►
From YouTube: 20191002 - 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
and
welcome
to
the
Wednesday
October
2nd
edition
of
the
cluster
API
office
hours,
a
sub-project
of
C
cluster
lifecycle,
just
a
reminder
that
this
meeting
does
adhere
to
the
kubernetes
community
code
of
conduct.
So
in
general
please
be
great
to
each
other.
And
if
you
do
want
to
comment
because
this
meeting
does
get
pretty
large,
please
do
use
the
raise
hand,
feature
and
zoom
to
avoid
people
talking
over
each
other.
A
B
So
the
first
one
is
promoting
some
documentation,
reviewers
motion,
if
you're
like
kind
of
volunteered
and
signed
up
to
be
our
like
hold
out
for
like
for
the
book
and
so
I
added
like
another
file
in
dot,
slash
book
with
their
name
in
it.
So
thank
you
so
much
for
stepping
up
and
if
you
also
would
like
to
contribute
in
this
area,
please
reach
out
we'll
be
able
to
make
promotion
processes
at
some
point
or
as
part
of
this
cycle
as
well.
B
B
C
D
B
Go
ahead,
please
go
ahead.
Okay,
so
you'd
shout
out
to
them's
like
have
like
push
like
the
the
cap.
G
work
to
kind
of
have
various
conformance
tests.
We
have
a
test
great
grid
dashboard
here
and,
as
you
can
see
like,
we
have
all
the
conformance
tests
running
for
cap
G,
so
in
the
GCP
providers
for
clothes
for
FBI.
As
far
as
I
know,
this
is
the
first
time
that
a
provider
in
college
API
is
like
like
they.
We
have
compartments
for
it
and
one,
for
example.
B
B
So
we
have
two
three
control
plane
in
tune
worker
nodes
all
running
in
TCP,
but
the
machine
side
of
things
two
of
the
control
plane
were
stuck
in
or
in
provisioning
phase,
and
this
kind
of
came
because
we
were
using
cube
idiom
from
from
master
and
it
kind
of
like
it
gave
a
signal
that,
like
something
was,
was
broken
in
the
inside,
which
I
believe
Chuck
is
also
fixing
today
or
there's
a
peer
out
already
so
working
on
it.
But
this
was
an
example
like
how
it's
great
to
have
end-to-end
tests.
B
E
B
One
in
in
scripts
is
called
CIN
e
treatise,
which
pretty
much
just
sets
up
the
environment
and
kind
of
like
run
the
end-to-end
conformance
tests
which
are
under
hack,
and
this
goes
and
like
sets
up
the
GCP
project
like
those
images
and
does
all
the
prerequisites
a
user
would
do
so.
It
like
really
test
the
whole
workflow
of
like
building
images,
creating
a
cluster
making
sure
the
cluster
is
up
and
kind
of
like
also
running
the
conformance
tests.
D
Alright,
the
one
thing
that
I
wanted
to
ask
here
for
people
to
think
about
is:
we
have
a
lot
of
things
that
have
variables
right.
What
is
the
version
of
kind?
What's
a
version
of
cluster
API,
you
know
boots
are
a
provider
for
cube
a
DM
GCP
image
builder,
the
cubed
M
itself,
then
the
cube
writers
that
is
installed
and
then
the
the
version
of
the
end
to
end
test.
So
what
are
the
combinations
of
things
that
we
should
set
up?
Ci,
for
you
know,
should
we
use
master
of
everything
or
you
know.
D
That
would
be
definitely
one
thing,
but
then
that's
more
likely
to
fail.
But
if
you
don't
have
it,
then
it
won't
catch
the
things
like
what
Vince
was
showing
before.
So
we
have
the
tools
we
have
the
scripts,
we
just
have
to
parameterize
it.
For,
however,
we
want
to
set
this
up.
This
is
just
the
start.
There
is
a
pre
submit
job
as
well
as
the
CI
jobs,
so
resubmit
is
currently
optional
as
in
it
runs,
but
it
is
not
a
required
job.
So
that's
the
way.
D
F
Yeah
so
Travis
and
I
even
talked
this
morning,
I
similarly
related
note
to
simplify
the
image
building
process,
to
be
just
a
single
script
that
is
callable
or
potentially
just
you
know,
make
your
life
everyone
who
wants
to
make
an
image,
make
their
life
easier.
So
in
looking
through
those
scripts,
there's
like
you
mentioned,
there
are
several
steps
that
can
be
automated
by
the
individual
component
or
e
to
make
the
overall
scripts
to
be
smaller
and
a
little
more
concise.
So
we
should
probably
make
I,
don't
know
exactly
we're
going
to
log
these
issues.
F
C
D
E
D
D
Right
so
on
the
slash
cluster
part,
you
know
we
have
to
do
some
research
on
how
we
would
integrate
this
with
cube
test
or
something
like
that.
That
would
be
the
thing
they
had
to
do,
but
then,
if
other
providers
have
an
interest
in
porting
this
to
their
own
cap
star,
you
know
that
would
be
a
interesting
thing
to
think
about
as
well
interested.
D
Absolutely
and
just
for
example,
the
Packer
thing
cannot
run
as
root
and
there's
a
bunch
of
code
in
there
to
create
a
fake
user
and
then
use
that
user
to
run
Packer.
So
if
we
are
able
to
move
that
kind
of
logic
into
the
image
builder
itself,
with
a
make
target,
that
would
make
this
script
really
smaller.
A
D
B
G
H
H
I
Sorry,
I
added
the
lower
in
the
doc,
but
I'll
say
here
so
that
for
a
cluster
upgrade
we
had
a
kickoff
meeting
yesterday
morning
and
the
goal
is:
will
solidify
the
goals
on
goals
and
motivations
by
Friday
of
this
week,
so
that
folks
can
review
them
on
Monday
and
agree
with
that
and
in
parallel
will
be
working
on
the
actual
implementation
details.
But
those
won't
be
ready
to
review
Wednesday
so
probably
tack
on
another
like
week
or
two
after
Wednesday
of
next
week.
C
So
last
time
we
spoke
this
time.
You've
really
been
fleshed
out,
but
listen
I've
been
doing
some
talking
and
thinking
and
writing
and
we
have
a
lot
of
stuff
in
the
link
there.
The
Google
Docs
link
that
I
think
is
ready
for
reviews.
I'd
love
to
have
some
folks
involvement
here
to
get
written
up
as
reviewers
in
the
kept
metadata,
and
then
we
can
open
the
pr.
After
there's
general
consensus
on
the
design.
B
H
B
C
A
B
Okay,
it
seems
like
we're
looking
at
all
be
peers
to
be
kind
of
at
the
end
of
the
month
to
be
out
and
which
leaves
us
with
like
a
few
months
of
iteration
on
the
possible
kind
of
action
items
on
those
yeah
yeah.
Something
I
just
wanted
to
make
sure
the
like
we're
all
aligned
on
the
on
the
same
page.
On
this
any
comments,
questions
concerns
okay,.
E
J
Don't
think
we're
on
any
definite
timeline
at
the
moment,
I
don't
think
we
need
any
particular
reviewers
at
the
moment
you
know
kind
of
when
this
thing
comes.
You
know
we're
cool
with
wherever
ends
up
landing
and
if
somebody
else
has
another
proposal
in
the
meantime
that
they
want
to
do,
you
know,
feel
free,
but
yeah
we're
definitely
planning
on
doing
this
cycle,
but
I
don't
have
a
definitive
timeline
of
it.
Okay,.
C
K
Also
see
a
mica
has
a
point
about
castle
to
Skyler
below,
so
we
should
probably
handle
them
together.
I
think
it's
like
two
positions
down,
so
we
had
enjoyed
by
the
way
I'm
one
of
the
maintenance
of
Casa
autoscaler.
We
had
a
discussion
on
the
sick-out
of
scaling
meeting
this
week
and
the
conclusion
was
that
we're
going
to
merge
the
pull
request
with
the
implementation
that
michael
has
linked
in
the
dog.
K
So
that's
one
thing:
there
is
just
like
some
minority
basing
stuff
and
I
think
there
is
a
discussion
on
V
1,
alpha
1
versus
V
1,
alpha
2
from
seek
out
the
scaling
point
of
view.
We
don't
really
have
enough
context
or
some
defense,
so,
whichever
one
you
decide,
I
think
we're
happy
to
go
with
and
another
point
I'd
like
to
race
here
is
I.
K
There
is
the
case
where,
basically,
the
crust,
API
controllers
and
crust
autoscaler
may
be
modifying
the
crust
at
the
same
time
leading
to
many
races,
especially
if
such
upgrades
used
silo
down
a
dog
with
list
of
issues
that
we
are
currently
aware
of,
and
and
it's
linked
here
and
basically
I'd
like
this
dog.
There
is
no
solution
there,
just
a
list
of
problems
and
I
like
this
dog
to
be
a
sort
of
starting
point
for
the
discussion
on
how
we
can
address
those
issues.
K
One
thing
I'd
like
to
finally
mention
at
the
end
is
I.
Don't
think
we
should
necessarily
just
say
because
I've
said
very
much
day,
two
issues
but
I,
don't
think
we
should
just
say
we'll.
Add
it
later,
because
I
think
that
sort
of
reason
well,
the
issues
are
coming
film,
is
a
difference
in
the
models
used
by
Krusty,
API
and
cross-sell
to
scale,
and
it
may
be
good
to
look
into
this
while
cross
the
PRS
in
alpha,
because
I
think
some
changes
to
cross
the
API
may
also
be
necessary
in
order
to
address
those
issues.
J
J
B
G
J
B
J
B
Know
I
would
agree
well,
the
foundations
are
good
and
we
want
to
build
upon
them.
It's
really
hard
to
say:
I,
guess
they
promised
B
went
on
for
one
and
throw
everyone
out
for
one
is
like
it's
completely
fair
to
sort
of
get
with
this
P.
Everyone
also
one-
and
you
mentioned
the
client
generated,
which
we
actually
don't
have
a
view
on
for
two
and
we're
not
planning
to
add
it
back
going
forward.
B
B
But
me
long
story
short
is
like
this
impacts,
a
machine
machine
set
of
machine
deployments
today
these
are
just
a
link
to
the
cluster
via
label
and
then
the
label
kind
of
sets
their
own
reference
back
to
the
cluster.
This
has
proven
linking
the
course
of
like
multiple
months
after
the
release
of
you
for
to
to
confuse
a
lot
of
folks
in
terms
of
like
UX
and
but
also
like
in
terms
of
debugging.
B
Moreover,
the
Machine
diploma
can
actually
function
without
a
cube,
config
secret.
The
you
config
secret
also
needs
a
cluster
name,
which
means
you
have
to
have
a
cluster
reference,
and
some
folks,
a
few
months
ago,
came
came
in
saying,
like
my
machine
deployment
is
naturally
scaling
up
and
everything
is
unavailable,
and
that
was
because
they
were
not
using
the
cluster
object.
They
were
just
using
the
machines
which
leads
to
conclusion,
so
the
general
feeling
around
this
like
to
make
everything
cluster.
That's
like
kind
of
opinionated
like
the
infrastructure
reference
optional.
B
So
you
know
you
don't
have
to
like
kind
of
create
the
AWS
cluster
or
like
whatever
you
provided
cluster
type,
if
you
don't
want
to,
but
at
the
same
time
make
sure
they're
like
the
core
types,
are
kind
of
like
a
little
bit
more.
Could
he
seven
like
they
actually
function
as
we
they
function,
because
there
are
assumptions
in
the
code
today
they
make
effectively
the
cluster
reference
required,
but
it's
not
written
anywhere.
M
Andy
thanks
I
wanted
to
emphasize
a
point
that
is
in
the
issue
that
Vince
has
up
on
the
screen
now,
which
is
that
the
cluster
object
does
not
really
have
to
have
anything
in
it.
Given
that
the
infrastructure
reference
is
optional,
so
it's
really
a
very
lightweight
rallying
point
to
associate
machines.
Machine
sets
and
machine
deployments
that
they
all
have.
They
all
point
to
a
singular
things
that
logically
represents
a
cluster.
So
hopefully
this
won't
be
super
contentious.
B
L
A
I,
don't
see
anything
so
what
at
what
I
would
suggest
is
is
if
we
are
looking
for
lazy
consensus.
We
probably
should
also
follow
up
with
a
item
on
the
mailing
list,
as
well
just
in
case
and
then
also
call
out
that
we're
gonna
do
consensus
on
the
github
issue
in
case
there
are
folks
that
are
trying
to
access
from
China
that
don't
necessarily
have
the
Google
Groups,
and
then
we
can
set
the
time
out.
J
A
J
J
J
N
G
So
we
just
cut
a
alpha
release
for
v30
of
cap
C,
which
is
the
V
1
alpha
2
stuff
I'm,
going
to
use
that
that
alpha
release
is
starting
point
to
clean
up
docks
and
the
release
process
and
make
sure
that
everything
is
vetted
properly,
we're
just
waiting
for
an
image
to
be
promoted,
but
that
is
available
when
the
docks
are
available.
I'll
send
something
out
wide
for
people
to
actually
kick
the
tires
on
it.
Super
excited
and
I'm
I'm
happy
that
everyone
worked
together
so
hard
on
it.
O
Yeah,
firstly,
thank
you
for
the
reviews
on
the
PR
I'm
just
here
representing
the
PR
for
Seth
I
want
to
make
sure
that
there's
any
other
changes
needed
other
than
the
bar
search
to
a
constant
I,
see
Chuck's
comment
now
on
the
Google,
but
I
believe
suss
should
have
time
to
said
this
morning
to
fix
it.
If
not,
I
can
for
sure
commit
to
fix
that,
but
yeah.
C
A
M
Right,
yes
of
didn't,
survivors,
cheering,
please
stop
and
I'll
go
ahead
and
flip
over
okay,
all
righty.
So
we
have
three
issues
for
cluster
API
with
no
milestone
and
two
for
the
cube
am
bootstrapper.
First
one
is:
could
we
commit
the
vendored
code
into
a
vendor
folder
in
the
cluster,
a
repository
which
we
used
to
do
and
stopped
doing
with
the
move,
to
be
one
alpha,
two
I
like
not
having
the
vendor
code
committed,
but
I
can
certainly
understand.
M
A
B
Given
that
the
vendor
folder
and
most
of
like
a
client
could
keep
it
removed,
it
really
simplifies
people
forgetting
please
like
update,
bender
or
like
having
a
whole
pipeline
to
checks
in
sync,
the
the
issue.
It
seems
like
more
of
a
user
problem
around
like
I'm,
not
having
a
working
proxy
which
will
solve
this
issue,
so
I'm
planning
to
push
back
here.
F
There's
a
bunch
of
github
is
down
scenarios,
there's
also
a
bunch
of
aircraft
environment
scenarios
where
the
repository
being
self
contained
is
a
useful
feature.
That's
kind
of
the
reason
why
most
of
the
repositories
and
key
proper
do
you
have
wonderful.
There
is
just
so
that
if
you,
if
you
get
a
fork
of
the
repository,
it's
a
self
contained
thing.
E
L
A
lot
of
I,
just
I,
think
this
particular
person
may
be
behind
a
country
firewall,
so
it
that
might
be
the
scenario
here
in
terms
of
not
being
able
to
have
a
working
proxy.
So
that
does
feel
like
us.
A
showcase
of
that
we
should
be
aware
of
I
was
wondering
where
they
can
spin.
It
have
the
vendor
in
a
different
branch
or
something
I
might
play
with
that.
A
So
I'm
also
kind
of
in
the
same
boat
as
just
in
there,
because
the
issue
here
is
this
is
one
of
our
Chinese
contributors,
so
they
may
be
chasing
working,
go
proxies
that
are
accessible
through
the
Chinese
firewall
and
I.
Don't
necessarily
know
that
asking
users
to
manage
their
own
go
proxy
might
be.
That
might
be
a
little
much
tasked
to
contribute.
So
since
we
are
trying
to
be
inclusive
of
the
community,
we
may
want
to
consider
that,
above
you
know,
just
standard
developer
experience,
David,
yeah,.
P
M
So
justin
has
volunteered
to
look
into
this,
and
that
is
a
very
valid
point
David,
because
we
have
similar
restrictions
or
requirements
so
we'll
check
under
that
on
our
end
too,
from
a
milestone
perspective,
I
think
at
this
point,
I'm
inclined
to
put
it
in
next,
while
Justin
does
some
research
and
then
when
we
have
some
more
information,
we
can
either
put
it
in
the
appropriate
milestone
because
we're
going
to
do
it
or
we'll
close
it.
Whatever
the
resolution
is.
M
B
B
Q
M
Q
Had
a
question
on
the
caps
that
folks
have
been
working
on,
will
there
be?
Will
all
the
caps
result
in
issues
being
created
in
the
repo
for
people
to
pick
and
work
on
if
I'm
asking,
because
I
might
not
be
able
to
participate
in
the
discussions
and
like,
but
I
would
love
to
contribute
like
in
whatever
way
I
can
so
I
wanted
to
know
if
issues
will
be
created
for
all
the
caps
yeah.
M
M
H
I
noticed
that
there's
documentation
for
creating
clusters,
but
not
but
kind
of
missing
documentation
for
cluster
CTL
delete
I
was
wondering
what's
the
community's
take
on
like
actually
adding
more
documentation
for
a
tool
that
is
kind
of
on
the
path
to
application,
but
then
there's
also
of
conversation
of
converting
plasticity
of
the
cluster
ADM,
just
kind
of
want
to
understand,
like
the
viability
of
cluster
CTR
on
the
future
and
contributing
more
work
towards
that.
Thank
you.
F
Too
soon
I
know,
Fabrizio
is
currently
working
unless
you
can
assign
it
to
me
for
us
to
at
least
to
triage
for
this
cycle,
but
I
don't
think
we
are
going
to
try
to
get
something
in
place.
What
we
call
it
is
TBD,
but
for
me
she
was
doing
a
lot
of
the
work
with
a
lot
of
feedback
from
other
folks
and
we'll
probably
have
a
demo
and
some
other
things
in
the
future.
F
M
From
my
perspective,
to
answer
Warren's
question
I
think
that,
because
we
don't
have
a
replacement
for
a
cluster
cuddle
that
if
somebody
wants
to
write
some
documentation
on
things
like
cluster
cuddle,
delete
that
it's
worth
it.
Even
though
we
know
that
the
tool
will
be
updated
in
the
future,
but
I
think
more
documentation.
Now
is
better
than
less
personally.
M
M
M
F
C
C
But
the
thing
is,
we
have
all
the
pieces
in
place
to
do
this
right.
We
have
a
run
scripts
before
Covidien,
which
can
go
fetch
binaries
from
somewhere
and
put
the
braid
you
want
them
to
be
so,
like
I
see
all
the
pieces
there,
I
don't
know.
If
the
request
here
is
for
a
change
to
the
API
to
make
this
easy
or.
A
What
yeah
I
think
the
challenge
here
is:
how
easy
do
we
actually
want
to
make
this
considering?
This
is
also
something
specifically
meant
for
testing
environment,
so
we
may
want
to
potentially
even
choose
another
route
and
just
have
a
helper
for
generating
the
more
verbose
config
that's
possible
today.