►
From YouTube: 20200715 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:
everyone
welcome
to
july
15
cyclist
rifles,
sub
project
meeting
cluster
api.
We
have
a
light
lightish
agenda
today.
I
added
some
things
here.
I'm
going
to
put
the
meeting
notes
link
into
chat.
If
you
want
to
add
your
name
to
the
attending
list,
feel
free
to
do
so.
Is
anybody
here
new
and
would
like
to
introduce
themselves
feel
free
to
mute
and
say
hi,
hello,
I'm
new
here
this
is
swan
hi.
Everyone
welcome
hi.
A
A
This
is
a
huge
milestone
for
the
project,
very
proud
of
the
work
that
this
community
has
been
doing
for
a
month
and
a
half
or
so
there's
a
lot
of
new
features.
So
if
you're
new
to
the
project
like,
I
feel
feel
free
to
dig
into
the
to
the
our
documentation,
the
book
should
be
up
to
date
in
terms
of
like
a
quick
start.
So
that's
like
a
good
way
to
start
and
yeah.
This
is
already
using
zero.
A
Three
seven,
given
that
it
auto
updates
every
time
a
new
pr
merges
in
but
yeah
give
it
a
try
and
if
you
will
find
any
bugs,
which
I'm
I'm
sure,
there's
gonna
be
a
few
definitely
found
an
issue
so
that
we
can
get
those
fixed
a
huge
shout
out
to
everyone.
Everyone
involved
in
this
release
again,
thank
you
and
now
we'll
look
forward.
A
One
psa
in
this
psa
is
that
cluster
cuddle
becomes
required
with
this
release
with
cluster
api.
It
was
required
before
for
most
of
the
infrastructure
providers
because
they
already
had
environment
variables
in
their
in
the
published
components
and
now.
Cluster
api
is
also
publishing
components
with
environment
variables.
In
so,
you
would
have
to
update
your
culture
code
to
the
latest
version
to
make
use
of
the
to
install
zero
three
seven
so
added
this
psa
here.
A
B
Yeah
I
have
one,
I
can't
raise
my
hand.
Sorry
yeah,
I
just
wanna
say
you
can
also
use
amp
steps.
Don't
have
to
use
clustered
cuddles
like
so
for
providers
in
your
like
make
files
and
for
development
purposes
to
install
the
cappy
components.
As
long
as
you
use
amp
stops,
it
should
be
fine,
you
don't
need
cluster
catalog.
A
Yep
gab
is
pointing
out,
the
tilt
is
failing
now
and
that's
actually
correct
and
there's
already
an
issue
open
to
fix
it.
Andy.
A
A
So
it's
like
definitely
like
some
wind
down
time,
given
that
the
huge
release
that
we
had,
but
if
you
have
like
anything
that
you
would
like
to
see
in
terms
of
features
improvements,
if
you
have
you
know
any
ideas,
we
definitely
need
a
lot
of
help
in
the
user.
Experience
side
of
things
definitely
feel
free
to
open
an
issue
and
we'll
try
ask
them
once
we
go
into
planning.
A
The
last
psa,
slash
shout
out
is
that
the
pr
to
promote
ccl
to
montana
has
been
merged
so
congrats.
There
has
been
a
lot
of
plus
ones
on
that
pr
and
very
well
deserved.
A
A
All
right,
so
if
there
is
no
demos
today,
I
added
the
these
here
we
don't
have
to
talk
about.
I
just
wanted
to
give
you
all
the
like
kind
of
like
a
status
report
that
we
have
four
more
proposals
open.
There
is
the
auto
scaler
infrastructure
capabilities,
machine
deletion,
hooks
external
remediation.
A
This
one
already
has
code
already
in
place.
External
remediation
needs
a
little
bit
more
review
again.
I
apologize
that
if
we
have
been
slow
but
we'll
definitely
get
to
it
like
either
this
week
or
next
week,
andy.
A
B
D
B
A
A
All
right
quite
a
bunch
today,
so
let's
go
to
discussion
topics
fabrizio,
take
it
away.
E
Thank
you,
so
I'm
I'm
working
on
an
upgrade
on
a
update
on
our
testing
guideline
and
conventional
forecast
api,
and
this
is
a
follow-up
of
some
recent
problem
that
we
are
going
to
test
some
flakes
and
also
some
facing
some
fake
client
limitations
that
are
documented
in
in
issue.
So
I
started
to
update
the
document
and
I'm
looking
forward
with
with
the
goal,
to
get
a
standardization
of
how
we
write
tests
and
and
to
address
the
problem
that
we
are
like
and
I'm
looking
forward
from
community
feedback
about
these
proposed.
F
Okay,
I
just
wanted
to
add.
I
just
read
that
proposal
this
morning
overall
plus
one
for
it.
I
noticed
that
currently
we
are
using
m
test,
primarily
in
our
ginkgo
suites,
but
from
my
understanding
it
we
are
sort
of
asking
to
lean
towards
using
m
tests
also
for
unit
tests
as
well.
F
E
F
A
Just
to
give
a
little
bit
more
context
to
to
this
group,
so
apart
from
the
flakes
and
etc
like
there
have
been
two
complaints
that
I've
heard
from
from
the
wild,
which
is
one
genko
test,
you
can
it's
really
hard
to
develop
them,
but
also
debug
them.
A
So
you
can
just
like,
for
example,
run
in
an
editor
like
on
an
ide
like
a
single
test,
for
example,
if
you
want
to,
you
can
do
that
from
the
command
line,
doing
regular
expressions,
and
so
the
developer
experience
is
kind
of
like
whatever
today,
and
the
other
thing
is
that
fake
client
has
a
lot
of
limitations
that
we
discussed
this
at
the
controller
runtime
meeting.
A
There
is
some
folks
that,
like
would
like
to
contribute
to
it,
but
given
that
we
rely
on
on
it
a
lot
and
it
actually
doesn't
behave
properly
most
of
the
time,
especially
with
regards
to
patching
and
things
like
that,
the
bridge
is
proposing
kind
of
pretty
much
said.
Let's
just.
E
F
Growing
yeah,
I
I
agree,
there's
also
some
places
I've
seen
in
the
code
base
where
our
fake
client
is
actually
like,
our
own
fake
client
with
like
maps
and
other
sort
of,
and
it's
not
the
controller
runtime
fake
line
so
be
really
nice
to
kind
of
standardize
on
a
consistent
behavior.
So.
A
Okay,
never
mind
any
other
questions
comments,
concerns
for
fabrizio
on
the.
A
C
Thanks
vince,
I
wanted
to
bring
this
to
everybody's
attention.
There's
an
issue:
that's
been
open
for
a
while
about
letting
all
or
more
of
the
kcp
spec
be
mutable,
and
we've
got
some
folks
who
are
interested
in
being
able
to
change
things
like
the
flags
for
the
various
control
plane
components.
Maybe
some
of
the
volumes
and
volume
mounts.
So
if
you
all
have
any
use
cases
around
doing
rolling
updates
where
you
want
to
be
able
to
change
really
anything
or
that
is
in
the
cube,
adm
config
spec
or
the
kcp
spec.
C
Just
let
us
know
add
some
comments
on
this
issue
and
we're
gonna
look
into
prioritizing
this
sometime
in
the
next
few
weeks
or
months.
So
the
more
use
cases
we
have
the
better.
A
A
All
right,
so,
if
there
is
anything
else,
we
will
just
go
into
issue
transaction
for
now
see
there's
a
few
with
no
milestone
mctill
to
work
with
ms
substitute
by
available.
So,
yes,
we
should
probably
do
this
soon.
E
A
Oh
okay,
perfect
yeah,
so
yeah,
I'm
just
gonna
bring
03x
in
here.
A
All
right
support
moving
custom
resources
in
cluster
curl.
I
think
there
is
already
a
pr
open
for
this,
so.
A
And
we
just
discussed
this
a
little
bit
last
weekend
for
folks
that
don't
know
like
what
this
is
actually
solving
is
talus
folks
that
are
installing,
like
other
components
or
crds,
alongside
like
cluster
api
with
cluster
cuddle.
A
C
E
C
For
cap
v
and
there
wouldn't
be
any
major
changes
to
cap
v
other
than
supporting
the
statically
assigned
ips,
but
there
would
be
custom
resources
that
this
other
controller
is
creating
and
maintaining
that
go
along
with
cluster
api
and
would
be
need
to
be
moved
as
well.
So
another
example
of
a
use
case.
A
What
else
do
we
have
flavor
tokens
flavor
nodes?
I
looked
at
this
briefly.
E
A
Yeah
thanks,
then,
for
asking
for
more
user
stories,
it's
a
little
confused.
What
they're
trying
to
do
so
for
now,
I'm
just
going
to
put
in
the
next
and
say
awaiting
more
evidence
here
unless
this
person's
on
the
call,
I
don't
think
so,.
A
Allow
docker
machine
template
to
be
immutable,
so
there
is
already
a
pr
out.
The
only
thing
that
I
I
pointed
out
here
is
that
so
template
in
general
have
been
immutable
for
quite
a
while
we're
trying
to
relax
that
a
little
bit
to
say
you
can
actually
mutate.
The
the
template
and
3283
actually
covers
like
some
of
this,
the
problem
with
like
relaxing
the
whole
requirements
that
like,
if
you
expect
like
a
then
for
in
this
case,
like
kapiti,
to
make
changes
on
it.
A
E
I
am
discussing,
there
are
two
issues
about
the
same
problem
and,
in
my
opinion,
that
there
is
something
underlying
which
which
should
be
clarified,
because
basically,
the
the
user
is
using
custom
image
to
to
use
kathy,
which
is
something
that
was
never
necessary
before
and
I
want
to
understand.
Why
is
doing
this
and
what
is
the
the
underlying.
A
Need
it
makes
sense,
I
think
yeah,
let's
put
in
a
way
more
evidence,
and
next
there
was
a
reply
to
that
so
yeah.
But
if
oh
go
ahead,
yeah
you're
here
all
right.
This
is
our.
G
Vendor
sure,
actually,
there's
a
there's,
a
separate
issue
of
youtube
going
through
that
list
that
you
were
looking
at
earlier,
the
need
is,
is
primarily
related
to
the
kind,
the
underlying
tool
that
we
use
for
percapti.
G
But
here
yeah
I
mean,
like
I
said
in
the
in
the
comment
there
is
that
it's
perfectly
it's
perfectly
fine.
If
you
know
we
wanted
to
make
this
mutable,
but
at
the
same
time
require
that,
for
example,
providers
like
capd
or
metal
cube
or
anyone
else,
then
go
ahead
and
actually
take
the
changes
to
the
template
and
apply
them
to
the
existing
machines,
or
we
can
leave
that
up
to
the
user
to
do.
If
that
is
desirable,
and
that
mechanism
does
not
exist
today
in
capti,
it
would
need
to
be
implemented.
A
Got
it
so
with
that
information
like
so
in
initially,
3283
was
open,
given
that
we
added
now
the
information
on
each
machine
where
it
was
cloned
from
so
with
that
information
like
capti
can
be
modified
to
reconcile
those
changes,
and
the
ask
there
was
that,
like
the
changes
would
be
reconciled
and
the
user
doesn't
need
to
have
more
intervention
other
than
changing
the
spec
field
that
you
would
want,
but
I
guess
like
I
don't
I
don't.
I
don't
have
like
enough
experience
on
custom
image.
A
A
G
Right,
that's
what's
been
documented.
Is
that
if
you
want
to
change
the
image
in
the
template,
you
would
basically
create
a
brand
new
template
and
patch
kcp
or
the
machine
deployment
accordingly,.
G
A
Now
that
makes
sense.
That's
a
good
question
to
ask.
That
is
a
good
question.
It's
like.
Is
it
necessary
anymore,
like
if
we
have
upgrade
after,
although
we
only
have
that
on
kcp,
I
guess
jason
or
any.
Do
you
remember
like
why?
I
guess
no,
I'm
forgetting
at
the
time
we
put
everything
as
immutable
and
we
didn't
want
to
change
the
template,
but
not
change.
Yeah.
C
Anything
it
was
because
there
was
like
no
way
to
really
know
like
if
we
wanted
to
support
deployment
role
rollbacks
on
machine
deployments,
if
everything
is
always
pointing
to
the
same
template
which,
which
becomes
mutable
there's
no
way
to
get
back
to
where
you
were
before
configuration
wise.
G
You
to
me,
if
I
understand
this
correctly,
the
objective
still
is
to
support
some
level
of
mutation
on
the
template,
but
it
would
be
left
to
the
providers
to
implement
appropriate
reconciliation
of
the
machines,
existing
machines.
A
Yeah,
so
I
would
say,
like
the
genetic
rule
we
document
this
is
like.
If
you
can
make
that
change
in
place,
then
you
can
make
the
template
immutable.
One
example
is
in
aws,
you
have
security
groups,
which
I
I
believe
that's.
Why,
like
andrew
raised
this
issue
the
first
time
and
you
want,
you-
can
add
new
security
groups
or
remove
them
on
the
fly
without
affecting
hopefully
affecting
anything
security
group
syntax.
A
A
A
This
is
one
other
of
yours,
fabricio
or
arbil,
arvin.
E
G
I
think
this
issue
is
actually
not
quite
the
same,
so
I
was
actually
working
with
kathy
a
little
bit
recently
and
again,
as
I've
said
in
that
comment,
there
is
that,
if
you
look
at
the
kind
release
documentation
they
for
a
particular
version
of
kubernetes,
they
recommend
not
recommend
they
require
that
you
use
an
exact
node
image
with
a
particular
digest,
and
if
you
don't
do
that,
what
I
noticed
at
least
when
I
was
working
on
it
is
at
least
a
very
flaky
behavior,
so
you
might
upgrade
from
let's
say
version
1.16
to
17..
G
Any
upgrades
wouldn't
work,
but,
however,
you
do
an
upgrade
from
17
to
18..
It
did
work
and
also
depends
on
it's
a
function
of
you
know
what
version
of
kind
you're
using
if
you're,
using
kind
version
0.7
versus
0.8.
G
E
I
want
to
replace
that
we
are
aware
of
difference
between
kind,
zero,
seven
and
zero,
eight
eight,
but
these
differences
are
relevant
only
for
the
management
cluster
that
you
actually
create
with
kind
wise.
We
are
not
aware
of
of
problems
related
to
the
diversion
of
kind
or
of
the
of
the
kind
or
the
kind
image
when
you
create
something
using
kappa
d,
kabri
internally,
kabdhi
only
relies
of
kind
image
for
having
an
image
that
contains
all
the
binary,
but
internally
it
does
not
use
a
kind,
so
it
should
be.
E
A
D
Yeah,
so
I
I
pretty
sure
that
you
know
all
it
would
take
today
is
just
to
delete
the
cube
cubeconfig
to
have
it
regenerate.
But
I
was
a
little
more
concerned
with
their
user
story
of
wanting
to
do
this
for
security
reasons,
which
I
I
don't
think,
there's
anything
that
we
could
do
to
address
that
until
we
have
some
ability
to
rotate
certificate
authorities.
C
Mainly,
I
mainly
because
I
agree
with
what
jason
was
saying
and
yes,
you
can
delete
the
secret
and
it'll
get
regenerated,
or
at
least
that
worked
in
the
alpha
two
days
might
be
worth
re-verifying.
A
Okay,
thank
you,
and
that
was
the
last
issue.
It's
great
any
other
last
minute
topic.
A
All
right
well,
thank
you
all
and
have
a
great
week,
yeah.