►
From YouTube: 2022-01-25 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
Yes,
aaron
brought
up
in
the
jvm
metrics
meeting
today
or
yesterday
that
wanted
to
chat
about
sort
of
where
instrumentation.
A
Should
live
sort
of
the
difference
between
libraries,
natively
instrumenting
themselves
versus
the
instrumentation
repo
that
we
have
versus
applications?
A
I
think
that
the
micrometer
is
kind
of
wanting
to
I'm,
not
quite
sure,
that's
why
it
was
at
the
end
of
the
meeting.
So
I
checked
if
she
could
join
on
thursday,
so
that
we
could
have
more
time
to
discuss.
A
A
That,
well,
I
think
it
has
hermitage.
A
I
think
it
has
something
to
do
with.
I
know
they
were
planning
micrometers
planning,
otlp
exporter,
so
I
don't
know
if
they're
kind
of
concerned
that
that
that
our
micrometer
instrumentation
could
I'm
just
guessing.
I'm
that
that's
part
of
the
question
there
or
behind
the
behind
the
concern
or
the
reasons
behind
the
question.
A
Mateish
did
you
you
were
there
this
morning
or
yesterday,
yeah
aaron's
question
about
the
micrometer
bridge.
C
A
We
have
a
couple
examples.
I
was
trying
to
think
of
like
lettuce
that
we
pull
in
from
others.
We
have
how's.
A
A
Any
other
we
had
one
example
on
the
supported.
C
And
there's
camo
versions,
three
point
something
they
also
have
built-in
open,
telemetry
support.
A
A
The
couch
on
rock,
who
is
you
were
working
with
somebody
at
one
point
or
there
was
that
couch
base
or
lettuce.
That
was
one.
A
A
A
C
B
A
A
Oh
that's
cool
yeah
I
was
thinking
we
would
have.
I
was
thinking
we
had
to
sort
of
build
a
feature
that
would
say
go
and
test.
B
A
Yeah,
I
would,
I
would
be
in
favor
of
testing
like
every
like
the
major
first.
B
A
Yeah
there
have
been
a
couple
of
cases
where
one
is
one
happened,
kind
of
recently
matish
it
was.
It
was
one
of
the
http
client
libraries
that
something
or
maybe
it
was
honorable,
maybe
you
were
it-
was.
A
B
B
Yeah
I
mean
if
we
wanted
to
make
it
a
pattern
where
we
do
test
the
this
point:
zero
or
whatever.
We
would
probably
want
to
add
some
tooling
to
make
it
easier
than
this
test
set
stuff
so
that
it's
like
right
now
we
have
latest
desktop
library.
Maybe
we
can
just
have
some
other
similar
method.
We
need
to
find
the
version
number
in
there
as
well,
and
it
sets
up
the
configurations
automatically
not
high
pretty,
but
we
can
do
at
some
point.
B
B
A
That's
sort
of
the
only
way
to
so
I
guess,
but
the
question
yeah
it's
a
good
question
is
what
the
difference
with
muzzle
is.
I
guess
this
is
one
way
to
find
out.
C
Oh
yeah,
I'm
in
progress
of
the
http
root
story
and
I
pushed
a
new
pr
yesterday
that
renamed
the
servers
from
naming
class
to
http
with
holder.
I
think
so.
You
can
review
this
one,
and
there
is
one
thing
that
I
wanted
to
just
discuss
here.
We
still
have
the
http
in
in
the
http
attributes:
extractors,
there's
still
the
root
method,
so
I
I
was
thinking
of
removing
it,
even
though,
in
some
cases,
almeria
and
maybe
one
more,
it
will
be
actually
a
little
bit
more
painful
to
call
the
holder.
C
Instead,
you
know
just
extracting
the
route
that
is
available
at
the
start,
because
it's
probably
a
bit
simpler.
If
we
have
exactly
one
way
of
setting
the
route,
so
you
don't
have
to
think
whether
oh
should
I
get
it
in
the
extractor
or
should
I
use
the
root
holder
or
what
you
have
to
use
the
http
holder
to
send
it
so
that
that
was
my
idea
about
it.
B
B
B
B
A
Consumers,
so
what
is
it
with
army?
Is
it
that,
when
like
because
a
lot
of
times,
I'm
trying
to
think
of
like
say
spring
like
say
spring
wanted
to
natively
use
the
instrumentary
api
and
they,
even
in
that
case,
there's
still
like
a
earlier
point
in
time
at
which
the
request
starts
the
request
life
cycle,
and
then
the
route
is
calculated,
maybe
near
the
beginning,
but
not
at
the
beginning.
A
B
A
A
So,
do
we
not
capture,
so
does
nettie
not
does
our
netty
instrumentation
doesn't
capture
the
army
area
requests.
C
B
B
A
C
B
Don't
want
to
get
rid
of
many
instrumentations,
but
I
don't
know
if
I'm
it
doesn't
work
with
http
2,
so
yeah
yeah
yeah.
I
I
don't
know
what
the
user
expectations
are
in
these
cases
I
mean
they
want
to
trace
their
business
logic.
I
think
I
don't
know
if
they
care
that
much
about
what's
happening
within
our
media
like
it
could
have
been
rpcs
or
something
but
just
standard
processing
of
the
request
into
root.
Maybe
that's
not
important
anyways.
B
C
Actually
javadoc
on
the
yeah
I'll,
probably
leave
that
for
the
last
as
make
it
the
last
item
in
my
to-do
list,
because
it
also
contains
like
adding
a
non-hacky
solution
to
http
server
metrics,
because.
C
No,
it's
not
actually
about
that.
It's
about
attributes
extractors
not
having
access
to
context,
so
you
can
in
attributes
extractor.
You
can't
get
the
root
from
the
root
holder.
So
if
I
were
to
add
the
root
2
matrix,
I
would
have
to
hack
it
around
in
http.
Server
matrix
like
use,
use
the
http
root
holder
as
another
source
of
attributes,
with
one
attribute
to
be
more
correct.
A
B
C
Yeah,
so
I
will
probably
leave
the
extractor
stuff
for
the
end
of
this
story.
A
Oh
honorary
and
I
did
review
this.
B
A
Oh,
I
I
thought
this.
This
is
even
yeah,
whether
let's
do
some
theoretical
escape
analysis-
theorize
yeah
yeah
on
this
one,
so
I
had
definitely
seen
some
I
hadn't
I
had
I
done
that
up.
I
did
this
optimization,
but
I
did
it
for
both
the
1r
again
2
arg.
A
At
the
same
time,
I
yeah
it
would
be
interesting
if
you
want,
I
can
add,
a
task
to
to
benchmark
my
two
arg
since
that's
not
super
pretty,
it
would
be
nicer
to
just
have
a
one
argument:
yeah,
I'm
not
optimistic.
I
take
a.
B
A
Of
quick
look
at
the
size
of
the
method,
that's
in
that
http
route
object
that
would
need
to
be
inlined
and
it's
a
good
size
plus
the
escape
analysis.
B
A
I
don't
mind
trying
to
throw
a
benchmark
in
there
to
justify
a
weird
api
hack.
A
A
B
B
A
Yeah,
I
think
one
reason
why,
because
I
looked
at
the
this
was
writer
a
little
after
I
had
committed
to
jfr
profiling
stuff,
and
so
I
think
one
reason
why
this
might
have
popped
out
in
those
profiles
is
that,
like
for
the
servlet
one,
we
call
that
a
lot
right,
mateish
yeah.
A
C
C
A
B
B
B
A
B
B
Instrumentation,
it's
it's
like
probably
a
daily.
I
would
do
it
except
right.
A
B
A
A
B
B
A
B
Right
we
do
like
it
does
sound
familiar
that
our
pattern
is
to
propagate
on
a
subscribe
time,
not
on
creation
time
right,
yeah.
A
B
A
A
Oh
yes,
I
had
a
question,
came
up
in
the
jvm
metrics
meeting
today
about
getting
approvals
on
jack's
spec
pr,
and
we
were
wondering,
do
you
know
if,
but
I
looked
jack
was
thinking
that
other
groups
like
messaging,
had
gotten
approvers
for
their
specific
piece,
but
I
didn't
see
in
the
code
owners
anything.
A
B
Even
if
it
was
just
that
it'd
still
be
okay,
but
like
I've
noticed
that
approver's
improving,
doesn't
really
speed
things
up,
it
seems
like
then
the
tc
comes
in
and
slows
things
down
from
there
anyways.
So
I
don't
know
how
beneficial
that
would
be,
at
least
that's
my
expert
impression
of
how
things
go
in
that
book.
A
Oh
and
I've
got
I
can,
I
can
wrangle
you
and
you
and
john
yes,.
B
A
I
need
some
green
check
marks
before
I
can
ask
tigrin
to
merge
it.
So
it's
like
there's
a
certain
a
number
of
steps
involved.
B
A
Oh
yes,
I
did.
I
was
curious
what
you
both
thought
of
instrumentation
scope.
I
thought
it.
I
was
surprised
it
got
so
much
buy-in
so
quickly.
I
guess
people
just
hate
the
logger
name
as
a
separate
thing,
and
so
this
was
the
only
compromise,
but
we've
had
instrumentation
name
has
been
so
confusing
for
users
of
what
they
put
in
the
get
tracer
already,
but
it
feels
like
this
just
makes
that
more
confusing.
B
If
this
was
what
our
method
was
named
from
the
beginning,
then
it
would
be
less
confusing.
I
think,
because
the
user
might
understand
what
the
scope
is
better
than
a
library,
but
since
our
method
is
going
to
be
called
set,
instrumentation
library
anyways
at
least
until
we
do
2.0-
then
it's
weird.
I
guess,
because
the
main
problem
for
users
has
been
like
if
they
want
to
make
a
tracer,
they
don't
have
a
library,
so
they
don't
know
what
name
to
put
in
it.
B
A
B
B
B
Like
I
having
the
logger
name
and
and
up
upward
scope
seems
to
fit
well
with
otp.
I'd
want
to
sort
of
compress
things,
because
there
will
be
many
log
messages
from
the
same
logger.
So
it's
nice
that
that's
all
compressed
but
yeah.
I
guess
there
isn't
a
place
for
the
instrumentation.
In
that
case,
it's
true.
A
B
A
B
B
A
The
path
is
going,
the
near
term
path
is
we
have
our
own
like
process
age
out
of
agent
box
agent
that
collects
other
metrics,
it's
sort
of
like
the
collector,
the
azure
version
of
the
collector
and
that's
going
to
have
a
that'll.
Have
an
otlp
ingestion
point.
A
B
B
A
Oh
yeah,
so
this
looks
good
to
go.
I
just
asked
for
one
comment
and
then
I'll
merge
and
try
to
make
I'll
initiate
110
one
tomorrow
with
the
two.
A
So
oh
yeah,
so
I'll
update,
hotel,
sdk
110
one,
and
then
we
have
two
prs
ourselves.
So
I'll.
B
B
A
C
Nice
and
with
the
latest,
the
local
change
the
load.
He
did
it's
actually
very
like
localized,
so
it
won't
be
triggered
by
accident
and.
A
A
Yeah,
that
was
the
last
comment
I
asked
for,
because
it
was
nice
I
and
laura
added,
like
the
essentially
pseudo
code,
that.
B
B
B
B
A
All
right
one
thing:
I
started
thinking
about
a
little
bit
that
is
triaging.
A
A
We
will
necessarily
get
to
sort
of
ever
that
we
would,
you
know,
accept
pr's
for,
but
I
wasn't
I'm
kind
of
torn
on
whether
to
just
you
know,
leave
stuff
tag,
it
leave
stuff
or
if
we
want,
if
it's
worth
closing
you
know,
tagging
and
closing
with
you
know
some
we
could
construct
some
nice
message
about
that.
We
would
accept
prs.
A
Yeah
like
say
this
one,
for
example,
I
mean
until
I
don't
really
know
like
if
it's
worth
until
somebody
gets
the,
is
interested
and
tackles
the
play
stuff.
B
A
Yeah,
I
mean
it's
always
there
in
the
clothes
but
yeah,
maybe
a
backlog
label,
so
that
for
for
us,
like
we
can
filter
out
that
as
far
as
things
that
are
maybe
more
actionable,
something
that
we
would
actually
pick
up.
B
A
B
B
A
B
B
A
Last
couple
days,
if
there
was
then
that's
yeah,
but
just
do
what
mattash
and
I
do,
which
is
said,
the
pr
real,
quick
to
fix
the.
B
A
B
B
A
The
way
you
saw
the
there
was
we
did
that
for
aws.
Recently,
the
aws
sdk,
I
meant
to
ping
you
to
make
sure
you
saw
that.
A
Yeah,
I
don't
know
either,
except
that
it
was
a
new
release
that
broke,
so
we
just
limited.
B
A
B
C
A
Yeah
kind
of
like
a
sporadic
test
failure
issues
just
pack
it
with
yeah.
What
latest
depth?
C
He's
out,
I
think,
he's
back
next
week:
okay
check,
yes,
he's
out
until
the
end
of
the
month,
so
all
right
next
week,
yeah,
he
should
be
joining
us
next
tuesday,.