►
From YouTube: 2021-09-22 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
Hi
everyone
we're
just
getting
just:
let's
wait
for
a
few
more
minutes,
hi
vishwa.
How
are
you.
B
A
A
I
think
I
think
we
can
get
started
because
again
we
have
only
a
few
items
but
all
good
items
and
then
you
can
possibly
and
give
some
time
back
to
folks
all
right.
A
All
right
so
richard
did
you
want
to
kickstart
the
discussion
in
terms
of,
and
we
can
all
look
at
the
doc.
Do
you
wanna
share
it
on
your
screen
or
if
you
want.
E
B
B
Okay,
you
can
see
the
dock
yeah,
okay
cool,
so
I
started
this
it's
by
no
means
close
to
what
we
want
to
do,
I'm
assuming
so.
Basically,
I
had
two
categories
in
mind:
one
is
the
data
produced
for
the
receiver,
another
one
is
the
configuration
consumed
by
the
receiver.
B
I
think
david
gave
a
feedback
saying
that
you
know
configuration
is
not
considered
actually
compliance,
but
there
are
quite
some.
B
Take
the
primitius
config
and
then
you
should
work
assess
you
know
for
the
for
the
receiver,
which
is
not
the
case
actually
today.
So
that's
why
I
brought
that
category
as
well
up
for
discussion
on
the
data
side,
I
basically
focused
on
very
very
basic.
B
You
know
cases
the
first
one
being
the
metric
types
and
their
magnitude,
because
I
knew
there
were
some
issues
with
counters,
where
we
were
actually
doing
a
difference
between
the
first
value
that
the
receiver
saw
delta
between
the
first
value
and
subsequent
values
for
the
same
thing
that
was
fixed
actually,
but
I
just
want
to
make.
D
B
That
you
know
all
the
metric
types
are
actually
covered
here
and
then
up
metric.
You
know
flowing
for
the
targets
that
were
discovered
and
then
external
labels.
B
To
the
time
series-
and
this
was
actually
on
the
remote
right
now-
it's
actually
in
the
receiver
as
well.
D
B
Site
as
well
and
then
job
and
the
instance,
labels
are
present
in
every
time
series
and
then
honor
labels
are
actually
honored
and
then
all
the
special
cases
like
nance
and
you
know,
stateless,
nands
and
plus
and
plus
or
minus
infinite
values,
are
actually
correctly
flowing
and
then
empty
labels.
B
Metric
renaming
currency
that
is
also
not
working,
we
cannot
actually
rename
metrics.
I
think
david
had
a
pr
to
block
renaming
due
to
a
known
part.
Basically,
I
went
and
harvested.
B
We
started
cleaning
the
what
issues
we
saw
and
also
a
few
more
things
added
to
it,
and
then
the
length
and
the
number
of
you
know
the
labels
that
you
know
prometheus
actually.
B
Supports
I'm
not
sure
if
this
actually,
I
was
looking
at
the
remote
traffic
compliance
test
as
well,
and
that
does
not
have
anything
more
than
this,
so
I
was
just
wondering
brian
rich,
you
know
you
guys
have
any
other.
You
know
requirements
that
you
think
that
will
make
sense
to
add
here
on
the
data
side.
F
C
Actually,
it's
not
that
bad,
because
you
can
just
spin
up
a
prometheus
and
send
whatever
you're
testing
and
the
promise
is
the
same
thing
and
then
query
them
with
prom
here
and
see.
If
you
get
the
literal
same
results
out
of
both,
there
might
be
some
timing
oddities,
but
beyond
this
that
should
actually
work
and
then
it
basically
tests
against
the
reference
code,
where
we
have
the
problem
that
the
reference
code
is
the
reference
documentation.
D
D
A
C
You
you
don't
need
grafana
here,
but
so
how
it
would
work.
Is
you
spin
up
prometheus,
which
now
allows
you
to
push
towards
it?
You
you
push
data
to
it.
You
push
the
same
data
to
whatever
other
system
or
systems,
you're
testing,
and
then
you
query
all
of
those
systems
with
a
predetermined,
prompt,
kill
period
and
just
see
that
the
json
you're
getting
back
is
literally
the
same
for
all
of
them.
C
A
C
B
D
B
Figure
out
the
testing
strategy
for
this,
how
to
make
it
quick
and
reliable
and
also,
at
the
same
time
like
rich,
was
saying,
have
better
fidelity.
C
At
least
for
the
case
of
cortex
thanos,
grafana
cloud,
aws,
most
likely
also
azure,
so
anything
which
has
a
single
binary
mode
or
anything
which
has
a
cloud
endpoint
which
you
can
just
run
against.
C
That
should
be
relatively
easy
to
to
just
to
just
fully
automate.
You
might
need
local
credentials
for
for
testing
against
certain
cloud
providers,
blah
blah
blah.
But
beyond
this,
that
should
be
relatively
easy.
B
C
Yeah,
I
agree,
there's
so
there's
test
instances
both
run
by
by
brian
and
by
prometheus,
so
you
have
to
one.
This
is
demo
with
perception.
I
o
and
one
is
demo
the
o
committee
or
I
think,
but
neither
of
those
just
accept
remote
right
from
anywhere.
C
Again
for
the
case
of
prometheus,
and
at
least
cortex,
but
I
think
also
for
thanos,
they
all
have
literally
a
single
binary
mode.
I
mean,
if
it's
kind
of
obvious
they
can
just
download
the
binary
start
the
binary
and
you
have
your
end
point
on
localhost
and
you
just
the
complete
thing
after
after
we're
done
running
the
test.
B
Yeah,
so
on
the
configuration
side,
does
anybody
have
any
other
feedback
like?
Certainly
if
you
take
the
prometheus
config
today
and
then
give
it
to
the
receiver,
it
doesn't
work
because
of
the
dollar
dollar.
A
B
So
because
today
prometheus
doesn't
support
enrollment
variables,
but
is
basically
doing
the
country
processing
every.
E
Dollar
is
looking
for
andromeda.
E
A
B
A
G
G
A
A
B
Take
the
data
to
different,
you
know,
stores,
yeah
and
the
company
is.
B
You
know
technically,
it's
not
about
you,
know,
matching
prometheus
one
to
one
and
how
do
you
configure
and
set
it
up
right?
Yeah
makes
sense,
whatever
david
says
so.
Okay,
I
will
update
this
and
I
will
also
put
some
put
some
examples
and
then
I
need
to
go
figure
out
how
we
would
test
this
in
a
quicker
and
reliable
way.
A
Yeah,
we
can
certainly
help
which
one
building
out
these
tests-
if
you
know
if
we
can
actually
identify
specific
examples
and
and
and
also
identify
what
is
the
expected
behavior
right.
So
I
think
that
that
would
be
useful.
A
There
other
label
tests
that
are
missing
or
need
to
be
added.
A
You
know,
because
I'm
looking
at
this
and
there's
obviously
a
whole
bunch
of
tests
around
labels,
so
you
know
that's
like
a
set
of
tests
around
labels
and
then
there's
verification
of
metric
types
and
then
there's
verification
of
you
know
up
and
nands,
which
is
unique
to
prometheus,
of
course,
but
the
point
being
that
there
are
three
categories
here
clearly
in
the
labels
tests
are
there
anything
else
that
we
want
to
look
at
josh,
maybe
from
and
prometheus
from
the
collector
slash
prometheus
receiver
side.
Maybe
you
could
give
us
some
guidance
here.
H
Yeah
last
week
I
asked
if
I
could
contribute-
and
I
had
that
on
my
list
and
didn't
make
any
progress
on
it.
So
I
I'm
listening
I'm
thinking
about
this,
but
I
don't
have
anything.
A
H
Yeah,
I'm
not
sure
what
to
spec
out,
but
I'd
like
to
see.
If
I
were
the
one
writing,
it
would
be
a
sort
of
test
that
that
stands
up
a
server
and
then
is
both
the
endpoint
that
the
server
is
probing
as
well
as
querying.
So
you
would,
you
would
open
up
a
prometheus,
metrics
endpoint
and
then
run
the
server
and
then
serve
data
to
that
prometheus
server
and
then
query
that
server
and
test
it.
H
H
Yeah,
I
was
thinking
those
through
as
well
in
the
last
week,
since
I
knew
this
quote
would
come
up.
I
mean
there's
things
about
reset
times
that
are
because
the
sidecar
is
reading
through
a
wall
and
it's
not
an
actual
prometheus
server
it
has
to
like,
and
when
we
are
turning
that
data
into
otlp
there
there
is
some
reset
detection
logic.
That's
the
only
area
where
I
can
think
of
anything.
That's
like
been
specced
for
the
otlp
layer.
H
G
I
do
see
metric
renaming
on
here,
but
I
figure
I'll
raise
this
while
we
have
brian
on
the
call
last
time
when
we
discussed
this,
it
wasn't
clear
that
it
was
actually
going
to
be
possible
for
us
to
support
this,
because
when,
by
the
time
we
actually
have
access
to
the
metric
name,
it's
using
the
new
name,
but
the
metadata
cache
that
we
can
look
in
only
has
the
metric
under
the
old
name.
G
G
G
F
Okay,
of
course,
there's
also
the
possibility
for
collisions,
but
that's
less
likely
in
practice.
Okay,
like
you,
could
be
smart
about
it.
That
would
be
an
improvement
and,
like
maybe
prometheus
itself,
will
have
that
improvement
as
well.
At
some
point
like
it
hasn't
been
an
issue
in
practice.
Basically,
okay,.
C
G
Makes
sense
yeah
at
the
end
of
the
day
the
behavior,
I
think,
might
be
surprising
to
users,
but
I
think
right
now
this
is
probably
the
best
we
can
do
and
as
long
as
we're
compliant
or
at
least
like
as
long
as
we
do,
what
the
prometheus
server
does,
I
think
we're
on
solid
ground
there.
F
G
So,
there's
a
few
things
previously:
when
someone
tried
to
rename
a
metric,
it
would
become
unknown
typed
and
then,
during
the
process
of
converting
from
open
census,
to
open
telemetry,
we
would
log
in
error.
So
it
was
like
you
would
just
get
log
spam.
If
you
tried
to
do
this,
so
I
think
what
we
need
to
do
is
first:
go
fix,
whatever
issues
we're
underlying
make
sure
that
we
can
correctly
support
untyped,
metrics,
which
I
think
someone
else
was
working
on.
G
So
I'm
going
to
go,
dig
up
that
issue
and
then,
once
we've
confirmed
that
you
get
some
reasonable
behavior.
When
you
rename
metrics
like
getting
an
untyped
metric
out
the
other
end,
then
we
can
remove
the
validation
and
start
supporting
it
again.
A
B
A
A
C
But
that's
also
not
something
which
we
I
mean
you.
You
can
just
take
the
reference
from
from
openmetrics,
which
has
a
ton
of
valid
and
invalid
cases
and
just
throw
them
at
the
thing
and
everything
which
is
well.
It
needs
to
make
its
way
through
somehow
and
everything
which
is
not
valid,
shouldn't
or
mustn't
and
done.
A
Yeah
I
mean
maybe
we
could
just
add
actually
exactly
that
suggestion
somewhere.
You
know,
maybe
in
the
in
a
different
section
where
we
just
call
out
that
you
know
the
recommendation
is
to
also
run
the
metric
open,
metrics
suite
of
validation
tests,
along
with
these,
from
a
compliance
test.
B
A
A
Examples,
what
do
you
need
do
we
need
to
handle
examples
here.
B
Well,
that's
chasing
and
matrix.
C
I
can
check
the
permissions
remote
right
spec,
but
I
think
it's
defined
so
it
should
be
tested,
of
course,
but
I
can.
I
can
check.
C
It
means
that
personally,
I
would
lean
towards
testing
for
them
already,
but
not
failing
if
they
don't
work.
Of
course
it's
I
mean
it's
it's
it's
self-evident
that
they
will
be
part
of
of
whatever
the
next
release
of
prometheus
realtorite
drivers.
C
C
Yeah,
I
mean
that's
something
which
we
don't
have
in
any
of
the
things
in
in
the
conformance
suit.
Yet,
but
it's
something
which
we
have
consensus
on,
that
we
are
okay,
doing
it,
where
we
discern
between
things,
which
we
require
and
things
which
we
just
note
and
and
rays,
but
don't
actually
error
out.
C
C
Other
stuff,
which
is
experimental,
it's
probably
better
as
an
info
thing,
because
it
might
just
go
away.
If,
if
we
realize
it's,
it's
a
bad
idea
or
we
need
to
change
design
or
what
have
you
so
I
would
avoid
warning
on
on
those.
A
C
F
A
C
A
B
A
B
Will
fail
because
of
the
known
issues
that
we
should
probably
put
it
here,
and
I
will
also
take
a
first
step
at
it.
A
I
mean
we
can
definitely
help
with
some
of
the
building
up
some
of
the
tests,
so
let's
work
on
it
yeah.
Thank
you.
A
Okay,
cool,
so
any
other
questions
on
this
visual.
I
really
appreciate
you
putting
together
the
first
cut
of
this.
It's
always
nice
to
have,
like
you,
know,
a
first
step
and
then
improve
upon
it.
B
So
one
of
the
questions
I
had
was
what
what
would
be
the
difference
between
these
receiver
conforming
to
prometheus
versus
the
remote
right
exporter,
conforming
to
from
ets?
I
think
they
are
all
very
related
right,
because,
whatever
data
the
receiver
produces
you
stitch
with
the,
but
I
think
there
is
a
few.
B
There
are
a
few
things
happening
on
the
remote
right
part
as
well,
so
if
we
can
actually
split
those
two
and
in
terms
of
like
you
know
telling
the
users
that
you
know
these
are
the
things
that
receiver
will
look
for.
These
are
the
ones
that
the
remote
right
will
specifically
look
for
then
that
would
be
great.
A
Yes,
we
do
so
I
mean
I
agree
with
you.
That
was
my
question
also,
because
I
think
that
the
receiver
today
does
more
than
just
remote
right
right,
so
the
tests,
the
number
of
tests
there
or
the
type
of
test
there
would
possibly
be
there
would
be
a
delta
between
the
thermodrug
exporter
and
the
and
the
receiver.
A
A
A
A
David
any
thoughts
there.
A
A
G
I
yeah
I
was
looking
so
grace,
has
a
pr
yeah,
sorry
to
change
topics,
no.
G
Metrics-
and
I
think
we
might
not
be
doing
the
right
thing
today
or
in
grace's
pr,
because
I
think
hers
has
us
convert
untyped
to
gauges,
and
I
think
we
actually,
if
we
can't
want
to
keep
them
untyped
and
should
probably
for
compliance
tests,
that
we
can
accept,
untyped
and
then
transmit
them
still
is
untyped
and
remote
right.
G
Sorry
I
missed
that.
What
was
the
first
thing
you
said,
so
I
was
saying
the
type
information
is
not
actually.
B
It's
not
stored
in
the
in
the
store
right,
so
it
just
so
so
what
you're
suggesting
is
just
send
it
asses,
rather
than
send
it
as
a
as
a
as
a
double
gauge.
G
I
I'm
not
sure
that
would
be
a
question
for
the
the
metrics
data
model
group
and
and
josh
and
riley,
probably
in
particular.
D
I
That
is
something
having
that
expressive.
Power
is
something
that
is
also
useful
elsewhere,
like
what
what
eventually
happened
with
the
the
staleness
martyr
or
the
the,
as
it's
now
in
otlp,
an
expression
that
this
data
point
carries
no
value
which
can
be
used
to
communicate
the
same
semantics
as
a
stainless
marker
and
for
previous,
and
that's
how
we're
choosing
to
use
it.
That
was
a
concept
that
had
broad
applicability
and
thus
made
sense
to
add
to
the
data
model.
I
Having
an
untyped
metric
is
something
that
we
would
need
to
consider
what
applicability
it
has
beyond
this
use
case,
particularly
if
there
are
options
for
us
to
embed
that
information
in
some
other
place
in
the
in
the
data
model
that
the
prometheus
remote
right
receiver
can
extract
it.
If
that's
going
to
be
the
only
thing
that
needs
that
information.
F
Yeah
so
like
the
reason
why
we
even
have
on
types
and
unknown
is
because
you
know
other
metric
systems
out
there.
Just
don't
tell
us
so
like
the
kernel
puts
out
stuff,
which
is
just
an
undis
disambiguated
set
of
counters
and
gauges.
So
the
only
thing
we
can
do
is
use
on
type,
so
it
does
unfortunately
come
up.
G
And
to
be
clear,
it's
important
that
we
propagate
untyped
as
on
type
for
compliance
purposes,
or
is
that
something
that
we
have
flexibility
in
and
could
do
gauges
or
something
like
that?
Instead.
A
But
that's
a
good
call
out
david,
because
I
think
that
we
definitely
need
to
validate
it.
Should
we
create
an
issue
on
our
backlog
to
yeah.
G
I
can
do
that
before
anthony
can
do
that.
G
A
I
Know
what
p
data
is
missing,
but
you
know
what
it
needs.
So,
if
you
can,
if
you
can
state
what
you
think
it
needs,
and
I
can
chime
in
and
and.
D
G
A
Yeah
contributors,
fine,
because
our
prometheus
components
are
all
there
is
fine
and
we
can
link
it
to
our
prometheus
web
group
issues.
A
A
All
right
good,
so
I
think
that
just
very
quickly
just
wanted
to
give
an
update.
Moving
on
on
the
collector,
we
have
a
release
that
we
cut
yesterday,
just
to
give
an
update,
for
you
know
the
backlogged
prs
that
have
been
sitting
and
not
been
reviewed
due
to
you
know
not
having
enough
bench
strength
from
the
approvers
and
reviewers,
and
my
maintainer
is
merging
these
components.
A
We
did
got
a
release
yesterday
that
will
actually
form
the
basis
for
collector
core
providing
tracing
stability,
so
we
will
be
announcing
tracing
stability
for
core
based
on
that
release
and
and
putting
up
an
blog
post
for
that
just
to
have
an
update
for
that,
and
hopefully
that
will
unblock
the
prs
that
are
backlogged
on
our
end
for
metrics.
A
All
right,
that's
all!
I
had
just
wanted
to
give
an
update,
quick
update
on
that
and
again
any
other
questions,
or
you
know
issues
I
will
go
and
go
through
all
the
issues
in
the
collective
act
log
as
well
as
contrib,
and
and
make
sure
that
you
know
those
are
something
we
queue
up
for
reviews
and
merging.
A
Cool
I'll
give
back
folks,
15
minutes
see
you
at
the
collector
meeting
thanks
thanks
everyone.
Thank
you.
Bye
thanks.