►
From YouTube: SIG Instrumentation 20210304
Description
SIG Instrumentation Bi-weekly meeting March 4th 2021
A
All
right
welcome
everyone.
This
is
sick.
Instrumentation
today
is
march
4th
2021,
and
we
have
a
couple
of
things
on
the
agenda
today.
First
thing
is
a
reminder:
code
freeze
is
march
9th.
That
is
five
days
from
now,
so
make
sure.
A
Right-
and
I
don't
know
about
the
u.s
but
monday-
is
a
public
holiday
in
germany,
so
part
of
the
world
part
of
the
christian
world
is
off
next
monday,
so
you're
you'll
need
to
make
sure
to
probably
get
things
in
before
the
weekend.
Is
I
think,
what's
what's
the
holiday?
B
Well,
well,
frederick
is
checking
what
holiday
it
is
on
monday,
which
we
don't
have
in
the
u.s.
I
will
maybe
talk
about
structured
logging,
but
someone.
B
This
is
what
I
get
for,
not
starting
the
meeting
okay,
so
hopefully
you
can
see
my
screen,
so
this
is
moot
now,
so
I
have
in
concert
with
merrick,
have
been
trying
to
coordinate
all
of
the
prs
around
structured
logging
and
finally,
I
was
just
absolutely
losing
it,
and
so
I,
oh
that's
in
another
tab.
B
B
So
as
far
as
what
has
been
done,
we
actually
do
have
a
bunch
of
these
that
have
been
merged,
which
is
great,
and
we
have
a
bunch
of
these
that
have
lg
tms
that
just
need
an
approver.
So
I'll
go
pester.
Someone
about
that.
We
also
have
a
lot
of
these
that
are
currently
waiting
on
the
author
to
fix
them
and
there's
two
things
that
are
not
specifically
like
a
structured
log
change
to
like
note
here.
B
So
one
is
adding
the
tool
for
verification
to
ci
and
the
other
is
fixing
the
panic,
because
currently
the
flag
doesn't
work
so
yeah.
Apparently
we
accidentally
them
earlier
in
the
release,
and
so
now
we
need
to
fix
them.
So
that's
kind
of
the
status
of
this
and
david.
I
don't
know
if
you
still
have
node
approver
powers,
but
for
anything
that
you
do,
I
might
pester
you
to
go
and
start
approving
things.
D
B
B
So
this
is
in
progress.
What
do
we
have
going
on
for
kubernetes
logging,
documentation.
F
E
C
B
C
B
C
B
B
B
E
E
If
there
is
a
direct
so
how
you
worked
on
tagging
prs
for
seeking
instrumentation
that
relate
to
metrics,
maybe
we
could
also
do
something
like
that.
So
like
link
to
the
documentation,
if
pr
touches
metrics
yeah.
C
Okay,
yeah
the
regex
thing,
but
yeah
the
the
only.
The
only
reason
why
it
did
it
won't
work
the
same
way
because
slugs
are
everywhere
and
we
would
just
have
to
literally.
C
B
So
I
think
we
have
an
action
item
to
update
those
docs,
so
we
can
go
into
that
for
next
thing.
Jstone
format
should
support
the
same
set
of
feature
flags
as
text
format.
Merrick.
Where
are
we
on?
That?
Is
that
done.
E
Someone
volunteered
and
assigned
the
issue,
but
I
I
think
this
is
pretty
hard
problem.
This
is
not
like
this
is
requires
someone
that
has
like
pretty.
E
We
can
like
this
was
put
into
beta
because
it
was
pointed
out
as
a
big
problem
in
alpha,
and
secondly,
if
we
like,
we
need
time
to
like
would
be
best
that
when
we
ga
we
have
already
deprecated
the
fla
the
flags,
that's
that
or
like
deprecation
may
be
removed.
The
flags
that
are
will
not
be
supported
in
future,
so.
B
I
think
that
it's
gonna
take
more
than
one
release
to
ga,
because
there's
like
given
how
hard
it's
just
been
to
migrate,
the
cubelet.
I
cannot
imagine
we're
gonna
migrate,
the
rest
of
the
codebase
in
one
release,
so
I
think
that
we
can
punt
this
so.
B
E
E
I'm
I'm
worried
that
the
distract
like
json
logging
will
not
be
usable,
so
we
could
like
here's
the
problem.
The
cap
touches
two
things
like
migration
and
json.
So
without
this
json
is
useless
technically,
because
you
cannot
use
anything
aside,
the
verbosity,
so
you
cannot
set
verbosity
per
module.
You
cannot
send
it
to
file
if
you
use
it.
B
Well,
in
theory,
this
is
the
point
of
the
cap
process.
Is
that
like
when
you
have
multi-sig
spanning
things,
they
should
be
coordinated
through
this
sort
of
thing
and
documented
in
one
place,
a
working
group
usually
has
like
more
than
one
topic.
They
care
about.
So.
B
C
B
Super
want
to
spend
time
like
talking
about
turning
this
into
a
working
group
when
we
have
five
days
to
code
freeze-
and
we
said
we
were
going
to
beta
this-
I
want
to
see,
if,
like
you
know,
it's
going
to
be
feasible
to
get
to
beta
and
if
we've
got
everything
written
down
that
we
need
to
do
or
not.
A
E
B
Oh
this
is
just
the
hublot
tracking
issue.
Okay,
the
title,
the
issue,
given
the
description
on
the
bug
didn't
make
sense
to
me,
so
json
format
should
support
the
same
okay.
B
But
we,
I
don't
think
we
can
put
like
there-
should
not
be
much
more
here
like
we
just
need
to
go
through
the
rest
of
this
list,
so
we
have
a
an
issue
for
that.
We
have
the
thing
for
that
which
is
the
cubelet.
This
is
a
to
do
that
I
will
try
to
either
take
or
assign
someone
to
and
then
what
so
mary.
I
think
you
added
this
about
the
end
to
end
test
scenario.
B
Okay,
let
me
add
a
card
on
here
to
remind
us
to
do
that.
B
And
han
wants
to
do
a
working
group.
So,
okay,
that's,
I
think
all
we
have
on
that
one.
So
I
will
stop
sharing.
C
C
So
my
proposal
is,
if
they
hand
this
stuff
over
to
us,
then
they
should
give
us
approvals
over
the
core
code
base.
A
C
Until
until
they
do
that
yeah,
but
is
everyone
on
board
with
the
the
events
thing,
because
it
kind
of
makes
sense
for
us
to
own
it.
B
A
Mean
like
these
things
kind
of
have
to
happen
when
a
project
is
big
right,
so
I
guess
we'll
need
to
just
figure
our
way
around
it
and
the
people
who
were
approvers
in
there
before
aren't
gone
right.
So
I
have
a.
I
have
a
good
relationship
with
forget
how
to
pronounce
this
name,
but
one
of
their
approvers,
no,
the
other
one.
Yes,
I
forget
one
of
the
two
people
who
were
approvers
for
the
v2.
A
Chelsea,
no,
it
doesn't
matter,
we
all
I'm
trying
to
say
is:
we
will
we'll
figure
it
out.
B
I
think
it's
okay
for
us
to
own
this
stuff.
As
long
as
we
have
approver
access-
and
we
like-
I
don't
want
this
to
just
get
dumped
on
us
because
I
don't
think
most
of
us
have
any
familiarity
with
the
code
base
and
there
are
existing
people
who
did
own
it.
And
so
I
don't
want
to
like
volunteer
and
say,
like
yeah,
of
course,
we'll
own
your
code
and
they
all
run
away.
A
I
I
I
do
think
like
it
logically
makes
sense
for
us
to
own
this
piece
of
code
and
like
since
people
are
like.
A
I
don't
know,
I
think
it's
the
right
thing
to
do,
and
I
I
think
I'm
happy
to
take
some
ownership
in
this.
C
B
A
Yeah,
so
the
the
exciting,
exciting
thing
about
this
tag
is
also.
This
is
the
first
time
that
we'll
be
using
automated
performance
tests
on
kubesave
metrics
lily
has
been
working
really
hard
on
automating
these
with
six
scalability.
So
I
mean
said
that
she's
not
here,
but
shout
out,
because
that
was
had
been
on
our
list
for
probably
years.
A
So
that's
really
awesome
that
will
now
regularly
test
performance
and
obviously
before
the
final
2.0
release
goes
out
as
well.
So
yeah,
that's
really
awesome.
E
A
C
One
more
thing
going:
circling
back
to
the
events
thing:
does
anyone
want
to
look
into
the
events
codebase
and
become
familiar
with
it
become?
I
don't
know.
B
C
B
Okay,
matthias,
we
have
an
item
from
huang
wei
about
scheduling.
We
should
do
that.
G
G
The
the
main
strength
is
just
where
helpers
cracked
correct
the
issues
of
running
the
scheduler
performance
test,
so
the
first
one
is
that
we
use
this
to
want
to
support
the
histogram
metrics,
but
in
the
upstream
we
change
that
some
metrics
from
historical
histogram
vector,
but
the
supporting
code
is
not
updated
properly,
so
this
first
first
one
is
for
that
and
the
first
one
unconditionally
returns
the
first
index
his
screen,
which
is
not
incorrect,
which
is
not
correct.
The
second
one
is
that.
G
We
call
some
dto,
oh
yeah,.
G
G
C
Yeah
yeah,
it's
on
me,
sorry.
I
have
been
out
of
office
and
I
am
currently
on
vacation
right
now
so,
but
I
will
take
a
look
later
today.
Yes
right,
I
have
to
show
you
that
yeah
no
problem.
A
Just
a
a
minor
comment,
just
glancing
over
this
right
now
that
the
the
one
pr
where
we
calculate
quantiles
and
stuff
like
that,
I
would
prefer,
if
that
wasn't
part
of
the
histogram
itself,
but
rather
a
separate
function,
but
that's
kind
of
minor.
I
can
comment
it
on
the
pr.
C
Yeah
we're
getting
two
two
metrics
into
stable.
C
There
are
two
pr's
out
for
it
and
I
have
been
contemplating
adding
a
third
which
is
so
right
now,
there's
a
pr
out
for
aps:
server
storage
object
counts
and
then
there's
there's
aps
over
request
totals
count,
makes
sense
by
the
way
I
was
okay
with
the
counts,
because
it's
actually
not
cumulative.
F
C
Seeing
yeah
so
I'm
okay
with
the
counts,
because
it
seems
to
signify
something
different,
though
it's
a
little
bit
confusing,
because
it's
a
counter.
It's
not
a
counter,
but
we
we
already
agreed
on
the
underscore
total
for
the
counters.
So.
A
Right
in,
in
that
case,
the
best
practice
would
actually
be
to
just
not
have
any
like
suffix.
It
would
just
be
api
server
objects,
okay
or
storage
objects
or
whatever
the
metric
would
be.
B
Thinking
about
the
whole
backlog
of
stuff
here
with
our
four
minutes
left
david,
what's
going
on
with
tracing.
D
So
I
was,
I
had
a
discussion
with
daniel
and
he
asked
for
some
changes
to
the
design.
So
there's
a
cap
update
that
I'm
waiting
for
review
on,
I
might
just
not
wait
for
review
and
actually
go
make
the
changes
I'm
proposing.
D
Most
of
them
were
kind
of
small
or
remove
functionality
that
we
were
planning
to
add
yeah.
I
just
need
to
give
given
the
tight
deadline.
I
should
probably
just
go,
implement
it
and
hope
that
everything
goes
through
some.
Some
good
news.
B
D
That
it
looks
like
etcd
was
able
to
resolve
some
of
the
issues
they
were
having
with
their
with
bumping
their
grpc
version.
So
I
think
they
got
it
to
like
it's.
Not
it's
not
all
the
way
solved.
So
I
I
it
might
be
enough
that
I
can
actually
implement
this,
but
I
would
have
to
use
a
slightly
older
version
of
open
telemetry,
which
is
unfortunate
because
a
lot
of
changes
have
come
in
since
then,
but
it
for
alpha
quality.
I
think
it'll
be
all
right,
but
yeah.
A
B
Okay,
so
I've
added
a
comment
just
so.
The
release
team
is
not
in
the
dark
on
what's
going
on
there
that
covers
tracing,
it
sounds
like
you
know,
we
might
slip
another
release,
but
it's
in
better
shape
than
it's
been,
which
is
good.
You
know
not
a
lot
remaining
metric
stability.
We
know
what
we
need
to
do
structured
logging.
We
know
what
we
need
to
do.
I
picked
clayton
about
this
one
there
to
graduate
to
beta
it's
just
a
matter
of
basically
updating
the
docs
we
don't
currently
have
like.
B
We
got
rid
of
the
phase
beta
on
metrics,
like
we
only
have
alpha
and
stable,
so
we
don't
need
to
like
bump
that
or
anything
like
that.
So
it's
really
just
a
dock
change
for
this
one
and
like
more
benchmarking,
I
guess
this
one
is
being
tracked
by
sig
security,
and
I
guess
the
only
one
we
got
left
is
metric.
Cardinality
enforcement.
I
know
there's
a
pr
up
for
that
han.
Is
that
making
good
time.
C
Yeah
yeah
dynamic
or
no
yeah.
B
C
Yeah
dynamic
cardinalities
but
yeah
the
the
doctor
lines
after
countries,
so
we
have
some
time.
A
Just
just
real
quick,
I
I
need
to
drop,
should
I
make
someone
host
or
do
we
wanna,
and
I.