►
From YouTube: KubeVirt Community Meeting 2021-08-04
Description
Meeting Notes: https://docs.google.com/document/d/1kyhpWlEPzZtQJSjJlAqhPcn3t0Mt_o0amhpuNPGs1Ls/edit#heading=h.vedtrrv0bjo2
A
A
B
A
It
is
you
we
were
just
waiting
for
the
first
person
to
to
ask
that's
fine,
exactly.
B
A
A
D
Hey
I
just
learned
that
stu
wasn't
joining
us
today
is.
Are
we
doing
this
sure?
D
Okay?
Let's,
let's
start,
then
let
me
share
document
in
the
chat.
B
I
don't
want
to
take
this
away
from
you,
david,
I'm,
sorry,
no,
no,
nothing
happened.
We're
good.
D
What
did
I
do
if
you
at
any
point
would
like
to
do
this
then
just
say
working
get
you
do
we
have
anybody
new
on
the
call
who
would
like
to
introduce
themselves
to
the
community.
C
Okay,
I
actually
subscribed
to
the
google
group
yesterday
and
my
name
is
leonardo
and
mostly
I
just
got
invitational
jobs
here-
to
see.
What's
going
on
in
the
community,
I'm
a
new
user
and
I'm
going
to
use
the
keyword
to
run
a
build
system
for
rpms,
and
I
think
that
that's
it.
A
C
No,
we
actually
have
a
service
that
builds
rpms,
but
it
relies
on
on
some
software
that
that
needs
to
spell
inspire
some
containers
using
the
the
low-level
kernel
api
and
that
and
that
that
that
makes
it
enable
to
run
on
a
bear
on
a
usual
container.
So,
but
we
still
wanted
to
use
kubernetes
and-
and
so
we
are,
I'm
I'm
I'm
basically
trying
to
run
this
particular
service
on
this
part
of
the
service.
Anyway,
using
keyboard
in
the
virtual
machine.
A
Excellent
yeah
we've
actually
so
this
use
case
isn't
that
uncommon
kuvert
for
build
systems
is
kind
of
one
of
the
surprising
use
cases.
I
would
say
that
kind
of
popped
up
quite
a
bit
in
the
community
in
the
past
few
years,
so
great
and
you're
not
alone
and
welcome.
D
All
right
anybody
else
would
like
to
introduce
themselves.
D
F
Sure,
hello,
everyone
yeah,
I
think
I've
been
working
on
this
this
vrs.
They
are
kind
of
a
long
standing
task
about
adding
more
documentation
about
the
ci
infrastructure
and
yeah.
They
are
up
for
review
on
and
yeah.
If
anyone
interested
could
take
a
look
and
if
feedback
is
is
very
appreciated,
there
are
three
of
them,
the
first,
the
first
one
I
put
the
links
in
the
document.
F
The
first
one
is
an
infrastructure
overview
of
what
we
have
in
all
the
the
systems
that
are
are
used
to
run
our
end-to-end
tests
and
all
the
all
the
jobs
that
are
are
run
in
the
prs
and
so
on.
The
second
one
is
this
is
came
from
from
a
question
from
how
to
how
to
scale
our
our
brow
instance
and
yeah.
This
is
this
is
kind
of
the
the
procedure
of
of
if
you
want
to
run
your
specific
keyword,
end-to-end
tests
in
our
infrastructure,
how?
F
How
could
you
prepare
first
and
then
add
your
your
the
secret
of
of
your
of
how
to
access
your
cluster
in
our
in
our
infrastructure
and
and
yeah?
The
last
one
is
about
how
to
if
you
are
interested
in
contributing
and
in
the
as
an
infrastructure
maintainer.
F
D
Okay
last
call
for
open
floor
agenda.
If
anybody
wants
to
erase
anything
before
we
jump
to
the
mailing
list
and
back
scrap.
D
A
Yeah
I
looked
at
this.
I
give
a
brief
summary.
The
idea
is
to
be
able
to
mark
volumes
being
attached
to
a
bmi
as
being
hot
plugable
right
out
of
the
gate.
So
today,
if
you
want
to
hot
plug,
be
able
to
add
and
remove
a
volume,
you
do
that
after
the
vmi
has
started,
and
you
can
only
remove
a
von
that
you
previously
hot
plugged.
So
if
you
want
to
have
something
that
you
can
dynamically
put
on
a
bmi
and
take
away
from
a
bmi,
it
has
to
be
hot
plugable.
A
So
that's
what
this
this
pr
and
this
feature
request
is
doing
it's
saying,
let's
be
able
to
mark
up
front
declaratively
when
we're
starting
a
vmi
that
these
sets
of
volumes
are
going
to
be
starting
a
hot
plugable
way,
which
means
that
they
can
be
removed
later
on
if
we
want
or
dynamically
re-added.
A
If
we
want
and
to
do
that,
the
implementation
requires
starting
this
hot
plug
attachment
pod
at
boot
time
with
the
vmi
and
allowing
these
volumes
to
be
added
to
the
vmi
through
this
attachment
pod
during
the
start
flow,
which
is
something
new
all
seems
totally
reasonable,
feasible.
I
reviewed
the
pr-
and
I
don't
know
the
person
who
wrote
pr's
here,
but
it
looks
like
like.
I
definitely
think
that
the
the
idea
makes
a
lot
of
sense.
A
The
execution
the
pr
needs,
some
just
shifting
of
where
things
are
taking
place
like
the
ordering
that
certain
things
are
taking
place
during
the
start
flow,
and
then
I
think,
it'll
work,
so
it's
fairly
close.
It
might
require
a
decent
amount
of
rework
in
the
pr.
But
it's
not
like
the
pr
is
totally
wrong.
It's
doing
the
right
things,
but
in
the
wrong
order,
I
would
say.
D
That's
pretty
cool
to
see
such
a
huge
pr
coming
in
nice.
Okay,
thanks
david
for
shooting
a
light
on
it.
That's
so
this
one
gets
more
than
enough
attention
and
then
there
is
dad's
comment.
I
have
to
answer
and
just
meeting
notes,
so
we
should
be
good
on
this
front
now,
let's
try
go
through
the
issues
for
the
last
seven
days.
Aba
there's
not
many
of
them
this
time,
just
awesome.
D
I
was
about
to
say
it's
surprising,
but
then
I
realized
I
always
run
as
an
admin.
So
I
that's
why
I
never
hit
these
issues.
A
Default
things
default
roles
to
the
excuse
me
back
up.
There's
default
users
like
admin,
editor
and
user.
Something
like
that.
I
don't
exactly
remember
and
we
aggregate.
A
To
give
give
access
to
virtual
machines
to
those
default
users,
and
as
long
as
a
user
like
an
actual
user,
inherits
those
they
should
be
authorized.
So
in
this
case,
my
guess
is
whatever
account
is
being
used
to
try
to
start.
The
virtual
machine
doesn't
inherit
from
one
of
these
standard
default
accounts.
D
A
Quickly
find
the
user
guide
for
that.
A
A
So
I
would
point
them
to
this
and
just
make
sure
that
the
user
trying
to
start
the
virtual
mainstream
as
the
expected,
our
back
permissions.
I
There's
a
similar
there's,
a
similar
bargain
in
the
bugzilla
that's
been
investigated,
but
yeah.
We
need
more
logs
from
that
specific.
I
D
A
D
A
Read
write
mini
pvc.
A
I'm
curious
how
this
got
mounted,
so
the
pvc
itself
must
be
read
right.
Many,
let's
see
we're
not
gonna
be
able
to
see
from
here.
Are
we
no.
J
Well,
read:
write:
you
can
have
multiple.
J
It's
node
based,
so
they
would
just
the
pods
would
just
end
up
on
the
same
node,
so
you
can
have
multiple
pods
axing
or
accessing
a
read,
write
once
they're
running
on
the
same
node,
okay,
yeah,
the
ones
first
many
is,
is
a
node
thing,
so
I
guess
la
vert
or
keemu.
It
sounds
like
puts
a
file
lock
on
yeah,
so
yeah
I
mean
I
I
mean
probably
be
better
to
import
the
nfs
volumes
in
their
vms
directly.
If
they
can.
I
I
Yeah
then
it
doesn't
have
to
contain
the
image
it
doesn't
have
to
be
the
same
image.
It
can
be
just
an
empty.
E
J
A
D
Okay,
I'm
hearing
many
alternatives
to
this,
so
this
won't
work.
J
That
won't
work
either
I
mean
you
can't
you
can't
have
him.
You
won't
allow
for.
J
A
Gonna
work:
let
me
find
the
user
guide
thing.
E
A
D
K
K
Yeah
so
I
mean
that
would
be
the
right
right
way
to
approach
it
if
you
were
trying
to
solve
the
underlying
problem
rather
than
find
a
use
case.
For
you
know,
multiple
sharing,
like
a
secondary
disk.
A
A
So
it's
trying
to
act
like
an
nfs
volume,
but
at
the
lower
level
and
yeah
it's
not
going
to
work
so
directly
connecting
to
the
nfs
server
within
the
guest
would
potentially
allow
them
to
get
the
behavior
they're
looking
for
it
wouldn't
be
treated
as
a
disc,
though
that's
kind
of
they're
trying
to
make
it
they're
putting
what
the
one
wanting
it
to
be.
Essentially,
a
hard
drive
attached
to
this
virtual
machine
and
the
hard
drives,
backed
by
nfs.
A
B
D
So
the
the
ephemeral
thing
wouldn't
help.
I'm
sorry
well.
A
Data
across
it
like,
if
they
just
want
an
nfs
mount
that
is
read
only,
but
I
mean
the
fact
that
every
right
means
that's
unlikely.
D
D
Convert
modules
are
using
more
cpu
than
requested.
Are
we
did
do
you
know
whether
this
was
brought
up
on
the
sixth
scale
cold?
Last
week,.
D
Sweet
there's
one
that
isn't
trashed,
so
I
think
we
talked
about
this
the
last
week.
Yes,.