►
From YouTube: KubeVirt community meeting 2021-03-10
Description
KubeVirt community meeting 2021-03-10
A
A
C
A
A
D
D
This
does
change
a
few
things
and
the
idea
is
that
now
we
have
naming
so
we
do
the
association
between
objects,
so
that
things
can
be
overridden
and
attached
as
part
of
the
pool
which
kind
of
makes
sense
because,
like
when
we
we
think
about
doing
these
likes
object
overrides
like
climate
data.
Virtual
machines
are
sorry.
Pvcs
would
be
another
one
like
if
you
want
to
attach
disks
to
individual
or
you
want
each
individual
vm
to
have
its
own
disk.
This
would
be
another
example.
D
I
don't
really
know
what
this
if
this
is
the
right
name
for
this,
but
the
idea
here
is
that
now
that
we
have
sort
of
names
for
these,
like
you
know
whatever
they
are,
I
have
like
gs012.
So
we
have
this
increasing
postfix.
You
know
how
should
it
scale
up?
You
know
if
we
have
think
about
like
the
staple
set.
If
you
want
three
things
in
a
staple
set,
we
create
the
first
item.
D
D
Let's
say
if
I
have
this
field
threads
and
I
set
it
to
one,
it
would
imply
that
we
create
the
first
vm
in
our
pool
and
we
wait
till
it's
running
before
you
create
the
second
one,
the
third
one
and
so
on,
but
doesn't
have
to
be
this
way.
It
could
be
that
we
do
multiple
threads
and
we
have
sort
of
multiple
api
calls
and
we
don't
do
waiting.
E
E
The
same
sort
of
issue
is
hit
with
deployments
on
pods.
I
would
imagine
as
well-
and
I
don't
know
how
that
is
handled.
So
if
somebody
sets
that
they
want
10
000
pods,
I
don't
think
they
create
ten
thousand
pounds
all
at
once.
I'm
pretty
sure
when
I
looked
at
the
well,
ultimately,
it's
a
replica
set
was
doing
it,
but
I'm
pretty
sure
it's
batched
and
I
don't
know
if
the
user
has
control
over
how
those
are
batched.
But
I'd
say
that
would
be
important.
E
D
Yeah
there
was
there
wasn't
in
jobs,
there's
two
there's
two
fields:
it's
configurable
there's
you
can
do
parallel
or
you
can
do
completion
and
parallel
is
like
non-blocking
and
you
just
kind
of
like
get
to
like
c105.
You
just
get
to
five
and
then
completion.
It's
like
okay
get
to
five
and
then
we're
done.
E
So
with
the
stateful
set
they're
trying
to
create
deterministic
ordering
like
how
things
are
started,
I
don't
know
if
we
need
that,
with
the
virtual
machine
pool,
I'm
kind
of
hesitant
to
start
with
that
we
could.
We
could
add
that
on
later
on,
if
we
needed
it,
but
I'd
have
to
see
a
really
strong
use
case
for
a
while,
that's
necessary.
No,
what's
up
somebody
who's
that
I'm
here,
let's
can
we
mute
whoever
that
is
so
they
don't
accidentally
say
something.
A
E
Let's
come
up
with
their
own
and
the
precedent
that
we
have
today
within
cubert
is
this
idea
of
like
batch
operations
like
a
batch
count.
For
example,
we
have
that,
for
how
many
migrations
will
and
vm
starts
will
allow
during
workload
updates,
and
things
like
that
of
something
I
had
recently
so
to
be
consistent.
I
would
probably
call
it
like
about
a
batch
size
or
batch
count,
or
something
like
that,
because
threads
might
be
confusing
in
the
context
of
like
actual
cpu
threads
or
something
like
that.
D
D
So
it's
something
that
I
think
makes
sense
to
be
configurable
and
something
that
people
may
want
to
use.
Okay,
so
maybe
then
so
then
let's
say
it.
Let's
just
say
it's
one
like
what
what's
our
behavior
then
so
non-blocking
do
you
think
like
we
just
kind
of
like
like
what's
one
and
what's
the
difference
between
one
and
two,
if
it's
not
blocking.
E
D
I
don't
know
if
I
wanted
that
correctly,
so
would
you
well,
let
me
try
to
read
it
back
like
so
this
would
be
like
we
have
created.
We,
we
made
the
api
request
so
like
we
have
to.
We
make
we
make
the
api
requests,
and
this
would
be
so
it
would
be
based
on
phase
so
so,
which
phase
is
it
that
you
considered
like
when
we
start
the
next
one?
When
we've
reached
running.
E
D
D
Implement
okay,
so
then
that
would
make
sense
for
one.
So
then
we
just
like
have
that
that
would
get
us.
This
kind
of
we'd
follow
this
ordering
with
with
one
and
then
anything
above
one.
It's
now
we're
doing
that
we're
doing
the
captioning.
So
it's
it's
kind
of
like
a
bowl
like
yeah.
No,
it's
actually
more
than
that,
because
we
need
to.
We
need
to
control
how
much
we
want
to
match.
D
Okay,
I
mean
that's
mainly
all
I
wanted
to
kind
of
bring
up.
I
think
we
could
talk
about
the
rest
on
the
mailing
list,
but
that
was
mainly
the
changes
and
how
in
pets
now
the
idea
of
pets
kind
of
changes,
the
the
dynamic
I
mean,
I
think,
like
the
other
one,
was
we
delete
a
virtual
machine
like
say
this
one:
should
it
be
replaced
with
the
same
name
or
not.
E
E
So
you
could
have
some
sort
of
policy
about
if
the
persistent
storage
or
secret
isn't
owned
by
that
virtual
machine,
it
would
stick
around.
It
was
owned.
Maybe
it's
cleaned
up.
If
you
delete
it
and
then
recreate
it,
I
don't
know
that
would
be
some
details
to
work
through.
But
if
we're
going
to
use
prefixes
and
we're
going
to
allow
that
to
vol
matching
for
pre-population
of
secrets
to
match
something's
going
to
be
created
by
this
this
controller,
then
I
mean
I
don't
know
what
else
we
would
do.
E
D
Yeah,
the
only
thing
other
thing
is,
I
could
think
of
just
increment
on
top,
but
yeah
it
doesn't
quite
get
us.
The
matching
doesn't
allow
us
to
match
the
the
pvcs
and
the
secrets:
yeah.
Okay,
all
right
I'll
do
a
little
bit
more.
Another
look
at
this
is
and
kind
of
incorporate
some
of
the
the
idea
of
pets
all
the
way
through
again,
because
it
does
it
also
defects
detach.
That
was
the
other
thing
I
mentioned.
D
A
A
A
A
Because
we
also
need
volunteers
to
to
attend
the
booth
and
and
answer
questions.
So
there's
that
lots
of
talk
about
so
our
team
went
through
a
transition
in
the
last
two
weeks
and
I've
been
very
busy
trying
to
move
our
our
work
blocks
around
and
get
organized
so
that
email
should
be
coming
within
a
the
next
day
or
two.
D
A
Yeah,
it's
there's,
gonna,
be
a
a
conference
platform
and
if,
if
you've
been
to
a
red
hat
summit
before
it's
basically
a
a
dashboard,
a
virtual
summit,
I
mean
it's
a
dashboard
where
you
can
wander
around
and
and
enter
various
video
conferences
and
participate
in
chat
and
the
new
platform
is
going
to
have
one-on-one
chat.
So
a
group
of
people
can
can
have
like
a
zoom
meeting.
A
A
A
A
A
A
B
B
Higher
consumption
of
ram
memory-
and
I
also
saw
that
the
ram
peaks
to
almost
140
gigs-
which
I
guess
is
the
run
capacity
of
each
node-
so
maybe
we
should
reduce
the
load
on
the
nodes
and
we
could
reduce
the
flex
that
we
see
in
the
tests
and
because
I
saw
this
behavior
when
I
was
working
on
the
srv
lane
stabilization
and
when
the
node
was
loaded
and
this
the
memory
peaked
to
the
top
and
there
was
a
huge
load
on
the
cpu
everything
slowed
down.
I
had
these
connections.
B
E
I
think
roman
introduced
something
like
that
fairly
recently,
where
there
was
a
request
to
request
more
memory
for
each
test
lane
or
something
something
like
that,
which
would
in
turn
reduce
the
number
of
parallel
test
lanes
that
could
execute.
So
by
requesting
more
memory,
it
would
reduce
kind
of
some
of
our
pressure
there.
E
F
As
you
say,
as
you
said,
this
increased
memory
would
make
things
better.
I
think
it
landed
at
the
end
of
last
week,
so
we
should
start
seeing
this
improvement
now,
even
and
and
also
the
now
that
we
don't.
We
don't
have
the
the
1.17
lanes
running
now.
The
pressure
is
much
lower,
so
yeah,
hopefully,
during
all
this
week
we
haven't
seen
this
test
failing
and
yeah.
F
Hopefully,
let's
keep
it
for
a
while
more
to
try
to
establish
some
kind
of
procedural
policy
to
to
take
tests
out
of
out
of
quarantine,
but
it
is
not
established,
but
yeah.
We
should
see
the
the
test
passing
consistently
for
for
a
while
and
but
yeah
this.
This
sounds
good
to
me,
but
the
the
explanation
from
from
all
about
what
what
happened
with
the
with
this
test
so
yeah,
let's
keep
it
there
for
for
a
while
and
then
bring
it
back
to
the
stable
suite.
B
F
I
think
that
there's
a
pr
also
for
increasing
it
and
I'm
not
sure
if
the
if
it
is
what
is
the
current
amount,
but
I
think
we
have
a
keyboard
to
beard
in
the
repo
pr
for
increasing
it.
To
I
don't
remember
exactly
how
much
is
it
in
the
in
the
in
the
jobs
we
are
now
requesting,
if
I
recall
correctly,
34
gigs
for
in
each
in
each
job,
for
for
all
the
all
the
ci
cluster.
B
F
B
B
F
This
is
the
the
the
notes
have.
We
are
running
now
in
parallel
kind
of
at
the
peak
like
10
10
jobs
per
node
and
yeah,
the
the
nodes
have
more
than
300
gigs
yeah.
Previously
before
this
increase
it
was,
it
was
more
than
12
jobs
or
something
like
that,
and
now
the
concurrency
in
each
of
the
notes
is
about
10
jobs.
F
F
A
G
B
A
A
B
A
A
A
E
B
C
E
Yeah,
so
does
anyone
have
any
bugs
they'd
like
to
bring
up
for
discussion?
If
so,
we
can
have
that
sort
of
thing
now
I
I
think
for
the
bug
scrub
we
might
want
to
skip,
or
at
least
be
a
little
bit
more
proactive
next
time
with
assigning
somebody
before
the
meeting
starts
to
to
do
it.
Otherwise,
we
kind
of
fumble
around.