►
From YouTube: 2022-06-14 meeting
Description
cncf-opentelemetry meeting-2's Personal Meeting Room
B
B
It
is
the
support
rotation,
that's
exactly
what
it
is
and
I
always
feel
kind
of
bad
because
I'm
like
I
don't
most
of
the
time.
I
don't
really
know
the
answer
to
these
questions.
I'm
like
well,
I
guess
I'll
help.
You
take
a
look.
You
know.
B
C
C
I
feel
like
now
we're
a
startup
that
has
like
a
bigger
bank
account.
Yeah,
there's
there's.
Definitely
like
a
lot
of
hiring
and
a
lot
of
stuff
just
like
going
on,
so
I
feel
like
yeah
yeah,
the
high
growth
phase
of
the
startup,
and
maybe
maybe
eventually
things
will
calm
down
and
we'll
have
have
what
we
want,
but
yeah.
B
C
C
C
The
only
thing
I
can
tell
you
is
they
were
quibbling
over
whether
the
error
message
should
be
optional
or
not,
and
I
think
in
the
quibble,
the
the
quibbles
that
it
should
just
not
be
optional,
just
send
an
empty
string.
If
you
don't
have
one
for
string
fields,
ludmila
was
asking
to
merge
the
attribute
requirement
levels
which
we
talked
about
a
little
bit
before,
and
somebody
had
merged
it
by
the
time
we
got
to
it.
C
C
In
this
fact,
like
attributes
can
be
like
an
array
of
primitive
types
or
a
map
of
key
value,
pairs
of
of
those
types
and
the
the
protoss
kind
of
allow
for
attributes
to
always
contain
these
nested
maps,
but
there's
some
spec
language
that
at
least
for
tracing
the
tracing
and
metrics.
It's
disallowed,
it's
technically
kind
of
here
for
logs.
C
So
I
think
I
think
there
might
be
some
rough
edges
in
the
spec
around
this.
I
don't
think
anything
is
changing.
I
hope
nothing
is
changing
around
that,
because
there
are
good
reasons
for
not
allowing
this
nesting
in
in
tracing
and
n
metrics.
A
C
Yeah,
all
right
so
tristan
has
a
pr
for
some
reason:
the
spec,
like
kind
of
forbids
x
exporters
or
the
export
methodology
exporter
from
being
called
concurrently
kind
of
explicitly.
Has
this
synchronous,
synchronous
mandate
here,
but
tristan
was
trying
to
just
write
some.
D
C
That
relaxes
this,
without
changing
without
changing
behavior.
C
C
And
this
was
the
discussion
that
kind
of
took
up
the
most
of
the
meeting
is
to
figure
out
what
needs
to
be
done
to
declare
otlp
has
1.0
stable,
so
this
would
include
json
stability,
amongst
other
things.
I
think.
C
Yeah,
I
think
I
think
you
are
interested
in
doing
this.
I
think
they
just
want
to
clearly
state
what
things.
What
does?
What
does
stable
mean
like
what
is
actually
a
breaking
change,
and
I
think
that
was
like
one
big
debate
and
I
think
the
clearest
suggestion
was
a
breaking
change
should
be
anything
that
changes
the
wire
format.
C
But
there
are,
I
guess,
like
a
number
of
changes
that
can
there
yeah
the
thing
that
was
complicating.
This
is
that
some,
depending
on
your
code,
generator
and
language,
it
seems
like
you
could
have
some
changes
that
don't
change
the
wire
format,
but
are
would
technically
be
like
breaking
changes
to
your
generated
code
and.
C
That
was,
I
think
that
was
where
things
started
to
get
complicated
and
I
think,
there's
definitely
some
viewpoints
from
from
the
go
perspective
that
the
generated
code
versions
should
kind
of
match
up
with
the
protocol
version.
Although
I
think,
towards
the
end,
people
were
very
much
against
that
as
a
goal
or
just
as
it
seemed
like
it
was
an
unnecessary
limitation,
but
yeah.
Like
one
thing,
I
as
a
concrete
example,
I
guess
is
feel
the
optionality.
If
you
like
change
a
field
to
make
it
like
optional.
C
It
does
not
change
the
wire
format
and
for
for
javascript,
for
example,
it
wouldn't
actually
even
change
the
generated
code,
but
for
go.
It
would
so.
That
was
like
a
huge
discussion,
but
I
think
ultimately,
people
kind
of
arrived
at
at
least
that
suggestion
anything
that
would
change
the
wire
format
would
be
incompatible.
C
And
then
the
last
thing
that
we
made
it
to
is
there's
a
profiling
somewhat
exciting
and
news
to
me,
I
guess,
is
that
I
think
there's
like
a
profiling
sig
that
is
spitting
up
and
there
is
a
channel
hotel
profiles
in
cncf
slack.
C
C
Yeah,
so
I
think
the
one
of
the
key
elements
is
getting
a
event
type
to
represent
this
profiling
and
I
think,
there's
kind
of
like
these
events
being
built
on
top
of
logs.
I'm
not
sure
if
the
profiling
is
being
built
on
top
of
the
events
that
are
being
built
on
top
of
logs,
but
something
like
that
is
probably
happening.
C
Scrap
that
comment
I
don't
know
where
I
was
going
with
it,
but
yeah.
I
did
at
least
see
this
comment
in
here
with.
F
Yeah,
it
looks
like
they've
actually
got
one
that
integrates
with
the
hotel
go,
so
it's
kind
of
like
their
proof
of
concept.
It's
very
cool.
A
B
B
I
was
that
got
me
thinking
like.
Oh
man,
we've
got
all
these
like
three
years
of
issues
or
whatever
it
is
built
up
in
hotel,
ruby
and
like
do.
We
want
to
migrate
the
existing
issues
to
hotel
ruby
contrib,
or
do
we
want
to
have
like
a
policy
for
which
issues
go
where
moving
forward
and
just
deal
with
the
issues
being
in
a
different
repo
that
might
be.
My
vote
is
like
from
now
on.
We
get
but
curious
if
others
have
thoughts.
C
Definitely
thumbs
up
on
from
now
on,
categorizing
them,
as
as
where
they
should
go.
I
guess
my
only
other
thoughts
is
how
clear
is
it
which
ones
should
be
in
which
repo
and
how
hard
is
it
to
actually
move
stuff
over?
It's
like.
E
F
F
That's
where
I
would
lean
towards
like
new
stuff
here
and
and
and
oh
and
things
that
are
still
open
or
in
progress
to
move
over,
because
we
have
to
move
over
some
of
the
open
pr's
anyway
and
get
folks
to
change
their
upstream
and
where
they
want
to
contribute
it.
So
you
have
to
communicate
a
process
to
people
to
update
those
two
as
well.
B
I
they
made
it
thanks
github,
you
made
it
so
much
easier
to
transfer
issues
from
one
repo
to
another.
This
is
so
much
better
than
the
last
time
I
tried
to
do
this.
I
can.
I
can
move
those
issues
over
almost
anyone
else.
Just
really
wants
to.
B
F
The
one
thing,
the
test
after
merging
that
pr
pr's
releases
and
making
sure
that
releases
actually
work,
because
it's
kind
of
like
fingers
crossed
that
everything
would
do
the
right
thing.
I
don't
know,
like
the
organization
secrets
actually
haven't,
looked
at
them
to
see
if
they're
still
there
and
we're
like
we
have
access
to
them
in
some
way.
F
E
I
can
check
where
would
that
be?
I
never
had
to
look
at
that
before.
F
Click
on
settings
and
then
down
by
security,
there's
a
in
the
security
section.
Second
to
the
bottom:
it's
a
secrets
and
then
it'll
have
action
secrets
and
right
now
I
don't
have
org
secrets
or
repository
or
environment
secret
set
for
our
actions.
E
Okay,
I
have
enable
releases,
get
user
email
get
username,
rubygems
api
key,
I'm
assuming
at
least
all
of
those
are
probably
required.
F
Yeah,
because,
if
we,
if
we're
going
to
continue
to
use
those
same
keys
and
contact
info,
that's
stuff
that
we'll
need
to
add
to
this
repo
in
order
to
enable
releasing
okay.
E
E
F
Or
you
know,
we
may
want
to
generate
new
secrets
specifically
for
this
repository,
because
you
know,
as
security
people
would
say,
don't
share
stuff.
E
Yeah,
honest
honestly,
I'm
not
really
familiar
with
like
this
part
of
jim
release
process.
It's
usually
like
I'm
leaning
on
something
that's
already
established.
So,
okay.
E
F
Right
so
the
the
rubygems
key
is
going
to
be
the
authorization
of
whatever
user
we're
using
to
log
into
rubygems
to
try
to
do
stuff.
So
if
we've
got
like
an
organization,
there's
opentelemetry.org
user
or
something
and
and
then
then
they're,
all
someone
will
have
to
provision
the
key
for
us
if
they
don't
want
us
to
have
direct
access
to
it.
Okay,.
E
B
Yeah
super
quick,
I
been
reviewing
the
pr
for
contrib
I
was
like
instrumentation
might
have
docs
and
we
might
want
to
publish
that
on
the
open,
telemetry
website.
I'm
not.
I
don't
think
this
is
necessarily
a
good
idea
or
a
bad
idea.
I
feel
like
having
it
in
the
gym.
Repo
is
totally
fine.
B
I
just
wasn't
sure
if
there
was
like,
if
other,
I
guess
I
could
have
done
five
minutes
of
work
and
checked
to
see
if,
like
other
languages,
instrumentation
packages
were
publishing
to
the
hotel
websites,
but
was
curious.
If
people
thought
this
was
good
or
bad.
C
So
right
now
we're
just
publishing
like
rubydoc.info.
This
is
true,
and
I
I
think
that
yeah,
I
think
that
the
I
think
people
would
be
really
happy
if
we
were
auto
publishing
some
stuff
to
the
doc
site.
If
that's,
if
that's
a
possibility.
F
F
That
we
need
like-
and
I
would
say
yes
I
I
you
know
we
just
I
don't
think
we've
ever
sat
down
and
described
what
we
want,
that
experience
to
be.
F
F
Can
we
derive
that
from
the
builds
like
there
was
some
chit
chat
between
rob,
rob
kid
and
myself
some
time
ago,
when
we
were
first
kind
of
like
fleshing
this
out
and
our
first
iteration
was:
let's
just
try
to
get
something
out
there
and
tell
people
to
use
the
all
gem
and
then
we'll
iterate
on
it,
and
then
hopefully
it's
like
a
stone
soup
thing
where
people
are
interested
and
they'll
want
to
add
some
more
color,
and
so
it
looks
like
you're
bringing
the
carrots
man,
because
because
we
just
haven't
gotten
around
to
it
right.
F
G
Some
of
it
could
some
of
it
could
could
just
be
one
of
the
things
that
ariel
and
I
did
was
experiment
with
just
writing
something
not
for
the
doc
site,
but
for
a
readme
and
then
so
that
it
gets
written
somewhere
and
then,
as
that
content
with
like
multiple
eyes,
and
this
group
can
go
like
yeah
all
that
makes
sense,
and
then
you
try
it
with
somebody.
Somebody
actually
tries
to
follow
those
instructions
and
confirms
that
they're
they
hold
together.
G
Okay,
then
that
stuff
can
get
copied
or
or
something
into
the
documentation
that
shows
up
at
open
telemetry.
I
o,
I
think,
some
of
the
developer
level
docs.
We
could
lean
on
the
readmes
for
individual
instrumentation
gems.
I
know
that
sometimes
I
reach
and
point
customers
to.
Oh
the
rack
instrumentation.
You
want
to
filter
out
some
of
the
endpoints
I
go
to
the
readme
and
the
readme
tells.
E
G
G
G
G
G
To
be
used
by
humans
that
aren't
open,
telemetry,
ruby
developers,
but
users
of
the
instrumentation
or
the
api
or
the
sdk,
the
the
methods
and
classes
in
there
that
we
would
expect
humans
to
use
versus
the
private
stuff
would
have
maybe
some
good
api
level
documentation
for
how
do
you
use
this
method
and
some
example
talk
some
example,
usage
and
stuff
baked
right
there
in
with
the
code?
F
I
mean
no
one's
going
to
say.
Oh,
we
do
have
an
open
issue
of
a
bunch
of
things
that
we
had
brainstormed
at
some
point
that
we're
kind
of
missing
or
a
little
lacking
like
you
know,
it's
like
very
minimal,
and
this
was
a
many
moons
ago
we're
talking
two
years
ago
at
this
point
that
we
had
talked
about
it
and,
like
you
know
life
is
you
know
what
it
is
man
so
775
right
there.
F
I
actually,
I
posted
it
in
the
in
the
slack
and
not
in
the
slack
room
so
that
for
for
faster
reference
or
whatever,
but
like
here's
all
the
stuff
that
you
know
we
wanted
to
like
get
in
there
and
so
rob
and
I
kind
of
like
brainstorm.
But
what
do
we
want?
The
doc
like
that
initial
experience
of
getting
started
quickly,
as
you
mentioned
already,
and
then
you
know
after
we
released
that
we
were
kind
of
like
well,
we
did
it
so
all
done
quick.
H
B
Cool,
well,
maybe
maybe
it's.
I
can
just
take
a
look
at
the
that
issue
and
kind
of
go
from
there,
but
it
sounds
like
people
wouldn't
be
angry
if
I've
made
some
pr's
to
do
some
ducks.
This
is
really
what's
going
on.
B
H
Okay,
I'll
try
to
do
that.
H
G
I
I
can
think
in
in
the
time
that
I
said
I
can't
think
of
anything
auto-generated
that
that
would
be
premature.
I
can
see
a
world
where
the
the
doc,
that
is
the
quick
start
that
says,
throw
this
snippet
into
your
gem
file,
so
that
your
dependency
resolve
would
maybe
have
version
numbers
updated.
If
we
even
put
version
numbers
in
there.
G
But
my
gut
tells
me
like
if
we
were
to
go
and
and
just
judge
the
readmes
for
individual
gems,
that
those
tell
you
a
little
bit
more
target.
We've
got
a
lot
of
boilerplate
in
there
and
if
we
could
add
to
it
what's
different
about
this,
this
particular
instrumentation
library
versus
another
one
would
be
super
helpful.
When
people
want
to
start
fiddling
the
knobs
on
the
right.
F
I'm
sure
we
could,
if
we
could
accommodate
some
tech
writers
to
help
us
with
this
process,
because
that's
the
biggest
that's
the
hardest
part
right.
It's
like
figuring
out
of
this.
Even
if
you
look
at
the
open,
telemetry,
io
docs
and
you
look
at
the
how
documentation
is
organized
and
written
per
instrumentation,
there
is
not
a
unified
voice
whatsoever.
F
And
you
could
say:
oh
because
one
language
is
different
than
another.
No,
it's
like.
We
just
lack
the
skill
set
of
being
able
to
write,
documentation,
I'll,
say
it.
Okay,
hopefully.
F
E
C
Yeah,
I
guess
the
only
thing
I
say
about
this
is
it
would
be
nice
to
auto,
generate
some
stuff
and
whatever
kind
of
framework
we
come
up
with,
it
would
be
nice
if,
like
I
don't
know,
whatever
things
we
decide
are
important
things
that
we
would
like
to
be
on
like
the
doc
site.
If
it
could
also
be
on
the
read
me,
and
if
the
thing
that
manages
this,
this
could
generate
a
doc
that
could
go
in
both
places.
That
would
be
super
awesome,
but.
C
It
seems
like
whatever
we
want
to
like
generate
things
off
of.
Maybe
we
will
end
up
with
some
sort
of
template
somewhere
along
the
line
with
like
the
critical
documentation
and
like
if
we
could
embed
that
in
the
readme,
so
that,
like
that
information
is,
is
in
the
readme
somewhere
and
then,
if
that
same,
information
can
go
to
like
the
doc's
site.
Just
that
stuff
is
staying
in
sync
and
we're
not
maintaining
like
multiple
copies
of
kind
of
like
the
same
thing
in
the
end,
but
right.
C
G
I
could
I
could
foresee
if-
and
this
is
like
a
hypothesis
I
can-
I
could
see,
for
instance,
probably
the
initialization
of
your
opencelebratory
ruby
configuration
so
like
initializing,
the
tracer
in
your
app
is
the
quick
starter.
How
how
do
you
get
things
installed
and
then
most
of
the
documentation
is
about
manual
instrumentation
and
configuring,
the
auto
instrumentation
yep
and
the
we
could
start
with?
G
This
is
my
hypothesis,
I
don't
know
if
it's
correct,
but
the
theory
is
that
we
could
our
doc
or
some
code
documentation
and
comments
format,
the
appropriate
places
for
manual
instrumentation
and
the
configure
call
for
the
instrumentation
so
that
those
little
situated
near
the
code
that
changes
when
you
add
new
configuration
options
is
how
do
you
use
this
thing
and
then,
if
we
use
the
standard,
our
docky
type,
docs,
api,
docs
generator,
and
that
goes
somewhere.
E
G
Right
versus
the
like:
here's,
how
you
install
gems
and
config
and
like
have
a
sort
of
standard
configuration
before
without
nuances
right
is
a
place
to
start
at
least
like
a
happy
path.
That
gets
you
everything,
but
you
aren't
really
customizing
stuff
and
then,
if
you
want
customizations,
it's
links
to
the
things
that
change
over
time.
G
Yeah
that
makes
sense
to
me,
and
then
we
can
see
how
that
fails,
then
figure
out
what
else
we
need
to
write.
Yeah.
B
C
B
Oh,
you
know
you
can
check
off
stretch
coal
contributed
micro
service
to
hipster
shop
andrew.
Did
that
didn't
he
there
you
go.
F
D
F
G
Excuse
me,
I'm
digging,
I'm
digging
the
microservices
in
the
demos,
because
that's
super
helpful
to
point
people
towards
and
say,
like
here's,
a
here's,
a
thing
you
could
spin
up
and
play
with
here's
the
thing
that
has
demonstrated
how
to
do's
it's
it's
good
stuff.
F
Here
we
got
some
open
pr's.
A
lot
of
this
stuff
is
kind
of
like
we're.
Gonna
do
like
this
freeze,
where
we're
holding
off
on
some
of
the
stuff,
because.
H
G
We
had
some
sqs
stuff
come
up
at
work.
That's
the
all
the
details
I
have
in
my
brain,
so
yeah,
I'm
going
to
go,
read
what
we
learned
about
it
and
then
I
might
have
some
comments
for
that
feature
for
the
sqs.
B
Oh
sorry,
oh
I
forgot
this
on
youtube.
I
think
I
believe
so.
I
believe
that
he
no
longer
works
at
expected
aspect.
G
All
right
well
I'll,
see
what
I
can
do
and
see
if
what
we
learned
about
sqs,
I
think
in
some
other
context,
I
think
a
different
language
hotel
implementation,
the
instrumentation
for
it
maybe
js.
Sorry.
This
was
a
useless
comment.
I
might
have
some
follow-up
to
the
sqs
pr
to
help
us
bring
that
home
or
not.
C
Yeah,
I
think
I
think
that's
useful-
I
think
amir
who
is
still
with
aspecto
is
definitely
would
like
to
work
with
us
to
get
this
to
pull
through
to
the
finish
line.
Yeah.
F
So
I
think
we've
also
had
users
that
come
into
the
channel
and
ask
us
about
it,
because
they've
been
running
into
trouble,
where
they're
seeing
traces
dropped
for
their
lambda
functions
and
they're
trying
to
continue
to
continue
the
traces
from
sqs.
So-
and
I
apologize
for
interrupting
that,
so
it's
not
just
them
that
want
it
right.
C
C
So
yeah,
I
think
the
consensus
was
pretty
positive
on
this.
At
last
we
looked
at
it.
I
think
we
just
need
to
if
I
recall
where
we
were.
This
is.
C
This
is,
this
is
adding
everything
has
links.
I
think
it's
adding
everything
as
links
or
adding
everything
as
children,
one
of
those
and
we
support
the
other
two
options
and
the
consensus,
as
I
remember
wrote,
was
to
probably
merge
this
as
is,
and
we
could
add,
support
for
the
the
other
two
if
needed
did
I
have
that
right
is
this?
Is
this
using
links
right
now?
G
H
C
We
can
talk
to
amir
about
tests.
I
think
my
question
would
be
how
how
comfortable
is
amir
writing
tests
in
ruby?
B
F
F
G
E
G
I
was
about
to
say
we
could
we
could
push
this
branch
to
as
as
one
of
us,
we
could
push
this
branch
to
the
contributor
close
this
pr
say
that
it's
reopened
over
there
and
then
we
have
some
control
over
it,
and
I
can't
imagine
amir
wouldn't
mind.
B
B
But
it
could
be
a
good
way
to
test
out
our
our
migration
to
the
contributor
has.
G
So
we
don't
want
to,
we
would
want
to
merge
it
there
anyway,
right
now
so
yeah,
let's
experiment,
let's
push
this
branch
to
the
trip,
repo
open
it
up
as
a
different
pr
that
sam
owns
and
and
we'll
swarm
around
it.
There.
F
B
F
Like
we,
I
pushed
the
I
I
basically
overwrote
the
contributory
bow
with
this,
the
history
of
this
and
then
added
a
pr,
and
I
gotta
fix
like
these
links
in
the
thousand
files
that
I
keep
misspelling
because
final
replays,
I'm
so
stupid.
G
F
H
F
G
It's
like,
then,
our
value's
done
people
curse
here,
because
sometimes
curses
are
needed.
G
Expressing
yourself
in
a
colorful
way,
sometimes
sometimes
your
scent
needs
some
spice.
I
F
G
F
I
I
F
F
Seen
so
many
animated
films
where
the
birds
do
have
arms
like
this,
like
like
angry
birds
and
whatnot,
don't.
F
B
G
G
There's
a
good
trick:
chloe
condom
has
put
a
little
emoji
in
your
name
and
then.
G
F
I
like
it
when,
when,
when
the
systems
don't
support
ucfa
and
before
so,
oh.
G
H
F
Anyways
all
right
well,
are
we
gonna,
do
this
parkinson's
law
thing
or
are
we
like
we're
just
hanging
out
until
we're.
F
F
Other
than
like
the
the
unhappy
reports
of
like
you
know,
struggling
through
like
trying
to
get
those
releases
out,
but
adding
new
instrumentations
for
the
monolith
has
been
working
out.
Okay,.
G
F
F
Yeah,
because
we
manually
try
to
control
volume
and
like
what's
getting
recorded,
yeah.
G
We
have,
we
have
hopes
and
dreams
for
improving
the
tail
sampler
in
the
collector,
so
that
so
that
you
can
do
more
auto
instrumentation
and
not
have
to
dial
it
back
in
your
auto
instrumentation
and
have
some
more
central
rule
sets.
F
When
I
was
looking
at
the
at
the
tail
sampler
and
in
particular
like
the
ones
that
are
super
interesting
to
me,
are
going
to
be
like
attribute
filters.
So
like
one
of
the
attribute
based
samples,
because
we
have
these
situations
where,
like
one.
D
F
Our
staff
members
right
wants
to
try
to
debug
something
in
that,
and
it's
like.
I
really
want
to
capture
a
trace
specifically
because
I'm
this
user
or
be
able
to
turn
on
a
feature
flag
and
say
like
enable
the
trace
for
a
specific
customer
like
sample
100
of
what's
going
on
with
them,
because
they
fit
a
specific
kind
of
profile
that
called
wreaks
havoc,
I
feel
tempted
to
say,
like
also
for
error
cases,
but
I
feel
like
that's
just
to
skew
the
results
so
bad
downstream,
because
it's
like.
F
Oh,
I
want
to
capture
100
error
traces.
So
then
my
error
rate
is
like
100
all
the
time.
G
Yeah,
unless,
unless
the
thing
that
you're
sending
it
to
knows
the
sample
rate
of
your
successes
and
can
mathy
math
representative
populations
and
say
I
I
only
receive
one
out
of
100
successful
thing,
I
received
every
error
thing,
but
if
your
back
end
knows
that
you
had
this,
one
represents
100
successes.
It
knows
how
to
do
math
to
tell
you
what
your
actual
error
rate
is.
Yeah.
G
G
F
So
much
fun
yeah,
you
know
so
happy
reports.
Is
that
that's
like
looking
good?
What
else
can
I
tell
you
about
happy
reports.
G
I
had
I
had
a
user
pleased
that
the
the
the
rack
instrumentation
library,
upon
which
most
of
the
other
stuff
is
based,
has
a
little
like.
No,
please
don't
make
me
spans
for
the
health
check
or
routes
right.
The
untraced
endpoints,
I
think,
is
the
configuration
option
on
that
instrumentation.
G
So
all
two
of
our
viewers
since
you've
liked
and
subscribed
know
that
you
can
tell
your
rack
instrumentation
and
not
be
so
chatty
about
endpoints.
F
One
thing
I
wished
I'm
trying
to
think
of
like
the
one
thing
that
I'm
wishing
for
so
after
enabling
active
record
in
action
view
and
the
monolith,
which
is
been
helpful,
I'm
still
yearning
for
more
details,
because
you
know
where
it's
right
now.
F
It's
only
doing
fine
by
sql,
and
so
I
would
love
to
see
a
little
bit
more
if
we
had
time
or
capacity
for
us
to
look
into
adding
more
instrumentation
around
active
relation
proxy
or
potentially
to
the
adapter
to
give
us
a
little
more
color
into
it
so
like
at
github.
What
we've
got
is
a
situation
where
we
got
multiple
databases,
primary
and
secondary
clusters.
F
So
our
setup
is
a
little
bit
complicated
right
and
we're
missing
some
color
as
to
like
hey
look.
This
particular
query
is
hitting
this
specific
net
pier,
but
in
your
in
the
rails,
configuration
there's
like
this
logical
host.
Essentially,
so
you
have
like
a
logical
host
which
then
eventually
resolves
to
a
net
peer,
which
is
either
going
to
be
a
primary
or
secondary,
or
it's
going
to
be.
F
You
know
basically,
a
read
replica
or
write
replica
of
that
particular
note,
and
that
information
is
lost
in
the
sql
span
has
to
be
deemed
as
a
read-only
or
a
writing
a
reader
or
a
write
replica.
F
G
Have
you
looked
at
the.
G
At
the
stuff,
that's
available
on
the
active
record,
active
support,
instrumentation
keys,
where
you
can
get
the
connection
object,
you
can
get
or
is
it
or
you're
not
going
to
get
it
in
the
connection
of
it?.
F
You
can
get
it
from
there,
but
then
you
now
we
play
a
game
about
what
span
is
gonna
get
annotated
with
those
attributes,
so
you
don't
have
access,
because
when
notifications
occur
runs
the
event
has
completed.
F
So
if
I
want
the
original,
if
I
want
to
enrich
the
original
auto
instrumentation
from
the
trilogy
gem,
it's
a
little
bit
too
late
because
I
don't
have
access
to
it,
because
the
current
span
is
not
in
scope
in
that
ar
notification.
That
span
is
closed
and
it's
popped
off
little
bit
up
in
the
stack.
E
F
An
active
record
it
could
be
in
the
database
driver
instrumentation
or
you
know
I
don't
know
depending
on
the
execution
scope.
If,
if
there's
no
instrumentation
for
active
record,
it
might
get
appended
to
the
the
rack
span
right.
So,
if
I
add
an.
F
Span
and
I'm
like
oh,
I
got
this
additional
span.
There's
no
way
to
continue
the
contacts
propagation
between
those
yeah.
G
B
For
real
yeah,
I
don't
know
man,
oh
my
god,
what
an
exciting
adventure
I
haven't
decided.
F
F
You
know
filet
mignon
with
a
little
bit
paul.
You
know
potatoes
all
groton,
you
know,
but.