►
From YouTube: Open Telemetry's Personal Meeting Room
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).
B
B
B
For
minutes
after
time
as
we
can
start
does.
E
Some
issue
is
that
in
I
see
like
let's
say
the
two
like:
if,
if,
if,
if
a
view
is
filtering
on,
let's
say
on
the
attributes
like
let's
say:
one
has
a
filtering
like
at
the.
If
the
resulting
attributes
at
is
the
same
like
two
two
different
views
with
different
attribute
filtering
and
then
the
resulting
one:
it's
it's
not
counted
properly
for
the
asynchronous
counters
it.
This
appears
to
be
an
issue,
for
you
know
many
other
languages.
B
Okay,
the
are
you
talking
about
these
comments
about
from
Riley
here,
the
python
implementation.
Someone
spoke
with
me
this
one.
E
D
There's
an
example
in
the
original
question
that
shows
the
the
missing
Behavior,
but
I
think
this
example
is
actually
with
a
gauge,
but
the
the
problem
still
seems
to
exist.
It's
it's
outputting
to
like
two
time
series.
Instead
of
a
single
one,
taking
the
last
value.
E
I
yeah
I
I
created
an
issue,
but
I
did
not
actually
run
with
it
as
encounter
with
up
double
counter.
I
can
guess:
I
can
go
back
and
try
it
out.
Yeah
I
can
I
can
verify
like
which
one
is
which
instruments
have
this
issue
and
which
don't.
E
The
this
is
there's
an
issue
like
the
first
issue
in
the
agenda
is
like
in
our
repo
that
I
created
just
just
to
make
sure
that
we
have
some
issue
in
tracker,
but
yeah
I'll
go
back
and
actually
verify.
What's
the
which
issues
are
there.
B
Oh
okay,
okay,
all
right,
I
guess
we'll
have
to
look
into
this.
D
I
was
surprised
that
this
was
the
case.
I
thought
we
handled
this
correctly,
so
it
might
just
be
a
special
case
for
the
empty
attribute
set
I'm,
not
sure,
but
yeah.
B
F
C
B
Have
to
look
at
it,
I've
been
multiplication,
that's
exponential,
histogram
stuff.
So
all
right,
I'll
take
a
look
at
a
second
all
right.
Let's
see
Next
Issue
says
here:
resource
detection
for
container
properties.
A
Yeah
I
have
added
this
one,
so
basically,
this
issue
is
related
to
two.
You
know
detecting
the
container
ID
in
which
the
instrumented
application
is
running
so
that
we
can
correlate
with
the
infrastructure
instrumentation
if
any
issue
Concepts
So.
Currently
this
particular
feature
is
only
available
in
you
know,
AWS
extension,
so
we
are
planning
to
make
it
available
to
all
the
supported
Frameworks.
A
B
Okay,
what
do
you
refer
with
container
here
in
this
context.
A
So
container
is
any
pod
or
any
Docker
container
if
the
application
is
running
inside
that
so
in
general
I
mean
so
every
container
has
this
ID
or
name
right
so
which
we
can
we
can
just.
If
there
is
an
issue
we
can
just
narrow
it
down
to
a
particular
container
or
the
power
it
is
running
on,
so
that
can
be
possible
if
you
can
pass.
You
know
the
container
ID
or
the
name
or
the
Pod
idea,
name
as
a
resource
attribute
or
a
span
integral
sorry.
A
So
so
that's
why
we
are.
You
know
trying
to
I
mean
we're
discussing
to
include
whether
this
feature
is
feasible.
With
you
know,
in
general,
I
mean
I
mean
out
of
the
box.
D
Oh
so
there's
actually
I
was
just
going
to
say
for
for
kubernetes,
specifically
there's
a
separate
set
of
conventions
and
there's
actually
a
spec
for
how
to
implement
the
resource
detector.
For
that
I
think
this
one
is
like
General
container,
which
would
be
like
I'm
starting
a
Docker
container
on
a
VM
or
something
like
that.
A
So
this
particular
feature
is
already
available
in,
as
mentioned,
is
java.js.net
and
go
so
it's.
It
will
be
great.
If
we
can,
you
know
digital
feature,
parity
with
in
Python
engine.
E
Yeah
I
think
we
can
import
the
similar
logic
here
we
already
have.
Like
you
know,
in
contributor,
we
have
different
propagators
different
like
for
something
that
does
not
fit
in
the
main
repo
we
have
like
proper.
We
can
start
adding
the
third
party
resource
detector
as
well.
Where
that
does
not,
we
will
I
think
yeah.
We
only
have
the
EnV
detectors
again,
we
detector
in
the
main
and
then
rest
of
the
components
can
go
in
the
country.
E
A
So
so,
for
now,
only
AWS
instrumentation
I
mean
which
is
available,
as
extension
in
the
country
report.
That
only
has
this
resource
detector
for
ec2
and
eks.
So
I
thought
we
can
make
it.
You
know
more
available
to
I
mean
we
can
make
it
available
in
all
the
frame.
All
supported
Frameworks,
not
a
particular,
not
limited
to
a
particular
extension
yeah.
E
I
think
I
think
you
can
start
a
new
detector
like
you
can
start
the
new
detector
package
and
you
know
put
in
the
logic
there.
That's
you
know
more
generally
applicable
just
like
how
other
other
click
country
packages
are
doing
like,
for
example.
Here
is
our
job.
A
A
E
It
will
still
be
in
the
country
breaker,
but
I'm
saying
it
as
a
separate
package.
Not
not
something
like
you
know
something
that's
already
baked
into
the
main
repo
it
will.
It
will
be
subject,
like
you
know,
a
container
detector
or
something
like
that.
E
Okay,
I,
don't
know
I'm,
just
suggesting,
like
you
know,
just
like
how
we
have
the
different
instrumentation
libraries
like
a
different
third
party
web
Frameworks.
Similarly,
we
will
have
detectors
like
resource
detectors
each
for
you
know,
kubernetes,
there's
a
subject:
detector
containers
there
is
a
software
detector.
You.
C
C
B
All
right,
I
agree:
let's
see
next
topic,
I
kind
of
think
this
one
is
yours.
Sorry.
E
Before
we
move
and
I
I
have
one,
let
me
add
to
the
notes
so
they're
they're
two:
there
were
two
issues
related
about
the.
Let
me,
let
me
add
them
to
the
let's
see.
Let's
talk
about
it.
E
Yeah
these
two
issues,
so
the
overall
like
just
of
it,
is
that
we
they
can.
They
can
provide
the
so
the
Tracer
provider
accepts
the
list
of
arguments
like
it
takes
the
resource.
It
takes
this
active
plan.
E
Processor
right,
the
active
span
processor
should
be
of
the
kind
like
you
know,
synchronous,
multi-processor
or
concurrent
multi-span
processor,
but
the
these
people
were
directly
using
the
simple
span,
processor
or
bad
spun
processor,
and
they
have
the
different
like
different
interface,
like
there's,
no
ad
span,
Plus
or
method
on
the
simplest
and
processor
app
that
span
processor.
E
So
the
the
that's
a
typing
issue
like
they
they're
running
into
and
the
other
user
also
like
where
they
mentioned
like
if
they
try
to
let's
say
if
they
try
to
pass
it
as
a
active
spanning
processor,
as
instance
of
the
batch
pen
like
our
simple
span
processor,
and
then
they
when,
if
they're
going
to
let's
say,
get
Tracer
provider.adspan
processor,
it
will
like.
There
is
another
issue
that
I
added.
Next
to
that,
you
can
open
that,
like
it's
going
to
error,
give
an
error.
E
No,
we,
we
are
not
okay,
so
you
could
say
we
are
missing,
but
it
does
not
make
sense
to
add
this
on
the
balance
span:
processor,
yeah,
so,
okay,
so
the
Tracer
provider.
Ideally
you
know
all
examples.
How
we
do
it
is.
We
only
provide
the
resource
and
then
the
in
the
next
line.
What
we
do
is
get
Tracer
provider.
We
call
on
that
method
and
then
call
the
ad
span
cluster
right
instead
of
doing
that.
E
If
they
pass
the
badge
plan
processor
as
an
argument
in
the
Tracer
provided
Constructor
itself,
then
let's
say
at
the
later
point:
if
they
try
to
do
the
getracer
provided.adstan
processor,
it's
going
to
give
the
error,
because
that
method
is
not
there
and
the
simple
span,
processor
or
best
pen,
Plus.
E
Yeah
it
it
internally
again
calls
to
the
this
side,
like
Tracer,
provide
again
calls
to
the
multi-span
processor,
but
that's
not
there.
Okay,.
E
If
they,
if
they
don't
actually
pass
it,
what
we
do
is
we
create
a
synchronous,
multi-stand
processor.
This
multi-span
processor
can
have
number
of
batch
and
processor,
but
if
they
are
passing
the
as
a
batch
plan
processor
to
the
Tracer
provider
itself,
then
it's
going
to
give
an
error.
B
E
E
E
E
So
the
the
first
issue
that
the
user
shared
like.
If
there
is,
they
bring
up
the
specification
and
then
talk
about
like
why
it
should
not
be
there.
But
the
specification
is
not
very
strict
about
like
what
additional
methods
we
can
have,
and
you
know
I
think
if,
if
we
yeah
I'm
not
sure
what
solution
to
take
like
how
to
address
this,
but
yeah
I'll,
give
it
a
thought.
D
So
in
the
sorry
I'm
trying
to
find
the
code,
the
actual
parameter
that
the
SDK
Constructor
takes
for
the
span
processor.
Is
it
requiring
to
be
one
of
those
multi-order.
E
Yeah
yeah,
it's
just
the
span
processor,
it
does
not
explicitly
say
it
should
be:
multi-span,
Pros,
multi,
multi
kind,
because
these
multi
kinds
multi,
like
synchronous
multi,
are
you
know,
concurrent
multi
are
based
on
the
span
processor
interface
only
so
it
does
not
explicitly
say
it
should
be
one
of
them.
It
just
says
the
span
process
that
that's
where
the
issue
right,
like
they're,
expected
to
be
working
like
span.
If,
as
long
as
the
span
processor
interface
is
satisfied,
they
expect
it
to
be
working,
but
that's
not
the
yes.
D
D
D
E
D
Yeah
I
think
that's
the
first
issue
and
then,
if
this
is
the
case,
then
the
whole
then
actually
the
code
is
correct.
C
E
D
E
D
I,
don't
know
if
it's
anything
in
respect
specifically
but
I
think
a
lot
of
people
use
this
composite
multi-span
processor
approach.
A
lot
of
the
other
sigs
I'd
have
to
look
in
the
spec
more
to
see.
C
D
I
don't
know
if
this
was
mentioned
anywhere,
but
another
fix
would
be
add
to
this
Union
type
just
like
or
replace
this
Union
type,
which
is
span
processor.
If
the
past
or's,
Bend
processor,
isn't
a
multi-span
processor
doesn't
have
that
add,
span
processor
method,
then
we
could
wrap
it
with
one
of
with
probably
synchronous,
which
should
be
a
no-op
for
most
cases
right.
E
E
Yeah
I
I
I
kind
of
agree,
but
do
we?
Is
it
really
restricting
to
say
that
this
multi-span
processor?
It
has
to
be
one
of
these
multi-span
processor,
because
I
think
they
I
kind
of
see
where
they
are
coming
from?
They
do
not
have
to
necessarily
add
the
like,
create
the
instance
of
the
multi-span
processor
and
then
call
AD
spam
process
run
that
to
initialize
badge
and
then
pass
it
as
an
argument
here.
D
Yeah
I
mean
you
just
leave
it
as
it
is
and
tell
them
to
just
wrap
it
with
the
synchronous,
multi-span
processor.
E
Yeah,
which
is
what
like
we
are
currently
doing
like
it's
just
the
empty
one
but
yeah
we
could
wrap
it.
That
makes
sense.
It's
also
not
really
it's
still
a
Backward
Compatible,
not
breaking
yeah,
I.
Think
yeah
I
think
that's
a
good
solution.
Let's
see,
we
will
try
to
negative
Dayton
has
any
other
things
to
say,
and
otherwise
we
will
we'll
add
something.
Yeah.
D
E
D
Right
right
and
then
so
in
in
both
in
in
the
latter
case,
the
the
issue
30
19,
that
the
poster
just
didn't
read
the
type
annotations
there
and
then
yeah.
E
D
E
Yeah
this
this
yeah,
this
I
think
for
agree
with
that
point.
But
does
anyone
else
have
any
you
can.
C
D
Hey
all
right,
I
just
said:
will
you
respond
to
those
issues.
B
Sorry
all
right,
Next
Issue
yeah
this
one
right.
Second,.
E
Yeah,
this
is
already
much
I
think
they
didn't
added
this
I'm
not
sure
like
to
get
the
reviews,
maybe
but
yeah,
okay,.
B
Three
eleven
great
job.
Second,
thank
you
all
right.
Let's
see
that
here
missing,
metrics
yep,
thank
you
for
your
reviews
as
well,
and
finally,.
C
F
F
So
basically
the
code
I
mean
the
injection
does
work.
Where
is
okay?
There
we
go,
but
there
is.
There
are
certain
situations
where
it
causes
the
circular
dependency
that
I've
never
seen
before.
So
because
of
how
entry
points,
work
and
entry
points
are
like
the
mechanism
by
which
we
like
I,
don't
instrument
everything
it
can.
It
can
create
a
circular
dependency
depending
on
how
the
user's
package
is
organized.
F
I've
found
so
in
my
testing.
I
didn't
run
into
this
issue
just
because
I
was
like
using
a
super
simple
package
that
basically
just
had
this
Factory
that
went
into
an
existing
sampler,
but
what
I
found
is,
let's
say
we
have
so
the
sampler
defaulting
logic
is
is
currently
in
SDK.
You
know
open
telemetry.sdk.trace.sampler,.
F
F
What
I
found
is
that,
because
this
is
in
Trace
anytime,
you
import
anytime,
you
do
entry
point
dot
load
in
sampler
and
you
load
a
module
that
itself
references
something
in
the
trace
package.
It
causes
a
like
runtime
circular
dependency.
So
even
though
open
Telemetry
SDK
doesn't
have
a
circular
dependency
and
the
user
package
doesn't
have
a
certain
dependency,
they
just
rely
on
something
like
readable
spam.
A
F
Example,
I
found
there
is
no
circular
dependency
until
you
actually
load
this
entry
point
and
then
it
seems
to
python
seems
to
when
does
entry
point
outload.
It
seems
to
reincorporate
the
loaded
module
as
a
dependency
of
the
current
module
so
anytime.
The
sampler
is
loading.
This.
It's
basically
saying
this
new
sampler
and
its
package
are
depend.
A
F
Dependencies
of
the
sampler
module,
which
is
kind
of
tricky,
because
if
if
the
sampler
is
for
something
adjacent
to
it,
is
relying
on
anything
in
the
sampler
or
it
seems
the
trace
module
at
large,
then
it
causes
a
circular
dependency.
F
I
didn't
I
mean
I
I,
like
I'm
gonna.
Do
the
entry
points
in
general,
but
no
run
to
a
circular
dependency
like
this
and
so
I'm
gonna
I'm
gonna.
Rework
this
to
see
if
I
can
move
the
custom
sampler
injection
to
configuration,
and
unfortunately
that
would
mean
that
you
can't
use
custom
Samplers
for
manual
instrumentation.
F
You
can
only
use
it
for
for
auto
instrumentation,
but
that's
the
only
way
I
can
think
of
that
where
we
would
keep
somebody
where
we
would
keep
the
existing
functionality
and
be
able
to
add
this
custom
sampler
functionality.
So
that's
that's
kind
of
a
very
big
overview
but
I'm
curious.
If
anyone
has
any
thoughts
and
if
they've
encountered
something
like
this.
B
C
F
F
See
if
I
can
so
yeah,
it's
really
tricky
in
is
in
the
standard
module.
So
this
is
ultimately,
let's
see
if
I
can.
Oh
boy,
there's
there's
a
lot
of
lines,
but
it
took
me
a
while
to
trace
out
everything
that
was
happening.
Yeah.
C
F
C
F
The
API,
it's
just
once
the
the
mod
the
entry
point
is
loaded
at
runtime-
is
that
a
circular
dependency
is
created.
So
it's
very
it's
it's
very
tricky.
B
Well:
okay,
first
things:
first
amazing,
debugging!
Okay,
thank
you
so
much
and
how
hard
must
that
have
been
yeah?
Well,
I
have
no
idea.
Aaron
has
raised
so
go.
D
B
D
Jeremy
I
was
gonna
ask
so
you
ran
into
this
with
readable
span,
specifically
right,
I'm
guessing
this
was
a
type
annotation.
Yes,.
F
F
D
B
C
F
I
think
I
think
it
is
only
being
used
as
a
type
annotation
or
anything.
It's
called
the
get
used,
yeah
yeah.
B
I
think
the
type
checking
in
in
all
caps
that
still
work
around.
F
Okay,
so
I
I
I
did
I
have
sometimes
seen
where,
where
people
are
using
like
this
I,
don't
know
if
they
do
I
guess
they
would
do
quotations,
I
know
I,
guess
we
would
do
quotations
around
here
that
that
this
might.
This
might
I'm
not
sure
this
will
solve
the
problem,
because
I
think
there's
another
type
that
if
it
didn't
run
into
circular
pregnancy
on
this
type,
we
might
have
done
another
one,
but
that
also
might
be
a
type
checking
one.
F
D
Import
this
yeah
yeah,
so
so
the
caveat
here
is,
if
you're
using
it,
for
example,
to
say:
hey
I'm,
subclassing,
the
sampler
class
and
you're
using
it
like
a
real
runtime
value.
Then
this
won't
work
so
like,
for
instance,
if
you
have
like
class
custom
sampler
and
then
in
parentheses
sampler,
then
you
can't
use
this.
It
only
works
for
type
checking
positions
like
where
I'm
assuming
you're
using
readable
span,
otherwise
yeah
you'll
have
to
actually
fix
the
circular
import.
Interesting.
C
F
C
B
And
yeah,
it
is
I
think
it's
it's
the
official.
It's
a
it's
an
official
solution
right,
it's
in
the
my
documentation,
so
we
are
okay
with
telling
anyone
to
catch
that
if
that
helps
with
the
circular
dependency,
which,
by
the
way,
is
a
very
or
horrific
problem
right
by
the
way.
This
is
the
second
time
this
has
proved
to
be
very
valuable
information,
I've
kind
of
wondering
if
we
should
add
this
to
our
documentation,
I'm
sure
we
should.
B
There
are
some
python
specific,
you
know
not
well
known,
yeah
things
that
may
help,
but
yeah.
F
F
Maybe
we
could
do
that
as
a
part
of
I
know,
there's
this
impending,
like
reorganization
of
how
we
do
three
points
to
use
factories
instead
to
allow
more
flexibility
on
the
user
side.
We
could,
since
I
mean
I,
don't
know
if
this
these
issues
only
come
up
with
entry
points,
but
if
they
are
very
much
linked
to
that,
we
could
kind
of
add
that
as
part
of
that
kind
of
documentation
rebound
for
factories.
D
I
think
I
don't
know,
I
think
you
might
still
run
into
the
issue
with
the
sample.
Are
you
extending
the
sampler
class.
F
Even
when
I,
when
I
change
this
so
important,
yeah
yeah,
so
I
I
created
yeah,
Creator
or
PR,
my
thought
is
that,
given
that
we're
I
don't
know
if
the
release
is
fully
gone
through,
I
know
we're
like
in
the
process
of
doing
a
release.
My
thought
is
that
we
should
merge
this,
and
just
just
allow
me
to
like
get
a
bit
even
though,
even
though
I
might
you
know,
I
might
be
able
to
solve
it
today
or
tomorrow.
F
E
I
agree:
I
would
like
to
see
this
rewarded.
Like
I
saw
you
had
a
PR
to
reward
the
documentation,
but
I
don't
think
you
know
we
should
let
it
go
like
people
might
still
look
at
the
code
and
see
why
I
think
I
would
like
to
see
this
rewarded
before
that
it
is
yeah.
That's
why.
F
C
F
Could
approve
this,
we
can.
We
can
merge
this
and
then
and
then
yeah
I'll,
rework
this
a
bit
well.
First
of
all,
first
I'll
see
if
the
conditional
importing
solves
the
whole
problem,
I'm
guessing
there
might
be
something
else
to
do
too.
I
might
rework
this
a
bit.
F
Oh
one,
one
remaining
question
is
also
so
the
reason
the
reason
the
code
is
structured
like
it
is
I
talked
a
bit
about
this
during,
like
I
think
like
I,
think
like
two
weeks
ago,
but
the
reason
that
sampler
defaulting
logic
is
in
sampler,
as
opposed
to
in
configuration,
is
in
the
open,
Telemetry
spec.
There's
this
specification
that
you
can
choose
the
sampling
logic
with
an
environment.
F
You
could
say
always
on
always
off
or
I
think
there's
some
other
ones.
F
Oh
there's
like
a
there's
like
a
trace,
ID
based
rate
sampler,
there's
a
few
built-in
options
that
environment
variable
is
being
used
for
and
so
I
didn't
I
thought
it
would
have
been
bad
design
to
put
more
entry
point
environment,
variable
Logic
for
custom
Samplers
here
that
where
it's
only
used
for
auto
instrumentation
and
then
we
kind
of
have
two
different
interfaces
for
sampler
like
for
manual
instrumentation,
you
can
just
use
the
built-in
ones
and
for
auto
instrumentation
you
can
you
can
inject
the
custom
Samplers.
F
C
E
F
F
All
right
yeah,
that's
it
for
me!.
E
E
B
Yes,
the
apparently
there
is
a
an
issue
with
okay.
Let
me
so
time
ago,
I
fixed
the
an
issue
and
apparently
that
fix,
create
an
average
and
I
think
these
PR
is
fixing
the
issue
that
my
effects
created,
apparently
I.
B
We
also
need
to
use
the
the
class
that
I
created
here
and
the
user
is
suggesting
that
we
check
for
the
connection
attributes
and
if
they
are
present,
we
add
them
artificially.
I,
don't
like
this
solution,
because
here
we
are
dealing
with
private
stuff,
yeah.
C
B
Yeah,
but
on
the
other
hand,
what
we
can
do,
maybe
is
that
we
can
check
dynamically
for
all
the
attributes
that
are
there
with
a
deer
or
the
magic
attribute.
Ptt,
RS
and
dynamically
add
them
here,
maybe
maybe
that
can
work
so
yeah.
That's
what
I
commented
here,
let's
see
what
the
heels
are
of
these
brses.
E
Yeah,
so
is
it
it's
still
going
to
be
risky
right,
let's
say
the
the
let's
say:
buy
my
secret
or
some
other
library
is
using
privates
like
let's
say
self
dot.
Underscore.Com
it,
let's
see
if
they
change
the
name
some
to
something
else
tomorrow.
This
solution
is
not
going
to
work
right,
so
you
are
suggesting.
B
So
but
let's
see
what
the
the
the
the
the
author
of
these
prsos
okay
cool,
all
right,
all
right,
I
guess
this
was
the
last
one.
Then
well
then,
thank
you
all
for
joining.