►
From YouTube: 2022-10-13 meeting
Description
cncf-opentelemetry@cncf.io's Personal Meeting Room
A
B
A
I
mean
unless
you've
got
like
an
elf
wire
going
up
to
the
the
surface
or
something.
A
Yeah
I've
got
so
much
of
it
now.
I've
Splunk
keeps
sending
me
hoodies,
even
though
I
don't
work
there
anymore.
So.
D
So
I'm
not
at
home
and
I'm
gonna
have
to
drop
after
20
minutes
so
hoping
somebody
else
can
share
and
drive
the
agenda
today.
D
C
D
Oh
right
yeah,
so
with
chatted,
I,
don't
know
a
couple
of
months,
a
few
months
ago
about
potentially
long-term
the
runtime
attached
and
the
static
instrumentation
living
in
the
instrumentation
repository
as
sort
of
alternative
onboarding
mechanisms
to
Auto
instrumentation.
D
And
so
I
think
we
thought
that
seemed
like
a
good
idea.
I
I
think
that
the
static
instrumentation
is
too
early.
Still
it's
very
experimental,
but
the
runtime
attach
is
fairly
stable.
We
have
customers
a
good
number
of
customers
using
it
now
and
so
I
think
it
could
be
promoted
to
the
instrumentation
repo
and
in
particular
it
would
allow
us
to.
D
This
was
kind
of
a
bad
bug
in
the
prior
release,
where
it
basically
didn't
work
at
all
because
of
a
packaging
issue
and
because
we
have
unit
tests.
But
we
don't
have
any
really
smoke
tests
in
the
Java
computer
repo
and
we
could
build
a
smoke
test
over
here.
But
if
we
did
move
it
to
the
instrumentation
repo,
we
could
potentially
just
leverage
the
existing
smoke
test,
harness.
C
So
I
haven't
looked
at
the
source
code
for
this
runtime
attachment
mechanism.
Does
this
allow
you
to
attach
the
Java
agent
to
a
process
that's
already
running
and
with
without
modifying
the
source
code,
or
do
you
have
to
modify
the
source
code?
As
is
shown
in
this
example?
D
Yeah
yeah,
you
have
to
you,
have
to
add
this
line
of
code
to
the
at
the
very
beginning
of
your
main
method.
So
it
has.
You
know
it's
sort
of
limited
in
scope,
it's
great
for,
like
spring
boot
or
like
a
caucus
or
something
where
you
have
a
main
method.
A
A
I
think
it's
ultimately
useful
if
you
want
to
manage
the
Java
agent
division
by
made
in
their
Gradle
dependency,
because
usually
you
would
have
to
download
the
separate
jar
and
keep
track
of
it
separately.
C
D
That's
true:
we've
seen
customers
kind
of
that
really
like
this
approach
of
adding
it
into
their.
You
know
as
managing
it
as
a
normal
dependency
and
especially
for
spring
boot,
where
people
like
this
having
this
single
jar
that
they
can
throw
anywhere
it
sort
of
bundles
the
agent,
essentially
into
the
that
single
jar.
E
Then
override
the
the
Run
command
this
way,
it's
much
simpler.
C
Yeah-
and
you
have
one
way
to
attach
the
agent
regardless
of
whether
you're
running
in
a
doctorized
environment,
or
you
know,
if
you're
running
just
locally.
B
I
mean
there
are
plenty
of
APM
products
out
there
that
attach
to
running
jvms
without
code
modification
like
dynatrace,
has
been
doing
it
forever
and
it
looks
like
elastic.
Has
that
coming
out
in
the
next
version
of
their
APM
agent
too,
but
I
haven't
seen
any
open
source
really
doing
it.
To
my
knowledge,
yeah.
D
And
just
to
clarify
this
we've
sort
of
limited
the
scope
here
to
attaching
at
the
very
beginning
of
main,
as
opposed
to
attaching
later
on,
because
there's
certain
things
that
we
do
like
say
when
a
grpc
Interceptor
is
created,
we
need
to
instrument
that
creation,
and
so,
if
you
attach
later
on,
it's
sort
of
too
late
for
us
and
we'd
have
to
do
a
lot
of
work
around
some
and
change
a
lot
of
the
way
our
instrumentation
Works.
In
order
to
truly
do
like
anytime,
attach
yeah.
C
Are
there
any
gotchas
associated
with
attaching
the
agent
this
way,
you
know,
assuming
that
you
can
make
this
call
at
the
beginning
of
your
main
method,
or
does
it
work
in
the
exact
same
way,
otherwise,.
D
D
D
B
D
A
A
D
Yeah
and
Jean
is
on
the
call
and
might
be
able
to
speak
a
little
bit
more
to
the
JRE.
The
reason
why
JRE
doesn't
work,
we
think
the
read-only
file
system
might
be
solvable
at
some
point,
but
I
don't
recall
if
the
if
the
jdk
is
like
a
hard
restriction.
A
I,
don't
remember
exactly
but
I
think
it's
of
a
file.
We
need
an
executed
file,
but
this
is
only
on
a
GDK
and
not
on
the
GI.
D
C
I
always
when
I'm
building
dockerized
applications,
I
always
use
the
JRE
versions,
because
they're
significantly
smaller.
E
Well,
with
with
Java
9,
the
notion
of
JRE
is
deprecated,
because
one
is
allowed
to
configure
their
environment
using
this
little
blocks
modules
and
therefore,
whether
it
has
a
full-blown
jdk
or
very
small
runtime,
it's
entirely
in
the
hands
of
the
customer.
E
So
that's
why
the
notion
of
JRE
is
deprecated,
but
of
course
we
don't
know
what
the
customer
is
using.
The
customer
can
have
Java
17
but
trimmed
to
the
bare
minimum,
and
we
might
need
some
to
use
something
which
is
not
there.
So
the
dependency
is
still
there.
It
isn't
just
not
using
the
term
JRE
anymore.
A
B
We
have
talked
to
several
security
companies
that
would
like
to
basically
get
run
time
lists
of
what's
actually
being
used
in
an
application
versus
what
is
in
the
jar
files
of
the
application,
which
a
lot
of
tools
do,
and
it's
kind
of
outside
of
the
observability
use
case
directly
so
like
when
you
kind
of
look
at
the
point
of
open
Telemetry,
it's
about
observability
and
less
about
security
use
cases,
but
the
idea
would
be
you
know.
Is
this
something
that's
possible?
B
What
are
the
thoughts
of
those
of
you
on
the
line?
Obviously,
it's
possible
to
program
it,
but
is
it
outside
of
the
scope
of
what
the
project
is
meant
to
do.
A
I
mean
I,
think
that-
and
this
is
an
observability
of
I
mean
of
of
a
different
sort.
I
guess
I
would
think
that
if
you
know
how
to
build
this
tool
and
whether
it's
relev
like
is
it
really
relevant
for
open
Telemetry,
like
I,
mean
I,
guess
I'm
not
sure
like.
Could
you
just
build
an
agent
that
you
know
intercepts
a
class
loader
and
you
guys
get
this
information
like?
Do
you
need
that
is
there?
Is
it
really
at
all
related
to
open,
Telemetry,
I?
A
C
And
you
could
use,
as
you
suggested
here,
you
can
use
otlp
to
transport
that
I.
C
B
I
would
assume
that
we
would
just
send
it
as
a
log.
You
know
generically,
but
going
back
to
what
John
was
saying.
The
the
purpose
of
doing
this
and
open
Telemetry
would
be
add
an
additional
set
of
use
cases
around
security
and
visibility
into
the
existing
instrumentation
framework
versus
going
and
installing
something
else.
B
B
These
kind
of
static
code
analysis
tools
that
want
to
understand,
what's
actually
used
versus.
What's
in
a
repo,
for
example,.
A
So
like
what
what
we,
what
would
be
the
actual
output
of
it,
because
the
your
comment
mentions
libraries
and
Method
names,
but,
for
example,
if
we
call
like
hibernate
session,
do
you
would
you
expect
the
agent
to
list
out
that
it's
hibernate
Dash,
core
version,
3.
I.
B
B
A
A
Mean
capturing
like
the
class
names
would
be
pretty
easy
method.
Names
would
be
probably
a
bit
more
difficult,
but
figuring
out,
which
version
it
actually.
Is
this
going
to
be
a
challenge?
Yes,
for.
B
B
A
I
think
that
perhaps
the
like
POC
version
of
it,
for
example,
just
exporting
the
class
names
or
methods
and
or
method
names,
could
be
implemented
as
a
division
expansion.
And
maybe
we
could
add
that
to
a
country,
Bravo
yeah,.
B
So
that's
why
I
kind
of
brought
it
up
was
potentially
the
way
we
could
do
it
in
a
way
that
allows
everyone
to
solve
this
type
of
problem,
because
what
you
have
now
is
some
of
the
proprietary
tools,
whether
it's
datadog
or
dynatrace,
or
app
Dynamics,
building
this
functionality
in
their
proprietary
agents
today.
To
do
this,
you
know
which
is
kind
of
part
of
the
rest.
The
runtime
application,
security
and
protection
use
cases
that
they've
been
building
around.
B
D
Yeah
this
this
use
case
has
come
up.
I've
seen
this
use
case
come
up
for
us
also
I'm,
very
supportive
of
trying
to
standardize-
and
you
know,
like
material
says
it's
hard
to
get
the
version
you
or
you
can't
get
the
version
reliably,
but
right.
D
You
know
we
could
do
a
best
effort,
there's
various
heuristics
that
sometimes
work
to
get
the
version
either.
So
pretty
sad
yeah,
and
even
if
we
yeah
the
class
names
or
package
names,
I
was
trying
to
think
of
something
to
make
it
not
as
noisy
yeah
like
every
class
but
and
I
I
think
I
was
mistaking.
When
I
first
read
your
comment
here,
I
was
thinking.
You
wanted
the
instrument,
every
single
method
and
capture
every
single
method,
but
just
capturing
what
classes
were
actually
loaded
at
runtime.
C
All
right,
thanks,
Jonah!
The
next
item
on
here
is
about
the
jmx
metric
inside
PR,
which
Peter
has
been.
E
Working
on
yeah,
so
since
our
last
meeting
I
changed
the
schema,
the
Yama
schema,
so
previously,
I
was
using
the
term
label
for
the
metric
attribute,
which
was
not
working
well
with
with
the
standard
terminology.
E
So
the
changes
only
in
the
documentation
in
the
readme
file
as
I
did
not
change
the
code.
Yet
I
would
like
to
get
the
group's
blessing
for
for
this
change
before
I
go
ahead,
so
not
to
repeat
the
the
work
several
times
if,
if
necessary,
and
also
I
would
like
to
know
what
what
are
the
next
steps
here?
E
What
what
the
group
is
thinking,
whether
if
we
do
this,
will
be
there
any
extra
tasks
to
be
performed
before
approval
and
merging
I
I
know
it's
a
lot
of
code,
there's
3000
lines,
so
it's
really
hard
to
to
review
and
if
I
can
help
in
any
way.
With
with
reviewing,
let
me
know
I'll
be
glad
to
to
assist
with
whatever
questions
you
might
have.
D
Yeah
sorry
for
I
I
was
playing
around
with
the
schema
a
little
bit
Peter
at
the
and
I.
Think
to
me,
that's
the
that's.
D
The
only
thing
really
that
we
have
to
resolve
is
the
skiba
and
then
I
think
everything
else
you
know
look
looks
good
to
me,
the
one
so
the
one
I
thought
I
had
that
I
haven't
had
a
chance
to
play
enough
with,
is
not
having
the
exploring
what
it
would
look
like
if
the
schema
had
static
like
left-hand
side
values
for
everything
so
kind
of
like
turning
those
Maps
of
things
like
where
the
left
hand
side
is
attributing
into
like
lists,
yeah
and
seeing
what
that
looks
like
I'm,
not
convinced,
that's,
better
or
not
I,
just
it
was
something
I
wanted
to
explore
before
before
you
go
further.
E
That,
if
we
reverse
the
order,
the
key
would
have
to
be
still
unique
right.
So
you
want
the
value
to
be
the
key
in
the
map
and
the
value
will
be
the
the.
D
More,
like
so
say
like
the,
if
you
take
the
example
of
the
attributes,
where
it's
just
kind
of
a
map,
key
value
pairs.
A
E
E
D
I
see
what
you're
saying,
let's
see,
yeah
and
that's
what
I
was
kind
of
I
started
and
was
going
to
force
myself
to
try
it
with
all
these
examples
that
you
have
in
here
already
to
get
the
feel
for,
for
that,
hey,
I'm,
really
sorry,
I
have
to
dropped
early,
but
I
will
I
will
look
more
at
this
I
promise,
but
maybe.
D
If
y'all
want
to
continue
discussing,
if
anybody
else
all.
C
All
right,
well,
Trask,
has
been
the
main
person
that
has
had
feedback
for
this
PR.
From
my
perspective,
it
looks
reasonable.
You
made
the
changes
to
consolidate
all
the
into
a
single
Library
module
in
a
single
Java
agent
module.
So
that
looks
good,
so
I
don't
have
anything
else.
To
add
to
this
I
think
we'll
just
wait
for
trasky's
trask's
feedback,
offline.
C
C
All
right
well
congrats,
on
the
Java
agent
1.19
release,
looking
forward
to
the
contribute
release
in
a
couple
of
days
or
tomorrow,
I
think
so
and
then
we'll
be
have
1.19
released
across
the
board,
so
great
work.
Everyone.