►
From YouTube: 2020-02-26 - 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
Today
is
February
26
2020.
This
is
the
clustered
API
office
hours
meeting
cluster
API
is
a
sub-project
of
CID
cluster
lifecycle
and
this
meeting
is
being
recorded
and
will
be
posted
to
YouTube
in
a
little
while
we
do
have
meeting
a
ticket
for
our
meetings,
which
is
basically
let's
all
be
kind
to
each
other,
and
please
use
the
raise
hand
feature
of
zoom
if
you'd
like
to
participate
in
the
discussion
within
this
document.
A
Please
add
your
name
to
the
attending
list
and
we
do
have
our
agenda
for
this
meeting,
which
I'll
be
going
through
in
a
minute.
If
you
have
additional
topics
that
you'd
like
to
add
or
PSAs
or
demos,
please
add
them
in
the
right
place
in
the
agenda
and
we'll
move
on
to
the
first
item
here.
So
we
like
to
welcome
any
new
attendees.
C
If
you're
not
talking
to,
please
mute
yourself
and
if
you
like
to
get
anything
prioritized
for
today,
like
feel
free
to
reach
out
and
any
slack
or
just
in
on
a
peer
review
or
an
issue,
the
tentative
release
date
for
0-3.
We
slipped
it
like
for
a
few
days
from
the
6th
to
March
10th,
which
is
the
Tuesday.
After
mostly
because
the
6th
was
a
Friday
and
we're
trying
to
do
friday,
reviews
Lisa's
going
forward.
C
C
Okay
regarding
the
reviewer
list,
I
just
wanted
to
kind
of
like
reach
out
to
the
group
and
like
if
anyone
is
interested
to
become
a
reviewer,
please
do
reach
out
like
we
can
add
you
to
the
reviewer
list,
which
you
can
see
here.
Thanks,
Andy
CCO
has
expressed
interest
to
work
more
upstream.
I
should
have
done
a
great
job
and
as
a
provider,
opening
an
issue
and
like
also
like
to
looking
at
at
some
of
the
PRS
in
copy,
so
I
think
we
yeah.
We
can
write
this.
D
D
The
other
thing
that
I
had
was
we're
doing
another
batch
of
cluster
API
t-shirts.
So
if
you
haven't
already
gotten
one
of
them,
it's
going
to
be
the
three
stack
turtle
design.
Please
click
that
link
and
we'll
get
one
out
to
you
as
soon
as
we
can.
That
said,
if
you
have
already
gotten
some
I
am
double
checking
that
list
against
lists
that
we
previously
fulfilled
from
so
we'll
only
be
sending
them
out.
If
you.
A
C
Sure,
usually
like
it
would
be
great
like
if
someone
it's
like
more
familiar
with
the
codebase,
you
don't
have
to
be
familiar
with
every
part
of
the
code
basic
given
its
kind
of
large,
but
sometimes
we
just
put
people
on
the
review
list
if,
like,
if
they're
like
even
have
a
little
knowledge
in
one
particular
area,
for
example,
close
Ricardo
is
our
CLI
tool.
If
you
know
really
well,
you
might
want
to
like
do
more
reviews
in
there
and
we
need
reviewers
in
all
areas,
especially
at
like,
for
example,
so
Docs
so
documentation.
C
Our
book
needs
reviewer
list.
After,
if
you
want
to
become
a
reviewer,
we
usually
ask
you
like
to
get
first
familiar
then
start
reviewing
on
your
own.
You
can
you
can
review
it
without
like
any
permissions.
So
that's.
That's.
That's
a
nice
to
have
and
then
after
it
like
a
while,
we'll
just
attitude
the
reviewer
list
and
the
bot
will
out
mighty
thick
reviewers
based
on
that
list.
C
D
I
was
just
gonna
say
if
you
are
also
interested
in
only
reviewing
a
subset
of
the
codebase,
it's
possible
for
us
to
make
accommodations
for
that
as
well.
We
can
leverage
the
per
directory
owners
files
so
that
you
would
only
be
tagged
for
reviews
on
specific
parts
of
the
code
base
rather
than
anywhere
yeah.
E
Yes
and
that
helps
actually
Prakash
here,
the
cluster
cuddle
I
have
been
dealing
with
the
integration
with
airship,
as
well
as
with
the
bare
metal
operator
in
the
metal
cubes.
So
I
think
this
is
a
great
place
to
start
I.
Believe
I
am
a
beginner,
you
can
say
as
far
as
kubernetes
is
concerned.
Certainly
I
would
like
to
review
and
I
will
need
some
help.
I
wiII
come
to
you,
but
I
think
you
can
add
me
and
I
do.
A
To
put
review
comments
on
pull
requests,
and
we
welcome
anybody
and
everybody
to
come
in
and
review
if
you
think
that
you
can
add
useful
comments
so
just
because
the
bot
isn't
assigning,
you
doesn't
mean
that
you
can't
write
comments
at
the
same
time,
if
there's
lots
and
lots
of
people
even
two
or
three
people
who
have
already
had
significant
review
time
and
a
lot
of
comments
on
a
pull
request.
There
can
be
times
where
coming
in
as
an
additional
reviewer
on
a
pull
request,
may
just
be
making
things
a
little
bit
confusing.
A
So
you
have
to
be
a
little
bit
aware
of
when
it's
okay
or
maybe
not
as
appropriate
to
review
but
in
summary,
feel
free
to
sign
up
for
notifications
for
pull
requests
and
issues
on
the
repository.
If
that's
something
you
can
do
all
that
yourself
and
github
feel
free
to
periodically
come
through
and
look
at
what
pr's
are
out
there
to
get
a
sense
for
what
changes
people
are
putting
in
and
like
I
said.
We
welcome
comments.
Yeah.
E
That
helps
certainly
it
will
take
some
time
learning
time
and
will
be,
or
at
least
the
new
one.
Like
me,
you
or
him
at
some
point
of
time,
so
I
will
seek
your
help,
but
I
will
try
comment
even
whether
you
appoint
me,
reviewer
or
not,
I
will
definitely
go
ahead
and
take
a
code
review
at
some
of
the
level
where
I
can
contact.
Thank
you.
A
All
right
last
PSA
here
was
that
yesterday
we
had
a
two
hour
marathon
API
review
session
with
Jordan
Liggett
from
Sega
cam
Sheri.
We
got
a
lot
of
good
pointers
on
things
to
look
out
for
with
our
v1
Elfa
to
be
one
out
of
three
changes
and
thank
you
Vince
for
filing
issues
related
to
most
of
those
and
thanks
everybody
for
attending
and
helping
take
notes.
I
thought
it
was
a
great
session.
A
F
Yep,
so
I
have
been
asked
to
look
at
doing
some
sports
and
stuff
and
in
particular
bringing
up
a
CA
ap,
say
a
EP.
However,
you
pronounce
that
yeah
for
the
spice
and
stuff
so
I
raise
a
couple
of
questions.
Are
they
in
the
stock
channel
and
Vince
got
back
to
me
and
suggested
that
we
bring
up
here?
I?
Think
the
thing
is
at
the
moment,
because
we've
already
done
a
proposed,
an
open
shift.
We've
already
got
a
phase
substantial
proposal.
A
So
the
the
flow
that
we
usually
like
to
follow
is
that
the
first
step
would
be
a
Google
Doc
for
collaborative
editing
and
commenting,
and
then,
if
we
feel
like
the
majority
of
the
comments,
are
resolved
or
non-blocking,
take
the
content
and
move
it
to
a
pull
request.
I
will
copy
out
that
by
saying
that
we
are
a
very
chatty
and
argumentative.
A
So
we
can
get
to
a
point
where
we
feel
like
the
Google
Doc
is
in
a
good
spot
and
then
open
up
the
pull
request
and
have
an
entirely
new
round
of
comments
and
I.
Don't
think
that
that's
anything
that
anybody
necessarily
wants
it's
frustrating
to
the
author
or
authors
of
the
proposal,
as
well
as
the
community
in
general.
So
I,
don't
know
that
there's
any
specific
improvements
I
have
on
that
process,
but
I
think
that's
probably
the
best
way
to
go,
especially
if
you've
already
got
a
fairly
involved
proposal.
F
A
F
F
I
guess
the
only
difference
is
whether
you
only
want
to
support
it
in
one
or
the
other
and
but
I'm
the
way,
I've
sort
of
imagined
it
is
there'll,
be
in
the
the
infrastructure
template
for
the
Machine
anyway,
so
kind
of
be
I've
either
one
will
work.
You
just
have
to
set
the
options
in
the
in
a
template.
C
Maybe
we
can
I
think
they
do
have
like
one
implementation
for
Michigan
for
in
Azure.
So
we
can
probably
look
at
that
together
later,
but
there
is
no.
It
sounds
like
because
the
proposal
process
like
kind
of
Lindsay
using
like
a
current
abstraction,
will
make
that
much
faster
and
we
can
probably
avoid
like
also
like,
if
we
have
done
in
the
past,
which
usually
takes
a
lot
of
time.
F
F
C
A
G
Hey
Andy,
can
you
hear
me?
Yes,
okay,
cool
yeah
I
just
wanted
to
bring
back
with
this
topic,
make
sure
this
concern
Suzanne
and
support
from
the
community
unaware
of
more
than
a
few
parties.
Now
you
seen
either
the
scallop
er
that
we
put
in
place
or
the
sing
approach
for
the
skin
with
the
cleft
API
I,
believe
both
communities
would
benefit
by
having
a
single
place
to
to
collaborating
in
a
common
solution
and
also
a
good
favorite
option
for
both
class
TBI
and
of
the
scaler.
G
So
I
can't
really
see
a
reason
for
this
not
to
to
happen
at
this
time.
We
are
in
a
good
position
to
make
the
efforts
to
get
the
the
PR
smart
and
also
we'll
be
happy
to
commit
ourselves
to
to
its
maintenance,
so
I
just
like
to
to
propose
a
battle
to
move
forward
which
would
be
syncing
up
with
the
autoscaler
Falls
and
get
the
PR
match.
G
Then
we
put
a
follow
up
here
right
away
to
support
any
version
of
the
cluster
API
that
we
could
easily
achieve
that,
and
there
is
already
a
work-in-progress
PR
by
using
dynamic,
informants
and
and
just
configuring
the
group,
personal
and
kind
as
an
argument,
or
also
so
so.
Yeah
I
just
wanted
to
make
sure
that
this
makes
sense
for
everyone
here
and
make
sure
that
we
are
aligned
and
that's
the
case.
We
will
make
the
remaining
work
to
make
it.
A
Thanks
so
in
terms
there's
a
few
things
here
in
terms
of
the
autoscaler
pull
request:
I,
don't
think,
hopefully,
I
speak
for
everybody.
I,
don't
think
anybody
is
against
this
going
forward.
Ultimately,
it's
up
to
the
PR
author
and
the
autoscaler
folks
to
see
it
across
through
the
finish
line
for
the
Machine
autoscaler
CRD
proposal.
I
haven't
had
a
chance
to
look
at
this.
A
I
saw
it
come
through,
but
you
know
I
think
we
talked
I
believe
at
the
last
meeting
that
we
definitely
are
interested
in
seeing
this
go
forward
at
some
point
in
the
future
as
well.
So
maybe
we
just
need
to
spin
off
a
few
folks
to
you
know,
set
up
like
a
mini
little
working
group
to
work
on
this
together
over
time,
but
yeah
I
mean
I,
think
we're
in
favor
of
it
Jason
yeah,
you're,
gonna,
yep,.
D
So
the
first
question
I
have
is:
is
the
post
v1
alpha
one
work
that
you
mentioned
lburrows,
that
is
that
also
going
to
include
support
for
machine
pools
as
well.
G
Hey
Jason,
so
did
you
look
at
the
code
at
the
moment?
It
doesn't
include
support
for
machine
pools
like
it
could
have
an
interface
node
that
we
call
scalable
resource.
Well,
you
could
fairly
say
forward
to
plugging
machine
pools
or
any
other
abstraction
that
we
had
in
the
future.
So
should
be
easy
to
support
that
in
the
future
cuz.
It
doesn't
make
sense
to
you.
Yeah.
D
I
think
that
makes
sense,
and
the
other
thing
is,
is
I
definitely
support
the
idea
of
a
like
a
mini
working
group
to
kind
of
plot
out
the
next
steps.
I
know
that
in
the
past
we
dealt
with
kind
of
the
impedance
mismatch
between
how
the
cluster
autoscaler
expects
to
interoperate
with
resources
and
how
cluster
api
exposes
those
resources
and
I'm
really
interested
in
seeing
what
we
can
do.
H
Yeah
so
I
just
wanted
to
say,
like
I
made
some
comments
on
the
autoscaler
PR
and
I'm
gonna
try
to
start
attending
those
sing,
all
those
scalar
meetings
to
help
I,
guess
kind
of
Lea's
or
bridging
the
gap
between
you
know
what
we're
doing
here
and
what
they're
doing
just
to
see.
If
we
can.
Oh,
you
know
what
it's
gonna
take
to
get
that
merged
over
there
and
then
like
what
we
need
to
do
to
coordinate
that
cool.
A
Thanks,
so
I
would
suggest
that
people
who
are
interested
in
working
together
on
this
self
organize
and
figure
out
a
time
to
sync
up
whether
you
put
your
names
in
the
agenda
document
here
or
some
other
way,
I'll
leave
it
to
you
to
organize
so
Mike
and
Jason.
Do
you
all
have
any
thoughts
on
how
you
want
to
do
that.
H
C
C
C
The
idea
is
that,
like
the
types
that
we
have
today,
you
copy
those
shouldn't
change
that
often
so
we
want
to
do.
Is
that,
like
we
want
to
have
these
type
like
work
on
the
side
so
that
they
then
target
cluster
API
types,
so
yeah,
like
I,
just
wanted
to
put
this
out
there
and,
if
you're
interested
in
putting
experimental
features,
WeChat
thanks.
D
Vince
Jason,
so
I
just
wanted
to
add
that
the
current
PR
for
feature
gates
is
had
solved
the
problem
for
functional
kind
of
features,
but
has
not
solved
some
of
the
outstanding
questions
that
we
have
around
the
actual
CRD
types,
and
things
like
that
that
we
may
have
as
part
of
one
of
these
features.
So
we
still
have
to
figure
out.
You
know
how
exactly
the
deployment
and
distribution
of
those
experimental,
C
RDS
is
to
be
accomplished.
I
Andy
yeah
I
hope
to
run
into
you
again
actually.
So
somebody
just
pinged
me
about
this
awesome
proposal
by
sedessa
vos
seems
pretty
recent
and
I
just
wanted
to
jump
on
to
I
realized.
I
hadn't
talked
about
the
work
that
cluster
addons
has
been
doing
in
this
similar
area
and
I
really
love
the
overlap,
and
so
the
link
to
the
proposal.
That's
on
the
cluster
API
side
that
goes
through
the
pr's
Google
document,
as
well
as
the
POC
PR.
That's
it
I've
put
up
and
I'd
like
to
just
kind
of
start.
I
Any
open
conversation
about
composing
and
sharing
some
struction
code
and
I
see
that
Michaels
writing
some
notes.
Thanks
for
helping
out
so
the
on
the
cluster
add-on
side,
we
have
an
API,
it's
a
component
config
style,
but
we
could
also
publish
a
custom
resource
for
it
if
you'd
like
to
store
it
in
the
cluster
and
that
has
this
add-on,
installer
configuration.
I
Oh,
the
links
got
a
little
bit
copied
here
there
we
are
I,
think
that's
the
same
thing,
so
the
add-on
installer
configuration
basically
just
provides
a
collection
of
groups
of
things
that
are
supposed
to
be
bundled
together
and
fetched
or
bundled
in
some
way,
and
the
idea
with
this
API
is
people
have
various
mechanisms
of
differing
levels
of
maturity,
to
bundle,
expand,
inflate
and
patch
things.
Customize
is
like
a
very
upstream
mechanism
for
doing
that
kind
of
thing,
and
then
we
have
like
coop
cuddle
apply
of
manifests
like
over
HTTP.
I
We
have
several
files
and
directories.
We
have
even
more
ways
of
inflating
things
using
JSON
njk
and
all
that,
and
what's
like
the
minimal
thing
that
we
can
start
out
with,
that
will
be
like
useful
in
the
80%
case,
and
then
how
can
we
formalize
the
grouping
of
those
things
together
as
a
collection
of
things
that
are
versioned
so
yeah?
I
This
API
is
currently
proposed
for
kuba
D,
hum
I'm
gonna
talk
with
Justin,
we've
talked
informally
about
doing
it
with
cops
as
well
as
ETS
cuddle
outside
of
kuben
high
score,
so
it'd
be
great.
If
we
could
find
a
little
bit
of
overlap
if
things
are
not
serving
the
use
case
properly,
what
can
we
do
to
change
things
so
that
we're
doing
it
in
the
same
way.
A
Thanks
Lee
I
know
we
talked
a
little
bit
about
this
in
one
of
the
past
couple
of
close
to
API
meetings
and
I
know.
Justin
was
here
for
at
least
one
of
those
discussions
and
we
wanted
to
make
sure
that
we
don't
duplicate
what
the
add-ons
working
group
is
implementing
and
I
think
we
probably
need
to
find
a
distinction
between
what
we
do
with
Saddam's
proposal
and
what
that
turns
into
and
what
which
all
are
doing
in
add-ons
so
Tim
I
see
you've
got
your
end.
Oh
yeah.
J
That
was
always
actually.
My
intention
is
that,
right
now
there
is
more
than
just
that.
There's
a
bunch
of
stuff
people
would
want
to
apply
for
policies
or
for
whatever
reason
you
can
imagine
for
every
cluster
that
they
create.
But
until
we
get
like
a
lot
of
mileage
and
I
want
to
get
a
lot
more
mileage
on
you
from
the
unit's
management
in
place
that
we
can,
we
can
just
use
this
mechanism
to
as
an
example
of
how
to
deploy
it.
I
J
K
B
I
L
E
E
Where
we
deal
with
the
clusters
and
four
clusters,
we
approach
there
is
the
airship
shuttle
come
on
and
there
is
a
cluster
or
whatever
we
have
cluster
cuddle
and
in
the
with
respect
to
the
two,
there
are
two
one
is
the
target
called
ephemeral
and
the
other
target
is
other.
This
label
is
called
target,
so
we
do
have
two
of
them
and
something
to
do
with
bare
metal.
E
I
wanted
to
have
some
type
cluster
type
being
bare
metal,
because
we
have
to
deal
with
metal
cube
that
is
metal,
three
dot
IO
from
cluster
cutting
from
a
ship
to
cluster
cuttle
to
bare
metal.
So
it's
some
kind
of
a
chain
which
is
not
clear
to
me,
but
I
want
to
see.
Is
there
any
cluster
type
called
that
metal.
A
M
E
E
A
The
the
data
model
for
cluster
API
is
there
is
a
type
called
cluster.
It
is
a
kind
in
the
kubernetes
api
parlance.
There
is
no
distinction
of
a
cluster
of
one
type
or
another.
It's
just
a
cluster,
and
every
cluster
has
a
field
in
its
spec
for
an
infrastructure
reference.
You
can
have
that
refer
to
a
specific
infrastructure
providers,
kind
for
a
cluster.
E
L
E
M
Michael
yeah,
if
we
could
just
get
everyone
to
stop
using
the
term
bare
metal
at
all,
because
there's
there's
no
bare
metal,
there's
some
provisioning
service
that
it
made
provision
we're,
but
there's
a
service
in
there,
whether
it's
some
law,
your
own
PC
service
or
some
kind
of
OpenStack
ironic
service.
There
is
some
service
there
in
calling
every
every
time
somebody
wants
to
provision
to
met
middle.
A
bare-metal
provider
is
terrible,
so
you
should
refer
to
things
as
so
like
the
middle
three
provider
should
be
the
middle
three
provider.
E
A
C
A
C
K
B
C
Last
updated
field
or
machine
status
is
not
populated.
We're
discussing
this
before.
Thank
you,
Joe
for
finding
this.
There
is
a
field
on
machine
status.
That's
called
a
stub.
Theta
is
actually
never
updated
and
I
don't
even
know
if,
like
we
updated
imbuing
offer
to
that
needs
to
be
checked.
So
we're
discussing
like
a
few
implementations,
one
would
be
to
just
said
last
updated
when
the
face
changes.
This
is
a
smaller
field.
The
other
one
would
be
kind
of
like
modified
to
help
her
path
to
understand.
C
F
C
F
Wouldn't
say
it's
blocking
it,
but
part
of
the
functionality.
Part
of
the
check
won't
work.
So
there's
like
a
little
block,
that's
like.
If
this
field
is
not
nil,
then
we
can
short-circuit.
If
a
node
never
comes
up,
it's
only
a
very
minor
part
of
it
and
it
doesn't
block
it
to
say
it
just
means
that
bit
isn't
working.
C
C
D
N
C
C
So
let's
try
to
get
this
merge
today
in
terms
our
reviews.
If
you're
interested
in
doing
some
reviews,
we
have
the
closed
circle
dropping
complete
plans
if
contract
changes
is
a
good
one
to
review
and
the
cost
for
pebble
into
and
that's
framework
both
of
these
need
more
reviewers
and
I.
Think
that's
it.
We
don't
have
any.