►
From YouTube: SIG Instrumentation 20200709
Description
SIG Instrumentation July 9th 2020
A
All
right,
I
am
starting.
The
recording
today
is
july,
9th
2020.
This
is
sid
instrumentation
bi-weekly
meeting
welcome.
It
looks
like
we
have
a
single
item
on
the
agenda
today.
B
Sorry
I
I
did
and
just
thinking
about
it
it
probably
would
have
been
useful
if
I
had
invited
sally
actually,
but
so
this
is
a
a
component
that
sully
initially
created.
This
is
the.
B
Metrics
api,
the
custom
metrics
api,
as
well
as
the
external
metrics
api,
and
we
thoroughly
use
this
as
red
hat,
for
example,
and
there's
a
we
know,
there's
a
big
community
of
people
using
this
out
there
and
we've
had
it
multiple
times
that
we've
had
the
situation
multiple
times
where
people
were
like
saying
or
questioning
whether
this
is
equally
official
as
any
other
adapter,
seemingly
only
because
it's
in
under
sally's
police
user.
B
I
have
discussed
this
with
sully
before
and
he
was
happy
to
to
to
move
the
repo,
obviously
I'll.
I
would
double
check
this
again
if,
if
we
were
to
decide
to
adopt
this
project,
but
that's
essentially
my
my
proposal
that
we
adopt
this
project,
especially
because
pretty
much
every
maintainer
of
this
refill
is
already
part
of
sick
instrumentation.
A
Yeah
so
point
of
context
that
not
everyone
might
have,
but
it's
always
actually
on
my
team,
so
I
can.
I
can
actually
just
talk
to
them
today.
I'm
sure
it's
fine,
there's
a
couple
difficulties
about
transferring
stuff,
but
given
the
license
of
the
thing,
it
may
be
slightly
legally
more
convenient
if
we
just
created
a
sig,
instrumentation
repo
and
maintained
a
fork
and
then
had
sully
update
this
one
with
some
readme
that
says
this
one
is
no
longer
being
maintained.
Please
use
the
other
one.
B
Lilly
wrote
something
in
the
chat
she
can't.
Okay,
I
mean
she
asked.
Where
would
we
move
this?
To
which
org
I
mean
my
proposal
would
be
the
kubernetes
six
or
because
that's
where
everything
goes.
I
believe.
Regarding
the
the
legal
things
I
mean.
Obviously
I'm
no
lawyer,
but
I
I
would
assume
that
this
is
has
to
do
with
with
google.
B
Essentially,
my
understanding
is
that
solid
created
this
all
of
this
before
joining
google,
and
so
I
would
assume
that
this
is
probably
not
an
issue,
but
I
again
I
don't
know
if
that's
the
actual
thing
that
we
would
be
referring
to,
but
yeah
I
mean,
if
I
I
guess,
all
that
we
need
to
answer
within
this
group
is
whether
we
would
be
okay
with
adopting
this
project
as
a
sick
project.
B
A
A
So
if
someone
wants
to
maintain
the
kids
prometheus
adapter,
then
I
think
it
would
make
sense,
but
I
should
probably
leave
this
for
when
frederick
is
here,
because
I
since
may
have
an
opinion.
How
does
what?
What
does
everyone
else?
Think
about
that?
That's
unreasonable
or.
D
D
I
know
that,
like
we
had
a
couple
of
discussions
around
also
taking
the
adapter
to
kind
of
like
a
cid-based
approach
in
the
future
and
like
a
couple
of
conversations,
so
I
I
like
I'm,
definitely
not
in
the
position
to
kind
of
like
do
it.
I
know
a
few
people,
but
I
also
don't
want
to
like
throw
them
under
a
bus
right
now.
D
So
I
would
need
to
double
check
and
maybe
like
frederick
as
we,
as
we
just
said,
knows
more
yeah,
but
like
definitely
like
from
our
side,
we
are
interested
in
and
like
keeping
keeping
the
component
updated
and
it's
like
essentially
also
a
crucial
part
of
the
openshift
stack
that
we
ship
so
we're
definitely
interested
in
kind
of
like
maintaining
prometheus
adapter
going
forward.
What's
the
question
frederick
like
who,
who
would
kind
of
be
able
to
shepherd
this,
and
I
think
we
we
still
have
somewhat
like
yeah
need
for
it
to
be.
D
So
I
I
don't
know
if,
like
anyone
from
our
team,
can
specifically
shepard
this,
but
we
definitely
want
to
want
to
go
with
it
into
a
future.
B
Yes,
so
sergio
is,
I
guess,
the
main
maintainer
of
this.
B
The
the
thing
is,
the
the
component
is
fairly
feature
complete,
with
a
caveat
of
there
is
a
larger
refactoring
plan,
but
like
right
now,
it's
it's
just
not
a
priority
for
for
us
to
work
on
those
features,
but
in
itself
the
component
is
pretty
stable
and,
like
it's
updated
every
now,
and
then,
when
there's
a
new
kubernetes
version
or
something
right
like
there's
not
much
to
it
right
now,
it's
just
that
people
would
like
to
see
it
in
the
more
official
space,
and
since
we
are
already
tech,
instrumentation
members,
we
feel
like
it's
a
valid
space
to
live
in.
A
Yeah,
that
sounds
that
sounds
good.
How
about
how
about
we
take?
How
about
I
talk
to
sally
today
and
then
I
can
see
if
there
is.
A
If
there's
any
roadblocks
there,
then,
if,
if
not,
then
I
can
just
create
a
sig
instrumentation
repo
and
just
push
the
thing
there
and
then
I
saw
it
update
the
repo.
D
That's
for
something,
if
there's
an
alternative,
so
right
now
like
there's
the
metric
server
which,
like
obviously
doesn't
talk
to
prometheus
and
then
there
are
some
some
other
adapters
for
like
cloud
provider,
specific
metrics,
but
there
isn't
really
an
alternative.
That
kind
of
like
implements
the
api
from
like
the
api
service
interface
for
metrics.
That
would
then
talk
to
prometheus.
So
it's
basically
that
one
component
that
talks
like
that's
the
glue
between
prometheus
and
the
metrics
apis
and
kubernetes.
C
So,
like
my
question
is
like
seek:
instrumentation
already
maintains
the
library.
The
question
is:
should
the
integration
into
specific
components
be
justified
to
move
into
like
kubernetes,
like
sticky
instrumentation
of
kubernetes,
instead
of
like
to
give
it
to
prometheus
community?
C
A
No
that's
a
great
question.
There
is
precedent
for
stuff
like
this
and
I
I
can
in
api
machinery
the
way
that
we
do.
Clients
is
it's
basically
a
little
bit
of
both.
There
are
officially
supported
api
machinery,
clients
like
the
python
wrestling
client
or
the
python
client
go
client,
kubernetes,
client,
there's
also
clanco.
A
You
know
there's
certain
languages
which
are
supported.
So
that's
basically
like
this
right,
like
a
python
client
for
the
api
server
is
basically
one
implementation
of
that
interface.
B
Right
yeah,
I
I
think
the
same
question
we
could
ask
on
the
prometheus
side
right,
like
the
prometheus
side
could
say
the
same
thing
and
then
we
end
up
in
kind
of
a
undefined
situation
right.
But
this
is
specifically.
B
That
that
and
that
that's
the
point
where
I
was
getting
to
like
this
component-
is
almost
like:
95
just
kubernetes
style
code
right
and
so
that's
why
I
feel
it
makes
the
most
sense
in
sick,
instrumentation.
A
For
me,
this
is
an
implementation
period,
it's
just
because
it
uses.
For
me
this
is
my
permutation
maintain
it.
This
is
like
kubernetes,
specific
stuff,
so,
like
probably
the
best
mental
model,
for
this
is
like
there's
going
to
be
categories
of
things
that
people
bolt
on
top
of
instrumentation
such
that
there's,
like
third
party
libraries,
that
people
use
or
somebody
just
writes
in
a
personal
repo,
but
if
it
achieves
a
certain
level
of
traction,
then
as
a
community,
we
should
probably
say
we
are
officially
supporting
this
thing.
A
This
is
the
one
that
you
should
use.
You
can
use
other
ones
like
that's
fine,
but
they're,
just
not
supported
by
the
community.
B
Yeah,
I
I
I
don't
mean
to
promote
this
in
any
way
here
like
and
and
say
that,
like
this
is
better
than
metric
server
or
something
right
like
that's
not
what
we're,
what
we're
trying
to
do
here
at
all,
it's
a
binding
yeah,
exactly
it's
a
shim.
A
Yeah,
I
I
think
it's
fine,
it's
just
like.
I
mean
the
fact
that
api
machinery
has
multiple
clients
and
that
we
maintain
multiple
clients
doesn't
mean
that,
like
one
should
use
python
or
go,
that's
not
like
that
subjective
claim
is
not
being
made.
It's
just
yeah.
These
are
things
you
can
do.
These
are
things
that
you
can
use
and
we
will
support
these
official
ones.
B
Okay,
I
mean
I,
I
I'd
be
interested
to
hear
someone
else's
opinion
as
well,
but
it
seems
like
I
can
go
ahead
and
at
least
write
that
email
to
the
mailing
list
and
we
can
have
the
formal
conversation
yep
right.
A
Yeah,
okay,
so
that
was
that's
our
only
item.
Today's
also
code
freeze.
A
Oh
yeah,
that's
right!
Eleno
will
be
sort
of
on
soft
hiatus
for
like
the
next
month
or
so.
If
people
are
wondering
she
has
notified
us
in
advance
of
her,
she
has
a
lot
on
her
plate.
A
So
it's
a
difficult
time.
I
think
for
everyone
so
yeah
does
anyone
have
any
other
things
that
we
should
possibly?
Oh.
D
Yeah,
I
quickly
put
something
there
kind
of
like
hijacking
it,
but,
as
you
are
asking,
I
don't
feel
that
bad
now.
So
essentially
there
was
this
issue
about
like
making
the
api
server
request
metrics
table,
and
I
commented
a
couple
of
weeks
ago,
I'm
just
wondering
if
there's
anything
to
be
done
on
on
that
issue.
A
So
yeah,
so
I
have
actually
been
thinking
about
this.
I
and
I
have
not
had
the
bandwidth,
but
I
am
planning
on.
A
Thinking
about
this
more,
but
we
have
a
bit
of
time
before
120
now
that
could
freeze
so.
A
D
Like
like
no
activity
there,
so
I
just
wanted
to
follow
up
like
if
it's
still
on
your
radar
and
you
just
didn't-
have
the
bandwidth,
then
that's
that's
totally
fine.
I
just
wanted
to
check
in
kind
of.
A
A
Give
me
I
can't
say
that
I
will
have
time
in
the
next
by
the
next
one,
but
in
two
in
a
month
I
should
I
should
be
able
to
have
something
and
then
we
can
there.
Maybe
we
should
just
discuss
it
offline
because
I
am
like
your
use
case
is
also
interesting,
because
I'm
sure
we
have
different
things
in
mind
when
we
are
thinking
about
our
concerns
for
this
thing,
and
it
seems
like
you
may
have
some
insights
that
I
do
not.
B
Yeah,
I
think
in
general,
this
is
a
really
good
discussion
to
have
to
have
offline.
I
think
we
can
just
start
a
google
doc
and
kind
of
start
hashing
out
how,
because
I
think,
a
really
large
part
of
before
we
wanted
to
do
this.
We
wanted
to
review
once
again
the
like
set
of
labels
that
are
on
these
metrics.
Last
time
we
even
considered.
Maybe
we
will
only
have
the
latency
metrics
because
they
already
contain
counts
those
kinds
of
discussions
we
can
have
in
a
in
a
doc.
B
I
think
we
don't
necessarily
have
to
have
that
in
person.
I
think,
and
then
we
can
kind
of
lay
out
the
things
that
we're
trying
to
solve
with
each
of
these.