►
From YouTube: 20200422 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
A
And
if
you
have
any
topics,
you
want
to
bring
up,
go
ahead
and
add
them
to
the
agenda
under
discussion
topics
so
to
get
started.
If
anyone
here
is
new
and
would
like
to
introduce
themselves
or
if
you
haven't
been
to
this
meeting
in
a
while
I'll
give
you
a
few
minutes
to
go
ahead
and
say
your
name
and
why
you're
here
and
what
you
want
to
get
out
of
this
meeting
go
ahead.
B
B
Car
I
am
I
work
on
a
product
called
ocean
for
containers,
and
we
are
particularly
interested
in
the
progress
for
faster
API
and
sort
of
how
the
development
is
going
for
different
cloud
providers
and
would
love
to
integrate
with
that
for
ocean
in
the
future,
so
looking
to
explore
sort
of
where
the
work
stands
right
now,
what
is
the
maturity
of
the
project,
particularly
after
the
launch
of
the
v1
alpha
3,
if
there
is
a
available
roadmap
already
and
potentially,
if
there's
an
opportunity
for
contribution
what
those
areas
are.
Thank
you
for
having
me.
A
All
right,
if
not
I'll,
keep
going
so
before
we
get
to
PSAs
I
just
want
to
point
out.
If
you
have
any
questions,
any
general
questions,
even
you
know,
basic
questions.
There's
a
section
under
demo
is
called
general
questions
and
provide
our
implementation
implements
your
questions.
You
can
go
ahead
and
add
any
questions
there
throughout
the
meeting
and
we'll
make
sure
that
we
answer
them.
Ok,
so
now
let's
go
ahead
and
give
the
PSAs
yeah.
D
Thank
you.
I
have
two
PSAs
the
first
one
is
we
try
to
revisit
the
issue
proposal
and
exclusively
focus
on
the
cap
process.
If
you
open
the
PR
that
I
link
there,
this
PR,
like
not
only
removes
the
the
issue
template,
but
it
also
like
kind
of
like
put
some
policies
in
place
and
in
the
contributing
guide.
There
has
been
some
minor
comments,
but
in
general,
like
we
have
agreement
that
this
is
what
we
want
to
do
going
forward.
The
gist
here
is
it
like
we
have
had.
D
We
have
right
now,
20-something
proposals
in
the
open
as
the
issues
and
some
of
them
have
very
light
details.
Some
of
them
have
not
been
commented.
Some
others
instead
have.
There
has
been
like
a
lot
of
back-and-forth,
but
90%
of
them
have
actually
moved
to
a
full
cap,
Google
Doc,
so
to
facilitate
this
process
and
also
to
review
to
reduce
the
burden
on
month,
a
nurse
to
keep
track
of
all
the
proposals.
D
There's
also
like
some
requirements
and
expectation
for
the
experiments.
Experiments
are
like
a
nice
way
to
get
a
PLC
in
closer
API
under
the
experiment
folder,
but
they
they
should
follow
these
rules,
but,
for
example,
I
could
not
change.
Existing
MPI's
have
a
clear
graduation
process
and
ultimately
get
to
GE
and
what
we
mean
by
that
is
to
get
married
into
a
mainstream
class
rep
I
in
the
next
release.
A
D
Second,
one
is
actually
kind
of
related
there's
an
issue
to
align
on
code
freezes
and
release
timelines.
We
have
been
changing,
like
kind
of
like,
for
example,
right
now,
we're
trying
to
focus
more
on
alpha
3
because,
like
there
is
like
a
large
community
there
like
wants
to
adopt
alpha,
3
and
like
we
want
to
give
more
time,
and
we
have
a
lot
of
priorities
as
well
in
alpha
3.
A
All
right,
okay,
so
I,
don't
see
any
demos
for
today
and
again,
if
you
have
any
questions,
feel
free
to
add
them
throughout
the
meeting.
I
don't
see
any
right
now,
so
we'll
keep
going
to
discussion
and
I.
Think
Jason.
You
have
the
first
one.
F
Yes,
so
for
a
while
now,
we've
been
a
bit
bottle,
necked
on
being
able
to
do
timely
reviews
and
maintenance
work
on
the
GCP
provider.
Gab
has
volunteered
to
help
share
some
of
that
burden.
So
there's
a
PR
out
there
to
add
gab
as
an
additional
maintainer
for
a
cluster
epi
provider,
pcp
we're
planning
on
giving
that
one
weak
lazy
consensus
timeout
period.
So
if
you
have
any
questions,
comments
or
concerns
or
concerns,
please
go
to
that
peeler
and
race
them.
Thank
you.
E
The
seal
I
would
like
to
add
that
if
anyone
is
also
interested
in
helping
maintain
the
GCP
provider
or
any
other
provider,
please
reach
out
here
on
slack
the
usual
communications
channels.
We
can
get
you
involved
in
pol
requests,
reviews
and
eventually,
if
things
go
well,
hopefully
over
to
reviewer
and
maintainer
status.
So
definitely
looking
for
additional
people
if
you're
interested.
G
E
A
All
right,
okay,
so
service
help
for
rolling
updates.
Oh
yeah,.
H
Basically,
the
the
problem
that
I'm
facing
or
that
I'm
trying
to
solve
is
how
do
you
do
a
rolling
update
of
a
cluster
while
maintaining
the
health
of
the
services
underneath
it
and
something
like
a
a
pod
disruption
budget
will
get
you
a
little
bit
of
the
way
there,
but
it
doesn't
really
understand
the
health
of
of
something
underneath
it
so
I'll
take
it
say:
elasticsearch
as
an
example.
So
if
I
having
an
elastic
search
cluster
of
the
three
nodes,
I
create
a
new
machine
deployment
and
roll
those
nerds.
H
F
I'll
go
ahead,
Jason,
so
I
think
there's
kind
of
a
shared
responsibility
here.
I,
don't
think
we
want
to
take
ownership
of
the
complete
problem,
but
there
is
a
certain
aspect
that
we
need
to
make
sure
that
we're
respecting
such
as,
if
pockets
rupturing
budgets
are
available,
we
are
doing
operations
in
a
way
that
respects
them
right
now.
We
are
not
overly
mindful
of
that
when
we're
doing
the
deletion
and
node
draining
process,
but
it's
going
to
be.
I
Yeah
I
just
want
to
react
to
the
the
point
about
the
Rhodey
flag.
If
you're
not
ready
to
continue
rolling,
then
don't
already
it's
my
advice.
We
do
that.
We've
pods
that
come
up
and
stay
non
ready
for
20
minutes
and
you
know,
works
great
to
halt
the
rollout.
You
have
to
adjust
your
parameters
so
that
the
rest
of
your
system
doesn't
go
batshit
at
that.
But
my
reaction
to
that
piece
of
the
problem.
H
I
H
H
I
F
A
Thank
you,
okay,
so
I'm
gonna
circle
back
to
questions,
because
I
think
this
was
the
last
discussion
item.
Oh
and
there
are
a
few
questions
so
the
first
one
is:
where
can
we
review
upcoming
Romed
items
and
I
see?
Someone
has
already
posted
the
link
perfect.
So
there
is
a
roadmap
page
in
the
book.
If
you
go
there,
it's
actually
been
updated
quite
recently
by
this
to
include
the
new
v1
aqua,
3
plus
plans.
D
Okay,
so
I
can
raise
my
hand,
like
is
my
host,
but
yes
like
we
have
a
list
of
providers
here
generally
like
AWS,
like
we
work
on
on
it
like
actively
Azure
as
well
and
like
the
other,
like
the
others
like
listed
here
like
it.
Bold,
for
example,
like
in
this
case,
is
the
best
ashram
Idol
3
in
vSphere,
I
think
they're,
the
one
that
like
are
known
to
support
alpha
3,
oh
yeah,
that's
that's
written
there.
D
J
The
one
detail
there
is
that
those
are
all
at.
What's
the
term
now
pivoted
or
self
hosted
they
each
cap
of
manages
just
to
the
one
cluster
that
it's
on,
so
we
don't
have
a
lot
of
visibility
into
how
the
scaling
goes
like.
If
you
have
one
management
cluster
that
manages
all
of
that
stuff,
but
I
know
other
folks
are
doing
that
and
I
think
it
works
pretty
well.
K
A
A
D
Necessarily
like
have
a
provider
that
does
that
today
there
was
one
that
was
posted
in
in
a
cluster
API
issue
that
like
run
kubernetes
in
kubernetes,
but
as
far
as
I
know,
it
was
mostly
for
testing
purpose
in
an
experiment,
and
only
if
someone
can
link
you
to
that,
but
I
can
find
it
later
as
well
give
consumer
the
option
to
do
both
like
this
is
actually
really
possible
today.
So
close
to
API
is
composable.
You
can
bring
your
own
bootstrapping
mechanism
right
now.
D
We
use
cube
idiom
and
you
can
bring
your
own
infrastructure
so
technically
you
like
it
so
there's
a
life
cycle
that,
like
it,
the
machine
controller
for
example,
follows
and
if
you
follow
that
life
cycle
you
can
create
possibly
like
a
community
process,
kubernetes
cluster
or
using
some
own
nodes
for
different
system
services.
So
the
there
is
extension
points
that
you
can
use
to
hook
these
things
up
together,
but
we
don't
have
any
way
to
do
that
today.
A
A
A
A
D
Organs
sure
I
can
do
you
want
me
to
share
you
wanna,
just
you
want
to
be
the
first
one
I'll
do
that
following
ones
so
in
structure
for
using
external
entity
are
incomplete.
Are
you
briefly
looked
into
the
settlements
that
have
been
like
other
comments,
but
there
was
a
PR
out
if
I
believe
Scott
is
here
but
yeah
that
we
should
definitely
fix
this
and
in
alpha
three.
So
what
I
would
do
is
I
want
to
set
the
milestone,
and
we
probably
add
you
to
the
monster
maintainer.
A
Help
I
smell
yep,
okay
and
yeah
I'm
sorry
go
ahead.
I
was
saying
like
that.
That's
the
Help
Wanted
yeah
I
was
just
gonna
say
in
the
chat.
There
is
a
good
point
that
we
should
there's
a
cap
that
proposes
to
odor
add
the
needs
triage
label
on
new
issues.
So
I
guess
that's
that
be
something
we
can
use
for
triaging
new
issues
that
come
in
as
well.
A
L
I
had
so
I
attached
the
cluster,
the
cert
manager
lifecycle
issue
on
this
I
just
thought
that
if
we
do
tackle
this
epoch
as
I
know,
it
was
kind
of
mentioned
in
the
roadmap.
So
whoever,
when,
whenever
I
read
to
tackle
this
sort
of
managing
source
manager
lifecycle
in
classical,
we
could
tackle
this
issue
as
well.
I
just
wanted
to
track
that.
A
A
E
E
D
A
And
then
last
one
also
pretty
big
one
allow
existing
kubernetes
clusters
to
join
a
management
cluster,
so
I'm
gonna
say
there
was
some
discussion
about
this
in
the
cluster
epi
provider
Azure
office
hours
last
week,
and
a
few
people
were
actually
very
interested
in
the
possibility
of
doing
this.
E
A
Make
sense
yeah
I
see
a
lot
of
people
want
to
react
to
this
so
I'm,
just
gonna
read
the
chat
first,
so
s
+,
1
can
of
worms
but
interesting
nonetheless,
and
then
Jason
says
plus
1,
maybe
telling
folks
how
they
can
do
it,
but
ones
who
actually
supporting
it.
So
something
like
here
are
instructions.
If
you
want
to
attempt
this
but
big
warnings
all
over
the
docs
saying
that
this
could
blow
things
up
and
then
yes
and
go
ahead.
I
think
you
were
next
yeah.
K
So
I
would
echo
what
Andy
said
this
might
work
on
the
only
happy
path
there
is
basically
you're
using
qadian
found
are
and
configs
are
consistent
with
foster
api
expects.
As
for
implementing
it,
I
wouldn't
see
this
something
there
is
strictly
supported,
but
classroom.
Api
I'd
expect
an
external
tooling
to
provide
such
thing
for
the
1020
percent
use
cases
that
we
can
support.
M
M
So
if
you
have
an
existing
cluster,
I
I
think
there
are
projects
that
help
you
migrate,
workloads
across
clusters,
and
so
you
could
use
those
together
with
cluster
API
and
certainly
I.
Think
cluster
API
could
talk
about.
You
know
the
use
case
of
migrating.
You
know
from
one
platform
from
one
cluster
to
another,
and
maybe
you
know
helping
to
orchestrate
that
or
thinking
about
that
that
use
case
may
be
integrating
some
some
way
but
yeah.
That's
those
my
reaction.
D
Thanks
yeah
I'm,
just
just
gonna,
have
like
close
with
one
comment
like
for
my
part
like
it.
We
put
it
like
into
the
non
goals
because
we
wanted
to
keep
the
scope
to
like
a
manage
instances
that
we
know
like
how
they
were
spun
off
and,
like
you
know,
to
go
in
for
kind
of
worms
like
an
SMD
mentioned
in
general,
like
I,
do
think
like
this
could
live
outside
of
class
trip.
You
could
be
some
project
on
the
side
that,
like
it's,
not
supported,
but
we
won't
stop
it.
A
Yeah,
it
makes
sense,
I
think
someone
else
had
their
Henry's
I'm.
Sorry
I,
don't
know
how
you
say
your
full-name,
but
as
filigree
no.
J
Yes,
SSSP
my
yeah,
so
what
we
found
is
that
there's
certain
operational
benefits
to
being
able
to
like
adopt
like
pausing
the
reconciliation
of
a
thing
by
deleting
a
managing
resource
or
turning
off
the
catholic
controllers
and
then
fixing
up
some
state
that
we
didn't
know
how
to
get
out
of
otherwise
and
having
the
stuff
adopted
back
into
management.
J
So
I
think
there's
definitely
like
I
agree
100%
with
what
Ben
said
that,
like
mostly
this,
should
live
outside
of
a
cluster
API
but
like
cluster
API
shouldn't
prevent
it
from
happening
right
like
if
there's
machine
deployment
isn't
adopting
existing
resources
that
otherwise
it
would
be
able
to
manage
perfectly
normally
I.
Think
that
would
be
a
bug
in
kappa
or
kappa,
rather
that
we
should
fix.
D
Okay,
I
can
comment
on
this
and
provide
feedback
and
probably
close
at
some
point
yeah.
Thank
you.
Vince.
A
L
I
did
a
discussion
topic
at
the
last
minute
here.
Sorry
about
that,
but
I
just
wanted
to
kind
of
just
inform
the
community
that
I'm
starting
to
take
a
look
into
the
cluster
cuddle.
Templating
work.
I
saw
those,
though,
if
you
open
for
that
and
some
discussion
on
it
and
I'm
just
starting
this
week,
like
looking
to
the
design
and
work
that
might
be
necessary
for
supporting
templates
or
templating
in
within
cluster
Carol.
L
One
of
them
that
was
mentioned
in
the
issue
is
called
ytt,
so
we're
looking
at
that
and
the
initial
work
right
now
is
just
to
create
a
sort
of
interface,
so
nothing
will
actually
change,
but
the
initial
work
will
just
be
like
providing
a
layer
like
an
interface
layer
where
you
can
have
a
pluggable
template
engine.
So
by
default
everything
will
still
remain.
The
same.
It'll
still
support
the
variable
substitution
templating,
but
that
was
like
my
initial
thought.
A
L
A
G
Yeah
I'm,
good
yeah,
so
cushion
is
example,
I
give
you
we
use
that
template
and
then
in
airship,
which
I
am
representing
here
and
we
do
use
customize
to
do
changes
to
it
all
the
substitutions
and
all
those
which
come.
So
if
we
have
a
template
and
if
we
are
going
to
change
other
than
the
various
substitutions,
how
will
it
impact
the
current
relations
with
customize
that
were
for
my
thinking.
L
Ideally,
if
we
design
this
right
and
we
create
this
sort
of
pluggable
sort
of
interface,
we
could
create
like
sort
of
a
customized
library
or
a
ytt
library
or
cues.
That's
another
tool.
I've
come
across,
so
it
should
be
able
to
work.
But
as
of
now
like
again,
nothing
will
change.
The
variable
substitutions
will
still
be
the
default
way
of
doing
things,
so
it
shouldn't
break
any
of
your
existing
zuv
customize
sort
of
operations
that
you
have
going
on.
Yeah.
G
L
A
good
point,
I
think
what
I'll
do
is
I
will
attach
the
issue
that
I'm
kind
of
looking
at,
and
if
you
have
any
comments,
then
I
can
try
and
summarize-
and
you
know,
take
all
that
into
account,
but
I
I'm,
hoping
that
this
will
not
break
or
you
know,
won't
diverge
too
much
again.
Customize
is
just
yet
another
templating
mechanism,
so.
E
I
wanted
to
address
Prakash
this
question.
We
use
customized
right
now
to
build
out
the
the
Amal
for
our
release
bundles.
So
when
you
are
building
cluster
API
or
any
of
the
providers,
we
use
customized
to
generate
the
infrastructure
provider,
Amal
and
so
on.
We
may
or
may
not
be
using
customized
for
the
cluster
templates.
That
are
the
examples
that
we
have
for
that
cluster
couple
can
use,
and
this
is
just
another.
This
is
making
templating
pluggable.
So
if
you
want
to
keep
using
customize,
you
can
keep
using
customize.
A
N
N
A
O
Came
back
to
the
cash
point,
I
think
that
we
should
always
remember
that
we
are
doing
this
change
in
in
the
zero
two
serious
doesn't
mean
that
we
ensure
that
everything
will
be
compatible,
and
that
means
that,
if
you
are,
if
you
today
about
array
or
post
by
the
line
that
we
use
customized,
this
will
work.
We
continue.
It
will
just
contain
more.
We
are
what
the
change
of
who
our
enemy
is
going
to
do
is
to
add
a
new
feature
and
a
new
possibility
to
eventually
improve
your
partner
in
the
future.
N
This
can
be
CN
eyes,
are
our
Beck
rules
or
anything
that
customers
need
to
apply
right
after
a
cluster
is
created,
so
I
just
wanted
to
let
the
community
know
that
I
am
going
forward
to
resolving
the
issues
raised
in
the
design
doc.
So
if
you
have
any
other
comments,
concerns
please
edit
to
the
design
doc.
That
is
in
the
link
that
I
posted
in
the
dark
so
yeah.
That's
all.
A
F
P
We
were
proposed
said
proposal
regarding
external
intermediation.
Last
week
yesterday
there
were
like
a
couple
of
questions
and
comments
thanks
everyone
for
the
review.
The
question
was:
how
do
we
move
next
with
this?
Do
we
need
to
answer
to
those
questions
and
then
open
the
actual
proposal
on
the
github
or.
P
D
D
Usually,
the
process
is
that
we
allow
for
like
a
week
or
two
for
folks
to
kind
of
like
go
backward
comments
and
we
make
sure
to
address
out
the
comments
and
that
all
parties
involved
were
like
okay
to
move
this
in
PR
and
then
once
it's
a
PR
like
it's
gonna,
be
another
round
of
feedback
on
it,
because
there's
usually
reach
a
wider
and
even
wider
audience.
If
you
do
feel
like
you
might
want
more
like,
like
a
an
extra.
D
Feel
free
to
organize
one
and
I
could
drop,
maybe
like
a
link
or
work
with
folks
that
you're
working
with
to
kind
of
schedule,
one
so
yeah.
That's
at
that.
That
would
be
my
suggestion
in
terms
of
the
proposal
I
did
review.
It
I
had
a
concern
that,
like
you,
might
be
trying
to
solve
like
two
things
at
once.
Q
G
Percussion
yeah
I
did
this
is
from
the
last
week
Ferriss
has
brought
in.
There
was
one
conclusion
which
I
wanted
the
conclusion
at
the
end
of
that
I
could
not
attend,
but
the
video
said
that
in
the
video
somebody
said
that
there
will
be
two
ways
of
doing
one
is
to
address
the
health
of
the
node
and
the
other
will
be.
That
will
be
for
the
specific
time
outs,
etcetera,
so
that
you
can
you
don't
get
totally
bogged
down
by
the
bare
metal?
Nor
remediation
is
that
correct.
P
So
if
you
look
at
the
proposal
so
the
first,
what
we
are
proposing
is
that
to
introduce
the
news
article,
for
example,
machine
termination
request
that
will
be
used
to
request
the
remediation
from
the
provider
specific
or
the
provider,
custom,
permutation
controller
and
then
timeouts
on
all
this
stuff.
It
can
be
like
specific
to
each
providers,
ammunition
controller,
like
whatever
you
want
to
how
you
want
to
build.
Your
relation
controller
is
up
to
you
like
basically,
but
we
we
just
for
now.
A
There
will
be
an
additional,
separate
meeting
to
discuss
this
specific
proposal,
so
I
encourage
you
to
attend
that
one
as
well.
If
you're
interested
all
right,
I
think
that's
it
for
today,
thanks
so
much
nadir
and
everyone
else
who
helped
take
notes
throughout
the
meeting.
I
forgot
to
ask
at
the
beginning,
but
that
was
very
helpful
and
thank
you
all
for
participating
and
I
hope
you
have
a
healthy
and
good
week.
I'll
see
you
all
next
week.