►
From YouTube: 2023-01-23 meeting
Description
Open Telemetry Meeting 1's Personal Meeting Room
A
B
C
A
D
D
B
Our
Nanny
couldn't
make
it
in
so
I
had
do
child
care
this
morning,
so
I
just
got
on,
like
you
know,
10
15
minutes
ago,
although
when
I
say
I
had
to
do
child
care,
I
set
her
down
with
her
iPad
and
then
the
GitHub
issues.
C
C
So
I'm
just
updating
the
document.
I'm
gonna
spend
most
of
the
time
kind
of
talking
about
the
PRS.
We
have
them
getting.
Maybe
a
new
release
out,
but
Miko
I
think
he
put
something
too.
A
Carter,
maybe
we
can
discuss
again
regarding
the
switching
off
the
hotel,
I
think
for
for
vendors
that
use
agents.
It's
a
nice
thing,
because
then
we
can
simply
by
environment,
variables,
disable
or
enable
stuff.
A
B
A
C
B
B
B
F
Lot
I
think
I
wanted
us
to
do
this
as
part
of
one
of
the
I
think
it's
the
accounting
service
or
the
fraud
detection
service
I
can't
remember
which
one
of
the
two
it
was
but
to
have
that
not
be
instrumented
at
all,
and
the
operator
takes
care
of
all
of
it,
but
it
would
just
be
to
a
specific
service
and
I
thought.
We
said
we
wanted
to
do
this
to
the
the
Kafka
consumer
written
in
Java.
A
I
think
we
we
mentioned
that
when
we
were
discussing
regarding
the
the
operator,
so
yeah
for
the
operator
would
make
sense.
I
do
understand
from
where
this
comes
from,
because
dynatrace
also
uses
a
agent
and
in
our
case
we
work
together
with
the
hotel.
So
if
we
have
two
open,
open,
Telemetry
and
one
agent
together,
we
it
will
work,
but
it
would
be
nice
if
other
vendors
could
simply
suppress
the
hotel
in
the
service
and
then
use
their
their
stuff.
A
So
then
everything
is
linked
together
like
their
stuff
and
then
hotel
to
kind
of
like
thinking
in
the
vendor
way
to
Showcase
how
their
their
instrumentation
works
with
Hotel.
B
B
B
This
is
more
like
a
deployment
time
thing
right,
because,
theoretically,
that's
assuming
we
haven't
done
something
incredibly
stupid
and
I.
Don't
think
we
have,
you
can
remove
the
hotel
SDK
dependencies
and
nothing
should
break
on
every
single
service
today,
right,
because
the
way
that
hotel
is
designed
is
that,
if
you
remove
the
SDK,
then
all
the
API
calls
turn
into
no
Ops
and
you're
fine.
B
You
know
there's
no
way
to
do
that.
Otherwise,
like
you're
kind
of
if
we're
building
it
into
the
docker
container,
then
there's
no
way
to
sub
it
out
easily
and
in
my
mind
it's
like
okay,
then
we
should
solve
this
problem
by
like.
B
B
F
B
F
B
B
F
I
think
what
we
wanted
to
do
is
show
different
ways
to
instrument,
so
I
wanted
to
show
fraud
detection
here.
Is
it
instrumented
via
the
operator
and
there's
also
one
where
you
just
did
all
the
docker
build
stuff
yourself
and
you
instrumented
that
build
time
instead
of
that
post
deploy
time,
we'll
say,
yeah
and
then,
if
so,
you
know,
but
just
to
show
a
different
variety
right,
I
I
would
prefer
us
to
have
different
ways
to
do
things.
F
Are
we
asking
to
do
this
to
all
the
Java
services
or
just
to
a
specific
ones?
I
guess
is
what
is
650
asking
here.
B
B
B
Yeah,
although
there
are
like
I,
think
he
did
raise
an
excellent,
he
did
raise
a
point
that,
like
node,
like
node,
is
failing,
because
API
returns
undefined,
if
spans,
if
his
key
is
not
loaded
like
there's,
probably
I,
think
this
would
be
a
good
project
to
take
on
in
the
sense
that
we
need
to.
B
F
B
Or
well,
not
necessarily,
it
would
just
it
wouldn't.
We
would
need
to
like
move
how
we
would
need
to
change,
how
we're
injecting
it.
We
would
either
need
to
do
the
operator
or
we
would
need
to
do,
have
like
agent
being
injected
at
runtime
through
sidecar
containers
or
whatever
right,
if
we
were
still
going
to
keep
multiple
ways
of
doing
this.
F
B
F
B
F
B
D
B
I
guess
my
point
is
I.
Think
there's
probably
there
is
some
smoke
here
that
we
we
should
figure
out
like
Hey.
How
do
you
like
it
should
be
possible
to
enable
and
disable
open
telemetry,
and
in
doing
so
it
should
not
break
the
application
right
and
that's
going
to
require
an
audit
basically
of
every
service
to
figure
that
out.
F
I
think
the
effort
to
go
through
and
add
this
environment
variable
before
we
initialize
the
sdks
and
all
services,
not
called
Java
should
be
done
anyways
just
through
like
what
you
said
just
to
make
sure
we're
actually
legit
yeah
I
think
we
actually
bro.
We
know
we
may
have
introduced
into
early
and
seek
dependency,
and
we
should
make
sure
those
are
not
there.
Oh.
F
I
gotta
go
yeah.
Thank
you.
B
D
B
Yeah,
maybe
it's
a
four-month-old
issue.
B
B
C
Yeah
it'd
be
great
to
call
what
PR
is
we
want
and
that
new
release
and
then
also
if
we
can
potentially
consider
doing
an
rm64
release
or
at
least
testing
it
that'd.
Be
nice,
though,
we've
had
a
couple
build
improvements.
B
B
Yeah
I
think
it's
it's
at
the
top.
It's
that
on
release,
I
think
you
also
do.
D
B
A
B
B
B
We
didn't
actually
ever
do.
It
did
I.
C
A
B
What's
the
problem,
the
problem
with
like
API
Scopes
and
like
what
the
keys
could
do
and
not
do?
Okay,.
F
So
it
was
a
GitHub
action.
Permissions
thing.
A
B
F
So,
except
for
the
builds,
the
rm64
stuff
and
the
docs,
which
really
doesn't
matter,
what
do
we
need
left
for
this
thing
with
Oscar.
C
I'm,
not
sure
I
think
he
was
either
out
for
a
while
or
maybe
just
didn't
have
the
time.
So
he
just
made
I'm
sure
you
may
have
saw
the
updates.
I
guess
see
how
often
participated
in
a
review,
but
you
guys
made
updates
last
night
after
about
you
know:
oh
okay,
three
or
four
weeks
or
so
so
I
think
it
just
needs
fresh
eyes
at
this
point.
C
I
think
he
added
both
of
us
on
it.
So
maybe
just
an
action
item
from
this
egg,
if
like
Giuliano,
herself
or
Mikko,
could
take
a
look
at
this
and
give
some
final
feedback
yeah.
D
B
Actual
change
is
like:
it
creates
a
request
counter.
F
F
Yeah
we
just
had
questions
about
it
and
I
think
we
we
determined,
like
you,
know
what
we're
gonna
we're
gonna
leave
it
like
that
right.
There
enabled
false
and
it's
because
there's
a
bug
yet
another
bug
and
the
JavaScript
Auto
instrumentation
for
file
server
for
file
system
that
just
broke
things.
C
F
C
Yeah,
let's
see
and.
F
C
I
think
he
moved
it
to
a
new
Branch
or
a
new
like
a
personal
repo.
Finally,
so
no.
C
Gotcha,
okay,
yeah
I
think
besides
that,
do
any
other
PRS
need
attention.
I've,
seen
reviews
I
think
for
pretty
much
all
of
these
or
participated
in
the
reviews
for
them.
So
I
think
do
we
feel
solid,
saying,
let's
just
clear
out
these
like
six
or
whatever
they
were
five
and
then
do
a
release.
F
F
Last
week,
I
built
the
latest
images
and
I
made
some
Helm
updates
and
it
did
not
work
completely
end
to
end.
I
did
not
have
time
to
really
debug
and
figure
out
why
it
was
not
working,
but
there
were
some
calls
like
from
the
checkout
service,
which
were
just
hanging.
For
example,
they
couldn't
figure
out
why,
when
you
clicked
on
the
actual
place
order
button
itself,
it
would
just
hang
I.
E
Yeah
the
first
item
here
is
the
link
about
the
when
I
tried
to
run
the
helm,
charts
I
think
that
the
experience
is
kind
of
in
my
environment,
quite
bad,
that
that
without
any
any
chance,
I
just
hit
on
me,
install
and
and
all
the
services
doesn't
come
up.
Oh
and
then
I
get
this
Auto
memory
killed,
and
this
grass
Loop
yeah.
F
G
B
B
F
Why
was
it
when
you
try
to
pull
in
your
own
images?
You
have
to
make
sure
you
do
something
to
do
it
and
then
you're
gonna
make
me
go
look
at
stuff.
Just
give
me
a
second
I
remember
how
I
did.
F
F
F
And
when
we
do
that
there
is
right
now
today
we
have.
What
we
do
is
we
take
the
the
way
home
works.
Is
that
there's
a
property
called
app
version,
We
prepend,
that
with
v,
then
we
do
app
version,
and
then
we
do
the
service
name.
When
we
build
up
the
release
tag,
so
we
would
just
take
out
that
pre-pen
V
from
the
help
chart
itself.
B
B
Going
to
change
starting
next
year,
there's
features
in
this
yeah
I'll
call
it
one
four
I
guess:
I.
F
Want
to
come
back
to
Josh's
point
there.
Different
K-8
platforms
seem
to
have
different
memory
issues,
which
is
scary
to
my
to
me,
because
I
should
not
be
the
case,
but
the
more
data
points
we
have
on
this,
the
more
we
can
understand
why
it's
happening
and
perhaps
for
fixing.
G
G
It's
really
fun,
but
I've
just
started
creating
my
own
values
files
with
all
of
the
limits
removed,
easier.
F
E
F
And
payment
service
for
me
runs
at
80
percent.
It's
been
running
for
19
days
and
it's
never
restarted.
It's
running
at
80
memory.
G
C
A
B
Okay,
wait
future
flag
future
flag,
erlang
or
rust.
F
D
B
F
Anybody
got
a
kubernetes
agent
that
collects
metrics
every
like
one
second
yeah
and
I
guess
I
guess
I
could
configure
the
honeycomb
want
to
do
that
as
well.
F
G
Yeah
I
definitely
will
will
take
a
look
at
that.
It's
I
mean
I.
Wonder,
though,
about
the
the
wisdom
and
trying
to
solve
this
just
for
this
one
service
like
isn't
this
going
to
be
a
common
thing
across
text
Stacks,
where
there's
some
kind
of
startup
cost,
that's
gonna
have
more
resource
requirements
than
the
ongoing
yeah.
That's
true,
yeah.
B
The
other
thing
just
to
throw
it
out
there.
This
seems
like,
as
Josh
said,
this
seems
like
a
pretty
common
problem
and
I
would
be
somewhat
remiss
to
suggest
that
shouldn't
there
be
a
way
in
kubernetes
to
tell
it
that,
like,
if
you
give
it
more
Headroom
at
the
beginning,
like
things
can
burst
above
the
limit
like
I,
think
there's
a
way
to
like
turn
on
like
bursting
right,
like
you
can
tell
kubernetes
like
hey.
B
B
F
B
F
E
B
G
C
G
I
just
had
two
little
things:
if
no
one
else
has
anything
one
is
I
mentioned
it
in
slack,
could
we
remove
the
the
hotel
exporter
endpoint
on
every
single
component
in
the
helm
chart?
Is
there
any
reason?
I
just
I
just
wanted
to
make
sure
I
wasn't
messing.
Something
I
can
make
that
PR
oud.
F
There,
the
end
point
is
slightly
different
for
each
like
some
of
them
go
to
traces,
V1
some
go
to
grpc.
Some
of
them
are
HTTP
export.
What
I
want
to
do,
though,
and
what
we
should
do
is
we
should
make
the
hotel
collector,
address
or
endpoint
itself
be
a
higher
level
variable,
and
we
use
that
in
a
downward
API
that
way
there.
Somebody
could
change
it
once
and
it
Cascades
to
many.
F
G
B
B
B
The
whole
thing
will
be
a
lot
more
resilient
if
we
can
make
it
so
that
component.
This
is
not
a
good,
stateless,
12
Factor
app
right
like
right.
If,
if
breaks
because
something
else
fails,
then
that's
not
good,
so
I
think
maybe
our
next,
maybe
just
as
an
idea
for
like
the
next.
You
know
next
release
cycle
would
be
to
like
spend
some
time
on
a
kind
of
Co
like
architectural
bits
and
making
sure
that
the
thing
is
actually
factored
well
and
that
these
kind
of
problems
are
taken
care
of.
F
So
the
coming
back
to
you
know
postgres
and
migration
to
Future
flight
service,
a
lot
of
people
who
might
run
this
like
myself,
you
may
run
this
on
spot
instances
for
your
notes,
meaning
your
node
is
going
to
restart
and
you
have
no
idea
about
it.
Yeah
if
your
node
restarts
and
that
node
happened
to
have
the
postgres
database
on
it.
F
It's
gone
yeah
options
here
is:
we
could
go
to
persistent
storage
for
the
postgres
database.
That's
one
way
to
think
about
it
and
that
would
probably
fix
the
problem
as
well.
The
reason
is
is
that
when
the
feature
flag
service
starts
up,
it
checks
the
postgres
database.
If
it
does
not
have
the
schema
there,
it'll
create
the
schema
on
Startup.
F
F
One
idea
is
to
harden
the
beach
bike
service,
so
it
on
any
error
for
any
database
thing.
It
just
restarts
itself.
Another
one
is
to
build
our
own
postgres
stuff.
That
has
the
initial
schema
stuff
in
there,
although
that
will
now.
If
we
do
that,
that
means
the
schema
that
we
generate
has
to
be
in
sync
with
what
the
feature
flag
service
is
expecting.
They
really
don't
like
that
idea
at
all,
because
we're
writing
ddl
statements
based
on
some
of
the
applications
own
way
of
doing
things.
G
Yeah
I'm
not
a
fan
of
having
the
whole
service
restart
just
because
it
needs
to
run
the
migrations.
So
if
there's
a
way
that
we
could
just
build
that
into
the
application
logic,
I
would
never
do
this
in
real
life.
But
since
we're
already
outside
of
the
tarot
career,
well,.
G
B
I
think
like
to
your
point,
Josh
part
of
the
goal
here
is
to
make
like
I
know.
We
are
Beyond
somewhat
beyond
the
bounds
of
like
actual
Production
service,
but
we
actually
should
like
at
least
somewhat
make
this.
You
know,
I,
don't
want
us
to
like
cut
so
many
corners
that
someone
looks
at
this
and
looks
at
it
and
says.
Oh,
this
is
authoritative
and
then
does
the
wrong
thing,
because
we're
like
hacking
around
what
you
should
actually
do
sure.
B
Right,
yeah,
okay
and
and
then,
and
if
we
had
a
version
of
this
right,
if
someone
took
this
whole
thing
and
said
I'm
gonna
like
if
you
AWS
this
right,
I'm
gonna
like
throw
this
in
Amazon
or
in
digitalocean
or
in
whatever
in
cloud
provider
X,
then
I
would
imagine
that
what
they
would
do
is
say:
okay,
postgres
now
becomes
a
managed
database
and
I
don't
have
to
worry
about
it
anymore
and
I'm.
Just
changing
the
endpoint
to
point
to
this
thing
over
here
where
it's
not
going
to
go
down.
You
know
magically
right.
F
F
It's
a
little
bit
of
a
bear
to
redo
the
helm
chair
part
of
my
crunch,
so
that
that's
that's
all
right
and
then
it'll
be
yet
another
sub
chart
dependency
for
the
helm
chart.
It's
already
got
four
of
them.
What's
five,
but.
F
Do
Prometheus
does
do
persistence
and
I
will
tell
you
when
I
had
to
do
an
upgrade
from
kubernetes
I
think
was
121
to
122
or
122
to
123,
something
like
that
on
Amazon.
The
CSI
driver
changes
that
actually
screwed
me
up.
I
had
to
uninstall
and
reads
all
the
chart.
I
could
not
get
that
persistent
volume
to
carry
there's,
probably
a
way
to
do
it.
I,
don't
know
how
and
I
just
ended
up.
It
was
easier
for
me
to
reinstall
from
scratch,
but.
B
Yeah
right,
that's
that
that
I
guess
is
my
point
in
my,
in
my
mind,
the
helm
chart
or
whatever,
at
whatever
we
are
using
to
package
the
application
here,
and
maybe
we
should
maybe
part
of
this-
is
thinking
about
like
hey.
Should
we
like
look
at
captain
or
something
as
a
way
to
sort
of
Define
the
application
rather
than
Helm
like
the
stuff
we
provide,
should
be
pretty
clean
and
let
you
chop
off
all
of
those
external
dependencies
and
put
them
on
a
managed
whatever.
F
Far
Kafka
and
postgres
are
not
able
to
be
chopped
off,
but
the
other
things
we
depend
on.
You
can
chop
them
off,
such
as
Diego
grafana,
Prometheus
and
I'm,
forgetting
something
there.
G
F
F
F
Go
to
Future
flag
service,
make
sure
the
environment
variables
kind
of
line
up.
So
we
could
talk
to
it.
Yeah
yeah
I've
found
that
the
way
we
do
releasing
stuff
and
naming
things
at
least
for
jaeger
I,
had
to
make
a
couple
updates
to
the
Jaeger
Helm
chart
because
they
don't
allow
you
to
do
what's
called
templating
inside
the
names.
F
It's
simple
thing
to
do:
I
just
you
know,
so,
there's
a
risk
there,
I
don't
know
if
the
postgres
chart
allows
you
to
do
templating
for
the
names
and
if
it
does
and
we're
great.
If
not,
then
we
have
to
probably.
B
I
think:
okay,
here's
my
my
yeah
just
as
a
suggestion,
then,
let's
think
about
making
the
next
version
a
pretty
thorough
pass
in
terms
of
like
application,
architecture,
code
quality,
whatever
right,
because
I've
feel
like
we
need
to
get.
We
need
to
come
back
through
and
we
need
to
do
stuff
like
we
need
to
get
logs
in
here.
Right
can.
B
But
what
I'm
saying
is
like
we
need
to
you
know
we
need
to
get
the
code
in
a
really
good
place,
so
we
need
to
get
like
any
kind
of
weird
little
bugs
or
whatever
squash.
We
need
to
make
sure
that
this
as
an
application,
is
architected
well
and
reliably
and
whatever,
and
so,
let's
figure
out
the
bill.
You
know.
Let's,
let's
maybe
say
these
are
our
three
things
you
know.
B
If
we're
about
to
cut
1.3,
then
for
1.4,
our
kind
of
three
big
rocks
are
like
application,
architecture
and
reliability,
making
sure
that
you
know
everything
Works
reliably
launches,
reliably
recovers
from
failure
reliably
make
sure
that
the
code
itself
is
like
not
that
there
aren't
weird
bugs
around
like.
If
you
turn
off
open
Telemetry,
then
something
breaks
or
or
whatever
right
that
we're
that
we
are
writing
code
well
and
three
kind
of
goes
with
one
or
two
like
make
sure
that
the
the
way
you
deploy
it
is
sort
of
that.
D
B
B
Like
are
we
trying
to
like
combine
too
many
things
here?
Are
we
trying
to
do
too
much
with
this?
The
way
it
is
like?
Is
there
something
that
we
could?
Could
we
make
this
easier
on
ourselves
and
everyone
if
we
like
narrowed
the
scope
of
what
is
actually
in
our
Helm
chart
right
like
do
we
need
to
deploy
from
atheist?
Do
we
need
to
deploy
that
or
should
we
have
something
where
it's
like?
B
B
B
We
have
to
get
everything
buttoned
up
and
if
it
feels
like
people
are
still
having
a
lot
of
problems
with
making
it
actually
run,
and
things
like
that.
So.
B
I'll
start
a
yeah,
I'll
start
I'll
start
one
four
discussion
thread
we'll
pick
out
a
clever
name
for
it.
It'll
be
great.
F
And
basically
we're
gonna,
like
you
know,
part
of
making
one
for
us
to
create
a
good
base,
so
we
could
add
the
other
few
things
we
want
to
do
like
vendors
bring
in
their
own
agents.
That
kind
of
stuff
yeah.
A
B
D
B
F
Let's
get
one
three
out:
one
three
will
make
a
couple
memory
changes
to
the
health
chart
by
the
way.
I
have
a
couple
hangings
that
I
think
will
help
optimize
a
little
bit
we
didn't
discuss
it
here.
One
of
them
is
Jaeger
continues
to
be
a
a
pain.
B
Well,
let's
talk
about
it
whatever
it's.
F
C
D
B
Actual
architecture,
what
we're
actually
shipping,
what
we
actually
should
be
yeah,
where
are
we
drawing
the
lines
effectively
so
we'll
talk
about
it?
I'll
put
I'll,
create
a
discussion
in
GitHub
and
I'll
link
it
and
let's
go
through
get
these
PR's
burnt
down
and.
D
A
F
I
will
not
be
able
to
look
at
it
until
oh,
look
at
that.
I
have
time
today.
I
just
had
two:
a
big
ass
Block
open
up
my
calendar
yeah.
It's
like
we're
not
ready
to
meet
I'll.
D
F
Poking
around
at
it
a
little
bit
more
today
for
sure,
okay,
there's
gonna
be
some
things
like
I
I
had
some
challenges
and
we'll
probably
have
to
fix
another
service
or
something
somewhere,
but
we'll
figure
it
out.
Yep.
B
All
right
thanks.