►
From YouTube: SIG - Performance and scale 2023-02-02
Description
Meeting Notes:
https://docs.google.com/document/d/1d_b2o05FfBG37VwlC2Z1ZArnT9-_AEJoQTe7iKaQZ6I/edit#heading=h.tybh
A
Okay,
sixth
scale:
this
is
February,
2nd,
2023.,
so
Brian
I
figured
we
could
talk.
Maybe
we
can
do
this
first.
Do
you
want
to
go
over
the
the
the
performance
cluster
changes
as
I
saw?
There
was
some
news
on
that.
B
That
DNS
update
has
been
done.
There's
a
couple
of
things
that
need
to
be
added
to
our
secrets,
repo,
so
they
will
hopefully
get
out
of
today
and
then
we
should
be
able
to
run
the
workloads
against
the
cluster.
There's
a
PR
in
Cuba
cuber
that
has
to
go
in
as
well
and
I
try
and
find
the
link
to
it.
For
you
now.
It's.
B
Service
just
changed,
so
it's
just
pointing
to
the
new
Prometheus
service
service,
so,
okay,
so
that
I
can
escape
the
metrics.
B
A
Yeah
I
saw
that
one
okay
yeah,
if
you
can
look
in
here,
that'd,
be
great
okay.
That
sounds
really
good,
so
like
what
were
some
of
the
changes
that
you
had
to
make
is
this
like?
A
Is
it
like,
so
we
upgraded
to
like
new
Centos,
I,
think
and
then
and
then
we
got
to
get
to
the
new
kubernetes
version
right?
Is
it?
Do
you
know
what
version?
What
version
is
it
on
now.
B
So
it's
on
1.24.,
it's
it's!
It's
a
rebuilt
cluster.
We
used,
we
use
openshift
to
build
the
cluster,
but
it
should
just
work
the
same.
Instead
of
that,
it's
based
off
of
I'll
tell
you
the
exact
version
now
in
a
second.
A
B
B
A
B
I
was
really
hoping
to
have
this
for
you
today,
but
unfortunately
you
just
got
the
DNS
update
there
this
morning,
so
yeah
a
few
more
bits
and
pieces
to
get
in
and
then
hopefully
we'll
have
it
for
next
week.
For
you
cool.
A
Yeah,
exciting:
okay,
that's
awesome!
Okay,
thanks!
Brian
cool
cool,
okay,
that's
awesome
here!
So
the
other
thing
that
I
wanted
to
do
so
we
we
had
this
discussion.
We
I
mean
me
and
lubo
at
the
last
scale,
on
yeah
before.
C
We
contact
switch
sure
I
had
a
quick
question
for
Brian
on
the
cluster
rebuild,
so
Brian
I
know
like
when,
in
the
past,
when
working
with
six
six
scale
and
some
of
the
monitoring
matrixes
I
was
using
a
dashboard
which
was
shared
by
marshmallow,
it's
with
IP
52117
69106.
B
C
I
am
not
actually
sure
I
can
share
with
you.
The
link
I
had
bookmarked
and
I
was
using.
B
Yeah
I
I
see
it
I
see
it
in
the
documentation
for
the
cluster,
of
course,
that
close
that
that
cluster
is
gone
now
completely
so
Deco
fan
instance
is
definitely
gone,
so
I
just
hope.
There's
I'm,
not
sure.
If
I'll
check
I'll
check
the
git
repository
deposit
game
for
repository
to
see
if
those
dashboards
are
there
I'm.
A
B
C
A
Okay,
all
right
thanks
a
lot
okay,
so
the
so
moving
on
to
the
the
performance
job.
So
one
of
the
things
that
we
discussed
with
ubo
last
week
was
the
basically
how
to
track
these
things.
A
How
do
because
kind
of
what
we've
been
doing
is
we've
been
kind
of
weekly
looking
at
these
things
and
and
it
works
for
sort
of
a
way
to
Monitor
and
if
there's
like
any
major
changes
and
so
on,
but
the
thing
is,
though,
when
from
our
discussion
is
that
when
we
look
at
changes
between
jobs
between
days,
there's
a
little
bit
of
variation
and
that's
okay,
but
what
is
challenging
to
see
is
when
we
look
at
a
pattern
over
time.
A
We
want
to
see
like
okay
is
this
job?
Is
this
you
know?
Is
this
PR
really
making
enough
of
a
change
from
what
we
expect
from
the
line
that
we
sort
of
the
normal
mean
that
we
expect
and
and
so
like?
A
A
We
can
basically
scrape
the
the
metrics
that
we're
doing
and
we
can
publish
them
in
into
graphs
and
then
and
then
look
at
this
data
and
that's
be
a
better
way
to
compare
retirement
instead
of
kind
of
just
looking
at
the
the
weekly
performance
job
results,
and
that
would
be
a
better
way
to
to
visualize
and
catch
changes,
at
least
until,
like
we
have
thresholds
and
other
stuff
in
there.
That's
good,
but
this
is
just
be
a
way
we
can
kind
of
get
patterns
across
a
release.
For
example,.
B
So
I
think
Daniel
has
done
something
similar
like
that
for
the
flake
finder,
so
he's
had
to
push
metrics
to
the
Prometheus
push
Gateway
in
our
prow
control,
plane
cluster
and
then
those
metrics
get
delivered
to
refinance.
So
you
have
to
create
a
dashboard,
then
using
those
metrics
that
could
be
something
we
could
do
is
if
we
could
find
a
way
of
well.
I
could
talk
to
Danny
and
see
how
he's
doing
that
with
a
fake
finder
and
how
he's
delivering
those
metrics.
A
Yeah,
the
the
thing
I
saw
was
this
repo.
This
Qbert
CI
Health
was
a
good
example
like
these
Char,
for
example,
were
like
really
useful
to
like
see
sorry,
my
daughter's
crawling
around
on
my
feet.
The
these
charts
are
really
useful
to
see
like
here.
A
We
can
see
jobs
in
there,
how
much
time
they're
taking
and
how
they're
doing,
and
basically
we
just
want
to
have
the
same
same
idea
so
kind
of
what
you're,
saying
and
I
think
so
lubo
said,
he'd
look
at
this
when
he
attacked
from
vacation,
but
I
just
want
to
let
people
know.
Let
you
guys
know
like
this
is
this
would
make
this
time
maker
lasts
a
lot
easier
as
a
direction
to
go
longer
term
and.
C
A
C
A
And
you
know
that's
the
way
we
can
kind
of
set
expectations
like
per
release
or
what
we're
currently
seeing.
You
know
if
it's
above
below.
A
Okay
yeah,
so
that
was
just
kind
of
what
I
wanted
to
share
is
like
that
I
think
So.
Eventually,
the
idea
is
that
this
just
gets
us
a
another
way
to
look
at
these
these
jobs
to
make
it
a
little
bit
easier
and
the
thing
is
we
could
do
the
same
idea
with
like
so
we
talk
about
like
the
Creator
running
time.
We
do
the
same
idea
with,
like
you
know.
A
The
HTTP
requests
that
we
see
the
post
gets
the
number
of
those
things
and
see
if
chart
them
and
see
how
they're
changing
and
then
that
same
idea,
like
you
know
how
many
it's
like
an
estimate
for
a
scale
and
then
anything
else
that
we
kind
of
come
up
with
would
be
the
idea.
A
Yeah,
well,
it's
something
that
we'll
have
to
look
into,
but
I
think
like
whenever
lubo's
back,
we
can
kind
of
discuss
this
and
maybe
even
bring
in
Daniel
to,
and
you
probably
have
some
like
me.
You
have
some
ideas
too
Brian
we
can
kind
of
get
together.
One
of
these
calls
and
see
what
we
can
come
up
with.
A
Cool
okay,
that's
all
I
had
I
owed
for
results.
I
had
already
looked
at
these
and
the
the
results.
Look.
Okay,
there
hasn't
been
much
change
so,
instead
of
reviewing
going
through
individual
ones.
Here,
the
girls
have
been
much
changed
so.
A
Yep,
so
that's
an
important
announcement
so
today
and
then
for
in
terms
of
the
release
schedule.
They'll,
be
it
should
be
a
few
weeks.
There'll
be
a
few
cycles
of
urc's
I.
Think
the
two
things
that
we're
waiting
on
are
one
of
them
is
the
the
126
gates
for
Central
S9
need
to
be
passing
and
then
the
second
one
is
the
the
odd
security.
A
C
And
this
might
not
be
related
to
perfect
scale
but
related
to
the
current
topic
about
feature
freeze,
so
I
noticed
one
thing
that
kubernetes
also
does
the
cube.
Community
is
that
when
these
kinds
of
feature
freezes
are
coming
up,
they'll
look
at
the
major
features
going
in
and
try
to
come
up
with
some
kind
of
blog
post
or
announcement.
C
A
Yeah,
that's
a
good
question.
I
think
yeah.
This
is
this
would
be
one
like
area
where
we
we've
been
sort
of
discussing
like
Andrew
burden's,
been
sort
of
trying
to
get
together,
and
you
know
figure
out
how
we
want
to
do
release
notes
and
organisms
to
a
way
that
they're
useful
just
for
people
to
see
like
what
features
are
changing
and
I
guess
what
I
it
sees
this
being
like
a
an
add-on
to
that
like
hey.
C
Yeah
yeah
I
again
today's
question
in
the
next
week's
community
meeting
and
see
if
we
can
have
more
discussion
around
this.
C
A
A
It's
like
the
the
new
Cadence,
a
bunch
of
things
are
changing,
and
so
the
thing
to
expect
is
that
they'll
be
there'll,
be
a
little
bit
of
growing,
so
it'll
have
to
occur
and
then
as
like,
because
I
like
I
guess,
the
biggest
thing
you
we're
going
to
notice
here
is
right.
We're
gonna
have
a
few
releases
of
release,
notes,
which
is
a
big
change.
So
it
sort
of
begs
the
question
of
okay.
How
can
we
highlight
the
important
ones
which
is
kind
of
what
you're
cutting
at
so
I?
Think
I?