►
From YouTube: KubeVirt Community Meeting 2022-11-23
Description
Meeting Notes: https://docs.google.com/document/d/1kyhpWlEPzZtQJSjJlAqhPcn3t0Mt_o0amhpuNPGs1Ls/
A
C
I
have
tried
the
link
for
the
cupboard
calendar
that
we
have
in
our
community
repo
and
it's
empty.
Is
it
me
doing
something
wrong?
Have
we
changed
something.
B
A
B
B
B
A
B
A
Good
deal
all
right,
then,
while
we're
here,
if
everyone
can
please
take
a
moment
to
pull
up
the
agenda,
notes
and
log
attendance,
that
would
be
greatly
appreciated.
A
Of
course,
if
we
have
any
new
members
joining
today-
and
you
would
like
to
introduce
yourself
say
hello-
love
to
welcome
you
hear
what
you're
using
cooperatform,
what
brought
you
in
today.
A
Please
be
sure
and
add
anything
that
you
would
like
us
to
cover
today
to
agenda
open
floor
or
if
you
have
specifications
bugs
bugs
you
want
to
look
at
or
PRS
or
anything
like
that.
We
can
go
ahead,
cover
those
in
their
respective
sections
to
call
out
anything
that
you're
interested
in
just
paying
special
attention
to,
and
we
will
get
to
that
in
a
bit
foreign.
D
Yes,
sure,
oh,
can
you
hear
me?
Yes,
okay,
cool
okay,
so
let
me
give
you
like
a
little
bit
of
context.
This
is
a
feature
I've
been
working
on
and
off
for
the
past
year
or
something
it
was
originally
stuck
on
on
maltes.
Basically,
we
want
to
add,
like
interfaces
in
runtime
to
running
virtual
machines,
but
since
the
virtual
machines
run
inside
pods,
we
need
first
add
interfaces
to
do
the
pods,
and
that
was
something
that
took
quite
a
lot
of
time
to
get
like
the
momentum
running
in
in
maltes.
D
We
eventually
did
it
and
now
we
can
unlock
the
the
part
of
cupert.
The
thing
is
there
was
an
approved
design
document
like
I,
don't
know
some
12
15
months
ago.
It's
linked
there.
That's
it
was
the
one
below
that
one
actually.
D
Yes,
that
one
original
design
document
and
yeah,
so
the
this
proposal
describes
both
the
part
in
maltes
but
and
the
part
in
keyword
what
needs
to
be
done
like
not
in
too
much
detail,
but
it
gives
like
a
broader
look
into
into
it.
If
you
could
like
just
scroll
down
to
the
bottom
of
it,
please
I
think
I
can
elaborate
a
little
bit
up.
D
Like
the
API
part
of
the
vmis
a
little
bit
further
down,
that's
it
like
the
thing
with
the
phases.
So
this
is
something
like
the
current
exactly
that.
Thank
you.
The
current
implementation
that
I'm
providing
in
an
open,
PR
I,
forgot
to
link
it
I'll,
add
it
in
in
a
while
it
basically
mimics
what
we're
already
doing
for
the
this
hot
plug
part.
So
it
mostly
has
like
a
all
the
components
of
Hubert
like
convert,
controller,
vert,
Handler
and
launcher
they
work.
D
They
try
to
figure
out
like
that,
the
desired
State.
What
we
have
in
the
in
the
networks.spec,
sorry
in
the
spec.networks,
is
different
than
what
we
have
in
the
in
the
current
state,
like
in
the
VM
in
the
VMI
status.
So
the
version
controller
notices
this
difference
and
it
starts
to
do
things
and,
as
like,
the
different
components
try
to
reconcile
the
state.
D
We
create
this
thing.
This
interface
hotbook
status,
and
this
has
a
phase
and
has
more
components
like
do
what
you're
doing
like,
because
we
need
to
First
hot
plug
into
the
pods.
Then
the
vert
Handler
needs
to
to
create
the
a
bridge
inside
of
the
Pod
a
top
device
configure
the
top
device
like,
as
those
things
happen
like
each
of
these
components,
would
update
the
state
of
the
of
the
VMI,
and
this
is
kind
of
like
a
state
machine
kind
of
thing,
and
this
is
something
that
well
I.
D
D
Yeah,
so
if
you
could
just
mute
please
if
the
idea
here,
instead
of
maintaining
this
the
state
machine
that
I
it's
more
error
prone,
would
be
to
have
like
an
approach
where,
if
the
controller
sees
that
there's
a
difference
between
the
desired
State
and
the
current
state,
it
would
just
like
get
maltes
to
Hot
plug
into
the
Pod,
and
it
will
continue
to
do
that
until
the
state
is
exactly
what
it
expects
it
expects
like.
The
the
Handler
would
afterwards
just
do
the
same
thing.
D
It
would
try
to
reconcile
the
state
and
if
it
doesn't
see
the
state
that
it
expects
it
would
try
to
create
the
bridge
and
try
to
create
the
tap
like.
If
everything
is
item
potent
it'll
just
like
do
nothing,
it
will
not
fail.
It
will
just
like
assuming
the
bridge
was
already
created,
it'll
try
again,
and
the
bridge
is
already
there
so
it'll
do
nothing
and
keep
on
being
happy
until,
like
the
launcher,
like
the
launcher,
pod
actually
manages
to
add
the
running.
D
The
the
new
interface
into
the
running
pod.
When
that
happens,
like
we'll
have
the
new
interface
and
in
the
current
state,
and
nothing
else
will
need
to
reconcile
like
everything
is
as
they,
the
the
everything
is
as
it
is
meant
to
be
so
I
guess
what
I'm
trying
to
get
at
is
that
I
really
want
to
get
rid
of
all
these
phases
that
they
basically
smell
of
a
state
machine
and
have
everything
stateless
at
the
cost,
Pro,
possibly
of
a
little
less
visibility
on
the
VMI
status
of
what
is
happening.
D
While
the
reconciles
are
happening
and
I
know,
I
spoke
a
lot
and
I,
probably
just
it
was
a
convoluted
explanation
and
I
I
would
welcome
any
feedback
and
any
questions
you
have
about
this
as
well.
E
Work
Miguel
can
I,
ask
you
how
you're
going
to
add
the
top
interfaces
without
the
state
not
sure
how
it
can
be
handled
for
the
virtual
machine
for
now
we're
running
the
binary
with
adding
the
interface
inside
the
pot
for
the
virtual
machine.
D
E
But
I
guess
that
at
some
point
we're
having
the
interface
inside
the
Pod,
how
we
can
edit
the
devital
machine
and
exactly
what
exactly
we
want
to
reconcile
there.
D
So,
okay,
this
thing
will
start
like
there
will
be
like
a
difference
like
you
wish
to
recommend
like
a
vertical
command,
exactly
like
it
is
done
for
the
disc
hot
plug
and
the
vert
cuddle
will
issue
a
command
to
the
API
and
it
will
change
it
will
update
it
will
mutate,
the
the
VMI
spec
and
it
will
add,
like
an
entry
to
the
networks
and
an
entry
to
the
interfaces.
D
Then
you'll
have
three
components
like
The.
Verge
controller
is
the
first
one
and
it'll
see
that
like
they
will
all
see
that
the
there's
a
difference
in
between
the
expected
State
and
the
current
state,
and
they
will
each
try
to
do
their
thing
so
first,
the
word
controller
will
see.
Okay,
I
need
to
ask
maltes
to
put
me
another
interface
into
the
pod.
I
will
afterwards
like,
send
you
more
information
about
how
we
got
models
to
do
this,
so
you
can
check
it
afterwards,
but
let's
just
focus
on
keyword
for
now.
D
D
The
vert
Handler
will
try
to
create
the
input
bridge
and
the
top
device
and
connect
the
Pod
interface
to
the
input
bridge.
If
the
until
the
Pod
interface
is
there,
it
will
fail,
but
it's
it's
a
reconcile.
Loop
it
will
try
to
keep
on.
It
will
keep
doing
try
to
do
the
same
thing
until
it
manages
to
so
when
the
Pod
interface
becomes
available
in
the
pods
this
next.
This
next
step,
like
the
step
that
word
Handler,
does,
will
succeed.
Like
you,
will
you
have
a
pod
interface
you'll?
D
Have
the
interface
connected
to
a
newly
created
input
bridge
for
discon,
new
interface
and
you'll?
Have
a
newly
created
and
configured
type
device,
the
first
launch,
which
would
be
the
final
component?
It
would
do
the
same
thing.
It
will
initially
fail
because
it
doesn't
have
the
tab
device
it'll
reconcile
until
it
finds
at
that
device
once
the
tab
device
is
there,
it
will
just
do
the
pr
that
I
will
link
afterwards.
D
Also
already
has
that
call
it's
something
like
attached
device
or
something
you
just
give
it
like
the
snippet
of
XML
and
it'll,
just
plug
it
into
the
the
running
virtual
machine,
and
once
that
is
in
the
virtual
machine,
it
is
reported
by
the
domain
and
it
will
be
listed
in
the
VMI
dot
status,
dot
interfaces
once
that
happens,
everything
will
stop
reconciling
because
the
desired
State
and
the
current
state
now
matches.
F
Welcome
Hey,
can
you
hear
me?
Yes,
hi
Miguel,
so
I
think
thanks
a
lot
for
bringing
this
topic
up.
I
think
the
overall
idea
of
removing
phase
sounds
very
like
it
sounds
in
line
with
the
conventions
that
kubernetes
Community
is
advocating.
I've
tested
the
source
and
code
from
there.
F
This
used
to
be
a
good
practice
when
kubernetes
started
in
the
early
versions,
but
we've
started
moving
out
of
phase
into
conditions
now,
as
far
as
I
understand
all
the
state
that
you
need
in
order
for
reconciling
reconciliations
to
be
idempotent,
you
have
that
in
in
the
form
of
pod
interfaces
and
BMI
interfaces.
The
only
question
I
think
you
have
is
that
if
we
remove
the
face,
there
will
be
some
sort
of
loss
of
verbosity
where
users
will
not
be
clearly.
Users
will
not
clearly
be
able
to
understand.
F
Have
you
considered
raising
events
warning
events
for
failure
cases
so
I
think
what
what
needs
to
happen
is
in
a
happy
case
when
things
work
it.
It
might
be
okay
to
not
have
that
information,
since
this
is
a
transparent
operation,
but
in
you
know,
failure
scenario,
it
will
be
critical
to
understand
what
is
failing.
F
I
would
so
just
looking
at
raising
events
so
and
just
to
backtrack,
where
this
comes
from.
So.
F
Mount
errors
in
pod
right
now,
if
PVC
is
not
able
to
mount
on
a
pod,
you
don't
see
it
in
the
Pod
API,
but
you
do
see
events
being
raised
like
Mount,
attach
error,
and
things
like
that.
I
see
this
very
analogous
to
that
and
you
could
consider
using
events.
D
Yeah,
that's
that's
very
interesting,
like
we're
my
current
implementation,
USE
events
but
I
think
it's
actually
does
the
opposite
like
it
does.
It
sends
events
when
everything
succeeds
which
is
kind
of
well
redundant
because
that
will
be
noticed
in
the
state
of
the
VMI.
That's
a
really
good
point.
F
D
Okay,
thank
you
a
lot
for
for
that
feedback
and
yeah.
That
was
like
the
the
initial
part
you
said
like
to
report
stuff,
not
via
phases,
but
using
the
conditions
is
something
that
yeah,
depending
on
the
feedback
I
got
today.
I.
That
was
the
next
thing.
I
try
I
would
try
to
do.
F
Yeah
I
to
I'm
not
sure
if,
if
this
hot
plug
will
be
rolled
into
one
of
the
existing
conditions
or
we
will
introduce
new
ones,
I
I
need
to
dig
into
it,
but
certainly
adding
events
will
give
you
visibility
into
what's
happening
with
the
reconcile
foreign.
E
D
Well,
thanks
a
lot
for
your
feedback.
Anyone
else
has
more
to
add
to
this
topic.
D
Okay,
if
not
I,
think
we
can
move
on
thanks
everyone
for
your
for
your
for
the
info
and
the
feedback
you
gave
me.
A
Right,
let's
check
out
state
of
C
group
V2
support.
You
want
to
take
that
away.
E
Yep
I
just
wanted
to
ask
I
started:
I
tried
to
use
Cube
version
latest
Ubuntu,
Jammy
and
I
found
that
some
features
are
not
working,
for
example,
attaching
how
to
plug
devices
block
devices
and
not
hot
black
hot
black
blog
devices
make
them
not
working
and
I
found
that
there
are
some
issues
with
the
c
groups
version
2,
mainly
because
of
the
current
solution
does
not
allow
you
to
see
the
current
rules
for
the
devices
and
I
just
wanted
to
ask
if
anybody
using
it
in
production
and
what
the
current
state
of
c
groups
version
2.
E
G
So
this
is
a
first
herd.
That's
a
real
surprise
to
me.
We
do
aim
to
support
c
groups,
V2
rel9,
which
is
out
ships,
c
groups
V2
enabled
by
default.
So
the
fact
that
we're
hearing
issues
with
c
groups
at
this
point
is
kind
of
yikes.
We
we
intended
to
be
supporting
it
already.
Are
there
bugs
for
this
Upstream
I
see
the
GitHub
issue
here?
Is
that
capture
the
sum
total
of
what
the
problem
is.
E
So
the
problem
is
that
there
are
some
rules
added
by
kubernetes,
for
example,
if
you're
using
block
devices,
it's
attaching
the
the
rule
for
a
low
Port
using
this
device,
and
when
we
running
our
C
group
manager,
it
always
overrides
these
rules.
B
G
E
Yes,
I'm
using
latest
version
little
bit
patched,
but
this
case
is
not
working.
There.
I
saw
somebody
started
working
on
that
issue
exactly
on
this
issue,.
E
H
This
and
I
was
by
chance
also
looking
into
what
we
can
do
to
improve
the
situation.
E
Yeah
also
just
thinking
what
kind
of
company
cases
it
can
lead
to
us.
We
can
handle
the
devices
because
we
know
the
pods
back
and
the
virtual
machines
back
actually
to
see
the
blog
devices
and
add
the
specific
rules
for
that.
But
is
there
anything
else
which
we
can't
control?
E
Because
currently
we
can't
find
what
the
current
rules?
The
c
groups
attached
to
the
epoxy
group.
H
Yeah
so
about
that
I
was
thinking.
Maybe
we
could,
because
a
lot
of
kubernetes
deployments
will
be
moving
to
systemd
as
a
manager
of
c
groups-
I'm
not
entirely
sure,
but
maybe
systemd
will
be
keeping
the
list
of
devices
that
are
allowed
for
a
unit
or
for
the
for
the
Pod
or
container
and
I
guess
we
need
to
check
if
that's
the
case.
If,
if
that's
the
case,
we
can
probably
integrate
with
systemd
and
take
the
devices
from
there,
but
it
would
be
only
solution
for
systems
that
are
using
systemd.
B
E
You
I
got
something
from
psyllium
chats
and
they
saw
that
there.
They
say
that
there
is
some
options
for
that,
but
I
guess
in
output,
you'll
get
the
low
code
which
is
not
easy
to
read
or
somehow
bars.
H
I
see
him
yes,
well,
you
no
I'm
sure
how
how
it
will
be
returned,
but
if
it's.
H
A
F
I
I
was
trying
to
dig
into
the
changing
Behavior
just
to
understand.
Api
compatibility
across
upgrades
and
I
was
not
sure
what
change
is
being
suggested.
G
So,
to
recap,
what
the
email
is
basically
saying
is
that
the
current
behavior
right
now,
if
you
have
a
data
volume,
that's
set
to
Auto
populate
and
you
delete
it.
It's
going
to
recreate
a
new
one
and
that
behavior
bad
or
good.
It
might
be
surprising
to
some
people
if
we
reversed
that
behavior,
which
is
why
she's
proposing
this
is
a
change.
But
the
new
behavior
that
she's
suggesting
is
that
the
data
volume
would
not
be
recreated.
If
you
then
did
a
delete
and.
G
F
Yeah
makes
sense,
so
I
I
would
recommend
that
we
not
break
API
behaviors
in
for
reasons
which
are
very
well
established.
I
think
what
I
would
suggest
is
ADD,
adding
a
spec
field
which
is
Backward
Compatible.
That
allows
us
to
define
a
policy
that
policy
will
default
to
what
it
is
default
to
the
current
behavior.
So
any
upgrades
will
continue
to
work
and
then
users
can
choose
to
have
a
different
policy
that
implements
the
new,
changing,
behavior
and
then
backups
or
use
cases.
F
On
top
of
that,
new
policy
can
continue
to
work
with
a
different
spec
value.
F
J
No
I
what
you're
saying
definitely
makes
sense,
but
Shelly
the
author
of
that
email
Isn't
in
this
meeting.
If
you
can
respond,
that
would
be
helpful.
Yeah.
I
I
Great,
that's
a
good
precondition.
I
failed
to
add
something
to
the
minute:
stop
because
I
failed
to
open
it,
but
I
would
like
to
quickly
speak
about
the
work
that
I'm
currently
doing.
I
send
an
email
a
few
weeks
ago
about
reordering
how
we,
how
we
don't
matter
to
prove
us,
but
how
we
give
a
growth
part
to
approvers.
Now
we
spread
the
load
more
and
I
just
wanted
to
use
that
opportunity
to
speak
about
it
if
they,
if
that's
fine,.
I
And
I'll
take
all
of
this
to
to
the
mailing
list
as
well,
but
I
thought
I
used
the
opportunity,
Well
everybody's
here,
to
ask
to
answer
questions
if
there
are
any
to.
A
I
More
than
what
I
intend
to
okay,
so
this
about
mainly
the
approvals
of
cubic
Hubert,
and
so
today,
let
me
just
to
reorder
that
okay,
so
today,
approvals
are
here
to
help
I
mean
approve
the
generally
here
to
help
anybody
right
to
get
stuff
into
cubic
cubert,
and
we
have
contributing
guidelines
of
how
you
become
an
improver
and.
G
I
Have
a
few
if
you
look
at
the
owners
files,
and
so
that's
good.
I
I
think
the
second
bullet
illustrates
a
little
bit
the
one
of
the
problems
that,
with
the
approvers
approvals,
are
focused
or
are
asked
to
focus
on
a
realistic
holistic
symptoms
of
contributions,
including
backwards
forwards.
Compatibility
adhering
to
API
and
flag
conventions
simply
performs
the
correctness
issues
in
Corrections
with
other
parts
of
the
system,
Edison
DC.
I
B
I
I
I
G
I
But
today
each
of
these
teams
is
actually
maintaining
their
own
knowledge
for
the
kubernetes
problems
right.
How
do
we
write
a
great
API?
How
do
we
ensure
that
that
scale
will
be
carried
forward
right
that
we
continue
to
scale?
What
considerations
do
we
have
to
do
for
live
migrations
for
for
backups
for
updates
right?
All
of
this
is
now
built
or
maintained.
I
Within
These
is
these
zigs,
more
or
less,
and
only
aggregated
or
reviewed
when
an
approver
comes
in,
so
one
of
the
ideas
is
to
say:
let's,
let's
actually
try
to
Define
these
areas
that
an
approver
is
taking
care
of
with
that
the
goal
to
say
now,
if
we
have
any
skilled
person
who.
I
Or
who's
really
motivated
to
build
up
these
skills
to
be
in
one
of
these
sub-approvals
or
subsystem
maintainers,
which
is
like
that
second
group
of
bullets
here
and
what
which
I'm
calling
smbs
right
now,
but
maybe
actually
subsystem.
Maintenance
is
better
yeah
and
we'll
start
with
defining
these
important
areas.
I
Introduce
these
super
progress
for
these
areas.
The
super
progress
of
subsystem,
Maintenance
I,
would
call
them
smes
and
smes
and
zigs
are
required
to
cooperate
voluntarily
right.
So
it's
not
mandatory,
for
example,
that
an
smes
is
is
approving
the
work
that
the
zik
did
or
the
other
way
around.
I
It's
it's
required
to
be
as
friendly
right
and
smes
should
be
like
the
the
first
gate
before
you
get
the
final
approval
right,
and
they
are
your
subject
matter,
experts
if
you
need
to-
or
they
are
the
go-to-peer
personally,
if
you
have
a
specific
request
about
one
of
these
important
areas
and
ideally
by
the
way
Zig
members,
the
Vic
members
volunteer
members
or
are
parts
of
these
SME
groups,
because
how
can
we
meaning
for
what
are
meaningful
areas
right
and
the
areas
that
when
I
spoke
to
a
couple
of
people
and
like
the
list
that
it
we
we
came
up
with
and
by
the
way,
that's
like
the
third
or
fourth
iteration
is
an
API
and
controllers
right.
I
They
are
related.
C
B
I
There,
like
all
the
discriminator
discussion
in
kubernetes
and
how
that
Maps
down
to
us,
that
is
something
where
I
think
somebody
needs
to
should
have
you
know
an
impression
of
how
our
API
is
built
and
designed
in
kubernetes
to
ensure
that
we
can
then
provide.
You
know
continue
to
have
a
consistent
API
across
all
parts
of
keyword.
Controller
patterns
like
shared
informers
and
I.
Think
there
were
some
changes,
I
think
Michael
Hendrickson
was
looking
into
that,
so
I
think
control
pan
is
also
right.
I
This
is
the
roughly
the
area
that
somebody
could
concentrate
on
scheduling
and
not
Plumbing,
so
they
also
also
both
relate
right,
because
scheduling
is
so
heavily
built
around
labels
and
label
selector
mechanisms
and
no
properties.
That's
why
they
were
grouped
together
and.
I
This
effectively
also
includes
cni,
CSI
and
device
plugins
specifically
CSI
device
plugins
also
denote
resources
right,
and
that
is
why
we
moved
them
into
one
bucket
right
and
by
the
way,
within
such
a
group,
you
can
still
have
your
expert
for
for
Niche
cases
right
specifically
for
cni
or
specifically
for
CNR
specifically
for
device
plugins,
but
it
would
also
work
out
double
cover
because
it's
no
Plumbing
acid
Linux.
Second,
all
these
are
anything
that's
happening
and
taking
place
on
the
Node
scale
and
observability.
So
Ryan
was
pretty
active
in
the.
B
G
I
And
what
it
actually
saws,
by
the
way
that
this
scale
work
or
to
make
sure
that
we
can
scale
was,
was
accompanied
by
actually
observability
work
right.
The
metrics
were
introduced
in
order
to
be
able
to
measure.
Do
we
actually
scale
and
that's
why
scale
absolutely
relate
to
each
other,
because
observability
is
always
needed
in
order
to
to
make
sure
yeah
that
we
can
measure
what
we
believe
we
can
by
the
way.
I
That
reminds
me
that
observability
is
also
an
important
tool
for
operations,
and
operations
is
not
represented
here
in
these
important
areas,
then,
with
the
old
CI
testing
framework
SME
group.
So
we
have
awesome,
dedicated
people
who
are
maintaining
our
CI
Let
It
Be,
delaying
from
keyboard
CI
and
by
the
way.
So
we
have
that
whole
maintain.
I
We
have
maintained
keyword
CI
to
you
know,
which
are
the
foundations
for
our
lanes.
People
are
doing
it
and
I
think
would
be
nice
to
form,
for
example,
formalize
this
group,
which
is
doing
this
full-time,
almost
cubert,
CI
and
pro,
for
example,
me
that
plays
into
the
testing
framework.
So
this
SME
group
would
be
you
know,
doing
everything,
except
for
writing
and
fixing
test
cases,
for
example,
or
fixing
bugs
which
which
arise
from
test
cases.
These
are
really
specific
to
the
group.
We're
just
writing
them
right.
A
little
bit.
I
Updates
that
need
to
go
through
Babel
right,
so,
if
they're
rendering
updates,
then
these
need
to
be
managed.
They
need
to
be
pushed
forward.
So
that
is
why
the
build
system
SME
group
was
suggested
as
well
and
in
the
end,
I'm,
not
sure,
but
VM
lifecycle
and
live
migration.
It's
a
life
cycle
specifically
to
say
through
What
stages,
does
a
VM
actually
run?
Does
it
make
sense
right
how,
for
example,
if
we
think
about
the
backup
and
restore
right
in
what
states
do
VMS
need
to
be
suspend
resume?
We
had
them
or
pause
VMS.
I
So
that's
the
life
cycle
and
then
that's
all.
But
to
me,
life
cycle
is
probably
pretty
weak.
I
I
still
struggle
a
little
bit
to
define
it
crisply
like
migration,
it's
like
a
broad
topic
and
why
doesn't
only
fall
on
computer,
and
that
is
because
it's
touching
everything
right.
How
we
do
we
deal
with
storage
during
live
migration
without
networks
right
and
Android
can
see
your
PR
by
the
way
about
the
unplug
and
replug
of
network
through
lock,
migration
I
think
that's
a
great
idea.
I
It
would
be
interesting
to
see
hope
that
can
be
made
if
it
makes
it
way
into
a
cubits
Itself
by
the
way.
So
these
are
like
important
areas
that
got
identified
and
I'll
be
sharing
that
that
I
hope
later
this
week
to
I
mean
I
shared
another
deck,
but
that
was
even
worse.
I
So
this
I
hope
is
better
I'll
share
it
towards
the
end
of
the
week
to
get
your
input,
and
so
my
proposal
would
be
to
complete
the
ZX
and
formally
introduce
them
in
our
owners
files
and
to
our
companies
by
smes
or
actually,
alternatively,
is
a
subsystem
maintainers
right
to
pick
up
the
kernel
term,
subsystemators,
so
I'll
be
going
forward
to
to
see
that
we
to
propose
this
change
also
by
the
way,
I
think
if
you're
a
subsystem,
maintainer,
there's
less
time
and
less
expertise,
you
need
to
invest
in
order
to
be
such
a
maintainer,
so
I
hope
with
such
a
change,
we're
actually
getting
more
approvers
right
or
more
people
who
feel
competent
to
to
approve
PRS.
I
What
is
going
on
simply
to
help
us
to
scale
right
and
to
make
sure
that
we
don't
drop
the
ball
in
an
area
any
important
area
and
to
move
forward?
So
that's
a
quick
update
from
my
side
on
this
here
to
hear
any
feedback
right
now,
but
also
later
on.
The
mailing
list.
F
I
I'm
a
like
thanks
a
lot
for
the
initiative.
I
think
it.
It
makes
a
lot
of
sense
and
yeah
I
agree
with
the
problem
and
the
attempt
at
solving
that
problem.
I
have
some
questions
so
on
the
important
areas.
I
saw
API,
review
and
controllers
I.
Think
one
of
the
things
that
we've
been
struggling
with
running
keyboard
is
upgrades
and
maintaining
backward
compatibility
during
upgrades.
F
I
First,
thanks
for
that
company,
that's
great
so
I
mean
before
I
answer.
Let
me
ask
so
what
specific
struggles
did
you
have
during
upgrades?
Was
it
like
word
API
things,
or
did
you
have
other
yeah.
F
It
was,
it
was
API
thing,
so
one
of
the
issue
was
that
default
was
changed
to
a
new
API
field
was
introduced,
while
that
API
field
was
itself
Backward
Compatible
the
behavior
defaulted
to
false
instead
of
true,
so
the
behavior
was
not
a
backward
compatible
and
that
led
to
some
failures,
and
then
there
were
other
two
other
cases
when
we
also
saw
failures,
but
to
be
fair,
we
were
upgrading
from
version
0.35250,
so
there
was
a
lot
of
lag
and
things
changed
so,
but
regardless
that
points
to
an
important
area
that
I
think
we
should
all
be
focused
towards
that
is
making
sure
that
upgrades
are
upgrades
are
handled
correctly
and
we
have
a
very
defined
process
for
how
many
versions
We
support
behind
the
latest
version
right.
F
So
kubernetes
s,
plus
minus
three
and
I
think
Ryan
is
doing
some
work
in
order
to
stabilizing
the
release
and
these
process
changes
can
follow
that
release
schedule.
So
anyway.
That
was
a
lot
of
talk,
but
just
wanted
to
get
your
thoughts.
I
Oh
yeah,
but
no
worries
about
the
the
length
of
your
of
your
contribution
right
now,
because
I
spoke
for
much
longer,
so
I
think
backwards.
Compatibility
is
so
it's
great
by
the
way.
I
think
it's
good.
If
you,
if
you
encounter
some
things,
I,
think
it's
good
that
you
start
filing
bugs.
If
you
see
like
backwards
incompatible
changes
file,
bucks
about
them,
because
I'm
not
really
sure
that
somebody
was
aware
that
we
break
backwards,
broke
backwards.
Compatibility
I,
know
I'd,
like
of
two
cases.
I
For
example,
we
drop
floppy
support
silently
at
some
point
in
the
past,
while
it's
still
in
the
API,
but
it
has
no
effect
effectively.
It's
blocked,
I
think
with
an
admission
controller
for
that
and
then
I
think
in
the
bridge
and
binding
and
the
network
binding
modules
that
were
probably
also
some
breakages
simply
because
it
was
so
messy
to
get
that
somewhere,
decent,
but
yeah
I
think
backwards.
Compatibility,
that's
something
to
consider
here
on
the
API
level.
I
H
I
For
example,
for
the
live
migration
side
and
node
Plumbing
as
well
to
a
certain
extent,
so
I
see
several
areas,
but
API
might
actually
be
the
most
important
one.
So
I
added
it
here,
yeah,
okay,.
F
Awesome
and
I
think
in
one
of
the
previous
Community
called
I,
raised
this
topic
of
enhancing
the
API.
Well,
adding
some
new
things
to
the
API
review
process.
You
don't
want
to
redo
the
whole
thing,
but
it
is
on
my
to-do
list
to
summarize
that
conversation
and
putting
put
it
on
the
mailing
list
so
I.
What
was
being
proposed
was
in
line
with
this
proposal,
and
API
review
could
be
like
one
of
the
first.
F
I
B
F
I
had
another
question,
so
if
the
last
Point
says
VM
lifecycle
in
live
migration,
while
live
migration
makes
sense,
do
you
think
VM
lifecycle
is
also
somewhat
largely
related
with
API
and
controllers,
and
if
there
is
a
significant
overlap
is
like?
Is
that
something
that
should
be
mentioned
there
as
well.
I
You
know
what
I'm
I'm
with
you,
because
I
mean
veeam
lifecycle
was
a
recent
comment
and
I
appreciate
that
comment.
But
to
me
it's
difficult
to
isolate
it
to
you
know
to
really
scope
it
I
I
have
made
challenges
with
it.
You
know
what
I
will
do
I'll
go
back
to
that
person
who
suggested
it
and
see
how
that
person
was
imagining
this
this
area.
You
know
how
how
that
person
is
matching
the
frame
of
that
area
and
then
I
would
get
back
here.
I'll.
I
Let
me
take
a
note
actually
because
again
it's
to
me
how
what
this
this
code,
yes.
F
Myself,
no
I
I,
think
where,
where
you
are
going
with,
this
makes
sense.
The
only
reason
why
I
bring
it
up
is
if
there
is
an
so
we're
introducing
a
new
process.
If
there
is
any
kind
of
ambiguity,
I
I
am
little
bit
concerned
about
the
implementation
of
that
process,
so
at
least
from
the
start,
I'm
trying
to
make
this
as
crisp
as
possible
so
that
all
the
stakeholders
are.
You
know,
on
the
same
page.
I
Much
appreciated,
thank
you
very
much
yeah
and
in
the
because
you
mentioned
it,
I
put
up
a
PR
this
one
871,
which
is
actually
about
reworking
the
resemd
file
to
to
reflect
the
new
release
by
Cadence
that
we
were
discussing.
And
finally,
there
were
some
rough
agreements
and
that's
why
we
put
it
up
here
for
for
review,
so
feel
free
to
determine
here
as
well
sure.
A
A
A
Okay,
there
we
go
issue
with
uploading
using
the
cube
cuddle
bird
plug-in.
E
Try
to
link
it
to
the
CDI
issue,
but
not
sure
if
it
will
fix
anything.
It
is
hanging
there
for
a
couple
of
times.
G
G
Like
so
yeah
I
guess,
I
guess
I.
I
Can
could
probably
assist
here?
Would
you
probably
see
this
to
me
then
I
could
take
a
look.
A
E
Is
there
any
and
sudo
does
not
work
as
well.
D
I,
don't
remember:
if
can
you
scroll
down
to
the
bottom,
like
that
network
configuration
cloud
in
it
using
network
data,
a
version
one
well,
first
of
I
am
I.
Don't
currently
remember
if
this,
if
0
West
actually
understands
any
of
that
right
and
if
it
does
I
mean
this
is
currently
being
tested
with
Fedora
I'm,
quite
sure.
A
D
I'm
not
sure
if
this
is
a
COS
issue
or
if
it
is
like
a
net
plan
version.
One
issue
not
sure.
E
I
B
D
F
So
I
have
a
slightly
different
point
two
days
as
well.
So
if
you
go
back
to
the
error
message,
it
says:
RT
net
link,
error,
I,
have
seen
those
kinds
of
error
when
when
there
is
some
kind
of
security
context,
permissions
that
are
not
allowed
to
execute
those
IP
route
commands.
So
could
it
be
possible
that
the
container
is
mounted
with
config
in
such
a
way
that
it's
not
it
does
not
have
enough
permissions
to
carry
out
those
network
operations
and
and.
F
D
Like
the
weird
thing
as
it
like
is,
if
this
is
Cloud
in
it,
which
it
looks,
it
is
from
the
configuration
that
he's
providing
in
the
bottom
that
happens
inside
of
the
guest.
It's
not
like
your
pod
configuration
shouldn't
impact
what's
happening
inside
of
the
guest.
F
A
That's
where
I
was
headed
with
the
comment:
I'm
not
familiar
enough
with
the
serious
image
she
even
know.
If
we
have
any
privileges
in
that
image
or.
B
A
I
think
I'm
gonna
run
with
this
one,
just
asking
for
them
to
try
something
else
and
then,
as
far
as
I'll
tell
you,
what
I
will
do.
B
E
Hey
I
forgot
to
say
you
know
that
image
about
the
image
uploading,
I
haven't,
read
it
fully
now,
I
read
and
I
see
that
this
is
different
problem,
so
this
is
just
about
Missing
error
message
that
Syria
is
not
installed.