►
From YouTube: 2022-04-14 meeting
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
D
C
A
Hey
leighton,
I
don't
think
geo
can
hear
us
how
about
now.
Can
you
hear
me.
D
F
F
G
Okay,
cool
yeah,
as
usual,
as
just
diego
said.
Please
add
your
social
attendees
list
or
any
other
issues
and
topics
and
pr's
that
you
want
to
talk
about
today
with
the
rest
of
the
sig.
Please
add
them
to
the
meeting
notes,
go
right
ahead
and
start
with
the
first
topic.
I
believe
so
it
needs
more
approvers
and
here
was
a
link
to
the
community
guidelines
on
what
an
approver
is
plus
its
requirements.
G
So
I
think
alex
put
up
this
link,
but
he
talked
to
me
briefly
about
this.
I
think
this
stem
from
we
were
just
like.
We
were
just
kind
of
going
through
the
approver
and
maintainers
list
and
like
seeing
some
like
pruning
it.
Since
peop,
some
people
were
in
inactive
and
we're
kind
of
relative
to
the
other
things,
we're
kind
of
lacking
a
lot
of
help.
G
G
So
it
would
be
useful
to
get
like
more
eyes
on
pr's.
Does
anyone
guess?
Does
anyone
want
to
nominate
someone
in
terms
of
like
anyone
that
might
be
a
good
candidate
or
anyone
in
the
chat
want
to?
Let's
try
this
out?
Maybe
what
do
you
guys
think.
C
I
think
I
I
don't
know
if
sanket
and
ashu
are
interested,
but
they
have
been
contributing
for
quite
some
time.
I
think
they
did
a
lot
of
work
on
the
country,
so
maybe
we
can
encourage
them
to
join,
approves
in
the
contribution.
G
It's
like
say
kid
is
in
the
chat.
Is
that
something
that
you
might
be
interested
in
or.
G
Okay,
awesome
awesome
yeah.
So
if
that's
something
that
you
want
to
be
doing,
since
you
seem
to
be
contributing
a
lot
in
the
contributor.
G
Please
take
a
look
at
the
the
link
in
the
in
the
topic
section
that
is
linked
for
the
community
guidelines
and
I
believe
we
have
our
own
kind
of.
H
G
G
G
Yeah
yeah
no
worries
yeah,
please
let
us
know
cool.
I
believe
that
was
everything
that
alex
wanted
to
talk
about.
If
not,
we
could
bring
it
up
next
week
cool
moving
right
along
metrics
progress.
I
will
be
let's
open
up
that
project
board.
G
F
Okay
yeah,
can
you
show
the
world
please?
Yes,
thank
you
so
great.
So
as
as
you
know
already,
the
idea
is
to
have
this
released
in
the
month
before
the
or
before
the
kubecon,
which
is
happening
in
october,
in
yeah
in
may
16
something
okay.
There
is.
F
Okay,
so
we're
not
that
far
so
there
is,
there
are
four
issues
right
now
that
are
still
unassigned,
and
I
am
working
right
now
on
this
vr
data
from
identical
instruments
must
be
created.
The
rest
of
the
issues
are
assigned
already.
I
plan
on
taking
probably
another
issue
today
from
the
unassigned
list.
F
Which
is
because
I
am
quite
close
on
this
issue,
I'm
working
on,
so
I
am
trying
to
at
least
put
a
pr
to
all
the
unassigned
issues
by
the
end
before
the
beginning
of
next
week,
so
that
we
can
hopefully
focus
on
or
reviewing
those
and
making
progress
on
the
ones
that
are
assigned
already.
F
So
so
yeah.
The
the
rest
of
the
issues
are
already
assigned
and
I'll
I'll
try
to
free
myself
up
next
week
so
that
if
someone
doesn't
have
bandwidth
to
take
on
their
assigned
issues,
I
can
help
our
country
ending.
That's.
G
Hey
diego,
I
think
so.
This
project
board,
I
think,
like
you,
need
to
manually,
add
issues
to
it
when
they're
created
the
like.
Recently,
some
new
issues
have
been
created
mostly
by
you,
are.
H
F
H
F
Mostly,
the
issues
that
are
being
left
out
are
the
the
ones
for
this
feature,
that
is,
to
print
a
view,
recipe
that
can
fix
a
configuration
conflict.
That
is
not
another
strict
requirement
in
the
in
the
spec
institute,
and
I
also
talked
with
folks
from
java,
and
they
are
also
not
including
that
feature
in
their
in
their
in.
A
Their
project
yeah
for
foreign.
G
Specifically,
is
this
resolved
like
we
chose
not
to
do
this
right.
B
F
Yeah,
I
don't
know
if
we
reached
an
agreement
on
that
one.
It
is
if
I
remember
correctly
that
that's
not
a
specification
feature
that
we
need
to
implement
it's
more
on
implementation,
detail
on
the
outside
the
name
that
we're
going
to
use
for
measurements
depending.
Oh,
it's
an
internal.
A
So
so
that
I
was
still
a
little
fuzzy
was
this
was
the
whole
cubecon
thing
discussed
during
the
maintainers
meeting.
F
A
F
Something
that
I
I
was
asked
from
carlos
who's,
pretty
much
managing
things
on
our
side
on
license.
Oh.
F
G
Projects
so
right
because
we
wanted
like
multiple
languages
or
something
right
to
like
spearhead
the
metrics
release
right
all
right
cool,
so
it'll,
be
like
it'd,
be
a
great
milestone
for
us,
but
it's
not
an
actual
official
ask
from
the
community.
F
Wanted
well
from
the
yeah,
probably
it
was
about
uh.net
java
and.
G
Right
right,
we
were
the
languages
to
like,
you
know,
be
the
most
ahead.
However,
the
specific
deadline
of
having
it
by
kubecon
is
it's
not
like
a
community
like
like
no
one
said
that
in
the
spec
community
or
the
maintenance
main
community
correct.
G
Yes,
I
don't
recall
anyone
saying
that
okay,
all
right,
but.
F
G
You
could
just
find
out
like
verify
where
you
heard
that
that'd
be
great,
like
we're
not
saying
we.
E
A
I'd
like
to
see
like
the
sort
of
what
other
people
have
committed
to
do,
because
I'm
a
little
nervous
about
the
deadline,
specifically
with
like
respect
to
having
people
try
things
out
before
we
mark
it
stable,
because
even
if
we
had
something
that
was
perfectly
implementing
the
spec,
I
think
we
should
have
you
know
some
waiting
period
before
we
declare
it.
You.
A
For
people
to
try
it
out,
you
know,
report
memory,
leaks
or
performance
issues
and
stuff,
but
but
also
just
like
I
want
to
understand.
If
for
cubicon
is
just
to
have
a
spec
compliant,
I
think
that's
one
thing,
but
actually
making
the
1.0
release.
I
think
it
took
us
several
months
if
I
remember
right
for
tracing
to
actually
make
that
stephanie
1.0.
F
A
F
At
least
I
guess
we
should
aim,
for
I
mean
a
strong
release
candidate.
I
guess
for
that,
for
that
goal
right.
It's
it's
another
thing
now
of
pretty
much
taking
care
of
small
details
in
the
in
the
implementation,
in
our
implementation
with
the
spec,
which
are
luckily
not
that
many
but
but
yeah.
Definitely
it's
it's!
Oh
I'll!
G
Also
aaron
yeah,
I
think,
for
this,
like
we
had
a
couple
of
agreements
to
it.
We
kind
of
just
got
sidetracked
with
like
the
measurement
discussion.
I
believe.
A
Yeah
yeah,
so
a
few
things.
This
is
also
sort
of
tied
in
with
the
issue
that
diego
is
working
on
for
the
multiple
instrument
callbacks,
because
if
we
were
to,
for
instance,
do
the
option
two
in
that
list,
which
was
basically
the
each
observer.
Sorry,
each
callback
will
receive
like
a
function.
It
can
call,
and
then
we
don't
even
need
a
name
class
for
this.
But
I
did
talk
with
diego
about
that
and
I
think
you
were
leaning
to
option
one.
It's
not
on
this
issue
but
yeah.
A
Basically,
this
is
part
of
the
metrics
api
and
once
we
know
like
what
the
what
the
async
instrument
api
is
going
to
look
like
for
sure
yeah
for
this
one.
F
Yeah,
just
for
the
record
the
the
feature
that
aaron
just
mentioned,
the
multi-instrument
callback-
that's
not
something
I
am
currently
working
on.
I
don't
know
if
you,
you
saw
the
slack
messages
that
I
sent
you
yesterday,
but
the
there
are
two
parts
for
that
they
are.
One
of
them
is
supporting
zero
or
more
callbacks
in
the
asynchronous
estimates,
which
is
a
requirement
in
the
api.
F
I
already
implemented
that
in
that
vr,
okay
now,
the
second
part
is
that
will
be
the
multi-instrument
call,
but
that
is
not
yet
part
of
the
spec.
It's
still
an
open
pr.
So
I'm
not
focusing
on
that.
A
Right
now,
okay,
fair
enough
yeah.
I
just
think
that
we
should
at
least
have
a
plan
for
for
that,
because
it's
almost
certainly
going
to
like
it.
It's
completely
necessary
because
otherwise
you're
gonna
you're
gonna
end
up
having
to
do
like.
The
canonical
example
is
like
the
proxtap
file.
If
you,
if
it
has
like
10
instruments
that
need
to
get
stuff
out
of
that
file,
you
don't
need
to
read
the
file
10
times.
You
only
need
to
read
it
once
so,
like
I
I'm
very.
F
F
I
agree
completely
that
the
pr
I
think
it's
as
several
green
checks
already,
let's
go
in
for
regarding
the
plan
that
that
we
need
for
that
that
yeah,
I
am
leaning
with
option
one.
That
is
something
I
discussed
recently
and
so
far.
I
think
we
can
implement
following
that
option.
F
Okay,
this
vr
sorry
support
for
zero,
more
callbacks,
it's
big.
F
G
Okay,
so
slightly
related,
I'm
working
on
the
what's.
It
called
the
the
issue
in
which,
if
there
is
a
exception
thrown
for
one
of
the
async
instruments
in
a
callback
like
right.
Currently
we
just
like
stop
processing.
G
G
Which
issue
are
you
referring
to
this
one
consume
measurements
of
remaining
insecurity
when
one
fails?
Okay,
so
yeah.
So
here,
specifically,
we
just
loop
through
the
async
instruments
and
like
if
one
of
the
callbacks
fail.
We
just
like
we
stop
processing
of
the
a
string
instruments
because
an
exception
is
thrown.
G
So
this
is
fine.
If
there's
only
I
mean
like
a
single
callback
like
so
we
we
we
continue
on
the
we.
We
continue
with
the
next
instrument
if
the
callback
fails
right,
very
simple,
but
if
there's
multiple
callbacks
now
like
do
we
come?
Do
we
fail?
If
a
call
do
we
continue
the
callbacks?
If
it
fails
or
do
we
could
just
skip
to
the
next
instrument
if
it
fails.
F
G
Okay,
yeah
yeah
me
too
looking
at
this
change
here,
though
it's
it
looks
like
it's
gonna
be
done
in
a
separate
place,
then
right.
G
All
right
correct,
but
if
you
add
this
right,
like
my
code,
would
would
would
be
added
here
right.
H
G
Would
continue
to
the
next
instrument,
but
if
you
add
this,
then
it
wouldn't.
Actually
it
wouldn't
do
that
continue
on
callback
it
would.
It
would
be
continuing
on
instrument
right
so
yeah.
What
I
mean.
F
Is
that
in
in
that,
assuming
that
this
vr
gets
merged
first.
F
G
Would
probably
have
to
wait
until
yours
is
march
first
and
then
add
it
to
the
respective
area
in
which
you
loop
through
the
callbacks.
F
Right,
correct
that,
in
that
line
in
line
92
is
where
we
look
for
through
all
the
callbacks,
so
I
it
should
be
pretty
simple
once
the
this
get,
this
gets
merged.
You
go.
Oh.
G
Sorry
yeah,
you
could
bring
it
up
if
we
have
time,
but
we
should
keep
going
nice.
Oh
sorry,
aaron
can
we
so
this
is
not
needed
for
the
release,
but
I
I
didn't
unders.
I
didn't
catch
if
we,
if
there's
like
a
blocker
for
this
and
what
you
were
talking
about
with
diego-
are.
A
A
That's
the
goal
here
is
just
to
avoid
that
confusion.
So
if,
if
nobody
has
objections
to
it,
I
think
we
should
probably
just
you
know:
do
it.
G
Yeah,
I'm
okay
with
it
also
question.
Oh
sorry,
good.
D
G
I
got
it
oh,
but
like
before
that
question,
so
this
this
internal
measurement
class
right
is
won't
be
used
by
users,
but
but
the
api
one
will
right
like
currently
yeah.
G
A
Yeah
so
so
I
we
have
like,
basically
a
really
inconsistent
messy
state
right
now,
so
we
have
some
things
that
are
prefixed
with
underscores
some
things
that
are
in
private
modules.
I
think
I
I
have
an
issue
open
for
resolving
all
that.
I
remember
that
yeah
is.
A
G
What
I
thought
this
turns
to
like
a
like
a
triage
thing:
yeah,
it's
just
slow,
come
on
github,
anyways,
all
right
I'll!
Just
do
this
later,
whatever
okay,
I
think
yeah
moving
right
along.
G
Is
that
everything?
Oh
sorry,
we
we
didn't
even
talk
about
like
individual
matrix,
diego!
Is
that
what
you
wanted
to
kind
of
cover,
yeah.
G
Okay,
cool
yeah,
I'm
working
on
this,
but
I'll
be
waiting
on
diego
after
that
conversation.
So.
F
Yeah
that
that
pr,
that
yours
depends
on
it's
already
opened
and
ready
for
review.
So
if
you
yeah
yeah
I'll.
G
Take
a
look
at
it
too
is
alex
here,
alexis
wouldn't
be
joining
us:
okay,
cool
cool,
all
right,
erin,
any
updates
on
the
issues
that
are
assigned
to
you.
A
So
I
discussed
the
timeout
mechanism
with
diego,
so
the
issue,
the
biggest
issue
here-
is
it's
sort
of
scoped
to
also
we
want
to
have
like
a
mechanism
for
the
tracing
sdk,
but
the
in
the
spec
says
you
may
optionally
specify-
or
rather
you
may
optionally,
accept,
like
a
timeout
duration
and
adding
that
in
the
tracing
sdk
would
be
a
breaking
change.
So
we
we're
sort
of
we're
gonna
have
to
decide
between
having
like
an
inconsistency
between
the
two
verse.
A
You
know
just
not
having
it
be
configurable
which
I
don't
think
is
very
good
either.
So
there's
a
few
options
potentially,
but
I
haven't
actually
worked
on
it
or
really
like
decided
on
the
implementation.
D
G
H
G
If
you,
if
you
have
a
couple
of
ideas,
if
you
could
do
something
similar
to
what
we
do,
what
you
did
for
the
multiple
callbacks
thing,
just
like,
let's
start
with
the
options,
we
can
have
a
discussion
about
that
sure.
As
long
as
as
long
as
people
read
it.
Oh
yeah,
yes,
correct.
A
Yeah
like
it
was
like
I
want
to,
I
want
to
have
a
discussion,
but
I
also
don't
want
to
write
a
block
of
text
and
then
it'd
be
too
daunting
for
anybody
to
want
to
address.
So,
oh
no.
A
Yeah
I've
got
a
few
other
ideas,
ranging
from
python
hackery
to
awesome.
I
love
python
by.
F
The
way
you
are
talking
about,
when
you
mention
all
these
ideas
that
you
have,
they
are
all
related
to
the
to
how
to
configure
the
timelines
right.
G
Okay,
cool
sound
good.
I.
H
G
That's
we're
in
agreement.
There
awesome!
Okay,
I
think.
That's
it
for
magic
stuff.
Let's
move
right.
H
H
Yeah,
so
actually
I
mean,
if
you
guys,
are
aware
so,
a
few
months
ago
we
had
that
lock,
4g
vulnerability
issue
all
over
the
world
right.
So
I
was
just
wondering
I
mean:
do
we
have
any
such
vulnerability
check
in
python
agent?
I
mean
hotel
python
engine.
H
So,
just
to
before
any
any
anyone
introduce
a
new
library
or
something
we
we
make
sure
that
the
version
that
we
are
using
it's
not
vulnerable.
Just
like
I
mean
can
we
add
something.
You
know
any
any
build
check
like
black
truck
scan
or
some
open
source
scan
vulnerability,
scan
in
vertical.
H
F
C
Which
which
basically
does
the
analysis
and
generates
the
reports,
but
so
far
there
weren't
any
no
major
typical
issues
reported.
There
were
some
issues
around
in
in
the
examples,
but
I'm
not
in
the
library
code
itself.
G
And
about
does
security
checks?
Oh
it's
just
like
checks
for
dependency.
F
Conflicts
and
stuff:
no,
it
does
security
checks.
I
I
remember
I've
seen
that
before
showing
us
a
warning
saying
one
of
your
dependencies.
H
G
G
All
right
cool
sounds
good
right
along.
G
I
think
that's
it
for
the
topics.
We
go
right
into
issues
aaron,
auto
instrumentation
logging
problems.
A
There
were
two
of
them
with
logging
that
seemed
important
to
address,
so
the
first
thing
is
and
street
con.
You
can
correct
me
if
I'm
wrong,
but
basically
what
happens
is
the
hotel
logging
handler
seems
to
get
installed,
regardless
of
whether
or
not
you
configure
a
logs
exporter
and
the
side
effect
there
is
that
the
standard
outlogging
default
behavior
of
python
goes
away
so,
like
this
code
doesn't
print
anything
out
right
here.
A
If
you
add
dash
dash
logs
exporter
otlp,
then
it
will
print,
but
it
will
be,
like
you
know,
in
the
otp
logs
format
that
the
console
exporter
uses
sorry,
not
otlb,
I
meant,
if
you
do
logs
exporter
console,
so
that's
not
ideal,
because
it's
not
you
know
like
super
readable,
but
I
think
this
is
probably
a
small
issue
to
fix.
Streetcar
would
know
better.
C
Right
we
updated
like
there
was
an
issue,
somebody
open
and
we
updated
the
auto
instrumentation
to
you
know,
configure
the
like
login
by
piping
pipeline
as
well.
Our
own
is
correct
that
we
are,
you
know
automatically
attaching
the
log
handler.
Another
behavior
is
no.
If
you
do
not
add
any,
you
know
exporter,
but
I
I'm
I
haven't
like
fully.
You
know
sure
like
why
this
is
you
know
like
interfering
with
the
the
standard
line.
Like
you
get
on
the
console,
so
yeah.
Some
people
think.
C
So
I
I
I'll
assign
this
to
myself.
I
already
started
looking
into
it.
I
haven't
gotten
a
you
know,
much
time
to
look
into
this
yeah
I'll,
be
looking
into
this.
A
So
the
second
one
was
really:
oh
sorry,
this
isn't
the
other
logging
one.
I
didn't
make
an
issue
for
the
other
logging
one,
but
the
I
raised
that
to
distribute
as
well
and
what's
happening
is
for
some
reason
any
logs
or
exceptions
from
within
the
like
trace
sdk
batch
span.
Exporter
are
causing
they're
getting
completely
swallowed.
A
So
even
if
you
do
capture
logs
with
otlp
for
some
reason,
the
logging
gets
suppressed
and
I
tried
to
figure
that
one
out
for
a
while,
but
I
couldn't
figure
that
one
out
I'll
probably
make
an
issue
for
that
offline.
B
C
But
like
both
issues,
my
you
know
probably
the
same
issue,
something
yeah
so
yeah.
I
I
think
I
have
the
repro
like.
I
don't
show
me
a
way
to
reproduce
that
I'll
be
looking
into
that
as
well.
A
Cool
yeah.
Sorry,
this
one
is
a
separate
issue
not
related
to
logging.
So
this
one.
Basically
all
these
things
came
up
because
I
was
updating
the
gcp
cloud
trace
exporter
to
work
with
auto
instrumentation
by
adding
the
entry
point
all
that
worked
well
in
the
entry
in
the
entry
point
for
the
environment
variables,
but
sometimes
depending
on,
like
your
credential
setup,
what
happens?
Is
the
library
used
within
the
trace
exporter,
the
cloudtracex
portable
called
the
gcloud
command
and
the
g
cloud
command
is
written
in
python.
A
So
the
first
thing
it
does
is
invokes
a
sub
process
and
what's
happening
is
the
sub
process
tries
to
get
auto
instrumented,
which
will
in
turn
try
to
create
an
exporter
which
we'll
try
to
call
the
gcloud
command
again,
etc.
So
it's
kind
of
like
fork
bombing
your
machine.
If
you
auto
instrument,
it
looks
like
if
you
open
the
site,
customize
file,
the
code
I
linked
there,
we
prune
the
pipe
the
site,
customize
files
directory
out
of
the
python
path
to
prevent
this,
but
we
only
do
it
after
the
actual
setup.
A
So
the
setup
is
in
the
try
block
here
so
like
basically,
the
exporters
will
get
instantiated
during
the
try
block
and
the
rest
of
the
code
will
run
after
the.
Finally,
so
it
seems,
like
somebody
saw
this
issue
before
and
fixed
it,
but
basically
to
fix
this
issue.
I
might,
I
would
propose
just
moving
this
code,
that's
in
the
finally
blocked
to
the
start
of
the
initialize
function.
A
G
But
yeah-
maybe
maybe
they
weren't
specifically
running
to
the
issue
of
like
a
python
subprocess
being
generated
here.
Yeah.
G
Anyways,
it's
okay,
yeah!
Well,
it's
a
very.
G
Interesting
scenario
that
you
found-
I
guess,
luckily,
that
you
did
that
the
gcp
explorer
does
do
this
right.
Is
this
for,
like.
D
G
A
See
it
you
got
to
get
it
yeah
and
to
get
the
budget.
I
was
going
to
ask
one
one
other
thing
here:
do
we
so
so
just
to
be
clear,
this
code
that
I
linked
here
it
makes
it
so
that
sub
process
will
not
sub
processes
will
not
be
auto
instrumented.
A
F
F
I'll
I'll
say
it
is
the
the
difference
between
manual
and
automatic
instrumentation
is
that
chromatic
instrumentation
is
more
convenient
in
the
sense
that
it
allows
you
to
introduce
less
changes
to
your
code,
but
at
the
same
time,
is
something
that
gives
you
less
control
on
how
these
conditions
are.
In
that
sense,
I
don't
expect
bonus
limitation
to
work
on
absolutely
every
situation
and
that
their
and
that
it
is
okay
that
in
some
situations
you
need
to
do
manual
instrumentations.
F
I'm
not
saying
that
that
is
the
the
that
what
you're
trying
to
solve
is
not
possible
with
our
instrumentation.
Just
saying
that
at
least
I
I
consider
some
scenarios
where
observation
doesn't
work
to
be
acceptable.
A
Yeah
yeah,
I
tend
to
agree,
I
think,
that's
reasonable.
If
somebody
like
really
wanted
to
make
the
auto
instrumentation
work,
I
guess
they
could
hack
their
python
interpreter
to
have
like
a
rapper
script
to.
G
Sub
processors
should
automatically
be
instrumented
or
not.
One
example
be
like
if,
if
like
someone
would
use
like
g
unicorn
or
something
like
in
their
in
their
production
environments,.
A
G
A
I
think
this
might
only
apply
to
things
that
are
getting
exact
right.
So,
if
you're
just
creating
a
sub
process
and
it
just
forks,
it
would
already
inherit
like
the
right
exactly
yeah.
So
I
don't,
I
don't
think
it
applies
there.
G
But
it's
only
if
you
like
explicitly
create
one
through
exec
yeah,
which
is.
A
A
G
But
anyways,
regardless
of
that,
I
agree
with
you
guys
in
terms
of
like
the
limited
behavior.
I
think
that's
the
stance
that
we're
taking
too
for
at
least
for
like
our
product
offering
it's
like.
If
you
want
to
use
auto
instrumentation
it
should,
we
foresee
only
flick
the
for
the
the
regular
use
cases,
if
that
makes
sense
for
autonomous
orientation,
diego's
correct
about
the
loss
of
control.
I
guess.
A
G
Thanks
awesome.
H
Yeah
yeah.
Actually
I
have
a
credit
recording,
stick:
ons
yeah!
Yes,
can
you
go
down
so
she
can.
You
mentioned
that
the
app
in
teardown,
but
in
starlet
there
is
no
an
instrument,
app
function
as
of
now,
so
it's
a
little
little
very
small
change.
So
I
was
just
my
queries.
I
mean.
Should
I
implement
that
on
instrument
app
in
this
pr,
or
should
I
raise
a
new
issue
and
as
a
pr
separately.
C
If
I,
if
I'm
not
wrong
the
instrument,
is
that
part
of
the
base
class
which
is
not
in
the
individual,
instrumentation?
Okay,
so
I
think.
G
Really,
I
think
it's.
E
C
G
C
Yeah
I
mean
the
uninstrument
should
also
take
care
of
the
app
you
also.
H
Whatever
framework
I've
gone
through,
so
basically
it
removes
the
open,
telemetry
middleware
from
the
list
of
middleware.
For
that
framework
and
an
instrument
is
not
so
it's
a
little
different.
I
guess.
C
No
so
there
you
see,
I
I
think
layton
is
the
one
who
added
this
capability,
like
he's
instrumented
by
open
telemetry
check.
Is
there
right
so
this
this
will
be
set
in
in
both
instruments
you
can
underscore
app
like
any
any
kind
of
instrumentation
and
it
will
unset
when
you
do
the
on
instrument.
So
that's
like
that's
what
I
mean
like
they're.
You
know
asking
you
to
uninstrument
it
an
instrument
in
the
teardown.
H
There's
an
instrument
I've
already
added,
but
apart
from
that,
you
have
added
an
instrument
app.
You
ask
for
an
instrument
app,
so
I
thought
maybe
it
is
required
and
and.
C
G
Right,
I
think
where's
an
instrument
app
yeah,
see
like.
G
Instrument
app
is
like
specific
to
certain
instrumentations
like
because
the
app
themselves
need
to
be
on
instrument.
I
think,
sir,
what
you're
saying
is
like
an
instrument
for
the
whole
thing
should
be
able
to
cover.
All
of
that
is
that
what
you're
talking
about.
H
Guess
that
I've
already
covered
in
my
year,
but
I
just
checked
that
in
fast
api
it
is
there
and
it
is
not
that
installed.
So
I
thought
maybe
we
should
add
that.
G
Well,
actually,
there
might
be
something
there
like.
I
have
a
question,
that's
like.
If
we
call
an
instrument,
does
it
uninstrument,
like
apps
that
have
already
been
created?
It
doesn't
look
like
it
from
this.
This
implementation
right.
G
So
if
I
have
a
fast
api
app
right
and
I
do.
G
C
No,
it
also
like
so
so.
We
we,
we
usually
add
the
wrappers
for
the
instrumentation
type,
so
we
remove
them
like
we
try
to
remove
them.
Oh.
C
Didn't
see
this
yeah,
and
also
this
is
like
that's
one
of
the
things
that
it
does,
and
this
is
also
there
too,
like
the
check,
is
there
to
prevent
the
double
instrumentation
like
if
they
by
mistake?
If
somebody
does
that.
G
H
C
So
yeah,
that's
all
right
if
you
are
adding
them
like
so
the
implementation
that
we
use
that
like,
if
you
are
doing
something
of
this
sort
in
the
instrument
app.
There
is
an
instrument,
but
if
you
are
only
just
adding
the
wrappers
that
you
can
change
back
in
the
uninstrument,
so
we
don't
have
the
separate
and
instrument
app
underscore
app
method
are
defined.
C
So
so,
once
you
call
the
an
instrument,
they
should
not
be
sending
the
telemetry,
that's
the
ideal
behavior,
but
if
it,
but
it's
not
possible,
so
to
do
that
or
if
there
was
something
you
know
that
you
try
to
initially
modify
the
app
right.
So
we
undo
that
stuff
for
management.
G
Right,
I
think
I
think
I
ran
into
this
problem
when
creating
this
feature,
I
think
because
it
was
impossible,
like
our
recommendation,
was
to
call
the
uninstrument
app
method
manually
like
that.
That
was
the.
C
Yeah,
so
that
yeah,
that's
why
this
this
has
done
instrument
underscore
app
method,
but
not
all
like
packages
have
this
kind
of
behavior.
So
there
may
be
that's
the
reason
that
maybe
we
don't
have
an
instrument
app
in
our
other
packages.
C
H
So
on
instrument
should
be,
you
know.
G
Yeah,
sorry
to
derail
that
combo
just
wanted
to
make
sure,
but
yeah
flask
has
an
instrument
method.
I
think
I
specifically
added
this,
because
there
are
things
that
we
have
to
modify
on
the
original
app
like
the
tear
down,
requests
and
stuff
like
that.
G
Okay,
anyways
second,
is
your
question
answered.
H
F
Yeah,
sorry
surreal,
I
I
I
didn't
have
time
to
open
a
proper
vr,
okay,
real
quick.
So
if
you
click
on
that
link
up
there
where
it
says
from
the
spec,
oh
well,
it
says
that
implementation
must
aggregate
data
from
identical
instruments
together
in
its
sport
pipeline.
So
what
I
understand
from
that
is
that
if
you
have
two
identical
instruments,
they
pretty
much
should
be
treated
as
one
right
so
that
their
stuff
gets
aggregated
and
not
counted
separately,
yeah.
So
in
this
pr,
what
I'm?
F
What
I'm
doing
is
that
I
am
making
the
the
view
instrument
match
object
to
be
to
be
the
same
when
there
are
instruments
with
the
same
unit,
name
and
description
and.
A
Yeah
all
right,
sorry
to
interrupt
you,
I
just
running
out
of
time.
I
think
this
was
addressed
when
leighton
made
the
pr
that
will
just
return
the
same
instrument
in
these
cases.
F
Well,
I
just
tested
that
we
end
up
creating
two
different
metrics
that
are
not
aggregated
together
when.
F
A
I
agree
with
like
the
behavior,
but
I
think
the
most
like
the
easiest
and
simplest
way
to
handle.
This
would
be
to
just
give
the
same
instrument
back
in
the
meter
when
somebody
requested
an
instrument,
and
it's.
G
You're
saying
you're
still
getting
multiple
measurement.
F
G
Because
different
instruments.
F
Yeah
correct
because
when.
G
F
We
end
up
creating
different
views
when
match
objects,
and
so
on
that
happens
because
a
view
instrument
match
object
is
created
when
an
instrument
matches
matches
with
a
view.
So
if
we
create
different
instances,
okay,
so
nice.
G
Instrument,
if
we
do
like
like
create
or
something
right,
correct.
F
But
but
what
aaron
mentions
is
it's
a
great
idea?
I
hadn't
thought
about
that.
I
can
try
implementing
that
by
doing
that.
I
think
that's
good
too,
but.
G
I
understand
how
you're
you're
running
into
that
problem,
though,
because
it,
the
returning
identical
instrument,
is
only
like
when
we
fetch
the
instrument
but
like
if
we
instantiate
multiple
instruments,
even
if
we
fetch
them
like
it's
still
in
memory
and
then
when
we,
when
we
first
do
the
whole
view
it
instantiate
things.
I
think
it
finds
multiple
of
them.
So
all
right,
so
we
should
instantiate
only
one.
G
Movie,
okay,
awesome:
all
right
seems
like
we're
out
of
time.
So
if
there's
nothing
else,
pressing
I'll
see
you
guys
next
week.