►
From YouTube: KubeVirt Community Meeting 2021-05-26
Description
Meeting minutes: https://docs.google.com/document/d/1kyhpWlEPzZtQJSjJlAqhPcn3t0Mt_o0amhpuNPGs1Ls/edit#heading=h.r01iejodmxrt
A
Okay,
hi
everybody
I'm
chris
caligari
and
I'm
your
host
for
this
week's
kubevert
weekly
meeting
and
as
always,
we
record
these
sessions
and
post
them
out
to
youtube.
A
Okay,
let's
get
into
it.
I'm
looking
at
the
list
and
I
see
familiar
names,
so
nobody
knew
this
week.
A
A
A
Okay,
while
we're
are
filling
out
those
agenda
items,
we
can
get
started
with
the
first
one.
Itamar
take
her
away.
B
Yeah
hi
everyone-
I
was
thinking
thinking
about
adding
another
lane
similar
to
make
generates
to
basically
ensure
that
no
one
forgets
to
update
their
dependencies
before
issuing
a
pr.
B
I
think,
that's
very
valuable,
because
not
long
ago
there
was
a
situation
that
somebody
updated
a
commit
manually
to
the
vendor
directory
and
then,
when
I,
when
I
tried
to
update
the
dependencies,
they
basically
override
this
this
commit
yeah.
So
I
think
it
could
be
very,
very
valuable.
C
I
agree
that
would
be
very
valuable
just
there
should
just
be
one
detail
here
that
we
cannot
do
make
depth
update,
because
it
would
really
update
go
dependencies.
We
would
have
to
just
use
this
sub
part
here
where
we
are
where
we
do
go
more
vendor
or
something
to
keep
the
tree
in
sync
without
updating
the
dependence,
but
it
would
be
very
valuable.
D
E
C
What
what
itamar
said
is
community
value,
so
there
is
a
command
for
go
mod
where,
where,
where
you,
where,
where
it
really
synchronizes
the
vendor
content
with
what
is
heart,
so
any
changes
you
would
do
manually
would
be
rewarded
on
the
job
and
we
then
like
would
make
generate
and
our
format
checking.
You
would
see
that
there
is
a
div
which
should
not
be
there.
F
F
D
And
to
clarify
just,
I
think
I
understood
like
50,
of
what
you
said:
we're
not
pulling
the
dependencies
every
time
in
a
lane.
Are
we
to
verify
this?
It
would
just
be
a
some
sort
of
validation
of
the
go
mod
and
vendor
directory
are
insane.
How
would
what
are
the
mechanics
behind
this
maybe
clarify
that
one
more
time.
For
me,
sorry,.
C
So,
to
be
honest,
I
don't
know
if
I,
if
it
would
detect
changes
at
least
at
first.
We
would
really
forcefully
delete
the
vendor
directory
and
then
run
the
command
to
sync
the
go
mod
file
with
vendor
vendor,
and
then
it
would
download
everything.
F
Yeah,
you
would
run
ideally
go
more
tidy,
then
go
with
vendor
and
check
if
there's
a
good
div
for
the
vendor
for
the
right
directory
maybe
deleted
before,
because
I
think
it
doesn't
delete
stuff
on
its
own
yeah.
F
Vendor
and
that
it
might
download
the
dependencies
newly
from
the
respective
go
proxy,
but
because
we
don't
have
a
shared
gold,
mod
cache.
But
it
shouldn't
be
an
issue.
D
D
F
D
Yeah,
I
don't
know
I
mean
that's
the
whole
reason
we
use
the
vendor
directory
is
so
we
aren't
necessarily
dependent
on
the
availability
of
all
these
repos.
D
What
if
we
just
created
a
when
somebody
run,
runs,
make
depths
whatever
update
we
just
hash
the
vendor
directory
and
commit
that
hash,
and
then
we
just
have
a
lane
that
checks
that
the
vendor
directory
matches
the
expected.
D
F
D
For
example,
somebody
going
into
their
printer
directory
and
modifying
something-
sometimes
it's
not
even
obvious,
you're
working
in
a
vendor
directory
if
you're
tunneling
through
the
code,
an
ide
takes
you
to
a
vendor
portion
of
code
within.
I
think
it
lets
you
modify
that
stuff.
H
D
Aware
I
I'm
fine
with
whatever
approach
we
take
here
to
guarantee
the
consistency
of
our
vendor
directory.
I
want
to
avoid
any
approach
that
requires
pulling
bits
over
the
network
from
external
sources
that
may
or
may
not
be
available
simply
because
that
has
the
potential
to
block
our
ability
to
merge.
D
So
at
no
point
do
we
want
somebody
else's
repos
availability
or
our
ability
to
even
reach
out
to
to
impact
our
ability
to
merge.
A
Okay,
thank
you.
Ermar
next
issue
follow
up
on
non-root
testing.
I
Hi
everyone,
so
we
discussed
the
last
time
the
donor
testing.
Actually
it
was
in
april,
so
I
will
just
remind
the
everyone,
the
final
version,
and
that
was
that
we
will
add
the
testing
lines
for
the
non-node.
I
After
that
I
was,
I
was
going
to
open
a
pr
on
the
project
ci
and
we
discovered
that
it
will
just
add
a
lot
of
lanes
that
we
don't
or
can't
actually
run
now,
because
the
ci
status
is
as
it
is,
and
so
we
discussed
this
with
frederico
and
we
decided
that
we
will
go
with
just
a
periodic
job,
the
dealer
daily
and
it
will
block
the
release,
and
I
want
you
to
just
discuss
this
solution.
If
it's
okay
with
everyone
and
yeah,
that's
it.
D
I
So
we
don't
want
to
make
a
default,
because
we
want
to
also
test
the
code
path
of
non-root
for
the
virtual
machines
that
are
running
the
production.
So
we
don't
bring
something.
I
D
I'm
not
completely
following
it.
We
would
test
both
root
and,
oh
so
here's
my
understanding,
there's
some
code
paths
like,
I
think,
maybe
huge
pages.
If
I
remember
your
pr
that
require
root
and
some
that
don't
so
the
features,
I
guess
my
expectation
would
be
that
the
features
that
require
root,
silently
at
least
right
now
have
root
just
get
enabled
on
them
and
the
ones
that
don't
require
root
by
default.
We
just
don't
use
root,
so
it's
just
a
toggle
on
our
our
pot.
D
I
D
That's
interesting:
what
was
the
first
one?
You
said,
I
get
the
devices
in
the
sidecar,
you
said
existing
workloads,
so
yeah.
D
D
D
D
D
I'm
a
little
concerned
about
the
release
blocking
because
it
lacks
visibility
until
somebody
pushes
the
button
to
make
the
release
and
then
it's
urgent.
So
what
I
mean
we
could
go
a
week
without
noticing
that
the
periodic
is
failing
and
then
trying
to
make
the
release
realize
that
it's
failing
and
then
I
have
to
unwind.
I
don't
know.
G
Email
alert
capability
of
test
read
to
get
emails
if
one
of
the
periodicals
start
failing,
so
that
we
don't
need
to
be
watching
our
slack
channel,
maybe
just
subscribe
to
some
released
and
get
an
alert,
or
I
mean
an
email
if,
if
the
test
starts
failing
or
something
like
that,.
D
How
much
noise
will
that
create
like?
What's
our
like
we're
talking
about
the
entire
test,
suite
passing
or
failing
so
what's
our
pass
rate,
I
I'm
not
talking
about
necessarily
like
we
have
a
successful
run.
I
would
expect
it
to
mostly
pass,
but
sometimes
ci
infrastructure
has
issues
and
other
stuff
that
would
create
noise.
So
I'm
saying,
if
I
see
this
a
lot,
I'm
not
going
to
go
investigate
it,
because
it
requires
time.
I
J
I
Yeah,
I
can
just
tell
that
I
mean
it
is
month
from
the
initial
pr
probably,
and
I
had
to
rebase
it
four
times
and
probably
two
times
something
got
broke
so
then
android
implementation.
So
it
is
a
pain.
D
What's
the
minimum
amount
of
testing
that
we
need
to
get
this
merged?
Can
we
add
just
some
test
cases
that
cover
high
impact
code
paths
for
non-root
and
get
this
merged
so
add
new
tests
to
our
existing
test?
Suites
the
exercise
stuff.
I
I'm
not
sure
I
mean
it
can
break
on
anything
right.
For
example,
cisco
misinformation
for
cisco
or
some
file
system
changes,
maybe
some
devices,
for
example,
new
features,
as
is
the
custom
kernel
yeah.
So
there's
a
lot
and
I'm
not
sure
what
should
we
cover
as
the
basics?
What
would
you
define
as
a
basics.
A
D
Okay,
I
would
consider
it
experimental
to
start
out
with
it's
an
opt-in
feature,
and
we
just
get
enough
test
coverage
to
ensure
that
at
least
the
basics
we're
at
least
holding
it
together
for
now.
So,
for
example,
what
would
we
need?
D
I
Probably
we
could
pick
the
compute
line
for
the
basics,
and
I
mean
networking
has
some
advanced
stuff
in
the
networking.
Also,
the
top
devices
cni
or
for
sroe
and
the
storage
have
some
other
kinds
of
storage.
D
So
can
we
take
these
tests?
This
tests
that
already
exist
today
make
them
a
table
and
then
pass
in
the
ability
to
have
a
test
case
that
runs
at
root
and
non-root.
That's
that's
what
I
would
that's
my
inclination
or
my
instinct
says
that
we
should
take
the
existing
tests,
identify
the
ones
that
we
really
want
to
verify,
work
on
non-root
and
not
duplicate
them,
but
just
create
a
table
to
toggle
both
running
them,
root
and
non-root,
and
let
that
be
the
initial.
D
Does
anyone
else,
I'm
talking
a
lot?
Does
anyone
else
have
any
opinion?
Other
opinions,
I'm
just
trying
to
avoid
the
scenario
where
we
I'd
like
for
this
to
be
a
pre-submit
figure
out
a
way
to
get
basic
validation
of
non-root
as
a
pre-submit?
So
we
don't
break
it
and
we
don't
end
up
in
a
situation
where
the
release
is
blocked
and
we
have
to
scramble
every
month
to
figure
out
what
happened
to
not
read.
E
Wonderful
yeah
the
tag
for
every
for
every
of
these
additional
tests
that
we
want
to
test
with
non-root
and
then
run
it
as
a
separate
job.
I
guess
I
mean,
instead
of
instead
of
creating
like
duplicating
all
of
this
code,.
E
D
E
H
E
D
D
G
D
E
I
guess
my
only
concern
is
that
how
do
we
define
what
is
it?
What
is
this
minimum
minimum
subset
that
we
consider,
because
pretty
much
every
corner
case
can
cause
an
issue.
D
D
D
No,
I
mean,
I
mean
sorry,
not
not
run
all
the
test
lanes
with
this
app,
I'm
saying
with
our
new
test
lane
when
we're
picking
the
subset
of
tests
that
we
want
to
to
execute
within
that
and
non-route.
Maybe
I
was
taking.
D
Maybe
what
we
want
to
run
everything
and
and
select
a
few
tests
not
to
run.
I
don't
even
know
if
we
can
express
that
with
our
current
test
suite
or
not.
But
if
it
sounds
like
everything
like
95
percent
of
the
tests
need
to
run
and
non-root,
then
that's
gonna
be
a
pain
to
add
that
label
to
every
single
one
of
them.
Maybe
we
can
add
a
label.
That's
opt
out
of
non-root
instead
to
the
subset
of
tests
that
can't
run.
K
We
just
want
to
make
sure
that
we
don't
break
the
feature
it
doesn't
need
to
so
I
mean
we
can
just
run
the
basic
positive
tests
for
every
feature
so
say
for
a
survey.
If
we
have
10
tests,
it's
it's
good
enough.
You're
in
one
of
them,
don't
really
care
about
all
the
details.
It's
just
to
exercise
the
positive
code
flow.
I
suppose
so
I
if
I
was
to
make
an
estimation,
I
wouldn't
say
it's
like
90
of
the
tests.
If
that's
what
you're
getting
it.
D
Okay,
well
maybe
we
don't
have
to
completely
decide
on
this
now.
I
think
the
thing
that
we
can
decide
on
is
that
we're
going
to
enable
a
new
test
lane
and
somehow
enable
subset
of
tests
or
the
whole
whatever
makes
sense
within
that
test
lane
to
exercise
non-route.
It's
just
one
test
line
that
we're.
Having
does
that
sound
accurate
to
everyone
at
least.
I
H
It
does
make
sense.
The
problem
is
that
you
you
want,
you
will
have
a
job
that
runs
everything
almost
except,
let's
say:
10
things
10
tests,
so
that
is
duplication
of
you'll
duplicate
everything,
that's
a
resource
problem,
but
you
can
you.
I
think
we
can
just
start
for
the
next
month.
Let's
say
we
periodic
and
monitor
it
and
that's
it.
D
No,
no,
no,
no
who's
going
to
monitor
this
periodic.
Is
somebody
gonna
wake
up
and
have
a
invested
like
task
to
investigate
this
periodic?
Every
time
no.
H
Because
it's
not
every
time
they,
I
think
that
we
should
care
only
about,
in
my
opinion,
either
you
care
about
one
one
feature
like
that:
you
define
something
like
conformant
test
or
something
similar
to
that,
and
you
have
a
periodic
that
runs
in
parallel.
That
will
do
cover
everything
that
you
expect.
H
D
C
C
E
France
from
what
I've
seen
at
least
this
morning,
it's
very
close-
there
are
a
couple
of
issues
that
we
can
resolve,
hopefully
very
soon,
other
than
that
it
seems
pretty
solid.
D
So
I
want
to
get
this
merged
as
soon
as
possible
that
this
is
going
to
be
a
huge
burden
to
maintain
so
what's
the
path
of
least
resistance
to
get
code
merged,
that's
executed,
we
have
a
pre-submit
executing,
for
it
sounds
like
just
a
a
new
lane
and
then
the
path,
the
transitional
path
for
us
after
we
get
this
merged
and
we
have
a
new
lane
to
exercise
some
of
it
whatever.
D
E
So
we've
split
the
our
lanes
into
several
look
at
several
groups.
Perhaps
we
can.
We
can
choose
from
from
these
groups
subset
of
of
tests
that
we
want
to
run
for
every
group
and
then
slowly
and
then
merge
this
pr
and
then
slowly
move
the
tests
from
existing
lanes
to
the
new
lanes
that
execute
exercises
the
known
route
and
then
leave
in
the
previous
lanes.
I
E
So
without
without
the
vertifs
I
mean,
srv
will
be
resolved
with
this.
Yes,
so.
E
I
So,
what's
the
outcome,
I
mean
there
were
multiple
options.
What's
the
consensus.
I
I
think
the
last
thing
last
option
was
to
have
the
pre-submit
right
with
the
features
we
can
right
now
run
as
numbered.
C
D
A
Okay,
thank
you.
Speaking
of.
D
G
One
that
runs
the
the
brow
control
plane,
and
this
was
a
problem
with
the
dns
resolution
internal
in
the
cluster
and
also
with
external
resources,
and
it
was
solved
like
the
three
at
the
night
yesterday,
yeah
it,
it
is
already
recovered.
The
the
support
people
in
in
ibm
cluster
responded
super
quickly,
but
the
resolution
was
not
that
quickly.
G
The
good
thing
is
that
is
that
this
this
cluster,
now
that
we
have
migrated
from
from
the
old
cluster
that
run
the
the
the
control
plane,
this
new
cluster
allows
this
solution,
the
node
local
dns
cache,
and
with
this
with
this
option,
there's
a
demon
set
that
is
running
a
port
on
all
the
all
the
nodes
and
the
the
queries
are
done
to
this
to
the
local
port
running
in
in
the
node
and
not
to
the
to
the
cluster
white
core
dns
parts.
G
And
then,
if
there's
there
are
queries
that
can
be
solved.
But
by
this
local
demon
set
running
locally
on
each
node,
then
they
are
not
forwarded
and
then
to
some
extent
this
can
can
be
mitigated
the
next
time
so
yeah.
Let's,
let's
see
if
we
can
improve
in
this
sense
and
be
less
affected
by
infrastructure
authorities.
G
A
Also
noticed
daniel's
been
posting
a
job
wreck
with
red
hat
on
linkedin
for
cooper,
ci.
G
A
Be
awesome,
I
think
great
to
you'll,
be
very
useful
of
for
our
non-red
hat
friends
to
see
that
job
wreck
and
yeah,
as
well
as
if
any
other
company
is,
is
hiring
for
for
cooper.
Related
work
feel
free
to
to
speak
up
and
talk
about
what
you
have
and
and
basically
use
our
forum
as
a
recruiting
mechanism.
A
Hey
speaking
of
the
control
plane,
is
that
a
a
donated
infrastructure,
or
how
does
how
does
that
work?
How
do
we,
how
do
we
acquire
that
infrastructure.
G
I
I
don't
yeah,
I
think
we
we
are,
we
are
paying
for
it
to
to
ibm,
I'm
not
completely
sure
what
what
is
what
is
a
status
here,
but
but
yeah.
I
think
it
is
it's
not
donated
or
anything
like
that.
A
A
Okay
cool,
it's
interesting
to
know
how
how
those
the
structure
bits
are
are
given
to
the
project.
A
Okay,
oh
thank
you
federico
that
takes
us
to
the
end
and
we
are
at
7
42,
pacific
time,
nothing
in
the
open
floor.
J
Just
chris
just
one
thing:
I've
just
posted
both
things,
one
the
thing
for
cupid
ci
principle,
software
engineer
offering
and
the
other
one.
We
are
also
hiring
for
openshift
virtualization,
which
also
involves
involving
in
kubrick
upstream
work.
If
anyone
is
interested,
so
just
so,
you
know.
A
Fantastic,
it's
good
to
know
that
there
are
excellent
job
wrecks
out
there
in
today's
bad
market.
A
Okay,
no
no
items
in
the
open
forum,
no
poll
requests
that
we
want
to
talk
about.
Do
we
feel
like
doing
a
bug
scrub
today,
just
10
minutes.
K
We
go
thanks.
There's
just
a
couple
of
issues
from
the
last
week.
K
All
right,
first,
one
invert
kathleen,
was
able
to
find
game
resources,
blah
blah
blah-
and
I
said
it
blue
answered
here
so
I'll
just
go
ahead.
D
D
D
So
does
the
help
task
help
text
display
the
namespace
option?
That's
what
we
need
to
investigate.
Somebody
needs
to
do
that.
You
can
just
yes
it
does.
It
does
okay.
I
Yeah
I
mean
you,
go
and
do
a
beer,
ctl
options
help
and
it
will
show
up.
D
I
D
K
K
Shall
we
just
close
it
I
mean
I
I
mean
if
somebody
doesn't
isn't
able
to
figure
this
out
after
using
cube
cuddle.
That's
I
don't
know
if
code
documenting
this
and
the
user
guide
would
make
this
easier.
If.
D
L
Queue
control
then
close.
It.
C
L
Good
luck!
All
right.
A
Oh
yes,
I
think
roman
answered
my
question
yesterday
in
mailing
list.
A
A
So
I
just
wanted
to
make
sure
that
that
that
that
dependency
list
was
still
accurate,
since
it's
been
a
few
years
yeah
it's
for.
C
Yeah-
and
I
mean
we
have
the
foster
check,
as
they
told
you
there,
where
you
can,
where
it
even
scans
our
reap
and
all
dependencies,
it's
just
pretty
funny.
I
think
we
installed
at
some
point
with
pip
j2,
just
for
one
test,
where
we
check.
If
a
template
is
ginger
too
compatible
and
fossa
pulls
in
into
the
report.
All
changer
2
dependencies
is
a
project
dependency,
which
is
pretty
funny.
C
C
C
A
I'm
not
sure,
I'm
still
in
just
going
through
all
these
bullet
points
and
just
saw
this
one
and
created
an
issue
on
it.
A
K
C
In
general,
we
tend
to
support
the
same
authentication
plugins
like
cube
ctl,
but
if
they
host
it
somewhere
and
download
the
cubesat
from
there,
there
may
be
additional
authentication
plugins
it
compiled
in
the
cube,
cdl
and
then
may
not
be
able
to
talk
to
the
cluster,
but
is
there.
It
can
also
be
something
completely
different.
C
K
C
No,
it's
it's
available
as
a
preview
in
0
41..
It
goes
in
0.
C
C
K
C
K
C
Oh
yeah,
the
super
hot
with
word
ctl.
C
K
C
K
Let's
think
it
does
this
whatever
and
close
it.
K
Okay:
let's
try
to
get
these
two
as
well,
so
we
get
it
to
one
week
how
to
make
a
snapshot
and
one
click
recovery.
Do
we
do
have
relevant
documents.
C
C
C
C
C
Yes,
for
the
rollback
I
mean
you
can
advise
him
what
to
do
for
the
rollback
apart
from
investigating,
if
there
is
a
real
issue
for
the
rollback,
what
you
probably
not
do
is
setting
the
rollback
version
in
the
keyboard
cr.
So
you
probably
just
roll
back
the
operator
manifest
which
should
not
do.
He
should
stick
with
the
041
operator,
but
specify.
C
K
K
K
K
K
K
K
A
Thank
you
peter.
Thank
you.
Okay.
That
concludes
this
week's
edition
of
kubert
weekly
meeting.
Thank
you
for
joining
us
everybody,
and
we
will
see
you
next
week.