►
From YouTube: 2021-03-12 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).
B
A
It
was
warm
enough
to
go
out
in
my
and
my
puffy
jacket
and
absorb
the
rays.
D
I've
been
in
meetings
but
see
your
slack
john,
and
it's
definitely
like
it's
a
known
problem
in
my
head
also
how
to
deal
with
these
spring
integration
tests
because
they
reinitialize
all
the
beams
every
time
yeah
I
mean,
I
think,
a
testing
only
method
is
always
going
to
be
required
for
these
cases.
So
I
doubt
it's
a
bad
thing.
B
Yeah,
I
was
thinking,
that's
probably
the
best
solution,
and
the
funny
thing
was
the
test
so
that
the
test
in
particular
and
for
those
of
you
don't
know
what
we're
talking
about
sleuth
related
was
verifying.
It
was
like
had
an
application
listener
that
was
getting
all
these
events
when
the
well
originally
when
the
baggage
changed.
But
in
this
case
now,
when
the
context
changes
and
then
was
asserting
that
all
of
the
events
when
there
was
baggage
had
the
right
key
in
them,
but
it
was
never
asserting.
B
There
was
anything
actually
in
the
list
like
it
like
it
was
always
empty
and
it
wasn't
ever
asserting,
not
empty.
So
this
was
has
already
been
broken
for
a
long
time.
So
I
can
certainly
just
put
it
to
do
and
or
create
an
issue
and
just
allow
put
enable
it
on
our
side
and
then
fix
it.
When
we
can.
B
Oh
so
it
so.
The
issue
is
that,
in
the
open,
telemetry
context,
you
can
add
wrappers
to
the
contacts
storage
to
enable
all
sorts
of
interesting
features
like
adding
callbacks
when
storage
changes,
and
things
like
that
very
very
nice,
very
handy,
except
that
those
wrappers
are
statically
defined
and
cannot
be
removed.
B
A
D
Is
our
open,
telemetry
one,
and
so
we
have
to
support
adding
wrappers
without
that
weird
class
mechanism?
Okay,
yeah
yeah,
like
we
have
the
spi
and
it's
doesn't
have
as
many
of
these
issues
it's,
but
we
also
add
the
static
methods
mostly,
for
I
always
forget
what
halo
4's
name
is
peter
or
I
forget,
halo,
4
requested
those
and
but
yeah
the
problem
with.
C
C
B
Well,
I'll
probably
put
in
a
pr
to
add
a
reset
method
to
content
storage.
For
that
then
I
initiated
so.
D
Maybe
we'll
call
it
from
the
openometry
testing
extension
thing:
it
seems
yeah.
No,
I
don't
think
there
is
anything
yeah
like
that's
what
I
told
other
people
like
if
you're
trying
to
write
a
spring
test
use
the
extension
at
least
that
solves
the
reset
for
test
the
global
aspect,
not
the
wrappers
yet,
but
the.
B
B
No,
no
I'm
just
doing
build,
I'm
not
even
bothering
to
register
the
global.
D
B
D
B
A
Any
thoughts
on
is
running
tests
in
parallel
with
the
global.
Just
point
never
gonna
happen,
yeah,
that's
what
I
told
the
azure
sdk
people.
They
parallel
test
testing.
D
D
A
D
D
D
B
A
A
A
A
Cool
any
any
interesting
questions.
D
No,
I
don't
think,
but
there
was
one
intro
like
so
there's
one
question
about:
can
we
use
task
local
storage
for
context
in
java,
like
some
other
languages,
and
so
in?
I
guess
in
c
sharp.
You
just
add
context
to
the
task
and
somehow
that's
just
automatically
propagated
by
the
runtime
itself,
and
so
I
did
share
the
example
that
in
kotlin
you
can
because
kotlin
supports
that.
So
it's
not
a
fact
of
the
jvm
that
you
can't
do
it
java
doesn't
have
any
real
mechanism
for
that.
D
But
I
had
this
insight.
While
I
was
answering
that,
like
the
executor
and
is
effectively
a
task
level
storage
in
some
sense
because
you
add
the
context
as
a
fuel
to
the
executor
and
then
it's
always
just
mounting
and
unmounting
the
third
local.
Somehow
that
works
out
to
be
a
task
level,
because
you're
always
using
that
context
for
all
the
tasks
that
go
to
that
executor.
D
D
D
You
always
have
your
threads
pre-initialized
and
then
you
reuse
them.
So
that
was
not
helpful
at
all
yeah,
but
they
they
thought
about
that
way.
Back.
That's
that's
interesting!
Yeah
there
was
automatic
propagation
using
inheritable
thread,
local,
just
no
one
uses
it
because
no
one
can
use
it
yeah.
That's
that's
funny.
D
A
Oh
mattesh
brought
up
so
he's
looking
at
spring
cloud
stream
or
the
spring
messaging
instrumentation,
so
like
those
annotations
and
the
listeners
and
and
it's
the
abstraction
on
top
of
I
don't
know
if
you
have
used
spring
messaging-
it's
just
like
other
kind
of
spring
stuff
abstraction
on
top
of
generic
messaging
libraries.
A
So
he
wants
to
instrument
that
but
running
into
the
the
nested
producer,
a
nested
consumer
problem
now
and
so
on
the
producer
side,
the
more
detailed
one
that
is
nested
so
at
the
spring
layer
like
we
don't
have
that
good
of
info.
So
he
wants
to
do
the.
What
was
the
word
basically
setting.
D
Yeah
yeah
yeah
enrich
was
that
the
word
that.
A
A
So
kind
of
the
suggestion
for
now
was
just
go
ahead
and
suppress
nested
on
the
consumer
side
and
go
ahead
and
try
the
enriching
plan
on
the
producer
side
and
he
was
wanted,
like
he
was
concerned
that
that,
wouldn't
you
know,
be
consistent
with
what
we're
doing
with
client
spans
and
but
I
think
we
all
sort
of
agreed
that
we're
kind
of
kicking
that
can
down
the
road
until
we
get
and
once
we
get
guidance
on
like
should
we
do
events?
A
E
C
Yeah
and
that
point
was
made
and
heard,
but
that
seems
like
a
back
like
it
seems
like
you
can
fix
that
problem.
C
D
A
A
Yeah,
it's
just
how
same
with
our
back
end
makes
that
assumption.
I
don't
even
want
to
see
what
things
would
look
like
if
it
didn't,
and
you
know
yeah
if
the
spec
says
you
know,
that'll
be
that'll,
be
my
ammunition
to
go
and
say
hey.
We
need
to
make
this
happen,
but
in
the
meantime,.
C
You
know
I
mean
it's
not
a
major
point.
I
was
just
thinking
like
it's
nice
to
be
able
to
see
both
layers
right,
like
I'm
troubleshooting,
a
problem,
seeing
that
a
spring
messaging
with
jms
underneath
versus
spring
messaging,
with
kafka
underneath
like
and
seeing
both
of
those
with
details,
I
think,
is
nice
for
somebody.
Looking
at
a
problem.
A
C
A
D
D
A
D
A
D
D
And
I
think
I
don't
know,
I
think
it
will
be
smart
yeah.
So
if
I
do
the
cherry
pick
myself,
it
should
be
fine.
I
think
the
workflow
would
still
be
able
to
cherry
pick.
It
just
then
would
fail
at
the
very
last
step
of
get
push
after
releasing
everything
but
anyways.
I
would
probably
just
do
the
cherry
pick
manually
also
in
that
case.
A
So
do
your
the
integration
test
that
you
run?
Is
there
any
thought
of
putting
bringing
those
upstream
or
do
they
rely
too
much
on
aws
infra.
D
A
D
I
mean
they
are
integration
tests
all
the
way
down
to
x-ray.
That's
one
reason
not
to
push
them
upstream,
like
our
test
export
and
then
download
the
trace
and
verify
the
trace,
looks
correct,
they're,
really
full
end.
A
D
A
Okay,
in
in
theory,
we
could
have
smoke
tests
that
did
do
real.
D
A
D
D
Yeah,
like
the
same
apps,
could
be
shared.
I
think,
like
the
app
itself
is
just
like
a
spring
boot
app
with
some
urls
that
are
set,
and
then
the
fact
that
he
uses
x
x-ray
is
just
the
collector
config,
it's
a
very
small
part
other
than
that.
It
would
be
easy
to
point
that
any
other
back-end,
so
smoke
test
would
be
pretty
easy.
A
Okay,
yeah
just
thinking
of
clearly
the
aws
instrumentation.
A
D
B
D
B
D
B
D
B
C
B
Released,
let's
say
we
released
1.1.0
and
it
accidentally
did
a
breaking
change.
I
think
we
would
want
to
release
1.1.1
that
undid
that
breaking
change
as
a
patch
release
is
a
bug
fix.
I
think
I
think
I
mean
it
might
be
that
we
actually
had
to
rev
up
to
1.2
to
fix
it
because
there
wasn't
any
way
to
I
don't
know
anyway,
I
mean
I
think
it
would
come
down
to
the
specific
case.
A
A
Had
a
good
point
that
since
we,
since
the
java
agent
should
always
be
as
new
as
the
hotel
api,
that
will
always
want
to
follow
that
with
a
java
agent
release
so
yeah,
I'm.
D
A
It
was
definitely
the
official
policy
that
we
had
agreed
on
up
until
1-0,
but
we
hadn't,
we
were
you
know,
and
these
discussions
probably
go
back
way
like
probably
even
before
you
started.
A
I
think
we
weren't
sure
what
we
would
do
after
one
zero,
but
I
I
think
the
discussion
this
morning
was
kind
of
like
yeah.
Let's
keep
doing
that.
D
D
The
agent
of
course,
as
we've
seen
like
agent,
so
much
development
like
we'll,
have
new
instrumentation
every
week
or
something
like
we
could
release
more
frequently,
but
that
might
involve
releasing
new
instrumentation
like
either
getting
the
versions
out
of
sync
or
releasing
on
patches,
which
we
don't
want
to
do,
in
which
case
we're
stuck
with
the
monthly
which
is
fine.
But
someone
could
say
that
oh
there's
so
much
development.
We
should
be
releasing
faster
and
we'll.
We
would
just
say
no.
D
D
C
I
think
if
there
were,
if
there
were
a
customer
pressure
around
a
certain
like
implementation
enhancement,
then
that
just
comes
out
in
the
regular
case.
D
C
A
Yeah-
and
we
we
can
know
like
I-
give
out
snapshot-
builds
of
our
distro
a
lot
yeah.
B
A
D
C
A
Stable
api,
okay,
instrumentation
repo.
C
Do
we
the
question
kind
of
is:
do
we
know
if
there's
a
reason,
a
good
reason
for
that
like
is
it?
Is
it
too
early
in
the
life
cycle
for
the
config
to
be
dialed
in,
and
I
don't
think
it
is
because
several
of
the
other
spi's
do
provide
config
and
I'm
thinking
about
overloading
the
at
least
one
of
the
methods
to
provide
config.
If
I
can
do
it,
I
haven't
explored
it
fully.
Yet.
B
D
C
B
B
D
Yeah,
the
resource
provider
definitely
doesn't
have
it
this
one.
Doesn't
this
one
does
yeah
yeah.
A
And
would
we
pass
in
this
as
opposed
to
the
I
mean
the
the
java
agent
config.
C
A
D
D
A
I
liked
the
idea
of
moving
property
source
into
auto
configure,
but
I
guess
part
of
my
thought
there
was
that
some
people,
some.
B
Do
you
remember
that,
oh
I
just
it's
mainly
that
it's
kind
of
some
weird
magic,
it's
just
a
bunch
of
you-
have
to
know
the
strings
that
are
valid
properties.
I
mean
it's
very,
very
powerful
and
it's
kind
of
cool.
I
wonder
whether
it
really
applies
outside
of
like
agent
distributions,
though.
D
Yeah,
that
was
the
question
right,
but
maybe
just
by
the
fact
that
we're
using
our
auto
configuring,
the
agent
we
need
to
be
acceptable
to
having
some
agent
specific
stuff
there,
also
because
otherwise
the
workaround
is
this
annoying
copy
thing,
which
it's
not
the
worst
thing,
but
it
is
like.
I
think
we
do
in
a
component
installer,
so
that
sort
of
screws
us
if
we
want
to
pass
config
properties
into
component
installer,
because
the
order.
B
B
A
So
yeah
jason
go
for
it
and
it'll
be
good
to
see
what
you
know.
B
The
submitter
to
do
is
to
first
create
some
benchmarks
that
he
thinks
is,
or
they
think
are
good
benchmarks,
because
they
were
bad-mouthing
our
benchmarks,
left
and
right
and
before
actually
trying
to
optimize
any
code
like
let's
get
some
benchmarks.
Let's
agree
that
they're
good
benchmarks
and
then
once
we
agree
that
they're
good
benchmarks
and
we
have
a
baseline,
then
we
can
go
and
do
some
optimizations
using
those
benchmarks.
B
Here's
some
good
benchmarks
and
here's
my
thing
that
it
fixes
all
your
problems
and
look
these
benchmarks
that
I
just
rewrote
prove
that
rather
than
kind
of
saying,
here's
some
good
benchmarks,
let's
baseline
and
then
let's
show
how
the
improvement
happens,
because
we
don't
really
have
a
way
to
do
like
it's
all
in
one
big
pr
at
the
moment.
B
So
I
think
I'll
feel
better
if
we
all
can
agree
on
what
on
the
benchmark
quality,
because
with
our
existing
benchmarks,
the
changes
are
a
no-op
they'd
make
no
difference
at
all.
If
our
benchmarks
are
bad,
that's
fine!
Let's
fix
the
networks
first
yep,
then
let's
fix
the
problem,
so
I
think.
D
That
is
for
this,
especially
since
such
a
complicated
change
I
mean,
I
think
I
often
make
small
micro
optimizations
and
have
a
benchmark
in
the
same
pr.
I
would
generally
be
okay
with
that,
but
for
this
bsp
it's
a
much
bigger
topic.
I
think
so
we
have
to
get
a
better
handle
on
the
benchmarks
right
yeah.
That
seems
to
make
sense
to
me.
The
particular
solution
is.
B
Pretty
I
mean
there's
a
lot
of
moving
parts
to
it
and
before
we
go
and
accept
it,
I
want
to
make
sure
that
we're
really
sure
that
it's
the
right
change
to
make.
Okay,
because
you
know,
as
we
talked
about
on
tuesday,
tuesday,
my
tuesday
or
wednesday,
you
can't
do
you
can't
keep.
You
can't
call
the
size
on
that
queue,
because
it's
a
linear
time
operation.
B
B
I
can't
I
had
a
really
hard
time,
looking
through
the
code
and
being
able
to
mentally
verify
that
the
race
conditions
were
benign
all
the
stress
tests
passed,
which
is
good,
but
anyway
it
was
complex
enough,
there's,
actually
a
third
atomic
in
there
that
he
has.
That
is
that's
counting
up
all
of
the
spans
that
have
ever
been
added,
rather
than
just
the
size
of
the
queue
which
he
which
is
used
by
to
trigger
an
export.
D
D
A
I
have
a
question:
I
haven't
looked
at
the
disruptors,
spam
processor
and
I've
never
used
disruptor,
but
I
thought
that
it
was
sort
of
the
microbe
and
the
micro
optimized
thing
to
do
this
kind
of
like.
So,
if
somebody
really
wanted
a
micro
optimized
version
of
this,
the
disruptor
would
be
the
way
to
go.
B
B
In
the
in
when
I
was
working
in
new
relic,
we
had
a
whole
ton
of
kafka
consumption
that
was
based
on
disruptors
and
tuning
those
tuning.
Those
disruptors
is
very
tricky
and
not
it's
a
very
non-trivial
operation,
so
I've
always
been
a
little
nervous
with
recommending
anyone
use
that
it's
they're,
not
a
power
user
who
really
knows
how
to
tune
the
disrupter.
D
Yeah
I've,
in
my
experience,
also
like
some
zipkin
people,
would
want
to
use
disruptor
but
like
for
such
a
simple
task
as
batching
spans
like
the
disruptor,
tends
to
just
not
help
that
much
like
it's.
I
think
it's
use
cases
are
a
bit
different,
I'm
not
sure
what
they
are
exactly,
but
it
even
like
in
the
preliminary
benchmarks.
I
wasn't
surprised
to
see
the
jc
tools
thing,
beating
the
disruptor,
because
the
disruptor
is
more
complicated
for
more
complicated
use
cases
than
just
a
queue.
D
B
B
Overkill
yeah,
I
mean.
Certainly
I've
never
seen
it
used
outside,
like
big
beefy
server
situations,
where
you
are
very
carefully
in
control
of
every
configuration
and
memory
setting
of
your
virtual
machine
so
yeah,
I
I
tend
to
agree
with
you.
I
have
a
hard
time
recommending
it.
Well,
especially
we
haven't
done
any
benchmarking
on
it.
We
don't
have
any
way
to
I
mean
I
don't
have
any
way
to
recommend
it
in
good
faith
to
anyone
right
now.
D
That
is
one
reason
why
I
mean
I
am
interested
in
applying
some
optimizations
to
the
bsp,
like
once
we
figure
out
the
benchmarks,
I'm
sure
we'll
find
something.
That's
not
super
complicated
and
it'll
probably
be
much
better
than
what
we
have
right
now,
so
definitely
interested
in
that,
rather
than
just
saying
use
the
disrupter,
I
think
to
answer
trusk's
question.
A
Cool
yeah
yeah.
I
definitely
had
some
miss
complete
lack
of
blind
spots.
B
D
D
B
Really
wanna,
I
like
I
just
saw
kent
beck
tweeted
that
he
wants
to
make
himself
into
a
non-fungible
token
and
just
sell
himself
for
50
million
dollars,
and
someone
can
claim
they
own
him.
I'd
be
done
with
it.
B
C
Yeah
but
like
I
could
see
people
wanting
you
know
fire
hose
like
kinesis
style
line,
oriented
span,
stuff.
D
I
remember
the
person
who
contributed
a
completable
result
code.
I
forgot
what
his
name
was,
but
he
said
he
has
some
scala
based
spam
pipeline
fully
asynchronous
something
something
somewhere.
I
don't
know
if
it's
open
source,
but
he
didn't
mention
that
maybe
people
just
aren't
contributing
them
yeah.
Somebody.
B
Asked
for
a
kinesis
exporter,
and
I
think
we
all
said
well
what
format
do
you
want
to
send
stuff
through
kinesis
like.
D
B
Cool
all
right.
Well,
I
think
I,
unless
we
have
other
topics,
I'm
going
to
take
off
and
have
a
good
day,
everyone.
D
B
D
B
Cool,
I
think
I
had
at
least
one
update
on
one
of
my
pr's
for
you
to
take
a
look
at
whenever
you
get
back
to
github
yeah,
nothing.
Incredibly,
urgent.
We've
got
a
couple
weeks
before
release
so
cool
all
right.
Take
it
easy
all
right
see.
D
C
D
A
Yeah
looks
like
we
probably
have
some
some
cleanup
to
do
on
some
old
pr's.
D
I'm
still
going
to
have
to
so
I
do
want
to
get
the
grpc
context
bridge
implemented.
Finally,
it
was
definitely
much
harder
than
I
was
expecting
because
I
thought
I
would
just
use
the
open
telemetry
interface
in
our
library
instrumentation,
but
I
mean
there's
so
did
you
understand
what
I
said
in
slack.
A
D
So
if
I
implement
the
context
provider
in
the
library
instrumentation,
it
still
get
like
by
default,
anything
under
io.open
telemetry
gets
shaded
and
added
to
the
bootstrap
class
loader.
So
when
that
library
instrumentation
is
included
into
the
agent
actually
that
contacts
override.
D
The
whole
agent
yeah
so
that
which
didn't
work
so
another
approach.
D
D
D
Another
would
be
to
have
a
separate
instrumentation,
not
gr
like
a
grpc
context,
instrumentation,
because
if
it's
a
separate
artifact
it'd
be
very
easy
to
exclude
it
from
the
java
agent
like
when
I
depend
on
library.
Instrumentation
like
don't
depend
on
the
grpc
context,
library
instrumentation.
D
A
A
D
D
D
In
the
library,
in
both
right,
yeah
yeah.
A
So
if
you
I
mean
how
would
we
for
the
we
couldn't
use
the
open,
telemetry
context
override.
A
D
D
And
I'm
not
sure
like
inst,
the
inst
folder
is
added
to
the
agent
class
loader
right
yeah,
which
is
also
there
for
this.
Maybe
if
you
fix
the
exporter
filter,
whatever
class
loader
like
I
was
always
thinking
of
maybe
using
the
io.jrpc
override
mechanism
in
the
grpc
instrument,
agent
instrumentation,
I'm
pretty
sure
that
would
affect
our
exporters
also,
and
so
I
I'm
probably
going
to
not
use
that
file.
D
A
I
see
and
you
would
all
you
would
want
it
in
the
agent
just
for
injecting
it.
A
D
A
Should
have
brought
it
up,
I
mean.
A
D
A
D
D
A
Okay,
yeah.
I
have
a
feeling
if,
like
if
you
with
your,
if
you
put
out
sort
of
like
that,
the
basic
idea
and
then
you
know
everybody
can
sort
of
start
to
try
to.
If
you
can
map
out
a
way
that
we
can
all
help
like
do
different,
because
there's
just
there's
too
many
instrumentations
for
if
one
person,
but
if
you
can
map
out
how
we
can
sort
of
tackle
them
and
then,
as
we
run
into
you,
know
these
weird
cases,
yeah.
E
D
A
Well,
yeah
have
a
good
friday,
yeah
good
weekend,
see.