►
From YouTube: 2022-03-11 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
Hey
john
this
summit.
Here
we
wanted
to
check
if
there
are
other
folks
who
might
be
joining.
B
A
I'm
doing
right,
I
haven't
met
you
before
yeah.
I
think
we
might
have
kind
of
yeah
exchange
couple
of
messages,
but
no
I
we
have
not
met
so
I
am
amit
good.
I
work
for
doordash,
oh
okay,.
B
A
Yeah
we
work.
We
are
very
happy
with
open
telemetry.
So
far,
the
way
it
has
gone
yeah.
I
I
had
few
issues
that
I
wanted
to
bring
up
to
see
if
I
can
get
some
help
on
that
awesome
sounds
good.
Where.
B
Do
you
work
out
of
I
well
I
work
for
splunk,
but
I
am
in
portland
oregon
nice.
B
B
B
I
don't
know
where
everybody
is
well,
we
could
get
started.
If
you
have
questions
or
issues
I
might
be
able
to
help.
I
don't
know
it
depends
on
what
they're
about.
A
No,
I
I
think,
I'm
I'm
much
more
interested
in
kind
of
getting
this
issue.
Four
one
six
nine
addressed
to
see
if
there
is
anything
that
can
be
done
to
so
good,
to
get
the
fix
for.
B
B
A
Yeah,
so
luckily
we
have
the
repro
now
which-
and
we
kind
of
I
think
know
roughly
the
pattern
that
causes
this
issue
but
and
looks
like
okay
analog
is
here
great.
A
So
I
was
describing
john
the
issue
4169.
A
And
I
was
trying
to
see
if
there
is
anything
we
can
do
to
get
a
fix
for
this.
I
know
we.
We
have
a
repro
now
and
we
kind
of
know
a
pattern
that
causes
this
and
and
also,
I
think
you
added
started,
adding
this
as
unit
tests
making
or
making
it
easy
to
add
the
repo
to
the
unit
test.
A
So
I
wanted
to
see
if
what,
if
there
is
anything,
any
pointers
or
anything
like,
I
can
help
with
to
trying
out
effects
or
something
so
just
to
give
you
context,
this
is
yeah.
We
have
been
hitting
this
for
right
from
the
beginning,
almost
right
from
the
beginning,
and-
and
this
doesn't
happen-
I
think,
like
the
pattern
suggests
here-
it
doesn't
happen.
A
Always
we
have
about
150
plus
services,
java
services
on
running
open,
telemetry,
fine,
but
out
of
those
nine
are
unfortunately
using
this
pattern
and
they
run
into
this
every
time
we
enable
grpc
instrumentation.
So,
unfortunately,
we'll
have
we
are
having
to
disable
jrpc
instrumentation.
That
means
traces
are
broken
and
kind
of
a
bad
experience
for
everyone,
so
yeah.
I
wanted
to
see
if
you
have
any
thoughts
on
how
we
can
proceed
here.
C
A
Yeah,
it
sounds
good.
I
think.
If
there
is
anything
I
can
help
with
I'm
happy
but
yeah
you
were
the
expert.
C
C
B
By
the
way,
speaking
of
micrometer
stuff,
I'm
on
the
micrometer
slack.
B
And
people
are
now
once
again
asking
for
a
micrometer
registry
that
doesn't
use
doesn't
bridge
the
api.
They
explicitly
only
want
micrometer
api,
so
they
want
like
a
otlp
exporting
registry
without
involving
the
like
the
full
bridge
that
we
have
right
now.
B
B
In
the
thread
taylor
talks
about
why
he,
I
think
I
don't
think
he
has
a
problem
with
sdk.
Necessarily
he
doesn't
want
that
open,
telemetry
api
to
be
there.
B
For
because
he
apparently
doesn't
want
people
bringing
in
the
api
and
like
doing
their
own,
their
own
metrics
when
they
have
their
own
micrometer-based
library
or
something
internally,
I
don't
know,
I
don't
know
why
it
would
be
harmful
to
also
generate
other
metrics.
But
who
knows.
D
Yeah,
it
sounds
kind
of
similar
to
a
micro
profile
concern
about
having
multiple
metrics
apis
just
on
the
class
path,
just
yeah.
So
it's
fair.
D
B
B
B
C
B
B
C
C
I'll,
probably
I'm
going
to
rename
that
folder
to
micrometer
one,
because,
like
there's
two
approaches
we
I
mean
we
can,
I
think,
we're
sort
of
assuming
that
micrometer
two
is
gonna
come
up
pretty
soon,
so
we
could
do
a
micrometer
now
and
then
micrometer
two
later
or
we
can
just
practically
the
micrometer
one
now
in
the
micrometer
two
later
and
I
think
we're
accomplishing
enough
microcontrast
two
is
coming
up.
That
would
be
proactive.
B
C
B
B
C
They
finally
fixed
a
really
annoying
bug
for
me,
where
I
would
scroll
up
to
the
top
of
the
list.
Basically
anytime,
I
did
anything
and
the
support
was
like
we
can't
reproduce,
but
so
I
thought
they
would
never
be
able
to
fix
it,
but
eventually,
after
four
months
they
finally
fixed
that
yeah
random.
Happy
note.
C
C
D
C
B
C
C
D
This
one
laurie
kind
of
just
asked,
I
think
somebody
had
pinged
on
the
issue
asking
what
our
plans
were
for
these
two
pr's.
D
That
well
yeah,
that's
the
bigger
question,
but
for
these
two
pr's
in
particular,.
D
C
C
C
C
D
D
This
one
at
least,
seems
very
simple.
I
wouldn't
mind
pulling
that
one
in.
I
don't
really
this
one,
this
one's,
oh,
it
just-
has
more
tests.
Maybe
it's
very
interesting
that.
D
C
D
D
Oh
john
asked
sorry
I
have
not
been
on.
I
had
was
not
on
github
hardly
today
and
so,
but
I
remember
seeing
john
asked
about
my
pr.
D
Yeah,
so
what
this
this
is
for
so
up
here
right,
it's
just
exporting
full
batches
and
then,
if
you
get
to
the
bottom
and
you
have
a
partial
batch
and
you
haven't
exported
any,
you
didn't
export
any
full
batches.
You
have
a
partial
batch
and
your
time
has
elapsed.
B
Yeah,
that
is,
that
is
not
oh,
bat
is
batch,
like
is
that
a
field
of
this
class
yeah
this
inner
class,
that's
kind
of
ugly
but
yeah.
I
know
why
it's
there.
B
Okay,
could
you
just
add
a
comment
then,
because
I
was,
I
missed
the
fact
that
well,
it
was
not
obvious.
I
guess
that
that
batch
was
what
was
being
used
inside
that
export
current
batch
and
then
getting
cleared
out.
D
B
D
B
It
still
feels
like
it's
more
complicated
than
it
needs
to
be,
but
it's
less
complicated
than
it
used
to
be,
but
I
think
it
still
feels
more
complicated
than
it
is.
B
So
the
one
question
about
that
was
auto
configure
for
that
option
like
what
are
we
gonna
do
with
that?
Are
we
gonna
make
it
an
experimental
until
it's
something
that's
specked.
D
D
Yeah,
I
mean
I
can
I'm
happy
to
do
whatever,
but
mostly
just
wanted
to
check
what,
since
it's
not
suspect
what
your
feelings
were
on,
adding
it
as
a
to
this
table
as.
D
B
B
Well,
I
like
that
it
defaults
to
one.
So
that's
good.
I
thought
there
was
actually
a
spec
at
the
moment.
That
said,
you
could
only
have
one
at
a
time.
B
B
C
B
C
C
B
B
C
B
D
B
So
when
I,
when
I
saw
your
pr
or
the
two
of
them
and
and
the
y
was
there,
I
was
actually
about
ready
to
make
a
comment
that
maybe
this
is
a
thing
that
should
be
implemented
at
the
exporter
layer,
because
the
exporter
layer
exporter,
knows
whether
concurrent
exports
are
okay
or
not,
whereas
the
bas
fan
processor
does
not
know,
necessarily
whether
the
exporter
is
wired
up
to
it
is
okay
with
concurrent
and
if
you
so,
if
you
had
an
exporter
that
was
okay
with
concurrent,
it
could
just
have
its
own
queue
and
immediately
return
and
do
everything
async.
D
B
B
D
Yeah,
that's
not
a
bad
idea
at
all.
I
think
the
only
trick
there
is
the
still
is
the
flushing.
I
would
need
to
coordinate
flushing
because
you
need
to
flush
the
back
span,
processor
first
and
then
flush
your
exporter.
D
D
I
think
that
might
yeah.
I
I'm
gonna
think
on
that
see
it
seems,
sounds
like
a
solvable
problem.
D
Cool
well
for
now
I
will
I'll
add
that
comment
just
so
that
it's
you
know
the
pr
is
there,
but
then
I
will
probably
close
it
with
end
quote
this
for
for
now
and
there's
no
reason
leaving
I
mean
I
could
put
it
into
draft,
but
I
don't
want
to
pollute
your
pr
queue.
B
D
B
B
Especially
with
all
of
the
there's
there's
now
almost
a
twitter
meme
about
linked
list
about
linked
lists.
D
Where
I
found
them
useful
is
not
native
linked
lists,
but
by
essentially
linking
sort
of.
If
you
have
a
bunch
of
objects
that
you
have
to
keep
together
anyway,
you
can
just
add
a
field.
You
know
a
next
field
into
them
and
attach
them,
and
that
I
mean,
if
you
have
to
iterate
through
them,
yeah
that
actually.
B
D
D
But
again,
if
you
can
stick
the
the
the
pointers
directly
in
your
objects,
then
you
don't
you
don't
lose
the
value
of
the
the
array
packing
the
pointer.
C
D
Yeah
yeah.
B
D
B
What
do
we
do
about
the?
I
know
there
was
a
bug
or
issue
around
prometheus
exemplars
being
out
of
order.
I
do
I
don't
understand
that
thing
at
all.
I've
been
ignoring
it.
C
C
So,
like
I'm
wondering
if
the
times
we're
recording
are
wrong
for
some
reason,
just
because
we
recorded
the
wrong
time
stamp
or
like
it's
supposed
to
be
after
some
other
thing
not
before,
but
so
that's
one
thing
I
sort
of
want
to
confirm,
though,
without
a
reap
road
dump.
It's
a
bit
hard
to
reason
that
out
so
maybe
awesome
again.
If
there's
no
way
they
can
make
some
scrubbed
dump
for
us
to
follow.
B
C
B
C
And
yeah
a
long
time
ago,
we
did
have
that
thing
like.
Should
we
be
creating
invalid
spam
context
with
those
valid
ids
inside
and
like
using
what
we
can
we
decided
not
to,
and
that
does
simplify
things
and
that
that
causes
this
use
case
to
not
be
supported,
which
I
I
mean
like
hotel
spec,
also
says
a
trace
has
to
have
a
trace
id
in
the
span
id
so
yeah.
This
is
just
not
an
hotel,
compatible
concept
yeah.
D
C
D
D
B
B
C
B
B
No,
I
don't
think
he's
working
on
that
at
all.
Also
istio
has
terrible
memories.
That's
what
I've
discovered
with
so
that
some
alerts
were
handed
off
to
our
team.
That
now
istio
is
like
paging
every
every
two
days
or
so
on.
One
of
our
in
one
of
our
realms
in
all
of
our
realms
saying,
we're
out
of
memory,
and
we
have
to
go
manually,
started
supervision.
B
C
C
Even
I
think
finally,
kubernetes
is
better
at
it,
but
before
it
wouldn't
even
try
to
even
now
I
don't
know
if
it
ever
reschedules
pods
when
they
get
unbalanced.
Just
because
like
as
you
restart
nodes,
then
your
deployed
containers
can
not
be
balanced
across
nodes,
so
at
least
in
the
past,
I've
had
just
automatically
killing
pods
just
to
at
least
be
able
to
balance
them
across
nodes.
No
there's.
C
C
B
B
B
B
And
it
uses
basically
like,
and
you
can
look
at
logs-
you
can
actually
hit
the
s
command,
just
hit
s
when
you
have
a
pod,
selected
and
it'll
shell
into
it
and
give
you
a
shell
which
is
super
good,
yeah
yeah.
So
you
don't
have
to
remember
where
the
it
goes
and
where
the
exec
goes
and
where
the
container
id
goes
or
the
pod
id,
which
I
never
remember.
B
C
As
far
as
here
light
tricks,
though,
just
today,
I've
tried
installing
zsh
for
humans,
which
is
a
very
convenient
way
to
get
a
pretty
good
zsh
setup
and
I'm
actually
liking
it
a
lot
better
than
oh,
my,
oh
my
zsh
yeah,
I
mean
oh,
my
getting
it
like.
You
have
to
configure
it
a
lot,
and
it's
so
slow
like
this.
One
is
really
fast
and
like
it
installs
a
font
automatically
to
make
sure
it
can
display
its
icons
like
it
it's
for
humans,
like
it,
tries
to
do
everything
interactive
and
automatically
as
much
as
possible.
C
That's
that's!
Pretty
cool.
C
Yeah
and
then
it
can
also
automatically
copy
itself
when
you
ssh,
so
you
only
have
to
install
it
on
your
sort
of
mac
and
then
it'll
also
load
itself.
Up
on
your
ssh
machine
has
some
cool
features.
That
seems
a
little
scary
yeah.
I
install
it
on
windows,
also
it
works
or
wso,
not
yeah
native
terminal
powershell.
Whatever.
B
C
B
Cool
all
right.
Well,
I
think
I'm
going
to
bail
out
and
I'll
see
you
in
a
few
weeks,
yeah.
B
A
D
I'll
probably
push
the
I'm
still
planning
to
send
release,
notes
or
change
log
update
today,
but
I
will
but
I'll
I'll
do
the
release
tomorrow
and
that'll.
D
C
C
C
The
only
hypothesis
I
have
which,
unfortunately
like
so
my
one
idea,
was
to
maybe
important
that
still
alpha
artifact,
we
have
with
the
debug
spam
processor
for
leak
spans
like.
I
wonder
if
there's
some
pattern:
they're
using
servlet,
that's
async
that
we're
not
closing
the
span
properly.
If
we
have
some
code
that
doesn't
handle
that
correctly,
that's
the
only
hypothesis
I
have,
and
otherwise
I
don't
even
have
any
more
debug
mechanisms
other
than
a
diva
garden
which
isn't
showing
anything.
C
D
I
I
just
I
I
generally
get
on
a
call
with
them
and
ask
them
to
you
know
if
they
won't
send
us
the
code.
Ask
them
to
show
me.
You
know,
walk
me
through
the
code
and
then
try
to
piece
enough
together
to
put
together
a
repro.
C
D
D
Oh
yeah
yeah,
I
always
yeah
go
for
it
otherwise
yeah
I
I
did
meant
I
did.
I
have
it
on
my
mind
now
that
they've
not
to
forget.
D
Hopefully
I
didn't
break
anything
with
the
yo
you'll,
if
it,
if
it,
if
the
release
notes
automation,
fails.