►
From YouTube: 20200722 Cluster API Office Hours
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
Thank
you
all
right
welcome
everyone.
This
today
is
july,
22nd
2020,
and
this
is
the
cluster
api
office
hours.
Cluster
api
is
a
subproject
of
sig
cluster
lifecycle.
A
A
Okay,
it
doesn't
look
like
anyone
is
unmuting,
so
I'll
keep
going.
I
think
everyone
here
has
been
here
before.
Okay,
so,
let's
start
with
psas
vince
preparing
for
yeah.
B
Yeah,
I
just
just
said,
like
we
found
a
few
bugs
in
zero
through
seven
expected,
but
one
of
them
kind
of
that
I
put
here,
is
kind
of
important
because
he
could
roll
out
your
cubitium
control
plane.
So
if
you
have
a
cluster
configuration
that
it,
there
was
nil
before
which
it
could
happen
that
you
have
danielle,
because
you
don't
have
to
set
any
fields.
B
The
kcp
would
be
stuck
in
the
loop
like
because
he
cannot
compare
it
wasn't
able
to
compare
before
like
a
nil
cluster
configuration,
because
the
jason
and
marshall
were
always
at
martial
it
and
it
would
still
be
an
empty
struct
in
that
case
and
nil
and
an
empty
structure
were
not
equal,
so
you
always
say
like.
Oh,
I
have
changes
and
I
have
to
roll
out,
so
this
was
causing
infinite
rollout
and
thank
you,
the
folks
that
kind
of
brought
this
up
last
week.
So
this
has
been
merged.
B
It
has
been
fixed,
probably
do
either
release,
probably
maybe
next
week
and
yeah,
the
other
one.
B
C
Andy,
not
a
question
just
a
comment
with
the
first
one
about
the
kcp
roll
out
being
a
problem.
There
is
a
workaround.
I
don't
remember
if
you
said
that
a
minute
ago,
events,
but
you
can
just
set
the
cluster
configuration
to
an
empty
struct
and
that
should
fix
it.
A
Good
call
out
thanks,
andy
awesome,
all
right,
so
I
don't
see
any
demos.
A
A
So
if
you
want
to
go
ahead
and
take
a
look
at
those
and
add
feedback
go
for
it
for
now,
let's
move
to
discussion
topics
and
I
think
there's
just
one
for
now.
You
went
off
before
planning
who
wants
to
take
that
one.
C
I'll
start,
so
we
have
a
lot
of
issues
that
are
just
in
the
repo
there's
over
180
right
now,
and
I'm
sure
that
several
of
you
have
probably
thought
about
things
that
you'd
like
to
see
in
the
next
api
version
around
restructuring
things
and
any
other
sort
of
breaking
changes
that
we
might
want
to
do
so
we're
probably
going
to
do
some
backlog
grooming
sessions.
C
We
don't
have
any
dates
in
mind
at
this
point,
but
soon
and
we'll
make
sure
that
we
announce
them
with
enough
lead
time
and
we
recognize
that
time
zones
are
hard.
So,
if
you're
not
able
to
make
one
of
the
grooming
sessions,
then
we'll
definitely
make
sure
that
we
have
a
way
to
have
some
async
feedback
as
well.
And
if
you
do
have
anything
that
you're
interested
in
seeing
cluster
api
do
in
the
future
and
there's
not
currently
an
issue
for
it.
C
Please
file
one,
and
if
anyone
has
any
suggestions
around
how
to
make
this
as
inclusive
as
possible
in
terms
of
timing
and
async
feedback,
would
love
to
hear
your
suggestions
either
now
or
offline.
A
Thanks
cool
anyone
have
anything
to
add
to
that
or
questions
comments.
A
Okay,
I
actually
have
a
question
have
any
of
the
infrastructure
providers
if
anyone's
here
have
has
anyone
started
thinking
about
v1
alpha
4
like
breaking
changes
for
specific
to
the
provider
apis.
C
And
yeah,
so
for
kappa,
this
is
really
vague
right
now,
but
we're
looking
at
potentially
using
a
an
external
or
separate
repository
that
can
manage
aws
resources
and
that
may
involve
some
changes
to
the
data
model.
I
think
that
repo
is
in
active
development
or
planning
or
somewhere
in
between,
and
so
it's
not
something
we
can
really
consume
just
yet,
but
we're
excited
about
it
because
it'll
move
a
lot
of
our
code
out
and
we'll
just
manage
the
aws
resources
declaratively.
C
But
again
I
don't
know
exactly
what
that'll
do
to
the
kappa
crds
themselves,
but
I
imagine
some
stuff
might
move.
A
Around
got
it
yeah.
We
have
a
similar
proposal
in
flight
for
cabsie
as
well:
cool,
okay,
yeah.
I
guess
it
will
become
more
concrete
when
we
get
closer
to
the
date
and
once
we
get
the
ball
rolling
for
cappy
itself,
but
hopefully
we
can
get
some
time
like
margin
after
the
cappy
api
changes
to
get
provider
api
changes
in
so
that
we
don't
like
release
right
after,
like
we
did
for
v1
alpha
3.
I
think
it
was
a
bit
too
tight
on
the
timeline.
A
Okay,
anyone
else
have
any
thing
about
this
or
anything
else.
Anything
you'd
like
to
discuss
any
general
questions,
if
you
don't
want
to
say
them
out
loud,
feel
free
to
message
me
your
message
on
the
chat
and
I
can
speak
for.
D
D
So
I
didn't
know
that,
because
I
am
I'm
coming
from
a
different
world
of
open
stack
where
we
use
git
review
and
get
done
with
that
here
I
had
to
deal
with
something
on
the
just
two
lines
of
code
change:
okay,
it's
not
a
major
thing,
but
the
problem
I
faced
were
like
this.
So
first
I
pulled
from
the
main
source
rather
than
going
to
my
github
then
ran
into
some
cla
licensing
issue.
D
E
D
A
So
if
you're
generally,
it's
it's
better
to
try
to
have
smaller
pull
requests
to
facilitate
reviews.
So
it's
better
if
you're
able
to
scope
each
change
to
one
pull
request,
but
you
can
have
multiple
commits
purple
requests
if
you're
doing
related
changes,
that
kind
of
depend
on
each
other
and
are
relatively
small.
But
if
you
want
to
post
in
slack
I'm
happy
to
pair
with
you
later
or
I'm
sure
someone
else
can
help
you
get
through
this
and
yeah.
We
can
figure
this
out,
but
okay.
D
D
A
A
A
The
most
of
the
providers
have
their
own
slack
channel
since
they
have
their
own
repository,
so
yeah,
okay,
I
appreciate
it
all
right,
yeah
thanks
for
bringing
it
up.
Oh
and
warren
posted
the
exact
name
of
the
channel
thanks
warren.
A
Okay,
any
other
questions,
if
not
we'll
move
on
to
the
triage.
A
Okay,
so
switch
screens.
A
No
milestones,
I
think
most
of
these
have
muslims
yeah.
So
just
a
few
here
this
looks
like
a
bug.
It's
pretty
recent
upgrade
using
kcp
hangs
with
cat
d.
Does
anyone
start
looking
into
this
okay?
Fabricio
said
he
can
repro
sounds
like
we're
waiting
for
more
evidence.
A
D
F
D
A
Let
me
follow
up
with
you
after
the
meeting-
okay,
let's
just
get
to
the
triage
for
now,
because
I
can't
look
at
the
pr
right
now.
I'm
sharing
my.
D
A
But
I'll
follow
up:
okay,
okay,
all
right!
So
in
terms
of
milestones
for
this
one,
I
don't.
B
A
Yeah,
okay,
great,
I
don't
think
I
actually
have
this
permission
yet
because,
oh
I
think
I
opened
the
pr.
A
F
One
yeah,
so
I
think
there's
a
bunch
of
discussion
regarding
cluster
carl's
sub
commands
in
general,
and
this,
let
me
see,
let
me
just
get
context
again
on
this.
F
Oh
yeah,
so
this
is
essentially
moving
commands
that
were
under
the
config
sub
command
to
a
new
sort
of
sub
command.
Name
called
generate
to
make
it.
I
guess
just
improve
the
user
experience.
F
There's
another
issue
that
I
had
opened
up
just
before
the
meeting
that
would
also
address
some
cluster
cuddle
user
experience
issues.
So
if
anybody
has
any
comments
on
why
we
should
not
go
down
this
path
of
sort
of
breaking
up
a
cluster
cuddle,
config
cluster
and
cluster
kettle
config
provider
to
generate
the
yamls,
please
put
that
information
in
the
in
the
on
the
issue.
F
Just
to
note
what
this
specifically
one
of
the
specific
things
this
is
doing
is
the
cluster
cuddle
generate
provider,
it's
going
to
be
taking
out,
or
rather
moving
functionality
from
cluster
cuddle,
config
provider
dash
o
yaml.
So
the
thing
that
the
command
that
actually
prints
out
the
yaml-
this
is
basically
going
to
be
extracting
that
into
its
own
sub
command.
The
previous
command
of
cluster
cuddle
config
provider,
where
it
sort
of
pretty
prints
the
the
information
that
will
still
remain
those
won't
be
removed.
B
One
just
just
one
comment:
yes
well.
This
is
probably
nothing
for
zero
four,
so
like
in
terms
of
milestone,
that's
probably
where
it
fits
given
that,
like
I
I
have
proposed
this
leaked
war
and
and
for
britain
like
last
week,
is
mostly
because
right
now,
it's
very
confusing
but
convict
us
sometimes
like
prints
out
stuff,
sometimes
actually
configures
things
and
subject.
This
is
just
like
a
way
to
improve
the
user
experience
and,
first
time
experience
when.
F
A
B
No,
I
would
say
like
so
to
get
the
current
configuration
for
a
provider
like
we
have
the
get
verb,
so
we
can
do
closer
call
get
providers
because
I
think
config
does
that
too
at
some
at
some
point
and
that's
like
where
the
kind
of
where
this
comes
from
and
then
generate
would
actually
generate
that
configuration.
A
Okay,
yeah.
That
makes
more
sense
because
I
think
config
never
actually
applies
anything.
It
always
just
generates,
and
it
just
depends
like
on
how
yeah
okay
got
it
so
so
this
was
actually
yeah
it
weren't
already
what
this
should
be
b1
alpha
four:
what
do
we
think
about
having
some
sort
of
mechanism
like
a
label
or
a
milestone
where
we
can
flag
issues
that
we're
opening
as
like?
We
should
consider
this
for
v1
alpha
4,
to
make
the
grooming
easier
when
we
come
to.
That
is
that,
instead
of
putting
everything
in
milestone
next.
B
A
B
Issues
they
should
be
able
for
api
changes
or
braking
change
or
something
so
I
can
look
and
put
it
in
okay.
B
C
Yeah,
so
this
was,
as
I
wrote,
just
putting
out
some
feelers
to
see
if
people
had
interest
in
doing
this.
So
it's
definitely
a
milestone.
Next,
and
I
know
jason
had
put
some
comments
about.
There
might
be
some
issues
here.
So
I'd
say
at
this
point:
it's
mostly
just
an
fyi.
Please
take
a
look
if
it
seems
like
a
good
thing
or
a
bad
thing.
C
A
All
right-
and
I
think
this
is
the
one
that
michael
just
added
michael,
do
you
want
to
talk
about
that?
One.
E
Sure,
well,
I've
been
working
on
some
things
that
utilize
lifecycle
hooks,
but
it's
probably
more
broadly
applicable.
Basically,
there's
there's
no
indication
on
a
machine
set
object.
When
the
last
time
the
replica
accounts
were
updated,
and
so,
if
some
other
component
wants
to
rely
on
that
information
being
up
to
date,
we
don't
really
have
a
way
to
do
that.
So
if
the
machine
set
is,
is
a
broken
and
those
counts
aren't
getting
updated.
Other
components
might
make
the
wrong
decision.
Based
on
that.
B
E
Yeah,
like
you
know
like
I,
can,
I
can
scale
a
machine
set
down
if
the
machine
set
controller's
broken
for
whatever
reason.
E
E
E
B
That
wouldn't
change
like
the
at
least
the
observed
generation,
because
the
spec
was
stayed
the
same.
Maybe
this
is
like
a
good,
a
good
use
case
for
a
condition
if
we
wanted
to
go
down
that
path
instead
of
adding
more
stuff
to
the
the
status
but
yeah
like
either
way,
this
could
be
done
in
zero.
Three.
A
No
that's
awkward.
Okay,.
A
All
right,
so
I
think
that
was
all
for
the
issues
that
were
not
already
milestones.
A
A
B
What
yeah,
I
think,
that's
what
andy
robert
mentioned
before
or
along
those
lines
like
we
will
have
to
like
go
into
the
the
backlogs
that
we
have
and
like
see
what
we
want
to
actually
do
in
zero
3x
and
what
we
don't
want
to
do
with
zero
3x
and
put
it
into
zero.
Four
okay,
I
would
say,
like
we
probably
have
another
quarter
of
bug,
fixes,
like
you
know,
just
maintenance
releases,
and
in
august
after
we
do
planning,
we
do
the
first
breaking
change.
B
Probably
we'll
do
the
in
the
main
branch
so
we'll
move
to
the
new
control
runtime,
for
example,
create
the
alpha
four
types
and
then
have
a
release:
zero,
three
branch,
so
that
that's
where
we
start
only
to
backboard
bug,
fixes
and
not
like
add
any.
A
Weeks,
okay
sounds
good,
just
a
heads
up
that
the
tilt
customized
substitutions
are
currently
broken
and
I
think
andy
has
a
pr
out
to
fix
it.
C
A
Okay:
first
gotcha:
okay,
thanks
so
yeah
all
right,
anything
else
that
anyone
would
like
to
add.
A
All
right
have
a
good
one,
everyone
and
see
y'all
on
slack.