►
From YouTube: 2021-03-10 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
B
It's
interesting
I
just
put
in
a
pr
to
sleuth
to
upgrade
their
server
tracer
to
use
our
semantic
conventions.
So
that's
good.
I
guess
I
mean
some
of
the
semantic
conventions
are
already
being
used
inside
the
the
base,
whatever
the
http
server
tracer,
that
is
in
the
instrumentation
api,
but
they
were
adding
a
few
extra
things
that
I
guess
our
tracer
wasn't
doing
by
default.
B
B
A
Okay,
yeah,
I
was
surprised
to
see,
concur
and
link
you.
I
thought
it
rained
like
blocking.
You
would
always
be
better.
B
B
C
A
A
A
B
A
B
B
A
B
Hey
yeah,
you
should
tell
alolita
to
yell
at
everybody
about
it.
A
B
A
B
B
Through
various
acquisitions
or
the
backbone
it's
complicated
and
I'm
not
sure
I
even
understand
it-
there's
at
least
something
that's
in
cassandra
somewhere
and
there's
something
that's
in
well,
I
mean
there's
the
old,
the
the
classic
splunk
log
storage,
which
is
some
custom
thing.
I
don't
really
know.
I
don't
understand,
don't
know
much
about
that
either
yeah
and
then
there's
the
span,
storage,
which
I
don't
remember
how
that
works.
But
it's
I
mean
because
it's
all
put
together
from
a
bunch
of
acquisitions.
B
We're
focusing
more
on
trying
to
build
a
because
it's
just
an
ey
than
worrying
about
making
a
consistent
back
end.
It's
the
exact
opposite
of
new
relic,
where
the
new
relic
has
one
back-end
to
rule
them.
All.
Everything
is
stored
in
the
same
big
distributed.
B
B
I
mean
there
was
something
like
I
want
to
say
for
times
time.
Slice
storage
was
something
like
150
sharded,
my
sequel
servers
yeah.
It
was
really
it
was
kind
of
a
crazy
back
end.
That's
all
been
replaced
with
cassandra.
Now
they
still
have
10
shards
of
mysql
to
store
non-like,
metadata,
tons
and
tons
of
metadata.
There's,
I
think,
there's
some
non-metadata
in
there,
but
it's
mostly
metadata.
B
B
A
B
Also
true
yeah
and
the
libraries
that
have
instrumentation
easy
instrumentation
points.
We
usually
pick
those
off
and
get
them
done
because
they're
fun,
because
they're
so.
A
B
A
A
Yeah
but
then
it's
like,
as
I
was
doing
it
like
the
first
thing
I
found
with
the
couch
base,
was
it
only
supports
one
version
of
couch
base
and
like
they
are
using
the
same
approach?
We
do
it's
like.
They
expect
people
to
align
the
versions.
It's
like,
that's
not
how
we've
written
instrumentation
so
far.
How
are
we
supposed
to
like?
A
A
Anymore
and
then
also
it's
just
so
hard
like
we're
talking
about
how
to
model
the
spans
like
couch
base,
has
the
fan
out
problem
like
nested
client
spans
and
what
we
do,
and
it's
like
all
of
these
are
still
not
well
solved.
So
maybe
it's
related
to
this
instrumentation
design.
Initiative
like
we
do
need
to
make
sure
all
of
these
are
well
written
out,
or
else
people
can't
write
good
instrumentation,
even
if
they
wanted
to
in
their
libraries.
B
B
Remember
whether
that
was
something
that
he
was,
that
was
that
was
going
to
be
in
there.
I
haven't
clicked
on
this
and
it's
good.
A
So
this
ecosystem
management,
where
does
instrumentation
live?
So
that's
the
interesting
topic.
I
just
randomly
started
digging
into
this.
The
past
couple
of
days,
yeah
like
as
long
as
couch
based
owns
this
instrumentation.
A
They
weren't
even
parenting,
oh
yeah,
because
they
have
their
like
couch
base.
Actually,
when
you
make
a
request,
you
can
specify
the
print
manually,
and
so
apparently
I
guess
they
were
expecting
people
to
use
a
wrapper
to
convert
context
or
current
into
that
and
send
them.
I
was
like.
I
don't
think
you
should
do
this.
A
It's
gonna
look
very
different
than
any
other
other
library
instrument,
but
but
then,
but
then
I
was
like,
but
actually
it's
your
call
like.
If
you
don't
want
to
put
implicit
propagation,
that's
also
a
decision
you
can
make.
I
don't
know
if
I
recommend
it,
but
that's
also
up
to
you
because
you're
writing
this
code.
B
B
That
for
weeks
now,
I
forgot
yeah
yeah,
it's
not
easy,
it's
a
cut.
They,
the
the
bridge,
is
extremely
complicated,
but
mostly
it
was
complicated
because
I
was
trying
to
test
it
and
nothing
was
ever
working.
So
I
finally
figured
out
why
today
and
the
reason
why
it
wasn't
working,
I
put
in
a
pr
to
fix
this-
was
that
the
they
were
the
way
that
their
composite
propagator
was
set
up.
B
B
And
they
were
calling
the
extract
on
it
and
then,
if
it
didn't
have
a
span
in
the
resulting
context,
they
were
throwing
it
away,
which
meant
that
baggage
only
would
you
would
never
see
it,
and
I
was
just
testing
sending
curling
in
some
like
just
baggage
headers
and
I'm
like.
I
never
see
anything
what's
going
on
nothing
works.
This
is
totally
broken,
but
it
was
because
they
had
tied.
They
basically
were.
B
The
code
was
saying
that
if
you
didn't
have
a
span,
then
you
shouldn't
get
anything,
and
I
convinced
marcin
today
that
that
isn't
the
way
that
open
telemetry
works
out
of
the
box,
you'll
get
baggage
if
baggage
comes
in
you'll,
get
trace
contacts
if
that
comes
in
and
they
will
kind
of
work
independently.
B
So
I
put
in
a
pr
to
make
that
change
it'll
make
it
a
lot
easier
to
test,
because
I
don't
have
to
cook
up
headers
for
trace
contacts
for
every
time.
I
want
to
try
to
test
something
I
can
just
put
in
baggage,
and
the
package
header
is
super
simple
unless,
unlike
the
trace
context,
editor
yeah.
So
anyway,
I
probably
spent
about
seven
hours
told
until
I
finally
figured
out
that
that
was
the
problem.
B
I
think
a
lot
of
it
is
that
the
way,
then
this
is
it's
a
dumb,
dumb
problem,
but
because
they
have
their
own
place,
where
they
store
all
their
snapshots
and
milestones
not
in
maven
central
like
I
never
can
get
the
source
to
anything.
I
can't
ever
see
their
source.
I
can
only
look
at
the
the
decompiled
bike
code
that
I
played
it's
just
really
hard
to
try
to
follow.
B
B
A
Always
have
trouble
debugging
agent
because
we
tend
to
have
like
a
million
versions
of
the
same
library
in
intellij,
and
then
it
doesn't
pick
the
right
one
yeah.
It
doesn't
know
what
which
one
you're
supposed
to
be
using
yeah.
I
thought
it
would.
It
could
just
give
a
drop
down
or
something,
and
then
I'm
sure
I
could
figure
it
out,
but
I
haven't
figured
out
how
to
get
it
to
pick
the
right
like
there
must
be
some
way
right,
but
I
have
no
idea
how.
B
A
B
B
B
A
A
Was
it
hasn't
been
too
flaky?
Yeah,
it's
so
weird
like,
but
that
day
this
is
obviously
causing
a
lot
of
flakes
and
murphy.
Just
so
weird,
that's.
C
B
I
actually
it's
interesting.
I
wish
makito
didn't
use
to
yell
about
this
right.
It
started
yelling
about
this
version.
B
Yeah
or
three,
I
don't
even
remember
what
version
we're
on
now
and
I
find
it
a
little
bit.
I
mean
this
is
handy,
but
I
wish
it
was
something
you
could
turn
on
rather
than
something
you
would
turn
off.
B
C
A
B
B
Yeah,
anyway,
let
me
go
back
and
see
what
else
is
open,
while
we're
both
here
on
the
call
all
right.
I've
totally
forgotten
about
that
one.
So
well,
my
extension
api
one
there's
the
whole
dash
spam
processor,
optimization,
which
I
honestly
think
is
not
going
to
make
a
difference
to
anyone
in
the
real
world
for
real.
B
B
Taskrabbit
yeah,
I
mean
I'm
I'm
fine
with
this.
It
seems
it
seems
totally
fine
to
me.
B
B
Everyone
I
mean.
I
think
we
all
want
to
recommend
people
use
the
agent,
but
people
seem
to
like
there's
it's
either
people
use
the
agent,
oh,
but
I
think
also,
we've
we've
refactored
the
documentation
for
the
agent,
so
it's
actually
really
hard
to
find
the
auto
configuration
properties
like
the
system,
properties,
environment
variables.
It's
like
four
clicks
away.
A
B
But
we
have,
we
have
something:
that's
really
hard,
I'm
having
a
hard
time,
wrapping
myself
wrapping
my
head
around
how
to
structure
the
documentation
to
guide
people
to
where
what's
going
to
be
useful
for
them,
because
there's
I
mean
there
are
really.
I
think
there
are
really
only
three
personas,
but
for
some
reason
they
don't
seem
to
self-filter
themselves
very
well
into
the
right
place.
B
B
B
To
be
in
everywhere,
because
people
will
run,
it
will
land,
I
mean,
and
maybe
even
in
the
slack
channels,
in
cncf
like
in,
like
a
pin
a
pinned
article
in
the
slack
channel.
It's
like,
if
you
want
this
go
here.
If
you
want
this,
go
here
like
just
that
same,
that
same
kind
of
flow
chart
on
how
to
get
like
how
to
get
started.
B
C
B
Yeah
so
yeah,
we
had
a
very,
very
long
conversation
with
somebody
on
cncf
slack
today
who
was
trying
to
configure
the
agents
in
kubernetes
and
didn't
understand,
really
how
to
use
either
of
them.
I
didn't
understand
how
to
actually
configure
kubernetes
period,
and
that
was
also
trying
to
figure
out
how
to
use
the
agent
which
they
didn't
understand,
how
to
use.
A
A
B
B
Anyway,
where.
A
B
A
B
Yeah
I
mean
my
goal
was
to
make
sure
that
if
somebody
broke
something
it
showed
up
in
ci
as
quickly
as
possible,
I
mean.
Ideally
it
would
be
even
better
if
there
would
be
a
way.
We
could
point
the
examples
for
ci
at
the
modules
like
have
an
actual
like
not
have
it
based
on
something
that's
in
maven
central
and
I
have
a
base
on
the
actual
project,
but
that's
definitely.
A
A
B
Yeah,
it
seemed
like
it
was
easy
enough
to
like
this
is
something
that's
copy-pastable
right
like
if
somebody
has
their
project
their
gradle
properties,
they
can
set
things
up
this
way.
Basically
not
worrying
didn't
seem
too
outside
the
realm.
Like
I
know,
I've
worked
on
projects
where
all
the
versions
were
actually
kept
in
in
the
gradle
properties.
B
A
B
Did
you
have
anything
else
outstanding
of
significance?
Besides
that
the
rap
of
the
static
wrappers
nope.
C
B
A
C
B
B
The
api
was
super
simple.
It
was
very
easy
to
use.
You
could
wire
it
into
your
whatever
message:
bus,
your
your
container
was
hooked
up
to
that
was
great,
but
then
there
was
also
another
classic
there
were
these
stateful
session
beams
like
you
could
have
like
they
built
a
whole
layer
on
top
of
the
servo
api.
I
think
there
was
a
lot
of
there
was.
It
was
just
crazy
or
engineered
like
a
web
framework
like
a
super
crazy
enterprise,
web
framework,
yep
yeah,
and
I.