►
From YouTube: CNCF SIG Observability 2020-09-01
Description
CNCF SIG Observability 2020-09-01
A
A
B
C
C
D
D
Okay,
while
we're
waiting
I'll
just
say,
welcome,
this
is
the
sick,
observability
first
meeting
of
the
fall
after
taking
august
off
as
many
people
were
on
holiday
and
kubecon
happened,
I'll
remind
everyone
as
we're
supposed
to
do
that.
This
is
a
cncf
event
and
the
code
of
conduct
applies.
Obviously,
so
it's
never
been
an
issue
before.
But
it's
always
worth
saying.
Bear
that
in
mind
with
any
content
in
chat
and-
and
talking
is
richard
here
as
well-
was
dropped
I'll,
be
back.
D
A
A
Unfortunately,
so
that's
work
for
the
agenda.
We
didn't
have
any
toc
updates,
obviously,
because
we
didn't
have
any
any
any
calls
in
orders.
But
what
happened
is
the
tlc
voted
on
both
cortex
and
thanos
and
actually
finished
voting,
and
so
both
are
now
incubating
projects,
as
of
cubecom
was
used
as
the
announcement
for
both
of
those
and
that's.
D
Basically
it
for
plc
there
was
one
other
update.
I
spoke
with
amy
today
it
looks
like
we
have
11
binding
votes
on
the
election
for
our
tech
lead.
So
someone
on
the
call
deserves
a
a
a
round
of
a
welcome
and
congratulations
and
all
of
that.
I
believe
they
should
be
putting
out
some
communication
in
a
bit.
D
C
Cool,
so
I
can.
I
can
start
with
with
next
topic
straight
away.
We
have
like
just
want
to
kind
of
get
get
back
in
kind
of
the
major
thing
we
did
like
before
the
break
in
august.
We
are
talking
about
analytics
apis
for
observability
signals
and
we
started
with
like
prometheus
and
how
to
get
some
analytic
use
cases.
On
top
of
that,
so
there
didn't,
I
mean
nothing,
major
changed.
C
Since
then,
however,
we
took
some
action
on
the
red
hat
side
and
we
started
the
short
like
poc
project
for
certain
apis
and,
like
the
the
report,
is
called
opslytics
on
the
thanos
community.
Please
join
us
and
and
contribute
the
idea
is
to
have
just
just
a
proxy
that
will
essentially
understand
the
promoters
and
thanos
export
apis,
which
is
remote,
read
and
store
api.
C
They
are
really
really
similar
and
it
will
expose
a
certain
app
analytics
apis
and
right
now
we
focus
on
parket
file
and
we
also
want
you,
for
you,
want
to
add
an
apache
arrow
flight,
which
is
a
jrpc
endpoint.
So
this
is
the
plan,
but
we
essentially
it
won't
be.
You
know
super
efficient
from
the
start,
but
it
will
do
the
work
and
enable
folks
to
to
just
use
metric
data
much
more
efficiently
in
those
systems.
C
So,
if
you
want
to
have,
I
don't
know,
if
you
want
to
add
end
point
to
to
to
to
to
you
know,
support,
maybe
other
observability
signals
or
add
any
analytic.
Apis
yeah
feel
feel
free
to
propose
that
in
the
github
issue
on
the
pr,
so
we
can,
we
can
look
on
that
and,
at
the
same
time,
a
kind
of
reminder.
We
are
going
through
use
cases,
and
this
is
the
doc
like
this
sick,
observability
data,
analytic,
use
cases,
high-level,
use
cases.
C
What
else
we
can
do
in
our
community
in
our
projects
in
terms
of
what
is
missing
essentially
for
analytic
use
cases.
So
I
think
we
can
get
back
to
this
at
some
point
for
the
next
meeting,
but
please
yeah.
Let's
re-read
this
on
its
own
time
and
and
add.
If
anything
is
missing,
I
think
we
are
in
the
middle
of
this
dock
or
something
so
from
the
next
meeting
weekend.
We
can
continue
that
so
I
think
that's
the
status
that
was
happening
on
our
side,
the
anyone.
A
C
A
C
Something
that
will
share
an
end
point
of
apache
flight,
for
example,
or
it
will
just
a
bad
job
that
will
transform
whatever
data
you
want
into
pocket
files.
So
this
is
kind
of
two
modes
that
we
are
aiming
for.
One
is
a
bad
job
and
one
is
like
service,
stateless
service
that
will
transform
the
data
from
one
format
to
another,
hopefully
without
not
much
overhead.
So
that's
the
plan
current.
A
D
A
E
D
A
D
Clear,
starting
with
all
the
all
the
hullabaloo
I
I
missed,
the
question
from
me
is:
is
the
idea
here
that
that
this
is
something
this
is
an
api
that
you're
implementing
in
in
thanos
to
to
help
export
durability
data,
or
is
that
just
expedient
to
create
it
there?
And
we
really
think
that
this
is
something
that
would
live
that
would
use
either.
You
know
some
that
would
work
on
prometheus
itself
versus
something
that
requires
thanos.
C
No
so
this
is,
this
is
only
well.
We
needed
that
for
the
thanos
community,
but
because
thanos
has
really
similar
api
to
promote
use
just
grpc.
This
is
also
working
against
against
thanos
right
and
for
now,
like
with
everything
else
like
we.
We
talked
about
use
cases,
and
you
know
what
apis
are
the
most
important,
but
we
actually
have
a
real
use
cases
on
in
such
red
hat
to
use
observability
data
for
parkette
well
for
something
that
consumed
pocket
files.
C
So
we
went
ahead
and
created
this
helper,
and
this
will-
and
this
is
actually
a
good
first
step,
because
then,
if
we
see
that
there
is
like
super
high
usage
of
it
and
and
very-
and
maybe
there
are
some
blockers
or
something
that
needs
enhancement,
then
it
will
be
a
good
step
to
to
forward
discussion
on
you
know,
maybe
promote
you
should
have
that
by
default
on
on
their
side.
So
you
know
this
is
actually
some
actionable
thing
that
we
chosen
to
do
so.
C
C
What
does
they
think
like?
We
are,
at
least
for
for,
let's
say
and
that's
a
good
point
like
what's
the
scope
of
this
of
this
tool,
this
tool,
we
want
to
make
sure
that
you
can
use
the
data
that
are
already
ingested
by
prometheus
tunnels
or
even
cortex
like
can
can
that
can
can
easily
export
this
api.
C
There
are
other
already
tools
that
kind
of
allows
replicating
in
different
databases,
but
we
want
to
kind
of
allow
using
the
data
that
already
are
consumed
and
stored
in
one
place
and
try
to
avoid
replicating
them
in
many.
Thousands
of
you
know
other
locations
that
are,
maybe
you
know,
have
different
apis.
That's
why
this
is
a
helper
for
like
a
pooling
method
like
pulling
only
the
data
that
you
want
for
for
analytic
purposes
and
that's
why
remote
right
doesn't
fit
here
yeah,
but
you
can,
you
can
totally
do
do
this.
C
D
No,
I
hope
to
have
some
better
updates
in
a
couple
weeks
within
our
own
team.
About
two
weeks
ago,
we
started
working
on
some
of
the
scenarios
that
are
captured
in
the
dock,
that
we
had
talked
about
before
around
machine
learning
and
model
training
and
and
we've
since
deployed
cortex
in
our
case
and
we're
looking
at
various
ways
to
do
the
same.
So
I'll
have
a
look
at
what
you've
posted
from
the
thanos
community
and
yeah
I'll
have
some
more
concrete
results
in
a
couple
weeks.
D
But
it's
it's
been
interesting
to
see
a
team
that's
used
to
having
to
ingest
data
and
then
build
an
entire
database
just
to
hold
things
and
they're
they're
they're,
looking
at
prometheus
and
the
data
model
for
the
first
time
and
and
they're
finding
that
they
might
be
able
to
prune
a
lot
of
work
just
due
to
how
prometheus
is
already
so
optimized
for
time.
Series
data,
so
results
so
far
positive.
But
we
still
have
this
core
need
to
be
able
to
bulk
bulk
read.
C
C
You
know,
writing
their
own
tools
from
scratch
and
then
having
and
just
reusing
the
the
data
that
is
available
in
front
side
and
for
particular,
your
use
case
like
make
sure
to
also
poke
cortex
to
enable
remote
read
or
we
can
talk,
how
to
do
it
internally,
because
the
data
model
we
have
really
have
the
same.
So
it
would
be
amazing
if
you
can
reuse
exactly
the
same.
The
same
tool
and
yeah
we'll
be
up
for
that
for
sure.
D
B
B
Cncf
is
moving
away
from
meetups
and
is
planning
on
self-hosting
the
entire
community
platform
along
with
bevy,
and
that's
what
the
concept
of
cloud
native
community
group
is.
So
the
first
suggestion
that
I
had
was
like
so
with
cg.
The
first
thing
that
we
have
is:
we
have
divided
the
whole
community
into
different
chapters
according
to
locations
right
now,
right
and
that's
just
a
horizontal
distribution
across
the
globe.
I
was
hoping.
B
Maybe
we
could
have
a
vertical
chapter
sort
of
for
observability
and
with
baby
you
can
add
sub
chapters
onto
it.
So
basically,
if
you
add
a
single
event
to
a
sing-
let's
say
a
parent
chapter,
it
shows
up
in
all
your
sub
chapters
right.
So
if
we
could
have
a
vertical
chapter
around
there
and
then
we
can,
you
know,
make
all
of
these
locations
of
chapters
as
location
chapters.
B
I'm
really
sorry
within
the
community
itself,
and
as
for
kubecon
itself,
we
had
keynote
passes
and
everything.
So
it
might
be
a
good
idea
if
we
could
have
these
keynotes
available
on
that
community
channel
itself,
because
that
makes
engaging
the
people
working
with
the
community
within
the
community
chapter
itself.
B
The
third
thing
that
I
had
was
it's
just
a
general
suggestion
that
we
post
these
videos,
or
these
talks
or
any
useful
links,
that
we
have
resources
within
that
community
itself,
and
you
can
do
a
bunch
of
different
things.
Yeah
you
can
have
newsletters
and
everything
over
there
within
the
single
chapter,
so
people
generally
have
a
common
place
where
they
want
to
talk
to
everyone
interested
in
observability
right.
A
B
Can
have
better
resources
around
that
and
the
last
thing
that
I
had
was
basically
so
what
I
found
when
I
was
trying
to
contribute
to
the
observability
community
and
sig.
Is
we
don't
have
a
lot
of
documentation
at
the
moment
around
it
right,
and
it
might
really
be
a
good
idea
to
have
some
sort
of
documentation
so
that
people
know
how
they
can
contribute
to
the
observability
community
on
a
broader
scale,
and
not
just
for
project
specific
like
contributing
to
the
community
as
a
whole
right
so
yeah.
B
D
Great
thanks
for
thanks
for
for
covering
that
I
put
a
couple
links
in
the
meeting
document
in
terms
of
the
this,
the
videos
from
the
from
our
sig
meetings
now
and
moving
forward
the
cncf
records
and
posts
all
the
the
videos
in
a
playlist
under
the
cncf
youtube
channel.
So
if
there
are
our
community
groups
or
or
other
places,
you
want
to
link
to
that,
that
would
be
the
authoritative
source,
at
least
from
the
sig
mechanisms.
D
In
addition
for
some
of
the
documents
that
we
review
moving
forward,
I
had
hoped
to
have
the
link
from
amy.
Today,
I'm
told
it's
been
provisioned,
but
we
have
a
google
drive
as
well
for
google
documents
moving
forward
that
we
can
have
for
sick
observabilities
use.
I
must
confess
that
I
had
somehow
missed
the
blog
on
the
community
groups,
so
I'm
I'm
not
an
authority
or
anything
does
anyone
else
have
feedback
or.
C
I
think
we
also
talked
about
some
entry
point
for
for
like
observing
topics
so
that
that
once
finished
we
can.
We
can
definitely
point
to
that,
and
my
question
overall
tutorial
think
like
it's,
not
amazing,
but
what
actually
actionable
is
here
from
from
from
this
sick.
Should
we
sign
up
for
for
sampling?
Should
we
announce
the
keynotes
or
have
ability
for
keynotes
look
for
presenters,
how
we
can
help
essentially.
B
So
what
I
was
essentially
looking
for
is
the
current
limitation
that
we
have
on
community
groups.
Is
their
location
based
right,
so
you
of
you
have
sigs
for
local
chapters,
but
you
don't
have
them
on
a
broader
level.
So,
let's
say,
for
instance,
if
you
are
in
mumbai
right,
you
wouldn't
probably
join
some
chapter
in
new
york
or
san
francisco
unless,
until
you
have
some
very
special
interests
there
right
or
you
won't
have
visibility
onto
it.
That's
the
primary
point
right.
B
So
it
is
predominantly
our
partners
that
we
have
new
relay
candidates
and
everyone
right,
but
they
still
don't
have
enough
community
around
it.
They
don't
have
enough
contact
around,
so
they
can
work
with
open
telematic
or
other
tools
like
that
right.
They
don't
have
they
just
don't
have
that
at
the
moment.
So
having
a
broader
community
might
also
help
with
that.
B
B
D
Yeah,
I
think,
oh
no,
I
think,
as
we
said,
it
might
be
helpful
if
what
was
being
asked
for
or
what
the
actionable
next
steps
were
captured,
either
in
a
document
or
or
ideally
in
a
github
issue.
In
the
sake
of
durability,
github
repo
and
then
we
can,
you
know,
then,
as
a
community,
we
can
have
some
some
feedback.
D
C
B
C
D
Sure,
yeah
and
and
in
terms
of
you
know
the
the
I
don't
know
it's
a
parochial
but
the
the
local
aspect
of
it.
You
know,
like
I
think
time
zone
makes
sense,
but
at
least
in
at
least
in
north
america,
or
for
some
outside
north
america
too.
You
know
with
the
current
state
of
affairs.
You
know
what
I'm
not
sure
what
what
it,
what
it
means
necessarily
in
practice,
because
at
least
you
know
most
of
the
local
meetups
and
such
like
that
are
in
person.
Are
these
aren't
happening
until
things
so.
B
B
G
Yep
hi
everyone,
I'm
ganesh,
so
I'm
currently
working
on
extending
from
ql,
starting
with
a
very
specific
case
of
extending
the
current
top
k
function.
So
before
I
talk
about
what
it
is,
let
me
explain
what
the
current
problem
is.
So
prometheus
just
allows
a
query
and
a
query
range
and
query
range
is
just
a
syntactic
sugar
or
query
where
it
runs
the
same
query
or
multiple
timestamps
and
returns
individual
results
for
those
individual
times
time.
G
G
The
use
case
that
I
am
trying
to
solve
is
you
might
want
to
look
at
some
aggregation
over
the
past
few
hours
or
past
some
time
and
just
pick
top
few
matrix,
let's
say
5
metrics
out
of
all
those
aggregations
and
just
plot
five
matrix
for
that
entire
range.
Instead
of
getting
thousands
of
metrics
for
every
evaluation
timestamp.
G
Currently,
that's
not
easily
solvable
with
current
prompting
or
current
apis,
because
in
the
query
range
you
cannot
use
the
start
and
end
or
other
parameters
of
the
query.
Range
and
promptql
doesn't
allow
any
such
extensions.
So
currently,
I'm
trying
to
solve
this
particular
problem.
I
am
currently
working
on
a
design
dock
and
exploring
different
parts
to
solve.
The
two
directions
that
are
possible
are
extending
the
prom
kill
itself,
which
sounds
more
promising,
which
might
even
solve
more
problems
than
just
x.
G
Extending
top
k
and
the
other
option
is
introducing
a
new
api
which
doesn't
have
restrictions
like
the
query
range
currently,
but
that's
not
so
good
solution
when
it
comes
to
extended
prompt
cure.
So
currently
I
am
focusing
on
extending
from
kill.
So
if
anyone
has
more
ideas
on
how
you
want
to
extend
from
ql
and
more
use
cases,
you
can
reach
me
out
on
slack,
my
name
is
ganesh
vernicker
other
than
that
I
don't
have
much
to
share
in
terms
of
solutions.
G
C
G
C
D
Yeah
yeah.
I
want
to
be
be
careful,
though,
that
you
know
this
really
is
something
for
the
prometheus
community
to
provide
feedback
on
and
has
governance
over
in
terms
of
prom
ql,
but
we
can
certainly
surface
it
here
as
a
a
call
for
you
know
if
anybody
else
is
interested
or
or
such.
So
thank
you
for
bringing
it
to
the
city.
D
There's
no
other
feedback
or
comments
on
this
one.
Let's
move
on
the
next
one's
mine.
I
actually
could
do
this
in
about
30
seconds.
It's
really
just
an
ask.
I
put
some
details
there,
but
we
had
talked
back
in
june
or
july
in
in
some
of
the
charter.
Discussions
around
you
know
providing
something
from
the
sig
that
a
new
developer,
that's
new
to
the
open
source
space
or
new
to
the
cncf
stack
could
have
quickly
working
on
it.
D
C
Yeah
yeah,
have
you:
have
you
seen
much
the
katakota
tutorials,
I'm
sure
they
were
katakota
tutorial
platform
like
actually
in
the
dark,
so
this
is
like
particular
course
for
thanos.
We
are
working
on
and
this
is
a
platform
that
literally
gives
you
opportunity
to
just
play
around
those
things
and
play.
You
know
kind
of
just
run
the
tutorials
across
different
components
like
right
now
it's
about
tennis,
but
you
can
actually,
there
are
already
tutorials
about
kubernetes
prompt
use,
other
projects,
and
it
essentially
works
in
our
browser.
C
So
maybe
this
is
what
we
want
to
use.
It's
kind
of
you
know
open,
oh
well,
free
to
use
for
open
source
projects,
so
it's
kind
of
much
much
more
ready
than
than
just
trying
to
set
up
the
yeah,
the
airport
actually
airplane
laptop
motor.
Something
like
that.
So
anyway,
I
will
add
that
to
the
other
issue,.
D
I
have
I've
found
again
some
like
docker
compose
files
that
have
been
particularly
useful.
That
are
a
good
starting
point
that
have
you
know
a
bunch
of
these
tools
put
together,
but
it's
a
bit
of
a
patchwork
now,
so
I
was
curious
if
having
something
a
little
more
curated
by
the
sig
for
new
contributors
to
the
space
might
make
sense.
So
maybe
we
could
do
this
on
one
of
these
other
platforms.
H
H
I
have
some
repo
that
I
can
share
with
you,
or
maybe
I
can
help
in
your
goal
to
create
some
kind
of
small
projects
or
examples
to
use
docker
compose
and
that
things
I
think
that
maybe
I
can
join
forces
with
you
with
the
sick.
So
I
I
think
that
I
can
share
here.
My
repo
is
it's
like
a
kind
of
well.
I
am
trying
to
create
some
space
for
the
students
in
the
cncf
ecosystem.
H
H
D
Yeah,
that's
that's
great
on
the
link
or
d
subject:
that's
what
my
team
runs
in
production
as
well,
and
we're
actually
we're
finding
that
it's
great
to
onboard
people
to
cncf
observability
stack
in
general,
because
the
the
link
rd
proxies
can
generate
trace
headers
as
well
as
time
series
metrics
in
prometheus
format.
So,
for
example,
on
the
on
the
tracing
side,
even
for
services
that
do
not
integrate.
H
Yeah,
I
have
some
examples
using
the
open,
telemetry
library
too,
so
I
can
integrate
that.
I
am
trying
to
create
some
very
simple
examples
for
the
students,
because
it's
really
hard
for
for
them,
because
it's
a
new
technology
and
it's
like
totally
wow.
What
what
is
this
thing
and
yeah?
I
think
that
I
can
help
in
that
way.
D
Great
thanks,
so
I
guess
the
last
thing
I
we
had
on
the
agenda
was
just
update
your
discussion
regarding
existing
tasks
that
are
already
as
marked
as
issues.
I
don't
want
to
do
necessarily
an
issue
walk,
because
that
might
be
a
little
torture,
but
is
there
anything
folks
wanted
to
discuss
that
they're
working
on
or
that
they
want
to
propose
or
that
they
want
to
just
get
some
put
some
visibility
on
from
the
sig.
F
Hey
this
is
actually
related
to
the
previous
question.
We
have
a
brand
new
tool-
it's
not
out
of
beta
yet,
but
it's
a
cli
tool
to
actually
bring
up,
grafana,
prometheus,
etc,
and
it
comes
with
a
helmet
as
well
posted
on
here.
It's
called
tabs
and
right
now
it
integrates
time
scale
db
as
the
back
install,
but
we
welcome
contributions
from
other
back
ends
as
well
so
kind
of
related
to
both
of
these
topics.
I
I
So
we
are
basically
changing
how
we
build
the
system
and
how
the
deployments
are
going
to
look
like
in
the
future
as
well.
So
the
the
let's
say
the
5g
is
not
going
to
be
the
usual
operator
idea
that
we
have
with
this
antenna
the
masses
outside
it
can.
They
are
going
to
be
much
more
like
smaller
and
complex,
and
one
part
of
building
these
systems
is
in
the
troubleshooting
and
observability
area.
I
So
you're
moving
from
a
more
system,
softer
type
of
tracing
observability
idea
to
a
cloud
native
and
one
of
the
ideas
that
I
was
trying
to
bring
here
or
discuss
more.
It
was
also
in
the
in
terms
of
white
papers
and
also
having
bringing
this
this
discussion
in
terms
of
the
5g.
Also
to
this
dig,
I'm
not
sure
I
had
the
impression
that
you
guys
are
more
focused
on
the
tools
and
some
things
that
you
are
building
and
improving.
I
Right
now
I
would
say
that
maybe
I'm
in
my
daily
life,
I'm
like
you
but
see
here-
I
probably
a
layer
above
maybe
trying
to
get
a
bigger
picture
and
trying
to
how
we
promote
this,
the
observability
in
different
areas
and
for
different
purposes
as
well.
That's
not
only
that
people
don't
only
think
about
yeah
certain
use
cases
that
are
associated
with
observability
right
now,.
D
That's
great,
actually,
that's
I'm
putting
a
link
in
the
chat
to
the
charter
that
we
recently
settled
on
a
few
months
ago,
and
and
scenarios
like
that
are
specifically
in
scope
for
this
sig.
You
know
how
people
are
using
observability
tools
and
data
methods
on
on
how
to
to
leverage
them
and
how
to
make
them
actionable
or
useful
for
various
domains
and
verticals.
D
That's
all
that's
all
in
scope,
and
so
we
have
a
lot
of
latitude
as
a
sig
about
how
we
spend
time
and
and
what
we
might
want
to
either
welcome
members
to
have
talks
or
webinars
and
orchestrate
that
or
to
provide
a
forum
to
get
feedback
on
on
ideas
and
or
approaches
in
that
space.
So
welcome
to
the
sig
and-
and
we
welcome
all
of
yeah-
have
a
look
at
the
charter,
but
but
no
that
that's
that's
in
scope.