►
From YouTube: 2021-01-14 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
So
it
seems
that
tyler's
having
some
power
issues
and
won't
be
available.
I
guess
I
will
do
my
best
to
get
us
through.
C
B
Yeah,
it
does
seem
that
there's
not
much
on
the
agenda,
do
remember
to
add
your
name
to
the
attendee
list.
B
B
I
think
puna
said
he
wasn't
going
to
be
able
to
make
it
as
well,
but
I
believe
that
the
the
driving
force
behind
this
is
to
set
us
up
so
that
we've
got
the
ability
to
have
stable
dependencies
or
that
we've
got
a
set
of
modules
that
we
we
feel
we
can
declare
stable
and
get
to
1.0
and
not
have
any
unstable
or
experimental
modules
that
depend
on
them.
B
So
this
looks
to
be
an
attempt
at
mapping
out
our
dependencies
and
seeing
if
we've
got
some
clear
lines
that
we
can
draw,
I
think
if
we've
got
the
the
hotel
trace,
propagation
of
baggage
and
semconf,
those
don't
depend
on
metrics,
which
I
think
would
be
the
first
thing
that
has
to
come
out
into
an
unstable
or
remain
as
a
zero
dot
rather
than
one
1.0.
B
So
I
think
this
is
probably
a
good
start
on
on
looking
at
that,
then
we
will
almost
certainly
need
some
issues
to
excise.
The
dependencies
that
do
exist
from,
for
instance,
hotel
and
global
on
meter,
meter
provider
and
those
sorts
of.
B
Things
so
just
ask
people
to
see
this
is
the
tooling
issue
right
yeah.
So
if
you
can,
please
take
a
look
at
this
and
if
you
have
any
thoughts
on
issues
or
concerns
that
might
come
up
with
this
or
how
we
should
address
it,
comment
on
that
issue.
B
The
other
thing
is
yesterday,
I
shipped
a
0.16
release
for
the
api
and
sdk,
while
updating
the
contrib
repo.
For
that,
however,
did
run
into
some
issues
with
tests
that
started
failing
josh.
Thanks
for
the
the
comments
on
the
cortex
changes,
I
I
I
changed.
Those
back
to
floats.
B
Okay,
I
will
look
at
this
again.
The
the
issue
I
was
running
into
was
in
the.
B
In
the
datadog
exporter
example
where
the
test
server
doesn't
appear
to
be
receiving
any
data
from
it,
and
it's
kind
of
surprising
in
that
I
didn't
think
I
had
changed
anything
that
that
would
have
impacted
that
right.
So
I
other
than
this
sketch
distribution,
which
I
thought
might
have
caused
it
to
not
send
the
data
that
that
I
was
expecting.
B
But
I
I
did
validate
that
even
with
the
inexpensive
distribution,
the
min
max
subcount
aggregator
is
used
and
the
the
exporter
tries
to
send
the
min
and
max
values
that
this
is
expecting
and
the
only
other
things
I
changed
were
getting
rid
of
this
double
close,
because
there's
a
defer,
close
right
up
here.
A
Yeah
I'm
looking
at
it
myself
and
I
couldn't
see
anything
wrong
in
the
diff.
So
do
you
is
this
reproducible
locally
for
you?
I
assume.
B
This
is
reproducible
locally
for
me.
Yes,
let
me
see
if
I
can
share.
B
A
And
try
and
reproduce
it
myself.
I've
had
trouble
with
udp
sockets
in
tests
in
the
past
for
mysterious
reasons.
Actually,
that's
the
first
hunch
I
have
is,
I
don't
know
udp
is
fishy.
Is
it
is
it
the
udp
yeah?
It
is.
I
believe
that
the
data,
the
dog
statsd
test,
that
I
put
there,
uses
a
file
type
for
this
reason
to
not
have
any
unreliability
from
udp
yeah
there's
a
type
yep.
It
might.
A
Not
do
that
it
does
use
unix
files,
but
that's
just
as
murky
a
territory
unix
socket
pairs,
that's
as
murky
as
using
gdp.
I
guess.
A
B
Okay
yeah,
so
I
I
don't
have
any
ideas
here
either.
My
initial
thought
was
that
maybe
it
was
never
hitting
this
block,
but
what
I
confirmed
it
was
indeed
hitting
this.
It
seemed
like
if
I
played
with
the
timeouts
on
the
exporter,
it
was
also
questionable
whether
it
would
actually
trigger
an
export
before
the
the
half
second
that
that
the
test
is
willing
to
wait
or
not,
but
even
when
I
I
caused
it
to
wait
long
enough.
B
B
One
with
the
unused
code
contribute
fairly
straightforward.
B
You
should
my
fork
is
public,
so
you
should
be
able
to
to
just
add
it
as
a
remote
and
check
out
that
branch
or
if
you've
got
the
the
github
plugin
on
on
vs,
go
that
actually
makes
it
fairly
easy
to.
A
A
B
Sure
so
we're
writing
that
pr
up
metric
exporter
bridge.
B
So
I
had
looked
through
this
and
seemed
good
to
me.
I
think
there
was
discussion
of
whether
it
could
make
use
of
the
exact
aggregator
type
that
got
changed
over
david.
I
don't
know
if
that's
something
you
want
to
go
back
and
update
this
with
now
that
I
I
think
that's
landed.
C
Yeah
I
forgot,
I
can
definitely
do
that
or
I
can
do
it
in
a
follow-up.
I.
A
Wouldn't
I
wouldn't
trouble
trouble
yourself
for
that
right
now,
given
how
large
this
pr
is,
I
started
looking
at
it.
I
it
so
far.
I
hadn't
seen
anything
surprising,
but
I
was
only
about
two
files
in.
A
Is
there
anything
particularly
tricky
or
worth
discussing
that
you'd
like
to
discuss
right
now.
C
C
It's
like
for
each
metric
for
each
time
series
for
each
point
run
a
conversion
function
it
I
I
think
one
thing
I'm
not
entirely
sure
of
is
the
mapping
between
some
of
the
data
types
I
let's
see
where
would
that
be
yeah?
It's
where
we
try
and
convert
between
like
type
gauge
and
64
to
the
int
number
kind
and
value
observer.
A
Yes,
I
think,
I'm
I
know
what
you
mean,
I'm
I'm
currently
working
on
a
prometheus
sidecar
where
I'm
facing
the
same
thing.
There's
the
those
instrument
kinds
are
not
actually
super
relevant
by
the
time
you
get
that
deep
into
the
export
pipeline,
and
I
had
to
write
myself
a
comment
to
remind
myself
why
they're
even
there,
so
there
is
a
in
the
hotel
exporter
model.
I
guess
you
have
a
chance
in
your
exporter
to
choose
your
export
aggregation
temporality.
A
So
there's
a
chance
that
you
could
have
two
exporters
doing
different
export
temporalities
and
then
you
could
make
the
call
in
the
exporter,
which
means
that
you
have
that
information
there
if
needed.
But
you
you
don't
need
it
given
that
you're
not
coming
from
an
hotel
api.
So
at
some
some
level
the
choice
you're
making
is
arbitrary
because
nobody's
going
to
use
that
information.
Since
we
know
what
the
exporter
looks
like
the
otlp
exporter
will
only
use
a
couple
of
those
bits
like
to
choose
between.
C
Okay
and
then
the
thing
I'll
probably
have
to
follow
up
on
this,
but
there's
a
couple
types
that
I
wasn't
that
did
make
sense
in
my
head
at
least
the
type
gauge
distribution
as
well
as
I
think,
there's
another
one
down
there
that
I
just
skipped
for
now,
but
yeah
I'll
have
to
come
back
to
those.
If,
if
anyone
off
the
top
of
their
head
knows
what
those
open
census
concepts
are,
then
I
can
add
that.
A
In
I
think
I
do
and
we
might
want
to
talk
about
it-
offline,
there's,
okay,
there's
going
to
be
a
section
on
data
model
in
the
workshop
on
on
metrics
tomorrow,
where
that
might
come
up,
but
if,
if
that's
the
only
problem,
let's
let's
put
that
to
another
pr,
that's
like
that's,
probably
not
gonna
work
for
a
while,
for
other
reasons.
A
B
Yeah,
I
think
in
general,
this
is
probably
in
good
shape
to
get
it
merged
and
try
to
get
it
in
front
of
people
to
to
use.
I
think
we
get
more
value
from
having
people
actually
try
to
use
things
and
figure
out
where
it
doesn't
work
for
them
than
trying
to
get
it
perfect.
The
first
time
yeah
exactly.
A
B
A
B
About
I
think
these
are
just
a
lot
of
older.
B
We
need
to
work
through
no
all
right,
so
one
thing
that
I
did
notice
when
preparing
the
release
was
that
we've
we've
made
a
lot
of
progress
in
the
in
the
main
repo,
and
there
was
basically
one
bug
that
was
fixed
in
the
contrib
repo,
which
I
think
is
probably
the
appropriate
place
to
be
at
right
now.
B
The
other
thing
that
I
noticed,
which
was
good,
was
that
other
than
the
breaking
changes
that
had
just
been
made
to
the
metrics
and
that
some
of
the
metrics
exporters
had
to
be
updated,
for
none
of
the
other
instrumentations
really
had
to
be
updated.
So
that's
good
as
well.
We
were
able
to
make
you
know
some
fairly
significant
improvements
to
the
api
and
sdk
and
all
of
the
instrumentations
continued
to
to
function
after
updating.
B
B
B
Yep
no
problem
it
had
been
a
while,
since
we
we
did
one
so
definitely
good
to
get
one
out.
I
think
there
are
a
few
other
things
that
are
just
about
ready
to
land,
so
maybe
next
week
we'll
do
another
one
just
to
keep
things
moving
great
right.
Well,
I
don't
want
to
keep
anybody
if
there's
nothing
to
talk
about
so
one
last
chance
for
anybody
to
chime
in
with
anything
else.
I
want
to
talk
about.
A
B
Okay,
thanks
yeah,
I
I
fixed
up
the
the
cortex
issue
and
I'll
get
that
pushed
up,
and
so
hopefully
it'll
just
be
the
last
day
to
dog.
One.