►
From YouTube: 2022-09-26 meeting
Description
Open Telemetry Meeting 1's Personal Meeting Room
A
A
Yeah,
it
seems
like
you're
having
some
issues,
but
we'll
give
some
more
people
in
you
know
some
time
to
get
ready.
C
A
Yeah,
let
me
let
me
look
at
the
thread.
I
think
I
have
had
some
experience
with
these
issues
before
so.
I
can
probably
take
a
look
at
it.
Just
I
just
woke
up
so
give
me
some
time.
Yeah.
C
I
I'm,
just
thinking
that
I'm
about
to
I've
got
some
time
today
and
there's
likely
going
to
be
a
few
PR's
come
up
for
some
service
readies
and
I'm
gonna
copy
the
framework
we
did
for
the
checkout
service.
That
was
you
think
that
kind
of
hung
me
up
and
to
be
great,
if
we
could
figure
it
out.
A
Yeah,
if
you
just
want
to
open
the
pr
with
the
broken
issue,
I'll
just
pull
it
real,
quick
and
fix
it,
and
then
you
can
just
pop.
A
Awesome
cool
Austin,
are
you,
are
you
about
good,
or
you
saw
some
issues.
A
What
was
them
still
issues,
but
this
is
kind
of
a
good
group
for
us,
though,
because
we're
only
really
talking
about
kind
of
our
B1
release
items,
it's
already
the
end
of
September,
so
it's
really
just
kind
of
finishing
out.
A
You
know,
whatever
finishing
polish,
want
to
put
on
top
of
the
V1
release,
so
I
put
a
couple
items
around
like
improved
documentation.
Here
it
sounds
like
you're
ready
to
talk
about
that
a
little
bit.
We
can
talk
about
our
potential
hero
scenario
and
then
maybe
we're
a
default
dashboards
could
just
improve
our
default
user
experience
and
then
maybe
the
Euro
scenario
experience
as
well.
So
I
was
just
kind
of
the
the
last
tracking
items
we
have
did
anybody
else
have
anything
or
anything
top
of
mind.
A
The
thing
that
I
want
to
talk
about
like
just
improving
debugging
facilities,
okay,
the
reason
is
yeah
I-
can
talk
it
once
we
get
there.
Yeah
no
problem.
C
C
For
the
vendoring
of
it
right,
I've
got
that
figured
out
if
you're
doing
a
Docker
build
or
Docker
compose
whatever
it
seems
yeah.
B
Yeah
I
can
talk
through
that
because
I
tried
to
push
something
we
can
get
through
that
later.
Yeah
I.
B
A
A
Okay,
under
the
services
yeah,
then
we
have
the
beautiful,
beautiful
checkout
service
and
depth
yeah
and
so
you'll
just
be
replicating
this
to
the
other
languages,
I
guess
or
service
Representatives
or
whatever.
That
is
and
I
guess.
This
is
probably
where
you're
having
the
issue
is
this
link
up
here,
yeah.
C
C
B
C
You
gotta
go:
go
pull
up
the
366
PR
uh-huh,
so
it's
closed
already,
but
if
you
just
pull
it
up.
B
C
B
C
B
A
B
B
Would
be
my
assumption,
but
if
I'm
looking
at
the
I'm
looking
at
Doc
FX
job
in
the
workflow
and
it
really
just
does
Doc
FX
build
dry
run.
So
it's
not
even
publishing
anything.
It's
not
transforming
anything.
It's
just
running
I,
guess,
stock
effects,
lint
base
or
it's
running
a
lint
basically,
but
there's
also
a
markdown
link
check
that
runs.
So
what
compute
like
what
is
Doc
FX
getting
us
so
that
that
isn't
well.
A
A
A
Yeah,
we
also
just
maybe
posted
a
slack
Channel
real,
quick
and
see
if
anyone
has
a
complaint,
and
maybe
Juliano
can
just
respond
real,
quick
too.
B
B
C
B
Yes,
it
makes
no
sense
to
me
that
we
are
that
this
is
the
problem,
that
it
is
the
other
thing
that
you
could
do.
One
thing
you
could
try
that
might
just,
but
to
spitball
like
what,
if
you
put
a
fully
qualified
hyperlink
in.
B
Because
my
only
like
the
one
thing
to
turn
around
on
this
documentation
discussion
is
that
comms
is
trying
to
build
out
a
new
subsection
of
the
site,
which
is
where
documentation
or
the
where
the
demo
page
would
live.
B
Basically
right
now,
if
you
go
to
community
up
here,
it's
not
it's
just
this,
but
we're
trying
to
make
this
something
that
can
have
subsections,
and
one
of
those
subsections
would
be
like
open,
Telemetry
demo
I'm,
not
I,
don't
like
I,
don't
like
using
sub
get
sub
repos
I,
don't
like
using
sub
repos,
so
I
guess
we
shouldn't
optimize
for
that.
B
I
think
we'll
just
have
a
link
to
the
to
the
GitHub
repo,
so
yeah
either
get
rid
of
Doc
FX
or
change
the
things
that
don't
like
relative
links
to
folders
to
fully
qualified
hyperlinks.
C
Yeah
I
think
we'll
start
with
that,
so
making
this
two
PRS
one
two
temporarily
remove
it
and
if
we're
satisfied
just
completely
nuke
it.
A
Yeah
want
to
screen
test
it.
It
sounds
like
a
documentation,
at
least
it's
not
a
good
point
pure
and
I,
like
the
the
checkout
service,
so
I
can
help
with
some
of
the
other
stuff
too.
But
just
let
me
know,
I've.
C
Got
a
wonderful
long
flight
this
evening,
I
am
going
to
Europe
and
so
you'll
see
me
on
banging
away
quite
a
bit
this
evening
and
and
if
you
want
to
just
hit
me
up
on
slack
and
we'll
just
pick
a
service
and
run.
But
that's.
C
C
B
A
Okay,
great
appreciate
it
before
you
go
to
Oslo,
okay,
so
the
draft
hero
scenario,
I
think
Austin.
You
put
some
thoughts
over
the
weekend
on
a
slack
Channel,
but
I
was
hoping
some
other
people
might
have
some
input
too,
but
yeah.
C
B
B
Okay,
I'm
quite
confident
you
can
tell
it
to,
and
also
admittedly
thinking
like
I'm
I'm
thinking,
mostly
kubernetes
versus
Docker
here,
but
I
I
think
the
problem
is
set
is
the
same
either
way.
C
So
an
idea
we
could
do
is
is
we
could
you
know,
pick
a
service?
Any
service
I
think
check
out
checkout
kind
of
feels
like
a
big
GateKeeper
it'll,
be
massively
noticeable,
but
check
out
just
builds
some
kind
of
internal
BS
cache
on
a
feature
flag
that
just
grows
which
each
request
and
it
never
shrinks,
and
eventually
the
checkout
service
will
will
wear
kubernetes
and
our
Docker
restart
policies
will
kick
in
to
bring
it
back
alive.
Yeah.
C
And
you
can
even
make
it
so
to
only
ads
to
the
cash
under
specific
scenarios,
but
we
would
we
would
have
to
hard
code.
Those
scenarios
type
deal
right.
B
Yeah
we
should.
B
Think
well,
I
think
met
I,
mean
I,
think
we're
so
right
now,
I'm
I
am
officially
waving
the
like
holy
flag
about
metrics
like
we
are
way
too
close
to
drop
dead
and
metrics
is
like
that
is.
The
kubecon
announcement
for
otel
is
metrics
GA,
so
we
one
thousand
percent
have
to
have
like
some
of
the
with
metrics
in
it
and
it
has
votelmetrics
who.
B
I
think
the
checkout
cache
is
fine
and
probably
easy
to
implement.
Even
though
that's
again,
that's
Alpha,
but
we
don't
need
like
we
don't
need
to
go
wild
here.
We
just
need
a
counter.
You
know
like.
B
The
kids
yeah,
but
it
can
be
a
really
simple
gauge
metric
of
just
like
cash
high,
like
cash
items,
even
I,
actually
think
the
most
interesting
thing
would
be
that
the
metric
is
it
your
all
tracking
logically,
the
same
thing
but
with
different
like
pieces
of
telemetry.
B
But
the
shared
attributes
are
a
long
service
like
service
and
operation
and
then,
if
you're,
using
Hotel
metrics,
you
should
be
able
to
attack
then
that
so
using
Hotel
metrics
and
you
have
an
Exemplar
aware
backend,
then
you
would
see
the
associated
Trace
metrics
or
you
would
see
the
associated
traces
because
they
would
be
exemplared
into
the
metric.
B
You'd
be
able
to
see
like
oh
here's.
This,
like
I,
see
I,
see
my
om
and
then.
B
C
Or
maybe
some
like,
you
know,
get
from
cash
and
it
would
return
a
hit
or
no
hit.
I
we'd
probably
want
to
work
that
one
out
a
little
bit.
Yes,.
B
A
B
B
We
created
all
right,
so
checkout
service
gets
the
cash
whenever
something
gets
written
to
the
K.
There's
a
gauge
on
that
cache
and
whenever
that
gauge
is
changed
or
whenever
it
gets
added
to
or
whatever,
then
we
use
otel
metrics,
which
will
create
an
Exemplar
which
will
look
at
whatever
the
current
span
is
and
then
link
it
as
an
Exemplar.
B
I
think
pretty
sure,
that's
how
this
works,
that's
how
it
should
work,
and
then
we
just
need
to
make
sure
that
whatever
we're
sending
the
metrics
to
is
Exemplar
aware,
which
might
mean
we
need
to
Sub
in
like
Loki
or
something
looking
in
cortex.
B
And
then
we
make
sure
that
and
then
we
make
sure
that
it
has
like
the
shared
resource
attributes
right,
so
it
has
like
service
version
or
service
name
and
like
operation
name
or
whatever,
and
that
should
be
enough.
I
think
forever,
like
that,
should
be
the
minimum
set
of
stuff.
If
you
know,
if
the
honeycomb
Fork
you
know,
honeycomb
vendor
experience
needs
something
else,
then
it's.
C
C
C
C
It's
kind
of
like
the
the
all-encompassing
path
or
Road.
C
We'll
do
that
we'll
cash
every
currency
code,
every
time
you
do
something
with
a
specific
currency
code,
for
whatever
reason,
we're
writing
cash
and
we're
done.
B
C
B
C
Don't
know
if
we'd
load
data
dot,
the
currency
conversion.json
at
startup
or
if
they're,
actually
hard
code
I
forgot,
which
one
it
was
but
either
case
you're
only
looking
at
like
33
currencies.
So
it's
far.
B
B
B
So
for
everything,
red
metrics,
then
host
metrics
and
then
like.
If
we
have
an
Associated,
flagged
scenario,
then
whatever
chart,
you
need
to
help
you
like
see
that
so
for
the
checkout
service,
you
would
need
the
right
error,
duration
and
then,
like
the
cache
size,
metric
right.
A
Yeah
I
think
that'd
be
a
good
one,
and
none
of
this
would
be
specifically
Trace
related
information.
Probably
it
would
all
just
be
vanilla,
metrics
or
hosts.
B
Unless
I'm
losing
my
mind,
auto
collector
config
I.
B
B
Anyway,
we
have
a
Spam
metrics
processor
that
I
assume
works
this
way,
metrics
exported
from
atheists
and
that
exports
from
atheist
metrics.
So.
B
I,
don't
think
this
helps
right
now
like
looking
ahead.
Okay,
so
like
looking
into
the
Future
Let's,
this
is
the
problem.
Is
this
is
more
of
a
version
two
thing,
but
let's
assume
that
inversion
by
the
time
we
you
know
act
like
at
some
point.
We
have
drop
in.
You
know
foreign
metrics
plugins
that
do
things
like
well
here
for,
like
oh,
you
have
grpc
cool,
then
here's
rate
error,
duration
for
all
your
grpc
endpoints.
B
You
know
as
histograms
like
at
that
point.
It's
like!
Oh
okay,
then,
let's
change
how
the
traces
work
and,
let's
start
like
exploring.
How
would
you
act
like
what
would
this
actually
look
like
if
this
wasn't
a
toy
scenario?
And
this
was
like
a
more
real
world
thing
and
you
were
talking
about
like
hundreds
of
thousands
or
like
millions
of
QPS
and
you
need
to,
and
so
you
have
to
do,
sampling
and.
A
B
And
that's
the
point
where
it's
like:
okay
now
you
know
the
metrics
plus
traces
stuff
works
a
little
better
because
you
have
the
metrics
for
True
numbers
and
then
you
have
the
traces
that
are
being
sampled,
that
have
you
know
the
actual
sampling
information
encoded.
But
you
know
again
it's
a
bit
more
real
world.
I
guess
yeah.
A
This
is
where
I
think
actually
Harris
unit
might
help
us
kind
of
mature
and
improve
our
own
kind
of
debugging
story.
Just
because
we're
actually
doing
it
I'd
be
curious
from
the
open
Telemetry
website.
Is
there
anything
you're
doing
from
like
a
a
dashboarding
perspective
or
maybe
tracking
a
signal,
so
we
could
just
maybe
replicate
and
the.
B
C
C
B
Well,
that's
the
thing
is
and
well
that
yes-
and
that's
that's
kind
of
the
core
Point
here-
is
that
most
people,
most
of
the
time,
actually
don't
have
high
enough
throughput
services
to
really
force
you
to
sample
like
when
you
start
getting
into
like
yeah.
There
are
people
that
are
running,
like
million
QPS
services.
B
My
own,
my
only
point,
is
this:
I,
don't
believe
that
we
will
have
those
drop-in
plugins
for
1.0
I,
think
for
1.0
we're
going
to
rely
on
metrics
from
spans
and
we're
going
to
have
like
a
handful
of
like
we're.
Gonna
have
our
hero
scenario
where
it's
like.
Oh
we
created
this
metric
right,
so
we
should
build
dashboards
with
that
in
mind
where
our
you
know
we're
kind
of
restricted
to
what
we
can
get
from
The
Collector,
giving
us
those
metrics.
B
C
A
B
B
C
A
Yeah
I'm
trying
to
create
some
action
items-
that's
probably
a
better
way
to
transition
this,
so
PR
is
going
to
work
on
documentation
for
the
services
for
the
hero
scenario.
Let's
go
ahead
and
type
out
some
items,
so
Matrix
sdks
or
go
Python
and
net.
A
A
Okay
and
then
we
will
need
to
create
the
default
or
create
a
metric
so
create
there.
You
go
catch
metric,
all
right.
A
And
then
create
the
default
dashboards.
Well,
let's
call
it
a
default
dashboard
for
now
create
default.
Dashboard
click,
basic
host
protocol,
slash
hero
scenario
metric-
is
that
pretty
much
all
the
items
we
have
or
major
items
we
have
for
the
hero
scenario?
B
Doing
if
you
can
put
multiple
scenarios
in,
if
we
can
have
multiple
I
mean
the
thing
is
I,
don't
think
we
need
to
show
metrics.
We
need
to
show
metrics
in
as
many
languages
as
possible,
but
we
don't
need
to
show
metrics
in
multiple
languages.
If
that
makes
sense,
although
I
will
say
if
we
wanted
to.
If
someone
wanted
to
go
to
product
catalog
and
add
in
like
the
Prometheus
Auto
instrumentation
for
go
or
whatever
that'd
probably
be
fine,
like
I,
think
there's
something
to
be
said
for
hey.
We
have
this:
let's
show
it
off.
C
B
A
B
A
Yeah,
so
I
think
Key
Well
I
think
has
proposed.
They
will
merge
in
they'll,
donate
their
Auto
instrumentation.
A
I
think
there
is
a
PR
out,
I,
don't
think
it's
much
easier.
At
least
last
I
think
it
was
not
merge
yet.
B
Because
my
my
point
is
we're
just
we're
getting
way
too
close
to
like
go
time
for
this
demo
for
me
to
be
comfortable,
like
waiting
on
other
cigs
to
get
their
together
like
if
there's
something
we
can
put
in
here
that
works,
then,
let's
put
something
in
here
that
works
today
and
demonstrate
like
well
hey
look.
This
is
how
you
use
like
this
is
how
you
scrape
Prometheus
metrics
and
then
send
them
to
otlp
endpoints
right
I.
C
A
A
Meet
it
now,
just
one
question
on
the
cart
ad
and
recommendation
Services,
we
just
be
adding
the
instrumentation
libraries
and
the
metric
or
meter
providers
for
now,
or
we
want
to
be
doing
more
besides
checkout
service,
go,
never
check
out
service,
we're
doing
native.
You
know
an
actual
open,
Telemetry
instrument,
metric
I'm,
not
sure
if
we
want
to
more
of
more
of
that
work
like
the
python.net.
A
One
thing
we
can
do
with
net
if
they
have
standard
metrics
from
the
runtime
now
and
I.
Think
cjo
is
one
of
the.net
developers
and
the
Sig,
and
also
on
my
Microsoft
team.
Let's
talk
about
doing
some
default
dashboarding
for
that
and
integrating
that
into
the
demo.
So
I'll
just
make
a
note
to
Ping
him
and
we'll
see
if
we
can
get
that
turned
on
and
then
also
just
integrated
into
the
default
dashboard
you
make
so
at
least.net
should
be
covered.
B
Yeah,
let
me
see,
are
you
all
cool
going
over.
C
B
Metric
support,
some
things
do,
there's
requests
has
metrics
you're
a
lib,
whiskey,
Falcon,
Django
I,
don't
see
no
grpc.
B
Anyway,
there
might
be
stuff
in
python
I
think
we
should
just
turn
it
on
see
what
we
get
yeah.
C
B
A
very
a
hacky
version
of
basically
replicating
the
functionality
of
the
existing
like
extras
like
merged
contlector,
config
and
Tyler's
kind
of
pushback
was
like
well,
we
should
do
it
like
we
do
the
other
collector
thing
and
I'm
like
okay
I'm
in
a
speed
mood.
Not
a
do
it
right
mood,
so
if
you
just
want
to
have
Tyler.
B
C
I
just
said
that
over
I
do
have
to
run
yep
I
will
chase
up
on
him
and
you'll
see
stocks
coming
from
me
later
on
this
evening.
Okay,.
C
A
B
A
B
You
guys
yeah,
we
can
just
talk
about
stuff,
say
Saint
Carter,.