►
From YouTube: 2022-06-06 meeting
Description
Open Telemetry Meeting 1's Personal Meeting Room
B
D
Is
I
don't
know
exactly,
I
think,
you're
east
coast
right
austin,
so
you
get
quite
a
few
squirrels
around.
D
C
C
E
Okay,
hey
everyone
yeah,
just
getting
the
notes
set
up.
I
think
we
have
a
pretty
full
agenda
today.
I
know
juliana
is
here
giuliana.
Do
you
have
james's
material?
I
know.
B
A
E
Yeah,
that's
fair.
I
think
james
has
some
developer
standards
items
he
wanted
to
cover,
but
yeah.
So
I
think
the
main
agenda
item
we
have
today
is
to
go
ahead
and
confirm
our
infrastructure.
I
think
or
confirm
our
architecture.
Excuse
me
a
little
bit
earlier
right
now.
I
think
our
repo
is
pretty
well
set
up.
I
think
we
have
a
lot
of
items
ongoing
and
with
gary
gary
thanks
for
joining
joining
us
as
well.
E
We
now
have
some
sort
of
owner
for
rust,
ruby,
php
and
c
plus
plus,
so
we
should
be
covered
completely
from
like
a
new
language
perspective,
so
we
can
go
ahead
and
either
confirm
our
architecture
or
raise
any
open
questions.
I
think
from
like
kind
of
a
foundational
perspective,
we're
almost
ready
to
go
so
austin.
Do
you
have
any
questions
you
could
call
out?
Maybe
that
we
should
discuss.
C
Yeah,
the
I'll
just
go
through
the
new
comments,
starting
at
the
top
there's
a
kind
of
a
java
specific
service
question
about
spring
boot,
spring
cloud:
jsp
server,
whatever
web
mvc
framework
to
cover
more
tracing
span
cases,
does
anyone
is
who
put
that
gian
gian?
C
E
C
So
there
was
a
discussion
about
this
in
the
slack
channel
this
morning
around
or
last
night
or
some
over
the
weekend
about
service
swaps,
and
I
think
my
general
we
can
just
show
of
hands
this,
like
my
general
gut
feeling,
is
that
we
should
push
towards
the
v1
as
pilot
to
be
polyglot
and
have
as
many
individual
languages
represented,
because
there's
a
real
dire
need.
I
think
for
sample
code
that
isn't.
I
know
every
language
has
example
code
right
now,
but
they
don't
there's.
Those
are
all
kind
of
very
point.
C
C
E
Yeah,
I
definitely
agree.
I
was
getting
a
little
bit
worried
about
supportability
of
multiple
languages
and
just
finding
owners
kind
of
across
this
spectrum
of
languages,
but
yeah.
I
do
feel
a
lot
of
the
current
open,
telemetry
examples-
they're
very
hello
world
at
the
moment.
So
if
we
can
have
a
more
complicated
example,
that
would
definitely
be
desired.
So
in
this
case,
are
they
asking
to
expand
like
other
services
to
java.
E
C
A
I
think
just
something
that
I
I
wonder,
maybe
what
gn
or
em,
I'm
not
sure
how
to
pronounce
his
name
mantis
like
spring,
is
widely
used,
so
maybe
replace
the
java
service
that
we
have
to
java
service
using
the
spring
framework.
Something
like
that,
but
yeah.
I
think
that
would.
C
C
F
How
plugged
in
are
we
with
the
like
spring
templates
and
open
telemetry
like?
Can
people
basically
just
import
it
the
same
way
they
might
import
a
rest
template
or
would
it
be
more
involved
because
I
feel
like
that
changes.
The
effort
that
you
would
have
to
put
in
to
actually
make
the
demo
out
leave.
C
C
Yeah,
like
there
might
be
a
more
spring,
springy
conventional
way
to
do
all
this,
but
I
don't
know
if
the
spring
community
has
really
built
one
of
those.
Yet
so
my
my
gut
feeling
is
that,
like
changing,
whatever's
there
to
spring
would
probably
be
pretty
straightforward.
It's
just
a
framework
so
anyway,
moving
on.
I
think
we,
I
think,
I'm
actually.
I
think
we've
resolved
this
last
time
about
kubernetes
versus
docker.
Is
that
we're
going
to
do
both
effective,
both
essentially.
B
C
C
There's
no
requirement
to
do
manual
context
prop,
but
I
feel
like
and
maybe
other
people
that
are
kind
of
end
user-facing
can
chime
in
one
way
or
the
other.
I
do
feel
like
it
would
be
useful
if
we
could
figure
out
a
way
to
show
off
like
trace
propagation
between
services,
where
you
don't
have
auto
instrumentation.
A
Super
useful
to
have
context
propagation
in
place,
but
just
for
the
the
ones
that
are
not
supported
out
of
the
box
like
everything
that
we
have
alto
or
with
the
agent,
then
it
doesn't
make
sense,
but
for
technologies
that
we
that
opentow
doesn't
support,
then
it
would
be
super
valid
because
in
the
hotel
docs
we
kind
of
lack
of
context.
Propagation,
real
explanation
on
how
to
so
maybe
having
a
live
example
would
be
useful.
C
D
Have
any
I'm
thinking
that
the
the
feature
flag
service?
I
thought
we
were
going
to
thinking
about
running
it
with
c
plus
plus
and
then
having
the
service
itself
being
erlang
elixir
thing.
What
about
the
context
propagation
from
c
plus
plus
into
early?
Could
that
be
mainly
done?
And
I'm
saying
that
because
I,
I
believe
the
auto
instrumentation
on
the
c
plus
plus
it's
quite
weak.
C
Yeah
my
I
feel
like
the
c
plus
thing
is
really
not
super
well
specked
out
right
now,
but
my
understanding
was
that
the
c
plus
plus
example
was
just
running.
I
thought
I
was
running
the
php,
the
admin
service
with
nginx
or
apache
or
whatever
I
could
be
wrong.
E
Against
it
when
they
were
or
double
check
with
them,
I
know
debash
was
proposing
some
like
apache
web
server
stuff.
I
don't
think
we're
pursuing
that.
C
Kind
of
deliberately
not
doing
auto,
not
not
instrumenting
the
route
so
that
you
can
show
the
manual
way
of
doing
context
prop
or
you're
just
creating
some
weird
novel
like
you're
doing
something
weird
or
you're
doing
something
dumb
like,
and
I
don't
know
what's
the
better
way
to
do
it
like.
C
I
could
certainly
see
novel
sorts
of
like
what
if
there
was
a
like,
if
you
needed
to
pass
context
across
a
pipe
like
if
you
did
some
sort
of
p
invoke
thing
where
you
needed
to
continue
a
trace
into
a
sub-process
spawn
like
that's
an
option,
but
is
that
too
esoteric?
Does
that
actually
like
meet
a
need
like.
E
C
E
All
right,
we
want
to
quickly
review,
so
I
think
we
have
some
service
owners
here.
Do
you
want
to
quickly
review
like
the
traces
expectation
just
so
everyone
is
aware,
I'm
sure,
like
gary's
new
to
the
effort,
yeah.
C
We'll
we'll
just
cover
the
trace
require
the
telemetry
requirements.
So
briefly,
it's
in
order
of
priority.
We
should
traces,
should
be
kind
of
the
primary
thing,
so
auto,
instrumentation
and
or
library
instrumentation
critical
path.
Rpc
should
be
traced,
everything
that
has
trace
instrumentation
like
automatic
or
otherwise
needs
to
extend
those
traces
into
the
business
logic
somehow
and
create
like
custom
spans,
so
not
just
the
grpc
span
or
whatever.
There
should
be
something
where
we're
creating
a
new
span
and
adding
you
know,
service,
attributes
or
events
to
it.
C
C
We
should
be
aware
of
when
trace
links
make
sense
right
now.
The
one
big
example
of
trace
links
is
linking
from
load
gen
to
front
end,
but
if
there's
other
kind
of
like
I
could
see,
maybe
the
feature
flag
stuff
like
that
would
be
maybe
trace.
Things
would
make
sense
there
somehow
or
if
we
have
a
fan
out,
fan
in
or
queue
or
something
we
should
avoid
filtering
at
the.
C
The
preference
would
be
to
push
filtering
out
to
the
collector
rather
than
doing
filtering
at
the
service
level,
and
then
we
should
do
use
baggage,
where
appropriate,
there's
an
open
issue
for
baggage
that
has
some
of
those
requirements
in
it.
Any
questions
about
or
comments
about
sort
of
the
trace
stuff,
either
the
expectation
that
it's
like
trace.
First
or
any
of
the
details
in
here.
D
H
D
But
that's
just
my
sense.
At
the
same
time,
if
a
trace
started
from
a
web
browser-
and
we
actually
instrumented
the
web
front
end,
I
would
be
all
for
that.
Continuing
it
through
just
load
generator
feels
weird.
That's
all.
C
C
C
Do
we
does
anyone
feel
strongly
that
we
should
put
that
in
for
v1,
add
in
browser
instrumentation.
D
C
G
E
E
I
mean
I
should
probably
wait,
then,
because
I
mean
the
load
generator
is
essentially
playing
the
role
of
the
client.
It's
just
doing
a
pretty
bad
job
of
it
and
wouldn't
look
like
you
would
see
it
in
like
a
real
world
scenario,
but
from
like
a
trace
entry
point,
we
could
consider
that
kind
of
like
okay,
you.
C
Yeah
I
mean
and
another
like,
maybe
a
to
look
ahead
a
little
bit.
Maybe
the
real
long-term
story
here
should
be
to
replace
or
to
kind
of
accentuate
the
load
generator
with
something
like
puppeteer
like
once.
There
is
a
better
web
tracing
story
than
using
like
puppeteer
or
something
to
have
actual
like
clickamabobs.
E
Yeah
that'd
be
cool,
there's
actually,
so
playwright
is
actually
a
microsoft
team,
which
is.
C
C
C
C
The
other
option
is
that
we
assume
that
this
will
not
be
run
at
100
sampling,
that
it
would
be
run
at
some
lower
amount
sampling
and
thus
to
create
accurate
statistical
data
about
throughput.
You
know
about
rate
error
duration
that
we
would
need
to
emit
those
as
metrics,
and
that
is
the
question.
C
The
two
two
two
things
one:
it's
a
demo
application,
so
you
know
I
think
saying:
100
sampling
is
fine.
The
other
is
that
you
know
obviously
slightly
biased
because,
like
tool-wise,
you
know
lightstep
does
this,
but
I
think
in
general
the
idea
is
that
if
we're
we're
trying
to
say
like
what
should
observability
be,
it's
like
well
hotel
is
kind
of
trace.
First,
we
kind
of
have
this
opinion
that
traces
are.
You
know
the
best
type
of
telemetry
for
a
distributed
system
to
let
you
see
things
like
this
ergo.
C
You
should
just
use
traces
for
stuff
like
that,
so
it's
sort
of
a
philosophical
like
what
matters.
E
Yeah,
I
think
that
kind
of,
like
austin,
said
we
have
an
issue
on
the
sampling
strategy,
so,
if
anyone's
super
passionate
about
like
kind
of
leading
the
design
of
that,
I
think
that
would
be
great.
I
think
right
now,
like
you
said
it's
always
on
sampling,
and
so
we
don't
have
a
sampler
actually
specified
at
any
of
the
provider
levels
currently
so
that's
kind
of
a
gap,
but
the
the
probability
of
sampler
that's
not
actually
developed.
Yet,
though,
that's
kind
of
a
forward-looking
sampling.
C
Yeah
and
we
could
always
do
like
because
there
is
a
trace
based,
there
is
like
a
tail
sampler
in
the
collector
like
there
are
options
here,
as
always
anything
with
hotel
is,
you
know,
moving
targets.
E
C
C
I
would
assume
apache
or
nginx
or
whatever
other
resourcey
components
we
throw
in
here.
We'll
have
metrics.
D
C
Yeah,
I
mean,
I
think
we
we
have
to
be.
We
have
to
show
both
right.
We
have
to
show
business
logic
metrics
and
we
have
so.
We
have
to
show
the
hotel
metrics
apis.
You
know
in
situ.
We
also
need
to
show
the
like
collector,
scraping
metrics.
C
C
C
C
Right
and
that's
why
to
me
this
is
asper,
it's
aspirational
to
say
like
well
in
the
future
you're
not
going
to
have
to
worry
about
this,
but
I
would
in
my
mind,
it's
easier
to
start
out
by
saying
100
percent
sampling
and
we,
when
we
get
all
these
from
the
trace
data,
then
and
then
when
we
need,
if
we
come,
if
we
come
to,
if
we
decide
later
on,
it's
like
okay,
we
need
to
turn
this
down.
We
need
to
add
sampling
in
or
add
a
flag
for
sampling.
C
Then
we
can
figure
that
out
later,
but
we've
got
quite
a
bit
of
work
on
our
plate
already.
I
think-
and
this
just
removes
a
this-
reduces
the
complexity
of
what
we
need
to
do
ahead
of
us.
E
C
Oh,
anyone
violently
opposed
to
implicit.
D
B
C
C
D
C
C
I
don't
know
if
it's
bugs
or
just
like
jankiness,
but
my
thought
is
that
basically,
this
is
saying
if
you
have
a
service,
if
you
write
like
an
actual
like
service
metric,
that
is
something
you
know
a
business
logic
metric.
Let's
say,
then
that
should
use
hotel
otlp
push
to
the
collector.
But
then,
if
I
have
like.
C
And
there's
all
same
thing
and
with
baggage
same
note
as
there
was
before,
like
there's
a
suggest,
there's
a
baggage
thread
in
the
issues
to
what
needs
to
get
the
baggage
met
or
baggage
attributes.
E
What's
our
default,
visualization
experience
be
for
metrics.
Will
this
be
graffana?
I
think
it's
a
big
grafana.
C
Okay,
I
think
we're
I
think
we're
gonna
have
to
ship,
especially
with
the
open,
slo
stuff
that
we
can
get
to
later,
but
I
think
we
have
to
ship
grafana
dashboards.
C
Logs
are
pretty
underdefined
right
now,
I
feel
like
in
otel,
and
so
I
don't
want
to
spend
a
ton
of
time
on
them.
I
know
like
the
otp
definition
is
what
it
is
but,
and
I
know
net
is
doing
what
net
is
doing.
But
you
know
logs
have
not
been
the
focus
area,
so
let's
try
to
keep
logs
light
and
mostly
limited
to
things
that
you
can't.
C
C
E
C
C
Last
yeah
and
by
the
way,
pierre,
this
is
collectors,
should
report
underlying
resource,
metrics
and
stuff.
So
this
the
collector
would
be
responsible
for
scraping
the
kubernetes,
you
know
or
apache
or
whatever,
and
doing
the
adding
in
the
service
name,
attributes
and
all
that
stuff,
so
that
everything
links
up
nicely.
C
C
C
Feature
flags
that
stuff's
in
progress
and
then
the
open,
slo
folks
are
interested
in
you
getting
in
on
this
and
having
the
web
store
demo
be
a
way
to
kind
of
show
off
open
slo
in
the
wild.
So
hopefully
we
can
get
some
additional
help
from
them
to
define
and
build
that
out.
So
that'll
be
cool.
E
E
C
A
Sure
if
it's
an
architectural
question,
but
I
got
in
a
discussion
with
andrea
hayworth
regarding
ruby
and
jrpc,
so
he
sent
out
a
pr
implementing
the
email
service
in
ruby,
but
he
changed
it.
B
A
A
C
A
Cool
because,
well
I
I
I
know
that
for
php
we're
gonna
go
to
http
because
grpc
simply
it's
just
for
client
side
and
we
need
something
different,
but
yeah.
Okay.
So
thanks.
E
Of
course,
okay,
so
it
looks
like
we
have
10
minutes
left
and
of
course
there
are
some
other
people
on
the
call.
So
we
have
like
51
open
issues
on
github.
I
don't
know
if,
like
you
want
me
to
open
it
and
we
can
go
through
it,
but
like
mikko,
ali
rhys,
we
have
you,
know
stuff
from
documentation
to
like
working
on
the
services.
Do
you
have
any
particular
areas?
Y'all
are
interested
in
not
to
put
you
on
the
spot.
J
Yeah,
I
I
heard
about
this
a
while
ago,
and
this
is
the
first
time
I've
been
able
to
attend.
So
I
just
kind
of
wanted
to
get
my
bearings
and
yeah
we'll
see.
What's
going
on.
E
Feel
free
to
describe
an
issue
if
you
can't
assign
yourself
just
let
me
know
just
at
me
on
github,
but
if
you
see
anything
of
interest
to
you,
you
know
go
ahead
and
take
it
and
run
with
it.
B
I'll
do
carter
as
well
I'll,
look
at
some
of
the
documentation,
some
of
the
issue
lists
and
go
ahead
and
assign
and
move
forward
with
it.
Thank
you
yeah.
Without
the
help.
I
E
Yeah
yeah
definitely
something
to
be
mindful
of
good
call
out
michael
and
then,
of
course,
gary
is
owning
the
rust
item,
which
we're
very
appreciative
of
welcome
gary.
E
But
besides
that,
I
think
we
can
probably
go
ahead
and
end
a
bit
early
are
any
of
y'all
planning
on
being
at
the
open,
telemetry
community
meetup,
I
think
austin.
I
think
you
mentioned
you
are.
F
C
F
Right
anyone
else,
so
the
one
on
the
20th
that
we're
talking
about
yeah,
I
don't
have
to
plan
on
attending
virtually
just
because
I
can't
attend
for
immuno
reasons.
Yeah.
Oh.
C
We'll
try
to
yeah
we'll
figure
I'll,
try.
C
We'll
try
to
do
as
good
of
a
job
as
we
can
on
that,
but
boy
howdy
cncf
loves
making
that
difficult.
D
D
I'll
try
to
attend
virtually,
though
that's
it's
a
honeycomb
day
off
and
we're
pretty
adamant.
I'm
not
doing
work
things
on
days
off,
yeah.
C
It's
like
250.,
it's
like
250.,
although
we
do
have
some,
I
need
to
poke
someone,
but
we
will
have
if
anyone
is
in
the
austin
metro
area
and
does
want
to
come
in
person.
I
think
we
have
like
five
tickets
to
give
away
to,
basically
whatever
I
say
so,.
D
E
B
E
For
joining
rhys
welcome
gary
welcome,
we
appreciate
y'all's
involvement
and
support
and
everyone
else.
I
appreciate
you
all
too
so
have
a
nice
day.
Yeah.