►
From YouTube: Community Meeting, September 20, 2022
Description
No description was provided for this meeting.
If this is YOUR meeting, an easy way to fix this is to add a description to your video, wherever mtngs.io found it (probably YouTube).
A
Foreign
to
the
community
meeting
for
kcpe
today,
September
20th,
and
we
have
the
agenda
here
in
front
of
us
issue.
2035
shift
topics,
please
add
seven,
so
we
are
pretty
soon
today
for
the
moment,
so
there's
plenty
of
space
and
time
today
I
think
yeah.
As
usual.
We
have
the
issue
hygiene.
We
do
attempt
I
added
one
topic,
just
a
link
to
the
document
started,
so
we
just
start
a
discussion
about
kcpio
which
we
acquired
a
few
weeks
ago.
A
That's
the
first
suggestion
what
we
could
have
there
and
basically
it's
pretty
much
resembles
what
we
have
the
readme
already,
maybe
minus
TMC,
so
TMC
is
something
of
course
you
want
to
have
in
some
way,
but
we
decided
to
eventually
separate
that
into
its
own
service
or
maybe
its
own
repository.
So
when
we
write
up
the
pitch
of
kcp
itself,
like
sap
core,
it
shouldn't
be
at
the
top.
So
that's
the
result
here.
A
What
I
suggest,
if
you
have
comments,
ideas,
better
ideas
about
wording
or
you
want
to
contribute
some
nice
drawings
or
something
like
that
yeah.
Please
join
this
effort.
That's
all
I
wanted
to
say
about
that.
A
B
Yeah,
just
two
super
quick
notes,
so
we're
still
very
slowly
making
progress
on
structured,
logs,
I
I
do
want
everyone
just
to
try
changing
like
one
file
a
couple
of
functions
over
get
into
the
headspace
of
attaching
logs
Etc
it's
a
useful
exercise
and
then
for
the
key
funks.
B
This
should
be
a
complete
no
up
to
everyone
generally,
but
in
the
process
of
moving
everything
over
to
like
strongly
typed
listeners
and
informers
that
require
you
to
put
in
a
cluster
when
you
need
one
part
of
that
has
been
creating
these
cluster
where
key
function.
Splitters.
B
The
format
is
slightly
different
from
the
stuff
that
we
had
in
the
cube
Fork
if
you're
doing
like
a
raw
get
by
key
you'll
need
to
format
the
key
manually.
So
if
you
look
at
geez,
probably
yeah,
that's
fourth
commit
format.
The
index
key
manually.
B
B
Yeah
we
have
a
suite
events
that
will
fail
because
you're
not
going
to
get
anything
from
your
indexer,
okay
and
I.
I
do
hope
that
this
will
be
resolved
like
fairly
shortly.
A
C
A
friendly
reminder
that
we're
hoping
to
hit
code
complete
next
Friday
on
the
current
Milestone,
our
last
Milestone
closed
on
time,
so
we've
got
a
lot
to
live
up
to
this
time.
So
let
us
know
if
you
have
any
risks
on
things
Landing
for
the
topics
you're
working
on
for
the
next
Milestone
planning.
You've
got
a
link
to
the
work
packages.
Document
We'll,
add
a
section
there
for
the
next
Milestone
I
I.
Think
we've
got
a
lot
of
long-running
topics
right
now.
A
All
right,
so
that's
going
to
be
a
short
meeting
today,
so
we
have
not
many
Works,
eight
of
them,
which
I
knew
so,
let's
go
to
them
one
by
one.
First
one
is
about
pre-condition
scales,
uid
and
precondition
and
I.
Guess
this
non-necessary
surprising
thing
Steve.
You
can
comment
on
that.
B
It
doesn't
seem
to
me
I,
don't
know
it
doesn't
seem
to
have
any
Downstream
effects,
I'm,
not
really
sure.
C
C
A
I
mean
of
course,
yeah
so
backlog
for
sure.
If
anybody
wants
to
pick
it
and
look
into
it,
just
self-assign
so
right
so
I
so
I
think
our
agreement
was
we
just
put
backlog
here
right.
That's
what
we
discussed
last
time:
okay,
yeah.
B
I,
just
and
maybe
this
is
a
general
comment
for
flake
hygiene.
It
probably
is
useful
to
log
these,
yes,
and
we
can
try
to
understand
how
often
they
happen.
People
can
link
to
them
when
they
retest.
D
Which
doesn't
really
give
you
what
you're
asking
for
like
you?
Can
you
can
go
into
test
grid
and
you
can
have
it
filter
and
only
show
failed
tests,
but
there
may
be
tests
that
show
up
there
that
fail,
because
somebody
was
working
on
a
PR
and
you
know
they
messed
up
something
and
broke
it.
So
you
really
would
want
to
look
at
failed
tests
against,
like
main
periodics
or
post
submits.
We
have
periodics,
but
not
post
submits.
B
D
A
D
Gotten
a
lot
better,
yeah
I
mean
I
would
say
like
if
you
are
writing
a
PR,
and
you
find
yourself
needing
to
comment.
Slash
retest.
The
first
thing
you
should
do
is
check
to
see.
Is
there
already
a
bug
report
filed
for
whatever
test
is
failing,
did
I
introduce
the
failure
in
my
PR
and
if
you
don't
think
that
your
PR
introduced
the
failure-
and
there
is
no
bug
report,
for
it,
add
a
bug
for
it.
Don't
just
hit
retest
over
and
over
again.
A
So
I
left
it
in
backlog.
Somebody
should
investigate.
We
have
to
pick
up
this
book
anyway,
because
it's
just
one
half
of
what
blocks
and
what
exec
so
I
think
we'll
come
back
to
that
when
it's
happening
or
more
often
at
least
next
one
something
about
shared
compute
when
placements
are
the
details.
We
edit
back
I,
do
not
see
the
employment
going
from
one
to
zero.
B
A
D
Yeah
I,
don't
remember
that
error
is
coming
from
our
yaml
patch.
So
if
we
need
to
change
it
because
it's
okay
to
have
a
leading
number,
we
can.
A
D
A
No,
no
wait,
I
think
there's
one
one
thing
that
we
had
it
in
secret
Machinery
some
months
ago:
DNS,
sorry,
CID
or
CR
names.
Actually
they
are
more
constrained
than
native
type
names
and
I
I.
Think
I
remember
it
was
about
DNS
segment
and
there
was
not
a
real
strong
reason
just
that
maybe
people
use
CR
names
in
DNS
contacts
and
it's
better
to
do
that.
Something,
like
that,
so
hand
waving,
but
we
cannot
change
it.
That
is
my
memory.
I
might
be
wrong,
but
I
think
it's
a
CID
restriction,
yeah.
D
C
D
A
Oh
yeah
yeah,
we
we
talked
about
that
today.
That's
a
common
problem,
people's
face
who
are
building
apis,
that
you
want
to
have
objects,
not
types
but
real
objects
shared
among
workspaces,
and
we
had
a
lot
of
discussion
around
projections.
Maybe
projections
should
be
a
generic
API
I.
Don't
think
this
is
a
good
it's
a
solution
for
for
that
use
case
here,
possessions
after
group
resource,
basically
like
application
is
used
in
queue,
hopefully
safe
in
a
safe
way
that
doesn't
take
down
the
system,
but
basically
they
are
like
aggregation.
A
D
A
A
Yeah,
just
because
it
needs
a
design,
it's
not
fear
of
this.
We
can
put
it
here
but
doesn't
mean
that
people
work
on
it.
True.
A
Definitely
anyway,
so
egglock,
so
quick
if
you're
here,
and
you
definitely
need
that
spiritual
talk.
A
C
Yeah,
so
it's
the
only
thing
and
we
discuss
that
audio
is
that's.
C
Available
to
the
end
user,
we
wouldn't
want
something
like
deployment,
which
is
a
resources.
That's
not
always
get
the
same,
at
least
what
is
generated
about
right.
So
when
we
sync
plugin.
B
Yep
and
so
the
work
that's
undergoing
right
now,
I
should
fix
that
so
you'll
be
able
to
opt
into
whatever
resources
you
need
on
that
particular
scene,
Target
and
then
having
to
point
to
the
same
underlying
cluster
seems
like
a
totally
reasonable
I.
A
All
right,
so
what
should
we
do
here?
Basically,
there's
something
missing:
I
commented,
but
we
are
near
right.
One
PR
is
missing
and
configure
map
Secret
Service
account
syncing
that
it's
selective.
A
C
A
Yeah,
this
was
a
discussion
some
minutes
before
the
meeting
Mike
was
bringing
up
the
question
about
all
those
static
endpoints
we
have
in
Cube
and
them
wasn't
that
in
those
confidence
leg.
A
Basically
we
identified
those
which
are
no
point
in
us
and
we
talked
about
that
before,
because
I
think
Fargo
CD
was
needing
any
of
that
or
something
like
that.
So
we
should
probably
map
some
into
workspaces
so
just
into
the
endpoint
of
one
workspace
and
maps
and
I
mean
encode
and
project
them
on
the
public
top
level
endpoint.
This
is
easy
to
issue
help
wanted.
If
anybody
wants
to
look
into
that,
it's
probably
not
very
hard,
but
you
see
the
I
mean
the
Handler
chain
and
another
filter
there.
A
So
that's
probably
it
must
be
done.
There
are
more
endpoints,
especially
open
API,
which
we
talked
about.
Maybe
last
time
I,
don't
remember
we
want
it.
It's
not
static
like
it's.
It's
not
uniform
of
our
workspaces,
so
we
have
to
do
computation
in
some
sense
and
I
think
so
yeah.
Basically
the
theory
outwards.
Yes,
we
want
it,
but
it's
not
trivial
to
do
it
right
anyway.
So
it's
still
open
and
the
other
one
metrics
we
discussed
a
little
bit
more
metrics,
which
are
workspace.
A
Labels
are
not
possible,
so
we
had
this
discussion
before
that.
The
workspace
label,
which
is
unbounded,
explodes
the
cardinality
of
metrics
we
have
so
we
cannot
do
that.
We
can
have
aggregate
metrics
over
all
workspaces,
that's
what
we
have
today
and
the
question.
What
Mike
was
basically
asking?
Can
we
have
metrics
in
a
workspace
just
with
those
counts,
offset
workspace
and
without
any
workspace
label,
because
it's
just
one
workspace
and
what
I
understood
and
so
I
goes.
If
he's
here,
he
can
correct
me.
A
This
is
something
we
can
do,
because
we
will
not
query
that
endpoint
for
every
workspace,
but
the
user
can,
if
they
want
to.
So
if
a
user
wants
to
connect
their
own
Observer,
so
the
commuters
instance
and
confine
and
everything
they
can
do
that.
That's
something.
A
C
C
Metrics
and
there's
also
APF
API
priority
and
fairness
has
a
debugging
point
that
is
needs
to
be
a
workspace
or
logical.
Cluster
specific
I
see
yeah.
C
I
guess
that'll
be
a
relatively
easy
one.
I
think
the
metrics
has
an
additional
difficulty,
which
is
that
the
Upstream
code
right
it
doesn't
have
any.
It,
doesn't
wrap
up
the
metric
collection
behind
the
interface
that
we
can
instantiate
per
workspace.
So
there's
a
real
code
based
problem.
A
A
All
right
so
much
about
that
Mike,
any
anything
I
forgot
you
had
some.
There
are
some
things
around
open
ideas,
so
right
well,
I'm
not
even
sure
what
it
means
when
we
have
those.
C
I'll
start
by
just
asking
the
question
yeah,
so
it's
a
question
that
needs
to
be
asked
right.
Those
are
currently
part
of
what
a
user
can
do
to
an
API
server.
So
it's
a
question.
A
A
A
Okay,
sorry,
let's
really
just
to
build
the
cue
cutter
and
other
Zips
or
Target.
That's
what
we
have
there
attached
to
leases.