►
From YouTube: CNCF SIG Observability 2020-05-12
Description
CNCF SIG Observability 2020-05-12
B
D
A
E
E
So
I
think
we
should
get
started
and
by
and
large
establish
a
precedent
of
starting
st,
not
CT,
to
be
mindful
of
everyone's
time
and
not
just
sit
here
for
like
five
minutes,
so
the
territory
apart
first,
the
first
as
you
can
see,
I
did
my
thing
in
just
right
and
stuff
which
we
talked
about.
This
bullet
points
you're
also
more
than
welcome
to
start
doing
this
if
you
want,
but
still
we
are
following
the
discussion,
as
is,
but
that
saves
typing
effort
later.
E
Cool
so
POC
status.
I
wasn't
a
few
seek
hold
last
week
this
week
last
week,
so
I
poked
them
again
to
please
vote
for
the
future
and
please
vote
for
detective
course.
We
had
zero
replaced
there
and
they
promised
to
do
so,
but
I
haven't
seen
a
little
check
today,
at
least
on
the
question
of
the
user
survey.
Basically,
it
was
not
fully
specified,
but
basically
here
we
can
do
you
service,
as
always,
if
we
want
to
if
we
they're
useful
and.
E
E
And
also,
as
we
were,
having
our
monthly
called
with
CN
CF
with
my
community
set
on.
Yes,
they
also
help
Chris
about
taking
this
to
do
back
to
TOC
on
the
internet.
Using
calls
to
please
get
on
third
chair
and
active
course,
as
per
the
official
documentation.
I
seen
see
if
we're
actually
not
a
real
working
group.
Yet,
of
course
we
don't
have
the
requirements
for
it,
but
that's.
E
E
Okay
from
the
people,
I
can
see
my
video
it's
four,
it's
five.
Okay,
so
that's
good.
Hopefully
it
will
become
more
important.
As
a
reminder.
We
are
trying
to
establish
a
practice
where
people
read
the
documentation
and
working
documents
and
such
before
the
call.
So
we
don't
have
a
case
of
existed
reading,
but.
E
E
F
G
E
E
E
E
H
E
E
E
Perfect,
so
the
next
is
about
about
having
a
code
of
conduct
and
I
would
propose
that.
We
also
put
the
comment
on
every
zero
point
and
my
proposals,
obviously
that
we
agree
that
there
is
a
code
of
conduct
course,
it's
literally
the
same
as
the
one
for
four
seeds.
I
mean
everyone
agree,
any
any
naysayers.
E
G
F
So
we
only
list
projects
in
like
companies
and
adopters
or
MD
if
they
are
using
it
in
production.
Further.
We
are
working
on
adding
case
studies,
so
we
did
one
case
study
with
go
check
and
this
or
like
yes,
this
week,
I
have
my
second
case
study
scheduled
with
the
reveille,
which
is
a
German
grocery
chain
and
slowly
we
are
going
to
expand
the
case
study
section.
E
C
C
I
C
B
One
more
comment
on
this
and
I:
don't
want
to
do
rail,
so
we
don't
have
to
discuss
it
but
I'm
curious.
What
is
the
process
for
listing
adopters
like
do?
The
companies
agree
should
be
they
be
the
ones
that
submit
this
PR
Dewey's
get
their
approval,
then
can
add
it
to
the
adopters
like?
What
is
that
process?
Look
like
so.
F
B
I
E
Think
for
the
for,
for
for
the
sake
by
default,
we
trust
unless
there
is
a
reason
to
mistrust
Chris
else.
We
have
issues
anyway,
but
it's
a
good
point,
I
think
that's
actually
within
within
the
policy
of
the
project,
how
they,
how
they
get
to
list
people.
The
requirement
is
just
that
they
that
they
confirm
that
it's
happening,
and
ideally
this
happens
with
with
names
and
not
just
with.
F
E
C
F
E
So
I
read
your
answers
to
this
and
basically
for
anyone's
benefit.
The
top-level
bullet
points
in
this
section
are
copied
from
from
CN
CF
documentation.
I
would
agree
that
all
of
those
are
met,
except
for
the
last
one,
but
that
is
basically
a
to-do
item
for
for
TOC,
but
in
my
opinion,
cortex
operates
cloud
native.
So
the
gist
of
the
question
is
also
positive,
or
the
answer
of
the
gist
of
the
question
so
same
for
this
I
would
propose
that
we
that
we
have
consensus
on
SiC
observability
agrees.
E
I
I
A
E
I
E
E
Incubate
I
would
say
that
both
my
knowledge
of
the
project
and
what
we
are
seeing
documented
here
is,
is
actually
positive.
I
left
some
comments
to
to
maybe
tighten
up
that
section
a
little
bit,
but
I
would
still
be
also
suggest.
We
agree
with
the
section,
as
is
with
the
request
for
for
cortex
two
and.
F
One
more
point
is
all
the
adopters
dot
MD
users-
they
are
all
running
in
kubernetes.
They
are
monitoring
their
Cuban
et's
deployments
with
cortex.
We
having
said
that,
we
are
also
working
with
a
new
adopter
who
is
running
it
on
bare
metal.
So
it's
not
tied
into
communities,
but
it
works
really
well
with
kubernetes
and.
E
C
Yeah
I
mean
having
done
a
fairly
deep
dive
through
the
architecture.
I
would
I
would
certainly
concur
that
it's
obvious,
obviously
architected
in
a
cloud
native
style.
That's
that's!
That's
clear.
I
do
agree,
though,
that
the
useful
feedback
for
the
TLC's
is
that
the
useful
moniker
is
not
terribly
useful
in
this.
You
know
in
terms
of
in
terms
of
due
diligence,
so
I
mean
it
almost
seems
like
there's
a
request,
implicit
in
the
question
to
have
a
little
market,
marketing
or
selling
of
this
solution.
C
I,
don't
know
if
I'm,
the
only
one
that
reads
that
you
know
is
this
useful?
Why
would
you
care
you
know
with
you?
You
know
if
you're
running
kubernetes
or
if
you're
running
a
cloud
native
data
center,
a
data
center,
architected
and
cloud
native
sort
of
styles
or
Oracle
eyeing
time
when
somebody
else's?
C
E
Tell
me
it
reads
as
if
most
of
this
document
and
the
questions
have
been
written
from
the
perspective
of
hey,
that
might
be
a
good
question,
but
not
from
the
perspective
of
if
I'm,
in
the
position
to
to
actually
answer
those
questions.
Are
those
well-sculpted,
useful
questions
and
again
useful
is
subjective
but
I
think
they're
just
a
little
bit.
I've
read
every
Power
issue,
not
really
battle
tested
yet,
but
we
can
feed
this
back
and
we
absolutely
should.
E
E
E
I
I
I
think
like
assuming
that's
what
it's
assuming
that's
what
they're
getting
up
getting
at.
You
know,
we've
done
Prometheus
keynotes
and
the
you
know
talking
to
press
about
prometheus
launches.
Like
you
know,
whatever
the
CN
CF
needs
them
as
I.
Think
we've
got
the
experience
doing.
C
Even
reinforce
that
I
mean
if,
if
you
just
throw
a
rock
on
the
internet
and
look
at
the
number
of
talks,
meetups
and
other
various
things,
just
on
YouTube
alone,
it's
clear
that
the
project
does
not
only
understood
the
spirit
of
that,
but
has
been
demonstrating
it
for
a
couple
of
years.
Now,
at
least
in
my
judgment.
Thank
you.
Matt
I
realized
that
dangerously
close
to
sunshine,
a
no
I
mean
just
even
on
the.
If
one
wanted
to
get
started
and
dive
into
what
is
cortex
and
how
is
it?
C
D
E
D
E
E
C
Healthy
numbers,
just
just
to
be
clear,
just
so
I
think
Michael
I
made
the
point
separately,
the
three
independent
and
users
and
in
terms
of
using
it
does
that
mean
since
we've
and
Griffin
a
cloud
are
kind
of
obviously
driving
the
project
that
they
don't
count
like
in
terms
of
those
three
I
think
there's
still
plenty
that
are,
but
should
we
just
call
out
that,
like
if
you're,
if
you're
paid
to
work
on
cortex
you're,
not
an
independent
end-user?
That's
all
yep.
C
G
Maybe
isn't
exactly
we
had
the
very
same
question
about
a
project
we've
reviewed
for
at
delivery
and
had
a
discussion
with
TOC
and
we're
still
in
like
the
final
wording,
but
the
interpretation
that
we
are
kind
of
stuck
at,
which
is
not
the
family.
What
is
obviously
an
end
user?
It's
like
it
actually
uses
us
at
the
start,
like
a
provider,
because
we
had
expected
that
very
discussion.
So
you
would
not
consider
like
we've
for
we've
come,
but
they
are
selling
it
to.
G
Customers
is
an
end
user
because
it's
really
somebody
who's
like
not
building
this
after
it,
but
when
you're
using
it
and
also
not.
You
would
also
consider
like
a
main
company
like
if
you
build
software
for
yourself,
but
the
CNCs
is
having
to
have
a
clear,
but
it's
really
something
not
sell
himself,
but
but
using
the
software
would.
G
C
In
particular,
for
something
like
cortex,
which
is
specifically
meant
to
be
multi-talented
and
something
that
someone
could
use
to
provide
as
a
service
to
others,
in
addition
to
someone
like
in
our
in
our
company,
where
we're
using
it
directly
as
an
end
user,
yeah,
again,
I,
don't
I,
don't
want
to
hold
this
up
and
I
think
the
requirement
is
clearly
met.
I
just
wanted
to
make
sure
that
we're
not
double
counting,
Griffin
and
weave
in
the
list
of
adopters,
as
as
counts
towards
end
users
for.
C
E
F
F
C
So
yeah
that's
very
mean
I
can
actually
say
like,
for
example,
right
now
today,
I'm,
actually,
you
know
in
addition
to
rolling
out
cortex
internally
in
you
know,
in
the
coming
in
this
quarter,
we're
also
a
end
user
by
that
definition
of
crow
fauna
clouds
hosted
Prometheus,
but
I,
don't
know
if
it's
actually
fair
on
the
merits
to
say
that
when
I'm
using
that
I'm
an
end
user
of
the
cortex
project,
you
know
to
me
it
looks
like
a
remote
right.
Endpoint
and
I
have
no
idea.
What's
behind
it,
I
don't
I
could
know.
C
I
can
know
yeah
I,
don't
even
need
to
know
that
it
is
cortex.
It's
just
a
previous
remote
and
a
query
endpoint.
So
I
agree.
This
is
really
more
of
a
yeah
another,
slightly
ambiguous
language
from
GNC
up.
On
the
other
hand,
it's
it's
stimulating
this
discussion,
which
is
probably
the
useful
part,
so
yeah
I'm
happy
to
move
on
I,
just
I.
E
C
That's
a
great
point
too,
but
I've
just
been
in
the
context
of
the
due
diligence
for
cortex
I.
Don't
think
we
need
to
sterilize
it
like
you
know,
I,
don't
want
to
wait
for
the
TOC
to
define
it.
If
we
again,
we
have
way
more
than
the
requirements
to
exit
sandbox
for
the
actual
projects.
I
see
them
as
well
slightly
orthogonal.
That
makes
sense.
Yeah.
E
H
E
E
H
E
E
F
E
E
Next
one
documentation
of
CCF
alignment
Bartok
suggested
to
just
copy
in
the
incubation
stuff.
Basically,
yeah
I
would
agree
with
this
as
long
as
there's
also
a
statement
by
my
cortex
that
all
of
this
still
applies,
I
would
be
content
if
anyone
from
the
cortex
project
was
just
very
confirm,
override.
E
E
E
E
H
E
F
C
Not
in
the
next
six
months,
I
think
for
the
first,
the
first
piece
of
this,
the
primary
target
called
native
use
cases.
You
could
actually
add
a
fourth
one
that
you
know
the
most
of
the
backend
services
for
both
object:
storage,
as
well
as
for
trunk
as
well
as
index.
You
know
work
on
major
public
cloud
backends
today,
like
there's,
not
a
need
to
run
your
own,
you
know
object,
storage
or
Cassandra,
or
something
like
that.
You
can
use
BigTable
or
that
I'm
going
to
be
or
GCS
or
s3.
E
As
everyone
is
probably
noticing
I'm
taking
Liberty
to
to
just
write
this
in
an
update
floor
for
the
court
exchange,
I
didn't
yes!
Thank
you,
okay,
that's
why
ideally
I
want
to
like
we're
like
halfway
through
this
document
and
I
would
actually
prove
it
or
disprove
it
before
before
this
Collins
so
or
agreed.
E
Would
also
tend
to
agree
with
this
one
all
agreed
yep
good.
What
exactly
are
the
failure
modes?
Are
they
well
understood?
Have
they
been
tested?
Did
it
format
the
part
of
continuous
integration
testing
already
appropriate,
given
the
usage
ieg
plus
the
wide
shared
services
need
to
fail,
gracefully
EPC.
I
E
I
C
Yeah
I
mean
a
bunch
of
engineers.
We
can
sit
around
and
say
how
could
we
make
this
better
like,
for
example,
you
know
one
could
one
could
write?
One
could
have
a
CI
or
CD
testing
negative
failure,
cases
that
simulates
or
mocks
out
s3
barfing
we're
having
operational
issues
from
the
cloud
provider
side
with
these
I'm
not
I'm,
not
sure
how
deep
we
need
to
go.
I've
read
what.
E
You
know
how
it
is
basically
our
V
with
our
sig
heads
on
happy
with:
what's
the
current
state
of
the
art,
to
let
someone
progress
from
send
books
to
incubating
that's
the
exact
scope
of
what
we're
currently
debating
and
within
that
scope,
I
think
it's
it's
battle
achieved
I
think
that
they
can
more
and
I
think
they
should
be
doing
more
before
graduating.
But
for
incubation,
I.
H
I
H
I
I
mean
the
the
the
simple
one-dimensional
trade-offs
are:
are
never
entirely
accurate.
You
know
the
consistency
versus
availability.
One
is
not.
You
know
there
are
cases
when
we
prefer
the
other
like
a
each.
One
of
these
is
tuned
to
the
use
case
and
configurable
like
you,
can
have
it
such
that
I
guess.
If
we
wanted
to
have
a
trade-off
generalization
here,
it
would
be
configurability
like
we,
we
favor
being
able
to
configure
the
system
to
behave,
how
the
user
wants
and
leaving
that
to
the
user
to
decide.
I
But
we
don't
I,
don't
think
that
necessarily
implement.
Sorry
introduces
complexity
anymore,
because
we
ship
default
configurations
and
we
make
recommendations
right.
So
you
don't
have
to
understand
these
trade-offs.
If
you
don't
want
to,
and
we've
helped
many
users
deploy
it
and
not
had
to
explain
every
single
configurable.
H
H
E
I'm
going
to
do
something
very
cherry.
That
section
is
currently
not
been
answered.
There
should
be
something
in
there,
but
I
think
we
can
expect
a
reasonable
answer.
So
I
would,
as
a
call
for
consensus,
maybe
say
that
we
expect
positive
feedback,
but
it's
what
we
expect.
We
expect
sufficient
feedback.
I
I
A
E
E
I
C
Think
that
there
wasn't
how
I
read
the
spirit
of
this
question
I
think
it's
sort
of
like
when,
when
the
project
is
leaving
sandbox
to
go
to
the
next
phase
like
say
there
was
no
a
chase
story
or
say
like
hey
this
works,
but
there's
no
way
to
deploy
it.
Unless
you
build
it
from
source
to
know
how
it
works.
Good.
C
Gonna
say
like
has
someone
looking
at
cortex,
like
you
know,
we're
taking
a
lot
of
time
to
go
figure
it
out.
It's
not
like
immediately
but
again.
That's
that
I
wouldn't
call
that
a
hole
in
the
architecture
or
a
hole
in
the
the
offering
I
don't
see
those
types
of
hole
prevented
from
leaving
sandbox
I
think
oh,
is
to
make
sure
that
we
don't
have
things
leave
sandbox
that
are
like
fundamentally
not
done
or
not
ready
to
be.
You
know
in
that,
in
that
middle
tier,
but.
E
C
E
E
E
F
E
E
So
call
for
consensus,
seek
observability
is
happy
with
the
documentation,
provided.
Everyone
agree,
yeah
good,
what's
really
small
immersion,
shame
every
depends
of
stability,
blah
blah
blah
again.
I
would
suggest
yes
for
the
consensus
and
I
would
move
partakes
comment
inline
to
make
this
part
of
this
stuff.
E
I
You
can
run
the
unit
and
integration
tests
yourself
like
wherever
you
want
to
run
them.
So
that's
that's
pretty
straightforward
d.
There
is
provided
what
we
call
the
test
exporter,
which
is
something
that
exports
a
sine
wave
and
then
queries
for
it,
and
you
can
run
that
in
you
know
that
would
be
the
integration
test.
You
would
run
in
your
own
environment
to
check
the
things
working
and
beyond
that.
I
That's
about
the
limit
of
what
we've
got
and
if
you
wanted
to
extend
it,
my
encouragement
would
be
to
contribute
back
to
masteren
and
have
that
benefit
everyone,
but
but
there's
no
reason
why
you
can't
build
your
own
test
on
top
of
integrate
your
integration
environment.
Does
that
answer
your
question?
Matt.
C
Sort
of
and
again
I'm
not
sure
that
this
is
maybe
we
take
this
up
in
a
subsequent
chat
but
like,
in
other
words,
if
I
wanted
to
reproduce
all
of
the
tests
or
to
debug
something
as
a
user.
You
know
again,
maybe
I
need
to
educate
myself
a
little
more
of
it.
You
know,
can
I
run
the
full
Suites
of
tests.
If
I
have
my
own
I
do.
If
I
have
my
own
circle
account
like?
Is
there
a
oh.
E
So
call
for
consensus.
Sic
observability
is
happy
with
documentation,
provided
all
agree,
yep
good.
What
licensing
restrictions
apply,
blah
blah
blah
blah
blah
Apache
yeah
yeah
can
confirm
by
definition
we're
happy
Carson
City
for
licence
of
all
things.
Yes,
but
just
to
make
it
explicit
call
for
consensus.
The
observability
semi.
Aren't
we
yep?
I
E
E
I
C
H
Just
to
be
sure
like
this
is
not
like:
there
is
no
rules
how
to
create
this
recommendation
from
the
seek.
So
we
just
grabbed
the
due
diligence
dog
and
just
you
know,
had
some
checklist,
but
I,
don't
think
like
that
was
required.
So
I
could
see
that
those
six
we're
just
you
know
composing
their
own
decision
recommendation
based
on
something,
and
then
that
was
that's
what
it
and
the
rest
was
done
by
the
TOC.
We
went
kind
of
yeah.