►
From YouTube: SIG - Performance and scale 2021-12-02
Description
Meeting Notes: https://docs.google.com/document/d/1d_b2o05FfBG37VwlC2Z1ZArnT9-_AEJoQTe7iKaQZ6I/edit#heading=h.yg3v8z8nkdcg
A
Hey
marcelo,
okay,
we'll
get
started
sixth
scale.
This
is
december.
Second,
well,
I
attached
the
dock
and
I
just
have
an
attendee.
A
Okay,
all
right
marcelo,
thanks
for
joining,
so
you
have
you're
the
first
set
in
here.
B
Yeah,
so
you
know,
I
don't
know
if
I
think
the
time
you
know,
maybe
winter
time
or
summer
time.
You
know,
I
don't
know,
what's
happened
it's
one
hour
later
now
for
me
and
it's
just
like
you
know,
12
a.m,
and
if
you
guys
well,
if
it's
not
too
bad
for
you
guys,
if
we
can,
you
know
get
one
hour
earlier
before
was
like
11
p.m.
For
me
in
japan,
which
better
you
know.
A
Yeah
I
so
I'm
not,
I,
I
don't
have
a
conflict
now
earlier.
I
don't
know,
do
you
have
a
conflict
david.
C
I
do
not
that
would
be
fine
with
me,
yeah,
okay,
there
will
be
some
days
like
once
a
month
that
I
won't
be
able
to
join
the
very
beginning.
So
we
have
this
sprint
like
review
session.
You
might
remember,
and
it
bleeds
over
until
9
30
every
three
weeks
or
four
weeks.
Okay,.
A
Yeah
I
mean
I,
I
think
it's
I
think
it's
okay
marcelo
I
mean
there
might
even
be
like,
like
I
was
telling
david,
like
I
think
you
mentioned
to
me
like
there
was
some
of
you
guys
showed
up
to
the
earlier
time
I
think
today
and
previous
week,
so
there
might
even
be
some
other
people
that
we
could
catch
as
well.
So
it
may
make
sense.
A
I
think
that's
fine,
so
I
can
talk
with
chris
and
I
can
do
the
logistics
for
it.
C
Oh,
by
the
way,
chris
is
no
longer
at
red
hat
or
working
on
cubert
directly.
At
least
he
went
to
work
for
nasa.
Oh,
he
did
wow
yeah.
He
was
all
a
very.
It
was
pretty
quick
and
it
wasn't
like
there
was
no
problems
or
anything.
He
just
found
an
interesting
opportunity.
C
Like
I
don't,
it
might
be
me,
I'm
not
sure,
okay,
what
do
we
need?
We
need
to
move.
The
calendar
invites
yeah
like
okay.
I
can
figure
that
out.
Let
me
just
assign
that
to
myself
in
this.
A
Okay,
all
right
we'll
go
to
the
second
item,
kubert
summit,
so
I
pulled
these
two
items.
The
bullet
point
two
and
three
from
last
week.
B
Yeah
I
saw
this
mail
in
this
f
part
here.
I
reply
the
mail,
but
for
some
reason
I
cannot
find
anymore
my
message
anyway,
so
I
wrote
here,
but
no
one
replied
so
david.
I
don't
know
if
you
know
those
guys
that
are
in
charge
for
the
kubrick
summit
or.
C
I
do
yes,
I
know
josh,
so
just
some
insight
on
the
red
hat
side
for
people
that
are
involved
with
organizing
the
community
josh
burkus
jay
burkus.
It
was
chris's
manager,
so
he's
kind
of
the
go-to
for
organizing
the
stuff.
That's
that's
kind
of
sponsored
by
red
hat.
B
Yeah
so
well,
two
things
one
is:
maybe
I
shouldn't
like
try
to
send
a
message
to
him
might
be
interesting
to
to
help
you
know
in
the
organization
it's
just
for
personally
for
me
and
the
other
thing
is
for
our
our
you
know
our
group.
B
It
might
be
interesting
to
present
something
you
know
some,
so
maybe
some
tasks
that
we
are
doing.
You
know
this
performance
test
that
we
did
and
also
the
trace.
The
trace
think
that
ryan
is
working,
I
think,
should
be
things
that
are
should
be
interesting
to
do.
C
I
think
there's
definitely
a
there's
two
audiences
here
for
q
vert
summit.
I
think
that
it
can
be
really
tailored
a
similar
presentation
focused
on
like
key
for
developers
and
users
on
how
to.
C
The
focus
is
more
of
how
we
tested
in
maybe
the
more
generic
sense,
how
we
tested
a
kubernetes
controller
and
the
things
that
we
looked
for
and
we're
interested
in
and
less
focused
on
the
virtualization
part,
even
though
that's
obviously
going
to
be
there
and
more
focused
on
you
can
apply
this
to
your
development
of
like
your
generic,
whatever
controller,
so
I
think,
there's
certainly
overlap,
but
I
wouldn't
say
it's
the
same
presentation.
I
would
kind
of
focus
it
differently.
C
C
B
Can
present
like
more
narrow,
you
know
down
to
the
in
the
convert
summit.
For
example,
you
know
the
delta
tool,
the
audit
tool
that
you
you
create
and
the
tool
that
we,
the
the
load
generator
tool
we
can
present
like.
You
know,
specifically
the
the
test
that
we
are
integrating
in
the
cicd
system.
B
C
Yeah
and
maybe
even
the
profiling
as
well
distributed
profiling
all
right
yeah,
even
though
I
I've
got
to
even
use
that,
but
something
well.
B
A
Yeah,
we've
done,
I
know,
we've
done
it
somewhat
like
on
at
least
internally
on
some
of
the
stuff
we've
looked
at.
We
have
some
we
have.
I
can
ask
so
tomas
is
the
one
who's
done.
This
he's
the
one
who
did
the
contributed
that
patch
and
he
was
using
it.
So
he's
got
a
ton
of
data
on
this
sophie.
If
that's
an
idea
like
I
can
ask
him
if
you
know,
if
he's
interested
in
talking
about
it,
and
you
know
marcel,
if
you're
interested
too,
I
can
pair
you
guys
up.
A
If
you
guys
want
it
or
ask
him
if
you
guys
want
to
do
something
together
or
whatever
yeah
there's
that's
another
avenue.
C
Yeah,
that's
going
to
be
a
pretty
dense
presentation.
We'd
have
to
really
structure
it
from
a
high.
It
could
possibly
even
be
two
presentations
depending
on
how
much
detail
we
want
to
get.
A
Yeah-
and
I
think,
like
you
know
with
this
one,
what's
interesting
to
me
like
we,
so
I
think
I
want
to
remember
is
like
one
of
the
things
that
we
we've
done
with
the
profiling
is
like.
We
just
see
it's
a
lot
of.
It
is
a
lot
of
the
a
lot
of
the
profiling
targets,
areas
where
we're
we're
marshalling
on
marshalling
json,
I
think,
was
some
of
the
biggest
areas,
but
I
mean
it
would
be.
Nice
is
like
to
take
because
I
think
one
thing
is
like
we
would
be
good
to
have
pictures.
A
It
would
be
even
better
to
like
also
have
like
some
a
conclusion
or
a
direction
that
we're
going
with
it.
That's
just
something
that
they're
not
nice
to
have
as
well,
but
yeah.
I
don't
know
we
need,
I
think
like
there
has.
There
needs
to
be
more
of
a
push
on
some
of
the
profiling
for
sure
yeah.
There's
a
lot
of
data
there.
A
Okay
and
the
density
test
too.
This
is
like
we
have
that
threshold
like
we
have
the
well,
we
don't
have
the
thresholds,
yet
we
almost
we
almost
do.
I
just
I
haven't
had
the
time
to
do
that,
to
finish
this
analysis
of
the
density
test,
to
fully
understand
what
the
results,
but
this
is
another
one
like.
I
think
we
need
some
more
of
a
push
to
like
to
complete
this.
A
A
B
Yeah
yeah,
we're
kind
of
I
think
it's
my
photo.
We
are
kind
of
waiting
for
the
performance
cluster,
I'm
I'm
working
on
that.
It's
it's
been
like
you
know
many
many
things
you
do
and
prs
that
don't
get
merged
and
things
get
delayed.
But
if
everything's
go
fine,
maybe
next
week
we'll
have
the
cluster
so
and
then
we
can
really
have
like
reliable
tasks.
B
A
Yeah,
no,
that's
definitely
a
factor
in
this.
I
think
that's
like
yeah,
so
in
and
that's
that
is
an
important
change,
we're
making
so
yeah
I
mean
that's.
That
makes
sense.
Yeah
I
mean
to
me
these
are
all
really
good
ideas
like
I
think
yeah
we.
I
think
they
need
a
little
bit
more
get
closer
to
like
a
conclusion
or
at
least
kind
of
under
more
along
the
track
of
like
you
know.
A
More
than
just
you
know,
here's
the
work
we've
done
and
let's
see
where
you
know,
here's
closer
to
where
we're
going
but
yeah,
I
think
they
they
all
should
be.
They
all
could
be
talks
and
for
cube
or
something
yeah.
A
Yeah
and
then
also
another
note
on
the
on
the
kubernetes
side.
We
you
know,
I
have
I've,
had
some
thoughts
about
this
one
like
so
we've
like
they.
We
submitted
our
presentation
previously
this
one
and
I
think,
we've
had.
I
mean
at
least
from
our
end.
Like
nvidia's
side,
we've
had
some
monkeys
like
we've
done,
like
we've,
we've
progressed
on
like
a
lot
of
our
scaling
and
performance
and
those
updates
that
we
could
do
there
and
a
lot
of
things
we've
found
too,
not
just
in
kubernetes
and
kubernetes
lots
of
things.
A
A
Well,
so
I'm
not
I'm
not
sure
yet,
like
I'm
like,
so
I
think
what
I'm
leaning
towards
is.
I
ideally
do,
I
guess,
doing
something
remote.
I
think,
like
it's
probably
probably
the
best
idea
yeah.
I
I
don't
know
like
I
don't
know
what's
going
on
with
you
know
everything
out
there
so,
like
that's
sort
of
leaning
towards
that's.
C
My
concern
yeah,
that's
why
that's
one
of
the
reasons
why
I
said
I'm
not
interested
in
kubecon
eu.
It.
B
B
You
know
system,
and
actually
I
was
talking
with
david
and
roman
before-
and
I've
probably
going
to
send
with
my
colleague
in
ibm
so
but
but
definitely
after
this
first
call
for
you
know
proposal
they
have
the
maintainers
proposal
and
I
talked
to
them.
I
sent
a
mail
and
they
will
open
after
it's
closed
the
first
and
call
for
proposal
and
the
maintainers
needs
it's
like
who
who
submits
the
proposal
needs
to
be
a
maintainer,
so
it
can
be
any
of
you
guys
and
and
if
it's
possible
I
can
also
participate.
C
C
The
way
it
was
last
time
we
did
get
a
maintainer
slot
one
time
and
it
was
because
somebody
else
yielded
theirs
to
us
like
another
project
or
something
like
that.
But
once
we
get
incubation
I
think
we're
guaranteed
that.
B
Well,
but
my
main
motivation
to
become
a
maintainer
was
to
be
able
to
submit
this
proposal,
but
actually
I
don't
need
to
be
if
you
guys
already,
are
and
can
submit
that.
But
of
course
one
of
you
guys
needs
to
want
to.
A
B
Ryan,
you
can,
you
can
do
that.
Maybe
we
can
do
that
together.
If
you
want
so
after
the
coffer
proposal,
the
regular
one,
it
will
be,
you
know
open
a
maintainer,
you
know
call
for
registration,
something
like
that.
I
don't.
I
don't
really
know
the
names
the
terms,
but
something
like
that
and
and
then
it's
for
some
it's
for
it's
specifically
for
projects
and
actually
cooperate
presented
that
in
2018
19..
B
C
I
think
we
got
that
in
2019.
We
probably
got
in
2018
as
well,
but
we
we
failed
to
get
it
in
2020
for
kubecon
in
a
yeah.
C
A
C
And
I
would
have
gotten
so
ryan,
we
would
have
gotten
a
slot
in
2020s
if
we
had
a
maintainer
slot,
we
didn't
so
we
failed
the
to
get
the
general
admission
slot
or
whatever
the
big
pool
of
people.
Once
we
get
incubation
we
would
have
actually
been
able
to
be
guaranteed
a
flight.
C
Here's
my
thought,
ryan
and
you
can
go
like
I
support
you
in
whatever
you
choose
to
do
with
eu
or
n
a
or
whatever
that
presentation
we
were
going
to
give
last
year.
If
there's
a
possibility
to
give
it
in
person.
A
Yeah,
I
I
I
hear
you're
saying
I
I'm
thinking
the
same
thing
and
that's
what
that's
why
I'm
also
like
in
some
way,
but
I'm
wearing
a
lot
of
things,
because
I
would
I
mean
I
would
love
to
obviously
get
one
accepted,
but
yeah
I
mean
not
going
in
person.
A
A
B
A
And
you
know
if
it's:
if
we
have
a
presentation,
that's
close
to
getting
into
the
general
session,
I
almost
want
to
wait
till
we
have.
We
can
do
it
in
person.
C
Here's
here's
my
thought:
we're
going
to
get
incubation,
we're
really
close,
we're
probably
going
to
get
it
in
the
next
few
months.
Maybe
it's
closer
than
that.
I'm
not
sure,
there's
a
very
good
chance
that
for
kubecon
in
a
that,
we
could
present
this
and
get
the
maintainer
slot
for
n
a
in
person.
A
C
C
About
getting
stuck
overseas,
I
don't
really
care.
That's
like
domestic
travel,
doesn't
bother
me
it's
detroit
wow,
that's
right!.
C
D
B
A
Yeah
I
mean
I
think,
like
marcelo
like
this,
like
I
think
yeah
I
mean
to
me.
I
I've
been
back
and
forth
a
lot
of
this.
Like
I,
I
was
first
started
like
when
I
talked
to
david
like
I
was.
I
was
like
okay
yeah.
Let's,
let's
go
travel
to
spain,
then
I
was
like
well
now
some
things
have
changed
and
then-
and
you
know
obviously
doing
it-
not
in
person-
is
the
thing
that
I've
been
I've
been
challenged
with.
A
So
the
yeah
I
mean,
like
even
kind
of
you
kind
of
said
david
is
the
you
know
the
the
travel.
That's
not
quite
the
same
as
as
it
was
yeah.
It's.
C
A
Do
you
think
like
if
you
know,
let's
say
this
time
around
so
you
which
is-
I
don't
know,
what's
the
date,
it's
it's
it's
the
winter.
Whatever
it's
a
few
months,
that's
the
eu
conference
and
in
the
na
conference
I
mean
you
know
we
could
I
I
mean
I
don't
know
what
he
like.
I'm
okay
like
if
you
want
to
present
like
with
me
and
david
on
this,
like
you've,
done
a
lot
of
work
in
sixth
scale
like
if
you
want
to
talk
about
you
know
with
us
on
this
presentation
for
na.
A
C
I'd
say:
go
for
it.
Marcelo
go
for
independently
of
what
any
of
us
do
submit
your
talk.
If
you
get
it,
you
get
it.
A
A
C
Go
for
it.
There
are
so
many
angles
to
this
topic
that
if
you
present
in
kubecon
eu,
we
could
still
present
kubecon
in
a
and
have
just
a
different
take
like
there's
just
so
much.
Then
you
only
get
like
a
maybe
a
30
minute
talk.
They
say
like
it's
a
40
minute
slot
or
four
or
five
minutes
long,
but
realistically
it's
about
30
minutes
of
talking,
there's,
probably
hours
of
concept
here.
B
B
Yep
yeah
yeah,
so
I'm
going
to
submit
like
anyway
a
presentation
for
that
and
then
we'll
be
up
to
you
guys
if
you
you
guys
want
to
when
it's
open
the
maintainers
call.
If,
if
you
are
up
just
submit
something
or
not,
we
can
talk
that
by
the
time
probably
will
be
january.
Only
okay
and
then
see
we
see
yep.
C
We'll
we'll
do
what
we
can
to
support
you
there
I'll
have
to
look
at
the
details
if
what
the
requirements
are
as
far
as
like
from
the
maintainer.
If
we
can
submit
on
your
behalf,
that
would
be
interesting
or
if
we
actually
need
to
participate.
I'm
not
sure.
C
Talk
about,
I
don't
know
what
the
criteria
for
a
maintainer
is
it's
possible
that
you
could
just
become
a
maintainer
by
that
time.
What's
maybe
we
can
review
that
real,
quick.
C
C
C
Maybe
reach
out
to
josh
about
that
and
gain
an
understanding
of
what
it
is
to.
I
think
that
we
have
to
sponsor
a
candidate
and.
B
C
But
we
don't
have
a
clear
understanding
of
what
that
role
like
the
responsibilities
of
that
role,
yet
I
don't
think
which
makes
it
a
little
bit
of
a
challenge.
I
think
that's
what
we
kind
of
seeded,
an
initial
group,
so
ryan
myself,
and
maybe
somebody
else
from
red
hat,
maybe
one
or
two
other
people
from
red
hat.
Maybe
somebody
from
apple
too.
I
can't
remember
certain
for
certain,
and
that
was
just
the
bootstrap,
the
process
that
we
haven't
created.
Yet
I
believe.
A
Okay,
all
right
marcelo
are
you
gonna,
so
you
can
talk
to
john
and
you
get
to
feedback
yeah.
We
can.
Let
us
know
what
like
decide
with
the
presentation.
A
A
Yeah
marcel
you
out
of
this
too.
B
B
B
Actually
it
has
all
the
emails
locally
and
use
bazel
to
you
know
to
deploy
things
in
the
clusters,
so
I'm
trying
now
to
do
to
use
the
same
scripts
and
the
same
technology
same
tools
that
that
exists
for
the
other
clusters.
So
it's
taking
more
time
that
I
was
expecting
to
to
learn
all
the
tools,
but
I
I
already
did
that
now
so
tomorrow
I
will
submit
the
final
pr
for
that
and
then
we
would
have
like
permitted
and
the
phone
in
the
cluster
and
then
we
can
actually
start
to
run
the
jobs
there.
B
A
B
Right
and
so.
A
B
B
B
Then
the
the
interesting
part
is
with
800
when
I
created
100
vmis
with
the
last
bump
well
less
curve
in
the
graph.
D
B
B
B
Even
worse
with
800,
you
see,
so
it's
definitely
some
scalability
things.
You
know
fine-tuning
this
api.
You
know
pairs
per
second,
which
was
a
little
bit
surprising
and
also
interesting
for
me.
So
and
I
I
didn't
see
if
it's
improved
with
something
else
we
can
check
now
I
took
his
snapshot
from
all
the
metrics
we
can.
B
A
B
A
B
Yeah,
I
need
I
need
yeah.
I
need
to
separate.
Otherwise
it
doesn't
take.
This
name
shot
yeah.
So
you,
if
you
go
down
a
little
bit
little
bit
more
yeah.
So
you
still
we
see,
even
though,
if
200,
we
still
see
some
rate
limit,
although
it's
very
low
six
millisecond
here,
but
it
happens,
many
many
requests
and
it
might
be-
we
can
still
see
like,
even
though
with
very
high.
You
know
pairs
per
second
200.
B
We
were
still
are
seeing
some
rate
limit
and
I'm
not
breaking
too
much
vm,
so
only
800.
So
I
was
wondering
for
a
really
scale
test.
With
hundreds
of
vms,
I
mean
hundred
of
nodes.
What
can
be
that
so
we'll
we'll
really
see
like
we
need
to
check.
You
know
this.
This
kind
of
kind
of
metric,
yeah.
A
Yeah
I
wanna
try
this
with
our
larger
clusters
and
see
what
this
comes
out
to.
Okay,
that's
interesting,
especially
yeah
I
mean
with
here
I
mean
the
I
mean
this
is
noticeable
phase.
Let
me
compare
the
latencies
again,
let's
see,
let's
go
look
at
these.
A
A
A
Failed,
so
what
what's?
This
is
the
what's
so
this
is
like
when
it
was.
B
C
A
C
B
B
A
A
Different
node
counts
to
see
800
yeah,
because
this
also
some
of
this
too
I
mean,
could
have
to
do
with
kubernetes
like
I
would
have,
I
mean
we're
doing
like
you're
looking
at
like
scheduled
transition
to
scheduled
state.
I
mean
this
is
mostly
going
to
be
kubernetes.
A
B
Okay,
so
another
question,
so
you
guys
you
know
in
the
beginning
of
the
this
six
killing
meeting.
I
don't
remember
if
he
was
your
colleague
ryan.
That
was
comparing
the
time
that
the
pod,
you
know
become
ready
to
the
vm,
become
ready,
isn't
it
so
do
you
know
when
a
way
is
an
easy
way,
maybe
that
we
could
create
a
graphona
dashboard?
With
that
I
couldn't
find
you
know
a
easy
way
to
correlate.
B
B
A
This
was
way
so
I
talked.
I
talked
about
an
add-on
to
this.
This
was
it's
somewhere
in
our
notes
here,
where
I
have
mentioned
different
ways
that
we
can
get
data
from
the
or
more
detailed
views
between
in
between
transition
times.
It
was
like,
because
you
can
there's
some
there's
a
bunch
of
data
on
the
the
pod
and
the
vmi
objects
based
on.
You
know,
different
updates
that
the
cubelet
or
qvert
has
done
to
those
objects,
and
you
can
get
timestamps
from
them,
but
there
isn't.
A
Well,
it's
in
it's
in
it's,
it's
not
the
logs.
It's
actually
it's
on
the
object.
It's
it's
you
can.
You
can
scrape
it
from
the
object.
It's
already
there,
there's
just
there's.
Just
no
like
I
said,
there's
no,
there's
no
metric
for
it,
at
least
that
I'm
aware
of.
A
A
Yeah
so
there's
a
bunch
of
there's
a
bunch
of
stuff
like
marcelo,
if
you,
if
you
do,
if
you
look
at
the
pod,
you'll
see
that
there
are
there's
a
bunch
of
timestamps
that
are
posted
by
the
different
controllers
for
the
work
that
they're
doing
the
changes
and
how
they're
modifying
the
object.
A
So
if
you
like,
if
so,
if
you're
after,
like,
for
instance
when
the
when
the
pod
or
when
the
containers
go
ready,
so
like
the
moments
that
the
so
right
before
the
moment
that
the
vmi
goes
to,
I
think
it's
scheduled
state.
That's
when
all
the
that's
when
at
least
every
pod
except
the
computer
doesn't
have
to
be
ready.
You
can
measure
the
time
that
that
happens.
It's
like
in
kubernetes
and
then
you
can
compare
it
to
the
time
that
we
already
have
this.
A
B
A
So
that's
that's
an
idea,
but
there's
a
bunch.
There
is
my
point
like
there's
a
bunch
that
we
could
look
at,
but
it's
not
necessarily
the
cuber,
but
it's
things
we
could
look
at.
I
mean
I
was
thinking
like
we
could
do
this
in
the
audit
tool.
A
few
ideas
about
that
I'll
see
if
I
can
find
it.
B
Yeah
we
can,
we
can
find
that
in
the
trace,
but
I
think
it
might
be
in
the
future
a
good
idea
to
have
some
of
these
informations
in
the
grafana
dashboard,
also,
which
would
be
much
easier
to
to
check
and
verify,
maybe
maybe
new
metrics.
We
can
ex
find
some
maybe
some
way
to
expose
those
metrics,
but
it
doesn't
need
to
be
if
it's
already
there.
A
Yeah,
the
other
alternative
I
was
thinking
is
like
the
auto
tool.
Right
now
reads
the
metrics
and
we
could
consider.
I
mean
we.
This
would
be
like
a
slight
divergence
from
the
what
it
currently
does
but
like
it
could
be
that
we
we
have.
It
also
read
objects,
so
we
just
you
know,
but
that's
that
would
mean
that
we
sort
of
change
the
the
principle
there
to
also
not
just
do
you
know
metrics,
so
I
mean
there's
that's
a
possibility.
A
I
guess
like
we
could
you
know
if
we
want
that
additional
data,
but
that
does
that
mean
this
doesn't
end
up
in
here
it
just,
but
it
would
give
us
an
avenue
to
test
it.
A
C
Yeah
this
one's
really
close
I
just
had
to.
I
don't
completely
understand
the
step,
because
it
seems
like
it's
not
doing
what
I
would
it's
doing,
I'm
not
sure
what
it
gives
us
the
way
the
step
is
structured.
Let
me
look
at
the
pr
yeah.
I've
lost
account
like.
C
I
reviewed
this
like
a
week
or
two
ago
and
then
I
forgot
to
hit
post,
so
it
said
and
my
comment
got
lost
forever
until
I
reposted
it
and
now
it's
like
in
some
comment.
That's
like
in
the
middle
of
all
this.
C
A
Yeah
there's
because
so
for
the
vm,
the
vm
is
different
from
the
vmi
tracing
because
the
vms
doesn't
have.
It
doesn't
have
one
of
the
either
update
settings
or
sync,
I
don't
remember
which
one
it
doesn't
have.
I
see.
C
B
C
If
we
put
the
defer
at
the
beginning
of
okay,
sync,
we
do
a
defer,
so
we
know
from
the
gang
of
the
trace
the
end
of
sync:
okay
got
that
and
then
we
know
from
the
end
of
sync
to
the
beginning,
at
the
end
of
update
staff.
Okay,
that
makes
sense.
That
does
make
sense
to
me.
C
Status,
that's
kind
of
I
mean
realistically,
it
doesn't
give
us
anything
just
because
that's
the
whole
length
of
the
execution
update
status
is
the
last
thing
that
executes.
So
essentially,
you
have
the
step
trace.
It's.
The
very
last
thing
that
executes
before
we
put
stop
trace
like
that
stop
trace
is
actually
the
next
thing.
That's
called
right
after
step
trace,
logically
yeah.
A
Yeah,
I
I
agree
so
I
my
thought
on
this-
is
that
so
what
is
so
like?
What's
this
getting
us
like,
so
the
I
would
say
so,
we'll
look
at
it
without
it.
So
if
it's
not
there,
we
don't
have
sort
of
a
label
saying
what
we
expect
to
be
like
what
we're
tracing.
So
that
was
kind
of
my
my
thought
on
this.
Is
that,
like
I'm,
using
steps
as
like,
I
say:
okay,
here's
this!
You
know
period
of
time
where
we're
labeling
something
so
it's.
This
means
the
time
between
start
and
update
status.
A
Now,
the
time
after
that,
like
I'm,
not
even
thinking
about,
but
it
is,
I
agree
it's
useless
like
because
it's
like
it's
sort
of
the
time
after
like
why
and
that's
that's
not
necessary.
I
don't
think
that's
your
point,
but
the
the
time
is
useless,
but
the
your
point
is
that
right
that
we're
basically
putting
a
step
in
here
and
it's
and
there's
no
point
because
it's
it's
99.9
percent
of
the
time
is
is-
is
like
update
status
like
it's
just
like.
Why
would
we
want
to
step
and
like
I
like?
It's
it's.
A
The
same
point
is
that
we
can
label
it
and
also
we
can
add
there
could
be
future
code,
and
you
know
that
comes
along
later,
that
we
want
to
trace,
and
so
we
can.
You
know
I'm
figuring
that
like
like
we
could,
you
know
we
could
still
have
it
and
then
you
know
maybe
someone
adds
could
like
that.
We
want
to
trace
and
we
could
just
add
that
there
too,
and
then
it
becomes
you
know
then
there's
more
things
to
trace,
and
then
we,
maybe
we
don't
have
this
problem
anymore.
C
Okay,
that
makes
sense
there's
a
place
we
can
put
the
equivalent
of
a
sink
for
vm.
If.
A
So
this
is
so
in
this
in
this,
if
block
you're,
saying
or.
C
C
End
of
this,
if
block
right,
where
that,
right
after
line
311
before
between
3
12
and
3
11,
okay,
that
would
that
would
give
us
a
trace
of
essentially
everything
that
led
up
everything
that
was
actionable.
C
So
you
could
call
it
a
sync
and
in
fact
we
could
abstract
out
line
287
or
288
to
311
and
call
it
sync
function.
In
fact,
that's
probably
what
we
should
do.
I
I
think
that
that
just
needs
a
refactor.
A
C
A
We'll
get
it
that's
good,
though
so
this
once
this
is
in
though
marcelo,
if
you
or
if
anyone
has
any
other
ideas
for
other
traces,
just
add
it
like
which
we
should.
This
should
give
us
the
ability
to
just
add
this
to
all
our
workings.
I
was
going
to
do
this
eventually,
but
you
know
if
there,
if
you
have
other
ideas
we
can,
none
of
this
will
unlock
a
lot
of
things
which
would
be
cool.
A
Okay.
The
I
haven't
had
a
chance
to
review
this
again.
Do
you
want
to
talk
at
all
david
about
your
virtual
machine
pool
changes
or
anything
that
you've
made
that
significant
things
that
you
want
to
bring.
C
Up
the
best
sure
the
only
changes
I've
made
since
I
posted
the
pr
were
your
changes.
So
you
had
some
comments
about
taking
away
the
attach
or
the
adoption
logic
which
I
stripped
out
so
now
we
only
have
the
ability
to
release
a
virtual
machine
from
a
pool
and
not
the
ability
to
reattach
it.
I
think
that
makes
a
lot
of
sense.
For
now
we
can
talk
about
attachment
in
the
future.
If
there's
a
strong
use
case,
I
don't
have
a
strong
use
case
and
it
complicates
it
complicates
the
scenario
quite
a
bit.
C
The
other
change
was
there
were
some
cosmetic
changes,
just
people
asking
for
some
better
co,
quality
and
roman
requested
that
I
add
permissions
and
some
functional
testing
for
the
default
permissions.
So
there's
a
view
edit.
An
admin
default,
our
back
role
that
we
automatically
apply
to
people,
so
they
have
access,
depending
on
their
role
within
kubernetes,
to
create
virtual
machines
or
view
them
or
whatever.
I
just
need
to
add
that
for
virtual
machine
pools,
that
was
this
should
be
good.
I
think
that's
roman
yeah
roma
gave
it
to
the
proof.
C
I
think
he
was
waiting
on.
You
brian
just
to
make
sure
that
you
felt
comfortable
with
the
changes
so
far.
Okay,.
A
So
yeah
it's
on
sunday,
yeah,
yeah,
yeah
I'll
leave
you
I'll
see
if
I
can
get
to
this
week.
Okay,
it's
thursday,
sorry
thursday!
Yeah!
Let's
see
if
I
can
do
it
this
week,
so
I
it's
probably
gonna
be
tomorrow.
If
I
do
get
to
it
this
week,
but
yeah
I'll.
I
have
this
and
I
saw
a
roman
improvement
so,
and
I
saw
that
I
saw
you
go
make
those
changes
so
I'll
go
through
it
again
when
I
can
hopefully
so
we'll
aim
for
the
next
meeting
that
we
have
this.
C
A
A
Okay,
all
right!
That's
alright!
Four
minutes
left,
but
that's
that's
all.
I
have
for
agenda.
Okay,
all
right
guys
thanks
a
lot
all
right
and
yep
talk
to
y'all
later
have
a
good
day.