►
From YouTube: SIG Instrumentation 20210513
Description
SIG Instrumentation Bi-Weekly Meeting
May 13th, 2021
A
Hello,
everybody
today
is,
may
13
2021,
it
is
a
thursday,
and
this
is
sig
instrumentation.
We
have
a
few
items
on
the
agenda
today.
Welcome
first
thing:
enhancements
deadline
is
today
woohoo,
so
if
you
would
like
to
enhance
kubernetes
in
the
upcoming
122
release
cycle,
today
is
the
day
that
you've
got
to
get
all
of
your
acknowledgements
in
and
approvals,
speaking
of
which
we
have
some
caps
for
this
cycle,
which
I
will
share,
and
we
can
super
quickly
go
over
there's
only
three,
because
one
got
dropped
from
the
milestone.
A
B
A
Aps
server
tracing
happening,
good,
structured
logging.
I
gave
an
ack
for
the
kept
to
go
ahead
because
it's
being
taken
over
by
the
working
group,
so
I
approved
an
lgtm
that
and
it
would
be
good
to
merge,
if
not
for
the
fact
that,
like
they
did
prr
last
cycle,
so
I
think
it's
fine.
They
don't
have
to
do
that
again,
but
I
was
not
sure
if
we
need
to
like
move
that
into
a
separate
directory
or
something
now
that
it's
owned
by
the
working
group
so
worst
case
scenario.
A
If
merrick
doesn't
get
back
to
me
by
the
end
of
the
day,
I'll
just
cancel
the
hold,
so
it
doesn't
get
kicked
by
the
release
team
and
then
this
one.
I
think
this
one's
probably
gonna
miss
the
milestone
from
my
reading
of
it.
A
A
So
all
I
had
for
that,
so
we're
tracking
two
things
for
122.
and
this
one
technically
isn't
tracked
by
us,
because
it's
gone
to
the
working
group
which
I
think
just
met
this
morning.
I
wasn't
able
to
make
the
meeting.
I
had
a
conflict,
but
I
hope
that
that
went
well
any
questions
about
caps
before
we
move
on
in
the
agenda.
A
C
Hello,
the
thing
is,
we
are
I'm
in
the
process
of
migrating
the
the
cluster
add-ons
that
we
own
into
the
sig
instrumentation
add-on
org
in
kubernetes,
and
I
don't
actually
know
who
to
get
reviewed
for
it.
I
was
hoping
that
in
the
process
of
doing
it,
andre-
and
I
who's
the
other
owner
of
the
add-ons-
would
have
the
ability
to
do
it,
but
we're
still
getting
flagged
for
not
being
members
of
the
org.
A
A
A
A
So
I'll
submit
a
pr,
it
hopefully
should
get
approved
and
then
you
will
get
an
invite
and
then
once
you've
been
joined.
Then
it's
all
good.
A
Great
next
on
the
agenda,
we
have
sully
with
a
quick
overview
of
the
cube
builder,
docs,
tooling,.
E
Hey
yeah,
so
I
heard
y'all
were
interested
in
like
learning
how
we
build
and
host
our
q
builder
site,
so
figured
I'd
pop
in
and
give
a
quick
overview.
Let
me
share
my
screen.
E
E
E
So
I
believe
there
are
at
least
one
other
project
that
kind
of
has
a
similar
setup,
that's
under
the
kubernetes
umbrella,
so
you
could
probably
talk
to
sig
docs
to
get
that
set
up,
I'm
actually
using
that
wifi
is
pretty
straightforward.
You
can
point
it
at
a
repo,
and
then
you
can
you
get
this
nice
configuration.
E
We
actually
have
multiple
subdomains
pointed
at
each
of
our
branches,
so
that
we
can
manage
multiple
versions
through
a
subdomain,
which
is
something
that
no,
if
I
supports
pretty
straightforwardly
so
you
can
see,
this
is
just
the
top
of
our
nullify
config.
E
We
pointed
out
a
subdirectory
in
a
repo
and
tell
it
how
to
build
like
what
the
input
directory
is
just
up
here
and
then
what
like,
what
where
it
should
take.
The
final
built
files
from
which
is
out
here
netlify
has
some
like
default
configurations
for
various
tooling,
but
we
kind
of
use
our
own.
So
the
actual
book
itself
is
built
using
a
tool
called
md
book.
E
If,
if
anyone
here
is
familiar
with
getbook
mdbook
is
vaguely
similar
to
getbook,
it
was
written
originally
for
the
rust
language
project,
so
all
of
their
non,
basically,
their
equivalent
of
godok
is
in
their
own
tool,
but
all
of
their
kind
of
more
prose
doc
is
all
written
in
this
tool.
E
It's
a
really
nice
little
tool,
it's
very
similar
in
structure
to
md
books.
So
basically
you
just
have
a
bunch
of
folders
and
markdown
files
and
those
folders
that
you
can
use
to
define
the
structure,
and
then
you
get
you.
You
write
this
basically
table
of
contents
that
links
to
all
the
structure
you
want
and
together
those
two
are
used
to
generate
your
book.
E
One
of
the
nice
things
about
md
book
is
that
it
kind
of
supports
executable
plugins,
which
we
make
a
heavy
use
of.
So
I
can
take
a
look
at
our
configuration,
so
we,
for
instance,
use
them
for
kind
of
doing
like
literate,
the
literate
programming
parts
of
our
book.
So
when
we
slurp
in
files
from
our
example,
projects
and
turn
the
comments
into
markdown
and
for
our
auto-generated
cli
docs,.
E
But
you
could
also
use
this
for
doing
stuff
like
auto
linking
caps
or
something
like
that.
Basically,
it
just
calls
out
to
an
executable
that
gets
a
blob
of
json
with
the
structure
of
the
buck
and
then
gets
to
do
transformations
on
it
and
spits
back
out
whatever
it
wanted.
E
The
configuration
is
also
pretty
straightforward.
It
supports
technically
supports
different
languages,
we're
not
using
that
right
now
and
google
analytics
and
stuff.
If
you
need
to
track
that
kind
of
stuff,
as
well
as
kind
of
custom
themes,
it
comes
with
a
default
theme
and
then
you
can
just
override
individual
files
in
here.
It's
a
lot.
E
The
structure
is
a
lot
simpler
than
the
tooling
that
the
main
kubernetes
website
uses
to
generate-
which
I
think
is
a
lot
nicer
for
managing
and
editing,
makes
it
just
like
simpler
to
jump
right
in
and
figure
out
what's
going
on.
Is
there
a
generator
for
the
boilerplate?
What
what
do
you
mean.
D
Like
you
know,
the
total
file
summary
the
summary
thing.
E
Yeah,
the
first,
the
the
first
time
you
run
the
first
time
you
run,
I
believe
it
it
spits
out
like
a
bare
bones
tomofile,
if
you
need
to
change
the
theme
parts,
there's
a
flag,
you
can
pass
to
md
book
that
will
kind
of
like
scaffold
out
the
entire
default
theme,
and
then
you
can
just
choose
to
edit
what
you
want
or
delete
the
files
that
you
don't
need
to
be
anything
different
than
the
default.
E
Basically,
the
default
theme
is
like
built
in,
and
so
if
it
doesn't
see
an
override
print
file
and
then
for
the
summary
that,
like
you,
you
get
like
a
kind
of
default
one
when,
when
you
scaffold
things
out,
I
believe,
but
like
you,
this
is.
This
is
a
part
you
write
by
hand
because
it
defines
this
defines
how
it
shows
up
in
the
sidebar.
E
Basically,
so
technically,
like
the
the
folder
structure
is
not,
I
necessarily
does
not
need
to
be
identical
to
the
bullet
point
structure.
So
I
can,
you
can
see
like
you
can
see.
This
is
exactly
what
shows
up
over
here.
Was
that
summary
file,
except
it
has
numbers
in
the
output
version.
E
D
Nice,
this
is
nice.
Do
you
have
like
a
link
to
the
thing
that
you
used.
E
And
I
will
also
drop
a
link
to
our.
E
E
D
Yeah
this
this
sounds
promising
for
us.
We
should
generate
some
scaffolding
for
the
instrumentation
repo
that
we
have.
E
Yeah
and
if
you
need
like,
if
you
get
or
probably
talk
to
sig
docs
about
like
getting
netlify
set
up,
but
I'm
happy
to
I'm
happy
to
answer
questions
or
help,
like
figure
out
the
details
for
like
configuring,
your
particular
instance
as
well.
If
you
want,
because
I've
had
to
do
a
decent
amount
of
poking
around
in
the
netlify
ui,
to
get
some
of
our
more
complicated
stuff
set
up.
A
F
I
just
wanted
to
mention
that
we
really
thanks
to
thanks
to
new
contributors,
especially
yun
yang,
we
moved
and
yeah
also
to
help
with
diamond.
We
move
metric
server
implementation
really
forward,
so
we
are
currently
targeting
and
to
migrate,
metric
server
to
new
and
metric
resources
and
points
in
front
of
this
format.
F
But
we,
when
started
testing,
we
are
finding
some
more
more
issues
related
to
that.
So
I
just
wanted
to
maybe
bring
that
yeah
that
topic
with
the
proposal
for
endpoint
and
because
this
is
so.
The
overall
proposal
wants
to
add
a
new
start
time
metric,
which
is
not
standard
premiums
metric
because,
like
usually
premiums
come
so
normally
prometheus,
doesn't
use
start
time
for
any
like
any
metric
calculation.
But
this
is
something
that
we
we
technically
need
for.
The
metric
server
itself
and
premises
is
just
a
format.
A
So
I'm
looking
at
this
cap-
and
this
is,
I
guess,
one
of
the
ones
it
looks
like
it's
been
languishing
in
alpha
for
a
while,
because
according
to
graduation
criteria,
it
met
alpha,
but
not
beta,
and
I
don't
know
where
it's
currently
at.
D
A
F
So
sorry
yeah
there,
so
I
linked
to
two
issues.
One
is
related
to
more
to
how
this
match
some
one
is
related
to
the
proposal.
Second,
one
is
related
to
problem
with
the
start
time
itself
that
we
found.
D
Okay,
I
mean,
I
think
it
makes
sense
to
have
the
start
times.
I'm
just
pointing
out
that
there's
gonna
need
to
be
some
special
handling,
otherwise
we're
going
to
leak
memory.
B
Yeah
we
do
that
with
all
the
metrics
on
the
metrics
resource
endpoint.
One
other
thing
we
could
consider,
although
it
feels
like
we
changed
this
end
point
fairly
regularly,
is
that
while
prometheus
format
doesn't
come
with
start
times
for
cumulatives,
the
openmetrics
format
does
and
they're
sort
of
closely
related.
But
that's
one
of
the
main
differences
between
the
two.
G
A
B
D
D
Also,
there
is
a
process
start
time
in
prometheus.
It's
process
start
time.
B
Yes,
that's
correct.
I'm
saying
that
in
openmetrics
format
a
cumulative
data
point
right
includes.
B
Time
of
that
cumulative
metric,
so
it's
more
expressive
than
the
prometheus
text
format
is.
B
Anyways,
I
don't
know,
if
that's
it
would
be
another
breaking
change
to
the
end
point.
So
I
don't
know
if
we
actually
want
to
do
that.
It's
probably
just
as
well
that
we
expose
the
start
time
as
a
separate
metric.
D
D
B
D
A
B
A
I'm
and
I
just
realized-
I
was
confused.
I
mixed
up
metrics
resource
and
metrics
summary.
So
that's
why
I
was
like:
isn't
that
a
json,
endpoint
yeah
different
thing,
so
there's
actually
been
just
to
maybe
like
look
at
this
from
a
slightly
higher
level.
There
have
been
a
bunch
of
like
requests,
slash,
prs
and
other
stuff
coming
in
to
try
to
improve
like
metrics
coming
out
of
the
cubelet.
A
I
really
think
that
it's
problematic
that
we're
trying
to
do
it
piecemeal
and
not
looking
at
the
cubelet
as
a
whole
and
everything
that
we
want
and
there's
also
a
big
change
coming
in
for
the
container
stat
side
to
stop
using
c
advisor
for
a
lot
of
stuff
and
to
use
cri
stats
instead.
So
I'm
wondering
if
maybe
a
thing
that
we
can
do
for
this
release
is
like
assign
someone
to
look
at
that
and
come
up
with
a
more
holistic
plan
than
just
trying
to
like
deal
with
each
one
of
these
things.
A
B
A
A
A
Okay,
so
there's
that
then
I've
seen
at
least
like
one
pr
come
in
to
add
more
metrics
for,
what's
being
tracked
in
the
cubelet
like,
for
example,
pods,
but
the
naming,
I
think,
is
really
confusing
and
now
we're
seeing
things
with
like
container
create
times
here.
This
may
also
intersect
with
some
of
the
stuff
going
on
with
the
cri
stats.
So
I
I
think
we
should
try
to
coordinate
the
effort
as
best
we
can.
B
Okay,
I
I
do,
I
think
we
can
probably
consider
them
all
piecemeal.
I
think
they're
each
asking
different
questions.
I
think,
as
long
as
we
don't
change
the
scope
of
the
endpoint,
which
is
for
serving
resource
metrics,
I
think
we
can
probably.
B
A
A
B
If,
if
you
do
decide
to
go
to
sig
node
and
want
me
to
attend
just
let
me
know-
and
I
can
because
that's
probably
where
this
discussion
needs
to
happen-
yeah.
A
I
agree
and
I
will
be
there,
so
we
have
two
minutes
left
and
nothing
else
on
the
agenda.
We
do
have
some
new
or
occasional
faces
on
the
call.
Does
anybody
want
to
say
hi.
H
H
Along
with
my
job
and
a
baby
boy
in
the
house,
so
if
anything
comes
up
and
I
can
be
of
help-
I
would
really
like
to
get
to
catch
up.
I
don't
have
I'm,
I
I'm
working
as
a
devops
engineer,
so
I'm
not
sure
no,
I'm
not
like
a
developer
100.
A
Sounds
like
me,
I'm
an
sre
turn
dev
you
can
attend.
It
might
be
helpful
if
you're
looking
for
things
to
work
on,
we
have
a
triage
session
every
other
week
on
wednesdays.
It
wasn't
this
week,
but
it'll
be
next
week
on
wednesday,
and
the
time
is
on
our
calendar
and
on
our
instrumentation
charter.
Page.
G
Hi,
my
name
is
christine:
I'm
from
iceland
just
curious
on
how
to
get
involved
with
with
the
seeing
instrumentation,
so
I'm
just
lurking
at
the
moment.
Thank
you.