►
From YouTube: CNCF SIG Observability 2020-07-07
Description
CNCF SIG Observability 2020-07-07
A
A
A
A
Is
the
due
diligence
of
sorry
no
there's
one
more
on
the
POC
voice
concerns
about
the
diversity
level
of
the
chairs,
so
they
actually
asked
us
to
to
look
for
four
different
nominations
which
obviously
it's
bad
for
Steve
we
reached
out
for
both
COC
and
also
behind
the
scenes.
But
anyone
who
wants
to
speak
up.
A
A
A
B
Nice
I
think
one
more
thing
I
got
from
from
the
call
we
had
literally
one
hour
ago
was
that
since
EF
is,
is
really
encouraging,
seeks
to
use
the
CNC
F
block
space.
So
we
are
welcome
to
produce
any
content.
We
want
relevant
to
our
our
topic
and
this
will
be
published
in
the
CNC,
a
blog
SS
CNC,
a
blog
post,
so
I
think
recently
seek
storage
or
assembly
that
expose
something
on
the
white
paper.
So.
C
What
do
you
think
as
an
idea
for
a
white
as
an
idea
for
a
white
paper
or
a
rather
a
blog,
to
identify
opportunities
for
new
contributors
to
come,
join
our
sig
and
do
stuff
I
think
we
have
a
lot
of
like
potential
future
ideas
identified
and
yeah.
In
some
cases
we
already
have
github
issues
that
are
marked
with
you
know
the
good
first
time
issue.
I
bet.
A
B
B
C
There
was
a
blog
that
went
up
from
I.
Think
that
see
a
little
while
ago,
a
couple
weeks
ago.
That
caught
me
a
little
bit
by
surprise,
so
I
didn't
actually
get
a
chance
to
respond
before
they
printed
it,
but
that
was
sort
of
from
the
outside.
In
perhaps
now
it's
time
to
do
a
blog
from
us
to
the
broader
community.
A
A
That's
the
inherent
problem
we
have
is
most
people
who
are
here
are
here
because
they
already
actively
watch
what
is
happening
within
CN
CF,
whereas
the
blog
might
be
a
chance
to
get
coverage
on
the
outside
of
CN
CF
as
well.
I
suspect,
especially
end
users
in
such,
like
most
of
them
on
all
of
them.
As
we
see
in
this
call,
but
most
of
the
end
users
will
not
actively
consume
anything
which
which,
which
is
happening
within
CN
CF,
so
we
can
spread.
This
more
is.
C
D
B
A
A
A
B
C
C
A
A
E
I
can
explain
a
bit
more.
What
I
would
like
to
see
is,
and
the
problem
we're
facing
right
now
is.
There
is
no
way
for
us
to
relates
all
this
thing
together.
Like
you
mentioned,
there
is
no
metadata
for
it.
Look.
He
didn't
exists
until
very
recently,
and
we
can't
offer
the
users
even
a
single
way
to
look
at
metrics
logs.
Traces
are
reporting
in
a
consistent
way
like
there
is
no
single
dashboard
for
all
of
this
things,
so
it
goes
further
than
just
anomaly.
Detection.
A
A
A
B
A
A
A
A
A
As
per
last
week
was
to
just
collect
use
cases,
get
quick
consensus
on
all
use
cases,
are
they
valid?
Are
they
not
on
it,
and
we?
We
have
quite
some
of
them
to
then
in
the
next
iteration,
without
needing
to
be
interactive
and
in
real
time
basically
comment
on
on
what
that
is,
and
that
means
obviously
priority.
The
priority
station.
A
So
basically
anything
here
we
just
have-
is
this
valid
to
be
agreed?
Do
we
need
to
be
more
specific,
less
specific
other
things
like,
for
example,
the
MTD
our
comment
or
the
multi-tenancy
comment?
This
is
exactly
the
kind
of
honing
the
statements
which
which
we
are
looking
for
and
then
just
accept
it
and
work
based
on
this.
A
What
does
this
actually
mean
like,
for
example,
to
have
a
specific,
to
have
a
specific
example
with
in
with
in
prometheus
multi-tenancy
is
deliberately
ignored,
but
then
it
makes
sense
for
with
my
premises
head
on
to
get
a
challenge
from
CN
CF
or
the
wider
community
or
whatever.
What
are
you
doing
about
multi-tenancy
and
with
my
premise,
you
said
on
the
answers.
We
deliberately
ignored
us
for
very
good
reasons,
but
then
at
least
we
we
make
this
more
explicit
and
I
know
with
my
Prometheus
I
found.
A
We
have
this
in
our
documentation
as
to
why
we
do
with
this
bla
bla
bla
bla,
but
then
it
becomes
a
checkbox
and
either
we
take
it
or
we
say.
Well,
we
don't
take
it
for
a
reason
X,
which,
with
my
sake,
hair
on
thick
hat
on,
is
I,
think
absolutely
valid.
That
projects
deliberately
ignore
use
cases,
but
then
we
can
say:
ok,
this
project
doesn't
follow
the
use
case.
A
A
B
A
This
is
what
I
just
put
in,
because
this
is
also
something
which
will
start
to
bite
more
and
more
data
sovereignty.
Laws
like
this
data
must
not
leave
a
certain
jurisdictions.
Singapore
China,
European
Union
come
to
mind.
This
might
be
soft,
with
just
deploying
in
a
different
zone
or
something
but
yeah
I.
A
C
A
A
A
A
C
Mean
this
is
Laurie
now
at
basic
Prometheus
queries,
but
if
I
assume
and
if
I
infer
that
they
really
mean
on
large,
that's
a
historical
data.
Like
you
know,
over
the
last
year
or
two
years
like
that,
that
gets
to
be
more
at
least
into
this
OLAP
or
analytics
these
case
versus
just
a
straight
forward
to
the
computer
rate
using
metrics.
B
E
A
H
One
thing
that
may
be
different
here
is
that
you
might
be
aggregating
over
dimensions
other
than
theories
ideal,
like
prom
QL,
you
can.
The
aggregate
of
a
series
ID
which
matters
for
aggregates
like
median.
Oh,
we
are
react.
Rotating
makes
a
big
difference.
So
one
of
the
thing
is
that
I
think
is
is
getting
at
here
is
aggregating
over
different
dimensions
than
just
here
is
ID.
C
He's
saying
that
one
of
the
one
of
the
harder
things
to
do
if
I
understand,
is
to
aggregate
over
over
things
that
are
not
series.
So
there
are
some
aggregations
that
involve
revisiting
all
of
the
data
or
reformulating
it.
Or
did
you
know
like
outside
of
just
a
prom
ql
box,
where
you
can
aggregate
on
unserious,
using
if
you
were
to
either
have
other
attributes
that
you
want
to
aggregate
on.
B
H
B
H
Sorry,
what
I
mean
is
you
can
impromptu
all
you
can
get
the
medium
by
series
ID
and
then
you
could
give
the
median
by
namespace,
but
getting
the
median
by
series
ID
and
the
median
by
name
is
different
than
getting
is
medium
by
namespace
to
begin
with,
and
that
you
cannot
do
it
from
cow.
There's,
always
an
underlying
lower
level
of
aggregating
by
series.
B
C
Yeah,
like
say
you're
using
say
using
like
time
scale
or
something
like
that,
so
it's
Postgres
underneath
and
your
remote
writing
metrics
to
it.
Alright,
you
might
yeah
it's
a
good
point,
I'm
happy
to
accept
it
and
move
on
and
just
take
an
action
to
to
write
this
up
and
more
in
detail.
If
you
want
to
do
that
directly,
you
seem
to
have
the
best
get
along.
A
Also,
just
like
I'm,
obviously
always
torn
with
with
20
different
heads
about
how
much
to
interject,
when
we,
when
we
diverge
in
interesting
discussions
from
the
actual
topic
at
hand,
which
is
oh,
it's
super
hard
to
juggle
and
any
feedback
as
to
how
much
how
much
we
may
versus
the
cracking.
People
want.
A
E
Can
generalize
it
a
bit
to
trend
analysis
or
predicting
time
series,
because
it
it's
not
only
about
capacity
tying
any
form
of
metrics
you'd
like
to
predict
what
its
gonna
look
like
in
20
cents
a
year
or
in
a
month
it
doesn't
have
to
be
just
how
much
can
I
join
you
to
scale
my
cluster
in
in
a
year.
A
A
E
A
E
A
A
G
A
A
A
If
we
face
it
like
this,
is
everyone?
Okay
with
this?
Anyone
disagree
that
we
like
or
maybe
for
context.
This
is.
This
is
one
of
the
tricks
from
from
ITF,
where
you
have
the
actual
intention
of
the
document
in
covered.
It's
in
this
case
use
cases,
but
then
you
just
have
a
consideration
section,
which
basically
allows
you
to
put
non-core
stuff
into
the
same
document,
while
clearly
keeping
this
distinction
between
what
you
actually
want
to
talk
about
versus
what
is
good
to
keep
in
mind,
so
I
would
tend
towards
doing
it.
A
A
A
Kind
of
is,
which
is
like
I,
think
honestly.
Anomaly.
Detection
is
largely
out
of
scope
for
this
effort,
for
the
simple
reason
that
if
your
data
is
large
enough,
you
will
always
have
correlations,
so
you
would
mean
actually
actually
domain-specific
things
which,
which
can
look
stuff
and
optimize,
and
those
things
are
usually
human.
Of
course,
machines
depend
on
on
protect
data,
even
cutting
edge
machine
learning.
You
still
need
humans
to
to
take
data.
So
all
of
this
is
the
company
way
to
complicated
way
of
saying.
E
B
Well
true,
but
you
know
there
might
be
systems
that
allows
easy
definition
of
how
animated
action
should
be
run,
and
you
know
if
you
can
at
least
create
some
integration
that
will
pull
this
data
from
where
you
want,
and
this
will
be
defined
by
humans.
This
is
already
for
the
system
to
make
that
to
allow
that
right
to
define
the
API
that
would
satisfy
this
need.
So
this
is
still
kind
of
valid
use
case
to
me,
I.
A
E
Think
it
covers
it
because
if
you
look
at
the
second
white
points
below
there
is
something
like,
for
instance,
symptoms
versus
versions
of
software,
and
you
could
have
a
human
saying
well
for
this
version.
We
expect
this
for
this
new
version.
We
expect
that,
and
certainly
we
don't
have
this
normal
state
anymore,
that's
kind
of,
fortunately,
what
you,
what
you
wrote
in
the
end.
A
But
yeah,
sorry,
let
me
restart
the
sentence.
My
point
is
I'm
trying
to
reduce
text
as
much
as
possible
as
per
usual.
Of
course,
writing
more
text
is
easier
than
writing
less
text,
so
I
want
to
really
condense
into
into
into
atomic
use
cases,
and
then
we
can
start
exploding
from
there
again.
So
the
question
I
think
which
I
was
trying
to
ask,
is:
is
this
covering
everything
in
here
or
did
I
miss
anything.
A
E
A
A
E
A
Thank
you
very
much
very
very
much.
Let
me
check
guidelines
for
new
metrics
and
list
of
projects
document.
Okay,
so
should
we
should
we
leave
off
here
and
actually
try
to
condense
this
down
into
into
into
use
cases,
because
these
are
more
like
these
are
literally
bullet
points,
not
atomic
use
cases.
B
I
kind
of
I
think
I
answered
everything
you
can
you
can
kind
of
check
like
below,
however,
what
I
would
try
to
kind
of
talk
about
during
this
meeting
is
how
we
can
make
sure
you
you
would
be
answered
without
our
help
by
we,
but
with
some
content
that
we
provide
right.
So
my
question
would
be
for
you
are
there?
What
were
the
sources
that
you
are
trying
to
read
before
this
question?
Did
you
try
to
reach
I?
Don't
know?
B
E
B
Anyway,
from
in
my
opinion,
like
we
already
have
topic
around
working
group
for
blog
posts
or
videos
or
even
white
paper,
so
those
questions
are
are
definitely
giving
some
kind
of
direction
that
we
should
follow
and
I
don't
know.
Do
you
have
any
today
to
the
kind
of
CN
CF
seek
team?
Do
you
have
any
suggestions,
how
we
could
take
it
and
make
it
fashionable
something
from
it.
B
Yeah
that
makes
sense
from
my
set.
However,
this
is
very
like
one
of
think
that
we
could
write
and
it
will
be
just
you
know,
forgotten
after
some
time.
So
generally,
it
would
be
nice
to
have
some
kind
of
yeah
Vicki
or
some
kind
of
space
that
someone
can
yeah
some
kind
of
index
of
knowledge
right.
This
is
what
sick
absurdity
could
try
to
maintain.
B
I
would
say
to
have
like
a
starting
page
for
for
those
kind
of
observability
features
for
for
like
CN
CF
application
right
and
then
you
know
for
metrics,
you
go
there
for
logs
you
go,
there
are
like
I
would
say
something
more
discoverable
than
blog
posts,
but
the
blog
post
is
better
than
nothing.
Yes,.
E
E
B
Is
growing
so
it
would
be
nice
if
this
feedback,
you
know
we
can
be
an
entry
point
as
a
sake
observability,
and
then
we
can
forward
this
feedback
to
the
project
and
improve
their
kind
of
documentation.
However,
we
should
be
kind
of
a
starting
point,
which
means
we
should
first
even
forward
those
people
to
those
projects
documentation,
but
that's
yeah.
That's
very
good
point,
because
you,
like
your
questions
like
they.
The
answer
for
those
should
definitely
be
part
of
the
prompt
you
stock
right.
E
So,
the
time
when
we
have
questions
around
this,
where
I'm
working
it,
we
always
refer
to
different
huge
documentation
when
considered
metrics,
because
a
lot
of
those
questions
are
actually
answered
already
issues
they
suffer
either
from
the
discoverability
problem
or
that
people
just
don't
immediately.
Think
about
ok,
how
can
I
figure
this
out
on
my
own
I.
A
Think,
within
the
six
phase,
having
and
I
think,
particle
was
looking
for
BCPs
best
growing
practices.
There's
this
knowledge
base
of
what
is
currently
the
thing
to
do
or
to
suggest
thing,
but
I'm
not
sure
if
it
makes
sense
for
for
Prometheus
and
others
to
basically
refer
back
to
a
shared
documentation.
It
has
its
pros
and
cons.
A
E
B
Well,
I
would
say
entry
point
would
be,
would
it
be
next
step,
I
think
art
or
goods
we
can
help
art
or
like
in
the
meantime,
in
this
issue,
but
entry
point
with
maybe
what
what
art
or
asked
about
it's
a
good
starting
point
port
for
entry
point.
So
entry
point
would
be
some
index
page
whatever
in
our
red
book.
Maybe
it's
a
good
idea.
B
B
This
landscape
map,
if
we
can
have
some
kind
of
filter
based
on
the
seek
relation
and
looks
like
Dan
kind
of
pointed
me,
that
there
is
observability
and
analysis
kind
of
topic
so
can
filter
by
that
and
I
kind
of
I
think
we
changed
our
repo
event
to
to
give
this
link
and
might
be
good
enough.
That
was
the
latest
kind
of
news
on
this
topic.
B
B
E
Missing
something
in
its
where,
basically,
it
says,
monitoring,
logging
tracing
and
then
chaos
engineering,
but
maybe
monitoring
is
a
big
category
right
now
as
and
you
can
split
that
further
into
metrics
tracing
I'm,
pretty
sure
there
are
things
that
are
makes
of
those
as
well.
So
it's
not
exactly
black
and
white,
either
I'm
on
those
categories,
either
those
I'm
thinking
about
a
vector,
for
instance,
which
can
do
metrics
and
logging.
At
the
same
time,
correct.