►
From YouTube: 2022-06-08 meeting
Description
cncf-opentelemetry@cncf.io's Personal Meeting Room
A
D
Yeah,
so
I
can
talk
about
it.
It's
a
work
in
progress.
I've
been
working
on
it
in
between
things,
but
basically,
what
I'm
doing
is
I'm
trying
to
follow
what
was
done
in
the
otlp
exporter
as
far
as
grabbing
the
the
open,
telemetry
protos,
and
so
following
that
model,
as
opposed
to
trying
to
reach
into
the
exporter
library,
to
try
to
extract
the
the
types
from
there
because
they're
marked
as
internal
so
just
simply
grabbing
the
source
of
those
protos.
D
I
grabbed
version
18,
which
I
think
was
included
in
the
last
release
of
the
exporter
and
from
there
I'm
trying
to
add
the
some
of
the
test
framework
bits
so
adding
a
listener
that
can
then
decode
requests
from
those
protos.
D
I
have
yet
to
write
a
test,
that's
using
it.
So
I
don't
quite
know
that
the
right
shape
is
there
yet,
but
that's
kind
of
the
the
work
in
progress
aspect
of
it
so
still
trying
to
keep
it
similar
to
what
we're
doing
with
the
mock
zipkin
stuff
that
we
have.
But
this
time
using
the
protos.
D
D
Yeah
good
to
know,
probably
the
the
the
questions
that
I
have.
The
most
is
what's
the
right
level
of
abstraction
for
the
filtering
that
we
wanna
do
so
with
the
zipkin
stuff.
We
have
the
filters
that
allow
you
to
filter
directly
on
the
spams,
but
with
otlp.
The
data
is
a
bit
more
hierarchical,
so
you
got
the
resources.
You've
got
the
instrumentation
scopes
and
then
you've
finally
got
the
metrics
and
the
data
points,
and
so
at
this
point
I've
exposed
the
the
top
level
thing.
B
Yeah,
so
so
that's
a
that's
a
very
good
point,
because
where
we
want
to
draw
that
line
kind
of
in
the
filtering
for
tests,
you
know
yeah,
that's
a
very
interesting,
but
I
I
think
we
even
have
a
issue
open
that
eventually
you
wanna
kind
of
the
default
be
otlp
and
the
other
ones
with
just
smoke
test.
E
Not
sure
chris,
if
it
will
be
helpful,
but
I
remember
that
in
sdk
there
is
something
like
in
memory
in
memory
matrix
exporter
or
something
like
that,
and
it
has
some
some
stuff
which
it
exposes.
D
Yeah,
when
I
looked
at
the
tests
in
that
repo
they're
really
they're
not
doing
the
same
type
of
filtering
that
that
we
have
been
doing
in
our
tests,
and
so
they
are
walking
through
from
the
resources
on
down
and
validating
what
they
need.
D
So
that
that's
where
it's
at
you
can
feel
free
to
take
a
look
but
yeah.
It's
very
much
a
work
in
progress.
B
Yeah
and
take
a
look
at
the
link
that
piotr
posted
if
it
helps
somehow.
D
A
A
But
but
based
on
the
one
of
the
issue,
paul
created
the
new
task
to
to
the
regression
test,
which
are
using
dots
net
core
3.1.
B
Yeah
dotnet
core
2.1
is
going
out
in
the
end
of
the
year,
but
whatever
you
support
image,
we
we
need
to
do
some
smoke
tests
to
ensure
that
at
least
the
basic
stuff
is
working.
We
miss
that
so.
A
I
think
that
we
have,
in
the
backlog,
a
bit
broader
task
to
execute
all
our
tests
based
on
the
dosnet
core
matrix,
which
is
supported,
and
it
should
be
done
across
all.
B
Yeah,
but
but
just
to
to
be
clear,
because
this
one
the
way
that
we
do
the
integration
tests,
we
run
them
on
the
vm
and
we
started
docker
with
the
whatever,
whatever
kind
of
back
end
we
are
talking
to
this
swing.
Just
came
about
kind
of
when
I
tried
to
run
in
the
docker
image.
F
So
I
was
out
for
last,
like
one
and
a
half
week
today,
just
started
back
so
I'll,
be
looking
through
these
tasks,
revisiting
things
so
no
new
updates
as
such.
So
I'm
just
following
up
the
pr
that
got
merged
in
the
past
week.
So
the
first
thing
I
will
be
I'll
be
working
on
the
documentation
update
for
the
metrics
for
the
usage.
F
If
that
is
missing-
and
there
is
a
pending
task
for
me
to
create
a
like
issue
in
open,
telemetry,
sdk
repo,
so
I'll
be
getting
that
done
today
and
all
of
the
other
tasks
the
work
is
complete.
The
only
the
doc
update
is
pending
so
I'll
get
that
done.
C
A
C
E
A
Yes
and
this,
this
pr
is,
in
fact
the
proposal
how
how
it
should
look
like
how
it
should
look
like,
but
still
it
needs
a
lot
of
well,
probably.
E
F
B
This
one
yeah,
so
what's
the
status
of
the
pro
meteos
exporter
for
for
metrics,
oh
on
the
sdk.
F
B
That
question
I
I'm
asking
about
the
status
of
the
pro
mutual's
exporter
on
the
sdk,
the
version
of
that
of
the
package
that
we
are
using
right
now
we
are
on
the
latest
right,
but
yeah.
What's
this
start
with
that?
That's
working
because
we
we
definitely
should
add
the
the
promuters
export
right
now
we
we
don't
have
that.
B
B
D
I
mean
people,
don't
necessarily
have
to
use
the
collector,
I
mean.
There's
I
mean
many
other
things
support
that
protocol,
which
is
kind
of
nice.
D
Oh
yeah,
for
that
prometheus
support,
I
mean
that
that
probably
makes
sense
for
zero
two
one,
as
opposed
to
trying
to
squeeze
it
into
this
next
release
is
my
guess:.
B
We
moved
the
date
last
meeting
to
or
the
the
meeting
before
I
don't
remember,
but
we
moved
the
day
to
21st.
I
think
we
should
check
what
we
want
to
be
for
for
that
date
to
be
sure
that
we're
gonna
make
it.
E
D
D
Yeah,
I
I
will
be
on
leave
for
a
few
days
between
now
and
the
21st.
B
Okay,
I
next
week
I
have
a
splunk
conference,
so
I'm
not
sure
how
it's
going
to
be
my
time
next
week,
but
yeah
just
a
heads
up.
Maybe
I
don't
show
up.
F
C
E
A
E
B
Yeah
and
robert
these
two
metrics
that
you
have
assigned
you,
you
think
you
can
take
care
of
them
right.
E
You
just
which
of
them,
oh.
E
B
Okay
yeah,
so
let's
keep
in
touch
via
slack
and
try
to
see
if
any
one
of
us
can
take
and
help
raj.
Also,
if,
if
appropriate,
we
can
break
down
some
of
this
stuff
in
smaller
steps.
So
then
more
stuff
can
be
shared.
B
B
B
Okay,
so
I
think
you
can
move
to
the
next
step.
What.
D
A
B
Yes,
so
there
is
a
slack
channel
that
is
started
for
folks
that
are
interested
on
that
the
the
thing
it
seems
that
now
the
natural
place
to
start
is
about
the
the
thing
that
priority
was
mentioned
about
how
to
send
and
code
the
signal
the
there
was
somebody
from
people
present
at
the
meeting,
so
they
that
is
the
start
point
probably
for
people
to
experiment
and
do
stuff,
and
the
other
big
thing
that
was
mentioned
kind
of
initially
is
how
to
correlate
the
signals
you
know,
so
you
wanted
the
stack
somehow
to
be
correlated
to
spends
and
traces
logs
of
that
so,
but
is
it
still
very
open,
so
anyone
that
wanted
to
have
some
participation
there
is
very
in
the
beginning.
B
D
Yeah,
the
correlation
concept,
I
think,
is
an
interesting
one,
but
I
don't
really
have
any
good
ideas
there.
I
just
thinking
about
how
this
profiling
has
been
done.
In.Net
in
the
past,
it
just
seems
so
disconnected
from
from
a
lot
of
the
things
going
on
in
the
runtime.
B
Yeah
but
but
if
I
recall
correctly,
they
have
that
concept
of
activity
id
that
migrated
in
a
sense
for
for
dotnet.
If
we
have
the
activity
context
now
kind
of.
B
D
A
D
Maybe
maybe
yeah,
I
think
it
all
depends
on
how
you're
doing
that
stack
sampling,
because,
depending
on
how
you
do
it
you,
you
have
to
actually
stop
all
of
the
runtime
threads
to
then
be
able
to
to
grab
the
stacks,
and
at
that
point
I
don't
know
if
you
have
the
ability
to
to
make
the
call
to
to
to
get
that
context,
I
mean
it's
just
a
lot
of
unknowns
to
me.
B
I
I
I
had
a
a
thread
about
this
with
noah
some
time
ago,
but
it
was
in
the
context
of
the
signal
effects,
profiling
and
basically
I
was
asking
for
api
to
be
able
to
be
notified
when
the
active
contacts
change.
B
So
then,
in
our
case,
we
call
back
the
native
and
update
a
table
that
the
profiler
then
tracks
that
right.
G
Yeah,
that's
we're
doing
something
very
similar
or
trying
to
do
trying
to
achieve
that
coalition
with
a
async
local
callback.
B
Yeah,
so
for
dotnet,
seven,
there
is
a.
There
is
a
oh
with
this
api
that
you
are
notified
about
changes
for
the
activity,
the
current
active,
so
you
can
update
the
profiler.
So
this
is
one
path
for
that,
but
I
I
think
the
the
the
big
question
that
chris
is
touching
is
the
falling.
B
This
stuff
happens
a
lot
on
the
dot
network.
Time
kind
of,
I
think
the
ideal
scenario
in
the
future
is
if
we
could
use
just
the
runtime
to
get
the
samples
they're
called
stacks,
but
then
we
need
to
make
the
the
the
code
there
on
the
runtime
aware
of
that
concept,
to
report
to
us,
you
know
and
right
now,
these
tags
are
not
aware
of
that
kind
of
content,
which
reminds
me
about
another
thing
that,
but
this
is
completely
dot-net
stuff
a
long
time
ago
and
nobody
blocked
it.
B
It
was
just
a
matter
of
things
that
didn't
happen,
but
the
idea
of
kind
of,
perhaps
separating
the
stuff
that
does
the
profiling,
like
the
the
trace
that
exists
nowadays
from
the
runtime.
You
know
because
of
cycles
and
deployment,
and
things
like
this,
I
I
missed
cycles
deployment
cycles
release
time.
So
then
you
could
release
the
runtime
and
a
little
bit
later
release
the
updated
profiler
that
you
do,
the
the
the
sampling
and
whatever,
but
that
completely
separate
conversation.
D
Yeah
with
profiling,
it
it'll
be
interesting
to
see
if
the
main
solution
for
open
telemetry
will
be
using
this
project
or,
if
it'll
be
yet
a
separate
thing
that
open
telemetry
is
going
to
be
integrating
with
ford.net.
D
D
Yeah,
it's
just
an
interesting
thought
about
the
longevity
of
the
the
profiling
apis
that
we
rely
on
for
the
the
bytecode
instrumentation.