►
From YouTube: SIG Instrumentation Usage Metrics Collector 20230125
Description
No description was provided for this meeting.
If this is YOUR meeting, an easy way to fix this is to add a description to your video, wherever mtngs.io found it (probably YouTube).
A
Hello
and
welcome
to
today's
special
edition
of
Sig
instrumentation,
we
are
having
a
kickoff
meeting
for
our
new
sub-project
usage.
Metrics
collector
and
I
have
the
pleasure
of
Hosting
today's
meeting,
but
I
would
love
to
go
around
and
do
intros.
We
have
a
lot
of
new
faces
lots
of
people
that
I
don't
normally
see
it's
the
instrumentation
meetings,
which
is
amazing
if
you
haven't
already
I,
dropped
a
copy
of
the
link
to
the
meeting
notes
in
the
chat.
Please
feel
free
to
put
your
name
on
there.
A
A
So
my
name's
Elena
I'm,
currently
one
of
the
co-chairs
of
Sig
instrumentation
I
work
for
apple
and
I'm
gonna
go
through
these
three
things
in
the
intro
section,
so
I've,
given
my
name
a
little
bit
about
my
background
with
kubernetes
I'm,
chair
of,
say,
instrumentation.
What
do
I
want
to
get
out
of
the
sub
project?
I
am
one
of
the
folks
who
has
been
working
on
donating
this
code
to
kubernetes
and
making
it
awesome
and
bringing
in
new
collaborators.
B
Hey
everyone,
I'm
David
I'm,
one
of
the
TLs
for
Sig,
instrumentation
and
I'm,
actually
just
here
to
listen
and
learn
and
see.
I
suppose
I
feel
like
this
is
a
big
unveiling
of
a
lot
of
work.
That's
gone
on
for
a
while,
so
I'm
very
curious
to
see
what
what
people
have
come
up
with.
C
Hello,
hopefully
you
can
hear
me
I'm
Louis,
actually,
I've
been
a
traveler
for
a
few
years.
I've
mostly
worked
on
building
super
kubernetes
clusters
and
the
tooling
to
build
those
at
scale
I'm
now
part
of
a
team
that
is
mostly
concerned
with
efficiency
and
auto
scaling.
So
you
can
see
the
interest
and
having
good
metrics
I
will
pass
the
Baton
to
Katrina.
D
D
Also,
these
days
a
tech
lead
and
chair
for
six
CLI
and
some
Project
Lead
for
customized
in
particular,
so
you'll
see
me
around
the
community
in
those
forums
and
my
interest
in
this
one
in
particular,
is
that
I
do
the
kubernetes,
open
source
team
at
Shopify
and
try
to
find
opportunities
where
the
projects
that
are
going
on
in
the
community
would
really
benefit
us
that
really
solve
a
problem
that
we
have
internally,
and
this
seems
really
promising
for
some
of
the
Auto
scaling
investigations
that
we
were
doing
in
in
last
year.
D
E
Whoops
I
don't
know
how
to
use
zoom
either
I'm
Paul
I
work
for
Apple
I've,
been
in
a
few
different
areas
of
of
kubernetes
over
the
years,
and,
let's
see
what
do
I
want
to
get
out
of
the
sub
project.
E
I
think
what
at
this
point,
what
I'm
very
interested
in
is
you
know,
as
we
share
this
technology
with
the
community
I'm
very
curious
to
know
like
areas
of
resonance
with
problems
that
you
may
already
be
trying
to
solve
areas
of
like
dissonance
right
like
if
there's
something
about
the
pro
the
approach
that
we're
taking
in
this
collector.
That
is
incompatible
with
problems
that
you're
trying
to
solve.
That
will
also
be
very
interesting
to
know,
and
you
know
I
think
those
are
are
the
main
two
things
but
comedy
comes
in
threes.
So
if.
C
E
F
Hi
I'm
ty
from
Shopify
I'm,
also
on
Louie's
team.
We
spoke
a
bit
part
part
of
the
reason
we're
interested
in
this
is
driving
good
recommendations
for
resource
requests.
We
spoke
a
bit
about
this
at
a
cncf,
I,
think
Sig,
ux
anyways,
so
just
interested
in
in
seeing
kind
of
high
resolution,
metrics
options
and
sorry
I'm,
gonna,
say
Richard.
Next,
sorry,
if
I'm
not
getting
that
correct.
Oh.
G
That's
right
hi.
This
is
Richa
I
work
for
Google
and
I
am
fairly
new
to
kubernetes.
I
just
recently
gained
a
Kate's
membership
with
some
small
contributions
to
the
community.
I
am
here
to
learn
more
reward,
kubernetes
and
try
to
use
my
membership
as
much
as
I
can
to
make
meaningful
contributions.
A
Thank
you
very
much,
Rita
and
I'd
like
to
add
to
that
briefly
before
we
pass
on
to
another
person.
If
you
are
looking
for
org
membership,
and
you
want
to
contribute
on
an
ongoing
basis
to
the
sub
project,
but
you're
not
really
sure
about
that.
Zig
instrumentation
is
a
great
place
to
get
started
with
kubernetes
and
we
are
very
happy
to
sponsor
you
if
you're
active
in
any
of
our
sub
projects
for
kubernetes,
sigs
or
kubernetes
membership,
as
makes
sense.
So
please
let
like
any
of
the
leads
know.
A
In
particular,
we
can
definitely
help
sponsor
you
and
share
the
next
steps
for
that.
Sorry
and
I
I
totally
took
over
from
you
so
Rita
did
you
want
to
pass
the
potato
to
someone.
G
Yes,
sure
is
it
well
yeah.
It's.
H
Wrong
hi
I'm
John,
currently
I'm
working
for
rabbit
in
on
the
monitoring
team
and
over
the
last
year,
I've
been
trying
to
contribute
to
Prometheus
operator
and
also
discovering
the
space
and
seeing
what
other
projects
would
interest
me
and
what
do
I
want
to
get
out
of
this
sub
project
I'm
guessing
it's
a
bit
like
like
David
I,
came
here
to
learn
about
the
project
and
how
we
could
use
it
and
I'm
passing
the
Hot
Potato
to
Kevin.
If
I'm
not
mistake,.
I
Hey
wait
and
then
also
for
you,
I'm
on
the
phone:
hey
I'm
on
the
Google
companies,
Team
working
on
observability
and
I'm
very
excited
about
this
project,
because
it
touches
a
few
areas
of
eye
for
my
team
has
been
previously
looking
into
like
high
resolution
resource
usage
data,
mainly
for
overscaling,
so
I'm
curious
on
how
the
different
ideas
align
or
how
we
can
help
where
I
can
help
to
to
make
this
happen.
I
A
Awesome
I
think
there
was
everyone,
so
without
further
Ado,
not
quite
everyone.
Oh.
C
J
Right,
let's
see
I
started
working
on
kubernetes
about
two
years
ago
at
Apple
before
that,
I
was
actually
doing
something
really
different.
I
was
doing
computational,
mechanics
and
machine
learning.
What
that's
the
story
for
another
day,
but
yeah
I've,
Been,
Working,
On,
The
Collector
for
the
past
year
and
changed
and
I'm
excited
to
share
the
cool
stuff
we've
been
doing
here
at
Apple
and
Hot
Potato
too
hot.
K
Great
okay,
yeah
hi,
I'm
Han
I'm,
a
Staff
software
engineer
at
Google,
I'm
I,
I
coach,
our
seek
instrumentation
with
Elena
we've
been
doing
that
for
a
while
I've
been
working
on
kubernetes
for
the
past,
like
four
and
a
half
years,
API
machinery
and
instrumentation
mostly
I,
also
specialize
in
storage
layer
stuff
but
yeah
I'm.
Looking
forward
to
this
usage,
metrics
collector
and
seeing
what
it's
all
about.
A
A
Okay,
then,
let's
take
a
look
at
can
I
make
this
go
away.
Please
no
I
just
can't
see
my
URL
bar,
because
Zoom
is
covering
it
and
I,
don't
know
how
to
make
it
like
move.
Maybe
I'll
try
to
move
it
to
the
bottom
of
my
screen,
yeah,
okay
or
in
the
smack
dab
middle
okay.
So
let's
take
a
look
at
our
repo
I
want
to
do
a
quick
sort
of
just
like
Code,
walkthrough
overview.
I'm.
Sure
folks
will
have
some
questions.
A
I
will
also
caveat
with
I
am
also
new
to
this
code
base
to
some
extent,
I
only
joined
Apple
in
September,
and
so
I
will
do
my
best
to
answer
questions
and
whatnot,
but
there's
some
members
from
our
team
that
could
make
it
today
I'm
hoping
that
we
can
meet
again
and
we
can
go
more
in
depth,
but
today
we're
just
going
to
do
an
overview.
A
I
am
going
to
show
you
first,
here's
the
repo
for
our
project,
so
you
can
find
it
under
kubernetes,
sigs
and
I
will
add
a
link
to
that
in
our
meeting.
Notes.
A
I
also
wanted
to
quickly
point
out,
because
not
a
lot
of
people
necessarily
like
you
might
know
where
to
find
code,
but
then
navigating
test
grid
is
kind
of
scary,
so
I
will
also
share
a
link
to
our
test.
Grid
results.
A
flaky
test:
oh
no
and
I
will
make
it
small.
A
Oh
these
are
PR
so
that
that
makes
sense.
Maybe
so
here
are
periodics,
which
I
hope
are
not
flaking.
K
E
A
Yeah
I
thought
I
deflaked
it
darn,
okay,
yeah,
there's
been
some
issues
with
these
tests.
I
ran
into
like
I,
saw
that
they
were
very
sneakily
getting
killed,
so
I
increased
the
the
the
memory
limits
for
them,
but
okay,
so
there
are
probably
people
who
have
never
seen
test
grid
before
as
well.
So
we
have
this
repo
where
our
code
lives
and
then
in
order
to
set
up
testing
for
it,
we
have
a
repo
and
kubernetes
called
test.
Infra
and
I.
A
Don't
know,
I've
gone
to
various
places
here,
oh
great,
that's
pretty
close
to
where
I
want
to
be,
and
so
there's
all
sorts
of
folders
and
it's
a
very
large
repo-
and
you
know
you
I
hope
you
will
not
get
too
overwhelmed,
but
the
place
where
all
of
these
definitions
live
are
in
here,
but
you
don't
have
to
know
where
the
repo
is
because
this
has
a
little
URL
to
the
code
that
backs
the
tests.
A
So
as
long
as
you
have
the
test
grid
link,
you
can
copy
that
and
you
can
go
to.
Okay.
Here
are
the
definition
of
the
test.
So
I
was
talking
about
adding
resources
like
to
the
requests
and
limits.
Well
here
they
are
they're
defined
right
here,
and
so
you
can
see
exactly
like
what
image
is
this
running
with
what
command
does
it
run?
In
our
case,
it's
running
make
test
which
then
you
know
like
okay.
Well,
let's
look
at
the
make
file
of
this
repo.
A
It
is
a
little
bit
messy,
but
what
is
it
doing?
I
set
it
up
recently
so
that
it
goes
and
goes
and
runs
our
tests
and
outputs
it
in
junit
format
and
that's
how
they
show
up
in
test
grid
so
hooray.
So
we
have
a
make
file
and
we
have
two
Targets
for
our
CI.
We
have
verify
and
test
if
you're
interested
in
more
targets
and
you're
interested
in
CI
things.
Please
talk
to
me
in
terms
of
the
code
itself.
A
We
are
generating
three
main
binaries
and
so
I
will
talk
a
little
bit
about
container
Monitor
and
node,
sampler
and
I.
Think
Paul
will
talk
a
little
bit
about
the
collector,
so
the
container
monitor
is
like
a
relatively
new
piece
of
code.
I
think
that
this
just
goes
and
imports
main
yeah.
A
So
this
launches
some
basic
Prometheus
metrics.
If
you
want
to
to
scrape
directly
off
of
a
given
node
or
something
like
that,
this
was
this
code
was
initially
added,
mostly
for
debugging,
so
you
can
launch
a
containerdy
client
and
it
will
scrape
some
metrics
directly
from
c
groups
off
of
a
node,
so
it
is
we're
using
Cobra,
for
it
is
configured
via
command
line
this
one.
A
We
also
have
any
questions
about
that
before
I
move
on.
So
very
briefly,
this
is
sort
of
a
debugging
app
that
we
used
question
y.
So
we
want
to
get
high
fidelity
metrics
for
kubernetes.
We
want
to
be
able
to
present
those
in
a
relatively
consumable
format-
that's
not
too
high
cardinality.
So
how
might
we
do
that?
Well
in
order
to
get
stats
on
various
processes
running
on
a
Linux
system,
we
need
to
look
at
the
containers.
The
containers
are
running
c
groups.
A
The
c
groups
come
with
all
sorts
of
fun
stats
with
them,
so
these
are
collected
in
other
parts
of
the
project
via
C
advisor,
for
example,
but
C
advisor
looks
a
lot
of
things.
We
don't
really
need
to
look
at
all
of
those
things
if
we
just
want
to
focus
on,
for
example,
CPU
utilization
and
memory
utilization.
So
there
are
inevitably
something
somewhere
is
going
to
have
to
go
and
look
at
c
groups.
A
But
if,
if
a
thing
is
a
container,
but
what
if
a
thing
is
not
a
container,
then
in
that
case
it
might
be
more
helpful
to
talk
directly
to
the
container
runtime,
because
in
the
container
case
it
will
go
and
maybe
look
at
the
C
group.
But
if
it's
something
else
like
a
micro
VM,
it
should
have
some
way
of
getting
the
stats
for
the
utilization
for
that
runtime.
A
So
we
get
samples
on
a
per
second
basis
that
we
then
buffer
into
a
collector,
so
container
monitor
include
some
code
for
experimenting
with
using
containerdy
and
not
just
looking
at
c
groups
and
the
sampler,
which
is
where
I've
primarily
worked,
is
where
we
are
so
you
can
see,
there's
sort
of
some
split
stuff
or
Linux
specific,
because
the
containerdy
libraries
only
work
on
Linux
other
systems
do
not
support
them,
because
containerd
does
not
support
running
on
those
systems,
but
we
currently
run
with
containerd
for
some
of
our
systems.
A
So
hence
wanting
to
talk
to
it.
So
this
is
some
of
the
code
there
for
setting
that
up.
Somebody
I
heard
make
a
sound.
Somebody
want
to
jump
in
I,
don't
see
any
hands.
A
Okay,
I
thought
I
heard
someone
unmute,
but
so
this
sampler
is
the
thing
that
will
look
at
doing
that,
like
actual
scraping,
specifically
of
getting
those
like
per
second
container
samples
on
the
Node
for
memory,
CPU
and
so
on.
A
The
actual
code
for
that
lives
in
package
and
I
haven't
actually
looked
at
this
in
a
little
bit.
So.
A
And
I
was
playing
around
with
integrating
the
containerdy
way
of
getting
information.
So
that's
some
of
the
stuff.
A
E
But
that's
okay!
It's
it!
We
only
really
have
a
few
minutes,
so
I'll
just
cover
cover
the
the
very
high
level
stuff.
Why
don't
you
open
up?
Pkg
collector
API,
all
right
and
oh,
no,
that's
the
wrong
package,
not.
E
E
H
G
E
One
yeah,
okay,
because
I
I
thought
you
know
before
diving
into
like
what
is
the
actual
code
for
The
Collector.
We
can
talk
about
what
are
the
things
that
you
are
allowed
to
express.
So
what.
E
Is
the
go
type
for
the
main
configuration
API
for
The
Collector
and,
let's
see,
let's
take
a
look
at
this,
so
we
have
C
group
metrics,
which
are
what
they
sound
like.
They
are
metrics
derived
from
information
that
we
get
from
scraping
the
c
groups.
There
are
so-called
cluster
scope,
metrics
one
of
the
least
favorite
things
for
me
in
software
engineering
is
naming
things
and
I'm
unhappy
with
the
name
clusterscope
metrics.
What
does
it.
G
E
It
means
it
is
about
taking
information
which
is
stored
on
some
cluster
scoped
resource
in
a
kubernetes
cluster
and
emitting
metrics
from
it,
and
we
can
talk,
maybe
in
a
future
meeting
about
exactly
what's
happening
there.
Next
we've
got
extensions,
so
these
are.
Are
these
are
user
defined
user
in
the
sense
of
the
person
running
this
code
in
cooperation
with
their
users
that
are
running
workloads
can
Define
certain
labels
on
the
workloads
that
are
then
transformed
into
metrics
labels
and
about
I'm
doing
my
best
here
if
I
say
anything
wrong,
just
holler.
E
So
next
we've
got
the
built-in
metrics,
which
are
what
they
sound
like.
There
are
some
metrics
that
you
get
that
are
just
built
in
that
use
some
Reserve
labels
and
stuff
like
that.
I
think
those
are
really
the
big
ones.
E
There's
some
nitty-gritties
in
here
as
far
as
the
metrics
go,
but
let's
take
a
look:
if
we
can
just
scroll
down
a
little
bit
and
look
at
resources
which
are
which
resources
are
you
interested
in
and
then
aggregations
and
aggregations
is
one
of
the
central,
essential
things
about
this
architecture,
which
is
that
the
the
rules
that
drive
the
emission
of
metrics
like
the
metrics
are
not
are
are
defined
as
aggregations
of
other
metrics.
E
So
in
the
version
of
the
component
that
we
had
to
collect
metrics
before
we
built
this,
we
did
not
have
a
config
driven
scheme
for
defining
the
metrics
and
aggregating
them.
So
we
were
doing
things
like
we'd
have
to
make
code
changes.
If
we
wanted
to.
You
know,
compute
a
new
layer
of
aggregation,
and
one
of
the
things
that
we
have
found
to
be
very
convenient
about
this
new
scheme
is
that
things
that
we
used
to
be
doing
in
code
that
we'd
have
bugs
in
code
are
now.
E
You
know
crushed
in
the
mind
Vice
of
a
few
people
that
built
this
collector
and
we
can
write
some
rules
which
are
much
easier
to
understand
than
the
code
and
actually
I
think
you
know
given,
given
the
amount
of
time
that
we
have
I
think
it
might
be
good
ehash.
If
we
take
a
look
at
one
of
the
integration
tests,
because
I
think
that'll
bring
that
together
for
folks
and
I,
don't
really
think
that
we
have
a
meaningful
amount
of
time
to
really
start
getting
into
the
code
of
The
Collector
itself.
A
We
actually
only
had
three
minutes
so
I'm
wondering
if
maybe
we
want
to
try
to
plan.
You
know
I
think
we
should
meet
again.
You
know,
half
an
hour
is
probably
not
long
enough
by
any
means
and
I'm
wondering
you
know
what
times
might
work
for
folks
and
then,
if
anybody
has
any
brief
questions,
because
we
haven't
really
had
a
chance
to
go
into
that
if
you're,
okay,
with
that
Paul.
A
Awesome
does
this
time
generally
work
for
folks,
and
when
would
we
want
to
meet
again?
Presumably
people
might
want
a
chance
to
go.
Look
through
the
code.
Ask
questions
that
kind
of
thing.
I
tried
to
pick
a
time
that
would
generally
be
workable
for
folks
in,
like
west
coast,
East,
Coast,
Europe
I
know
we
can't
go
much
later
than
this
for
folks
in
Europe.
So.
A
Does
this
work
like
straw
poll?
Yes,
no,
okay,
so
this
this
approximate
time
seems
to
be
okay.
Shall
we
meet
in
like
another
couple
of
weeks?
Does
that
sound
good?
We
normally
don't
meet
on
a
weekly
Cadence,
because
it's
a
little
frequent.
J
Okay,
yeah,
we
can
try
to
do
a.
This
was
a
really
good
overview
by
the
way
for
ihas
and
Paul.
You
can
do
a
deeper
dive,
the
next
time
into
like
how
the
collector
works,
because
it's
a
manager-
and
it's
also
a
custom,
metrics
exporter.
We
can
go
into
those
details
and
also
about
the
aggregation
Logic
for
that
and
how
the
node
samples
also
push
to
The
Collector
and
how
the
aggregation
is
actually
done
in
the
background.
So
we
can
go
over
the
details.
I
A
I
Yes,
sir,
let's
go
tasks
or
issues
that
would
be
interested
in
looking
into
like
a
roadmap
of
what
should
be
worked
on
next
or
should
we
discuss
it
next
time
or
is
this
something
people
already
can
look
at
if.
A
I
say:
let's
discuss
it
next
time,
we're
missing
at
least
one
person
on
our
side
in
order
for
us
to
be
able
to
talk
definitively
about
that
which
I
apologize
for,
but
amazing
question
totally
anticipated
it
sorry,
I,
don't
have
a
better
answer.
Next
meeting.
A
Yeah
and
feel
free
to
yeah
use
the
Sig
instrumentation
slack
room
to
ask
questions
feel
free
to
reach
out
to
any
of
us.
I
think
we're
all
on
kubernetes
slack,
although
you
may
not,
our
usernames
may
not
be
here,
but
I'm
e
hashman.
You
can
definitely
reach
out
to
me
or
post
in
the
channel
for
folks
who
are
like
what
channel
pound
Sig
instrumentation
on
kubernetes
slack
and
the
link
is
actually
at
the
top
of
our
dock
here
somewhere
I
think
yeah.