►
From YouTube: 2021-04-16 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
B
B
C
B
D
E
C
The
only
update
is
that
I
think
the
plan,
so
the
plan
is
that
riley
is
opening
hoping
to
have
a
solidified
api
specification
for
the
new
metrics
api
in
the
next
three
weeks,
three
to
four
weeks
by
the
end
of
may.
Basically,
things
should
be
settled
and
then
the
plan
will
be
that
languages
will
start
implementing.
C
Based
on
that,
I
think
that
what
we're
going
to
do
in
java,
because
we
do
have
current
usages
of
the
existing
api
out
there
in
the
wild,
some
of
them
in
the
instrumentation
repo-
is
that
we
will
basically
copy
our
module
over
to
a
new
module
and
then
rip
out
what
and
basically
use
it
as
the
basis
for
all
the
new
one,
because
logan
thinks
most
of
the
internals
are
going
to
be
reusable,
seem
to
change
some
names
and
do
some
things
like
that,
but
provide
a
a
gentle
route
for
people
who
are
using
the
existing
api
to
then
have
a
chance
to
convert
to
the
new
api.
C
C
C
Just
because
they
have
other
considerations
but
probably
for
go,
excuse
me
go
java.
Python
might
be
something
similar
to
that.
D
Is.Net,
integrating
they're
using
the.net
core.
E
Can
I
can
I
talk
about
this,
so
we
I
have
a
I'm
going
to
call
it
an
agent
extension,
but
it's
it's
an
add-on
to
the
splunk
agent
and
it's
just
experimental.
This
is
a
just
a
prototype
and
what
it
uses
a
component
installer
to
set
up
a
context:
storage
like
a
custom
context,
storage,
and
in
that
context,
storage.
We
track
the
attaching
of
context
and
then
later
we
emit
that
in
the
form
of
log
messages
over
otlp.
E
To
the
collector
and
there's
no
exporter
right
now
for
logs,
so
we
have
this
really
hacked
up.
Grpc
based
our
miria
based
grpc
exporter.
That's
sending.
E
Yeah,
I'm
talking
your
language
now
and
starting
with
this
new
version,
with
the
last
update.
What
I'm
seeing
is
that
very
early
in
the
application
life
cycle,
so
I'm
starting
up
a
just
a
spring
boot
rest,
a
spring
pet
clinic
rest
and
early
in
the
life
cycle.
E
When
the
agent
is
hooked
up,
I'm
seeing
con
context
attaches
that
are
outside
the
scope
of
a
span
and
it's
it's
making
it
all
the
way
to
the
context,
storage-
and
I
guess
I'm
just
curious
if
there's
something
I
should
be
doing
differently
to
prevent
that
from
happening
or
if
that's
just
okay,
and
we
should
just
detect
those
kind
of
edge
cases
and
ignore
the
context.
Or
is
there
a
valid
use
case
for
that?
B
Probably
because
I
punted
on
making
sure
this
override
is
not
in
the
agent
class
loader,
because
I
didn't
expect
any
real
problems
so
what's
happening
is
I
think
the
exporter
is
so
using
the
tlp
expert.
Also.
E
All
right,
no
we're
using
our
own,
I
think
we're
using
the
protobufs,
but
I
think
it's
just.
I
think
we
have
our
own
thing.
I
think
we're
just
making
the
protobuf
object
and
telling
the
armyria
client
to
send
it.
B
E
B
E
B
B
Think
so,
okay
yeah
I
mean,
I
think,
like
I
know
what
we
just
need
to
filter
out,
that
one
class
file
from
the
jar,
that's
basically
what
we
have
to
do
after
we
make
the
change.
You
can
probably
verify
with
snapshots
or
something
yeah
perfect
which
which,
which
class
io.grpc.overwrite
context
storage
override.
E
E
I
mean
yeah,
I
would
not
yeah,
don't
kill
yourself
over
this,
but
yeah
yeah
cool
and
it's
I
mean,
like
I
said
it's
just
it's
it's
being
invoked
with
like
outside
of
a
span
which
is
like
maybe
problematic,
but
then
also
you
know
it's
instrumenting
the
exporter,
stuff
and
stuff
itself
right
and
that's
that's
not
what
we
want.
Yeah.
D
D
D
B
B
B
So
I'd
either
like
the
two
approaches,
I
might
filter
it,
like
I'm
planning
on
filtering
from
the
jar
just
happily,
another
approach
would
be
to
separate
out
the
artifact
for
the
context
storage
override
so
that
I
could
exclude
it
from
the
agent
using
gradle
or
whatever.
Those
are
the
two
approaches,
but
I
think
I'll
just
exclude
the
file
from
the
jar
that
shouldn't
be
too
hard.
B
B
E
Okay,
no
one's
no
one
has
even
built
like
an
experimental
otlp
log
exporter.
Yet
have
they
like?
Do
we
don't
have
that
in
extensions
or
anything
anywhere?
I.
E
B
C
Yeah
and
I
actually
chatted
with
bogdan
about
whether
we
should
update
for
the
when
they
released.
There
was
a
recent
pro
release,
though
I
think
is
there.
A
B
C
I
mean
there
is
that
extension
that
has
a
very
weird
model
that
doesn't
match
anything
else
in
the
sdk
like
it
doesn't
use
the
attributes
like
the
normal
attributes.
It
uses
its
own
implementation
attributes
and
it's
a
little
it's
a
little
squirrely
like.
I
think.
If
we
want
to
really
make
the
logs
sdk
it
needs
a
lot
of
work,
and
I
think
it
has
an
exporter
interface
in
it.
But
it's
not
one
that
I
would
want
to.
B
Okay,
yeah,
I
actually
don't
have
in
my
head.
What
the
latest
state
of
that
folder
is.
I
don't
think
anyone
does
yeah,
but
so
jason
you
want.
Are
you
using
that
the
logs
sdk.
C
B
C
It
should
be
an
sdk
extension
since
there's
no
api
for
it.
It's
purely
just
sdk,
okay,
which
is
almost
certainly
always
going
to
be
the
case
like
there's
we're
not
going
to
try
to
invent
a
new
logging
api
that
makes.
D
C
C
E
C
C
C
E
The
thing
that
we
hacked
up
doesn't
have
the
it
doesn't
have
a
log
processor
or
the
thing
that's
analogous
to
the
span
processor
like
we
don't
have
that
we're
just
as
soon
as
the
log
comes
in.
We
send
it
like
we're
doing
the
dumbest
thing
possible,
so
yeah
I
mean
there
would
obviously
be
need
to
be
a
bunch
of
sdk
type
work,
an
api
definition
yeah.
So
if
you
were
to
add
an
exporter.
B
C
There
is
a
log
exporter
interface
in
there
there's.
C
E
C
Yeah
having
it
be
so
tied
directly
in
on
the
create
method,
is
a
little
weird
like
it
probably
should
be
create
with
a
with
a
should
be
a
builder,
and
it
should
have
options
like
spam
context
or
something
I
don't
anyway.
There's
a
builder
there's
a
bag.
Okay,
it
should
probably
be
in
there
there's
a
bunch
of
stuff.
D
Jason,
are
you
doing
auto
instrumentation
of
loggers
or
are
you
providing
like
logging
appender,
like
log4j
appenders,
that
map
to
it?
We.
E
Had
done
that
as
our
first
approach
and
it
was
just
kind
of
for
our
purposes,
it
was
like
unnecessary
overhead
to
try
and
adhere
to
any
logging
framework
or
logging
api.
So
we
just
build.
We
just
we
just
log
very
stupidly
right.
So
we
just
have
we're
just
logging
strings
and
they
might
have
some
attributes.
E
C
D
E
D
Good
discussions
this
morning-
oh
you
probably
saw
this
already
the
code
kind
of
stuff,
john,
updated
us,
and
then
we
talked
about
the
spring
cloud,
sleuth
and
sort
of
john's
experience,
updating
it
to
110
and
some
of
the
breakages
that
that
we
had
caused
in
the.
D
Or
api
in
the
instrumentation,
the
tracer
api,
I
think
primarily
what
he
ran
into
in
this
update,
was
when
we
moved
from
passing
spans
around
to
passing
context
around.
C
I
don't
remember
exactly
what
they
call
some
tracer
customizer
tracer
parser
request:
parser,
that's
http
server,
request,
partially
known
as
the
equivalent
on
the
on
the
client
side,
any
it
needs
to
be
able
to
update,
span
names
and
add
events,
and
if
you
have
a
builder,
you
can't
do
that.
D
This
this
matches,
though
I
mean
in
our
existing
tracer
world.
This
is
what
we
did
for
other
instruments
or
other
tracers,
where
it
needed
to
do
some
extra
stuff.
We
there's
a
lot
of
these
start
spans
in
all
the
tracers
that
do
some
extra
stuff
recall
before
or
after
calling.
D
Super
but
so
the
discussion
was
sort
of
like
what
can
we
do
to
break
it
less
I
mean
obviously
we're
gonna
completely
break
it
with
the
instrument
or
api,
so
one
I
mean
it
was
certainly,
and
I
don't
think
we
intend
to
do
any
more
work
really
on
the
tracer
api
side.
So
I
think
it's
sort
of
addressed
in
that,
but
there
was
a
good
suggestion
of
like
if
there
was
a
way
that
we
could
build
sluice
in
our
repo
just
so
that
we
can
know.
D
D
And
matej
got
volunteered,
I
think,
by
his
so-called
manager.
B
D
C
B
E
C
Then
grab
the
span
out
of
the
context.
My
point
was
more
that
actually
I
had
more
confusion
on
the
client
tracer
side,
because
the
constructor
for
the
client
tracer
now
requires
a
new.
C
Net
peer
attributes
and
there's
no
documentation
about
what
you're
supposed
to
pass
in
or
how
it's
supposed
to
work
or
why?
Why
it's
even
there?
So
my
main
point
was
hey:
if
we're
going
to
change
stuff,
that's
fine,
but
we
should
document
our
quote
on
quote
public
apis,
so
people
know
what
to
do
with
them.
C
B
D
And
then
we
discussed
what,
because
marcin
had
just
asked
about
the
road
map
sort
of
so
I
wanted
to
get
thoughts
on.
E
D
The
whether
we
can
release
instrument
or
api
one
zero
stable
before
metrics
are
stable,
and
I
think
that
the
people
seem
to
be
on
board
with
that.
It
can
that
we
that
we
can,
that
you
know
we
can.
D
And
the
alternative
is
going
to
be
pretty
painful
to
wait
till
the
end
of
the
year
for
metrics
one
zero.
B
I
wonder
how
well
we
can
shield
the
biggest
problem.
Is
that,
like
the
library
has
to
decide
which
metrics
are
if
it's
an
http
library
it'll
have
http
metrics?
If
it's
database
library
it'll
have
some
different
metrics,
is
there
a
way
to
expose
metrics
without
exposing
it
in
the
api?
I
want
to
think
about
that,
but
I
was
expecting
we
would
have
a
class,
something
like
http,
metrics
or
something,
and
so
the
library
instrumentation
would
have
to
like
add
metrics
new
http
metrics.
Something
like
that
to
indicate
that
it's
an
http
library.
D
B
Oh
because
the
jdbc
instrumentation's
not
going
to
have
http
dot
active
requests.
It
could
I
mean,
like
there's
no
reason
not
the
metric
itself
is
the
same
for
any
library,
but
the
semantic
conventions
are
a
bit
different
because
some
start
with
http,
some
don't,
and
so,
even
though
it's
still
require
recording
the
duration
of
a
request.
B
D
Okay,
but
the
oh
right
so
then
I
see
you
have
to
pass
in
right
because
we
don't.
I
keep
thinking
where
we
have
this
a
trade
instrument
or
per
library,
but
that's
not
how
it
works.
There's
a
set
of
extractors
only
one
http,
but
we'll
have
like
an
http,
client,
instrumenter
or
no
just
the
instrumenter,
and
that
there's
http
client
extractors.
F
B
D
B
Guess
hacking
it
like,
I
don't
think
hacking
into
http
extract
because
extractors
are
for
getting
attributes
but
they're,
not
for
recording
metrics.
At
the
same
time
they
do
get
called
on
starting
on
and
so
technically
for
dimensions
yeah.
But
since
they're
called
in
the
right
time
in
the
life
cycle,
they
could
record
metrics
as
an
implementation
detail
and
that's
something
that
could
work.
So
that
might
be
one
way
to
do
it.
It's
not
great,
but
at
least
at
least
that
would
hide
it
completely.
B
Yeah
yeah.
In
that
sense
we
know
we'll
have
http
metrics
that
api
we
could
probably
commit
to,
even
if
the
details
of
how
metrics
are
defined
change,
because
that
will
be
in
a
way
that
yeah,
I
guess
that's
correct,
even
if
the
word
metric
appears
in
our
api.
That's
okay!
As
long
as
it's
a
very
small
api
yeah,
that's
true!
A
D
Work
to
be
done
to
yeah
see
if
that's
feels
right.
I
don't
exactly
know
what
to
tell
spring
cloud
sleep.
Folks,
though,.
A
D
E
D
E
F
D
Loaders
and
bite
buddy,
and
there
was
an
action
item.
Why
did
we
think
we
needed
to
list
of
library?
Instrumentation
probably
spring
wants
to
know
it,
because.
D
E
D
E
Yeah,
that's
and
that's
what
it
was
and
I
was
like,
but
we
have
it.
We
have
document
not
documentation.
Well,
it's
not.
On
the
this
morning
before
I
was
fully
awake.
I
said
that
it
was
on
the
main
dock
site,
but
it's
not
really.
What
I
was
referring
to
is
the
fact
that
we
like
have
an
index
of
which
instrumentations
are
represented
right,
so
which
which
libraries
there
exists
instrumentation
for
that's
on
the
main,
the
main
site.
E
And
you're-
and
I
think
the
response
was
yeah,
but
we
should
document
the
library
instrumentation,
which
was
I
was
talking
about
something
completely
different
than
you
had
originally
been
talking
about
so
yeah.
So
this
list
also
exists
on
the
main
site.
The
open,
telemetry
site
doesn't.
E
E
Let
me
find
the
thing
I'm
talking
about,
because
I
think
it
does
well
it,
and
I
guess
I
should
be
careful
not
fully
documentation,
it's
more
of,
like
the
enumeration
of
which
libraries
are
supported
by
the
tree.
I.
A
C
A
E
E
E
E
B
E
D
E
Yeah,
I'm
just
I'm
trying
to
put
my
user
hat
on
like
let's
say
you
just
heard
of
open
telemetry
today
and
you
come
in
and
you're
like
wow.
I
wonder
if
that
would
help
me
answer
the
question
of
how
how
long
x
is
taking
to
do
why
right
like
if
you
have
some
things
that
you
use
and
you're
new
to
this
and
you
land
on
that
page
you're,
like
I
wonder
if
they
support
my
thing,
I'm
guessing
that's
the
use
case,
but.
D
Ask
nikita
he's
he's
heading
up
the
instrumentation.
E
F
C
E
B
E
D
F
D
Honorable
or
I
guess
nikita
oh,
I
can
ping
him
if
anybody
else
wants
to
reply
to
the
spring
cloud.
Sleuth
folks.
D
Yeah,
and
also
there
were
jonathan,
was
asking.
D
D
C
B
B
C
F
C
D
Yeah
nikita
seemed
to
be
sort
of
leading
this
desire
to
test
on
their
their
versions.
I
don't
know
if,
if
he's
been
getting
some
kind
of,
do
you
all
have
customers
that
are
kind
of
driving
this
john?
No
yeah,
no,
not.
As
far
as
I
know,.
C
D
I
still
haven't
done
the
type
initializers
I
have
that
on
my
actually
have
been
look
at
this.
I
I
started
going
back
in
the
notes
and
looking
at
action
items.
B
D
D
B
E
F
D
D
D
D
E
D
In
in
the
hotel
community,
or
where.
E
C
C
B
B
So
the
java
stuff
is
mostly
implemented
in
american
node.js,
okay,
which
is
better
java,
lambda
sucks,
so
it
takes
so
we,
the
java
agent
I
mean,
I
haven't
tried
to
disable
instrumentation,
I'm
sure
there's
some
tricks,
but
just
in
the
default
configuration
this
cold
start
time
with.
The
java
is
around
50
seconds.
B
B
D
Do
you
have?
Are
you
able
to
profile.
D
If
you
ever
want
to,
I
can
hook
you
up
with
a
hacked
glow
root
recently,
so
that
it
I
can
because
normally
you
can
pro
glory,
has
a
profiler,
but
it
wouldn't
it
didn't
profile.
Pre-Main.
E
D
I
fixed
it
so
that,
as
long
as
you
do
the
jaw
the
glow
root
java
agent
first
in
the
order,
it'll
start
up
and
it'll
profile,
the
open,
telemetry
agent.
F
D
D
D
B
D
I
wonder:
do
you
sign
your
agent
jarg.
D
F
D
Do
you
sign
it,
meaning
is
it
a
signed
jar
file.
B
D
F
D
Yeah
yeah
you've.
Never
I've
never
dealt
with
this
before
it's
not
common,
but
it's
super
painful
because
in
our
and
that's
why
I
was
wondering
in
your
actually
let
me
look
at
your
because
microsoft
automatically
signs
it
when
I
upload
to
maven.
B
F
F
It
then
the
jvm
verifies
the
signature,
like
it
hashes
the
jar
file
at
startup
or
something
yeah,
and
what
sucks.
F
D
B
It's
yeah,
it's
really
slow,
even
like
I'm
wondering
what
that
50
seconds
yeah,
I'm
only
wondering
but
not
wondering
enough
to
try
running
a
profile.
I'm
sure
that
would
explain
it
quite
well.
B
D
Yeah,
I
would
I
wish
that
the
I
wish
the
I
mean
that
that
sounds
horrible,
but
even
the
even
the
you
know,
10
second
overhead
is
is
not
great.
I
wish
that
we
didn't
have
that.
F
D
Yeah,
yes,
yes,
I'm
thinking
with
my
with
my
signed
jar.
No,
it
does.