►
From YouTube: KubeVirt Community Meeting 2021-10-06
Description
Meeting Notes: https://docs.google.com/document/d/1kyhpWlEPzZtQJSjJlAqhPcn3t0Mt_o0amhpuNPGs1Ls/edit#heading=h.psxl2bj2v84l
A
Hello:
everyone
welcome
to
this
week's
coobert
community
meeting.
It's
wednesday
october
6th
2021.
A
I'm
your
host
chris
caligari,
and
this
is
your
opportunity
to
talk
to
core
developers
and
other
users
of
the
project
with
anything
you'd
like
to
talk
about
regarding
using
or
developing
for
coovert.
A
I'm
going
to
paste
the
meeting
notes
into
chat.
Everybody
could
open
that
up
and
add
your
attendance.
We
do
track
that
and
please
add
any
items
you'd
like
to
speak
about
into
the
agenda
or
open
for.
A
I
think
got
a
few
folks
here
that
I
don't
recognize.
Would
you
like
to
say,
hi
and
and
introduce
yourself.
C
Okay,
I
joined-
I
did
a
few.
This
is
my
second
call.
Actually
I
joined
a
few
weeks
ago,
my
first
one
and
I've
been
using
covert
for
to
run
I'm
actually
so
I'm
using
it
to.
I
have
a
service
that
runs.
Our
pen
builds,
so
I'm
currently
using
it
to
bootstrap
virtual
machines
on
demand
to
run
those
builds
and
get
the
resulting
rpm,
artifact
and
upload
somewhere
else
in
a
kubernetes
operator.
C
I
actually
had
a
sorry
I
I
was
going
to
put
in
the
chat,
but
since
I'm
talking
here
so
my
team
has
an
api
that
creates
virto
machines
on
on
a
vm
vendor.
So
I
was
wondering
how
hard
it
would
be
to
create.
I
don't
know
if
it's
a
would
you
call
it
convert
plugin
or
something
like
that.
That
could
create
a
virtual
machine
on
that
vendor
rather
than
on
kvm.
If
that
would
be
even
possible,
because
I
know
another,
there
there's
aws
support
or
other
runtimes.
A
Yeah
we
usually
for
the
for
deploying
to
cloud
providers.
We
usually
just
do
embedded
virtualization
okay,
and
we
have
several
large
customers
doing
that
exact
thing
with
some
of
aws
larger
offerings
and
even
their
bare
metal
offerings.
A
That
makes
sense
yeah,
that's
definitely
a
great
topic
for
further
on
in
the
agenda.
So
if
we
don't
have
anything
else,
let's
talk
about
that
more.
If
anybody
else
has
more
opinions
on
that.
A
What
does
everyone
think
about
using
alternative
virtualization
aside.
A
Yeah
there's
somebody
else
was
asking:
if,
if,
if
cooper
could
do
a
different
virtualization
mechanism
beside
qemu
or
kvm,.
A
So
now
now
we're
talking
about
doing
embed,
more
embedded
virtualization
in
a
cloud
provider.
D
A
D
A
That's
what
the
the
ddesk
guys
are
going
to
do
and
see
what
johnson
controls.
D
A
Just
yeah,
yeah
and
generally,
I
think
this
is
just
something
we
need
to
contemplate
on
because
it's
it's
come
up
several
times
now.
E
Yeah,
well,
that's
fine
yeah
and
I'm
not
opposed
to
anything.
The
the
only
question
I
have
is
we
we
do
use
libvert
currently
is
a
you
know
as
the
layer
of
the
shim
layer.
If
you
will
the
to
communicate
with
the
virtualization
engine,
if
you
will-
and
so
that's
I
mean
it's
an
interesting
topic,
but
you
know
there
may
be
some
technical
difficulties
that
we
need
to
think
on,
because
we're
very
we're
very,
very
closely
tied.
E
Excuse
me
for
my
stammering
to
deliver
it
at
this
point,
so
I
I
don't
know
I
mean
that,
would
it
would
just
require
some
I've
seen
your
fault
in
the
past
for
like
oh
can
can
I
create
you
know
an
ec2
vm
with
cooper,
and
that
is
a
hard.
No.
A
Does
does
livford
have
any
hook
into
sun.
E
No,
I
don't
believe
that
it
does
and
I'm
speaking
out
of
turn
I
haven't
actually
checked,
but
zen
is
a
very
different
thing
than
so.
Kvm
is
kind
of
it's
everything's
exposed
through
the
system
device.
If
you
will
zen
actually
has
a
different
kernel,
and
so
it's
not.
A
Yeah,
the
reason
why
I
was
asking
about
that
is:
there's
large
zen
installations
out
there
still.
A
F
Sorry
for
interrupting,
but
I
I
think
that
I
mean
the
answer
is
pretty.
I
don't
know
why
it
get
complicated.
We
are
covered
currently
supports
only
kvm.
We
leave
it
with
kvr
and
chemo,
and
it's
not
it's
not
possible
to
put
there
something
else.
I
guess
a
feature
request
to
decouple
this
and
create
some
interface
that
can
allow
others
to
integrate
may
be
possible,
but
we
are
really
far
from
it
at
the
moment.
F
Is
it
is
what
I
said
is
it?
Is
it
wrong
or.
A
Oh,
no,
that's
it's
it's
good!
This
is
like.
I
said
this
is
probably
the
third
time
in
two
weeks
that
this
is
that
this
topic
has
come
up.
So
it's
worth
talking
about
and
contemplating
on
and
seeing
if
there's
a
a
potential
new
feature
to
come
out
of
this,
the
discussion.
F
E
Well,
no,
we
we
shouldn't
be
how
to
say
we
we
build
the
containers
that
are
cube
vert,
and
so
it's
not
somebody
else's
problem.
It's
ours!
If
we're
going
to
support
this
so
but
but
as
as
edward
was
pointing
out
like
as
of
today.
This
is
not
something
we're
targeting
or
able
to
support.
E
I
think
the
question
we
need
to
ask
ourselves
is:
how
much
adoption
would
this
drive
because,
like
yes,
it's
come
up
three
times,
and
so
we
should
have
the
discussion,
but
but
why
would
like
if,
if
I'm
building
a
web
service
that
runs
in
virtualization
or
something,
why
do
I
care
what
the
engine
is
like?
Why
is
that
such
a
sticking
point.
E
Okay,
so
this
the
use
case
here
is
some
other
cloud
doing
the
virtualization
and
we're
simply
managing
the
virtual
machines.
A
E
A
A
Yeah,
like
I
said
this,
this
topic
has
come
up
three
times
now
in
two
weeks.
So
that's
what
I
thought
it's
worth
discussing
discussing
here.
A
A
And
it
it
arose
through
that
that.
A
H
It's
it's
hit
me
again.
I
don't
know
why.
But
every
time
I
start
zoom,
my
volume
is
decreasing.
Ten
percent
yeah.
What?
What?
Actually?
What
problem
are
they
actually
trying
to
solve?
What
was
the
thing
that
they
wanted
to
to
solve
in
general?
H
Do
they
have
a
current
some,
some
some
alternative
to
ktm
running
somewhere
and
they
are
looking
at
whether
they
can
integrate
that
or
what?
What
do
they
want
to
do?.
A
A
And
so
they
we're
very
interested
in
p
to
v
and
v.
A
But
nevertheless,
they
also
talked
about
this
alternative
virtualization
options
with
embed
using
embedded
virtualization
on
a
cloud
provider.
H
A
A
lot
of
these
companies
are
really
hesitant
to
get
him
get
directly
involved
with
open
source.
A
Even
after
25
years
of
using
the
linux
kernel,.
H
A
I
agree:
okay,
well,
let's
table
this,
this
topic
and
think
about
it.
If
there
is,
if
we
want
to
formalize
this,
we
can
create
an
issue
and
have
the
conversation
through
github
or
the
or
even
the
mailing
list.
Of
course,
I
don't
think
we're
gonna.
This
is
gonna,
be
the
end
of
this
of
this
topic
for
sure.
A
A
I
H
Yeah,
I
didn't
I
didn't,
can
I
can
I
just
I
just
didn't
fill
in
anything.
So
just
a
quick
heads
up
that
yesterday.
I
think
the
122
sick
lanes
got
to
run
every
time
so
yeah
one
step,
one
small
step
forward
at
least
yeah.
What
we're
also
wanting
to
do
is
to
disable
the
always
run
on
the
1.198,
but
this
is
still
not
not
finished
yet.
H
Yeah,
it's
like
we,
we
have.
We
always
test,
have
test
lanes
for
all
the
zigs
for
the
latest
three
providers
that
we
have,
and
currently
we
are
in
a
state
where
we
have
four
providers,
but
1.19
should
fade
out
and
yeah.
We
are
currently
in
the
process
of
running
the
122
lanes,
always
and
yeah
over
time.
The
1.19
lanes
will
go
away.
A
I
guess
that
since
we're
talking
about
ci,
we
added
check
linting
to
the
website
ci
a
couple
weeks
ago.
A
We
had
to
retract
that
on
friday
because
of
a
bad
image
and
it
was
causing
a
big
blockage
and
new
pull
requests.
So
we'll
try
again
sometime
soon.
A
The
markdown
in
the
website
and
and
the
user
guide
needs
a
lot
of
massaging
to
comply
to
to
standard
markdown.
So
that's
why
we
wanted
linting.
D
Yeah,
I
actually,
I
was
one
that
was
going
to
take
on
the
pr,
but
I
got
caught
up
with
a
lot
of
work
because
I
was
transitioning
from
apartments
and
everything,
so
I'm
also
I'm
okay
to
take
that
on
again
I
just
that
was
my
fault.
I
guess
I'm
saying.
A
No,
it's
okay,
another
community
member
took
it
on
and
and
he
slipped
something
in
that
that
didn't
work.
Well,
we
had
to
take
it
out.
A
We
have
linting
in
the
in
the
local
make
file,
but
it's
up
to
the
user
to
to
run
it
locally,
and
so
we
we
want
to
get
that
into
into
ci,
and
the
image
that
we
were
using
didn't
actually
didn't
even
have
podman
installed,
so
that
was
that
messed
things
up,
we'll
get
it.
It's
we're
very
close
to
getting
it.
A
I'll
I'll
I'll
provide
you
the
the
issue
after
the
meeting.
If
you
don't
mind
perfect
thanks.
B
A
Oh
yeah,
please,
let's
move
you
up
is
that
the
pull
request,
yeah,
okay,
sure.
B
So
yeah
it's
just
a
pull
request
about
the
live
migration
policies,
so
I
talked
about
it
in
one
of
the
community
meetings.
B
So
basically,
there
was
a
design
proposal
and
many
conversations
about
it
and
I
decided
eventually
to
like
have
a
poc
of
the
initial
implementation
without
the
controversial
stuff
and
just
so,
we
have
something
working
and
then
maybe
the
the
other
continue
discussion
will
be
easier,
so
yeah
it's
there.
If
somebody
want
to
take
a
look
and
that's
it.
B
So
if,
if
you
don't
remember
it's
basically
migration
policies
for
migration,
that
can
be
used
per
vm
or
as
I
mean
right
now-
it's
per
namespace.
Basically,
it
overwrites
the
configurations
that
exist
in
cubic
cubert,
cr
and,
and
that
way
you
can
find
grain
configurations
regarding
migrations
and
adjusted
to
a
set
of
virtual
machines.
Currently
it's
just
to
name
spaces,
but
I
guess
that
in
the
future
it
will
support
other
ways
of
defining
sets
of
virtual
machines.
A
A
Well
enjoy
your
time
with
her
because
they
grow
up
fast.
A
Okay,
thank
you
itamar.
Hopefully
we
had
some
attention
on
that
and
I'm
glad
I
can
help
you.
A
Okay,
I've
been
do
you
want
to
take
on
unprivileged
vms
talk
about
computer
virtue.
J
Hey
sure
hi
hi,
so
there
was
a
discussion
actually
two
discussions
about
different
uid
or
user
for
unpredicted
vms,
and
I'm
going
to
probe
this
topic
and
see.
If
there
is
a
really
need
for
this,
and
so
we
probably
need
some
adjustments,
for
example,
with
the
control
risks.
As
roman
pointed
out
there
in
the
comment
so
yeah
to
my
question,
is
there
someone
who
want
to
have
different
id
or
user
for
ample
hvms.
E
Sorry
about
that,
I
always
just
stop
talking,
because
it's
the
safest
thing
to
do
so
is
the
question
about
exploring
using
uids
other
than
107.
Was
that
what
we
were
talking
about.
E
J
Yeah
that
I
mean
that's
what
I'm
trying
to
see
if
there
is
actually
motivation
to
use
another
uid
or
the
user
I
saw
vaseline
was
asking
for
this.
Maybe
he
has
an
idea
or
is.
I
It's
not
about
a
different
uid,
it's
about
hard-coding,
some
arbitrary
one,
let's
say
so.
We
faced
a
problem
some
time
ago
when
building
our
custom
images
that
there
was
no
user
with
uid
107
in
our
base
image.
I
But
to
be
honest
for
now,
I
don't
know
how
to
mitigate
this
issue.
So
it's
in
a
long
term.
Let's
say
only
if
we
implement
something
like
suggested
by
david,
I
guess
with
init
container
and
word
launcher,
but
yeah,
it's
a
bit
of
quite
some
work.
I'd
say
to
do
it.
I
I
And
yeah,
since
this
also
goes
further
to
contain
your
disks,
that
even
in
the
user
docs,
it's
referred
that
it's
better
to
change
ownership
right
away.
To
avoid
some
weird
things
with
the
with
some
storage,
some
storage
issues
so
yeah,
it's
kind
of
a
bit
more
complicated.
I
J
Okay,
thank
you.
I
mean
yeah.
I
see
the
problem
here.
We
also
stumbled
upon
it
that
the
handler
did
have
different
id.
I
think
so
from
this
discussion.
Maybe
there
could
be
a
point
that
we
should
move
for
from
this
mechanism
of
discovering
the
kimono
id
from
handler
to
the
specific
launcher,
which
should
be
not
a
major
issue,
because
I
wish
science
we
can
probably
probe
the
local
database
of
the
launcher
and
yeah.
So
there
is
no
need
for
the
different
uids.
Then
probably
this
would
be
sufficient.
G
I
think
it
would
be
good
if
we
would
be
able
to
consolidate
all
of
this
all
of
these
definitions
of
few
ideas
in
one
place,
because
right
now,
it's
like
spread
across
the
code
from
what
I've
seen.
J
A
A
Let's
get
some
comments
in
into
this
pull
request,
so
we
can
get
get
moving
along.
A
This
is
a
an
important
feature
to
coobert.
A
Okay,
I'm
gonna
take
over
for
a
little
bit
and
talk
about
some
compute
community
things.
A
A
We're
going
to
follow
alicia's
framework
for
from
the
kvm
forum
and
sam
walker
also,
they
used
a
slide
deck
in
the
catacotta
scenarios
that
we
have
posted
to
o'reilly.catacoda.com
and
they
had
a
great
success
with
that.
They
didn't
have
to
worry
about
any
any
met,
bare
metal
servers
and
deployment
problems.
A
It
was
just
copy
and
paste
into
into
the
scenarios,
and
it
was
made
for
a
very
fast
and
impactful
demonstration
of
cooper,
so
we're
gonna
do
that
for
both
kubecon
and
all
things
open,
which
has
for
all
things
open,
stu,
sam
and
I
are
working
like
crazy
now
to
get
our
raspberry
pi
demo
up
and
running.
D
D
I
will
say
the
good
news
is
sue
was
able
to
use
my
open
soup
image
and
jordan
just
fine.
So
I
think,
as
far
as
using
tested
images,
I
haven't
worked
on
debian
11,
so
there
might
be
differences
between
that
and
it
would
do,
but
he
was
able
to
join,
keep
proxies
that
we
started
like
it
was
yesterday.
So
if
you
do
the
same,
you
could
probably
get
it
joined
in
or
you
can
progress
from
there.
Okay.
E
A
All
right,
so
the
next
two
weeks
are
going
to
be
foreign.
After
that
we
have
no
events.
If
anybody
thinks
of
can
think
up
of
an
event
or
they
come
across
a
a
major
event
in
their
internet
wanderings,
and
they,
you
think
that
kubert
would
be,
would
have
a
valuable
presence.
A
Okay,
we
are
putting
together
the
due
diligence
for
graduation
to
cncf
incubation
and
came
across
some
really
interesting
community
statistics,
and
I
wanted
to
show
you
this.
A
So
we've
been
doing
bug
scrubs
and
it's
having
an
effect
and-
and
you
can
see
this
in
our
in
in
dev
stats,
which
is
a
this
graph-
is
a
measurement
of
how
long
issues
sit
in
github
before
they
have
a
response,
and
the
average
is
four
days
which
is
down
significantly
from
from
last
year.
A
A
But
this
is
really
good
stuff.
We
don't.
We
don't
look
at
these
metrics
very
often,
and
the
questions
came
up
when
we're
talking
about
community
health,
so
this
was
pretty
jaw-dropping.
A
So
this
is
issue
response.
This
is
pull
request,
time
to
engagement,
which
we
were
just
talking
about
a
pull
request
that
had
23
days
since
the
last
engagement.
This
is
clearly
lower
as
well.
A
We
have
the
number
of
companies
contributing
to
cooper.
A
So
this
is
a
really
good
metric.
Unfortunately,
not
all
30
of
these
companies
allow
us
to
publish
their
their
names
on
our
website.
I'm
working
on
that.
A
So
over
the
course
of
two
years
we
have
a
a
nice
linear,
increasing
trend,
and
those
are
a
few
more
positive
notes
for
to
back
up
our
our
graduation.
A
A
We
have
yet
another
due
diligence
document
and
this
one
is
way
more
extensive
than
the
others.
If
you'd
like
to
review
it,
I
posted
a
link
out
into.
A
The
meeting
notes
don't
do
any
direct
editing,
because
just
like
make
comments
for
suggestions,
this
is
a
format
that
elena
with
the
cncf,
wants
to
follow.
H
Yeah,
to
be
honest,
this
is
just
more
some
to
get
a
little
bit.
Awareness
of
that
one
thing
is
that
what
we're
what
we
are
using
for
testing
keyboard
is
that
that
we
are
just
generating
local
or
virtualized
clusters
of
kubernetes
right,
and
this
is
kuber
ti
and
kuber.
H
Ci
has
a
couple
of
add-ons
like
cnao
and
cdi,
which
are
required
for
testing
certain
things,
and
what
I
did
was
just
I
just
optimized
the
installation
of
the
add-ons,
so
this
should
shave
a
little
bit
of
a
cluster
up
time
for
that
which
might
be
beneficial
at
least
for
for
testing
besides
ci
also
for
local
testing,
and
if
anyone
is
interested
into
looking
into
that,
probably
has
some
ideas
on
how
to
even
make
it
make
this,
probably
even
better
or
something
that
I
would
be
happy
to
to
receive.
The
feedback
of
that.
H
A
H
So
can
I
go
back
and
sorry.
H
Yeah,
this
is
this.
This
is
actually
a
real
issue.
A
H
The
introduction
of
the
whereabouts,
we
have
some
problems
on
the
czech
bus
drop
that
is
run
whenever
we
we
try
to
provision
the
new
clusters
and
then
check
whether
everything
is
coming
up.
So
everything
is
okay.
H
There,
the
whereabouts
has
a
crown
job
that
is
installed
which
launches
some
some
worker
parts,
and
these
are
getting
into
the
way
of
the
of
the
check
that
that
checks,
whether
everything
is
okay
and
yeah.
H
My
quick
fix
was
just
to
patch
the
crown
job
so
that
they
have
a
ttl
of
zero
which
in
general
directly
deletes
spots,
but
they
sometimes
are
still
there
and
the
check
fails,
and
this
makes
the
checks
in
the
and
the
lanes
makes
makes
flaky,
and
so
what
I
did
was
I
just
tried
to
delete
periodically
the
ip
reconciler
ports
so
in
order
to
make
the
check
run
again
so
or
work
again,
yeah.
A
A
Hey
daniel,
I
noticed
about
a
month
ago,
red
hat
was
hiring
to
help
with
cooper.
Ci
did
anything
ever
come
about
that
from
that.
H
That's
a
good
question:
to
be
honest,
I'm
not
really
sure
about
that.
A
How
are
you
feeling
are
you
feeling
overwhelmed
with
all
the
sea
ice
work
that
has
been
going
on.
A
H
A
A
So
that
takes
us
to
7.48
did
we
do?
We
did
a
good
bug
scrub
last
week
and
it
looks
like
we
have
one
more
user
that
came
late
and
would
like
to
make
an
introduction
brock.
Would
you
like
to
talk
for
a
little
bit.
H
If
you
don't
hear
yourself
talking,
you
might
also
get
caught
by
this
10
volume
issue
that
zoom
issues
on
every
time,
starting
zoom
again,
maybe.
A
A
A
Okay,
yes,
all
in
favor
of
skipping
a
bug
scrub
this
week,
I'll
return,
10
minutes
to
you.
I
A
Okay,
we
have
one
eye
and
that's
a
super
majority.