►
Description
Walkthrough of CEP-2: https://cwiki.apache.org/confluence/display/CASSANDRA/CEP-2+Kubernetes+Operator
A
Okay,
we
are
recording,
so
this
I'm
gonna
start
out
with
just
the
easiest
one.
Hopefully
we
can
somewhat
agree
on
the
or
we
could
agree
that
it's
it
easier.
What
Frank
brought
up
a
really
good
point
on
on
the
mailing
list?
And
the
point
was
hey,
you
know
we're
having
two
meetings
and
it
doesn't
seem
like
that's
a
good
idea
having
one
meeting
so
that
we
could
all
talk
about
you
know.
So
we
can
have
this
group
conversation
without.
Having
is
disconnected
conversation
would
make
better
sense.
A
Did
I
did
I
capture
that
right,
Frank,
exactly
yeah,
that
I
won't
lie,
it
won't
be.
It
won't
be
as
pleasant
if
we
do
that,
we
already
do
that
with
the
Apache
Cassandra
committer
or
a
contributor
meeting,
which
kind
of
sits
in
the
middle
of
two
like
twelve
time
zones
and
that's
what
makes
it
difficult.
So
it's
usually
run
around.
We
rotate
it.
It
does
10:00
11:00,
12:00
and
1:00
p.m.
A
Pacific
time,
which
makes
it
on
some
on
those
some
of
the
rotations,
that's
great
for
people
like
New
Zealand
and
it's
in
Australia
and
Japan
as
some
days.
It
isn't
because
they
have
to
get
up
early
and
then
for
for
everyone
in
Europe
it
it
gets
progressively
later,
and
then
it
goes
back
so
I
mean
that
that
that's
something
that
it
isn't
the
best.
Of
course
it's
not
a
tremendously
late.
It's
not
like
one
o'clock
in
the
morning
or
3:00
a.m.
A
A
A
A
For
the
next,
what
I
can
do
is
propose
the
change
on
the
dev
mailing
list.
If
there's
any
dissent,
that's
a
good
place
to
air
it,
because
you
know
you
can
you
can
type,
but,
more
importantly,
we'll
give
it
a
shot.
You
know
if
there's
no
strong
dissent
and
then
I
will
set
up
a
new
schedule
and
we'll
give
it
a
try
and
then
everyone
can
feel
it
and
say
that
this
is
terrible.
Will
never
do
this
again
ever
bye.
A
What
I
really
want
to
make
sure
I'm
optimizing
for
is
participation
if
it's
a
call.
That
only
has
the
same
group
of
people
and
they're
all
in
the
u.s..
That's
not
gonna
help.
That's
not
what
we
want
right.
There's
a
lot
of
great
efforts
going
on
in
most
of
the
world.
We
want
to
be
able
to
cover
that,
but
I'm
willing,
I'm
willing
to
go
down
that
path
and
try
to
schedule
it.
Okay,
all
right.
A
A
A
So
you
can
follow
along
on
the
chat
all
right,
so
this
is
this
is
Anna
Cassandra,
C
wiki,
and
what
Ben
and
I
tried
to
do
is
capture
just
the
basics
of
the
conversations
that
have
already
been
had
and
create
some
framework
here.
What
is
missing
right
now
is
a
really
big
graphic.
This
I
put
in
there
if
I,
don't
know
why
it's
not
showing
up
on
my
screen,
Firefox
decided
to
die.
You
know
it's
probably
the
rest
code.
Oh
did
I
say
that
the
let
me
just
walk
through
what
we're
doing
so.
A
The
CP
has
a
certain
format,
there's
a
there's,
a
template
that
the
pmc
has
created
so
we're
following
that
format.
Of
course
you
know
this.
The
the
status
of
this
right
now
is
this
is
brand
new,
so
we're
doing
the
discussions.
There's
no
we're
gonna
be
doing
this
on
the
dev
list
for
now
and
inside
the
Cassandra
JIRA.
A
So
obviously
this
is.
We
just
have
one
motivation,
but
I
think
this.
This
could
probably
be
stuff.
Look
that
it's
just
powder
food.
You
know,
make
it
easy
to
read
Cassandra
in
communities,
but
that's
what
we're
trying
to
do
we're
primarily
looking
to
make
the
lives
of
people
who
are
running
cassandra.
Better
I
love,
Frank's,
point
of
view
there,
where
it's
like
I
I,
don't
want
the
the
ops
team
come
in
to
ask
me
how
to
use
Cassandra
I'd
love
for
it
to
just
be
working.
A
A
A
Github
you
so
silly,
it's
like
it's
not
like
not
like
half
of
the
world
runs
their
code
out
of
github,
so
the
the
goals
for
this
is
around
trying
to
frame
it
around.
This
maturity
scale
that
the
operator
SDK
works
wrong
and
it
links
really
nicely
with
other
things
like
if
you
go
to
operator
hub
dot
IO.
Whenever
you
look
at
an
operator,
it
says
what
what
level
it
is
level
one
two
three
or
four
or
five,
and
let's
see
if
I
can
just
bring
it
up
this
way,
you.
A
Okay,
there,
it
is
there's
the
graphic
that
I
was
trying
to
do
is
the
graphic
working
outside
of
my
browser.
Is
anyone
looking
at
the
page
right
now?
Okay,
yeah,
all
right,
so
my
browser
just
barf,
that's
great
anyway,
so
the
the
five,
the
the
five
levels,
the
level
one
is
just
basically
install
and
yesterday
this
was
good.
This
came
up
and
I
think
this
is
also
for
today
is
one
of
the
goals.
A
I
think
that
we
should
have
is
defining
what
in
a
cassandra
mean
what
what
in
the
world
Kassandra
world
means
level
one
like
automated
application.
Provisioning
configuration
management,
that's
very
basic
for
for
kubernetes,
but
what
does
that
mean
for
Cassandra?
Seamless
upgrades
also
has
a
very
different.
It's
very
opinionated
in
Cassandra
a
level
three
full
lifecycle.
A
You
know
what
does
that
mean
for
a
Cassandra,
and
you
know
backup
failure,
recovery,
full
life
cycle,
the
think
there's
you
know
that
seems
pretty
reasonable,
but
going
back
to
this
P,
we
think
you
know
really
what
we're
trying
to
do
is
lower.
That
impedance,
like
don't
make
Cassandra
and
kubernetes,
keep
them
from
fighting
each
other,
because
there
is
some
big
differences
between
those
and
really
try
to
achieve
a
level
three
operator
compliance
and
the
level
three.
Is
this
full
life
cycle?
A
Getting
to
that
point
level,
four
and
level,
five
sound,
absolutely
fantastic,
but
you
know
I
think
from
a
scoping
of
just
what
and
looking
what's
already
out
there
level.
Three
would
be
a
really
cool
thing
to
have,
and
you
know
the
with
the
level
four
I
think
we
can
say
that
there
we
should
plan
for
that,
like
that,
should
be
a
part
of
the
we
shouldn't
get
in
the
way
of
doing
that,
but
that
shouldn't
be.
A
The
goal
is
to
ship
something
a
full
level
for
compliance
and
autopilot
that
gets
into
the
AI
ops
thing,
which
a
whole
different
ball
of
wax,
and
we
don't
want
to
impede
trying
to
get
level
three
done
by
saying.
Oh
we're
gonna
make
this
fully
automated
setup
for
Cassandra.
There's
a
lot
to
do
here.
However,
the
horizontal
and
vertical
scaling
the
horizontal
scaling,
yeah,
that's
kind
of-
should
be
in
level
one
and
that's
why
I
go
back
to
this?
We
have
we
have
these
different
levels
in
kubernetes,
LAN
that
we
should.
A
A
D
Yeah,
just
so
in
the
meeting
last
night
there
were
some
discussion
around,
as
they
were
just
to
reiterate.
As
far
as
configuration
management.
We
it
quickly
got
into
some
questions
about.
You
know
what
exactly
that
entails
means
obviously
miss
Cassandra.
They
can
get,
gets
non-trivial
fast,
we're
talking
about
consenting
up
configuring
encryption
dedication,
authorization
it's
more
than
just
templating
variables
in
the
animal,
for
example,
so
I
think
it
would.
It
helps
to
probably
kind
of
narrow
in
on
that
and
with
to
have
more
specificity
around
that.
You
know
I.
A
A
E
F
Well,
I'm,
currently
working
on
that
I'm,
integrating
Medusa
at
the
moment.
Looking
good
of
that,
it's
a
little
bit
tricky
because
it's
not
not
yet
to
be
running
in
kubernetes
as
well.
It
works
at
least
I
got
to
reinstall
yeah.
It's
a
bit
tricky
working
around
that
stuff,
but
it
kind
of
works
at
the
moment
could
be
improved
a
lot
and
actually
actually
looking
into
the
whole
decrypted
cluster
backup/restore
stuff.
The
user
doesn't
support
that
yet
so.
E
G
Yeah
I
would
add
travelers
like
I.
Did
you
like
you?
We
are
like
Wesley
building
something
wrong
like
kuru,
that's
like
kind
of
a
universal
declarative
operator,
and
then
we
have
a
kind
of
approach
that
is
called
plants.
So
basically
it's
kind
of
imagine
like
a
CRE.
Well,
we
can
execute
some
kind
of
operator
side
operations
so
for
us,
for
example,
repair
replace
and
you
know,
like
backup
all
kind
of
plants,
and
we
saw
issues,
for
example,
not
just
for
the
backup
and
restore,
but
also
for
node
replace
right.
G
So
it's
easy
to
do
not
replace
when
you're,
using
your
like
some
kind
of
EBS
storage,
but
imagine
like
we
have
customer
or
interest
in
local
storage,
also
see
if
that
is
gone.
Like
a
operator,
intervention
is
necessary.
One
part
of
kubernetes
I
think
like
that
is
also
like
to
be
work.
Seamless.
We
want
to
make
it
part
of
our.
You
know
like
operation
lifecycle,
where,
let's
say
community
node
is
gone.
G
We
are
able
to
recover
it
repair
it
and
pack
it
at
again
PDA
working
Cassandra
cluster,
so
yeah
like
if
somebody
is
interested
in
to
like,
like
that,
like
we,
we
can
go
more
deeper
into
that.
But
basically
it's
come
from
the
idea
from
the
plans
where
we
have
a
state
to
come,
let's
repair,
backup
or
restore
replace,
and
we
try
to
make
it
achieve.
Some
of
these
operations
are
like
already
like
in.
A
A
Although
some
people
do
that's
fine
but
kubernetes,
it
should
be
I
mean
the
we're
gonna.
Do
this
work,
we
should
have
a.
We
should
have
a
place
at
the
end,
where
it's
just
easy
to
run
a
Cassandra
cluster
because
we
run
around
kubernetes
just
and
just
moving
along
just
keep
it
in
keeping
track
of
time
the
non
goals-
and
these
are
really
just
trying
to
get
outside
of.
Let's
make
sure
we
set
some
boundaries
here-
yeah
removing
the
need
for
any
Cassandra
administration.
A
I,
don't
I,
don't
think
that
that's
where
the
goal
and
I
think
we
could
all
agree
that
that's
that's
an
ideal
that
would
be
wonderful
but
I.
Just
we're
not
gonna.
Try
to
do
that.
You're
still
gonna
need
to
administer
your
Cassandra
cluster
at
some
level,
also
providing
like
a
service
facade
for
Cassandra.
That's
also
not
a
goal.
Its
it'd
be
easy
to
go
down
that
rabbit
hole,
but
no,
and
then
there
is
a
concept.
This
whole
thing
around
supplying
official
docker
images
again
another
discussion
on
the
mailing
list,
which
is
sure,
get
heated.
A
A
The
what
what
I
would
mention
here
is
that
this
C
wiki
is
available.
Anyone
can
work
with
this.
You
can
edit
it.
You
can
go
in
just
and
I'll
put
these
in
some
notes.
So
you'd
see
how
to
do
this,
but
if
you're
interested
in
adding
things
and
one
of
the
sections
I'm
going
to
add
on
here
after
we're
done
with
these
meetings-
is
a
section
for
defining
Cassandra
level.
A
A
It's
it's
easy
and
free,
but
you
would
just
need
to
let
one
of
us,
like
myself
or
Ben,
know
that
you
want
to
have
edit
and
it's
really
easy
to
do.
Actually
anybody
with
a
there's
a
lot
of
us
if
you
go
on
the
dev
list
or
on
the
dev
slack,
you
can
ask
the
Kay
can
I
get
edit
rights
on
the
confluence.
This
is
wide
open.
If
you
want
it,
then
it's
not
that
big
deal,
but
that
way
you
can
go
into
the
stock
and
add
your
parts.
A
A
So
this
is
the
proposed
change
here
for
the
Cassandra
project
is
this
is
going
to
be
a
new
sub
project
for
Cassandra
and
that
I
think
that's
something
that
the
PMC
is
gonna
have
to
sort
out
it
will
be.
It
will
be
interesting
because
there's
a
lot
of
different
sub
projects
out
there,
including
the
sidecar.
A
How
does
where
does
a
seed
provider
happen
like?
Is
that?
Because
this
isn't
service
discovery
for
kid
for
kubernetes
and
what
happens
if
you
have
to
restart
your
cluster
or
when
you
start
up
a
cluster,
or
how
do
you
specify
where
your
seed
note
is
or
seed
nodes
are
I
should
say
and
yeah
I
mean
if
there's
I
think
there's
plenty
of
things
to
go
through
here,
but
we
listed
the
first
three
that
came
to
mind
compatibility
targeting.
A
This
is
Cassandra
versions,
targeting
Cassandra
version,
3,
Docs
and
above
going
much
further
back,
and
this
will
be
like
probably
very
closely
related
to
what's
happening
in
the
sidecar
project.
If
you
don't
know
anything
about
the
sidecar
project
ID,
but
you
go
check
it
out,
the
it
should
I
mean.
We
should
look
at
that.
Just
because
majority
users,
as
I've
found
through
polling,
are
writing
some
version
of
3x
and
as
we
move
into
four
dot,
oh
I'm
sure
we'll
have
people
on
3x
for
a
long
time.
A
If
we're
trying
to
show
that
this
official
kubernetes
operator
is
really
all
that,
then
we
should
be
doing
using
it
ourselves
and
that
operator
should
be
used,
as
should
be
a
part
of
our
CIC
D
pipelines
for
consent,
the
Cassandra
Project.
There
are
three
things
out
there
that
that
could
you
know
as
we
automate
the
test
like
D
test
and
there's
a
there's,
a
fuzz
testing
tool
called
hairy
that
which
will
be
open-source
soon
and
then
fallout,
which
was
just
donated
by
data
stacks.
A
A
And
there
was
a
kind
of
an
open
question
around:
is
there
an
acceptance
framework
for
kubernetes
operators
and
we
don't
yeah
I
mean
it.
We
know
that
they're
on
operator
hub
die.
Oh
there,
they
just
ask
like
what
level
are
you,
but
we
couldn't
find
anything.
This
is
test
us.
You
know
the
here's,
the
test
for
your
operator
does
anyone
know
about
that?
Is
there
one.
C
C
A
A
G
Can
speak
a
bit
for
the
Kudo
like
for
we?
Are
we
developed
it
like
for
testing
the
operators
and
we
are
using
that
and
in
the
last
meeting
of
the
open
SDK,
they
agreed
that
they
are
going
to
start
using
kudo
testing
framework
also
echoed
out,
and
we
get
to
present
it
in
the
community.
Sleep
testing
also,
and
they
are
also
like
trying
to
make
it
part
of
covered
and
sick
testing.
So
what
is
the
good
about
thing?
This
is
like
no
like.
G
If
you
guys
don't
know,
you
should
give
it
a
look,
but
it's
basically
about
will
make
sure
that
we
are
driving
the
development
of
an
operator
in
a
way
that
is
not
a
wave.
So
if
I
want
to
make
sure
that
a
if
Cassandra
is
deployed
and
I
want
to
test
that
is
it
deployed,
I
will
just
look
for
the
kubernetes
subjects
and
not
just
you
know,
I
look
into
the
data
quad
that
is
running
and
process.
That
is
not
exposing
and
yeah.
The
other
is
like
yeah,
it's
just
without
any
go
code.
G
C
G
A
A
Well,
there
is
one
okay,
well
I
mean
we're
at
the
bottom
of
the
hour,
and
you
know
just
keeping
our
timing
in
check
here.
I
have
the
other
requests
that
I
had
was
turn
this
into
an
hour
meeting.
If
we're
gonna
debate,
which
I
think
it
makes
sense,
but
other
comments
right
now
or
questions
about
like
what
we've
created
here
or
things
you
think
should
be
there.
I
had.
C
A
couple
of
questions:
mm-hmm
I,
don't
I,
I,
guess
I
hear
you
that,
like
we
can't
expect
everyone
to
use
like
this
exact
docker
image,
I,
don't
understand
how
the
project
can
make
some
things
and
call
it
official
without
a
that,
a
docker,
a
docker
story.
You
know
it's
like
hey.
This
is
the
reference.
If
you
need
to
deviate
because
you
just
have
to
run
Oracle,
Linux
or
or
whatever,
if
you're
I,
don't
know
anything
really
that
has
to
be
deviated
like
this
is
sort
of.
C
We
can
outline
as
much
of
a
spec
as
you
have
to
follow,
but
you
know
there
should
be
a
good
reference
implementation
of
that
right
and
most,
if
you
want
to
make
it
easy,
like
the
docker
image,
is
the
entry
door
to
easy
like
rolling
your
own
is
not
something
small
shops
should
be
trying
to
do?
No.
A
A
If
you
go
to
docker
hub
the
cassandra,
the
one,
that's
that
everyone
you
know
when
you
say:
docker
pull
Cassandra
yeah
it's
from
docker,
not
from
the
Apache
Cassandra
yep
folks,
but
this
is
I
mean
that
should
be
something
that
we
we
address
from
a
project
standpoint,
not
just
this
little
sub
project.
So.
D
I
think
that
reason
interesting.
Is
it
that
itself
reason
in
question
because
mm-hmm
what
about
the
the
Cassandra
project
itself
maintain
the
image
you
know?
I
I've
seen
plenty
of
projects
where
you
know
there.
Maybe
you
have
a
doctor.
Multiple
sets
of
images
set
images
that
are
targeted
specifically
for
kubernetes
and
then
used
for
docker
outside
of
communities,
and
I
think
that's
a
question.
I
I
think
it's
a
legitimate
question
of
maybe
start
out
that
way,
but
there
may
be.
D
A
A
yeah
you
brought
some
good
points
there.
I'm
gonna
stop
my
share,
so
we
could
talk.
You
brought
up
some
good
points
about
like
there
is
not
just
one
right.
There
could
be
a
variety
for
sure,
especially
when
it
comes
to
dev
and
different
things.
I
there
is
there's
a
thread
going
on
a
dev
list
right
now,
but
it
may
be
a
good
idea
to
start
a
different
thread
on
the
devlins.
A
It
isn't
because
right
now,
it's
kind
of
mixed
into
a
different
thread
and
so
yeah
I
I'm
with
you
I,
would
I'm
kind
of
sad
that
you
know
that
when
I
go
to
docker
hub
that
it's
not
an
official
Cassandra
like
Apache
Cassandra
project,
docker
image
and
the
one
that's
out
there
is
pretty
generic
I
mean
it
works
right,
but
is
it?
Is
it
what
we
as
a
project?
Think
is
the
right
way
to
run
it.
C
Yeah
I
brought
it
up
just
because
I've
seen
every
operator
work
around
this
and
we
did
too
and
it's
silly
and
it's
not
working
around
the
Cassandra
build
it's
not
like
the
guitar
GZ
should
be
that
it
shouldn't
be
a
debate
about
that.
But
it's
all
this.
It's
all
the
scripting
that
comes
up
with
your
image
and
the
environment.
Very
it's
all
in
it
like
that.
First
layer,
you
hit,
when
you
start
docker
with
the
shell
scripting
and
what
environment
variables
are
expected
to
be.
C
F
C
A
A
That
makes
sense,
yeah,
I,
think
all
right,
so
we
can
think
that's
worth
debating
during
this
project
for
sure,
and
it's
probably
bringing
that
topic
a
little
more
up
in
the
project.
The
project
has
been
going
along
without
needing
a
docker
to
worry
about
it,
but
I
think
the
time
is
up
right,
not
anymore.
A
All
right,
so
what
I
and
I
hope
everyone
sees
now
is
like
we
have
a
framework
to
start
working
with,
I
will
send
out.
Have
the
recordings
I'll
put
those
up
into
the
sea,
wiki
send
out
an
email
to
the
mailing
list,
but
essentially
we're
at
the
point
now
where
we
need
to
do
some
discussion
and,
if
you're,
not
on
the
ASF
slack,
there's
no
sub
slack
with
a
channel
for
standard
dev.
I.