►
From YouTube: KubeVirt Community Meeting 2021-09-08
Description
Meeting notes: https://docs.google.com/document/d/1kyhpWlEPzZtQJSjJlAqhPcn3t0Mt_o0amhpuNPGs1Ls/edit#heading=h.pp57ii1x9dan
A
Okay,
hello,
everybody
I'm
chris
caligari
and
I'm
the
community
organizer
for
project
kubert
welcome.
This
is
your
chance
to
meet
your
developers
and
contributors
and
talk
about
community
issues.
A
Okay,
let
me
share
my
screen
and
I'll
share.
Our
meeting
notes.
A
Okay,
a
lot
of
I
have
to
take
the
first
two
items
on
the
agenda
this
week.
I
want
to
bring
everybody's
attention
to
the
incubation
status
via
poll
111.
A
It
has
been
merged
and
we
have
we've
submitted
the
that
document
to
the
toc
and
I
actually
forgot
to
post
that
link.
So
give
me
just
a
second
here
and
I'll
find
it.
A
A
And
I
think
elena
has
already
made
some
comments
for
me
to
take
care
of
she's
fast.
A
Okay,
sorry,
that
was
really
fast.
Moving
on
to
events
do
we
have
alicia
with
us.
A
Creature,
we
need
you
to
talk
today.
Sam
have
you
talked
to
alicia
in
the
last
day
or
two.
B
Let's
see
talked
with
her
last
meeting
wednesday,
but
I
have
not
heard
from
her.
I
got
an
email
to
read
from
her,
though,
but
nothing
that
needed
my
attention
yet.
A
Okay
yeah
same
she
was
asking
about
a
project
roadmap.
A
Speaking
of
project
road
maps,
david
vossel,
if
I
could
grab
some
of
your
time
today
to
do
a
last
review
of
our
roadmap.
I
know
we
spent
a
lot
of
time
on
it
before
so.
It
might
just
need
a
little
polish
and
shine.
Do
we
have
a
do?
We
have
a
link
to
our
road
map
we
can
post.
Maybe
we
can
just
I've
been
using
the
road
map
that
we
put
into
community
poll
111.
A
We've
been
doing
some
work
on
the
cooper
cooper
and
putting
tech
enhancement
tags
on
on
issues
and
then
creating
milestones
so
that
that
work
is
just
getting
started.
C
Okay,
I
think
we
have
a
future
up
there.
Yeah,
I
think
very
went
over
that.
I
guess:
what's
your
concern
about
submitting
that
to
incubation
yeah.
A
And
alicia
is
gonna
advertise
that
list
also
to
kvm
forum.
C
Okay,
it
looks
it
looks
accurate,
I'm
sure
that
items
will
get
added
to
that
list,
but.
B
Yeah,
I
put
some
images
up
last
night
that
may
help
okay.
A
Yeah,
it's
gonna
be
a
priority
for
us
this
week.
Sorry,
I
didn't.
I
didn't
connect
with
you
late
in
the
day
yesterday
they
gotta
hold
the
see
it
oscar
got
a
hold
of
me
and
said:
let's
roll
with
the
toc
hole.
So
now
that's
done.
I
can
switch
gears
and
help
you
guys
out
and
of
course,
anybody
else.
Who's
got
a
raspberry,
pi
or
extra
x86
64
laying
around
that
wants
to
help
us
out
with
this
demo,
of
course,
you're
more
than
welcome.
A
C
Hey
yeah
sure
there
was
a
google
group
mailing
list
post
that
none
of
us
responded
to
and
I
looked
into
it.
I
didn't
actually
have
a
great
response.
That
was
part
of
the
reason
why
it
probably
went
on
unresponded
to
so
trying
to
regather
my
thoughts
around
this
one.
C
D
So
we
fetch
we've,
we
just
fetched
the
version,
because
there
are
no
special
privileges
needed
it's
kind
of
a
license
we
had
in
the
past.
I
don't
I'm
not
sure
how
much
of
an
issue
it
is
right
now,
but
in
the
past,
a
lot
of
issues
that
we
were
actually
connect.
It
looked
like
we
are
connected
from
the
kubernetes
client
perspective.
D
The
the
components
were
running,
but
we
had
in
practice
lost
connectivity
to
the
api
server.
C
C
Well,
the
email
is
mostly
related
to
why,
when
our
connectivity
with
the
api
server
fails
and
this
healthy
endpoint,
eventually
the
health,
the
endpoint
will
start
returning.
An
error
which
will
cause
the
radius
probe
or
lightness
probe
will
fail
at
some
point
and
then
it
will
restart
yeah.
D
So
this
is
the
purpose
of
it,
so
it's
not
about
the
api
version
as
such
sure
that's
the
purpose
of
the
so
the
only
question
remaining
is
why
we
also
expose
it
in
the
in
the
response.
E
C
Why
do
we
expose
it
in
the
response
and
why
do
we
keep
track
of
it
at
all,
like
what's
the
purpose
of
us?
I.
D
F
D
This
is
okay,
so
this
is
about
the
optimization
which
markets
did
so
that
we
are
not
querying
so
often
api
server.
C
Okay,
yeah,
okay,
a
version
is
nil.
Okay,
so
that's
why
we're
keeping
the
version?
Maybe
I
see
yeah.
F
D
C
That's
interesting,
so
our
behavior
looks
like
it's
changing
then
in
the
newer
version.
Where
do
we
actually
still
detect
when
the
version
endpoint
yeah.
D
It's
just
cached
for
some
time,
so
the
the
or
not
there
is
now
a
an
extra
loop
which
is
periodically
updating,
checking
if
it
can
still
connect
to
the
api
server
and
the
health
check.
Just
in
the
and
the
health
check,
just
fetches
that
then
from
internally
and
exposes
it,
and
if
it's
not
there,
then
the
post,
the
pot
goes
to
hold.
C
D
So
yeah
the
health
check
can
have
a
lower
period
with
that
and
can
also
check
internal
stuff,
and
basically
just
the
liveness
probe
to
the
api
server
is
decoupled
a
little
bit
more.
It
can
have
a
lower
frequency.
That's
the
main
purpose
of
this.
D
Yeah,
so
this
is
the
historical
thing
that
we
had
a
lot
of
issues
with
were
tenders
which
looked
healthy,
but
we're
not
and
we
had.
No,
then
there
were
a
lot
of
different
cases
where,
where,
where
parts
of
the
network
were
reachable,
other
parts
were
not,
and
so
you
could
sometimes
reach
other
parts,
but
not
the
api,
server
and
so
on,
and
what
proved
to
be.
The
only
really
effective
method
was
to
just
restart
the
handle
at
some
point.
D
Okay
and
it
was
not
in-
and
this
was
related
to
a
lot
of
different
issues
like,
for
instance,
in
openshift.
They
were
for
a
very
long
time,
some
dns
issues
where
the
dns
servers
and
the
nodes
were
sometimes
reconfigured
or
something
and
then
already
running,
parts
lost
connectivity.
But
it's
not
only
openshift.
There
are
a
lot
of
things
like
this
and
we
could
catch
all
of
this
with
just
restarting
in
this
case.
A
Thanks
david,
I
was
actually
slightly
late
to
the
meeting
because
I
was
reading
this
exact
mail
thread
so
I'll
remove
it
from
down
here.
A
B
Yeah,
hey
sam
here,
going
back
to
all
things
open,
real,
quick.
I
just
want
to
kind
of
draw
a
bit
of
attention
to
the
thread
I
found
a
few
products.
I
think
we'll
kind
of
tie
the
glues
together
between
you
know,
using
a
wire
guard
and
such
there's.
G
B
Plug-In
that
was
actually
made
by
started
by
a
guy
right
hat,
but
you
know
it
went
full
open
source
and
it
works
with
existing
cni.
So
if
we
end
up
using
flannel
updates
too
or
you
can
replace
the
whole
thing,
I
encourage
you
all.
Do
you
give
a
giveaway
playing
with
it,
because
I
think
that
that
that
c
and
I
will
definitely
be
the
easy
glue
together
to
get
a
cluster
working
across
different
gnats,
even
which
is
surprisingly
makes
it
super
simple.
B
As
far
as
annotations,
it's
called
kilo-
and
I
it's
in
the
thread
but
I'll
all
really
really
get
here.
G
A
B
A
A
B
My
mind
pick
up
my
slide,
I
mean
this
is
this
is
relates
to
is
a
related
project
and
that's
kind
of
reported
uploaded
to
called
tier
k
and
working
on,
which
essentially
does
a
lot
of
this
side-to-side
networking,
but
also
in
a
sense
where
you
can
route
to
your
pod
network
free
through
your
default
gateway.
So,
like
I
could
ping
an
ndx
pod
for
my
pc,
for
example.
B
Oh
sorry,
not
too
great,
it's
here
k
t
that
probably
I
like
it
too,
but
a
lot
of
that
work
and
the
research
that
was
done.
That
kind
of
relates
to
this
actually
kind
of
falls
in
hand
with
what
they're
trying
to
do
what
you
all
are
trying
to
do
doing
all
things
open.
A
A
Have
a
look,
thank
you
ramen
and
he's
doing
some
good
stuff
he's
he's
going
through
our
catacota
scenarios
and
making
sure
that
they're
they're
in
line
with
what
we're
advertising
on
the
website
in
a
due
diligence
documents.
And
so
I
think
alicia
is
going
to
be
using
some
of
this
stuff
on
the
jvm
forum
demo.
A
Okay,
we're
at
7
22
and
we
haven't
done
a
bug
scrub
in
quite
a
while.
So
it's
been
going
to
be
a
big
one.
This
week.
I
D
C
Is
that
going
to
be
as
simple
as
exposing
the
field
on
the
pod,
like
just
a
transparent?
Take
the
thing
that's
in
the
vmi
put
on
the
pod
and
we
get
it
or.
D
Okay,
it
may
mean
that
we
have
to
tweak
our
dhcp
responses
and
so
on,
so
that
it's
passed
through
properly,
but
it
can
also
be
that
if
it
just
sets
the
hosts
entries
and
rather
interested,
the
http
server
picks
the
correct
things
up
automatically.
But
we
have
to
check
that.
D
And
yeah,
and
the
difference
is
that
if
you
don't
set
this
field
as
far
as
I
understand
it,
then
you
get
really
host
name
in
this
case.
Wd
h1,
and
if
you
set
it
to
true,
you
would
get
wdh1
dot,
namespace
dash,
subdomain
or
node.subdomain.namespace.service.
Whatever
returned
yeah.
Okay,
that
makes
sense
yeah
and
can
be
the
drift
pick
up
different
things
here
in
dhcp.
C
So
how
it
sounds
like
we
would
approve
this
if
it
got
written.
How
are
we
marking
issues
as
like?
This
is
something
we're
going
to
do,
or
somebody
can
pick
this
up
and
I
think.
A
D
A
Do
you
know
who
is
usually
in
the
sig
network.
C
Well,
I
went
ahead
and
I
pinged
some
people
on
it,
so
we
should
be
good.
Okay,.
G
D
C
We're
generating
it
out
of
the
bmi
uid
yeah,
so
ever
so,
for
example,
if
there's
a
vm
that
has
cloud
or
nets
associated
with
it
every
time
we
start
the
vmi
for
that
vm
to
restart
the
vm,
it's
going
to
execute
the
cloud
and
net,
which
is
that
bad?
I
guess
why
would
that
be
bad?
I
guess
people
typically
would
want
it
cached
to
understand
that
it's
already
been
executed.
D
D
B
C
So
we
would
have
to
generate
a
uid
for
the
cloud
and
net
store
that
uid
and
the
bmi
status
or
the
vm
some
status.
I
don't
know
and
regenerate
it
any
time.
The
cloud
net
is.
C
Oh,
I
see
you're
talking
about
that.
What
if
we
just
took
if
it's
back
well,
if
it's
backed
by
a
vm,
then
we
should
have
the
vm's
uid
somewhere
in
the.
So,
if
we
just.
C
Is
that
good
enough,
though,
because
that
that's
quite
the
unexpected
configuration
to
get
consistent,
cloud-net
behavior,
to
have
the
set
explicitly
your
firmware
uuid
and
that's
somehow
associated
with
whether
cloudnet
runs
twice
or
not?.
C
Anyone
know
if
the
uuid
for
firmware
is
generated
automatically
in
the
vm
like
when
we
post
a
vm.
If
the
template
has,
this
does
not
have
a
uuid
set
from.
Where
did
we
generate
one?
There
yeah.
C
C
Yeah
I'll
do
that.
Thank
you.
D
So
with
with
sidecar,
where
the
lfs
could
give
us
that.
D
C
So
I
don't
know
how
far
we
want
to
go
with
this
people
always
have
the
ability
to
modify
pods
themselves,
so
they
can
create
a
mutation
web
hook
of
some
sort.
That
would
inject
this
empty
disc
and
all
of
that
into
their
sidecar
container.
D
C
Yeah,
well,
they
don't
even
necessarily
need
to
eject
the
sidecar
in
the
web
hook.
It
would
just
be
the
volume
mount.
So
I
look
at.
C
I'm
hesitant
to
the
psych.
Let
me
look
at
how
we
can
figure
the
sidecar
list,
real,
quick,
just
to
make
sure
it's.
C
Yeah,
I
don't
want
to
completely
recreate
a
pod
or
a
container
spec
in
there
yeah.
I
think
I
think
we're
gonna.
Let
me
let
me
I'll
write
it
I'll
inquire.
If
it's
good
enough
to
use
a
mutation,
webhook.
A
C
J
D
Empty
disk
is
normally
just
a
cue
card
without
anything
special
in
there,
so
I
guess
yeah
I
mean
we
could
I'm
not
sure
what
we
would
do
right
now
there?
I
guess
it
will
not
work
with
the
fs,
but
we
could
make
it
work
by
just
exposing
the
path
a
path,
but
we
don't
have
that
right.
Now,.
J
C
Oh,
I
see
so
it
seems
like
this
request
is,
don't
we
have
do?
We
have
a
bot
today
that
is
syncing
keeper
ci
with
the
key
for
key
vert
repo,
or
is
that
still
a
manual
process.
D
D
C
C
No
not
yet
at
least
okay.
C
A
A
And
I
guess
we
can
conclude
this
week's
meeting.