►
From YouTube: Community Meeting, February 14, 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
Welcome
welcome
to
the
February
14th
community
meeting
for
kcp
dropping
the
issue
into
chat
and
I'm
going
to
go
ahead
and
share
my
screen.
B
A
Okay,
so
I
believe
Andy
Anderson
made
the
first
issue.
C
Sorry
I
was
on
you
hi
everybody,
so
a
number
of
different
developments
happening
in
kcp
Edge
this
past
week
or
so
we've
gotten
some
critical
mass
around
a
reading
list
that
we've
developed
on
medium.
C
So
if
any
of
you
are
interested,
we're
starting
to
proliferate
proliferate
a
lot
of
information
having
to
do
with
experiments
that
we
did
last
year
and
incorporating
those
into
the
kcp
edge
point
of
view
and
we're
starting
to
see
results
in
indexing
in
Google
for
things
for
SEO
and
so
forth.
So
we're
starting
to
do
a
lot.
Oh,
that's!
That's
terrible!
C
Maybe
that
the
reading
list
doesn't
show
up
there,
it
is
so
in
the
top
in
the
top
collection
there
in
kcp
Edge.
C
So
there's
a
number
of
articles
that
have
been
picked
up
by
other
providers,
we've
written
some
internally
and
there's
some
that
were
shadowing
from
folks
like
Chick-fil-A
and
Cisco
and
others
that
are
starting
to
do
experiments
in
this
space.
So
we
also
came.
If
you
see
the
logo,
we
came
up
with
a
logo
this
weekend.
I've
ran
it
by
some
of
the
folks
over
kcp
I.
Think
it's
pretty
cool.
C
It
pays
homage
to
kcp
itself
so
when
we
think
we
think
it
it
expresses
consistency,
heterogeneity
and
scalable
Edge
configuration
management,
so
those
are
some
of
the
the
developments
that
are
happening
there.
So
that's
item
number.
Two
item
number
one:
we've
we've
started
to
we
created
what
was
what
we
call
a
coding.
Milestone
I
just
dropped
a
link
in
the
in
the
chat
here.
So
I
am
shamelessly
soliciting
help
to
try
and
complete
this
coding
Milestone.
C
And
so
the
top
half
of
the
document
describes
some
of
the
spots
where
we
need
some
help.
These
are
more
kind
of
high
level.
There's
issues
now
being
created
from
this
in
the
kcp
edge
repository
itself,
and
if
you
scroll
down
a
little
bit,
one
of
the
one
of
the
areas
where
we're
trying
to
emphasize
is
good.
C
First
issues,
I
think
it's
a
good
practice
and
help
people
get
like
you
know
acquainted
with
how
kcp
Edge
operates,
what
we're
focused
on
and
where
we're,
where
we're
starting
to
do
experiments
and
some
of
the
spaces
where
we
could
use
some
help,
specifically
from
kcp
I
was
coming
on.
This
call
to
you
can
stop
me
from
talking
if
you
volunteer
for
one
of
these
items,
so
I'm
just
going
to
filibuster
here
so
yeah.
So
we
have
a
few
items.
You'll
see
one
and
specifically,
is
in
underneath
kcp
issue
number
2682.
C
It's
a
it's.
Basically,
a
way
to
have
the
Sinker
do
different
behaviors,
based
upon
some
switching
internal
to
the
code
and
we'd.
Like
your
help,
there
it'll
help
us
change.
The
behavior
of
the
Sinker
Mike
can
elaborate
a
little
bit
more
on
what
that
is
if
you'd
like
them
to-
and
we
also
have
another
item-
that's
closely
related
to
this.
That's
also
on
that
list
and
do
some
help
from
kcp
as
well
there
it
has
to
do
with
denatured
objects
and
for
the
purpose
of
synchronization.
C
So
we
were
thinking
about
the
possibility
of
alternative
object
types,
but
the
general
gist
of
these
of
these
issues
is
that
we've
got
for
denaturing.
Is
that
when
you
send
down
an
object
to
be
throughout
the
through
the
topology
or
wherever
the
kcp
API
machineries
exist,
you
don't
necessarily
want
the
API
Machinery
to
act
upon
it.
You
know,
in
other
words,
take
a
deployment
and
turn
it
into
pods.
You
may
want
to
just
use
it
as
a
place
to
stage
and
pass
it
on
to
the
next
level.
C
For
then
for
it
to
be
acted
upon,
so
we
call
those
denatured,
workspaces
or
denatured
behavior,
and
we
could
use
some
help
there.
So
on
those
and
there's
one
other
in
there,
we've
got
some
large
scale,
testing
that
we'd
like
to
test
out,
and
we
thought
maybe
Steve
might
be
helpful
there.
So
those
are
the
three
items.
D
F
You
know
resource
involvement,
because
we
we
already
you
know,
use
those
resources,
but
about
the
Thinker
at
least
just
wanted
to
notify
that
there
is
ongoing
work,
at
least
for
the
I
think,
one
or
two
next
weeks
about
adding
support
in
The
Thinker
for
a
multi-shot,
and
this
will
obviously
change
a
bit
to
refactor
of
the
structure
of
the
you
know,
main
Sinker
flow,
so
yeah
that
address
could
be
useful
if
we
can
sync
together
and
maybe
avoid
starting
some
sort
of
refactoring.
C
G
D
Right-
let's
just
put
a
little
bit
finer
point
on
that:
David
I
think
there
are
two
points
of
diversions
that
we've
identified
between
the
TMC
Sinker
and
the
EMC
Sinker.
F
E
D
B
D
Then
we
can
submit
a
PR
that
does
the
work
there.
The
other
thing
is-
and
we
haven't
had
time
to
discuss
this
yet
in
any
detail,
but
we've
been
talking
about
in
the
denaturing
certain
object
types,
but
the
Sinker
will
need
to
renature
them.
D
We'll
need
a
kind
of
a
plug
point
where
we
can
plug
in
something
that
will
renature
these
object
types
and
again
you
know,
maybe
just
you
know
if
you
just
put
your
finger
on
where
that
plug
Point
belongs,
you
know
we
can
do
the
pr
to
actually
make
it
so
yeah.
F
I
mean
I
I'd
say
the
first
one
is
is
quite
simple,
and
even
if
there
is
some,
you
know,
restructuring
of
the
main
Sinker
flow
for
multi-shot
support.
This
should
not
really
hurt,
even
if
you
do
that,
if
any,
if
even
if
we
do
the
separation
first
for
the
second
Point,
probably
it
requires
a
bit
more
thought.
I
mean
I
I'm,
not
100,
clear,
because
it
also
touches
the
the
type
of
the
resources.
D
Thought
we're
willing
to
do
most
of
the
work.
It's
just
you
know
you're
the
expert
on
the
code,
so
it
would
help
us
a
lot
if
you
could
just
kind
of
point
us
where
we
need
to
be
looking
yeah.
F
No,
no
problem,
and
just
we
can
possibly
you
know,
set
up
some
meetings.
You
know
just
working
sessions
or
anything
like
that
sure.
A
E
D
C
A
Great
thanks,
Andy,
Mike
and
David
and
I
believe
Mike
is
next,
not
sure
if
you
want
to
continue
the
discussion
on
denaturing
objects.
D
Yes,
this
is
again
picking
up
something:
it
really
happened
in
the
edge
community
meeting
last
week,
but
you
know
Stefan
and
Andy
and
David
were
really
getting
into
this.
So
for
this
community
for
Edge,
we
have
this
issue
of
what
we're
trying
to
do
is
use
the
kcp
workspace
as
a
container
for
a
description
of
the
desired
and
reported
state
of
a
workload
that
runs
self-sufficiently
at
the
edge,
so
we're
using
a
workspace
as
a
container.
D
But
the
problem
is
that
some
kinds
of
objects
they
affect
the
API
server
Behavior.
So
we
can't
just
use
a
regular
ordinary
workspace
as
a
container
for
any
kind
of
object
right
if
you
want
to
put
our
back
object
in
if
we
want
to
put
you
know
a
crd
in
you
know,
these
things
affect
the
API
server
Behavior,
where
you're
putting
them
so
for
cids.
Actually,
this
is
an
easy
case
because
we
want
their
behavior,
but
for
the
our
back
objects,
for
example,
we
don't.
D
There
are
other
things
like
admission
control,
plugins,
also
that
affect
API
server
Behavior.
We
want
those
running
at
the
edge
not
in
the
center,
so
we
need
a
way
to
have
denatured
objects
in
the
center
for
these
kind
of
objects.
But
we
want
stuff-
that's
been
shifted,
left
to
view
them.
As
the
ordinary
types
Stefan
proposed
a
nice
solution,
which
is
the
stuff,
that's
shifted,
left
Works
to
review
so
the
stuff
that's
shift
to
the
left.
D
D
And
sorry
API,
export
and
binding,
so
that
sort
of
regular
workspaces
would
bind
to
the
regular
definitions
of
our
back
and
admission
control
plugins.
But
we
would
also
in
this
case
be
able
to
bind
to
denatured
versions
of
these
I.
Think
that's
I
I
think
that's
more
mechanism
than
we
need
I,
don't
think
we
need
a
a
new
kind
of
API
export
and
binding
we're
talking
about
a
fixed
set
of
type
object,
types
that
are
known
at
development
time.
D
So
all
we
really
need
is
alternate
object,
types
and
the
view
can
just
translate
have
a
fixed
built-in
development
time
translation
but
anyway,
I
just
wanted
to
kind
of.
Maybe
if
we
could
reach
some
conclusion-
and
you
know
on
the
and
then
you
know,
proceed
Andy.
I
D
Yeah
by
denatured
I
mean
simply
that
this
is
an
object
that
is
stored,
but
has
there's
no
no
controller.
That
does
anything
about
it,
so
our
back,
for
example,
has
controllers
and
the
problem
is
for
whether
it's
controllers
built
into
the
API
server
right.
So
the
API
server
has
controllers
to
act
on
our
back
objects.
It
has
stuff
that
acts
on
the
admission
control
plugins
right.
D
So
if
someone
who's
developing
for
the
edge,
they
have
our
back
objects
or
admission
control
plugins
that
they
want
to
run
in
the
edge
they
don't
want
it
to
run
in
the
center.
So
they
need
a
way
to
you
know
all
the
code.
That's
shifted
left
that
thinks
it's
developing
our
back
objects
and
Mission
Control
plugins.
You
know,
and
it's
gonna
you
know,
do
a
coupe
cuddle
create
or
whatever
of
our
back
objects
and
emission
control
plugins.
It
needs
to
not
actually
affect
the
behavior
of
the
kcp
workspace.
F
Sorry
it
was,
it
wasn't,
mute,
yes,
I
was
about
to
say
maybe
it's
a
technical
detail,
but
it
might
not
be
in
the
Sinker
the
same
way.
You
might
have
a
view,
for
you
know,
end
users
to
to
put
their
objects
as
with
the
normal
apis.
The
same
way
that
you
could
have
a
view
that
the
Sinker
points
to-
and
this.
D
F
H
D
Another
possibility
that
occurred
to
me,
probably
not
preferable,
is
we
could
add
a
controller
in
the
edge
clusters
that
does
the
renaturing,
so
the
Sinker
propagates
the
denatured
objects
and
then
separately
on
the
on
in
the
edge
cluster.
They
get
copied
into
renatured
objects.
But
that
sounds
like
you
know:
that's
two
copies
of
the
objects.
That's
that
sounds
like
it's
more
trouble
than
what
you
suggested.
B
A
I
was
going
to
ask:
are
these
views
going
to
be
like
hard-coded
or
defined
in
the
code
versus
dynamically
generated
via
an
eight
by
API
export?
Is
that
you're
thinking.
D
Well,
I'm,
not
sure
I'm
not
really
familiar
with
the
view
technology
here
you
know
I'm.
So
this
is,
you
know
for
the
edge
MC
repo
I
think
we
would
be
responsible
for
the
view.
I
would
need
to
learn
more
about
what
it
takes
to
actually
Implement
a
view.
F
J
F
Assume
that
you
know
I
assume
that
the
the
current
virtual
workspace,
which
is
Synchro
view
if
we
can
sit
so
with
the
new
term,
would
be
quite
an
at
least
parts
of
the
the
simplest
parts
of
it
would
be
quite
interesting
as
a
as
an
exercise
to
to
look
into,
because
mainly
it
takes
the
list
of
the
resources
exposes
them
based
on.
You
know
what
exists
on
the
scene
Target
and
then
exposes
only
the
the
objects
that
are,
you
know,
planned
to
be
synced
to
a
given
Synchro.
F
So
obviously,
all
the
inputs
you
could
take
you
know
with
your
own
rules
of
of
agency,
but
the
main
principle
of
having
a
view
that
only
exposes
a
subset
of
apis.
Possibly
you
know
changes
in
your
case.
You
would
probably
change
the
group
name
or
something
like
that,
but
the
Shema
would
be
the
same.
So
probably
you
could
start
from
the
the
syncope
virtual
workspace
if
I'm
not
mistaken,.
A
And
there
was
a
question
in
chat
from
sorry
if
I
pronounce
my
name
wrong
but
mangierdas,
asking
if
there's
a
write-up
or
a
doc.
D
Yes,
in
the
edgemc
repo
I
do
have
a
PR
that
hasn't
merged
yet
introducing
an
outline
of
the
POC
we're
trying
to
work
on
it's.
It's
like
still
a
work
in
progress
because,
as
you
can
see,
you
know,
there's
still
design
being
done
about
these
handling.
These
denatured
objects.
D
J
A
J
A
Yeah
I
was
gonna
say
it
sounds
like
the
action
is
to
investigate
the
Sinker
View
and
see
how
that
translates
or
it
could
be
used.
Yes,.
D
Andy
asked
for
that
link
to
that
PR.
Let
me
just
get
that
for
you
guys
and
drop
it
in
chat
and
then
I'll
I'll
do
that
offline
go
ahead
and
proceed
in
the
agenda.
Okay,
Fabrizio.
G
Yeah,
thank
you.
Hi
everyone,
I'm
Fabrics,
yeah,.
G
Before
with
beans
and
at
the
Costa
cycle
and
since
I
missing
to
annoy
them
with
my
request,
I'm
now
looking
at
kcp
so
yeah,
but
apart
is
so
what
happening.
I'm
starting
ramping
up
in
TCP
I
found
I'm
experimenting
and
initially
I
started
doing
everything
by
command
line.
But
then
I
started
to
look
at
the
end-to-end
test
following
suggestion
from
David,
which
I
thank
you
and
then
to
and
test
are
really
useful
to
learn
how
things
works,
and
so
I
figured
it
out.
G
But
there
is
a
problem
for
people
learning
is
they
tend
to
untest?
Basically,
it
goes
through
only
one
scenario
and
they
do
not
allow
you
to
easily
experiment,
and
so
I
come
up
with
something
that
I'm
going
to
to
share
to
you
and
see.
If
you
are
interesting
or
not,
is
basically
are
things
that
I
built
on
top
of
a
kcp
end-to-end
test
framework
can
I
share
for
a
to
show
it
very
very
quickly.
B
G
Okay,
here
it
is
so
what
what
I
built
is
another
plugin
which
I
call
the
kcp
playground
what
it
does.
It
takes
an
input,
a
config
file
and
we
will
have
a
look
in
in
a
second
and
basically,
it
spin
UPS
everything
that
is
written
in
the
config
file
using
the
the
framework
utility
in
this
case.
It's
building
up
for
me
ashard
a
pair
of
kind
clusters
and
then
it
is
creating
workspaces,
sync
Target,
whatever
defined,
and
then
basically,
if
I
want
to
use
my
plugin.
G
My
environment
I
can
use
a
single
Cooper
config,
which
actually
means
it's
a
uber
Uber
config.
That
sum
UPS,
all
the
shards
and
and
the
kind
whatever
so
I
can
do
stuff
like
cubicator
use,
a
Shard
or
Cuba
cattle
use,
P
cluster
and
each
and
quickly
switch
from
one
to
another.
So
I
don't
even
have
to
need
to
add
the
cognitive
effort
to
follow
about
work.
The
configure
and
just
to
look
at
this
is
these
the
config
file.
G
For
now,
it
is
allows
you
to
Define
shards
for
now
to
support
only
one,
but
that's
a
detail
within
the
chart
that
our
species,
which
could
be
nested
and
within
workspaces,
you
can
have
a
sync
Target
placement
and
yeah.
Then
you
can
easily
add
whatever
you
want
and
just
to
show
you
that
this
is
backed
by.
G
If,
if
you
could,
if
you
remove
all
the
scaffolding
to
get
a
Cubist
cuttle
plug-in.
Basically
this
is
the
code
that
spin-ups
everything
is
not
a
big
amount
of
code
and
yeah.
As
you
can
see
it
uses
a
framework.
Give
me
a
kcp
cluster.
Give
me
a
workspace,
install
the
C
framework,
what
it
is,
install,
sync
and
so
on
and
so
forth.
So
yeah
I
I
find
it
nice.
So
I
decided
why
do
not
go
back
and
show
to
the
team
and
see
if
they
are
interested
in
it
or
not?.
D
Yeah
I'll
tell
you
my
experience
running
through
the
quick
starting
kcp
I,
actually
like
issuing
the
commands
by
hand
and
I
would
go
through
it
more
slowly
than
the
quick
start
is
I.
It's
really
interesting.
What's
going
on,
I
looked
around
at
objects
before
and
after
each
command,
so
I
understood
in
more
detail
what
the
commands
are
doing.
D
D
It's
it's
nice
to
have
something.
That's
kind
of
one
button,
automated
or
just
a
series
of
button
pushes,
but
it's
also
I
think
important
people
be
able
to
dig
in
in
all
the
details
and
poke
all
the
way
around,
and
that
reminds
me
of
another
issue
which
I'm
not
sure
how
it's
reflected
right
now
in
issues
in
the
repo.
But
there
is
you
know
an
issue
with
the
current
quick
start.
Is
it
conflates
the
use,
conflates
some
different
roles
into
and
uses
the
work?
D
The
root
workspace
for
in
unusual
ways
right
in
sort
of
the
usual
scenario?
You
would
have
more
workspaces
and,
and
the
quick
start
you
know,
overloads
the
workspace.
Sorry,
the
root
workspace
for
things
that
we
would
never
Advocate.
We
don't
advocate
so
I
think
in
terms
of
just
the
scenario,
regardless
of
how
it's
presented
my
educate,
my
position
and
I
think
someone
else
said
too.
The
scenario
should
be
more
articulating
so
that
it
uses
other
workspaces
in
a
more
typical
fashion.
D
F
Maybe
just
just
a
point
on
this
last
Point:
yes,
there
is
a
pending
Pier
about
changing
the
readme
and
the
quick
start
in
order
to
use
two
workspaces,
one
for
location,
workspaces
and
one
for
user
workloads.
At
this
time
someone
raised,
you
know
some
some
Community
member
raised
the
fact
that
it
might
be
a
bit
too
complex
at
the
very
first.
You
know
read
me
where
you
just
want
to
execute
your
workload
inside
a
single
workspace,
so
I
mean
I,
think
we
we
should
continue.
I
I
brought
back.
F
G
Yeah,
just
just
want
to
answer
to
the
the
question
about
following
step
by
step:
I
think
that
that's
a
great
point
at
the
same,
so
maybe
I
want
to
really
fast
in
in
the
yam
or
where
I
Define
the
scenario.
There
are
a
bunch
of
common
describe
what
what
happened?
G
What
makes
so
this
scenario
basically
allows
you
to
repeat
and
change
it
right
very
fast
and
experiment
which
yeah
it
is
helping
me
to
understand
kcp.
So.
G
Are
value
in
both
method,
as
I
would
say,
as
a
people
approaching
and
I'm
also
trying
to
help
some
other
colleagues
to
look
at
this
as
well?
G
It's
kind
of
hard
dealing
with
kcp,
a
bunch
of
kind
clusters,
etc,
etc,
and
and,
in
the
meantime,
try
to
figure
out
your
mental
map
of
our
space
come
up
together
and
this
L5
can
help
you
to
build
this
metal
pipe
so
yeah
there
are
prime
cones
and
I
understand
your
your
comment.
I
and
I
learned
a
lot
about
by
going
comma
by
command,
but
then,
when
I
wanted
to
iterate
fast,
this
kind
of
conflict
driven
things
helped
me
a
little
bit
so
yeah.
D
Yeah
can
I
just
add
to
that
also
yeah
I
yeah
I
mean
I
I
do
think.
There's.
Another
idea
that
has
occurred
to
me
is
I
think
we
could
use
a
picture.
You
know
if
someone
drew
a
picture
of
what's
going
on
in
this.
D
Of
you
know
all
the
API
exports
and
bindings
and
the
workspaces,
and
just
how
things
relate.
You
know,
I
think
that
that
would
also
help
it
make
it
much
more
comprehensible.
I
This
is
awesome
for
retail,
thanks
for
working
on
it
and
showing
it
to
us.
I
think
we
should
merge
it
so
I
mean.
Maybe
we
call
it
something
like
test.
Bed
I
mean
playground
is
fine,
but
whatever
we'll
come
up
with
a
name
and
bike
on
that,
and
then
I
think
this
is
just
a
yes
and
like
you
can
do
step
by
step.
You
can
do
this
so
they're,
both
extremely
valuable,
so
kudos
to
you
for
Regios
is
super
awesome
and
I'm
looking
forward
to
seeing
a
PR.
F
Just
a
thought
about,
you
know
the
demos
that
once
existed
in
into
kcp
that
we
dropped
because
they
were
much
too
complex
to
maintain
and
quite
a
big
bench
of
Bash,
but
one
idea
was
that
you
know
you
and,
and
then
we
switch
obviously
to
extensions
extensively
use
end-to-end
test,
which
is
the
the
right
direction.
Obviously,
but
you
know
in
some
cases,
I
don't
know
you
at
least
from
me
for
myself,
I'd
like
to
test
manually
with
my
own.
F
You
know
according
to
where
I
am
in
the
development
things
that
finally
will
be
tested
in
the
end-to-end
test
as
well.
You
know,
especially
if
you
think,
of
high
level
end-to-end
tests
like
the
deployment
splitter,
like
you
know,
the
the
complete
test,
where
we
check
that
the
workspace
is
Downstream
are
correctly
isolated,
how
Services
can
communicate
or
things
like
that.
These
are
really
high
level
features
that
we
also
mainly
you
could
just
do
a
demo
with
exactly
the
same
thing
as
what
you
have
in
the
end-to-end
test.
F
For
such
things,
it
seems
to
me
that
it's
very
it
could
even
be
have
a
great
value
that
we
have
the
same.
You
know
bootstrapping
method
for
all
the
contexts
and
then
that
would
be
used
both
by
the
end-to-end
test
and
by
the
manual
test
that
you
would
like
to
do,
for
example,
to
debug
simply.
G
Yeah
I
totally
agree
when
I
was
thinking
at
these.
Is
that?
Okay,
when
I
got
these,
let
me
say
we
can
ideally
rewrite
end-to-end
test
to
use
the
same
Machinery
that
I
use
in
the
playground
to
spin
up
then
to
then
test
scenario.
So
you
will
basically
have
a
configuration.
You
can
use
it
in
the
end-to-end
test
or
you
can
spin
it
up
locally
and
play
around
it
manually
so
that
there
could
be
some
interest.
D
So
it
needs
to
work
on
Mac
OS.
It
might
need
to
work
in
sigwin.
You
know
it
needs
to
work
on
all
of
our
development
machines.
C
J
D
Yes,
I
have
colleagues
who
use
sigwin
I've
I've,
gotten
feedback
from.
A
Okay,
with
that
I'm
gonna
see
if
Andy
is
available
for
his
Flakes
and
docs
updates.
I
Yeah
thanks
Nolan.
So
over
the
past
couple
weeks,
we've
spent
a
lot
of
effort
in
trying
to
deflate
the
test
failures
that
we
were
having,
and
there
have
been
some
major
bug
fixes
that
went
in
towards
the
end
of
last
week
that
hopefully,
will
significantly
reduce
the
amount
of
serious
test
failures
that
we
were
seeing
in
our
end-to-end
tests.
I
So
I
would
say
that
the
two
big
ones
were
an
issue
where,
if
you
had
multiple
workspaces
and
each
workspace
to
find
the
same
crd
in
terms
of
a
group
resource,
so
you
know
widgets.kcp.io,
but
they
served
different
versions.
Then
we
had
a
problem,
and
this
was
really
an
order
of
operations
thing
so
depending
on
which
test
ran
in
which
order
either
all
the
tests
would
pass,
or
they
wouldn't
and
that's,
been
fixed.
I
So,
if
you're
working
on
a
test
that
works
with
API
bindings,
for
example,
when
we're
in
a
multi-sharded
environment-
and
we
do
have
a
CI
setup
for
this-
the
works
or
exports
API
resource
schemas,
API
bindings
land
on
different
workspaces
when
we're
out
land
on
different
shards
and
different
workspaces
and
a
multi-shorted
test
setup,
and
there
is
some
increased
latency
in
replicating
the
resource
schemas
and
the
exports
to
the
cache
server,
which
is
then
used
by
various
controllers.
I
So
when
there's
this
increased
latency,
you
may
create
an
API
binding
and
expect
it
to
work
quickly,
but
it
might
not
because
of
the
time
to
replicate
data.
So
there's
a
lot
of
places
where
the
tests
need
to
be
more
verbose
in
terms
of
doing
the
eventually
blocks
on
trying
to
create
instances
of
API
bindings
starting
to
create
instances
of
Widgets
or
whatever.
So
if
you
are
working
on
a
test-
and
you
see
it
flaking,
please
feel
free
to
reach
out
on
slack
or
GitHub,
we'll
help
you
out.
I
But
if
you
look
in
test
grid
for
our
periodic
jobs
that
run
I
think
every
three
hours
just
against
the
main
branch.
It's
a
nice
wall
of
green,
which
is
really
good
to
see
so
that's
Flakes
and
then
for
documentation.
I
am
working
on
replacing
our
Hugo
based
stocks
with
make
docs
using
the
material
theme.
It's
a
lot
easier
to
maintain.
It
has
support
for
multiple
versions
easily,
so
I
know
Mike.
This
is
something
that
you've
been
looking
for
so
either
this
week
or
next
week,
hopefully
that'll
land
and
for
anybody
involved.
I
It's
in
my
opinion,
a
whole
lot
easier
to
work
with
so
I
have
a
preview
site
that
is
hold
on
I'm,
going
to
get
the
URL
I.
I
So
this
is
what
it
looks
like
I
have
done,
minimal
effort
in
terms
of
trying
to
style
it
I
have
the
same
blue
that
we
use
on
our
current
live
website.
I
have
the
logo
up
in
the
top
left.
I
It
does
support
multiple
languages.
So
if
anybody
wanted
to
do
translations,
we
have
the
ability
to
add
additional
languages.
You'll
see
the
link
to
the
GitHub
repo
in
the
top
right,
which
I
think
is
live
data
in
terms
of
the
latest
release
and
stars
and
whatnot.
It
has
search
built
in
and
it's
pretty
easy
to
work
with
so
I'm
I
have
a
couple
things
left
to
do,
I'm
trying
to
make
it
super
easy
so
that
you
can
test
out
serving
documentation
locally.
I
So,
to
avoid
asking
everybody
to
install
Python
and
dip
packages,
I'm
gonna
do
a
container
image
so
that
you
can
just
run
this
inside
of
your
favorite
container
runtime
and
then
once
I
do
that
I
need
to
get
the
GitHub
action
set
up
to
publish
when
we
push
updates
so
look
forward
to
this
and
if
anybody's
interested
in
helping
with
styling
or
content
or
whatever
once
we
get,
this
merged
definitely
welcome
any
contributions
and
that's
all
I
got.
A
And
regarding
this,
you
are
still
working
on
getting
the
actual
version,
support
right,
yeah,
I,.
I
Just
I've
only
in
this
preview
site
I've
only
published
the
one
from
Maine.
So
what
I
have
to
do
is
take
all
of
the
changes
that
I
made
to
the
main
branch
so
that
it
publishes
the
with
what
you're,
seeing
right
now
and
I
have
to
back
Port
those
to
the
release,
0.10
branch
and
once
I've
back
ported.
Those
I'll
run
the
same
commands
that
I
ran
to
generate
and
serve
the
main
branch
for
0.10
and
then
you'll
see
a
version
selector.
I
Yeah,
it
uses
the
GH
Pages
Branch,
so
it's
pretty
straightforward
in
terms
of
publishing
to
GH
pages
and
we
only
build
the
content
that
changes
as
needed.
So
if
we're
not
changing
0.10.
whatever
we
never
have
to
rebuild
that
content.
If
we
do
change
it,
we'll
rebuild
it
and
they're
distinct.
So,
like
the
main
branch
builds,
the
main
branch
and
so
on.
A
Sweet
and
thank
you
for
tracking
down
those
flakes
I
have
a
question
that
this
may
be
premature,
but
should
we
default
to
testing
ede
based
on
shards
versus
one
instance.
I
We
do
in
CI,
so
prowl
and
get
up
actions
run
I
believe
with
two
shards
today
and
the
most,
but
not
all,
of
the
ede
tests
will
create
workspaces
on
random
shards.
The
transparent
multi-cluster
tests
were
waiting
for
the
Sinker
view
to
support
being
backed
by
the
cache
server
that
merged
yesterday,
and
so
now
the
folks
working
on
TMC
can
update
the
EES
to
work
in
a
multi-shared
environment.
I
There
are
other
tests
that
we
specifically
place
workspaces
on
root,
Shard
or
on
a
specific
Shard,
but
those
are
tests.
Specific,
there's
a
reason
for
doing
so
in
general,
when
you're
writing
an
end-to-end
test
and
you're
creating
workspace
fixtures,
you
shouldn't
need
to
to
give
it
an
option
that
says
put
it
on
the
root
Shard.
But
if
you,
if
you
know
that
you
want
to
test
like
purposefully
test
that
workspace,
one
goes
on
Shard
one
and
workspace.
Two
goes
on
chart
two
there's
ways
to
do
that
as
well.
A
Cool
Mike,
you
had
your
hand
up
first
yeah.
D
I'm
sorry
I
just
realized.
Remember
it's
something
I
wanted
to
put
on
the
agenda.
The
the
plan
for
v0.11.
A
Okay,
let's
jump
to
David.
If
he's
got
something
on
the
same
topic
and
then
we
can
discuss
zero
11.
F
Yeah,
it
was
just
just
about
to
say
that
that
for
the
TMC,
for
you
know,
shotting
and
and
twin
tests,
there
is
there's
short
support
in
the
Sinker,
but
there
is
obviously
more
than
that
in
the
sense
that
we
have
to
revisit
mainly
the
TMC
related
fixtures
as
well,
because
the
Sinker,
you
know
you
have
to
to
point
to
the
right
scene,
Target
in
the
right
chart
as
well,
so
I
mean
mainly
what
was
done
for
the
API
exports.
F
You
know
when
you
have
some
helpers
in
the
end-to-end
tests
to
take
to
to
point
to
the
right
chart
and
get
the
objects
where,
where
you
should
you,
we
have
to
do
something
like
that
as
well.
For
all
the
TNC
related
feature,
many
The
Thinker
feature,
and
so
that's
probably
one
of
the
apart
from
from
the
singer
aspect.
That's
also
probably
one
of
the
reason
why
all
the
TMC
tests
would.
A
I
Yeah
so
I
we
tagged
alpha.0,
not
because
we
were
trying
to
release
it,
but
because
some
of
the
work
that
Vince
is
doing
in
controller
runtime
and
the
controller
runtime
example
repo.
It
was
easier
to
have
a
known
tag
instead
of
trying
to
go
off
a
get
Shaw,
so
I'm
not
aware
of
anything
off
the
top
of
my
head.
That
would
prevent
us
from
releasing
0.11
or
we
could
release
alpha.1
if
we
wanted
to.
I
I
Yeah
we
have
16
issues
that
are
in
the
zero
to
11
Milestone,
but
I,
don't
know,
I
mean
a
lot
of
them
are
sort
of
multi-release
issues.
F
Like
just
I
posted
one
here,
it's
a
pending
peer,
it's
probably
quite
easy
to
to
review.
I
didn't
have
time
to
look
into
it.
It's
related
to
you
know.
There
was
a
problem
when
you
do
a
bind
compute
and
it
was
not
correctly
choosing
between
the
path
or
the
cluster
name
aspect
of
it.
So
you
were
not
able
to
bind
in
some
cases
if
I
summarize
the
issue
correctly.
F
This
also
the
same
as
as
what
I
pointing
out
the
last
time
it
might
increase,
notably
the
you
know,
support
requests
in
the
in
the
community
Channel
just
because
it's
a
basic
bug,
but
that
requires
searching
everyone
each
time
everyone
hits
it.
B
I
I
mean
we
have
so
we
have
the
alpha
tag,
so
you
could
work
off
of
that.
If
that's
what
you're
looking
for
or
are
you
looking
for
something?
That's
more
I,
don't
know
publicly
announced
in
terms
of
hey
we're
ready
with
0.11.
D
I
think
it's
in
some
sense
a
multi-part
question,
so
the
issues
I'm
thinking
about
are
well
first
off
what
tags
are
imminent?
How
does
this
relate
to
the
rebase
onto
later
versions
of
kubernetes?
D
What
level
of
maturity
can
we
pers?
Should
we
move
our
edgemc
work
on
to
the
alpha
or
some
other
tag?
Now
these
are
the
kind
of
things
I'm
wondering.
I
So
the
rebase
is
an
easy
question
to
tackle.
First
that
will
come
after
we
release
0.11,
so
it'll
be
in
0.12
rebasing
to
kubernetes
126,
or
we
could
backport
it
or
do
it
in
a
zero
11
future
release.
If
we
wanted
to
I,
don't
see
any
reason
to
hold
off
transitioning
to
the
current
Alpha
zero
tag
or
a
final
tag
if
we
cut
one.
So
if
it's
easier,
we
can
just
cut
0.11.0
and
then,
for
you
know
additional
things
we
can
have
new
patch
releases
foreign.
D
I
would
imagine
that
you
wouldn't
want
big
changes
for
patch
releases.
I
wouldn't
put
a
rebase
in
a
patch
release,
but
you
said
not.
That
would
be
12..
So
let
me
just
ask
this
Alpha
zero
tag.
I
wasn't
aware
of
it.
How
new
was
it?
What
what
is
it
does
it
have
any
meaning
or
maturity.
I
It
literally
is
a
get
tag
that
we
created
so
that
the
controller
runtime
example
repository
could
use
it
in
the
go
module.
I
D
So
it
was,
it
would
be
if
we
use
it,
it
would
be
something
that
should
show
up
in
the
doc
site.
Obviously
it's
gonna.
It
includes
the
The
Logical
cluster
V3
yeah.
So
if
it's
going
to
show
up
in
the
doc
site,
I,
okay,
if
it's
not
going
to
show
up
on
the
dock
side,
I
would
want
to
hold
off.
I
It
will
I
or,
like
I,
said
with
the
docs
update,
it's
going
to
take
me
another
couple
of
days,
probably
just
to
work
through
the
remaining
tasks
that
I
have
and
then
I
can
get
it
published.
K
Hey
hi
everyone,
so
I
need
help
with
the
issue
so
currently
I'm
looking
to
I
mean
this.
Is
the
issue
currently
I'm
looking
to
gather
some
metrics
related
to
kcp,
so
our
team
is
primarily
interested
in
analyzinger
like
how
much
latency
is
required.
You
know
to
sync
all
the
Exorcist
from
kcp
workspace
to
or
AWS
cluster.
So
during
the
try
I
have
noticed,
there
are
only
like
few
metrics
and
I
tried
expanding
Sinker
logs,
wherein
I
can
see
some
of
the
latencies,
but
are
these
metrics
getting
published
somewhere
so
that
we
can?
K
Let
me
know
like
you
have
to
if
I
have
to
give
more
context.
B
F
At
least
for
me,
I
didn't
take
time
for
now
to
to
look
into
this,
so
I
I
should
probably
look
deeper
into
what
you're
doing
and
the
requirements
before
you
know.
Okay,.
K
Metrics
related
to
the
latencies
like
how
much
time
it's
taking
for
resources
to
get
synced
from
kcp
workspace
to
or
a
client
cluster
yeah.
Okay.
So,
okay,
can
you
point
out
me
through
the
right
links
so
that
you
know
I
can
go
through
them
in
detail.
H
F
H
F
Mean
I'm
not
I,
don't
think
the
logs
were,
you
know
designed,
or
you
know,
put
initially
with
this-
is
my
this
in
mind.
You
might
find
things
interesting
in
the
logs
to
to
paste
your
metrics
on.
On
the
other
hand,
maybe
we
we
could
think
how
to
integrate
metrics
into
the
Sinker
in
a
more
you
know,
native
way,
possibly
maybe
have
a
dedicated
session
for
that
and
brainstorm
on
that
could
be
useful.
K
Okay,
so,
as
of
now
like,
if
I
wanted
to
get
I
mean
any
metrics
related
to
kcp
like
what
would
be
the
right
thing
to
do,
like
you
know,
I,
what
we
are
trying
to
do
is
we
want
to
use
kcp
with
one
AWS
cluster
versus
typically
AWS
clusters
and
see
and
get
all
the
metrics
for
both
the
cases
and
see
like
how
they
are
performing.
You
know
just
to
see
if
we
find
any
Matrix
interesting,
you
know
to
analyze
it
from
the
performance
standpoint.
D
Okay,
yeah
I'm,
not
sure
I
stood
clearly.
Are
you
saying,
I
haven't
looked
but
I'm
a
big
fan
of
parenthesis,
metrics
I
I
would
assume.
Are
you
saying
the
synchron
right
now
doesn't
have
any
particularly
designed
Prometheus
metrics.
K
D
B
F
K
Yeah
I
mean,
apart
from
that
I'm
trying
to
understand
if
there
are
any
other
metrics
which
I
can
collect
so
yeah
I
understand
like
Sinker,
is
not
currently
publishing
metrics
to
Prometheus
in
order
plus
cluster
but
yeah
in
future.
It
might
like,
as
we
are
talking
about
the
roadmap,
but
apart
from
thinker,
do
you
have
any
do
we
have
any
other
metrics
that
we
can
use
for
analysis
purpose.
I
I
would
say,
rather
than
asking,
are
there
any
metrics
I
would
encourage
you
to
suggest
what
sort
of
metrics
you're
looking
for
and
then
we
can
go
add
them.
I
We
I've
experienced
with
other
projects
like
cluster
API,
for
example,
where
we
filed
issues
saying
we
should
add
metrics
and
nobody
could
really
point
to
what
metrics
would
be
useful
to
collect.
So,
if
you
are
looking
for
particular
and
for
pieces
of
information,
please
let
us
know,
and
we
can
either
get
them
at
it
or
if
you've
got
time,
that'd
be
cool
as
well.
Yeah.
K
Sure
I
mean
for
now.
My
team
is
primarily
interested
in
single
metrics,
but
as
I
mean
as
an
external
I
have
asked
for
the
other
metrics
as
well,
but
yeah
I'll
check
back
with
my
team
and
you
know
be
specific
on
what
metrics
we
are
interested
in
apart
from
singer.
A
I
also
mentioned
I,
don't
think
he's
on
the
call
anymore,
but
sergius
put
this
PR
in
for
getting
Prometheus
metrics
out
of
the
ede
tests,
but
I'm
honestly
not
sure
what
his
source
is,
so
that
this
PR
2774
may
be
worth
looking
at.
K
Yeah
I
took
a
look
at
that
pair
as
well.
I
tried
it
out
in
my
laptop
like
so
I
wasn't
able
to
see
any
promises
were
getting
published,
so
I
ran
the
tests
it's
Advanced
properly,
but.
H
K
So
here
is
the
thread
link
which
I
tried
out
the
pr
and
I
wasn't
able
to
see
in
Prometheus
data,
so
I
mean
I
got
another
suggestion
from
sir
I
mean
then
I.
K
Well,
I
mean
but
feel
free
to
suggest
me
like
if
you
are
able
to
see
any
problem
with
this
data
related
to
interest
in
20s
I'll.
A
It
looks
like
we
have
six
minutes
left.
Do
we
want
to
go
through?
Some
of
the
open
issues
do
put
triage.
J
I
Okay,
you
can
actually
just
click
on
the
link
and
it'll
open
up
to
the
side.
I
I
mean
if
it's
a
bug,
just
we'll
put
it
in
the
backlog,
yeah.
I
A
Documentation
looks
like
Drake.
Has
this.
I
Frederick,
are
you
working
on
that
now
or
we'll
work
on
it
in
the
future?.
B
On
that
I
I
was
basically
working
on.
H
B
Internet
test,
but
I'm
kind
of
blocked
there
because
of
I
was
asked
to
use
a
private
cluster
products.
But
it's
not
good.
B
And
my
id1
was
too
full
open,
so
there's
a
documentation
afterwards,
because
I
have
limited
time,
but
so
yeah
I
could
start
working
on
it.
It's
that
would
take
me
I'm,
not
sure
to
have
something
completely
Facebook.
B
Yeah
we,
we
still
have
the
announcements
proposal,
which
provides
some
level
of
documentation.
So
it's
not
like
that.
There
is
nothing
and
manifests
are
part
of,
and
it's
not
proposal.