►
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).
A
A
A
Yeah
yeah,
that's
cool
all
right.
Let
me
share
my
screen
so
for
topics
we
release
the
version
1.14
and
0.35.
A
Of
the
SDK
and
the
contrib
repo
last
week,
or
earlier
this
week
there
was
a
hot
fix
that
we
want.
We
got
in
related
to
DB,
API
I
think
she
can't
released
the
related
instrumentations
manually,
thanks
for
calling
for
doing
that.
A
Did
you
want
to
talk
about
that
process?
Like
is
this?
Is
this
manual
kind
of
thing,
yeah,
maintenance
burden
and
everything.
B
B
It
was
not
very
good
experience
and
also
I
I
I
made
a
mistake
of
publishing
it
like
the
DP
API,
with
like
Dev
dot
like
we
have.
We
started.
B
C
B
Yeah
yeah.
Finally,
like
I
I,
like
there
were
two
people
who
I
think
like
started
on
it
like
they
reported
that
they
wanted
to
use
it
and
yeah
I
think
it's
working
fine
for
them.
Now.
A
I
see
awesome
yeah.
We
should
probably
I'll
create
a
tracking
issue
to
try
to
automate
just
just
copy
the
automation,
workflow
yeah.
B
Repo
yeah
I
was
thinking
about
working
on
that
yeah.
If
you
create
an
issue
like
assign
it
to
me,
I
can
work
on
it.
I
also
wanted
to
spend
some
time,
I,
think
updating
the
tooling
or
you
know,
country,
getting
guidelines,
I
think
people,
some
people
try
to
come
for
help
and
then
it's
not
sometimes
clear.
B
B
C
A
B
There
was
one
issue
that
we
discussed
last
saying:
I
think
you
were
not
there.
Oh
yeah
I
think
I
can
link
it.
If
you
haven't
gone
through
them.
Please
take
a
look
at
it
and
is.
A
Yeah
so
I
think
I,
so
I
I
read
through
this
and
I.
Think
I
made
a
comment
in
the
slack
Channel,
but
I
think
it
got
lost
or
something,
but
was
there
a
discussion
related
to
this
and
was
like
was
wondering
like
like
what
what
happened
like
what's?
Where
are
we
at
right
now.
B
Yeah
we
discussed
this
last
time.
I
we
were
thinking
about
like
not
like
not
be
very
strict
in
terms
of
the
the
typing
on
you
know
wrap
around
like
if
they
give
a
like
a
simple
spam
processor,
a
bad
span
processor,
we
will
wrap
it
around
monthly
spam
processor
and
then
do
it.
A
Like
I
have
I
have
no
I
kind
of
have
to
see
this,
but
I
I.
Don't
really
have
any
opinions
on
it
right
now.
It
seems
to
make
sense
to
me.
So
was
there
like
a
discussion
related
to
this.
A
Okay,
cool
yeah,
if
we
can
like
just
continue
the
discussion
in
the
in
the
issue,
that'd
be
great.
Nice
get
everyone's
thoughts.
I,
don't
think
Diego's
here
today,
either
so.
A
Does
anyone
else
have
any
thoughts
on
these
two
issues.
D
Yeah
we
talked
about
it
last
week.
I
think
I
basically
said
I.
Think
there's
nothing
like
the
types
are
correct.
D
They
are
restrictive,
but
you
know
if,
if,
if
we
changed
it,
then
this
ad
spend
processor
wouldn't
work
unless
we
wrapped
it
in
one
of
those
two
so
I
think
it's
reasonable
to
just
say:
hey.
We
have
to
leave
it
as
it
is
for
this
ad
spend
processor
method,
please
pass
in
a
wrapped.
D
If
you
just
have
a
single,
then
bronzer,
you
want
to
pass
in
just
wrap
it
in
one
of
the
multi-span
processor
implementations.
A
A
A
Okay,
cool
yeah,
we'll
just
continue
the
discussion
there
I
think
another
issue
was
added
here:
Aaron
is
a
suit
yeah.
C
D
Opened
this
one:
basically,
the
we
have
a
sequence
in
the
data
points
for
each
of
like
some
histogram
engage,
just
a
reminder:
sequence
would
be
like
list
or
Tuple.
D
On
the
other
hand,
the
actual
type
is
a
generator
on
which
you
can't
call
when
or
access
like
arbitrary
items.
Also,
you
can't
iterate.
C
D
It
multiple
times
you
can
only
iterate
over
a
generator
once
so
I
think
we
should
just
fix
it
and
not
adjust
the
the
type
which
would
be
the
other
option
and
yeah
the
reason
being
I
put
at
the
bottom.
But
if
somebody
doesn't
consume
these
generators,
then
it's
going
to
hold
the
SDK
block
until
the
generator
is
consumed,
which
is
not
great
so.
D
A
D
A
Right
also,
I
remember:
was
there
any
updates
on
like
Josh's
pass
through
through
our
metrics
APN
SDK.
D
I,
don't
think
I
might
have
missed
it.
No,
there
wasn't
and
like
he.
He
dropped
a
comment,
but
I,
don't
it's
probably
off
his
radar.
I
mean
I
can
remind
him
again,
but
honestly,
it
might
be.
D
A
Sure
yeah
that
sounds
good,
but
we're
getting
Community
feedback
anyways.
So,
oh,
are
we
like
like
we're
just
like
fixing
bugs
like
where
each
of
us
are
also
testing
out
the
SDK
and
using
it.
So
like
that's
another
good
Channel,
you
know
sure
yeah.
A
Okay,
cool
any
other
comments
on
this
issue.
Does
anybody
have
any
questions
or
have
a
problem
with
this
kind
of
solution.
A
Okay,
going
right
into
PRS
here.
E
All
right,
so,
basically,
this
is
a
fix
to
the
custom
sampler,
injecting
PR
that
I
that
I
did
last
week.
The
the
issue
essentially
was
the
way
python.
Does
entry
point
loading
I.
C
E
Yeah
I
think
last
week
last
week,
I
actually
did
a
little
presentation
explaining
this
a
bit,
but
essentially,
when
you
load
an
entry
point,
it.
E
Oh,
weird,
okay,
all
right,
I'll
start
over
so
yeah.
Basically,
I
did
a
presentation
about
this
last
week,
but
essentially,
when
python
loads,
an
entry
point,
it
actually
like
loads
that
entire
module
as
a
dependency
of
the
current
file.
So
if
we
the
way
I
previously
did
it,
the
sampler
module
in
open
Telemetry
would
load
point
which
could
potentially
be
a
module
of
a
user
that
then
uses
like
trace
or
sampler,
which
means
that
it
would
create
it.
E
Could
it
has
a
very
high
likelihood
of
creating
a
circular
dependency
when
users
actually
use
it
so.
E
Did
is
I
moved
this
logic
to
configuration,
which
is
one
of
the
places.
I
was
thinking
about
doing
it
before
anyways
and
that
essentially
means
that
if
something's,
using
like
Auto
instrumentation
or
really
just
if
they're
using
configuration,
they
can
inject
a
sampler.
But
if
they're
using
just
like
regular
manual
instrumentation
the
regular
defaulting
behavior
is
there,
so
it's
just
they
still
have.
They
still
can
use
that
those
environment
variables,
but
it's
just
for
the
built-in
samplers
and
if
they
really
want
to
watch
still
they
can.
You
know
they
can
pass
it
pass
it.
E
So
it
basically
moves
the
custom,
sampler
injection
to
yeah
to
configuration
and
leaves
the
previous
defaulting
to
built-in
Samplers
intact.
So
there's
backwards
compatibility
yeah-
and
this
also
this
also
solves
one
of
the
issues
that
people
identified
before
where
they
didn't
really
like
that
the
import,
config
components
was
moved
to
utilities
so
because
I've
moved
this
logic
back
to
configuration
that
can
be
moved
back
to
configuration
too.
E
So
part
of
this
is
kind
of
like
it's
kind
of
a
revert
of
that
original
PR.
It's
really
a
redo.
C
E
And
I
saw
Leighton
recently
left
some
comments
so
going
through
and
taking
a
look
at
those
but
yeah.
So
I'd
love
some
more
more
feedback
and
approvals
on
this
yeah.
Let
me
know
if
there
are
any
questions.
A
Hey
Jeremy,
I'm,
curious,
I
I
think
you
already
made
the
change
that
I
asked
to
remove
this,
but
your
comment
says
this
still
needs
to
remain
for
the
defaulting.
Behavior
I
was
curious
about
what
that
means.
Yeah,
okay,.
C
C
E
It's
just
I
need
to
I
need
to
like
find
in
the
actual
code
to
well.
Actually
you
know
it
probably
should
be
in
that.
So
basically,
if
you
look
in
under
like
the
init
directly
under
Trace,
which
I
think
we're
in
now,
if
you
look
at
the
Tracer
provider
class,
which
it
looks
like
we're,
probably
in
right
now,
what.
C
E
Is
I
moved
that
to
oops
good
you
can
just
go
back,
you're,
actually
you're.
Actually,
there
I
just
I
thought
you
were
in
a
different
class
yeah.
If
you
scroll
down
a
bit
I
say
if
not
sampler
sampler
equals
sampling,
dot
get
environment
variable.
So
basically
I
thought
you
were
saying.
Can
we
remove
this
defaulting
logic?
Oh
okay,
I
I
moved
it
there,
because
when
I
was
testing,
I
noticed
that
actually,
especially
because
we.
C
E
A
variable
Trace
sampler
that,
like
that
refers
to
this
function,
we're
actually
copying
that
function
every
time,
even
when
the
sampler
argument
is
initialized,
and
that
was
causing
issues
and
also
it's
just
like
extra.
You
know
extra
processing
for
no
reason
so
the
way
I
set
it
up
now
we're
only
calling
defaulting
when
it's
none,
as
opposed
to
calling
it
getting
the
default
sampler
and
then
only
assigning
it
when
it's
not.
A
I
think
my
comment
was
just
that,
like
I
think
when
I
reviewed
it
Trace
sampler
still
existed
so
yeah
you're,
removing
it
anyways
so
that
that
was
it
was
as
simple
as
that.
Yeah
okay,
I
see
yeah
because
this
was
being
used
anywhere,
but
that's
fine.
A
Okay,
so
a
second
question
to
clarify
so
for
for
manual
instrumentation.
If
someone
wants
to
use
a
custom
sampler,
they
have
to
create
a
custom
Tracer
provider.
E
A
Like
when
the
trade
supervisor
instantiated,
you
mean
right,
yeah,
exactly
correct
me
if
I'm
wrong,
but
doesn't
this
Oh
you
mean
like
so
the
defaulting
doesn't
happen.
If
the
sample
is
passed
in
exactly.
E
E
E
You
can
also
just
use
your
environment
variable
to
refer
to
the
built-in
Samplers
and
that's
what
that
get
get
from
and
for
default
method
is
what's
new
about
this
and
the
last
PR
the
functionality
I've
been
adding
is
making
it
so
that
you
can
add
a
custom
sampler
through
a
configuration
or
Auto
instrumentation,
and
so
that's
what
I
did
in
the
the
configuration
in
it.
D
A
I
think
in
this
PR
there
was
a
circular
import
like
a
dependency
issue
that
could
I'm.
D
Saying
line
one
one,
one
eight
was.
E
The
previous,
so
the
reason
yeah
the
reason,
I
changed.
This
is
both
because
it
was
just
it
wasn't,
had
some
issues
like,
for
instance,
the
I,
don't
know
if
you
saw
leighton's
comment
about
like
the
ID
generator,
but
it
was
like
inconsistent
typing
there
for
this.
The
reason
why
I
did
this
is
to
make
it
so
that
we're
not
always
calling
this
get
from
environment
method
or
default,
even
when
we
already
have
a
sampler.
D
C
A
Okay,
so
the
only
kind
of
quote-unquote
feature
Gap
is:
if
a
user
wants
to
do
manual,
instrumentation
and
use
a
custom
sampler
through
the
environment,
variable
right.
A
E
A
D
Wait
wait?
Why
not.
E
Oh
whoops
I
keep
oh
I,
see
what's
happening,
I'm
pressing
a
key
to
mute
myself,
yeah,
so
I
yeah
I
wasn't
able
to
find
a
way
to
do
that
without
a
circular
dependency
which.
E
Make
sense,
given
that,
like
the
rest
of
the
time
we're
doing
entry
point,
you
know
entry
point
style
injections,
we're
doing
as
like
the
configuration
level.
C
E
E
There
we
go
okay,
so
here's.
So
this
is
the
API.
This
is
the
SDK.
Before
what
was
happening
was
I.
We
had
I
I,
had
the
entry
point
upload
happening
in
in
sampler
or
really
in
like
utility
these,
but
essentially
in
essentially
the
same
sample
was
calling
utilities,
and
so
that
previously
worked
if
you're
looking
at
open,
Telemetry,
SDK
and
open
Telemetry
API,
there's
not
actually
any
circular
dependency
happening
here
and
so.
C
E
E
Is
that
it's
almost
like
almost
overwhelmingly
likely
that
that,
when
a
user
uses
this
they're
going
to
use
some
sort
of
sampler
that
refers
to
Something
in
the
trace,
module
or
or
just
a
sampler,
that's
in
the
same
module
as
something
that
refers
to
trace
the
trees
module?
Because
when
entry
point
download
happens,
it
loads,
the
entire
module
of
that
custom.
Sampler.
Is
it,
and
so,
like
the
the
example
that
I
was
going
to
change,
was
in
I
was
like
a
full-fledged?
Let's
name
it
that.
E
Sampler,
that
was
in
the
same
modules
like
an
exporter,
and
these
files
refer
to
like
readable.
B
E
Trait
the
Tracer
provider
and
stuff
I'm
there
for
just
border
logic.
We
had
and.
C
E
Of
that
has
a
dependency
on
sampler,
and
so
it
would
create
a
circular
dependency.
Only
one
century
point.load
is
called
so
I
haven't
made
another
diagram
to
show
how
it
works
with
the
new
style,
but
essentially
because.
C
E
Moved
all
the
entry
point
loading
back
to
configuration,
it's
fine,
because
configuration
is
essentially
a
edible
than
everything
else
like
configuration
relies
on
Trace,
which
relies
on
sampler.
So
if
configuration
is
loading,
this
user's,
sampler
and
exporter-
and
you
know
a
custom
ID
generator
or
anything-
that's.
C
E
Relies
on
sampler
and
it's
calling
something
else
that
lies
on
the
sampler,
so
there's
no
circular
there's
no
circular
dependency.
There.
A
So
Jeremy
I
think
the
question
was:
why
are
we
not
able
to
specify
custom
Samplers
if
we're
doing
manual
instrumentation
so,
like
I,
understand
how
your
circuit,
this
PR
fixes
the
circuit
dependency
but
like?
Why
is
it
so
that
we
we
still
have
that
feature
Gap
like
what
would
the
dependency
structure
look
like
if
we
try
to
add
that
feature
back
in
to
get
from
n
or
default.
E
Yeah,
so
so
this
is.
This
is
the
diagram
for
the
the
previous
change,
the
change
before
the
fix,
where
I
did
try
New
Haven,
so
you
could
use
custom
Samplers
for
me,
instrumentation
too,
and
to
be
clear:
that's
that's
actually
not
the
norm.
We're
usually
only
doing
like
entry
point
injection
at
the
configuration
level
and
the
other
times
we
use
custom
stuff
where
we're
allowing
people
just
pass
it
in.
A
For
audience
right
right
so
like
the
only
difference
with
that
so
like
you
should
just
like
pull
the
sample
or
entry
point
that
load
into
the
configuration.
That's
the
only
difference
right,
pretty
much
say
that
again,
so
you
see
that
square
that
your
mouse
is
on
sampler
entry
point
that
load
stuff
like.
E
So
basically
the
change
is
that's
right.
This.
E
Well,
but
it's
it's.
A
D
Okay,
but
once
it
Imports
that
that
will
import
configuration,
so
that's
like
where's
the
cycle,
I,
don't
understand
what
cycle.
E
Is
it's
like
the
cycle
is
if
sampler
is
like
in
order
in
order
to
load
it
for
manual
instrumentation,
the
entry
point,
loading
would
have
to
be
in
in
sampler
or
potentially
trades,
but
either
way.
You're
then
loading
something
that
will
likely
I
mean
it's
it's
not
it's
not
certain,
but
it
will.
It
will
very
likely
rely
on
elements
from
trace
and
sampler,
and
so
that
creates
that's
a
circular
dependency
between
our
you.
C
D
E
User
is
sampler
module.
D
Okay,
I
mean
I,
think
there's
definitely
ways
around
that.
If
that's
the
feature
Gap
like,
for
instance,
if
you,
if
you
tried
doing
the
you,
can
move
the
import
out
of
the
global
out
of
the
module
scope,
and
you
could
do
it
like
in
right
before
you
call
the.
E
No
no,
but
it's
it's
even
even
if
it's
in
you
mean
which,
which
thing
are
you
opposing
moving,
because
I
did
try
the
last
time
someone
posed
I
forgot,
which
import,
but
someone
propose
making
I
think
it
was
the
spam
kind.
No,
it
was
the
readable
Spanish
word
someone
proposed
making
the
readable
Spanish
report
in
inline
import
and
that
didn't
ultimately
solve
the
problem,
because
this
is
still.
E
This
is
still
a
circular
dependency
that
arises
like
at
runtime.
So
whenever
we
need
readable
spam,
it
would
then
create
a
circular
dependency.
C
C
D
Sorry,
maybe
I
don't
understand
completely
I
can
just
I
think
we
are
but
like
I.
A
Have
this
as
just
a
known
feature:
Gap-
and
we
didn't
have
this
feature
before
anyways,
so
at
least
not
a
buggy
version
of
it,
so
maybe
create
a
new
issue
and
we
can
have
a
discussion
there
to
better
explain
why
Jeremy
thinks
this
is
a
can't
be
done.
This
way.
E
E
Be
something
I'm
missing,
but
I've
tried
a
bunch
of
different
ways,
including
some
inline
importing.
It
didn't
seem
to
solve
the
issue
so
yeah
there
might,
there
might
be
something
else:
that'd
be
pretty
cool.
A
E
Used
well.
E
A
No,
the
issue
of
currently,
we
cannot
Define
custom
samplers
via
manual
instrumentation.
E
A
D
D
Good
yeah
and
like
also
maybe
a
minimal
Repro
that
doesn't
involve
any
of
the
code
in
the
repositories
like
you
know,
just
like
a
toy
example
that
exhibits
the
same
problem
but
for.
C
A
Okay,
cool
any
other
questions,
comments
on
that
PR
ere,
at
least
for
now.
Please
take
a
look
at
it,
we're
keeping
the
scope
kind
of
small,
so
at
least
we
can
try
to
fix
this
bug.
First,
I'll
take
another
look
at
it
as
well:
nice,
one.
E
A
A
Whatever
solution
you
come
up
with
we're,
probably
going
to
apply
this
to
other
components
as
well,
so
like
propagator
or
ID,
generator
or
stuff
like
that
yeah.
This
could
definitely
be
something.
A
Cool
I
put
this
up:
is
this
still
relevant
Jeremy.
E
No
I
thought
I
actually
closed
it
too.
I
thought
I
didn't
keep.
Oh,
maybe
I
just
revert
it
back
to
a
draft
Yeah,
so
basically
I
initially
wanted
to
revert.
The
change
I
think
suggested
that,
because
I,
when
I
noticed
the
bug,
I
wanted
to
be
able
to
fix
it
before
it
hit
the
release,
but
then
it
made
it
into
release
anyways,
so
I,
just
I
just
did
this
PR
and
went
ahead
with
the
with.