►
From YouTube: 2020-11-25 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
It
looks
like
a
lot
of
the
same
books
over
on
the
right
over
over
your
left,
shoulder.
A
Got
it?
I
have
only
one
shelf
like
that
in
my
house:
it
has
most
of
lone
wolf
and
cub
in
english.
A
B
C
A
C
C
C
Yeah,
it's
not
like
dead
like
at
some
places.
I've
been,
but
it's
definitely
like
way
less
meetings.
C
C
D
C
C
Running
running
our
tests
with
the
agent
john
give
john
overview
what
you're
doing.
B
So
we've
had
these
smoke
tests,
just
a
few
smoke
tests
that
do
verify
the
agent,
but
most
of
our
tests,
actually
don't
verified
very
well.
So
now
I
am
reconfiguring
so
that
gradle
package
is
an
agent
jar
and
runs
an
actual
agent
with
the
tests
and
with
all
the
shading
and
stuff
and
things
that
we've
found.
I
think
two,
maybe
three
bugs
or
something
that
we
weren't
finding
before.
B
C
B
C
B
C
B
Yeah
we
figured
out
what's
going
on
with
java
15
by
the
way,
that's
so
weird
like
why
just
that
one's
so
slow,
no
clue
without.
C
One
issue
you
opened
honorable
about
the
retries,
the.
C
I
realized
because
I
was
watching
one
of
the
when
we
have
failures.
We
auto
retry
the
whole.
B
D
B
I
think
something
happened
that
messed
up
our
cache.
Probably
maybe
it's
the
tool
chain
stuff
I
don't
know,
but
that
yeah
just
have
to
dig
deep,
but
for
now
I've
been
doing
deep
in
this
other
topic
instead
yeah,
but
I
think
I'm
ready
to
call
it
and
hand
it
over
to
you
for
the
last
bits,
because
the
remaining
tests
are
java
agent,
folder
tests,
also
testing
common.
A
lot
of
those
tests
aren't
actually
testing
common
tests.
B
They're,
like
java
agent
tests
from
what
I
could
tell
like
how
it
does
the
instrumentation
context
and
whatnot
it's
pretty
hairy,
and
I
think
if
you
were
able
to
help
with
that,
you
could
do
it
much
faster
than
me
detangling
them.
A
lot
of
them
also
see
maybe
they're
not
even
necessary
anymore.
I'm
not
sure,
though
yeah.
B
Yeah,
so
the
tracer
those
unit
does,
we
can
figure
out
later,
I
think,
but
yeah
there's
still
some
hairy
java
agent
deep
tests
like
class
loader,
blah
blah
blah
type
of
things
where
I
think
you
can
probably
knock
them
out
much
faster
than
me.
So
if
I
can
get
your
help
on
that,
that
would
be
nice.
C
B
C
Definitely
I'll
definitely
have
a
time
over
the
weekend
to
poke
around
and.
B
B
C
C
Smallest
time
to
review
review,
metrics
pr's.
A
Yeah
that
was
kind
of
the
smallest
smallest
thing.
I
could
write
that
actually
did
something
end
to
end.
That
was
actually
usable.
I
mean
usable
in
quotes
the
interesting
thing
today.
So
I
started
because
I
think
I
had
by
inspection,
found
a
race
condition
in
one
of
the
pieces
and
put
it.
A
So
I
had
to
I
had
to
hack
on
that
for
a
while
and
get
an
actual
stress
test
that
actually
did
like
aggregation
cycles
and
then
once
I
did
that
it,
it
actually
took
a
long
time
to
get
the
the
race
condition
exposed.
I
had
to
tune
the
numbers
very
carefully
and
then
it
happened,
but
so
that
was
where
I
spent
a
lot
of
time
today.
It
was
actually
writing
stress
tests
that
exposed
race
conditions,
which
is
always
fun.
C
Yeah,
the
metric
stuff
is
always
super
concurrency
stuff,
as
opposed
to.
I
don't
know
that
the
tracing
stuff
is
tends
to
be
less.
I
mean,
obviously,
because
you're
not
aggregating.
C
A
B
A
Solved
the
race
condition
by
just
introducing
unbounded
memory
usage,
which
is
not
a
great
solution,
but
anyway,
I
think
it's
a
it's
the
same
solution
we
kind
of
have
today
in
the
sdk.
So
how?
How.
A
If
people
have
unbounded
label
cardinality
like
in
their
instrumentation,
then
it
would
that
it
would
be
that
basically
that
would
that
is
what
would
be
causing
it.
A
Not
the
time
series
but
aggregators
that
could
aggregate
them
if
the
if
the
labels
show
up
again.
This
is
I
mean
this
is
actually
an
open
question
that
I'm
not
sure
how
the
sdk
is
supposed
to
deal
with
like.
If
you
see
a
label
combination,
a
label
set.
B
I
just
happened
to
review
this
pr
for
the
staff
city
receiver
in
the
collector,
where
they're,
adding
aggregation
in
the
receiver
and
it
seemed
like
it
could
have
like
one.
It
wasn't
doing
anything
about
thread
safety
and
was
looking
the
stuff
from
a
map,
and
I
was
like
this
looks
broken
but
then
also
yeah.
It
seemed
like
it
could
just
explode,
but
didn't
have
any
recommendation
for
that.
B
A
A
And
especially,
if
someone
has
some
instrumentation
that
somehow,
like
you,
could
imagine
adding
like
a
time
stamp
as
a
label
as
a
label
value
or
as
a
label
key,
which
would
be
terrible,
obviously,
but
someone
could
easily
do
that
and
then,
if
then,
basically
you're
gonna
have
labels
that
are
only
ever
used
once.
A
C
A
A
Anyway,
it's
a
it's
actually
a
fairly
tricky
problem
and
I
don't
have
a
real
answer
for
it.
I
think
the
way
that
the
existing
sdk
does
it
is
it
it
like
has
every
label
set,
has
a
basically
keeps
track
of
the
number
of
times
it's
been
used,
and
if
that
number
of
time
drops
to
zero,
then
it
gets
removed,
but
I'm
actually
not
sure
that
it's
even
correct
right
now
like
I,
I
want
to
go
back
and
rewrite
the
stress.
B
B
B
A
A
A
Never
used
before
yeah
if
it's
cumulative,
it's
all
super
simple!
It's
when
you
need
to
be
able
to
do
deltas
that
things
get
tricky.
Why.
A
So
new
relics
counters
are
all
deltas
and
then
the
back
end
does
the
accumulation,
for
example,
so
they're
different,
I
mean
different
back
ends.
Just
expect
different
concentrates
of
data
yeah
delta
is
hard.
Yeah
delta's.
Do
I
mean
cumulative
is
trivial
except
except
detecting
resets?
That's
that's.
A
Get
difficult
on
cumulatives,
but
there
are
yeah.
There
are
many
back
ends
that
only
want
deltas.
C
Saw
your
community,
your
yeah,
your
issue,
your
comment
about
prometheus,
that
we
all
need
to
be
using
prometheus
for.
B
D
A
B
C
A
Yeah,
that's
a
tricky
one.
I
think
I'll
an
a
ping
bogdan
on
internal
slack,
like
yo
dude.
I
say
something
here:.
A
You
know
he's
in
florida
at
the
moment
he.
C
A
We
had
some
sort
of
some
sort
of
internal
emergency
that
he
had
to
leave.
He
had
to
come
back
from
the
beach
he
was
like
he's
like
do.
I
need
to
come
home.
He
took
a
picture
of
the
beach
and
he's
like
all
right
I'll.
C
A
Is
that
yeah
we
use
that
we
use
that
a
lot
at
new
relic?
I
think
it's
a
really
good.
It's
a
you
know.
We
just
the
other
term
is
tracer
bullet
and
we
tried
to
kind
of
avoid
the
military
terminology.
If
we
could,
and
so
steel
thread
became
the
what's.
C
A
Steel
thread,
so
it's
basically
an
implement
like
the
first
implementation.
That
goes
completely
end
to
end.
Rather
than
trying
to
do
it
like
horizontal
layers,
you
do
one
one
complete
implementation,
all
the
way
through.
So
with
my
metrics
implementation,
I've
got
the
long
counter
working
end
and
all
the
way
through
to
export.
A
So
you
that's.
What
we
said
is
the
steel
thread
going
through
the
system
that
you
implement
first
and
then
you
can
start,
you
know,
building
the
actual
full
featured
system
on
top
of
it,
but
we
we
really-
and
I
actually,
I
really
prefer
favoring
something
that
works
end
to
end,
even
if
it's
just
a
very
narrow
use
case,
so
that
you
can
deliver
something
right
away
that
actually
functions.
C
Oh,
did
you
attend
the
maintainer
meeting
yesterday.
A
Yeah,
so
it's
worth
reading
through
ted
had
a
link
out
to
a
proposal
that
he's
put
together
about
how
we
should
track
versions
and
what
he
basically
is
saying.
We
should
get
rid
of
the
term
ga
altogether
and
that
shouldn't
be
a
thing
we
talk
about
and
hit,
so
his
proposal
is
open.
Telemetry
1.0
will
be
tracing
fully
functional
and
released
and
with
long-term
support.
So
he
kind
of
wants
to
replace
this
ga
term
with
a
long-term
support.
A
A
Yeah
this
was
part
of
the
discussion
that
happened,
so
it's
worth
looking
at
taking
a
look
at
his
doc.
I
think
he
wants
to
have
a
slightly
different
view
on
what
long-term
support
means
in
that
he
because
other
people
are
like
what
does
this
mean
back
porting
patches
to
the
one
to
the
1.0
branch,
and
he
said
he
doesn't
want
to
go
down
that
road,
but
he
wants
it
to
mean
the
api
will
be
stable
and
we
will
not
break
apis
when
we
tag
it
as
lts.
A
So
if
we
tag
the
tracing
apis
at
1.0
as
the
long-term
support
api,
is
we're
not
going
to
break
those
apis,
we'll
be
backward
compatible
with
everything
going
forward
into
the
future
for
some
number
of
years.
A
Question
it
was
like
what
does
2o
means,
I
mean
we
can
break
all
the
tracing
apis
and
he's
like,
and
his
response
was
no.
My
my
goal
is
that
when
we
say
we've
tagged
this
set
of
tracing
apis
as
the
ones
we're
going
to
support
long
term.
We
won't
introduce
breaking
changes
for
some
amount
of
time.
After
that,.
C
C
And
the
goal
here
is
of
decoupling,
tracing
and
metrics
is
to
be
able
to
release
tracing
soon.
A
I
mean
within
a
month
or
so
I
don't
think
there
were.
There
were
no
dates
on
his
document.
It
was
just
more
like
here's,
the
goal
and
what
we
want
to
do.
A
A
A
C
I
didn't
have
any.
I
didn't
have
anything
on
didn't
put
in
anything
on
the
agenda
to
chat
about
nikita's.
Not
here
we
could.
C
It
would
make
nikita
happy
to
mark
some
triage
some
issues,
but.
A
B
A
And
I'm
happy
to
report,
we
have
a
lot
fewer
breaking
changes
than
we
did
in
the
release
before
last.
A
A
A
B
B
C
B
A
D
B
A
B
A
A
A
Trace
date
is
purely
supposedly
supposed
to
be
purely
associated
with
the
intrinsic
attributes
of
tracing
trace
id
span
id
and
then
any.
If
you
have,
if
you're
a
vendor
any
customer
specific
tags
you
need
to
put
on
that.
But
baggage
is
a
more
general
purpose.
A
user.
C
A
A
B
C
C
B
A
C
B
A
So
span
samplers
and
mod
by
the
tri-state
yeah,
because
I
think
that's
the
only
point
unless
they
really
work
at
it
yep
they
can
really
access
and
modify
the
trace
data.
D
Good
good,
so
you
you
proposed
last
time
solution
for
systematics.
You
remember
this
conversation.
C
D
D
C
Oh
yeah,
are
you
able
to
use
the
bin
tray
how
the
snapshots
suck,
because
they
they
expire.
D
Yes,
yes,
we
can
use
snapshot,
but
the
problem
that
snapshot
will
be
deleted.
Yes,.
C
Yeah,
we're
kind
of
stuff
we're
having
release
problems
currently
still
and
nikita
was
gonna
spend
some
time
working
on
it.
I
think
he's
just
been
consumed
with
some
other
stuff,
so
I
don't
think
we
can
promise
a
release
time
frame
right
now,
but
if
I
check
back
early
next
week,
hopefully
we'll
have
some
better
news.
C
The
most
recent
one
nikita
spent
like
eight
hours,
trying
to
get
it
to
work
and
has
just
been
trey,
has
been
getting
more
and
more
flaky,
I
think,
is
the
main
problem
over
time.
They've
just
been
getting
more
flaky,
so
that's
where
we
need
for
potentially
we
will
move
to
a
hosted
artifactory
instance,
which
would
hopefully
give
us
better
reliability
than
the
the
free
massive
bin
tray
that
we're
using.
C
No,
I
don't
think
it
would
be.
I
mean
to
switch
over
to,
I
think,
like
artifactory
has
hosted
instances
where
you
can
just
sign
up.
I
mean
I,
I
can
expense
that
cut
the
monthly
cost
for
a
few
months
until
we're
able
to
transition
it
over
to
cncf,
because
that
part
does
take
some
time,
but
technically
we're
just
waiting.
Hope
nikita
was
going
to
reach
out
to
the
jfrog
folks
to
explore
what
what
they
thought.
The
ideal
solution
was
for
us
here.
D
C
C
Are
you
needing?
I
mean
if
you're
needing
a
to
build
your
own
release.
You
can
always
do
that.
I
mean
that's
what
I
do.
I
just
build
the
the
open
telemetry
repo
install
it
in
my
maven
local
and
then
I
can
build
something
from
that.
C
Course,
yeah
I
mean,
if
you're,
if
that's
something
that
you
know
about,
we
would
definitely
you
know,
could
use
help
with
the
publishing
stuff,
and
I
was
just
hoping
to
get
some
guidance
from
nikita
first
before
we
go
off
pavel
had
also
offered
to
help,
because
one
of
our
solutions
involves
not
publishing
all
of
the
java
agent
instrumentation
modules
since
there's
so
many
of
them,
so
that
is
contributing
to
the.
C
C
I'd
say:
probably
not
this
week
unless
nikita
yeah-
I
do
you
john,
do
you
all
have
thursday
friday
off
at
splunk?
Do
non-us
people
have
it
off?
Do
you
know.
A
C
Yeah,
probably
so
yeah,
let's
I
mean
you're
welcome
sergey.
Maybe
the
best
thing
is
just
to
post
issue
asking
you
know
about
the
next
release
there
or
in
getter.
C
You
glad
that's
working
out
for
you,
though
the
the
oshie
stuff.
I
know
that
that's
yeah.
D
D
C
D
C
D
C
C
Yeah,
I
think
I've
only
noticed
it
recently
because
it's
been
failing
right.
C
Been
there
or
just
didn't
notice
it,
although
you
know
something's
different
and
actually
you
know
it
might
have
been
when
we.
If,
when
we
upgraded
recently
to
the
latest
bite
buddy,
there
was
some
kind
of
significant.
D
B
C
Once
once
we're
in
the
muzzle
code
and
in
the
the
generation
of
the,
but
here
this
maybe
we
could
try
reverting
of
course
chat.
We
could
try
reversing
this
is
the
one
that
updated
yeah
adjust
to
new
bite,
buddy
gradle
plugin.
C
And
I
think
this
is
also
the
issue
you
had
submitted
about
the
the
failing
locally.
C
Asked
ago,
to
take
a
look
at
that.
C
D
C
I
would
be
well
we'll
have
to
see
if,
if
nikita
I'm
fine
with
it,
it
might
be
affecting
some
splunk
customer.
C
But
in
which
case,
I'm
still,
okay,
rolling
it
back
and
asking
splunk
yeah
to
fix
it.
A
Oh
yeah
yeah
yeah
yeah
plumber,
yeah
yeah.
They
are
kind
of
on
a
mission
to
backfill
a
ton
of
missing
app
server
instrumentation
in
the
next
month.
The
next
couple
months,
as
you
probably
have
noticed,.
C
Server
instrumentation
random
episodes
yeah
I
saw
yeah
that
we
had
somebody
was
had
submitted
some
well
the
spec
updates
to
even
capture
what
app
server
is
running.
A
Yeah
evo
evo-
and
I
have
been
talking
about
that
so
evo
was,
I
don't
know
he
was
ceo
plumber
cto
plummer
anyway,
he's
a
now
the
product
manager
for
all
of
instrumentation
at
splunk,
so
he's
kind
of
running
the
show
on
what
instrumentation
work
we're
all
doing,
yeah
evo
maggie,
I
don't
know,
I'm
not
sure
how
to
I
don't
know.
What's
done,
no.
C
This
is
a
different
this,
the
one
who
submitted
this
is
ago
ago.
C
Oh
sorry,
node
the
bite,
buddy
changes,
yeah.
A
A
A
When
and
not
have
it
look
completely
java
specific
and
he
and
I
have
been
talking
a
lot
about
a
lot
about
what
what
it
means
and
like
how
do
you
like?
He
would
like
to
have
kafka
like
give
the
version
of
kafka
you're
running,
show
up
as
as
this
and
the
version
of
cassandra
you're
running
if
you're
instrumenting
a
cassandra
server
like
the
actually
cassandra
implementation.
A
So
he's
thinking
really
big
and
I'm
like
no,
no,
no,
we
need
to
think
really
narrow
in
the
specs.
If
you're
going
to
get
anything
done,
and
then
you
can
build
out
from
that,
so
he
and
I
have
been
trying
to
figure
out
the
right
approach
to
getting
that
stuff
in
there,
because
we
all
agree
it'd
be
really
good
to
be
able
to
know
what
version
of
websphere
you're
running
or
what
version
of
tomcat
you're
running
or
any
of
those
things.
A
Okay,
so
trying
to
find
the
name
for
this
is
actually
quite
tricky.
What
is
this
thing?
What
he's
still
thinking
about
proposing
is
this
is
an
engine
on
which
your
code
is
running,
but
then
python
is
a
g
unicorn.
Is
that
the
same
thing
or
yeah?
This
is
the
question
like
in
python.
A
D
C
C
Though,
to
to
start
small
like
drive
a
wedge
in
there,
like
start
with
something,
and
then
I've
noticed
those
initial
pr's,
that's
where,
like
all
that
stuff
takes
place
and
then
when
people
want
to
add
you
know,
just
extra
things
tends
to
go
more
smoothly.
A
C
So
you
want
to
submit,
if
you
don't
submit,
that
revert
that
bite
buddy
pr.
B
C
B
C
B
C
Yeah
for
quiet-
I
was
just
thinking
today.
I
was
so
like
so
nice
to
have
like
nobody
or
nobody
bother
like
nothing
all
afternoon
made
me
remember
that
that's
something
I
should
work
harder
at.
A
All
righty!
Well,
I
guess
I
have
a
meeting
I
I
wonder
whether
anyone
will
show
up
tomorrow
morning
but
we'll
see
the
9
30
java
meeting.
C
Cool
well
have
a
john,
have
some
have
a
good
time
off
honorable
sergey,
if
you're
not
taking
time
off,
hope
it's
quiet
for
you
at
least.
C
Hand
next
I
hand
john
the
the
agent
tests.
A
Hey
I
did
get
one
of
I
put
in
a
pr,
so
my
first
instrumentation
vr
yeah.
A
We
have
only
have
nine
to
go
to
have
my
goals
for
the
quarter.
Complete.
A
Yeah
see
it
feels
like
a
little
bit
of
cheating,
but
hopefully
I'll
get
to
do
some
real
instrumentation.
Sometime
soon,
all
righty
have
a
good.