►
From YouTube: 2022-11-21 meeting
Description
Open Telemetry Meeting 1's Personal Meeting Room
A
B
B
B
E
D
We
can
probably
keep
this
one
short
I.
Imagine
a
lot
of
people
are
out.
D
A
I
went
through
all
of
them.
Okay
kind
of
edited
comments
were
required,
approved
the
ones
that
looked
good
to
me.
I
just
didn't
want
to
merge,
because
I
was
the
only
one
approving
so.
E
C
C
I
get
it
now,
I
see
it
now:
okay,
yeah,
yeah,
I,
don't
know
if
you
ever
wrote
rust
code
before
but
rust
code
as
soon
as
something
uses,
yeah
yeah,
it's
it's,
the
one
that
owns
it
and
I
guess.
Wrapping
your
head
around
that
took
takes
me
way
too
long.
B
D
D
B
B
C
I
have
somebody
on
my
team
who
wants
to
do
a
lot
more
with
rust
and
last
week
the
whole
zip
code
PR.
We
did
he's
going
to
help
me
figure
out
all
the
rest
code
to
do
it
and
we
did
that
PR,
because
we
want
really
a
scenario
where
you
could
break
a
product.
Shipping
to
Canada
and
my
team
is
gonna.
We're
gonna
write
that
scenario
we're
right
into
a
future
flag.
C
One
of
the
things
you
want
to
do
with
the
feature
flag,
Service
as
well,
is
initialize
certain
Flags
based
on
an
environment
variable
so
we'll
be
diving
into
erlang
code,
we'll
get
really
gnarly
and
nerdy.
But
that's
something
we're
going
to
take
on
the
team
like
we
want
to
do
it
anyways
to
build
our
demos
scenario
and
we're
going
to
Upstream
as
much
as
we
can
yeah.
C
C
D
C
D
Yeah,
it's
it's
through
now:
okay,
good
5.98,
oh
yep,
going.
A
B
D
A
C
When
people
are
updating
service
code,
we
need
to
make
sure
that
they're
updating
the
docs
accordingly
and-
and
maybe
you
know,
we
don't-
really
have
much
of
a
contributing
guide,
but
we
should
say
in
a
contributing
guide
like
you're
out
of
span
or
magic
if
you're,
adding
any
piece
of
Television
you're
modifying
any
piece
of
telemetry
make
sure
you
dock
that
thing
on
the
back
side
and-
and
we
have
various
stocks
to
do
it,
we
should
probably
improve
our
contributing
guide
a
little
bit
for.
D
That
yeah
I
think
we
have
a
contributing
to
MD,
but
I
don't
know
if
it's
really
no
yeah.
C
C
598,
so
basically,
what
it
is
is
it's
building
all
the
generated
classes
and
copying
the
protobuf
file
to
various
different
Source
folders,
and
then
it
runs
a
bunch
of
protobuf
or
proto-c
gen
commands
for
you
know,
for
for
each
language
or
just
copies
the
files
over
so
that
you
can
run
it
locally
or
not.
Get
IDE
red
squigglies.
D
Maybe
a
good
follow-up
for
this
would
be
seeing
if
that
makes
that
does
anything
to
builds.
If
we
were
to
do
this
as
part
of
the.
C
Generations
fast,
it's
compiling
that
takes
forever
generating
code
is
easy.
I
I
know
what
you're
saying
there's
another
PR
I've
got
where
I
add
the
proto-generation
to
the
feature
flag
service
inside
the
docker
file.
It
only
adds
like
20
seconds
10
to
20
seconds
to
the
Future
flight
service,
build
that's
actually
quite.
B
D
Does
this
that's
path
and
Docker
file?
Did
you
also
change
the
context
in
the
GitHub
action.
B
D
C
D
C
I'm
also
going
to
now
update
the
feature
flag.
One
two
do
pull
the
protogen
stuff.
Now
that
we've
got
598
merged
I'll
be
able
to
I
got
to
do
another
PR
for
5.99
I
gotta
do
another
push
to
that.
One
I
mean
to
basically
make
that
part
of
the
whole
entire
Dev
environment
life
cycle,
yeah.
D
So
the
one
thing
I
would
almost
one
thing:
if
you
wanted
to
expand
5.99
a
little
bit,
because
it
would
simplify
the
action,
the
actions
file
is
normalize.
All
the
contexts.
C
Okay,
let
me
play
around
with
this:
I
might
make
it
as
a
separate
PR,
because.
D
Yeah,
either
way,
if,
if,
if
you
do,
if
we
do
normalize
the
context,
then
that
then
you
can
simplify
the
get
up
actions
file
a
bit
too,
which
is
nice.
C
D
I
am
so
I,
don't
really
understand
what
this.
C
Probably
because
the
original
code,
just
never
really
did
what
perhaps
the
design
was
meant
for
it
to
do,
but
instead
of
passing
in
so
each
product
has
a
cad
one
or
more
categories
associated
with
it.
Instead
of
passing
into
the
actual
category
keys
for
the
search,
it
takes
the
product
code,
splits
on
on
each
character
and
uses
each
character
as
a
separate
thing
for
its
for
search
context
so
effectively
what
you
get
back
is
only
random
ads
recommendation
service.
C
So
if
we
actually
pass
in
categories
it'll
do,
as
the
thing
was
intended
to
do.
I
have
not
tested
this,
but
that's
how
I
understand
a
code
or
I
remember
seeing
the
code
do
that
before
and
that's
what
I
understand
this
person's
fixing.
A
I
had
yeah
I
haven't
reviewed
this
one,
but
so.
A
C
I
think
we
just
need
to
run
the
damn
thing
and
just
really
look
at
it
make
sure,
and
if
we
could
get
consistency
on
the
recommended
products
that
would
probably
be
better
yeah.
Probably
you
could
see
a
day
where
we
add
more
products
as
well,
but
that's
that's
a
yeah.
D
Yeah
I
mean
I
think
that
the
ad
product
stuff
would
no,
no,
it
didn't
really
come
up.
We
can
talk
about
this
and
like
sort
of
discussing
the
discussion
but
like
it,
it
does
feel
like
it
would
be
great
if
all
of
the
product
stuff
was
accessible
through
like
a
UI,
and
it
was
something
that
I
think
we
if
we
built
that
like
there
seems
to
be
room
to
build
out
sort
of
the
admin
view
right.
The
control
panel
view
for
like
creating
products
just
more
surface
area.
D
D
C
D
D
D
D
D
D
C
D
Then
we
can
build
on
the
we
can
build
arm
on
arm.
We
can
build
xa6
and
x86
and
then,
as
long
as
we
tag
the
images
appropriately
based
off
of
where
they
were
built
when
we
pushed
them,
Docker,
Hub
or
or
container
Registries,
will
correctly
merge
the
manifests.
It's
just
it's
a
question
of
like
you
have
to
tag
things
in
a
special
way.
D
D
I
mean
technically
you
don't
even
I
mean
technically
it's
one
of
those
things.
We
don't
really
need
the
I
mean
it
doesn't
have
to
be
like
bare
metal
hardware.
It
could
be
like.
C
D
With
part
of
the
challenge
is
that
it
needs
we
need
to
have
a
cluster
that
has
an
xa6
node,
and
so
we
actually
need
like
two
things
right.
We
need
a
cluster
because
we
need
so
we
need
kubernetes.
We
need
to
be
able
to
dynamically.
We
need
to
be
able
to
run
GitHub
actions
on
kubernetes.
We
need
to
be
run.
Github
actions
Runners
on
kubernetes.
B
D
B
D
You
need
it
so,
the
way
when
you
go
when
I
like
if
I
had
to
go,
and
let
me
show
you.
D
C
D
C
D
E
C
C
Believe
so
yeah
I
believe
so.
I
am
gonna
triple
check
on
that
one
I
believe
we
actually
entertained
doing
that
at
honeycomb
before
we
did
a
graviton
move,
but
we
ended
up
not
doing
it
because,
like
I
asked
for
the
complexity,
we're
just
gonna
go
all
graviton,
but
I
do
believe.
In
kubernetes
you
can
have
a
worker
Group,
which
is
a
x86
another
worker
Group,
which
is
arm-based,
and
then
maybe
you
know
you
have
those
as
down,
so
they
could
run
down
to
zero
nodes,
so
they
could
completely
shut.
B
C
D
C
D
D
Like
now
that
we
released
1.1
of
the
demo,
I
think
that
people
shouldn't
have
prob
I
mean
I,
think
that
you
can
pull
and
run
the
whole
thing
and
it
works
without
having
to
get
the
without
having
to
rebuild
it,
because
we
fixed
whatever
that
jit
compilation
thing
isn't
early,
but
so.
C
D
Do
we
even
need
to
bother
with
this
fair,
fair
right,
like
that's
kind
of
like
I'm,
still
gonna,
explore
the
the
reason
I,
don't
think
there's
any
the
thing
with
the
AWS
stuff
is
like
I
know,
I,
believe
Amazon
announced
some
big
pile
of
credits
or
whatever
for
the
cncf,
but
like
I,
don't
know
how
you
get
at
that
right
now
and
I.
Don't
know
if
it's
like
real
I
mean
I,
know
it's
real,
but,
like
I,
don't
know
if
they've
set
up
a
way
to.
B
D
Like
there's
this
community
infrastructure
lab,
which
I
think
is
where
you
would
get
access
to
this
stuff,
and
it's
equinix
only
right
now,
like
that's.
D
Has
a
million
bucks
a
year
like
into
this
so
like
I,
certainly
think
this
would
be
fine
like
I,
don't
really
feel
like
a
theme,
but
if,
but
if
in
the
future,
you
know
those
Amazon
credits
like
become
available
for
use,
and
that
would
be
an
easier
way.
Then
maybe
we
could
do
that,
but
again
this
all
might
be
moot
if
we're
able
to
just
get
the
build
times
down
enough
that
everything
runs
on
Free
Runners
and
there's
just
no
problems.
C
D
A
Actual
we
discussed
a
bit
because
we
didn't
want
to
get
redo
totally
of
grpc.
Then
we
just
saw
yeah.
Maybe
we
could
replace
in
shipping
and.
D
The
big
takeaway
was,
there
was
a
general
consensus
or
general
agreement
I
think
last
week
that
the
end
goal
should
be
to
effectively
flip
it
on
its
head
right
now,
grpc
is
the
primary
RPC
mechanism.
It
should
be.
Http
is
the
primary
RPC
mechanism
and
then
have
like
a
little
cluster
of
grpc,
but
mostly
use
like
HTTP
apis,
basically
right
and
you
Swagger,
or
something
to
kind
of
Define
those
contracts
kind.
D
E
C
Could
see
you
know
grpc,
you
could
use
interceptors,
for
example,
while
HTTP
it's
it's
middle
layers,
the
handlers
middlewares
so
I
think
that's
that's
totally
fine
and
and
I'm
agreeing
jrpc
is
also
a
lot
harder
to
debug.
Quite
frankly,
yeah
I'm
guessing
we're
going
to
stick
with
protobuf,
though
right
for
for
the
actual
payload.
D
D
E
C
C
C
I'll
give
you
some
some
insights
on
it.
They're
definitely
committed
to
it
that
I'm
sure
you
realized.
Last
year
the
whole
team
kind
of
blew
up
yeah.
D
C
It's
it's
been
for
the
large
part
we
built
I
think
they
still
have
a
couple
openings.
So
a
couple
holes
left
in
it,
but
it
seems
like
there's
still
commitment
from
above
to
improve
it.
Yeah
I.
D
E
D
They're,
just
if
people
haven't
already
posted,
oh
there's
some
new
posts,
great,
let's
get
everyone
to
go
through
this
brainstorming.
I
think
that's
really
be
cool
if
we
could
get
every.
If
we
could
maybe
pick
like
January
is
the
okay.
We're
gonna
like
go
through
and
actually
like.
First
meeting
in
January
is
where
we're
gonna
go
through
this
and
make
some
decisions
set
the
goals.
C
Yeah
and
the
goals
for
2023
yeah
timing
goals,
kind
of
PM,
it
yeah
so
I'm,
seeing
a
lot
more
requests
for
logging
support
as
well,
so
I
feel.
D
D
B
C
C
A
D
A
A
A
Yeah
I
played
around
when
I
started
and
it
worked.
It's
been
a
while
that
I
should
do
but
I
guess
it's
still
working.
C
A
D
Here,
here's
my
take
in
general
I,
don't
really
have
a
huge
problem
with
adding
services,
even
if
it
increases
the
resource
load
and
I.
Think
that
there
is
enough,
like
there's,
certainly
enough
people
trying
to
trace
async
stuff
through
Kafka
or
whatever
that
it's
something
that
we
absolutely
need
like
in
the
demo.
D
C
Yeah
I'm
all
for
it
I'm
just
a
little
concerned
on
the
memory
footprint
of
java.
That's
all
the
go
service
is,
is
minuscule
and
in
fact,
when
you
look
at
our
services,
there
are
go,
Services
stay
under
20
megabytes,
each
and
then
C,
plus
and
Ross
are
even
better
for
memory
management
there,
but
Java
was,
is
the
is
kind
of
the
pig
of
them
all.
C
D
Yeah
I
mean
I'm,
also,
okay,
saying
like
this
comes
in,
as
is
like
once
we
once
the
pr
once
everything's
reviewers
are
happy
and
then
the
corkus
stuff
is
kind
of
a
follow-up.
Yeah
I'm,
not
super
worried
about
like
I'm,
not
like
Maine,
doesn't
have
to
be
super
duper
fantastic
at
all
times.
You
know
we
can.
D
C
I
will
get
this
built
and
tested
I'll
get
it
I'll
get
it
deployed
over
on
my
eks
cluster,
so
I
can
start
running
some
long
long
term
tests
for
memory,
so
we
can
start
figuring
that
out.
Eventually
we
got
to
put
resource
limits
on
all
these
things.
Anyways.
C
And
I'll
come
out
of
the
box
and
tell
you
more
than
likely
to
go.
Services
will
be
probably
12
megabytes,
I.
D
D
Going
to
be
fine,
like
I
I'm,
not
even
extreme,
like
there's,
probably
a
bunch
of
things,
you
can
do
the
job
all
right,
like
Beyond,
even
caucus
like
there's,
probably
container
stuff,
that
can
be
done
to
remove
things.
There's
I
yeah.
D
B
A
Get
can
I
share
my
my
screen
to
discuss
something
on
the
586
sure
I
was
just
testing
out
why
we
were
in
the
call
and
I
don't
know
if
that's
the
actual
the
current
behavior
that
we
have,
but
just
to
to
discuss
here
so
prompt
hand,
is
calling
get
ads
and.
E
A
D
A
C
There
well
that's
this.
B
C
You
should
have,
it
could
be
included,
so
the
load
generator
doesn't
call
the
same.
C
Your
web
browser,
when
it
initiates
a
trace
from
your
web
browser
it
might
make
three
or
four
different
API
calls.
It
knows,
you'll
be
long
in
one
big,
long,
Trace.
Well,
the
low
generator
is
going
to
call
each
one
of
those
things
individually.
So
I
would
be
hesitant
to
try
to
compare
a
load
generator
initiated.
C
D
B
C
But
if
you
click
on
the
the
top
level
span,
I
think
we
have
a
attribute
that
we
add
sorry.
B
C
Oh
boy,
so
if
you
click
on,
if
you
click
on
the
root
span,
there
should
be
an
attribute
on
there.
It
says
if
this
thing
came
from
a
low
generator
or
not
yeah.
D
C
End
it'll,
be
you
know,
this
is
probably
something
we
should
add.
We
don't
actually
add
an
attribute
to
non-synthetic
traces.
We
don't
do
anything
to
them.
They
just
come
in
one
way,
I'd
like
to
look
for
them
is
language.sdk.
C
Telemetry.Language.Sdk
or
telemetry.sck.language
equals
web.jx.
That's
a
good
attribute
to
look
for
all.
C
A
D
A
C
Original
one
didn't
it
didn't
do
any.
How
do
I
say
this
so.
C
D
D
I
I
right
like
what
would
the
I
mean
I
know
what
I
would
think
it
should
be
right,
like
you
should
have
a
root
span
for
from,
like
page
load,
starts
on
the
client
on
the
JS
or
on
the
browser
side,
and
then
once
all
important
async
calls
finish.
Then
that's
the
end
of
the
load
span,
but
that's
not
how
it
works
right
now
and.
D
I
feel
like
it's
going
to
take
more
doing
to
I
know.
Actually,
there's
did
you
all
see
the
Splunk
donated
there
like
rum,
ORS
yeah.
They
have
their
their
real
user
monitoring
thing
that
I
guess
they're
putting
in
the
hotel.
D
C
A
Interview
well,
I,
I
know
that
we
we
got
over
it
and
you
guys
should
have
other
things,
but
just
one
thing
to
wrap
my
head
around.
C
C
D
D
D
I
load
the
page,
and
then
you
know
a
bunch
of
stuff
comes
in
cool
like
how
do
we
relate
all
of
the
individual
Ajax
loads
that
happen
there
to
each
other,
especially
when
it's
something
like
I
gotta,
you
know
I'm
scrolling
down
the
page
or,
like
then
I
click.
Something
right
like
that.
That's
kind
of
the
missing
link
is
that.
B
D
I
load
GitHub,
you
know
and
then
I
do
a
bunch
of
things
and
I'm,
never
actually
navigating
away
from
the
page
I'm
on
like
I
could
stay.
You
know
if
I
go
to
like
the
notifications,
tab
and
I,
just
like
keep
changing
filters
or
whatever
then
like
I
am
I,
guess
changing
the
query
parameters,
but
there
needs
to
be
a
way
to
sort
of
like
associate
all
that
and
that's
the
stuff
that,
like
hotel,
is
missing
and
there's
been
a
working
group
on
it.
D
That's
been
going
for
a
while
that
they
don't
really
and
I,
don't
know
what
the
end
result
of
all
that
is
going
to
be,
but
that's
that's
kind
of
the
missing
part
here.
Realistically,
all
of
those
page
loads,
like
every
Trace
from
the
front
end,
should
have
just
some
sort
of
like
session
ID
or
whatever
that
is
stable
across
me.
D
Me
loading
that
page
and
then
I
would
be
able
to
go
into
whatever
and
look
at,
like
oh
show
me
all
the
traces
relating
to
this
session
ID
so
anyway,
I
think
we're
over
time.
We
are
super
overtime,
but
thanks
everyone
for
coming.
Thank
you.