►
From YouTube: SIG - Performance and scale 2022-11-03
Description
Meeting Notes:
https://docs.google.com/document/d/1d_b2o05FfBG37VwlC2Z1ZArnT9-_AEJoQTe7iKaQZ6I/edit#heading=h.tybh
B
All
right,
let's
take
scale,
it's
November
3rd
2022.,
and
let
me
just.
B
There
we
go
okay,
Andrew,
selfless,
attendee
everyone
and
then
let's
go
right
to
the
agenda.
So
let's
review
the
performance,
periodic
job
results
so
so
Brian
you
have
Okay
so
start
with
the
so
which
one
was
it.
The
the
periodic
was
the
one
that
said
it
was
failing
right,
I
hear
that's:
okay,.
C
Yeah,
periodic
and
and
the
occasional
policemen
as
well
so
the
same
kind
of
122,
Sig
performance
job
runs
against
the
pre-submits
as
well.
Of
course,
I'll
try
and
take
out
the
link
for
you.
So
the
periodic
is
this
one
here.
B
Okay,
here
we
go
so
okay
this,
so
we
have
the
this
one's
been
failing.
The
president
one
has
been
failing
since
28.,
okay,
so
all
right,
so
we
had
a
few
intermittent
values
and
now
he's
constantly.
C
B
Okay,
so
we
need
to
move
on
to
this
yeah
yeah,
okay,
yeah.
We
should
I
think
the
next.
We
should
just
move
to
125
I
mean
I,
think
that's
we
just
haven't.
We
just
haven't
taken
the
time
to
update
to
the
to
go
ahead
and
make
the
change
in
the
in
the
lane,
but
I
mean
we're
not
really
tied
to
122
or
anything
I
think
it
makes
sense.
Okay,
cool.
B
B
B
B
C
Now
maybe
I
do
have
this
DNS
results
cash,
but
it's
returning
for
me.
Okay,.
B
I
think
that's
just
a
VPN.
Okay,
all
right!
Let's
see
here
all
right!
This
is
less
than
those
two
failures.
A
B
We
run
with
this
machine
We,
Should,
Skip,
news,
I,
don't
know
what
that
is
the
okay.
This
is
so.
What
we
did
do
is
we
add.
We
did
add
this
change
recently.
I
I,
don't
know
if
this
is
the
reason
why,
but
we
did
I'm
wondering
if
it
could
be
related
to
this,
because
now
what
we're
doing
is
we're
actually
launching
instead
of
100
VMS
as
part
of
the
density
test,
we're
actually
doing
300
with
so
we're
just
running
the
same
one
three
times
with
different
combinations.
B
So
that
could
be
why
I
just
I
don't
see
we
were
seeing
it
work,
though
that
was
the
thing
it
was
working
and
so
like
he
had
a
few
like.
We
had
a
few
passes
with
this,
and
so
that
that's
why
I'm
not
really
certain,
but
when
I
ran
it
after
this
merged
and
didn't
didn't
work.
B
B
C
Right
I
was
looking
at
the
changes
interesting
enough
on
the
27th.
There
was
a
change
to
bump
the
memory
in
some
of
the
components
I.
A
B
B
Is
it
eight
six
one
three,
this
one
looks
so
that
sounds
similar.
This
is
like
okay,
so
it's
on
a
control
plane.
Is
that
what
it
is.
B
Yeah
so
I
guess
well.
So
if
this
is
on
the
control,
Point
sign
I
guess
it
could
be
that
we
are
allocating
more
space
to
the
control
plane
and
we're
taking
a
little
bit
away
from
our
compute
side
and
then
we're
kind
of
hitting
our
head
on
the
ceiling
here.
We're
not
getting
enough
space
that
could
be.
B
Okay,
the
thing
that's
weird
about
this
log
is
I,
really
don't
see
a
lot,
you
know,
I,
don't
there's
like
there's
actually
not
much
here
like
there.
It
just
seems
like
it
like.
We
have
I,
don't
know
what
this
is
like.
This
is
I,
don't
know
what
this
is.
This
is
weird
to
me:
I,
don't
know
what
we
don't
run
this
like
this
test.
We
should
be
skipping
and
I.
This
doesn't
look
like
a
skip
to
me.
I,
don't
even
know
what
that
is,
and.
B
A
B
B
All
right
we'll
have
to
look
at
this.
Okay,
I
I
know
what
I'm
thinking
here,
but
I'm
I'm,
not
looking
at
property.
Okay,
I
thought
this
was
the
density
test.
This
is
not
the
density
test.
This
is,
for
some
reason,
I
mix
this
up
with
the
density
test.
The
density
test
is
the
name
of
the
one
that
is
that
I
said
earlier.
We
create
100.3.
C
B
B
I,
don't
remember
what
I'm
looking
for
is
I,
don't
remember
the
name
of
it.
I
think
it's
I
thought
it
was
density,
there's
like
a
presume
intensity
or
something.
C
B
I
thought
it
was
there
for
a
little
bit.
Let
me
let
me
just
check
the
dam,
so
oh
actually,
this
isn't
where
the
names
are
right.
This
is
It's
the
community
repo
or
is
it
the
CI
refill
or
something
project
info
I
think
is
foreign
okay,
so.
B
Okay,
so
this
is
the
one,
but
this
is
121,
then
I,
don't
so
I,
don't
remember
changing
this
to
122.
I,
don't
know.
Maybe
this
got
changed
and
let's
take
a
look.
I'll
just
do
a
quick
comparison
here.
B
B
B
B
B
B
Okay,
I'm
not
sure
I
I,
don't
know
what
I
think
some
things
got
added
into
this,
because
this
was
originally
for
using
as
part
of
the
hundred
or
the
103
times
the
Mi
test,
and
now
we're
a
few
things
in
here.
So
let
me
I'm
gonna
have
to
revisit
this
because
I
think
this
part
we
need
I,
think
it
should
be
in
its
own
job,
and
then
we
have
the
density
in
its
own
plane
in
its
own
job,
because
it
doesn't.
This
doesn't
quite
make
sense
to
me.
B
B
And
then
we
need
to,
you
know,
fix
the
performance,
I
think
I,
don't
know.
I'm
gonna
have
to
evaluate
this
I
think.
Maybe
we
could
decide
if
this
should
be
separate
or
not.
I
don't
know
I
I.
Unfortunately,
I
don't
have
the
without
looking
at
all
of
the
code
for
these
tests,
I,
don't
I
can't
really
I,
don't
know
if
we
can
make
any
accurate
judgment
as
to
which
direction
we
should
go
on
this,
but
it
just
seems
like
to
me
at
least
when
we
originally
did
this.
It
was
meant
to
be
density.
B
B
B
All
right,
let's
go
back
to
the
density
tests,
because
this
one
is.
Why
is
this
one
failing.
B
Like
that,
that's
that
seems
to
be
so
here's.
So
what
was
the
when
did
this?
Let's
see
this
started
scaling
on
the
let's
see
today,
10
28.
B
Now
guys,
so
we
can
go
and
then
I
don't
think
this
affects
this.
This
density
tests,
these
ones
use
a
different
Lane.
They
use
a
totally
different.
They
they
don't
use
the
test
path,
they
have
their
own
script,
they
run
okay
I.
If
this
one
just
seems
like
we're,
yeah
I,
don't
know
this.
One
just
seems
like
we're
failing
to
install
Qbert,
so
I
don't
know
why
this
is.
This
is
weird:
okay,.
B
Maybe
we
need
to
do
is
we
need
to
move
some
of
these
Lanes,
all
of
them
probably
to
the
newer
provider
and
then
maybe
try
and
fix
this
from
the
newer
provider?
What
do
you
think
about
that?
Brian?
Maybe
that's
probably
it's
probably
a
good
time
to
do
it
since
I've
already
seen
some
failures
on
122
and
we're
going.
B
C
Yeah
I,
wouldn't
I,
wouldn't
change
things.
I,
try
and
I,
try
and
see.
I
might
I
might
try
and
have
a
look
and
see
if
I
can
see
anything
in
the
artifacts
here.
The
reason
why
it
mightn't
be
starting:
it's
pretty
it's
a
pretty
recent
one:
okay,
there's
not
much
there
and.
B
B
B
He
I
mean
that's
the
that
lines
on
success
to
me.
It
doesn't
seem
like
it's
okay,
so
we
get
okay,
we
can
okay,
we
can
so
we're
getting
cubert.
It's
just
timing
out
waiting
for
okay,
so
yeah.
It's
the
key,
we're
not
ready
in
time.
Let's
see
okay,
so
the
control
plane
is
just
not
quite
going
up.
I
guess.
B
B
B
C
B
C
One
actually
uses
the
external
provider,
so
I
think
this
is
actually
using
the
performance
cluster
itself
as
an
external
provider.
So
I
think
it
is
connects
using
the
the
performance
clusters
cubeconfig
and
then
tries
to
install
keywords
directly
onto
that
cluster
and
then
use
that
cluster
rather
than
virtual
cluster.
B
B
A
B
B
All
right,
that'd
be
really
helpful.
Okay,
thank
you,
Brian,
okay,
that
I
think
it's
probably
the
same
for
both
of
those
I
I.
Wouldn't
imagine
the
atmosphere
and
it's
just
the
same
thing
on
the
same
cluster,
okay,
okay,
so
I
would
say
like
so
for
all
right
on
this
end,
all
right-
we're
just
gonna-
have
to
do
some
digging,
that's
to
make
a
decision
so
on
this
one,
let
me
just
put
together
a
better
picture
of
what
what
we
have,
because
this
job
has
changed
a
little
bit
over
time.
B
So
I
think,
let
me
just
put
together
some
ideas
as
to
like
where
we
should
go
with
this
and
if
it
should
be
maybe
two
separate
jobs,
one
job
or
and
then
also
the
prospects
of
kind
of
moving
to
one
of
the
newer
providers.
At
the
very
least,
I
think
we
should
be
on
124,
but
I
think
it
would
make
more
sense
to
someone
2025.
B
I
think,
maybe,
let's
let
me
just
put
together
a
kind
of
a
bunch
of
those
items,
and
we
can
let's
discuss
this
and
in
our
next
meeting
when,
when
I
have
all
that
together.
B
B
All
right
all
right
we'll
do
that
as
an
extra
and
I'll
take
that
one
and
since
unfortunately
we
don't
have
any
other
results,
we
did
I
think
we.
There
was
a
little
bit
of.
Let
me
go
back
to
the
periodics
I
think
so
we
had.
B
This
was
the
pass
yeah.
Okay,
so
I
wanted
to
see.
B
B
B
This
one,
oh,
he
linked
it.
Okay,
good
this
one.
B
B
B
So
he
what
he
did
is
he
added
some
profiles,
the
the
the
profiles
and
we're
using
VMS.
B
So
we
have
a
bunch
of
more
metrics
in
here,
which
you
can
tell
us
a
ton
more.
We
have
so
we
still
get
our
view
modulations
and
then
we
should
see
a
bunch
of
virtual
machines
changes
and
see.
We
have
some
VM
lists
what
else
custom
resource
counts
here
we
go,
get
virtual
machine
preferences
count.
So
here's
something
that's
new
virtual
machine
instance
types
count.
This
is
new.
B
And
then
yeah,
so,
okay,
so
we
get
looks
like
two
new
ones:
controller
revisions
that
also
looks
new
okay.
So
this
is
so
everything
again
is
the
same
with
this
now
we're
gonna
get
we're
gonna
get
us
some
new
ones
like
this.
So
what
we'll
do
is
we'll
when
we
eventually
get
this
fixed,
we'll
have
to
do
some
analysis
and
see
if,
let's
see
how
these
hold
up
over
time
looks
like
roughly
about
where's
our
Curry
ponds
right
roughly
about
three
to
one
for
for
each
VM.
B
For
these
API
calls,
it
looks
pretty
consistent
too.
So
that's
that's
good!
Okay!
Well,
that's
eventually
what
it
looked
like
when
we
get
it
working,
but
but
that
was
good
to
see
that
we
we're
going
to
have
some
more
data
in
our
jobs.
Now.
B
Okay,
all
right
I
think
we
have
some
good
action
items
from
this
meeting.
I
think
I
think
that's
all
I
don't
know.
Do
you
guys
have
any
other
items
you
want
to
bring
up
or
you
must
blend
early.