►
From YouTube: 2022-06-17 meeting
Description
Instrumentation: Messaging
A
A
A
Real
quick
dimitra,
I
was
working
on
the
presets
for
the
home
chart
again
and
ran
into
an
interesting
scenario
where
we
were
relying
on
null
in
the
in
users:
value
stat,
gammas
to
unset
items
right,
like
that's
just
a
helm
feature.
A
A
Why
did
we
lost
it?
Because
I
don't
want
to
talk
too
much
about
this.
We
can
talk
about
it
in
the
issue,
but
basically
helm
when
it
was
merging
the
user's
values.yaml
with
the
default
values.yaml.
B
I
understand
I
understand
I
don't
like
that
feature.
I
don't
like
to
rely
on
that
because
it
wasn't
even
available
on
every
home
version.
I
remember
it's
like
some
versions
work
with
us.
Some
doesn't
so
it's
okay
to
just
ignore
it.
So
we
don't
rely
on
that
anymore
because,
with
the
previous
sales
we
can
you
can
disable
all
the
presets
and
write
your
own
config.
It's
not
needed
anymore
right.
B
A
They
all
do
that
so
like
if
you
were
using
traces
and
you
wanted
to
use
the
otlp
receiver,
but
not
the
zip
code
or
jaeger
receiver.
There's
not
a
way
right
now
to
turn
off
only
zipkin
and
jaeger,
but
keep
the
presets
otlp.
B
Yeah,
we
need
a
way
to
like
remove
everything
that
was
stated
by
default
and
provide
the
ability
to
use
their
to
provide
their
own
config
like
from
from
the
scratch,
and
you
do.
A
A
B
A
I
prioritized
yesterday,
basically
I've
seen
some
other
issues
in
the
transform
process,
so
people
want
to
be
able
to
say
you
know
do
this
if
this
is
new,
so
it'd
be
nice
to
have
like
a
nail
check
in
the
in
the
grammar,
so
I
started
implementing
that
yesterday.
C
Yeah
this
was
another
flaky
test.
Oh
so
this
yeah,
this
one
has
been
disabled.
C
C
A
C
C
Them
I'm
gonna
at
least
document
it,
but
there's
just
so
many
different
ones,
and
I
don't.
I
don't
think
that
they
come
from
any
individual
cause.
I
think
they
just.
A
I
mean
most
of
the
time
to
have
a
flaky
test.
It's
pretty
hard
to
write
a
unit
test.
That's
flaky,
because
it's
like
kind
of
not
a
unit
test
anymore
right,
you
have,
you
would
have
had
to
do
something
that
was
like
sub
process
or,
like.
I
guess
channels
are
like
the
exception
to
that,
but,
like
we
have
so
many
like
what
are
all
these
tests
doing,
that
that
make
them
so
flaky.
C
D
C
Yeah,
I
don't
know,
I
think
that
probably
the
most
obvious
thing
we
should
do
is
is
get
more
strict
about
requiring
things
to
pass,
because
I
think
what
happens.
C
A
C
People
like
all
of
us,
maintainers
proverbs,
everyone
is
so
used
to
seeing
things
fail,
that
we
just
let
it
slide
we're
like.
Oh
that's,
probably
something
that
we
know
about
and
we
just
keep
going.
But
if
we
get
a
little
bit
more
strict
about
it,
then
we'll
have
to
work
through
these
things.
I
guess
it
slows
things
down.
Sometimes,
when
you
gotta
like
stop
the
vr
and
get
something
else
merged
first,
but.
B
Also,
we
used
to
have
code
collaboration
enabled
I
remember
we
disabled
because
of
some
because
of
code
car
some
security
issue.
B
But
it's
enabled
in
in
course,
or
I
believe
we
should
enable
it
in
countries
as
well
anyway.
There
must
be
an
issue
about
enabling
back.
I
just
want
to
ask
if
you
know
about
it:
okay,.
B
D
C
A
That
one
I'm
actually
working
on
it
right
now
honorary
gave
me
some
feedback.
We
came
to
an
even
better
solution,
so
yeah,
that's
all
good!
That's
awesome!.
C
C
This
one,
I
think,
is
basically
a
duplicate
of
this
other
one,
but
I
haven't
had
a
chance
to
read
through
in
detail
yet
but
I'll
I'll
take
a
look
at
this
one.
So,
let's
assign
it
to
myself.
B
B
A
P3
yeah
yeah,
since
we
don't
have
another
triaging
label,
we
kind
of
have
to
put
a
priority
on
everything
that
we
look
at,
or
we
won't
know
that
we
looked
at
it.
C
So
this
is
a
good
example
of
a
flaky
test
that,
like
it's,
not
really
a
unit
test
because
it's
exercising
like
a
b-bolt
database
but
that's
kind
of
the
point
of
the
of
the
extension
I
mean.
So
I
guess
maybe
it's
an
integration
test.
If
you
want
to
call
it
that,
but
pretty
important
thing
to
do
and.
C
A
good
like
another
bucket
that
runs
free
more
frequently.
A
C
C
Okay,
the
exporter
here.
B
A
A
A
A
A
B
Someone
from
spark
like
this
is
by
design.
B
Don't
know
yeah,
I
don't
know.
Actually
he
probably
works
at
the
back
of
the
city
engineer
but
and
probably
uses
telemetry
for
internal
reports.
So.
C
Yeah
from
my
experience,
there
are
sometimes
back
ends
that
sort
of
needs
you
to
like
pre-define
the
data
before
you
send
data.
The
actual
data
points
to
it.
Google
cloud
used
to
do
that,
but
they've
gotten
away
from
that.
I
think
no
just
in
general.
It's
not
it's.
Obviously
a
kind
of
painful
thing
to
have
to
require.
A
C
Yeah
I
mean
it
feels
like
this.
I
mean
I
wouldn't
advise
anyone
to
architect
their
system
in
a
way
that
requires
it,
but
I
have
seen
it
and
you
sometimes
have
to
basically
just
iterate
over
the
metadata,
but
you
can
do
that.
Sometimes,
as
the
data
comes
through,
you
know,
say
well
defined,
if
not
defined
enough
now
to
send
the
data,
but
it's.
C
B
B
Yeah
just
a
question
so
he
wants.
They
want
to
add
metric
new
metric
right.
B
A
B
A
A
C
I
basically
volunteered
it
was
already
pre-existing.
Do
you
think
some
folks
from
my
company
enhanced
it
and
are
probably
using
it?
So
you
know,
I
guess
I
should
consider
myself
a
real
sponsor
on
this,
so
yeah,
but
the
thing
is
I
just
I
don't
like
how
do
I
handle
stuff
like
this?
Like
I
don't
yeah,
I.
A
Don't
necessarily
feel
obligated
to
yeah
like
tell
them
if
they
wanna,
if
they
want
more
documentation
and
have
them
add
documentation
right,
because
that's
what
they're
asking
for
right
yeah,
they
want
examples.
It's
like
okay.
Examples
are
great,
please
add
them.
That
was
like
the
person
that
was
messaging,
dimitri
and
I
and
slack
and
all
over
corn
can
trip
in
the
helm,
charts
that
was
like
hey.
Why
isn't
this
documented
like?
Why
doesn't
it
work
this
way?
It's
like.
I
don't
know
man
we're
not
perfect,
like
please
help
us
improve
our
docs.
B
B
B
A
A
Was
the
let's
look
at
the
stack
tricks
real
fast?
What's
the
actual
error
on
the
stack
trace,
because
I
want
to
make
sure
it's
not
something
generic,
that's
not
actually
them
or
like
they.
They
didn't
like
set
up
their
batch
size,
big
enough
or
whatever,
like
invalid
temporality
and
type
combination.
B
Yeah,
it
looks
like
statsd
just
does
some
transformation
and
into
creating
valid
metric,
and
that's
my
guess.
I
didn't.
B
A
Didn't
in
the
in
the
leads
channel
weren't,
we
just
talking
about
some
of
the
components
that
were
in
the
core
distribution
being
moved
to
be
maintained
by
the
maintainers.
Wasn't
prometheus
remote
right,
one
of
those.
A
Ownership
was
moved
to
the
collector
core
maintainers.
I
thought
I
don't
know.
I'd
have
to
look
at
that
thread
again
in
the
in
the
channel,
but
that
one
was
called
out
of
out
as
one
that
was
that
should
be
owned
by
a
maintainer
like
a
full
maintainer.
C
A
A
Yeah
see
the
the
prometheus
rewrite
had
anthony's
name,
not
not
dimitri's,
name,
though
no
I
was
talking
about
that's
d:
oh
statsd,
oh
yeah,
okay,
gotcha,
nevermind,
sorry,
you're
on
the
hook,
oh
yeah,
yep.
All
right.
A
I
saw
that
alex
updated
the
readme
asking
for
triagers,
that
was
nice,
hopefully
there's
some
takers
for
that.
We
can
find
some
people
through
the
sig
or
something
make
this
easier.
B
Do
this
nice
and
we
do
we
plan
to
add
locked
support
to
the
like
this
end-to-end
correctness
pipeline.
There
is
correctness
a
lot
of
correctness
tests,
but
they
are
mostly
for
metrics
these
days
and
I
believe
it
would
be
nice
if
we
have
golden
data
for
logs.
We
can
add
correctness,
test
vlogs
as
well,
but
probably
not
not
like
near
term,
but
it
would
be
nice
to
have
in
future.
C
B
A
Cool
is,
is
scraper
test
something
for
testing,
or
is
it
something
that
gets
used
as
a
library
in
our
scrapers?
It's
for
testing.
C
Okay,
basically,
oh
it's
actually
right,
oh,
go
go
ahead!
Sorry,
it's
basically
it
re.
It
reads
and
writes
basically
the
pmetric
structure
to
disk
as
a
json
file,
and
then
that
way
we
can
check
those
in
and
then,
if
something
changes
it
has
to
validate
against
that.
So
we'll
end
up
having
to
validate
the
json
file
and
we'll
get
to
trace
the
json
file
when
it
changes
too,
which
is
nice
right,
it's.
B
C
Okay
sounds
good,
well
we're
actually
over.
Is
there
any
thing
like
that?
You
guys
wanted
to
get
on.
A
There
were
some
massive
like
there's
another
big
round
of
issues
opened
up
by
someone
I
think
yeah.
I
got
to
find
it
again.
It
wasn't.
C
C
C
B
C
B
How
do
how
it
happened
that
we
are
getting
our
linder
issues
for
deadlines
only
now.
I
believe
that
job
has
been.
C
A
A
It's
it's
relatively
new
in
the
last
month
and
contrib
it
was
in
core,
it
might
have
been
disabled
and
contributed
re-enabled.
I
forget,
but
it
only
checks
files
that
were
modified
in
the
pr
so
like
if
no
one
updated.
D
A
And
we
used
to
use
the
the
core
and
contrib
both
used
to
use
the
action
version
of
it,
but
there's
like
a
issue
with
the
action
version
of
it.
So
now
we
use
the
package
itself.
C
A
Right,
are
you
two
going
to
community
today
on
monday,
are
you
gonna
go
virtually
or
attend
in
person.
A
B
A
A
B
A
C
A
Post
pandemic
and
one
of
my
teammates
went
over
to
valencia
for
eu
cube
gun
that
would
have
been
a
way
better
place
to
go.
Hang
out
for
the
first
post
pandemic
conference.