that
front:
I've
seen
that
in
the
past
it
might
help.
If
we
track
a
spreadsheet,
or
I
mean
it
doesn't
have
to
be
complex,
but
just
so
we
can
coordinate
who's
trying
to
do
what,
in
terms
of
proposals,
sure.
A
I've
also
got,
I
created
a
board
in
github
also
for
community
activities,
so
that
might
it's
just
a
basic
trello
like
kanban
board,
so
that
should
help
us
keep
track
of
things.
Also.
A
Okay,
so
we're
all
done
with
that
topic,
we
can
move
on
to
shirley's
topic
regarding
cooper,
metrics.
A
So
david,
do
you
want
to
reach
out
to
her
offline
and
and
and
see
if
there's
anything
else
that
she
needs
to
get
this
pull
request,
moving
sure
and.
A
G
A
D
Oh,
I
already
merged
this,
so
what
she's
done
is
she
created
a
kind
of
policy
around
the
naming
of
our
metrics
and
how
we
introduce
alerts?
So
when
we
introduce
alerts
now
we
want
to
couple
that
with
a,
I
guess:
we're
calling
it
a
run
book
which
is
pretty
standard
which
kind
of
describes
okay,
so
this
alert
happened.
Here's
what
you
can
do
to
gather
information
and
potentially
solve
it
or
investigate
it.
D
Yeah
so
I
think
from
an
operational
standpoint,
this
policy
is
fantastic,
yeah,
absolutely
because
we'll
have
alerts
with
more
descriptive
like
priority
likes
or
severity,
and
then
somebody
can
look
up
the
alert
in
our
runbooks
repository
and
see
you
know
more
information
about
what
it
means
and
how
to
to
resolve
it.
So
it
didn't
really.
It
read
a
little
bit
more.
D
Like
a
proposed,
I
mean
a
not
a
proposal
but
more
like
a
guidelines
similar
to
what
we've
done
with
our
ci
guidelines
and
stuff,
like
that,
so
I
recommended
that
you
just
go
ahead
and
contribute
that
to
the
key
for
docs
repo
when
it
kind
of
popped
up
into
our
our
design
discussion,
and
I
agree
with
everything
she
had
in
there.
So
I,
the
other
big
notable
thing,
is
our
metrics.
We,
we
just
need
to
follow
the
conventions
that
already
exist
today
in
kubernetes.
D
So
when
we're
creating
a
metric
use
like
the
q,
vert
prefix
or
the
q
for
underscore
vmi,
it
was
specific
to
a
vmi
workload.
So
that's
the
big
thing.
We
need
to
use
those
prefixes,
it's
kind
of
a
name
space
for
us,
so
we
know
people
understand
like
what
the
metrics
are
related
to
and
when
we
pick
a
name
for
those
metrics
pick
a
name
that
it's
similar
to
something
that
already
exists
in
keyboard.
So,
for
example,
we
have
a
container
that
has
network
receipt
packets
total.
D
Then,
if
we're
going
to
have
something
similar
for
a
virtual
machine
instance
representing
how
many
packets
has
been
received,
go
ahead
and
say,
q
for
vmi
network
received
packets
total.
So
if
we
have
similar,
metrics
choose
the
name
that
already
exists
and
put
our
prefix
on
it
yeah,
but
the
rest
of
the
policy
has
already
been
merged.
You
can
look
at
our
docs
and
it's
called
monitoring
guidelines
and
that's
a
good
place
to
kind
of
get
yes
more
information
here.
A
D
It's
like
a
development,
it's
more
of
a
development
thing
so
for
people,
I
think
it
needs
a
user,
guys
presence
I'll
delineate
the
differences.
This
document
is
a
guideline
for
developers
who
are
creating
new
metrics
and
alerts.
It
tells
them
kind
of
what
we
expect
and
how
we
expect
it
to
be
structured
from
the
user
guide.
I
think
we
need
to
start
talking
about
if
we
haven't
yet
the
metrics,
the
kinds
of
metrics
we
expose.
D
I
don't
know
if
we
need
to
enumerate
all
of
them,
but
at
least
kind
of
describe
like
here's,
how
you
can
find
you
know
what
metrics
we
expose
and
here's.
You
know
what
happens
if
you
see
an
alert
you
can
go
to
this
directory
and
this
repository
to
get
run
books
and
things
like
that,
so
the
user
guide
should
be
from
the
perspective
of
an
operator,
whereas
this
document
she's
introduced
is
from
the
perspective
of
a
developer.
D
A
And
does
anybody
have
anything
to
talk
about
for
the
open
floor.
H
A
A
H
H
H
H
A
This
one
here,
that's
enable
nested
verb.
A
Well,
so
missing,
missing
how-to
for
nested
virtualization,
installing
kubert
and
sharing
virtual
gpu
to
guest
virtual
machines.
I
A
I
Means
that
vgpu
can
be
nested.
I
know
that,
like
pci
devices
can
be
nested
for
this
you,
you
will
need
the
the
the
l1
virtual
machine
to
have
a
virtual
immune
device
without.
I
H
No
problem
at
all,
but
this
is
something
that
we
are
working
on
on
bare
metal.
This
works,
let's
say
smooth,
but
on
gcp
we
are
getting
some
air
horse.
Some
time
works,
some
doesn't
sometimes
doesn't
work
and
you're.
J
I
You
would
like
to
look
so
we
have.
One
of
the
tests
in
our
keyboard.
Ci
is
is
basically
doing
something
like
that.
It
creates
a
vm,
a
vm
assigns
a
virtual
mmu
device,
and
then
we
reassign
it
to
a
test
vm
that
we
start
on
that
on
that
l1
node.
I
H
I
Keep
in
mind
that
every
every
layer
will
require
additional
memory
overhead,
so
I'm
not
entirely
sure.
A
Well
feel
free
to
reach
out
on
slack
and
and
work
with
us
in
in
real
time.
I
have
people
that
can
that
can
help
you
in
european
and
eastern
central
and
and
on
the
pacific
time
zone.
So
we
have
a
pretty
pretty
good
worldwide
time
zone
coverage.
A
A
G
A
Let's
make
sure
that
we
have
an
issue
for
this
actually
I'll
just
do
this
after
the
meeting.
Okay.
H
On
my
google
drive,
google
drive
is
not
updating
like
on
real
time.
I
don't
know
why,
on
the
document
on
the
document.
A
I'm
going
to
chalk
that
up
to
google
not
moving
the
bits
around
the
world
fast
enough.
A
H
H
H
A
Go
on
sorry
yeah,
we
can
you
and
I
can
stick
around
after
after
we
finish
the
meeting
and
and
figure
that
out
sound
good,
great
okay.
I
think
that
was
the
only
topic
we
have
here
in.
A
A
Sure
looks
like
it
yeah
we're
we'll
definitely
be
happy
to
help
you
with
figuring
out
that
gpu
problem
in
gcp,
okay,
anybody
else
for
the
an
open
floor
topic.
F
She's
unmuted,
but
we're
not
hearing
her,
so
she
might
still
have
audio
setup
issues.
K
K
K
K
K
K
And
the
second
part
is
basically
I'm
adding
a
new
command
to
your
ctl.
This
is
basically
running
a
pod
with
the
deep
just
first
tool
and
attaching
this
importance.
D
K
Yeah,
so
basically,
this
tool
is
creating
a
pot
with
the
image
that
I
was
talking
before.
So
he
has
that
as
the
all
day,
loop
is
the
first
one
inside,
and
this
pod
is
also
a
pvc
mounted,
and
I
am
basically
attaching
standard
input
standard
output
to
this
part.
So
this,
for
example,
is
very
useful.
If
you
want
to
debug
a
pvc,
so
maybe
you
create
some
wrong
partition
or
your
disk
is
not
booting.
K
And
you
want
to
check
what's
going
wrong
with
the
disk,
so
this
is
the
way
you
can
interactively
check.
It
is
gonna,
be
easy.
D
K
No
it's
so.
This
is
already
a
little
bit
what
cuber
cuber
tecton
tasks.
K
K
So,
as
a
first
step,
I
just
introduced.
Is
this
command
because
with
an
interactive
command,
I'm
checking
before
creating
the
pod
if
the
pvc
is
in
use
by
other
by
other
part?
K
D
K
Yeah,
the
use
cases
are
various
and
there
is
resize.
There
is
debugging.
There
is
also
customization,
but
before
also
digging
in
all
the
use
cases.
My
first
question
would
be
how
to
build
this
this
image,
because
you
are
already
using
actually
in
detect
on
repository.
K
If
you
look
at
the
docker
file,
you
are
basically
download
a
tarball
that
contains
the
kernel,
the
inita
key
and
the
root
for
the
deep
guest,
fest
appliance
and
the
problem
with
those
is
that
it's
like
two
years
old
and
it
doesn't
seems
the
virtualization
team
in
reddit-
is
going
to
publish
new
new
version.
K
D
How
often
I'm
trying
to
see
where
the
so
I
see
all
these
rpms
are
we
sourcing
and
your
pr
lib
guests
from
the
fedora
live
guest
fs
from
the
fedora
repos.
K
Yeah,
if
you
look
at
the
pr,
you
can
see
that
I
have
in
the
hack
directory
dockerfile
it's
necessary
to
have
a
docker
file,
because
we
need
to
run
some
post
scripting
install
yeah.
So
this
is
basically
in
in
the
main
fight
I'm
extracting
the
appliance
yeah.
So
that's
exactly
the
main
file.
I'm
talking
about.
D
K
So
yeah
there
you
can
see.
I
am
running
this
deep
classifies
make
fixed
appliance,
and
this
has
to
be
running
a
docker
file
because
with
bazel
the
postscript
steps
are
not
run
so
the
appliance
is
not
actually
generated.
K
K
K
Just
scroll
a
little
bit,
there
is
a
bean
basin
yeah.
There.
D
Is
there
a
way
to
avoid
that
dnf
install
of
libgas
fs
in
that
docker
file.
D
K
K
I
don't
want
to
steal
all
the
entire
time.
I
just
wanted
to
point
out
that
the
appliance
the
new
gustafest
appliance
is
the
common
place
of
this
work
and
the
tecton
because
detect
on
repositories
using
a
pretty
old
sfs
appliance.
D
Word:
okay,
so
just
taking
a
step
back
from
all
of
this,
I
I
think
that
makes
sense
to
do
that
in
kuvert.
If
we
have
solid
user
stories
for
how
the
kubert
repository
is
going
to
utilize
this
appliance.
So
if
we
have
like
vert
control
executing
this
for
some
reason
or
it's
used
in
one
of
our
workflows
within
qvert
or
or
whatever,
we
can
justify
that
that
we
can
justify
shipping
and
publishing
this
appliance.
K
D
Okay
and
if
it
doesn't
make
sense
in
the
kuver
kubert
repo,
I
think
there's
a
good
chance
that
cdi
might
accept
something
like
this
easier,
but
you
know
we'll
see
but
yeah.
I
think
that
there's
definitely
a
way
for
us
to
share
the
same
appliance
across
multiple
repos
and
do
that
in
a
way
that
gets
an
a
regular,
regularly
updated
appliance
with,
like
sync
to
some
schedule
that
we
we
already
produce
releases
for.
K
A
A
So
hopefully
this
is
a
start
in
that
direction.
K
Yeah,
actually
installing
some
additional
kernel
module
or
a
tool.
That's
one
of
the
use
cases.
A
D
M
Thanks
all
right,
we
have
a
lot
to
go
through
and
we
have
12
minutes.
So
let's
start
should
report
error
when
bmi
course
is
less
zero
or
equal
to
zero.
If
somebody.
J
D
Yeah,
okay,
if
somebody
explicitly
said
zero,
I
have
no
idea
what
I
don't
even
know
what
that
means.
D
I
We
have
a
validation
for
this,
at
least
we
will
override
it
with
the
with
one.
I
think
so.
D
I
F
Yeah
we
we
actually
do
in
our
mutating
web
hook,
if
course
is
equal
to
zero.
We
just
assume
the
default.
Okay
and,
and
the
reason
that
works,
I
think,
is
because
it's
not
a
pointer,
so
the
default
zero
already.
I
Yeah
and
then
in
the
in
the
template
in
service
template
where
we
create
the
code,
we
actually
also
have
a
some
kind
of
a
validation
that
it's
not
zero,
because
otherwise
we
won't
calculate
the
memory
correctly
yeah
I
mean
I'm
pretty
sure
we
do
the
right
thing,
but.
E
M
M
Okay,
the
second
one
I
cannot
create
vms
in
open
ci
using
gcp,
so
all
our
gcpds
are
failing
and
they
fail
to
create
vms
and
the
error
message
is
synchronizing.
The
virtual
machines.
That's
failed
because
failed
to
configure
vmi
network.
D
D
M
M
M
M
Seriously,
yeah
a
surfer
skipping
that
is
anybody
familiar
with
block.
I
o
dune
element.
I
I
I
Maybe
there
is
some
kind
of
a
logic
that
we
can
implement
to
make
this
right.
The
question
with
the
problem
is,
you
know,
maybe
we
can
ask
if
they
been
able
to
tune
this
correctly
and
and
then
from
this
we
can
derive.
M
So
you
want
to
see
if
the
poc
using
this
to
see
if
it
really
solves
the
issue.
I
G
Okay,
thanks.
M
All
right,
the
same
person
reported
this
one
unable
to
compile
and
run
and
convert
locally
for
development,
and
I
checked
it
offline.
I
think
it
has
some
direction.
M
D
Yeah,
this
should
be
resolved
so
they're
running
on
open
shift
with.
I
believe.
D
Be
some
way
would
work
and
kvn
device
doesn't
exist.
I
believe-
and
it
causes
work
handler
handler
to
crash,
because
the
node
labeler
can't
retrieve
information.
It's
been
addressed.
M
D
Can
you
look
for,
let
me
see
what's
his
name,
his
name's
carl,
but
I
forget
his
last
name:
yeah.
M
D
D
See
if
it
got
into
0.40,
it
should
have
been
back-ported.
M
M
Awesome:
okay
data
volumes
has
plugged
to
vm
do
not
update
their
status
to
running
and
alexander.
G
G
M
This
one
gaming
on
convert,
I
was
looking
at
spinning
up
kubernetes
cluster
and
separate
gaming.
Pc
does
anybody
know
can
confirm
they
have
been
able
to
game
on
convert
instance
robotic.
Have
you
tried
with
a
posture
cpu
posture,
our
gpu.
I
D
Oh,
can
you
add
the
full
key
vert
dev
mailing?
Oh.
M
Yeah
yeah,
I
sent
him
to
the
floor
all
right
and
we
are
on
top
of
the
hour,
so
I
think
we
can
rip
it
up.
We
made
the
week
into
the
bus,
so
I
think
that's
pretty
good.
A
A
It's
not
really
not
really
effective
in
in
communicating
that
that
that
forum.
A
Okay,
top
of
the
hour
meeting
is
over.
Thank
you,
everybody
for
for
joining
us
and
we'll
see
you
on
in
slack
and
and
the
mailing
list,
and
otherwise
we'll
see
you
next
week.