►
From YouTube: KubeVirt Community Meeting 2022-05-04
Description
Meeting Notes: https://docs.google.com/document/d/1kyhpWlEPzZtQJSjJlAqhPcn3t0Mt_o0amhpuNPGs1Ls/
A
All
right
well
welcome
to
the
may
the
fourth
edition
of
cooper
weekly
community
meeting.
Can
everyone
hear
me?
Okay,
all
right,
I
just
dropped
the
document
link
in
the
zoom
chat.
If
everyone
wants
to
go
ahead
and
pull
it
up
and
log
attendance,
real,
quick.
A
Of
course,
as
usual,
it's
open
to
adding
items
to
the
agenda,
notes,
open
floor
and
noting
prs
and
mailing
list
review
convos
or
bugs
that
need
attention.
A
B
Hey
okay,
my
start,
I
just
developing
one
interesting
feature
and
I
found
that
I
don't
like
how
the
network
is
working
in
kubert
and
I'm
trying
to
implement
it
different
way.
So
I
used
magnet
app
mode
to
bind
to
the
pod.
Networking
got.
A
It
hang
on
just
second:
do
you
have
a
pr
or
something
like
that
to
add
it
to,
or
do
you
wanna
adam?
Yes,
yes,.
B
B
A
We
can
either,
if,
if
it's
a
formal
pr
or
something
like
that,
you
can
go
ahead
and
just
there
there's
a
link
in
the
zoom
chat.
That
brings
you
to
the
community
meeting
notes.
C
If
you
can't
edit
the
document,
I
think
that
you
are
not.
You
have
not
registered
somehow
catherine.
How
was
the
process
again,
yeah.
A
All
right,
the
general
community
information
is
on
this
cooper,
io
community
page
and
then
to
have
edit
access
to
the
meeting
notes.
You
can
join
the
google
group,
and
that
gives
you
access
to
be
able
to
add
items
to
the
agenda
and
things
like
this.
A
We
can
go
ahead
and
discuss
it
real,
quick.
I
don't
see
anyone
adding
anything
else
to
the
agenda,
so
I
see
no
reason
to
do
this
in
any
other
order.
B
Me
yeah,
so
that's
it!
I
was
trying.
I
was
making
some
benchmarks
of
the
latency
and
I
found
that
using
the
mac
setup
mode
is,
has
less
latency
than
using
bridge
and
actually
yeah.
I
haven't
attached
any
benchmarks
here,
but
they're
in
slack
thread,
so
it's
actually
has
the
same
latency
like
using
standard
which
interface
or
I
was
also
doing
some
research
by
removing
batch
interface
and
replacing
this
with
the
top
on
the
host
namespace.
B
C
C
Okay,
so
I
think
we
okay,
I
guess
that
he
has
probably
microphone
issues.
Let's,
let's
have
patience
with
him
and.
B
Yeah
implementing
this
book
was
kind
of
funny,
because
macro
type
interfaces
are
slightly
different
than
standard
top
interfaces.
The
first
thing
I
wanted
to
ask:
why
do
we
have
the
special
handler
which
is
creating
the
interfaces
inside
the
pod
name
space
I
mean
I
mean
virtue
or
something
like
that.
B
Miguel
just
yet
so
anyway,
the
main
the
main
implementation
was
about
creating
the
magnet
app
interface
and
there
were
some
problems
with
the
device
which
is
creating
for
it.
So
I
was
adding
it
to
the
c
groups
for
the
pot,
and
now
it
is
working
as
a
charm.
D
B
D
B
So
that's
it!
I
did
the
implementation
of
heading
tap
into
the
top
mukba
tab
interface
inside
this
handler.
If
the
parent
is
specified,
it
will
create
mac.
What
app
instead
of
standard
tab-
and
there
is
also
some
additional
additional
implementation,
which
is
switching
on
the
permissions
for
this
device
to
allow
it
to
use
inside
the.
F
Last
yeah,
I'm
very
sorry.
Do
you
join
this
lights?
I'm
very
interested
on
this
on
this
subject,
but
I
lost
the
beginning
of
of
the.
B
B
So
this
is
almost
the
native
performance.
I
guess.
F
Yeah,
like
the
let's
say,
the
advantages
of
using
it
are
quite,
are
quite
easy
to
grasp
and
like
the
thing
that
concerns
me,
the
most
is
not
what
you're
trying
to
do
is
literally
that
we,
when
when
well,
basically,
I
developed
the
device
plug-in
and
the
and
the
cnip4
that
we're
using
for
this
for
the
secondary
networks
and
like,
as
I've
said
to
you
like,
we
never
consider
this
to
develop
what
you're
doing
right
now.
F
F
No,
I'm
just
trying
to
say
that
it's
good
that
you're
finding
like
new
use
cases
for
this
and,
like
I'm
just
a
little
bit
concerned
because
we've
never
tested
this
like
this
thoroughly
or
meant
to
do
it
like
this,
for
instance,
this
scenario
where
we
like,
I
don't
know
how
this
will
conflict
with
the
dhcp
that
is
running
on
the
on
the
host
on
the
lower
device.
That's
on
the
host,
like
these
sort
of
things,
we've
never
tested
them.
F
We
assumed
there
would
be
like
one
single
dhcp
server
on
the
entire
like
on
the
unload
device,
and
that
was
it
so.
B
Okay,
look
right
now.
The
this
pr
is
supports
both
modes.
If
you
use
mac
with
apps,
you
know
there
is
just
small
if
check.
If
the
host
device
created
for
port
is,
is
it
marked
in
the
tab
device?
If
it's
so
it
will
just
simply
use
it
like
we
do
that
right
now
and
if
it's
not,
it
will
create
a
mac
whatsapp
device
on
top
of
it
like
the
parent,
and
it
will
also
also
create
them
aqualand
device
for
the
dhcp
I
tested
it
with
the
silum
and
with
the
bridge
plugin.
B
It
is
working
fine,
and
I
also
wanted
to
try,
because
I
implemented
this
logic-
that
it
is
possible
to
run
ghcp
inside
the
port
for
the
mac,
what
up
c9
plugin,
but
it
is
not
working
because
probably
mac
vertapsina
is
not
exposing
the
ipam
block
into
this
into
the
c9
argument.
F
That's
quite
yeah,
if
I
remember
correctly,
that
cni
does
not
understand
like
it
will
not
call
the
the
chained
ipam
plugin.
B
B
Just
experiment-
I
I
just
wanted
it
to
make
it
working
with
my
pr,
but
actually
I
don't
need
that
because
I'm
considering
using
both
like
this
mode
for
binding
pod
network,
actually
we
use
cilium
and
using
macrotop
cni
to
building
normal
network
and
using
independent
dcp
server.
There.
F
Right
like
one,
what
one
of
the
things
that
I
did
not
have
yet
the
time
to
go
through
your
pr,
but
like
one
of
the
things
that
I
like
implementation,
wise-
and
I
think
I've
said
this
to
you
over
slack-
is
like
the
way
you're
granting
access
to
the
to
the
tap
to
the
tab
device.
So
if
I
understand
correctly,
you're
just
like
creating
like
well
a
new
device
with
the
same
major
and
minor
number
and
well,
it
works
because
of
that,
but
doesn't
it
require
to
be
like?
B
It
is
not,
and
that
is
it
is
done
by
handler
privileged
of
course,
yeah
yeah.
First,
we
create
device
using
the
tab
device
maker.
If
I
remember
correct,
and
then
we
use
special
c
groups
handler,
which
is
allowing
the
c
group
of
this
port
using
this
device,
and
not
only
after
that,
the
bird
will
allow
to
attach
it
to
the
network.
F
B
That's
done
by
the
witch
handler
first,
it
is
invokes
on
the
this
special
binary
for
creating
tab
inside
the
port
namespace
namespace.
Then
it
is
also
creating
a
desktop
number
device
and,
and
then
the
next
function
is
creating
the
c
group.
F
B
No,
no,
there
is
no
changes
in
from
the
user
side,
it
will
just
allow
the
using
macro
tab
for
in
for
the
winning
default
network
and
any
other
c9.
F
I
need
to
take
a
look
at
your
code
like,
but
anyway,
the
what
I'm.
The
only
thing
that
concerns
me
seriously
is
that,
if,
if
we
use
like
with
when
we
wrote
the
device
plug-in
approach,
we
actually
considered
doing
this
just
create
a
new
device
with
the
same
major
miner
and
give
access
to
it.
Our
concern
was
that
something
on
the
kubernetes
api
would
overwrite
the
c
group
allocation.
F
B
F
That
I'm
excuse
me,
that's
the
thing,
I'm
most
concerned
on
your
code,
but
yeah.
Let's
take
the
discussion.
They're
like
I
think
it's
very
valuable,
like
the
feature
itself
and,
as
you
said,
like
the
advantages
are
pretty
well
obvious,
like
it
performs
a
lot
better
in
terms
of
latency,
so
I
will
take
a
look
at
your
pr
whenever
I
can
find
time.
B
Thanks
yeah,
I
was
also
did
some
research
wise,
your
plugin
is
working.
Why
is
mac?
What
up
c9
is
working
out
of
box
because
I
think
it
is
because
it
is
using
standard
c9
functions
to
moving
device
into
the
port,
and
it
is
handling
this
all
these
things
by
itself,
but
since
we're
doing
that
from
the
keyboard
by
bit
handler,
we
should
do
that
by
ourselves.
F
That
was
one
discussion
we
also
had
in
the
beginning,
like
should
this,
this
functionality
be
located
on
a
different
project
or
in
the
well
in
hubert
in
the
handler
thing
itself,
and
I
managed
to
convince
the
community
back
there
that
we
should
actually
have
all
this
functionality
in
a
third
party
location
and
just
leave
the
the
binding
parts
on
on
cubert
and
that
the
the
the
reason
was
that
we
could
just
leverage
the
the
device
plug-in
architecture,
not
architecture.
The
framework
to
to
do
this
and
leave
kubert,
not
caring
about
the
device
at
all.
B
Yeah
anyway,
we
need
to
make
some
top
devices
to
be
able
using
it
with
a
virtual
machine
and
all
these
bridges.
So
this
is
just
another
method.