►
From YouTube: Kubernetes SIG Instrumentation 20191017
Description
Meeting notes: https://docs.google.com/document/d/17emKiwJeqfrCsv0NZ2FtyDbenXGtTNCsDEiLbPa7x7Y/
A
B
A
B
Of
in
progress,
as
far
as
like
basically
I,
don't
think
that
we
made
any
progress
on
the
metric
stability
of
stuff.
This
release,
unfortunately,
there's
a
very
small
enhancement
window
and
with
cube
con
coming
up
and
various
leaves,
and
that
kind
of
thing
both
Han
and
I
did
not
have
a
chance
to
work
on
that.
B
I
had
looked
last
released
and
to
try
to
throw
in
like
sort
of
a
feature
flag
for
enabling
the
hidden
metrics,
and
we
decided
to
not
go
forward
with
that,
because
we,
according
to
the
speck
in
the
cap
anyways,
we
need
to
be
able
to
turn
them
on
and
off
by
a
specific
like
metric
name.
That
was
going
to
take
more
implementation.
That
could
have
happened
at
this
release.
It
did
not
so.
A
B
A
B
Mean
it's
possible
that
we
could
walk
that
back,
which
would
maybe
make
our
lives
a
little
easier,
and
do
it
just
like
all
hidden
metrics
are
no
hidden
metrics
per
binary
like
that,
would
enable
the
feature
flagging.
A
pro
I
could
pick
up
like
with
my
last
PR,
but
maybe
if
we,
if
we
want
to
do
that
or
if
we
want
to
propose
that
we
should
probably
look
into
updating
the
cap.
A
C
Merrick
my
lecture
coverage
set
out
years,
I
think
is
working
with
Hannon
days
as
well.
I
think
he
is
going
to
take
over
this
work,
but
he's
unfortunately
not
attending
today
as
well,
but
in
general,
like
my
understanding,
is
that
there
is
cap
and
he's
going
to
basically
implement
whatever
is
in
this
cap.
Yeah.
A
A
B
Basically,
we
need
to
add
a
flag
whether
or
not
it's
like
on
a
per
metric
basis
or
just
show
all
the
hidden,
metrics
or
not.
We've
got
to
add
those
flags
to
then
at
a
Kuebler.
The
API
server
I
think
you
prop
like
all
of
the
various
binaries
basically,
and
they
need
to
be
wired
up
and
every
single
one
of
them
it
turns
out.
The
code
is
subtly
different
in
special
and
exciting
ways,
and
so
I
was
kind
of
working
through
that
in
the
work-in-progress
PR
I
threw
up.
B
I
was
able
to
do
so
in
the
cubelet
and
had
started
work
on
the
API
server
and
a
few
others,
but
did
not
get
that
through
to
completion.
So
if,
if
your
colleague
wants
to
like
sir
a
theist
wants
to
sync
with
me
and
I,
can
give
some
sort
of
an
update
on
that,
there's
also
been
a
fellow
who's,
been
pinging
me
a
bunch
on
the
kubernetes
slack
Rainbow
mango,
but
I.
Don't
he's
not
on
this
meeting,
so
I'm
not
sure
he
seems
to
be
interested
in
helping
out
with
this.
Well.
A
I
talked
to
him
a
couple
of
times
on.
Unfortunately,
the
meeting
time
is
not
good
for
him,
because
it's
like
the
middle
of
the
night
for
him
I
believe
he's
in
China
which
actually
like
don't
need
to
discuss
this
right
now,
but
maybe
we
can
have
an
alternating
meeting.
I
think
some
some
SIG's
do
that,
but
yeah
that's
not
go
off-topic.
Yeah
he's
definitely
interested
in
things.
I
feel
like
if
we
can
come
up
with
a
way
where
we
can
first
introduce
a
flag
for
all
or
nothing
and
then
extend
that
mechanism.
A
B
The
plumbing
is
slightly
different,
depending
on
whether
you
do
it
with
a
feature
fly
versus
a
CLI
flag
and
the
reason
that
I
was
looking
into
doing
with
a
feature
flag
initially
was
that
you
can
remove
feature
flags
from
the
binaries,
but
once
you
put
in
a
CLI
flight
I
think
you
can
never
get
rid
of
it
kind
of
thing.
So
that
was
why
I
was
looking
at
that
approach,
and
the
implementation
is
a
little
bit
different.
Okay,.
A
D
A
C
This
is
something
that
I
wanted
to
discuss.
Last
time
like
as
I
was
aware
that
there
is
a
deadline
for
caps
and
and
and
and
placement
features.
Unfortunately,
we
didn't
talk
too
much
last
time.
However,
it
seems
that
we
are
good.
Yes,
I
ensured
that
you
know
listen
context
of
this
matrix
stability
effort.
We
have
proper
caps
and
we
can
move
on
with
the
implementation.
I
think
this
is
the
case.
C
I
think
you
should
also
start
thinking
a
little
bit
about
like
long
term
of
the
roadmap,
like
maybe
something
2020,
and
you
know,
probably
like
cube
con-
will
be
a
good
opportunity
like
contributors
Sammy
during
cube.
Con
will
be
a
good
opportunity
to
discuss
this
like
to
have
some
long
term
plans
so
that
you
know
we
can
utilize
people
better,
because
I
think
there
is
plenty
of
people
interested
in
contributing
in
to
sig
instrumentation.
At
the
same
time,
we
don't
have
a
good
backlog
of
the
things
to
do,
and
it's
hard
to
like.
C
If
you
are
new
or
you
are,
you
know
not
very
hands-on.
It
is
area
like
it's
it's
hard
to,
like
you
know,
start
from
like
a
high-level
idea,
so
does
make
sense
to
have
a
kind
of
answer
like
what
is
the
status
of
the
agenda
for
contributors
summit,
but
maybe
like
we
can
have
a
kind
of
at
least
informal
discussion
around
this
area.
Yeah.
B
I
I
still
have
all
the
meeting
notes
and
then
everything
that
needed
to
get
followed
up
on
from
like
coupe
con
North
America
last
year,
I
didn't
go
to
Barcelona,
so
all
I
know
for
this
year.
I
will
be
attending
the
computer
summit,
so
I'll
be
on-site
for
that.
I
know
that
there's
a
cig,
instrumentation
meet-and-greet
or
something
that's
happening
in
the
afternoon
I
volunteered.
For
that.
So
you
know
I
guess
I'll
get
to
go,
meet
all
the
excited
new
contributors
and
have
nothing
to
assign
them
to
in
terms
of
like
I.
B
Think,
there's
also
supposed
to
be
a
couple
of
like
cig
instrumentation
sessions
that
are
on
the
agenda
like
there's
an
intro
level,
one
and
intermediate
level
one
or
something
like
that:
Archy,
our
esteemed
chairs.
Are
you
attending
and
running
those
like
what's
going
on
with
those,
because
one
of
them
I
know
is
scheduled
during
my
talk.
A
B
A
So
intro,
probably
I,
haven't
talked
to
her
about
this
just
yet,
but
I'll
probably
do
the
same
one.
That
I
did
last
time,
which
is
basically
what
have
we
worked
on
so
far?
How
can
you
get
involved
just
like
high-level
things,
because
it's
really
meant
for
people
who
may
have
never
contributed
to
kubernetes
may
be
completely
new
tube
coop
con
communities
so
that
that's
supposed
to
be
like
super
super
low
level
like
low
entry
and
then
the
deep
dive
is
going
to
be
on
describing
the
metric
stability
framework.
B
A
Would
make
sense
yeah
specifically,
because
probably
we
should
give
be
giving
each
other
a
shout-out
in
both
talks
and
if
they're
happening
at
the
same
time.
That's
unfortunate
for
the
attendees
yeah
that'd
be
great.
If
you
can
do
that.
Otherwise,
I
can
also
try
to
reschedule
the
intro,
maybe
I
believe
Tariq
you're
on
the
you're
on
the
call
I
think
you
said,
set
up
some
things
in
terms
of
the
contributor
summit.
Is
that
right.
A
D
Cool
I'll
try
to
speak
as
loud
as
possible,
see
a
your
question
is
around
the
contributor
summit
events
I
shouldn't
cry:
oh
so
yeah
I
did
post
honestly
instrumentations
luck,
I!
Think
couple
months
back
asking
for
volunteers
for
the
same
meet-and-greet,
see
ya
and
I.
Think
as
Anna
said,
she
was
one
of
the
volunteers.
I,
don't
remember!
Who
was
the
other
one
that
board
kid.
D
A
A
Yeah
I
don't
know
in
case
it's
not,
let's
not,
let's
not
occupy
multiple
rooms
and
that
are
short
and
short,
apparently
anyways
already
so
yeah.
If
you
could,
let
us
know
that'd
be
fantastic,
and
if
that's
not
the
kind
of
format
that
that
we
would
need
for
discussing
amongst
us,
then
I
would
try
to
get
a
room
through
Chris
or
someone
else
on
the
CNC
F
I
already
have
another
threat
going
for
a
room
anyways,
so
I
could
do
that,
but
yeah.
If
you
could
figure
out
the
format,
that'd
be
fantastic.
D
A
A
A
A
A
A
That
could
have
allowed
us
to
detect
this
earlier
and,
as
a
matter
of
fact,
I
started
looking
around
and
we
don't
even
have
events
or
anything
telling
us
about
human
kills.
So
I
was
wondering
if
people
like,
maybe
maybe,
if
people
have
other
experiences
or
other
methods
of
detecting
this.
The
best
thing
that
I've
come
up
with
is
using
some
of
the
coop
state
metrics
metrics
for
this.
A
A
No
comment
so
one
thing
that
I
thought
about
is
at
least
on
a
per
node
level.
The
couplet
would
be
able
to
to
detect
these
kinds
of
things.
That
could
be
a
possibility,
but
it'd
be
actually
really
fantastic
if
we
could
even
surface
this
up
to
users
and
let
people
know
on
a
per
namespace
basis
or
something
like
that-
an
aggregate
of
codes
that
have
happened,
but
for
that
we
would
actually
need
an
event
or
something
like
that
yeah.