►
From YouTube: 2021-06-04 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
C
A
How
are
your
fellow
portlanders
doing
with
the
heat
this
week.
A
A
I
was,
I
was
texting
my
family
to.
Can
we
turn
the
ac
on
on
tuesday,
like
please.
A
A
E
C
D
A
A
D
D
E
A
We
want
we
want
metrics
so
that,
because
currently
we
cap,
we
calculate
on
the
metrics
on
the
back
end,
but
we
do
sample.
We
do
recommend
sampling
for
cost,
because
we
charge
by
the
gigabyte
and
so
yeah
your
metrics
are
not
quite
perfect.
I.
C
D
E
A
Depends
on
I
mean
the
the
sampling
unless
you're,
using
that
that
async
sampling.
D
A
There
was
some
when
I
used
jfr
to
profile
glow
group
before
there
was
some
option
flag
that
you
could
enable
the
async
profiling.
Interestingly,.
E
A
So
that
it
wouldn't
be
sample
what
buy
it
biased
by
the
state
point.
D
That's
that's
that's
kind
of
what
we
said
that
I
was
just
I
just
calling
out
that
there's
a
little
bit
of
an
issue
with
everyone.
Every
time
we
add
a
new
one
into
in
there,
there's
a
little
bit
of
cost
like
startup
cost
in
running
that
detector,
especially
if,
like
I
think,
the
aws
and
the
gcp
ones
need
to
actually
reach
outside
the
jvm
and
potentially
outside
the
host
to
do
some
stuff
yeah.
So
that's
that's
the
only
reason
why
I
think
I
do
think
we
need
a
better
strategy
for
how
to
pick
and
choose.
D
I
mean
we've
talked
about
this
already
right,
we're
going
to.
We
want
to
name
name
the
resource
detectors
and
make
it
easier
to
turn
them
on
and
off,
but
yeah.
I
think
it
sounded
like
josh
was
probably
leaning
towards
that
his
main
goal
is
he
wanted
to
be
able
to
do
end-to-end
integration,
testing
using
gcp
and
didn't
exactly
know
how
to
do
that
when
it
was
all.
E
A
And
also
his
other
concern
was
when
it's
in
another
repo,
he
felt
that
he
had
to
review
every
line
of
code
in
the
entire
repo
that.
E
A
D
D
E
A
Yeah
he
was
epic,
so
he'll
do.
The
his
plan
was
to
do
the
main,
build
and
artifact
in
their
repo
and
then
have
like
just
a
small
wrapper
that
pulled
that
depends
on
that
artifact
in
the
sdk
repo
and
he.
A
The
point
is
for
it
to
be
visible.
His
concern
was
visibility
like
that
people
would
find
it
wouldn't
find
it
if
it
was
in
their
repo
and
then
also,
I
think,
maybe
to
then
get
into
the
auto
configure.
E
But
an
rsdk
is
pretty
good
on
not
having
external
dependencies.
I
would
rather
the
code
just
being
the
repo,
not
this
weirdness,
I
guess,
but
I
mean,
and
if
josh
keeps
on
contributing,
he
can
become
a
maintainer
too
right,
so
that'll
kill
so
many
birds
with
one
stone.
This
might
be
the
best
chance
for
that,
because
I
guess
the
concern
is
being
a
maintainer
of
that
code
right
and
for
aws,
it's
okay,
because
I'm
a
maintainer
and
for
gcp
unless
josh
becomes
a
meeting
or
the
repo
they
will
lose
control
of
that
code.
E
Is
that
the
concern?
I
don't
think.
A
So
he
needs
to
connect
to
gcp
to
do
the
testing.
E
I
mean
you
can
like:
we've
talked
about
using
doing
database
integration,
testing
and
the
instrumentation
repo
we
just
never
did
I
mean
we're.
We
have
an
account
and
we're
okay
with
it,
and
just
someone
has
to
write
the
test,
but
that's
sort
of
on
our
reader.
I
don't
see
any
reason
not
to
have
gcp
tests
in
our
repos,
I
mean
as
long.
E
E
E
Yeah,
it
was
really
easy
to
get
a
gcp
account
for
zipkin
integration
tests.
Maybe
that's
a
point
to
mention
you're,
also
like
actually
when
this
there's
a
spring
sleuth
team
within
google.
That
was
helping
with
this
integration
with
pcp
and
they
just
give
us
a
account
credentials
and
we
run
our
integration
tests
there.
So
there
are
some
precedence
for
running
gcp
integration,
tests
from
random
repos.
A
I
think
one
of
I
think
one
of
his
concerns
was
that
it
would
potentially
introduce
like
you
didn't
want
to
put
that
burden
on
the
sdk
repo
like
if
those
tests
are
flaky
or
linking
that.
E
C
D
D
A
Right
yeah,
so
nothing
currently,
but
I'm
sure,
once
the
other
languages
start
adding
the
need
to
do
that.
For
java
cool.
E
Zip
contest
anymore,
like
I
I'm
remembering
one
conversation
which
I
can't
remember
where
it
went
where
there
was
some
discomfort
in
putting
those
credentials
into
github
actions.
For
some
reason,
we
were
using
travis
before
that.
C
E
D
A
D
E
I
don't
know
about
e
kubernetes
and
specifically,
but
even
aws.
Our
resource
detectors
have
a
well-defined
order
by
well-defined,
I
mean
a
defined
order
like
you're
supposed
to
detecting
this
order,
just
because
some
of
these
things
are
layered
on
top
of
another
thing.
So
if
you
don't
do
it
in
the
right
order,
you
detect
wrong.
A
So,
do
you
have
different
detectors
for
each
one
or
just
one
that
can
do
them
in
the
right
order.
E
E
D
A
A
B
E
I
think
so:
okay,
it's
fairly
complex,
I
mean
still
it's
like
a
month
project
for
intern.
I
guess
it's
not
the
end
of
the
road
to
write
it.
If
you
really
wanted
to
and
then
maintaining
it
I
mean,
as
we
add
more
mapping,
we
have
to
keep
on
maintaining.
I
think
that's
a
bigger
problem
yeah.
So
there
still
isn't
any
direction
to
use
the
open
company
semantic
conventions
directly,
even
with
the
x-ray
model,
if
we
did
that
could
at
least
help,
but.
A
Well,
the
semantic
attributes
need
to
go
stable.
A
Yeah
so
other
more
josh
josh
topics,
he
was
just
kind
of
getting
up
to
date
on
kind
of
curious
about
the
metrics
api
stuff.
D
A
Go
and.
D
Why
were
we
talking
about
this?
I
think
it
was
all
just
related
to
the
same
thing
like
he
wants.
I
mean
if
he
were
going
to
put
resource
detectors
in
in,
and
it
was
hooked
into
auto
configure
like
all
of
this
is
kind
of
the
same
thing
as
he.
They
really
want
stability
to
get
people
off
of
open
senses.
So
anyway,.
D
In
the
specs
to
get
resource
get
like
the
resource
attribute
convention
stabilized
as.
D
Happened
to
the
shim,
I
think
they
kind
of
don't
work.
Do
they
just
give
up
on
it?
Well,
I
think
there's
there's
fundamental
problems
between
the
models
and
I
think
they.
E
D
He
did
talk
about
that
actually
in
the
metrics
api's
sake
today,
where
there's
some
fundamental
things
that
just
won't
mesh
up
and.
D
D
D
E
D
A
Now
open
tracing
x
sends
the
exporter
is,
does
it
have
exporters.
A
A
E
A
E
E
E
I
did
have
one
interesting
thing,
like
our
agent
doesn't
work
with
light,
bend,
instrumentation
or
something.
So
I
guess
it's
the
other
direction,
but
what
was
it
there's
a
issue
of
it
in
a
private
repo
lightbend,
I
guess,
is
a
scala
framework
that
has
some
open
tracing
based
instrumentation
inside
and
it
creates.
E
E
But
like
so
what
was
interesting,
they
have
a
page
and
it's
like
they
already
found
that
the
elastic
apm
agent
also
it
would
cause
the
app
to
crash.
You
know
they
added
a
setting
to
like
bend
like
disable
thread
local
span
or
something
like
they
had
their
own
span,
implementation
that
wasn't
an
open
tracing,
and
that
was
causing
problems
with
the
elastic
eight
games.
They
had
a
flag
to
disable
that
I
need
okay,
so
now
so
the
context
it
wasn't
our
agent,
it
was
our
shim.
A
E
A
A
I
did
realize
that
just
today
that
this
I
don't
think
this
affects
us
right.
This
only
affects
otlp
because
you're
getting
bite,
arrays
and.
D
C
So,
with
respect
to
the
cadence
thing,
because
I
was
actually
on
the
call
at
this
point,
the
way
that
I
interpreted
nikita's
thing
about
additional
optimization
potential
was
that
he
might
want
to
try
and
sneak
them
into
the
next
release.
Is
that
what
you
gathered
or
no
yeah.
D
E
Having
the
31st
of
a
month
on
a
monday
makes
the
monthly
kittens
hard
to
understand.
D
D
D
This
he
has
not.
Let
us
know
what
he's
cooking
up.
C
E
D
E
E
D
Well,
there's
there's
three
ids:
two
of
them
are
trace
ids
and
one
of
them.
Just
I
don't
know
two
of
them
are
standardies.
One
is
a
trace
id
right.
Okay,
that's
right!
So
four,
it's
definitely
getting
up
to
around
2.5
k
or
something
I'm
trying
to
think.
If
you
have
the
parent,
because
you
have
the
parent
stain
context,
even
though
it's
the
same
trace
id,
it
still
goes
through
the
same
path
and
still
has
to
do
the
same.
So
it's
actually
two
trace
id's
and
two
spanish
in
general
per
real
life
span
right
yeah.
D
E
E
D
D
A
E
E
B
E
A
D
E
D
Happens,
oh
I've.
I've
been
unfortunately
deep
deep
in
that
code
in
my
past
when
trying
to
figure
out
how
to
instrument
it
and
realizing
that
they
won't.
Let
you
instrument
it
yeah.
The
old
lambda
mana
factory
is
a
little
is
some
definite
magic.
D
A
And
it
has
to
pass
so
the
lambda
can
implement
an
arbitra.
The
interface.
D
A
D
A
D
Have
a
name
like
there's
no
byte
code
corresponding
to
new
supplier
right
right,
because
supply,
because,
because
remember,
supplier
is
just
a
it's
kind
of
more
of
a
signature
than
a
thing
right.
When
you
have
one
of
those
interfaces,
you're,
basically
saying
here's
the
signature
you
have
to
match,
I
don't
care
what
it
is
that
matches
it.
It
could
be
a
method,
it
could
be
a
method
reference.
E
And
I
think
that's
required,
because
otherwise
all
suppliers
are
the
same
like
hotspot,
can't
do
anything
as
long
as
you
have
a
different
implementation
for
each
lambda,
then
it
can
be
in
line.
Then
there
can
be
escape
analysis
and
whatever
it's
doing,
but
if
it
was
just
initializing
supplier
like
a
class
you'd,
be
sort
of
screwed
so
yeah.
That's
probably
why
it
does
that.
A
I
wonder
if
it
was
not
like
a
not
a
lambda,
but
just
did
an
anonymous
inner
class.
That
would
be
the
test
to
show
if
it
was.
D
E
A
D
A
A
B
C
A
I'm
very
happy
with
the
I
mean
I
know
john
and
I
know
john
and
nikita
don't
like
prone
and
maybe
I
will
learn
to
not
like
it,
but
I
I
certainly
like
the
the
cleanups
like
it's
a
bunch
of
it's
fixed,
two
real
bugs.
That's
always
nice
to
see
it.
D
E
E
A
I
use
a
save
action,
the
format
on
save
so
it
optimizes
the
imports.
D
Yeah,
but
you
can't
it's
really
hard
to
get
exactly
the
google
java
format
with
idea
right,
you
should
use
the
plugin
you're,
not
using
it
yet
the
plugin.
I
have
never
made
the
plugin
been
able
to
make
the
plugin
work.
I
mean
it
works,
but
it
doesn't
match
what
the
actual
gradle
plugin
does.
It
does
different
things.
E
E
C
A
Oh
here's.
I
was
looking
for
other
performance.
A
E
E
A
Cool
I'll
make
a
note
and
follow-up
on
those.
Once
we
have
a
new.
A
A
A
I
recommended
they
were
on
one
one
zero
and
there
were
so
many
changes
in
the
grpc
instrumentation
in
one
two:
zero
with
the
bridge
that
I
recommended
trying
one
two:
zero
and
letting
us
know
if
it
was
still
happening.
E
E
A
Yeah,
so
we
shade,
like
the
tracing
client
instrumenter,
his
library
instrument
is
our
library
instrumentation,
so
we
do
shade
it
in
the
agent
in
case.
Somebody
also
brings
the
library.
E
E
E
E
E
A
E
D
A
A
E
E
A
E
E
E
E
A
E
E
A
Yeah
yeah,
I
was
hoping
they
would
submit
an
issue
but
yeah.
It
was
like.
I
want
honor
x
thoughts
on
that
one.
A
E
A
A
D
A
E
E
C
E
A
Okay,
throw
an
issue
in
next
time.
You
see
that,
because
that
that
was,
that
was
what
failed
on
this
one.
This
was.
C
A
A
A
E
I'm
never
gonna
get
over
the
latest
steps.
Issues
like
I
not
only
was
it
not
running,
but
then
once
it
started
running
it
got
even
more
flicky
like
I
have
no
idea
what's
going
on
there,
they
must
have.
Maybe
they
added
a
cancel,
call
in
the
latest
version
and
that's
why
that
flakiness
only
happened
in
the
latest
step
test.
I
guess
that
must
be
what
it
was,
but
yeah
some
kind
of
optimization
of,
but
it
was
just
my
dependency
management
change
that
brought
that
in
for
some
reason.
So
weird.
E
A
Also,
at
the
same
time
ago
had
added
like
had
like
quad
quintupled,
the
number
of.
E
E
But
yeah
first
of
all,
latest
step
tests.
Just
there
was
obviously
compile
failures
with
latest
steps.
So
oh
yeah,
okay,
okay,
definitely
wasn't
getting
around
honestly.
What
did
I
do?
I
don't
know
whatever
yeah
there's
some
weird
things
you
see
and
then
that's
probably
one
of
the
weirdest
I've
seen
so
far.
A
E
B
A
E
E
A
Oh,
it
was
so
nice
today,
yeah
yeah
yep
a
little
actually
a
little
warm
in
here
right
now,
but
yeah
tuesday
night.
I
was
like
all
sweaty.