►
From YouTube: 2020-02-19 - 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,
this
is
Wednesday
February,
19th
2020.
This
is
the
cluster
API
office
hours.
Meeting
clustered
API
is
a
sub-project
of
CID
cluster
lifecycle.
This
meeting
is
being
recorded
and
will
be
posted
on
YouTube
later
and
we
do
have
meeting
etiquette,
which
is
basically,
let's
all
be
nice
to
each
other,
and
if
you
do
have
something
that
you
want
to
discuss,
please
add
it
to
the
agenda
down
below
in
this
document
and
please
use
the
raise
hand
feature
in
zoom
and
I
will
do
my
best
to
keep
track
of
that.
A
Also,
please
add
your
names
to
the
attending
list
and
let's
move
on
to
the
agenda.
First
thing
we
like
to
do
is:
give
new
attendees
a
chance
to
say
hello
if
you're
interested
in
doing
that
and
you're
new.
So
if
there
is
anybody
where
this
is
your
first
time
or
first
couple
of
times
that
you
want
to
say
hi,
please
feel
free
to
do
so
now.
B
C
D
A
A
G
So
yeah
we
have
too
many
PSAs
for
today.
Well,
actually,
it's
really
I'm
gonna
say
that
like
we'd
have
our
c0
that
has
been
released.
This
is
mostly
for
developers.
People
they're
like
a
very
familiar
with
the
code
base
or
like
they
want
to
try
the
bleeding
edge
of
technology
feel
free
to
use
the
new
version
of
closure
API.
G
There
have
been
some
issues
and
we're
trying
to
cut
our
c1
today
and
move
to
weekly
cadence
for
our
C's
until
we
get
to
release
what
you
should
be
like,
probably
like
first
week
of
March
or
second,
that's
still
TBD,
but
for
now
like
every
Wednesday
like
we,
you
might
expect
a
new
RC
release
and
that's
like
we
have
no
changes
which
really
go
for
infrastructure
providers
and
booster
providers
and
any
other
provider
control
plane.
We
have
two
really
important
changes
that
went
in
the
first.
G
One
is
C
or
D
D
1,
C,
I,
D
b1
is
the
new
GA
version
that
came
in
with
116,
which
also
dictates
like
the
minimum
version
for
your
management
cluster.
You
can
see
the
cluster
KP
or,
if
you
want
to
look
at
what
changes
need
to
be
made
to
update
your
provider.
The
other
one
is
for
the
multi-tenancy
issue
we
discussed
this
last
week.
G
There
were
a
lot
of
changes
that
went
in,
and
these
are
kind
of
like
very
intensive,
and
if
you
want
to
open
up
the
the
link
you
can
probably
show,
but
the
customize
directory
has
been
pretty
much
refactor
from
scratch
and
to
support
most
I
can
see
when
you
use
web
box.
So
the
way
we're
doing
this
now
is
that
Web
books
like
are
now
in
the
Kaveri
web
box
system
namespace,
and
this
namespace
will
also
like
have
Web
books
for
providers.
G
This
is
probably
related
to
these
the
customize
changes.
So
if
you're
using
the
framework
to
run
your
tests,
you
definitely
want
to
use
the
RC
0
as
your
module,
which
then
will
build
the
customize
from
the
correct
directory.
Otherwise,
I
don't
think
it
will
actually
work,
but
yeah
we
can.
We
can
debug
later
offline,
but
I
think
that's
that's
a
good
first
step.
H
G
Yeah,
if
checks
on
the
colic,
maybe
you
can
speak
to
that,
but
I
think
we
did
fix
that
for
the
in
the
best
framework.
G
For
KP
k
zero
one
six,
so
thanks
for
reminding
me,
so
this
is
a
new
release
that
we
had
to
do
for
B
1,
alpha
2
providers
and
any
other
provider
out
there.
That's
not
like
doing
this
change
if
we
actually
have
to
do
a
new
release
before
you
can
update
to
2
or
3
in
more
details,
we
found
an
issue
with
here:
the
V
1
beta
1,
where
the
unknown
fields
were
preserved
by
default.
G
G
G
This
will
disable
it,
and
then
this
will
allow
you
to
actually
operate
to
be
one
on
for
three.
If
your
provider
does
this,
you
would
have
to
first
update
release
a
new
version,
and
then
these
this
will
be
the
new
required
version
before
you
upgrade
to
be
mastered,
see
two
hands
raised
and
drew
+
Seth.
I
A
G
A
A
I
So
when
the
cube,
idiom
control,
plane
provider
now
so
basically
I
was
reviewing
the
Jason's
upgrade
pr4
qbm
control,
plane
and
I
kind
of
started
digging
into
how
changes
are
detected
and
it
uses
a
hash
function.
The
hashing
function
behind
the
scenes
against
a
couple
of
fields
in
the
queue
ATM
control
planes
spec
right
now
it
specifically
looks
at
the
kubernetes
version
and
the
reference
for
the
infrastructure,
template
and
I
kind
of
raised.
The
question
like:
would
there
be
any
issue
with
adding
the
cube
idiom
control
planes,
cube,
ATM,
config,
specs,
join
configuration
I.
I
Think
I
said
that
right
to
the
list
of
fields
in
that
hashing
function,
so
basically
things
that
would
trigger
a
quote
unquote
upgrade
would
be
changing
the
infrastructure
reference.
So
in
the
example
of
like
Kappa
say
you
want
to
change
your
ec2
instance
that
would
trigger
an
upgrade
changing.
The
kubernetes
version
would
trigger
an
upgrade
and
now,
if
I
say,
I
wanted
to
add
a
cubelet
argument
to
my
control
plane
machines.
I
That
would
also
be
something
that
could
trigger
an
upgrade.
This
is
super
important
for
our
use
case,
because
we
consider
machine
replacement
to
be
the
true
way
to
achieve
these
kinds
of
changes.
We're
not
super
interested
in
trying
to
like
do
anything
on
the
live
control.
Plane
machines
like
if
the
queue
medium
control,
plane
object,
is
going
to
replace
machines.
I,
it's
perfectly
reasonable
to
have.
It
also
be
able
to
affect
flags,
and
things
like
that.
J
A
J
So
I
can
say
when
we
were
crafting
up
the
original
proposal
for
the
Canadian
control
plane.
We
had
initially
said
that
we
wanted
to
try
to
make
that
use
case
out
of
scope
and
and
the
big
reason
why
we
wanted
to
scoop
that
out
was
basically
we
wanted
to
avoid
creating
a
situation.
That's
too
easy
for
somebody
to
shoot
themselves
in
the
foot.
That
said,
as
we've
nailed
down
the
implementation
for
the
control
plane,
a
lot
more.
J
We've
been
building
in
a
lot
of
safeguards
so
that
we
won't
do
any
type
of
mutation
of
the
control
plane
if
everything
isn't
healthy
to
start
with,
so
we
won't
upgrade
unless
the
health
checks
pass.
We
won't,
you
know
scale
up
if
the.
If
the
health
checks
aren't
passing,
we
won't
scale
down
if
they're,
not
passing
so
with.
J
That
said,
I
think
it's
pretty
safe
for
us
to
go
ahead
and
remove
the
immutability
of
that
gooing
configuration,
because
if
somebody
does
introduce
some
type
of
mutation
and
it
causes
an
issue,
we're
going
to
block
all
operations
and
we
shouldn't
get
into
a
situation
that
the
control
plane
and
especially
NCD,
loses
quorum.
So
I
don't
see
an
issue
with
opening
that
up.
I
Would
say
sure,
but
like
would
that
just
be
such
that,
like
just
so,
we
would
have
a
release
that
didn't
do
it
because
I
think
it's
actually
one
line.
It's
just
something
in
the
hashing
function
that
needs
to
change.
So
I'd
be
curious,
like
what
we.
What
would
we
be
trying
to
get
out
of
having
it
in
a
dot
one
or
two.
G
Things
my
first
worry
is
like
we
are
like
a
few
weeks
away
like
less
than
two
at
this
point
from
your
Twitter
release
or
like
what
was
scheduled
to
be.
It
is
a
design,
change
and
I
would
I
would
like
we
take
more
time
to
think
through,
like
what
can
that
said
like
if
you
are
like?
Okay
with
it,
like
I
trust,
your
judgment,
I
haven't
reviewed
the
PR
yet
so
like.
G
J
Validating
webhook
to
allow
mutation
of
that
field,
but
with
that
said,
compared
to
the
other
changes
that
we
have
in
flight,
around
control,
plane
and
still
outstanding.
This
is
a
pretty
trivial
change
and
I
don't
see
a
lot
of
impact
to
the
change
other
than
we
would
allow
somebody
to
configure
something
that
we
don't
allow
them
to
configure
today.
I
Do
you
think
it
would
be
more
come
like
more
comfortable?
If,
like
we
don't
do
the
change
now
up
like
in
inner
release
but
like
if
we
make
the
change
on
like
I,
don't
know
either
master
for
a
dot
one
or
like
we
just
do
it
in
a
fork
and
run
like
a
bunch
of
different
test
scenarios
like
would
like?
Would
that
help?
You
know,
increase
comfort.
G
My
preference
would
be
the
like
for
the
first
iteration.
No,
we
just
like
to
keep
addresses,
and
then
we
open
the
door
to
like
this
matrix
of
possibilities
that
they
can
change.
We
can
also
use
the
experimental
feature
flags
that
were
working
on
that
there's
a
different
PRS,
which
we
could
just
say
like
we
add
the
feature
like
you
have
to
enable
it,
and
then
that
will
allow
mutability
to
going
it's
just
an
idea,
though
I.
A
Also
think
at
some
point
we're
going
to
need
to
solve
how
to
do
rolling
updates
for
changes
of
things
like
the
API
server
configuration
which,
strictly
and
technically
is
not
part
of
the
joint
configuration
changes
and
long
term
would
probably
be
going
to
config
map
somewhere,
but
in
the
short
term,
I
think
if
we're
gonna
be.
If
we
need
to
modify
API
server
flags,
like
that's
the
sort
of
thing
that
I
could
see
in
a
similar
class
to
modifying
the
join
configuration.
A
A
All
righty,
so
I
wanted
to
give
a
brief
update
on
the
API
review,
with
Jordan
Liggett
for
the
alpha
2
alpha
3
changes.
If
you
haven't
seen
it
I,
have
a
doodle
poll.
I
will
grab
a
link
to
it
and
stick
it
back
in
the
doc
in
a
minute
if
nadir
doesn't
get
there.
First,
we've
had
several
people
respond
like
close
to
20
or
so,
and
the
right
now,
the
time
slot
that
has
the
most
available
participants
is
Tuesday.
A
I'll
also
add
that
during
this
this
time,
Tim
Hawken
is
available
and
has
responded
that
he
can
make
it,
which
was
one
of
the
few
slots
that
he
had,
which
I
think
would
be
useful
to
have
his
insights
as
well
for
an
API
review
for
these
changes.
So
I
know
we're
not
going
to
be
able
to
get
everybody
accommodated.
Given
there's
so
many
of
us,
so
I'm
gonna,
I'm,
gonna,
try
and
pick
a
decent
time
slot
and
I
apologize
in
advance.
A
G
Yes,
I
have
something
else
for
it.
I
opened
an
issue
today
about
cold
freezing
policies.
This
is
something
that
like
I,
was
looking
at
kubernetes
and
how
they
do
it
and
it
it
is
kind
of
interesting
and
also
like
a
more
complicated.
It
probably
should
be,
but
if
you
have
any
thoughts
like
I
open
this
issue,
so
you
have
experience
just
like
I'm
doing
code
free
just
before
a
release,
something
like
I
have
noticed
personally,
as
I
get
things
on
every
p,
VI
reviews.
G
Lately,
it's
like
we
keep
adding
to
the
milestone,
even
when,
like
the
milestone,
it's
like
pretty
much
should
be
almost
there
and
like
I
would
like
to
put
something
in
place.
That's
like
we
kind
of
like
a
merge
feature
request
with
this
time.
Frame
proposal
can
be
open
like
at
any
point
in
time.
It
doesn't
matter
like
when,
when
they
get
open
but
yeah
I
would
just
I
would
just
like
to
gather
ideas
and
put
a
document
out
there.
A
A
All
right,
well,
I,
think
I'm
going
to
skip
it
for
today,
given
that
we're
all
fairly
busy
with
reviews
and
code
and
whatnot.
So
if,
if
you
all
have
some
spare
cycles
to
review
pull
requests,
please
take
a
look
at
the
ones,
especially
that
are
for
any
of
the
0.3
milestones.
We
do
have
this
probably
worth
pointing
out.
A
So
in
the
milestones
we
created,
we
have
three
that
are
for
v1
alpha
3
this
first
one
that
I'll
mention
zero.
Three
zero
is
things
that
we
want
to
have
done
by
ultimately
by
March
sixth,
which
is
when
we're
trying
to
get
our
release
out
the
door.
The
RC
releases
RC
dot,
one
since
mentioned
at
the
top
of
the
meeting
about
trying
to
do
weekly
release
candidates.
A
So
if
we
have
things
that
we
know
that
we
can
specifically
get
done
sooner
or
that
we
want
to
get
done
before
the
March
6
deadline,
we'll
stick
them
in
one
of
the
release
candidate
milestones
and
then
for
things
that
we
know
we
want
to
do
in
0.3
that
are
not
incompatible,
API,
breaking
changes
rate
or
otherwise
breaking
change,
but
we
don't
necessarily
have
to
have
it
done
by
March
6th,
we'll
put
it
in
the
zero
3x
milestone.
There's
nothing
in
there
right
now.
F
A
Eventually,
we
will
probably
have
some
things
in
there
so
again,
I
would
highly
encourage
you
to
take
a
look
at
some
of
the
open
PRS,
especially
though
the
ones
for
the
milestone
the
machine
health
check.
One
I
know
I
mentioned
this
last
week.
It
would
definitely
be
worth
looking
at
if
you've
got
time
to
do.
A
large
review.
I
know
Joel's
put
a
lot
of
effort
in
there
and
is
largely
just
waiting
for
any
additional
feedback.
I,
unfortunately,
have
not
had
time
to
look
at
it
and
I.
G
A
I'll
also
add
for
for
newcomers,
if
you
are
looking
to
find
places
where
you
can
can
contribute,
it's
not
always
about
coding.
Documentation
is
always
something
that
is,
unfortunately,
a
bit
of
an
afterthought
or
second-class
citizen
when
it
comes
to
getting
done.
So
there
are
a
lot
of
documentation
issues.
There
are
a
lot
that
are
in
the
milestone
and
if
you've
got
some
spare
time,
we'd
love
to
have
some
help
there
as
well
alrighty.
Let's
end
this,
one
early
y'all
can
have
33
minutes
back
and
I
will
post
the
recording.
Once
it's
available
thanks,
everybody.