►
From YouTube: 2020-12-09 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
A
A
A
A
A
B
B
C
C
F
All
right,
we've
got
a
few
things
on
the
agenda.
I
I
put
my
stuff
a
little
bit
later
a
little
bit
earlier
this
morning.
Did
we
want
to
go
over
this
first
item?
First.
G
F
F
F
So
we've
been
focusing
our
triage
first
on
ones
that
are
labeled
bugs,
so
those
are
self-labeled
by
the
people
who
created
the
issues
just
to
at
least
you
know,
have
some
concentration
on
making
sure
that
anything,
that's
breaking
an
existing
release
is
at
least
looked
at,
and
then
we
cleared
off
all
the
bug,
issues
and
started
going
after
the
non-bug
issues.
F
But
that's
where
we
are
right
now.
Good
news
is:
we've
got
no
p
zeros
that
we
triaged,
which
means
there's
nothing.
This
is
like
the
world's
on
fire.
We
do
have
three
identified
p1
issues
in
the
collector
and
we
have
assignees
for
every
single
one
of
them
and
the
collector
contract.
There's
no
p1
issues.
We
have
p2s
p3s.
F
I'm
also
going
to
take
as
the
standing
item
time
box,
seven
minutes
for
triaging
collector
issues.
We
have
that's
been
open
and
I
will
start
again
with
bugs
first,
which
looks
like
we
got
like
four
to
begin
with.
B
Yeah
I
saw
this
it's
it's
not
really
a
bug,
it's
more
of
a
limitation
of
how
the
patch
processor
works
when
it's
used
with
exporters
which
have
cues
it
actually
is
not
an
issue,
so
I
would
set
it
at
somewhere
p3
and
re-label
it
as
a
feature
request.
F
A
B
B
B
B
And
the
area
is
complete,
spec
doesn't
apply
here
there.
He.
B
B
G
Someone
can't
it,
I
think
there
is
a
code
owner
for
that
for
that
exporter.
B
G
I
can
take
a
look.
You
can
go
through
the
next
item
and
I
can
look
it
up.
F
B
Seems
like
about
the
same
thing:
okay,
yeah,
why
is
very
separate
yeah?
I
don't
know
why.
It's
that's.
Why
there's
two
issues,
but
they
are.
F
B
B
Okay,
this
is
that's
the
receiver
and
I
guess
the
issue
is
that
the
type
of
the
metric
is
not
correctly.
E
A
B
You
remove
one
of
the
labels,
eliminate
one
of
the
labels
and
just
everything
is
enhancement.
F
Yes,
I
I
put
in
actually
a
a
bunch
of
suggestions.
I
need
permissions
to
be
able
to
edit
the
labels
in
order
to
I'm
happy
to
do
this.
We
can
go
over
it
a
little
bit
later
if
you'd
like,
but
I
need
permissions
to
be
able
to
do
that.
B
F
F
Okay,
so
I'm.
F
Or
I
can
really
push
the
sharing
in
order
to
cover
the
the
next
topic.
Let's,
let's
keep
going.
Okay.
B
G
You
want
to
talk
about
it
for
us,
what's
the
sure
yeah.
So
basically,
we
have
this
the
tail-based
sampling
processor,
on
on
the
contributor
right
now,
and
we
have
a
set
of
other
process
tours
that
that
were
created
with
the
idea
of
replacing
parts
of
the
tail
base
sampling
such
as
the
the
group
by
trace
right.
We
also
have
the
load,
the
the
load
balancer
with
the
trace
id
our
load
balancer
to
front
that.
G
G
G
So
I
sent
that
for
review
and
kevin
made
a
good
comment
there,
basically
about
allowing
other
people
to
create
new
policies,
and
I
think
we
can.
G
We
can
actually
go
one
step
further
and
instead
of
having
one
policy
process
or
policy
sampling
processor,
we
could
actually
split
all
those
policies
into
individual
processors
so
that
we
end
up
having
a
rate
limiting
process
or
an
attribute
sampling,
processor
and
people
can,
then
you
know,
create
their
own
sampling
processors,
perhaps
perhaps
based
on
these
ones.
G
B
Yeah
yeah,
I
think
it's
it
makes
sense
generally
as
an
idea
to
decompose
things
into
smaller
processors.
Just
keep
in
mind
that
the
processors
today
only
work
sequentially
and
if
I
remember
correctly,
intel-based
sampling,
processor,
the
different
policies
they
work
in
parallel
in
the
sense
that
they
split
the
allocated
budget
off
of
the
boundary.
If
I
remember
correctly,
I
may
be
wrong
on
this,
so
it's
not
always
possible
to
substitute
it
with
multiple
chained
processors.
B
I
think
it's,
it's
probably
difficult
to
make
this
decision
on
the
spot.
It
would
be
beneficial
to
see
maybe
a
quick
write-up
on
what
the
proposal
looks
like
and
what
the
end
goal
is
what
we
end
up
having.
But
yes,
I
agree:
that's
that
that
was
the
way
right.
We
wanted
to
have
processors,
which
could
be
composed
separately
in
creative
ways.
If
you
want
to.
G
Yeah,
can
you
go
back
to
the
to
the
comment
that
I
left
on
the
on
the
issue?
Yeah.
I
think
that
kind
of
shows
what
is
the
the
final
final?
Not
not
the
goal
for
the
whole
same
plane
or
the
tail
base
simply
but
a
goal
for
splitting
tail
base
sampling
to
other
processors,
so,
basically,
two
other
processors.
G
Only
one
is
the
rate
limiting
and
the
other
one
is
attribute
sampling
processor.
In
terms
of
you
know
the
final
or
the
the
end
goal
or
the
final
picture.
It
is.
A
G
I
don't
know:
are
we
okay
or
are
we
willing
to
consider
the
possibility
of
deprecating
the
tail-based
sampling
before
ga?
B
Close
right,
it's
right
we're
leaving
very
short
notice,
and
we
try
to
do
that.
Let's
say
we
declare
it
deprecated
today
we're
likely
to
make
the
release
in
a
couple
weeks.
It's
just
very
short
notice.
I
feel
comfortable
doing
that.
I
think
it's
fine
if
we
declare
it
deprecated,
but
we
probably
need
to
keep
it
around
for
a
bit
longer,
which
is
okay
right.
It's
it's
a
separate
processor.
We
can
keep
it.
B
H
I
I
like
the
way
you're
going
going
here
now
yeah.
The
other
thing
that
you
didn't
mention
that
that
I
was
thinking
about
is
you
know
all
these
different
enterprises
got
weird
stuff
and
there
there's
going
to
be
a
whole
bunch
of
them
to
say.
Oh,
we
have
this
special
thing
that
only
us
could
do,
and
you
know
I
just
wanted
to
get
it.
G
B
I'd
still
prefer
to
see
a
big
picture,
maybe
maybe,
if
it's
just
just
even
a
one-page
document
which
describes
what
is
it
that
we
want
to
achieve
in
the
end,
if
it's
scattered
across
different
github
issues,
it's
a
bit
difficult
to
understand
would
be
beneficial
to
have
a
google
doc
where,
even
if
it's
just
one
page
that
we
can
read
and
comment
and
have
a
discussion
around
it,
you
mean.
B
Yes,
for
the
future
of
what
the
sampling
looks
like
what's
the
combination
of
processors,
we
want
to
have
what
do
we
deprecate
right?
All
of
that,
what
we
just
discussed-
okay,
but
in
maybe
in
a
bit
more
confidence
in
a
detailed
way
in
a
single
document
that
we
can
review
and
ask
others
to
review
or
were
interested
and
then
decide.
That's
that's
the
roadmap
make
make
it
part
of
the
roadmapping.
G
Okay,
may
I
suggest
an
a
a
better
place
for
that.
There
is
something
that
I
just
discovered
today
called
you
know,
github
discussions,
which
is
probably
very
suitable
for
that
kind
of.
B
So
as
an
action
in
the
country
repository.
G
G
The
country,
that's
culture
yeah
so
actually
for
me,
then,
is
to
create
this
document
and
and
discuss
with
you
all,
and
perhaps
one
one
thing
that
we
have
to
to
decide
kind
of
quickly
is
whether
that
initial
draft
is
enough
to
deprecate
this
detail
base
simply
now
everything
else
can
be
figured
out
later.
I
think
yep.
I
I
will
just
add
two
like
quick
comments
on
on
this,
because
we've
been
researching
using
telesampling
and
we
found
it
being
missing
some
capabilities,
so
one
of
them
is
that
we
found
that
great
limiting
goes
with
definition
of
the
attributes,
so
you
might
want
to
have
different
rate
limits
for
different
attributes,
and
the
second
thing
is
that
those
attributes,
as
they
are
defined
right
now
as
policies
and
you
need
to
you're
creating
separate
policy
for
each
type
of
the
attribute
so
separate.
I
For
let's
say,
numerical
attribute
separate
for
string
attribute-
and
this
is
all
also
better
to
be
if
it's
possible,
to
put
it
in
the
single
place
as
a
single
policy.
And
the
third
thing
is
that
what
I
have
found
is
that
you
can
have
different,
let's
say,
classes
of
of
traces
present
together
and
you
still
want
to
have
some
global
rate
limiting
capability,
and
this
makes
things
much
more
complicated.
So
actually,
what
I
did
is
I
have
taken
a
sampling,
processor
and
built
a
prototype.
I
I
called
it
the
cascading
filter
processor
and
which
seems
to
work
just
fine
and
solves
those
issues,
so
maybe
I'll
just
put
it
in
the
comment.
If
anyone
would
like
to
review
it
and
and
see
if
it
makes
sense,
my
original
idea
was
to
propose
it
after
it's
really
another
way
to
to
handle
this.
This
problem.
H
G
Except
that
what
I
understood
is
that
each
one
of
those
sampling
processors
would
then
have
a
its
own
rate.
Limiting
attribute
and.
I
I
But
let's
say
that
you
have
a
budget
of
ten
thousand
or
one
hundred
thousand
spans
per
second
or
say
whatever
one
hundred
thousand
pounds
per
second
and
then
you
can
assign
budgets
to
different
rules,
and
then
you
can
say
that
I
want
to
have
a
rule
that
takes
any
available
budget.
If
there
is
something
so
you
can,
for
example,
have
some
specific
types
of
traces,
for
example,
traces
with
errors,
traces
with
extra
duration
and
things
like
that,
and
then
let's
say
some
budget
for
traces
that
are
just
a
random
representation
of
anything
that
came.
G
Okay,
so
at
the
end,
like
a
probabilistic
sampling
for
the
remaining
budget,.
I
Yeah-
and
the
thing
is
that
it's,
it
also
supports
probabilistic
sampling
defined
by
the
maximum
rate.
So
you
can
say
that
I
want
to
have
no
more
than
1000
pounds
per
second
and
it
will
like
for
each
and
process
the
package
each
second,
it
will
calculate
the
current
sampling
rate
by
selecting
this
proportion
of
spans
or
traces
that
fit
within
the
budget.
I
G
We
can
discuss
that
in
the
in
the
new
issue,
but
I
think
those
are
that's
actually
a
good
feedback
to
have
for
the
new
processors.
F
Administrative
tasks:
do
you
really
do
you
really?
Okay?
Well,
I
I
got
some.
This
will
help
reduce
the
number
of
labels
and
also
help
with
the
reporting
and
the
redundancy
of
such
things.
F
I
have
suggestions
on
refactoring
of
labels
in
the
collector
repo
just
giving
the
redundant
log
label
and
then
deleting
approved
and
working
progress
labels
and
getting
rid
of
the
redundant
feature,
request
label
and
I've
given
reasons
for
each
one.
Subsequently,
for
the
contrib
repo,
we
have
unused
labels
trace
metrics
logs.
We
also,
we
already
got
these
spectre
spec
metric
spec
logs,
a
simple
rename
and
deletion
of
the
working
power
samples.
F
B
Yeah
make
sense:
let's
see
if
we
can
give
you
the
permissions.
I
don't
know
why
you
don't
have
them
if
there
is
a
reason
why
you
can't
have
them,
then
just
let
me
know
I
will
do
the
changes.
Whatever
changes,
I
need
to
say,
but
I
I
prefer
you
to
be
able
to
do
that.
F
Okay,
sure
yeah
and
let
me
see-
I
guess
I
haven't
put
down
as
a
gen
item,
but
are
we
happy
with
how
we're
progressing
with
the
triaging
of
the
issues
we
were
able
to
achieve,
like
100
triaging
for
spec
repo,
we're
at
33
about
32
percent
through
the
collector
and
collector
contract
repos?
B
B
I
think
it's
good.
We
made
this
significant
progress
last
time,
but
probably
we
need
to
allocate
more
time
regularly.
So
maybe
seven
minutes
is
not
enough
and
given
that
we
have
time
today,
maybe
spend
a
bit
more
time
and
and
also
regularly
maybe
make
that
10
minutes,
at
least
because
I
think
you're
right.
We
have
a
very
long
scale
of
long-standing
issues.
There.
A
F
F
Yeah
so.
B
F
Back
to
the
collector
issues-
and
let
me
see
these
are
the
untriage
ones.
Let's
start.
H
F
I
also
see
it
an
appropriate
spec
or
area
label
for
this.
This
tracing.
B
This
is
a
processor
if
it's
in
the
batching,
if,
as
we
do
as
bogdan
suggests,
this
is
the
feature
of
patch
processor,
then
it's
an
existing
processor.
Otherwise,
it's
it's
a
new
process
we'll
need
to
see,
but
it's
a
processor
either
way
all.
B
B
B
B
G
H
G
So
it's
a
great
need
some
clarification
here.
So
if
they
want
to
change
on
the
fly
like
you
say,
then
then
yeah
and
touching
a
file
is
one
way
of
doing
that,
even
though
I
don't
think
that's
the
right
one,
but
if
they
they're
complaining
that
there
is
no
way
to
change
the
log
level
for
the
service
can't
pass
command
line
parameters
to
a
service,
I'm
not
sure
for
windows,
but
for
for
linux,
sure.
H
C
G
That's
that's
what
they
should
do
there
yeah.
I
agree.
So
I
think
I
think
it
needs
a
a
clarification
here,
so
whether
they
mean
changing
why
the
server
is
is
running
or
if
it
is
about
configuring.
You
know
the
service
facility
like
system
g
or
I
don't
know
what
it's
called
yeah.
G
Yeah,
I
can
ask
the
question
there
2255..
I
can
ask
the
question
there.
If
you
can
just
let
leave
it
waiting
for
all
yeah,
okay,.
F
But
this
is,
it
is
an
enhancement.
F
F
Must
be
something
that
needs
more
discussion.
F
F
B
F
This
is
an
enhancement,
yes,.
B
Yeah
this
is
anyway
we
we,
we
can
maybe
triage
this
better
after
that
discussion,
but
for
now
I
suggest
we
mark
it
as
after
ga,
p2
p3
or
whatever
yeah,
and
it's
it's
a
processor.
F
H
G
Yeah,
I'm
watching
this
one.
So
as
soon
as
tyrone
adds
a
discussion
feature
to
the
repo
I'll,
add
a
link
here.
F
Okay,
an
associated
spec
label.
G
B
B
G
It
is
similar
to
what
we
have
in
the
authentication
authentication
mechanism,
so
we
use
a
a
stream
interceptor
or
a
message
interceptor,
but
in
this
case
here
it's
what
they're
asking
is
basically
a
more
generic
feature
than
the
authentication
mechanisms
that
we
have
that
we're
thinking
about.
So
it's
also
something
that
we
have
right
now.
It
might
be
something
that
we
might
add
in
the
future.
G
D
F
Okay,
so
we
are
closing
in
on
the
october
ones.
We
can
switch
over
to
the
contra
repo
if
you'd
like
in
order
to.
F
Is
it
okay.
We
spend
like
another,
maybe
five,
ten
minutes
on
this.
J
B
B
B
Ideally,
we
should
also
sign
the
the
concrete
ones
to
the
code
owners.
I
don't
remember
the
code
owners
from
the
top
of
my
okay,
so
I.
G
So
exporter
aws
x-ray
exporter
right.
Yes,
then
kevin
is
one
of
those
owners
who
is
an
approver
right.
F
B
I
I
know
I
know
what
it's
it
depends
on
the
on
the
the
easiest
expo
and
the
core
repository
so
yeah,
whatever
we
did
with
that
one.
It's
to
be
done
the
same
way
here.
G
Very
ticking,
so
there
we
go
the
side
note.
Then
there
seems
to
be
liking
on
and
entering
the
code
owners
for
this
exporter.
For
that
the
last
one
which.
G
C
B
C
B
Okay,
stackdriver
james
is
already
assigned
yeah,
make
it
p3.
G
So
this
is
interesting
because
we
had
a
similar
discussion
on.
I
think
it
was
even
for
the
authentication.
No.
That
was,
for
the
better
token
feature
for
for
exporters,
I
think
and
back
then
we
we
thought
that
we
would
have
a
generic
feature
on
the
collector
to
read
properties
from
files
and
that's
why
I
haven't
implemented
yet
so
is
it
still
a
thing
or
are
we
really
or
are
we
now
free
to
load
things
from
files
with
properties?
G
What
do
you
mean
instead
of
having
credentials
path
here?
We
would
have
credentials
and
then
some
notation
like
file
and
then
open
parentheses
and
then
file
name.
G
Value
from
a
sourcing
from
a
file,
not
the
configuration
snippet,
but
I
enact
the
value
for
a
specific
configuration
option.
B
B
G
Like
we're
talking
from
a
file,
it
was
said
that
such
a
feature
does
not
belong
yet
on
the
collector.
We
should
have
a
this
facility
first,
but
if
we
are
having
like,
if
we
we
are
allowed
to
load
from
files
and
I'll
change,
the
data
receivers
as
well.
B
I
don't
think
we
can
prevent
the
components
from
doing
things
like
that,
but
we
should
recommend
right
some
uniform
way
of
handling
these
issues,
but
we
don't.
Since
we
don't
provide
the
mechanism
right
now.
I
don't
think
we
should.
We
can
force
people
to
to
do
it
in
a
way
that
does
not
exist
yet.
Okay,.
E
I
just
want
to
bring
up
one
other
thing
related
to
this,
which
is
that
james
and
niall
are
planning
to
roll
off
of
the
open
telemetry
team.
Here
at
google,
and
I'm
david
I've
recently
joined
and
I'm
going
to
be
trying
to
pick
up
any
work
that
relates
to
the
stock
driver
exporters.
E
C
B
Okay,
if,
if
you
did
reviews
and
if
you
qualify
for
being
the
member,
please
submit
the
request.
J
D
You
can
oh
great
morgan's
here,
yeah
yeah,
we're
definitely
still
interested
right
like
this
is
just
to
be
clear.
This
is
the
basis
of
all
of
our
our
vm
based
agents
going
forward
yeah.
So
we're
still
extremely
interested
in
that
james.
A
D
I
mean
yeah
technically
the
back
end
in
not
nothing
to
open
telemetry,
but
yes,
cloud,
ops,
cloud
monitoring
cloud
logging
now,
rather
than
stackdriver,
whether
we
rename
our
components
in
open
telemetry-
I
don't
know
I
mean
people
are
already
using
them.
So
probably
not.
B
Okay
yeah:
this
should
be,
I
think,
still
after
ga,
unless
you
want
to
do
it
earlier,
in
which
case
we
can
relabel,
but
for
now
make
it
after
you.