►
From YouTube: 20190624 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
Hello
today
is
Wednesday
June,
26
2019.
This
is
the
standard
cluster
API
office
hours
discussion.
Just
as
a
general
reminder,
we
have
a
quota
kind
of
policy
which
is
basically
sums
up
to
you,
don't
be
a
jerk.
There
is
inside
of
our
document.
We
have
because
this
meeting
gets
awfully
large.
We
often
recommend
that
people
use
the
raise
hand
functionally
just
because
that
allows
me
the
moderator
to
try
and
give
equal
time
to
anybody
who
raises
their
hand.
A
Otherwise
it
can
have
a
tendency
to
generate
so
with
all
that
preamble
in
mind,
we
have
a
if
you
want
to
guide
your
D
to
the
doc.
That
would
be
highly
useful
and
then
official,
the
other
people
who
are
doing
the
play
at
home
game.
If
you
have
topics
that
you'd
like
to
discuss
that
currently
are
in
the
list,
feel
free
to
amend
a
pen
to
the
end
of
agenda
before
the
issue
triage
area
and
let's
get
this
party
started,
that's
you
have
the
first.
B
Topic,
yes,
well:
I
didn't
notice
that
like
get
up,
package
registry
was
enabled
for
kubernetes
board,
and
so
it
just
like
kind
of
like
more
heads-up
like
if
we
start
like
looking
into
it.
If
the
community
will
be
fine
to
kind
of
be
using
this
going
forward
and
before
he
went
out
for
two
releases
to
tell
me
more
about
github
packet
registry
I
know
nothing
about
it.
It
just
like
a
docker
registry
built
into
github.
It
seems
that
like
gerber
neighs
or
is
enabling
and
I
can
all
be
both
TPD.
B
B
C
I
we
just
did
their
working
group,
kids,
infra
or
Nia.
Today,
I
have
an
hour
ago.
It
ended
there.
There
will
be.
We
have
unlocked
buckets
at
this
point.
If
you
send
a
PR,
you
can
get
staging
buckets.
There
will
be
a
promoter
process.
It
is
not
in
place
yet,
so
we
don't
yet
have
promotion
from
those
staging
brackets
to
the
official
Cates,
GCR
ones.
C
The
challenge,
I
believe,
will
be
that
that
promoter
process
as
currently
written
does
not
support
anything
but
GC
r
is
the
source.
I
do
not
personally
think
it'd
be
too
hard
to
add
another
source.
It's
like
it
doesn't
really
matter
where
the
staging
buckets
are
right,
but
the
final
promotion
I
think
we
should
have
it
be
under
the
criminalize
infrastructure,
and
today
that's
gonna
be
GC
r,
but
in
future,
should
you
help
us
other
things,
but
yeah
we
would.
C
D
B
Alright,
so
the
peeler
is
actually
linked
in
the
in
the
agenda,
but
I'll
link
it
in
chat
as
well.
B
So
the
TLDR
here
is
that,
while
printing
control,
the
runtime
and
control
tools
to
the
second
version,
which
brought
us
like
a
lot
of
breaking
changes,
we
actually
noticed
that
it's
kind
of
like
unsupported
or
like
issue
it
shouldn't
be
done
that,
like
we
embed,
object
meta
from
the
metope
one
package
into
like
other
fields,
so
this
VR
hasn't
been
back
for
a
year
to
be
one
up
for
one
because
kind
of
like
its
brings
a
breaking
change.
That
said
like
if
you
open
the
pure
at
him.
B
Actually,
the
new
object
is
a
shrink
down
version
of
what
you
were
supposed
to
set
on
an
object,
meta
field,
so
the
original
authoring
matter
actually
had
a
lot
of
flow
that
were
read-only
and
the
validation
was
actually
done
by
the
API
server.
So
it's
an
object
that
has
like
a
lot
of
special
things
in
it.
Given
that
it's
in
the
middle
you
want
package.
So
this
was
just
like
an
effort
to
capture
the
fields
that
you
actually
are.
A
B
A
So
that
gets
a
little
weird
with
the
innards
of
API
machinery.
Is
this
step?
Do
you
sync
with
like
someone
from
API
machinery
to
verify
that
this
is
what
their
practices
for
all
the
inner
goo
that
exists,
because
there's
there's
tons
of
innards
to
the
to
the
API
server
that
does
type
checking
and
it
does
like
everything
comes
from
object,
meta
and
all
of
that
stuff
does
first
checks
for
changes
and
everything
else
at
the.
F
F
There
is
a
an
embedded
nested
metadata
field
that
is
not
at
the
root
level,
and
so
generally,
we
don't
want
to
keep
this
going
forward
and
we
have
ways
to
work
around
it
for
you
1,
alpha
2,
but
for
B
1,
alpha
1,
assuming
we
want
to
keep
V
1
alpha
1
in
master,
while
we're
working
on
V
1
alpha
2,
then
we
may
need
to
address
this.
Although
I
think
Vince
with
the
split
into
two
different
API
groups,
does
this
go
away
or
is
it
still
problem?
No,
it's
still
a
problem.
B
F
B
F
B
B
A
Great
next
up,
just
as
a
PSA
that
I
I
talked
briefly
during
grooming,
that
I
wanted
to
collect
all
the
details
with
regards
to
cluster
cuddle.
So
I've
opened
up
an
umbrella
issue
about
like
what
are
what
does
cluster
cuddles
job?
What
are
the
roles
and
responsibilities
of
these
different
objects
and
right
now
I'm,
just
kind
of
soliciting
feedback
from
user
stories
for
multiple
people?
So
if
you
have
thoughts,
questions
comments,
complaints
concerns,
please
add
that
to
the
issue.
The
issue
has
grown
pretty
long,
so
I'm
gonna
request
at
this
point.
A
What
and
one
of
the
problems
we
face
is
how
do
we
do
a
unified
way
of
managing
the
lifecycle?
For
multiple
providers
and
one
thought
there,
it
is
to
think
about
the
idea
of
creating
an
operator
for
cluster
API
that
could
ride
across
providers,
but
that
Africa
operator
could
also
serve
several
responsibilities
so
right
now
it's
it
would
allow
us
to
get
rid
of
quest
rebuttal.
But
again,
this
is
an
idea
factor
at
this
point.
So
if
you
have
ideas,
please
put
them
in
the
issue
and
we
can
we
can
discuss
them.
D
Right
next
up,
Jason
yeah
we've
had
quite
a
few
changes
land
recently
that
have
been
back
for
generator
to
release
your
about
one
branch
and
I
think
we're
about
ready
for
zero
dot,
one
dot
for
release.
Now
we
have
some
updates
to
new
events
for
the
Machine
sets
and
machine
deployments,
as
well
as
the
remote
node
references.
F
F
G
F
Do
whatever
provider
you
you're
using
will
need
to
support
a
new
convention
that
we
established
for
remote
node
references?
So
it's
it's
a
two-part
thing.
The
first
part
is
that
when
your
cluster
is
created,
the
the
provider
is
responsible
for
creating
a
cute
config
secret
that
contains
a
cute
config
with
a
specific
name
naming
convention
and
then
cluster
API
is
zero.
Point
one
point:
four:
when
we
release
it
we'll
be
able
to
set
the
node
reference
for
you
and
get
the
status
of
the
node
to
make
sure
that
the
Machine
set
replica
counts
are
updated.
G
D
David's
not
able
to
make
it
today,
but
I
told
him
I'd
bring
up
this
topic
for
him.
He
just
wanted
to
highlight
and
make
sure
that
folks
were
aware
that
master
the
node
graph
changes,
the
node
graph
controller
has
landed
and
the
old
node
controller
is
now
in
a
deprecated
status,
and
he
also
wanted
to
make
sure
that,
as
we
introduced
breaking
changes
into
master,
that
we
also
have
associated
documentation
land
for
those
changes
in
the
gift
book
as
well.
F
D
I
think
there's
two
portions
of
this,
because
the
documentation
right
now
is
geared
towards
provider
implement
provider
implementers
today.
So,
ideally,
we
would
have
documentation
available
for
how
to
consume
what's
in
master,
as
we
start
gearing
towards
the
next
release.
But
we
also
probably
have
to
establish
some
type
of
strategy
for
having
released
versions
documented
separately
for
a
master
as
well.
B
Vince
and
then
I'm
gonna
call
Pierce
afterwards,
I
was
just
going
to
say
just
to
echo
I
think
the
human
Dacians
should
for
now
at
least
point
to
B
1
alpha
1
whenever
possible.
Given
the
demand
for
2
is
not
a
new
consumable
state,
there
is
a
question
that
I
had
is
like:
how
can
we
support
like
a
both
version
at
the
same
time?
Think
wait
for
David
for
that,
but
for
now,
I
would
prefer
that
if
the
dogs
like
actually
refer
to
a
stable
version,.
B
E
D
No
get
raises
ready.
Yep
next
topic
was
is
when
we
cut
releases,
we
should
use
annotated
tags
rather
than
the
lightweight
tags
like
we've
been
doing
to
date,
and
we
should
also
make
sure
that
the
versions
that
were
tagging
or
preface
prefixed
with
the
V
to
better
align
with
go
module
support
is,
is
that
in
the
release
documentation,
it
is
not.
I
was
planning
on
going
ahead
and
submitting
a
PR
for
that
today.
Yeah.
D
Other
than
that
next
topic,
I
just
wanted
to
highlight
again
that
breaking
changing
changes
are
starting
to
land
in
master.
If
you
are
consuming
the
closer
API
repository
for
some
reason
or
another,
specifically,
if
you're
creating
a
downstream
cluster
API
provider,
you
should
definitely
be
venturing
from
a
specific
tag
or
using
the
release
here
about
one
branch
and
not
been
during
it
master.
C
D
I,
don't
think
it's
a
huge
deal
right
now,
since
most
of
the
providers
right
now
we're
still
using
death.
I
think
the
vSphere
provider
is
the
only
one
using
go
modules
today,
but
when
we
switch
to
V
1
alpha
2
and
the
new
Q
builder
stuff,
I,
think
more
of
the
providers
are
going
to
be
using
go
module
so
to
be
more
important
going
forward.
Okay,.
C
We
can't
I
said
later,
if
we
just
like,
we
need
to
and
yeah,
although
currently
I
do
the
tag
and
all
of
that
I
don't
think.
There's
any
reason.
I
have
to
do
that.
If
you
wanted
to
drive
that
I
could
be
your
your
cloud
builder,
your
human
cloud
builder,
and
only
do
the
make
and
the
push
type
of
things
if
you
can
create
a
tag
to
release
yeah.
D
F
I
just
wanted
to
mention
that
we
have
a
new
repository
in
github
for
the
medium
bootstrap
provider
doesn't
have
any
code
yet,
but
we
will
be
using
this
repository
to
pad
code
for
for
Cuba
and
bootstrapping,
based
on
the
proposal
that
was
merged
last
week.
So
if
you're
looking
for
it,
it's
right
there
and
as
we
get
the
V
1
alpha,
2
types
PR
merged.
Hopefully
soon
we
should
be
able
to
start
working
on
the
fixed
route
provider
as
well.
So
if
you're
interested,
please
come
file
issues
or
find
us
on
site.
A
A
C
A
A
H
B
A
Maybe
we
should
live
in
teams.
I
do
think
we
should
try
to
get
this
done.
I
don't
want
this
like
in
lower
bids.
We.
C
Can
certainly
so
we
could
one
thing
we
could
do
is
we
can
put
that
the
docker
file
I
guess
to
generate
that
image
into
cluster
API?
For
now,
we
can
today
request
a
cluster
API
staging
repo
and
I
think
we
can
run
it
from
that
station
Rico
if
need
be.
I.
H
C
A
A
Had
there
was
an
issue
that
we
logged
as
part
of
the
code,
walkthrough
I
didn't
actually
say
it
earlier,
but
shout
out
to
Vince
for
doing
the
Code
walkthrough,
hopefully
focused
on
the
useful.
Apparently
somebody
has
set
a
PR
already
for
update
for
some
of
the
items
in
the
glossary
that
we
had
that
we
notice
we're
missing
this.
We
were
talking
about
things
walking
through
cope
with
the
global
eval
at
API
groups.
It
looks
like
somebody
had
already
triaged
this.
A
C
A
C
A
H
H
Respect
this
different
configuration
and
to
and
to
what
extent
and
and
how
you
know,
how
do
we
satisfy
the
goal
like
the
two
goals,
on
the
one
hand,
to
allow
providers
to
be
opinionated,
for
example,
to
you
know
not
you
know
to
enforce,
maybe
some
paint
or
to
you
know,
to
prevent
some
paint
from
being
applied.
On
the
other
hand,
to
you
know
make
that
transparent
to
the
end
user.
You
know
when
you're
using
a
particular
provider.
H
A
A
A
A
I
I
Yeah
I
can
talk
about
this
briefly.
Essentially,
you
know
you
could
have
say
100
machines
and
all
in
various
states
and
it's
kind
of
an
exercise
for
the
user
to
figure
out.
What's
gonna
be
successful
and
what's
not-
and
you
don't
have
anything
today
that
tries
to
inform
the
user
that
hey
this
thing's
been
going
for
a
while,
and
you
probably
need
to
look
into
it.
I
So
for
me,
the
most
logical
place
to
do
that
in
today's
model
and
the
V
1
alpha
2
miles,
probably
the
Machine
controller
and
basically
have
some
confederal
times
where
it
detects
set
if
it
hasn't
progressed
past
a
certain
point
for
whatever
reason
to
to
send
up
some
signal,
my
preference
would
be
labels
because
you
can
filter
on
labels,
and
this
is
an
exercise
for
the
administrator
to
quickly
identify
things
that
are
not
working
and
CR
D's.
Don't
currently
support
filtering
on
other
fields.
So
that's
what
this
is
about.
J
Hi
yeah
I'm
a
little
bit
worried
about
putting
this
in
the
machine
controller,
because
some
aspects
of
the
timeouts
may
be
provider
specific
and
so
I
would
just
want
to
make
sure
that
the
provider
is
able
to
say
yeah.
This
is
still
going
on
I'm
still
working
on
doing
whatever
it
is
that
it's
supposed
to
be
doing
and
not
make
assumptions
based
solely
on
time.
A
J
F
I
wrote
this
in
a
comment
down
below,
but
I
think
that
it's
important
to
think
about
this
from
a
requirements
perspective
and
not
an
implementation
perspective
initially.
So
if
we
can
come
up
with
use
cases
and
specific
examples
like
what
you
just
said,
Doug
about
the
interactions
between
an
infrastructure
provider
and
cluster
API,
proper
I
think
trying
to
look
at
it.
That
way
will
help
us
figure
out
where
it
makes
the
most
sense
to
implement
this.
I
Michael
yeah
I
mean
I'm
happy
to
implement
it
in
the
infrastructure
provider
itself.
I
mean
that's
kind
of.
If
you
kind
of
look
at
today's
model,
that's
machine
controller
of
your
structure,
provider
or
the
same
but
I
think
there's
room
for
doing
the
machine
controller.
In
any
case,
if
we
make
it
a
tunable
thing,
for
instance,
I
know
bare-metal
previously
said
well
hey
this
might
you
know
take
an
hour
for
us
to
actually
provision
something
so
I
think
you
know
having
a
tunable
would
kind
of
help,
support
that.
G
Just
a
just
a
general
comment
or
or
if
I
may
ask
for
a
favor,
it
sometimes
gets
will
be
it's
a
confusing
call
if
there
are
like
to
Michael's
or
to
Jason's
or
to
like
other
person
names,
maybe
just
like
maybe
first
letter
of
the
last
name
or
something
or
I'll
be
fine
was
Michael
B,
that's
that's
cool,
but
that
just
every
single
time
that
right
now
was
Michael
and
I.
Go
me!
No!
No,
not
me!
If
that's
possible
sure
I
can
I
can
try
to
do
that.
Thank
you.
A
I
A
A
There
are
no
more
issues
to
triage
for
those
who
are
doing
the
ad-lib
coat.
Do
you
want
to
alpha
two
grows
there?
Also,
there
are
a
number
of
issues,
that's
inside
of
you
and
alpha
2
that
are
either
marked
as
good
first
issue
or
have
Help
Wanted
labels
attached
to
them.
So
if
you
are
interested
in
helping
out,
please
feel
free.
That
would
be
very
highly
beneficial.
As
another
note
just
to
reiterate,
for
those
who
are
new,
there
is
a
separate
grooming
dock
which
I
can
paste
in
the
cluster
API
channel
afterwards.