►
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
Well,
good
day,
good
evening,
good
morning,
welcome
to
the
office
hours
of
cluster
API
provider
AWS.
A
This
is
a
cncf
kubernetes
sub
project
called
kubernetes
6
sub
project
and,
as
such
we
buy
by
the
the
code
of
conduct,
which
basically
means
be
nice
to
each
other.
A
If
you
could
add
your
names
as
a
list
of
attendees
in
the
document,
I've
shared
it
in
the
in
the
chat
window
feel
free
to
to
put
a
message
into
the
chat
window.
If,
if
you
don't
see
the
link
or
if
you
need
access,
if
you
don't
have
access,
you
just
need
to
join
a
Google
group
and
then
you'll
get
access
to
it.
Also
when
you're
in
the
dock.
A
If
there
is
anything
you
would
like
to
put
on
the
agenda
for
today
feel
free
to
to
add
that
now
as
well,
I'm
gonna
put
something
on
as
well
as
it
is.
A
B
Okay,
I'll
try
hello,
I'm,
Leonard,
I'm
search,
reliability,
engineer
working
for
Kong
and
we
started
to
gain
interest
in
cluster
API
provider
AWS
recently.
So
in
order
to
like
track
the
project,
I
decided
to
start
attending
the
community
meetings.
A
Well,
welcome
thanks.
Actually,
one
of
maybe
one
of
your
colleagues
was
one
of
the
original
contributors
to
the
eks
functionality
in
Kappa,
Michael,
Beaumont
I,
don't
know
if
you
name.
A
So
yeah
so
Michael
when
we
originally
did
the
the
eks
implementation
yeah,
so
he's
a
good
person's
name
as
well
cool.
A
So
let
me
just
add
this
note
in
there
so
moving
on
to
the
agenda,
so
there
is
one
PSA
unless
anyone
else
wants
to
add
any
so
the
first
well
and
only
PSA
is
version.
2.1.4
was
released
at
the
end
of
last
week.
A
Basically
just
a
small
number
of
bug
fixes
in
there.
We
still
have
one
issue
reported
with
this
release,
if
you're
using
Alpine
and
trying
to
use
cluster
AWS
ADM.
So
that's
probably
something
we
need
to
we'll.
We
will
have
to
investigate
and
do
another
patch
release,
probably
in
the
next
week
or
so.
A
So
also,
if
you
have
any
I
forgot
to
say
it's
right
at
the
beginning,
obviously
we
use
raised
hands
feature
in
in
Zoom.
If
you
want.
If
you
would
like
to
comment
on
anything
that
said
completely
forgot
the
script
it's
been
one
of
those
days,
but
so
we
can
move
on
to
the
agenda
items
so
address.
You
have
the
first
item.
C
Thanks,
that's
basically
what
we
had
in
May,
I
guess:
I
can
quickly
recap
what
it
was
about,
because
I
had
two
topics
back
then
so
this
one
is
about
machine
pool
notes
not
being
rolled,
so
no
new
service
being
created
if
the
conflict
changes.
C
So
if
a
machine
pool
refers
to
cube
ADM
config,
there
are
several
fields
in
the
qpa
demo
config,
which
can
change,
let's
say
files
or
scripts
or
whatever,
and
then
in
most
cases
you
want
new
servers
to
be
created,
how
it
works
copy
whenever
it
detects
the
change
to
cube
ADM
config
such
as
spec.files
marking
it
here,
it
will
not
regenerate
the
bootstrap
data
secret
immediately,
but
only
later
so
when
it
detects
say
it's
time
to
refresh
the
bootstrap
token
got
the
token.
So
new
notes
can
join.
C
So
it
makes
this
build
the
config
and
the
bootstrap
token
in
one
cloud
in
it
or
yeah
in
the
usual
case,
Cloud
init
script
and
puts
that
in
a
secret,
but
it
doesn't
update
that
immediately.
But
that's
what
we
want
and
then
also
Kappa
specifically
does
not
trigger
an
instance
refresh
meaning
rolling
out
new
nodes
through
the
order
scanning
group
in
AWS
when
there
is
a
change.
So
the
problem
is
both
in
copy
and
not
updating
the
secret
in
a
reasonable
time
frame
and
Kappa,
not
rolling
the
notes
and
we
discussed
last
time.
C
Does
it
actually
make
sense
that
a
machine
pool
refer
to
a
cubeadm
config
and
then
several
people
investigated,
including
Cameron
thanks
and
so
there's
some
quotes
from
Cecile,
for
example,
in
chat
that
says
yeah
the
cube,
ADM
config
is
used
for
all
the
instances,
so
this
is
on
purpose,
so
we
do
not
use
a
cube.
Edm
config
template,
but
just
one
Cube
ADM
config
object.
So
that's
fine,
that's
expected
and
Cameron
I
think
you
found
multiple
points.
Any
GitHub
comment.
So
basically
just
your
conclusion
was
that
it's
intentional.
D
I'm
trying
to
refresh
my
memory,
but
when
I
looked
at
this
I
was
looking
at
the
original
proposal,
doc,
which
is
what
I
linked
to
here
in
this
comment.
But
basically
the
proposal
for
machine
pools
says
that
the
Kuba
ADM
can
be
shared
for
all
of
the
instances
in
the
Machine
Tool.
So
that
sounds
exactly
right.
C
C
A
C
Basically,
come
here
again
to
just
discuss
what
you
want
to
do
so
I
do
have
this
partial
fix
for
a
couple
mentioned
in
our
meeting
notes
what
it
does
in
very
short
it
detects.
Is
there
a
difference
and
bootstrap
data
or,
as
we
call
it
in
AWS,
the
ec2
user
data?
Is
there
a
difference,
apart
from
only
the
bootstrap
token,
meaning
did
anything
in
the
node
configuration
change
if
so
trigger?
And
it's
refresh.
C
And
it
does
so
by
looking
at
yeah
the
difference
between
previous
and
the
new
value
of
the
secret,
essentially
doing
a
div,
and
if
it's
more
than
the
token
value
which
gets
refreshed
every
convinced,
then
yeah
hey.
Let's
roll
the
notes.
It's
really
not
perfect,
because
if
we
do
have
another
format
from
cloud
in
it,
let's
say
ignition
or
something
encoded
like
base64
or
whatever,
then
it
will
not
work
anymore.
So
it's
definitely
not
a
great
fix
to
be
honest,
I,
don't
like
it
that
much
and
it
only
fixes
part
of
the
problem
right.
C
A
C
A
Yeah,
it's
an
interesting
one.
Yeah
I
guess
do
you
know
if
there
is
a?
Is
this
the
only
issue
Upstream
in
Cappy
around
this
area?
Do.
A
D
Fixed
a
very
similar
issue,
not
with
the
Kube
ADM
config,
but
with
the
eks
config
object,
which
also
can
be
referenced
inside
of
a
machine
pool
if
you're
using
the
the
AWS
speed,
basically
the
AWS
version
of
the
service,
but
because
that
resource
is
managed
by
Kappa,
it
doesn't
have
the
same
problems.
It's
not.
We
don't
have
the
split
logic
like
we
do
here.
Yeah.
D
A
Yeah,
although
I
so
this,
this
reminds
me
of
another
that's
reading
this
issue
here
in
Upstream.
Capital
reminds
me
another
old
proposal
where
again
from
nadir
who
around
splitting
at
home,
especially
having
a
a
node
agent
machine
ADM
that
would
do
the
do.
The
bootstrapping
on
the
Node
yeah
I
can
see
the
difference
to
the
qubit
authentication
authentication
as
well
yeah,
so
I'm
I,
don't
know,
I,
don't
get
to
see
it
but
yeah.
So
we
should
obviously
Drive
the
change
Upstream
to
to
refresh
the
actual
secret.
A
C
C
It's
it's
tested,
but
I'm,
not
confident.
Yet
so
90
done,
let's
say
so:
I'm
ready
to
discuss
it
with
them.
I
think
and
then
the
other
question
is
so.
Let's
say
there
are
different
formats
and
in
one
of
them
we
cannot
diff
against
the
bootstrap
token,
because
it's
a
64
or
something
then
we
would
probably
also
talk
to
copy.
So
they
provide
us
a
way
to
to
check
hey.
C
D
What
would
be
the
ideal
way
to
detect
the
change
in
in
the
resource
when
the
resource
gets
modified
and
the
like?
The
web
hook
that
the
admission
web
hook
detect
the
change
from
the
resource.
C
D
D
C
C
A
Yeah
sounds
like
a
plan
Wellness
internal
discussions.
A
Oh
yeah
and
then
we
could
hopefully
then
chat
I'm
gonna
get
about
this
in
two
weeks
time
or
before.
I
think
subjects.
D
A
On
Wednesday,
brilliant,
thank
you
cheers
and
over
to
you,
Cameron.
D
Similar
sort
of
topic
I
have
a
PR
that
I've
had
open
for
a
little
bit.
We've
done
a
lot
back
and
forth
and
how
to
improve
it
and
workshop
it
and
I
think
it's
in
a
pretty
good
State.
Now
it's
got
a
LG
TM,
but
I'm
not
sure
what
else
we
need
to
do
before
this
can
be
considered
to
be
included
in
a
release.
D
If
there's
anything
people
are
concerned
about,
if
you
haven't
taken
a
look
at
this
is
fixing
a
couple
different
bugs
with
the
launch
template
diff
detection
and
we
ended
up
just
rewriting
the
way
it
does.
Diff
detection
entirely
came
up
with
a
much
better
solution.
I
think
the
reviewers
and
I
were
able
to
brainstorm.
Some
really
good
ideas
in
here.
So
I'm
just
wondering
what's
needed
to
get
this
included
in
the
next
release.
A
So
if
we
got
one
of
GTM
who
who
did
the
LG
TM,
okay,
yeah
cool
stage,
basically
just
need
a
an
approval,
then
so
we
need
to
get
I.
Guess
me
and
kitter
or
Daniel,
to
review
and
improve.
D
A
Good,
and
does
anyone
have
any
inputs
on
this
one.
A
A
Okay,
thank
you.
Well,
that's
that's
the
end
of
our
agenda.
It's
short
and
sweet
today.
Is
there
anything
else
anyone
would
like
to
talk
about
that
isn't
on
the
agenda
today.
B
I
might
want
to
talk
about
something
that
bothered
us
a
bit.
Not
I.
I
haven't
opened
an
issue
yet
about
that,
because
I
don't
know
if
it's
worth
mentioning
since
the
version
2.0,
we
lost
a
little
bit
of
capabilities
for
the
customization
of
networking
if
I
understood
correctly,
and
we
don't
have
a
way
to
selection,
to
like
select
only
a
set
of
availability
zones
across
like
to
create
the
VPC.
B
Unless,
like
we
have
two
ways
to
influence
that
it's
availability,
Zone
usage
limit
and
availability,
Zone
selection,
which,
like
say
home
home
many
availability
zones
we
can
select
and
how
to
choose
them
like
either
random
or
alphabetical.
But
something
that
will
be
nice
would
be
a
way
to
just
like
give
a
list
of
azs
just
to
create
a
VPC
like
hello,
a
bit
more
granularity
on
the
creation.
B
The
use
case
we
have
behind
that
is
that
we
are
trying
to
to
put
a
little
bit
of
context,
we're
trying
to
match
azs
for
our
customers,
so
we'd
like
to
have
like
a
VPC
that
only
spans
the
ACs
over
of
our
customer
as
well.
We
know
that
we
can
restrict
the
way.
B
The
subnets
in
which
we
deploy
the
nodes,
as
well
as
the
subnets
like
Carpenter
crates,
because
we
leverage
fargate
and
Carpenter
and
not
machine
poles,
but
directly
being
able
to
create
a
VPC
that
only
targets
specific
cases
will
be
much
simpler.
In
the
end.
D
B
B
It
means
like
using
node
labels
and
selector
and
selectors,
which,
like
adds
a
lot
of
overhead,
while,
if
we
only
have
vpcs
that
spans
the
ACs
that
interest
us,
we
can.
We
do
not
have
the
overhead
of
managing
labels
and
selectors
down
the
road
for,
like
both
Target
and.
A
All
right,
so
we
I
guess
we
got
a
longer
answer
or
longer
solution
to
this
this
problem,
so
there
is
discussions
going
on
around
rethink,
rethinking
completely
how
we
specify
the
network
in
Kappa
and
providing
more
I
guess
customization
out
of
the
box,
because
a
lot
of
times
at
the
moment
is,
if
you
want
to
customize
it,
you
then
have
to
sort
of
bring
all
of
your
own
infrastructure.
D
A
Which
is
yeah,
it's
not
great!
So
well,
it's
okay!
It's
you
know
it
is
what
it
is,
but
yeah.
We
want
to
provide
a
bit
more
of
a
middle
ground
and
more
customization
options.
So
that's
that's
the
longer
answer
and
there
is.
There
is
a
proposal
around
it.
I
can
I
can
forward
on
that,
but
I
guess
that's
going
to
be
a
longer.
A
That's
going
to
be
a
long
way
down
the
road
because
it
affects
everything,
but
there's
probably
something
we
can
do
I
it's
a
long
time
since
I've
looked
at
that
part,
okay,
but
there's
probably
something
we
can
do
shorter
term.
A
short-term
fix.
Does
anyone
know
off
the
top
of
their
head?
The
API
around
this.
A
C
B
What
I
was
thinking
is
if,
since
availability,
Zone
selection
is
a
string
like
a
hack
might
be
to
a
low
like
arbitrary
values
that
maybe
like
a
comma
separated
list
of
Aziz
and
then
pass
that,
if
it's
not
like
a
random
or
alphabetical,
it
might
be
like
a
quick
hack
without
extending
the
API.
A
It's
we
need
to
look
at
more,
so
we
could
add
an
option
to
the
availability,
Zone
selection,
which
could
be
I,
don't
know
if
one
of
the
bad
word
explicit
and
then
a
new
field,
which
is
which
is
the
actual
availability
zones,
maybe,
as
a
will,
be
a
string
string
array,
so
you
can
do
them
individually,
Maybe
I'm,
just
thinking,
and
that
would
be
that'd,
be
worth
a
change
to
the
let's
change
to
the
values
here,
but
that
could
work,
but
that
would
work
potentially
do
that
shorter
term.
A
Yeah
well
Leonard,
if
you
essentially,
could
you
raise
an
issue.
A
So
that
we
could
then
start
sort
of
the
discussion
on
you
know
the
short
term
and
the
longer
term
fix
for
it.
One
question
I
would
have
is,
if
we're
adding
an
enum
option,
does
that
class
as
a
class
is
a
backwards
compatible
change
from
the
API
point
of
view,
I
mean:
do
we
need
to
bump
the
the
API
version
for
that
we'd
have
to
investigate
that
I?
Don't
know
off
the
top
of
my
head.
C
I
think
yeah.
We
have
to
be
a
bit.
A
A
A
good
point:
that's
a
very
good
point.
Let's
find
that
was
it
being
on
beta,
3
I.
Think
it's
the
yeah
issue.
A
Okay,
yeah
yeah,
so
if
we
did
decide
to
do
and
a
bundles
of
the
API
version,
yeah
Leonard,
if
you,
if
you
create
your
issue,
then
we
can
link
it
into
this,
so
that
we
know
that
when
we
bump
okay,
we
could
get
it
included.
A
B
Sure
I
will
create
the
issue
later
today,
right.
A
A
Cool
well
I'll
show
anyone
else.
Have
anything
they'd
like
to
discuss.
A
Cool
well,
I'll
I
was
recorded
and
see
you
all
again
in
two
weeks.
Sometimes
we
stay
on
after
this
as
well
for
something
we
call
after
hours
and
sometimes
we
do
informal
help
on
issues
and
and
talk
about
and
do
some
bug
triage.
So
if
anyone
fancies
doing
that
feel
free
to
to
stay
on,
but
I'm
going
to
pause,
the
recording
call
admit
into
an
end.