►
From YouTube: 2020-06-30 Cassandra Kubernetes SIG
Description
Final meeting where John Sanda covers proposed changes. Next meeting will go over all of the combined changes and next steps.
A
B
B
The
I
think
I
have
it
written
somewhere
so
that
so
here's
the
doc
there's
some
some
minor
just
changes
initially,
since
I
put
some
just
to
clean
it
up
a
little
bit
some
background.
I
won't
spend
time
on
that.
You
can
go
back
and
look
if
you
sure
did,
but
the
the
what
I
can't
what
I've
considered
that
the
the
broad
goal
is.
B
The
operator
should
provide
sane
out-of-the-box
defaults
for
configuring
volumes,
for
your
persistence
and
then
for
other
persistence,
namely
for
those
files
under
Barlow
Cassandra
and
then
should
provide
some
same
defaults
for
other
files
and
categorizes
log
files.
Some
of
the
there's
a
couple
of
things
that
I
would
lump
in
there
that
aren't
necessarily
log
files
or
application
logs
like
to
keep
dumps
or
if
you
need
to
run
a
profiler
profile
or
output
it,
but
in
that
same
category,
and
that
if
you
lose
those
lose
that
data,
it's
not
going
to
affect
running
Cassandra.
B
B
B
So
the
sections
I
have
here
are
the
same
as
last
week
with
respect
to
the
examples
of
cast
operator
in
casts
cop
and
and
and
that's
that
was
what
got
me
started,
thinking
that
they
do
the
same
thing
with
respect
to
configuring.
The
logging
sidecar
in
terms
of
what
wait
with
the
end
result
cast
operator
just
doesn't
expose
it
in
the
spec
cast
cop
does
and
so
I
kind
of
spend
a
bunch
of
time.
B
So
you
have
the
the
fields
here
for
specifying
in
both
of
these
types,
both
lease
trucks,
the
volume
name.
The
name
of
the
things
are
the
for
the
name
of
the
volume
out
the
path
for
the
persistent
data
you
have
one
of
things,
I
saw
with,
like
so
I'm
trying
to
build
off
ideas
in
both
the
trans
operator
in
casco
in
calves
operator,
I'm.
Sorry
in
kescott,
you
could
for
the
data
volume
varlyn
Cassandra,
the
persistent
persistent
volume
claim
is
not
directly
exposed.
B
Instead,
there's
there's
properties
for
specifying
the
capacity
in
the
storage
class,
which
is
nice
on
the
one
hand,
because
it's
its
provides
a
nice
abstraction
if
all
I
care
about,
particularly
if
I'm,
just
getting
started
and
and
I
just
want
to
spin
something
up
quickly,
say
well:
I
I
need
a
10
gigabit
volume
with
you
know
its
SSD
sort
and
I'm
done.
Don't
have
to
worry
about
the
vaad
content.
On
the
other
hand,
if
I
know
a
little
bit
more
I
have
a
little
bit
more
requirements
and
want
to
have
the
flexibility.
C
B
So
let
me
scroll
down
to
come
back
to
log
data
and
I'll
answer
that,
since
you
raised
the
question:
okay,
so
so
I've
got
a
or
just
config
and
I've
you
have
so
I'm
gonna
go
so
you'll,
see,
there's
a!
Let
me
start
with
the
you.
Have
this
Cassandra
dated
her
and
that
would
correspond
to
varlyn
Cassandra,
so
the
on
the
one
hand,
I'm
striving
for
providing
maximum
flexibility
where
to
allow
somebody
to
specify
its
minimal
information
as
possible,
while
also
being
able
to
giving
them
maximum
configuration
maximum
configurability.
B
So,
by
default
the
operator,
if
nothing
is
else
so
notice.
These
are
all
pointers.
So
if
nothing
is
specified
well
that
the
operator
by
default
would
initialize
this
field,
the
Cassandra
data
dear
to
what
the
operators
basically
do
today,
let's
say
the
operators,
whether
you're
talking
Kaz
operator,
but
at
this
point
not
too
conservative
the
implementation.
So
whether
it's
the
way
Kaz
operator
sets
up
the
persistent
volume
or
the
way
casts
cop
or
something
or
or
the
way
it's
the
cluster
or
something
between
it's
just
there's
a
default.
B
Your
defaults,
you
don't
go
anything
else
and
then,
if
you
want
to
configure
the
default
there,
you
can.
The
I
was
chatting
with
Jim
and
he
pointed
out
to
me
chat
image
of
him
over
the
weekend
and
he
pointed
out
the
need
we
do
at
a
minimum
do
need
well,
you
do
in
a
minimum,
need
to
specify
storage
class,
so
I
added
a
default
storage
class
that,
if
you
just
want
to
use
so
that
we
could
be
used
globally
for
the
vines
or
it
could
be
overridden
for
each
of
these
fields.
B
B
I
think
you
do
I
hope
I
do
okay.
So
no,
let's
say
we
don't
want
to
that.
Let's
say
we
want
to
put
this
TDC
we
want
to
do.
We
want
to
do
use
the
change
data
capture
and
we
want
to
put
the
CDC
log
on
a
separate
volume,
and
so
here,
I
initialized,
the
CDC
Roger
property,
where
and
I'm
only
specifying
the
name
of
the
volume,
the
past
capacity
in
storage
class.
But
again
you
could
also
set
Z.
B
And
then
so,
if
I
did
so,
if
I
did
that,
then
the
operator
should
look
at
that
and
say
know
that
it's
basically
no,
you
still
use
the
default
data
directory,
except
for
the
CDC
raw
directory
and
put
and
create
the
separate
volume
for
it.
And
then
you
know
do
the
same
for
each
of
the
persistent
data
directories.
B
B
And
then
the
oh,
nothing
I,
don't
think
I!
Listen
here
you
know,
and
then
you
know
the
operator
I
think
you
could
do
some
validation
checks
to
make
sure
that
the
the
mount
points
you
don't
have
overlapping
mount
points
so
that
you
know,
if
there's
any
problems
you
find
those
out
earlier
rather
than
deploys
things
and
see
they
met
where's
my
direct
where's
my
date,
my
my
directory.
B
So
talk
about
the
first
property
or
the
sidecar
enabled
so
Cyril
in
the
last
meeting
spent
some
time
talking
about
with
the
the
logging
sidecar
that
they
have
and
again
kaz
operator
does
the
same
same
thing,
just
it's
not
exposed
in
the
manifest
and
so
think.
Let's
and
I
think
well
gosh.
You
know
and
and
I
spent
some
time
I
think
been.
There's
some
good
discussion
around
blogging
in
general
I
thought.
Well,
you
know
I
used
to
be
able
to
simply
have
a
flag,
I
could
say
well
for
the
a
this
particular
log
file.
B
You
know
I'd
like
that
sidecar,
so
that
I
can
use
standard
tooling
for
accessing
logs.
You
know
when
I
say
standard
tooling,
joob
CTL
in
particular,
I
think
I
mentioned
I.
Think
the
ideal.
The
most
robust
logging
solution
is
going
to
be
some
sort
of
a
whether
it's
LK
or
if
you
like
your
GK
or
if
you
like,
stackdriver
and
gke,
would
you
know
some
centralized
logging,
but
if
I
want
to
be
able
to
use
the
tools
available
with
your
cheap
CTL,
then
lock.
The
logging
side
card
makes
that
the
easiest
way
to
do
that.
B
B
Again,
so
to
provide
the
flexibility,
let's
say
so,
if
I
want
again
well,
not
just
well,
I
need
to
be
able
to
have
to
mount
the
volume
to
make
it
episode
mount
the
the
logging
log
files
in
a
volume,
that's
accessible
to
other
sidecar,
whether
it's
the
long
side,
car
or
or
something
else
that
I
mean
made
accessible,
that
volume
source.
That
could
be
a
persistent.
You
know
it
could
be
a
any
supported
volume
source.
So
you're,
not
you're,
not
limited,
limited
there
and
similar
to
what
the
persistent
data
explicitly.
B
Where
is
it
oh
here
it
is,
there's
the
Cassandra
log,
dear
property,
which
just
is
the
default
vara,
log,
Cassandra
and
and
whatever,
and
that
would
if,
in
the
absence
of
any
other
properties
being
set,
then
the
operator
would
my
thinking
is.
The
operator
would
initialize
this
to
the
default
settings
and
again
for
the
purposes
of
where
we
at
where
we
are
at
today
with
the
discussions.
That's
more
implementation,
I'm
thinking
about
design,
and
so
it's
whatever.
B
And
I
wanted
to
expose
the
cassano
debug
log.
So
here
is
an
example
of
how
I
configured
the
debug
log
and
mounted
that
to
a
an
empty
dear,
empty
dirt
volume
so
that
maybe
I
can
do
some
additional
processing
on
it
and
then,
if
I
wanted
to,
then
I
could
set
the
the
sign
card
enabled
flag
to
true.
If
I
wanted
to
have
the
logging
sidecar
for
it.
B
Oh
and
part
of
another
motivation,
I
was
thinking
about
with
the
the
having
with
how
I
kind
of
said
structured.
This
is
in
one
of
my
first
experiences,
which
was
with
using
Sandra
and
gratis,
which
was
going
back
incriminate
easy
in
kubernetes
years,
a
while
was
pre
stateful
sets
and
your
things
were
a
bit
rough
rough
in
terms
of
understanding
of
things
and
how
to
set
things
up
and
configure
them.
B
We
had
it
ran
issues
with,
so
it
was
a
struggle.
There
were
times
we're
trying
to
diagnose
problems,
and,
let's
say
the
Cassandra
node
was,
was
crashed
without
a
memory
errors
and
asked:
can
we
get
a
heap
dump
and
we
did
had
we?
We
had
very
little
ways
of
configuring
things
so
that
we
could
put
the
heap
dump
somewhere.
So
we
wouldn't
lose
him
when
the
pod
was
killed,
and
that
was
critical
to
try
and
try
to
troubleshoot
the
problems.
And
so
you
know
we.
B
It
was
kind
of
like
playing
whack-a-mole
trying
to
get
a
hold
of
files.
You
know
before
before
they
disappeared
so
wanting
to
provide
that
flexibility
for
The
Situation's
or
doing
any
kind
of
maintenance
and
operations,
and
then
I
have
the
last
property.
So
before
I
get
to
the
last
property,
there
may
be
other.
B
A
Was
probably
the
controversial
thing
from
last
meeting
was
the
whole
log
directory
and
how
many
other
operators
handle
log
directories.
D
B
C
Thing
yeah
and
I'm
not
disagreeing
with
the
approach.
I'm
just
saying
kind
of
what
I've
seen
I
couch
base
is
similar
couch
base
has
a
ton
of
different
logs
because
there's
a
ton
of
different
Damons
that
are
all
running
together
and
one
thing
we
knew
is
we
in
our
in
our
image
for
that?
We
basically
cap
them
all
out
to
standard
out,
basically
the
ones
that
we
want
and
so
we're
just
kind
of
like
piping
them
all
just
a
weekend,
because
we
do
ship
everything
out
to
a
centralized.
E
So
this
is
like
one
of
those.
This
is
what
I
was
trying
to
chime
and
say.
This
is
like
one
of
those
places
where
Cassandra
itself
is
not
very
friendly,
so
like
way
to
describe
couch
based
doing
that.
That
sounds
like
they
have
like
sort
of
the
minimum
bar
hit,
but
all
these
decisions
start
to
run
together.
So
what
happens?
Is
we
have
in
the
cast
operator?
We
use
the
management
API,
so
the
management
API
Owens
standard
out
and
we
were
debating
what
to
do
with
that
is
like.
E
Do
we
like
passed
internet
down
and
then,
like
you,
kind
of,
don't
know
which
process
some
of
the
logs
came
from
exactly
it
just
started
to
get
confusing
so
yeah.
So
if
you
don't,
if
you
don't
enable
some
minimum
config,
you
can't
even
start
those
side,
caller
sidecar
containers,
because
you
at
the
least
have
to
setup
like
the
empty
durval
um--
that,
like
all
your
sidecar
containers,
can
share
and
and
drop
the
logs
there,
but
yeah.
If
Cassandra
I,
don't
know
I,
don't
know
how
Cassandra
could
change
that
much
like
just
writing
a
standard.
A
The
log
back
XML
file
I,
mean
this
is
what
I
hate
this
answer,
but
it's
like
that.
The
the
portability
the
the
configurability
is
kind
of
built
into
just
how
Java
does
logging
like
if
you
need
to
dump
it
into
a
syslog
directory
or
syslog
server,
you
can
do
that
if
you
want
to-
and
actually
this
is
I
think
this
is
a
data
dog
works
as
a
matter
of
fact.
Splunk
has
a
listener,
you
know
so
think
of
places
where
people
collect
log
files
law.
A
A
A
top
problem
is
and
John
I
know
you
do
a
lot
of
consulting
in
open
source
side
yeah,
it's
like
what
do
we
do
with
our
logs
cuz,
unfortunately,
most
of
the
bad
things
that
happen
show
up
in
the
logs
first
and
they
gotta
pay
attention
to
him.
So
just
saying
well,
well:
dump
it
into
a
ephemeral
directory
on
a
container.
Just
doesn't
seem
like
a
good
idea.
B
I'm,
not
sure
I'm,
not
sure,
well,
I,
think
so
the
way
my
kind
of
view
of
the
world
is
you.
Let's
say
you
don't
do
anything
and
you
just
have
var
log
to
Sandra,
and
you
know
that's
kind
of
like
the
doing
the
least
amount
of
work
to
do
anything
to
improve
logging.
So
you've
got
to
you
know:
exact
use
commands,
maybe
with
keeps
detail
exact
to
copy
the
log
files
out
of
the
container,
which
is
certainly
not
not
great
yeah.
B
If
it's
not
understand,
what's
going
to
spit
out
yeah
more,
but
the
and
I'm
not
sure
like
the
cumulative,
it
does
store,
like
the
the
current
logs
and
I'm,
not
sure,
if
for
how
long
or
how
much,
because
it's
a
human
level
configuration
but
then
as
well
as
it
stores,
if
the
the
previous,
if
the
container
has
restarted
it'll
contain
the
last
set
of
logs
as
well.
So
that's
better
than
just
the
out
of
box
a
little
bit
better
and,
of
course,
sending
your
logs
to
a
logging.
A
A
We
have
a
lot
of
activity,
that's
already
run
happen
around
Cassandra
and
kubernetes,
but
as
where
it's
almost
like
we're
looking
at,
like
you
know,
looking
at
all
the
technical
debt,
that's
accumulated,
one
of
the
things
that
I
think
is
really
important
to
recognize
is
the
fact
that
there's
really
been
very
little
talked
about
and
changing
Cassandra,
and
this
is
a
good
example.
Cassandra
logging
facilities
were
pretty
much
baked
back
in
2010
ten
years
ago.
A
This
has
not
changed
a
bit
and
that's
back
when
I
used
like
Splunk
to
dump
all
my
Cassandra
logs
and
I.
Think
that's
just
something
that
we
need
to
think
about
from
a
project
standpoint
from
a
Cassandra
project.
Standpoint
is
what
needs
to
modernize
in
the
way
that
it
logs
new
facilities,
and
is
you
know
this
is
a
change
that
maybe
needs
happen
on
Cassandra
side.
E
E
Yeah,
it
would
be
like
why
do
you
need
all
this
like
just
right
to
standard
out
and
set
up
your
logging
infrastructure
with
your
kubernetes,
so
I
think
it's
really
that
say
why
is
there
system
log
and
a
debug
log
and
and
then
you
got
the
Java
stuff
underneath
it's
like?
Could
we
make
something
that,
like
scoops,
those
Java
logging
events
into
the
cassandra
log
like
try
to
unify
up
this
these
streams,
and
maybe
they
go
to?
Maybe
they
should
be?
Maybe
it's
better
treatment
of
standard
out
versus
standard
error,
I
think
I.
E
Think
kubernetes
offers
some
pretty
good
support.
For
that
and,
like
I
said
there,
others
sort
of
I,
don't
wanna,
say:
Oh
old
are
necessarily
bad,
but
there's
other
stuff
that
was
like
embedded
well
free,
kubernetes,
that's
adapted
a
little
bit
like
if
you
just
look
at
the
standard
engine,
X
container,
like
they've,
adapted
access,
log
and
error
log
to
fit
so
makes
me
think
it
could
be
done.
E
I
guess
kind
of
a
product
concern,
but
I'll
mention
it
because
it
came
up
was
also
the
occasions
when
maybe
sensitive
data
gets
logged
and
I,
don't
know
how
other
databases
are
handling
that,
probably
not
that
differently.
I
mean
I,
know
my
sequel
really
well,
and
if
your
put
somebody's
social
security
number
in
a
unique
key
and
you've
inserted
twice,
you
get
a
unique
like
the
logs.
Clearly
printed
out,
it's
like
you
know,
unique
key
violation.
You
know
print
it
out,
but
I
know
that's
something.
E
We've
been
concerned
about
and
that's
sort
of
another
logging
subject,
but
I
think
I
think
just
better.
You
know
making
it
so
this
isn't
necessary
is
like
it
would
be
a
nice
thing
to
get
upstream
and
Cassandra.
So
it's
just
like
hey,
maybe
Cassandra
does
more
like
at
the
more
system's
level
to
like
get
those
GC,
vlogs
and
and
JVM
logs
kind
of
sucked
into
some
kind
of
standard
error,
configure
ibly
and
then
there's
no
need
for
system
log
and
debug
log.
E
A
Think
it's
funny
that
we
haven't
is
like
almost
like
as
an
entire
ecosystem.
People
who
run
infrastructure,
we
really
haven't
moved
on
from
logging
as
what
it
is
I
mean.
Syslog
is
still
kind
of
the
way
that
it
works.
Even
though
you
have
agents
like
Danny
dog,
okay,
Splunk,
let's
this
log
is
really
that
but
I
guess,
maybe
they
we
don't
need
much,
or
this
is
modernized.
A
B
Me
no
I
I
think
it's
a
great
idea,
however,
seeing
how
some
discussions
on
the
mailing
list
seem
like
they
turn
into
long
chapter
books
be
good
to
make
sure
like
have
a
solid
some
ideas.
That
least
set
the
tone
for
the
conversation
versus
just
being
open-ended,
whether
whether
it's
hey
for
like
to
see
logging
changes
and
we'd
like
to
see
these
changes.
Here's
a
bullet
point
list,
for
you
know
better
support,
creating
huge
cases.
B
A
B
A
On
well
alright
I,
just
to
kind
of
put
a
cap
on
that,
then
I
think
that's
something
that
we
just.
We
need
to
consider
like
what
is
that?
What's
the
opinionation
of
the
old
kubernetes,
the
new
kubernetes
posse
rolling
into
town
about
logging,
like
what
kind
of
changes
would
be
done
to
Cassandra,
to
make
it
more
friendly
in
that
regard?
I'll
just
leave
it
at
that.
B
Well,
yeah
just
try
to
keep
just
the
slack
channel
active
any
any
thoughts
in
there.
I'll
make
a
point
to
keep
monitoring
that
around
the
unplugging.
Anything
for
me
to
look
at
I'm
happy
to
try
to
drive,
drive
or
consolidate
information
discussion
as
much
as
as
much
as
I
can
so
with
that
I'll
switch
gears
to
configuration.
B
The
with
a
brief,
very
brief
description
of
each
and
then
so
goals,
so
spatial
sets
one
of
the
I.
Think
one
of
the
things
that
the
areas
where
they're
kind
of
come
up
short
is
around
configuration
labels.
That's
provide
a
way
to
configure
persistent
volumes
for
the
pods
that
make
up
that
stateful
set,
but
they
don't
provide
any
such
or
similar
constructor
mechanism
for
config
files.
B
So
that's
was
a
good
gap
and
so
some
prior
meeting
when
we
were
going
through
looking
at
examples
of
the
examples
of
specs
with
different
operators,
we
were
looking
at
how
we
looked
at
an
example
of
how
insta
cluster
allows
you
to
configure
settings
an
example
of
how
kescott
and
then
also
with
Kaz
Lambert.
There's
a
lot
of
really
and
and
because
it's
left
out
from
statement
sets.
People
have
to
come
up
with
some
really
innovative
approaches
for
doing
things
and
I
didn't
repeat
prior
examples.
B
This
is
like
citizen
early
drafted,
but
I
do
have
an
example
from
Kaz
operator,
because
I
think
that
Kaz
operator
kind
of
really
hits
the
nail
on
the
head
with
respect
to
what
the
end
user
should
see
or
idea.
I
have
in
here
the
ghouls.
You
know
I
think
a
goal
when
I
think
of
if
I'm
running,
Cassandra
outside
of
kubernetes
and
I
need
to
update
a
I
need
to
provision
Cassandra
I,
simply
whether
it's
by
hand
in
VI
or
using
whatever
tools,
puppet
chef,
Ansel
or
whatever
the
case
may
be.
B
I
can
simply,
however,
I
need
to
do
it
edit
and
update
my
config
files.
Of
course,
that's
not
the
way
the
world
works
in
kubernetes,
so
that's
the
ideal
on
the
one
hand,
and
Kaz
operator,
gets
us
really
close
to
that
with
the
the
CRD,
where
we
see
here
where
you
just
simply
almost
a
one-to-one
mapping
of
property
names
from
Cristiano
to
it,
from
the
animal
file,
for
example,
from
the
Cassandra
animal,
so
I
asked
Jim,
and
then
you
also
see
the
JVM
options
listed.
B
E
E
We
we
thought
and
and
in
some
ways
I
still
think,
there's
there's
some
merits
using
config
maps
for
this,
but
it's
just
sort
of
another,
while
it's
probably
better
for
some
cases,
it's
definitely
like
a
much
bigger
barrier
to
entry
and
makes
reconciling
things
a
little
trickier,
and
it's
just
another
point
where
things
can
fail
right.
So
if
we're
pushing
down
the
config
config
Maps,
it's
more
slightly
more
robust
than
config
maps,
which
are
ultimately
eventually
consistent
at
the
end
of
the
day.
E
So
how
do
you
want
the
config
Maps
like
if
you're
changing
the
seeds
and
then
you
want
to
like
go
and
say
hey
the
seeds
are
updated,
like
you-you-you
need
to
start
building
a
lot
of
stuff
to
figure
out
if
your
config
map
is
updated
and
most
of
these
services
in
kubernetes,
which
I'm
sure
probably
most
people
here
know.
But
it's
fine,
if
not
like.
There's
a
bunch
of
operators
that
just
deal
with
this
one
problem
of
like
hey:
when
should
I
bounce?
E
When
should
I
bounce
my
pods
to
make
sure
they
get
a
new
config,
Maps
or
what-have-you,
which
we
don't
want
to?
Do?
We
don't
want
to
get
into
that?
In
fact,
we
can't
restart
a
known
necessarily
to
get
some
of
the
simplest
config
trains
that
are
read
on
the
fly,
maybe
especially
seeds,
to
be
considered
here
so
anyway,
so
we
built
so
we
started
from
there
and
we
built
down-
and
we
said,
hey,
let's
just
kind
of
expose
the
values
where
they
are.
This
means,
like
all
the
Cassandra
documentation,
is
also
a
documentation.
E
Basically
for
this
dialect
or
whatever
have
the
configuration
like
we're.
Not
there
there's
a
couple
things
in
here
that
are
there
shortenings,
especially
as
you
get
outside
of
Cassandra
amel,
but
largely
you
know,
it
corresponds
directly
to
the
same
files
from
upstream.
So
if
you
discover
an
answer,
for
you
know,
vanilla,
Kassandra,
tarball
on
your
laptop,
your
answer
will
apply
here
and
you
can
easily
check
that
what
else
so
some
of
the
glue
of
how
this
is
is
assembled
is
there's
a
in
it
container
that
fires
and
Pross.
E
So
the
operator
passes
this
data
plus
some
other
stuff,
like
the
data
center
name
and
rack
name,
we
don't
we're
not
requiring
like.
There
are
values
that
the
operator
knows
how
to
push
down
per
rack
and
we
passed
that
all
via
environment
variables
to
the
content
container
in
the
connect
container
comes
up.
Read
that
uses
a
bunch
of
other.
So
this
was
like
I
mentioned
in
a
previous
one
of
these
calls
we
had
some.
E
We
have
a
lot
of
code
around
this
aspect
of
starting
the
database,
so
we
repurpose
some
of
it,
and
so
we
say,
oh,
we
know
the
exact
version
of
Cassandra
are
on.
We
know
what
the
defaults
are,
so
we
render
out
the
full
defaults,
it's
like
a
template,
and
then
we
put
push
this
down
as
more
template
data.
E
For
like
your
back
to
hacking,
you're,
not
really
a
user
anymore,
you
have
to
become
a
developer
or,
as
this
I
think
we
could
keep
people
firmly
in
their
user
seat
and
say:
hey
I,
know.
There's
this
JVM
argh
I
want
applied.
I,
know,
there's
this
minus
D,
Kassandra
parameter
I
want
applied
and
like
I,
just
put
it
in
an
additional
JVM
ops
and
it
just
sales
sales
through
a
bunch
of
layers
and
I
can
stay
firmly.
In
my
like
just
reading
the
docs
and
changing
the
config
mindset,
yeah.
C
E
That's
a
great
point:
yeah
yeah,
the
more
we're
pushing
more
and
more
configuration
declarative
as
well
for
our
our
stuff,
it's
bubbling
through
a
straw,
but
it's
bubbling
through
the
operator
as
well
so
like.
If
you
look
at
cast
operator,
we
just
merged
something
we're
gonna
cut
it
release
today.
I
need
to
get
off
the
call
and
finish
it,
but
we're
letting
users
declare
that
they're
Cassandra
users
declare
specify
their
Cassandra
users
declaratively,
and
we
want
to
get
key
spaces
coming
and
grants
in
there
too.
E
So
more
and
more
of
your
like
test
test,
harness
can
just
be
like
this
is
how
it
should
be,
and
then
you
don't
really
have
to
it
make
sure
it.
So
maybe
you
don't
have
to
like
worry
about
when
the
server
is
up
to
apply
it's
just
more
stuff,
more
glue
that
you
know
the
developers
don't
need
to
worry
about.
B
Yeah
I
need
to
update
the
goals
here
too
I
think
I
could
have
summarized
what
I
have
here
says,
make
it
declarative
and
that
would
suffice
for
the
goal
and
the
the
the
background
that
was,
it
was
really
helpful,
helpful,
Jim
and-
and
it's
so
oh
I
do
have.
I
did
put
some
questions.
You
know,
as
I
was
kind
of
thinking
through
and
I.
So
are
there
properties
that
the
shouldn't
be
allowed
to
set
here,
for
example
like
ports
and
what
happens
to
so?
B
E
Ports,
you
can
probably
change
without
harm
in
general,
yeah
I
want
to
just
block
all
that
is
you're
in
kubernetes.
There's
no
sense
of
you,
like
micromanage,
unless
someone
can
show
me
a
great
example
of
when
it's
useful,
maybe
under
certain
security
constraints
or
whatever,
there's
a
good
reason,
but
I
mean
we're
not
binding.
The
port's
under
a
thousand
I
think
there's
some
restrictions,
even
though
you
can
just
do
that
kind
of
freely
in
a
container
some
some
of
these
runtimes
make
that
prohibited,
but
yeah
in
general.
E
There's
plenty
of
places
where
you
could
like
try
to
overthink
this
and
just
break
everything
like
I
think
we
were
chatting
John
I
said
like
if
you
use
the
initial
tokens
like
good
luck,
I,
don't
think
it
will
work
like
we
should
we
should
let
blacklist
or
you
know,
deny
reject
certain
thing.
You
know
like
think
things,
it's
just
why
you're
doing
this
yeah
I
have
some
outstanding
stuff
for
that
for
validation
that
we
haven't
gotten
to,
but
that
yeah,
what
I
would
my
approach
would
just
be
like?
B
Should
there
be
an
off
property
like
I
know,
I
lay
in
it
like
the
Postgres
proper
one
of
the
Postgres
operators
as
a
where
you're
off
property
section
for
for
divining
for
configuring,
whatever
earth
related
settings,
it
allows
you
to
configure
so
that
wasn't
want
the
other
question.
I
had
you
know
like
I,
don't
know
other
certain
properties.
It
makes
more
sense
that
she
be
called
out
more
explicitly.
E
E
A
couple
of
those
or
just
not
in
these
examples
but
like
specifying
the
garbage
collector,
is
kind
of
a
complicated
mess
of
like
one
or
two
or
three
JVM
arcs.
We
need
to
put
up
better
docs
for
that,
but
yeah
so
like
things
like
that,
do
get
abstracted
into
something
more
like.
So
what
you
see
under
javion
JVM
options
here,
where
it's
like
we're
writing
out
max
heap
size.
You
can
also
use
minus
X
and
M
X,
but
I
think
probably
better.
E
Yeah,
you
don't
have
to
use
those
nicely
named
ones,
but
I
find
it
a
little
easier
to
read.
The
this
SH
and
bash
scripts.
Estar
Cassandra
are
are,
are
aware
and
yeah.
So
I
I
need
to
look
at
the
scripts
that
the
config
builder
emits
out
for
Cassandra,
I'm,
pretty
sure
they're
exactly
the
same
as
upstream,
because
we
just
kind
of
take
the
upstream
stuff
and
insert
some
templating
placeholders
into
it.
But
I
would
probably
want
to
double-check
that
I
look
at
the
SE
a
little
more
often,
but
but
about
half
and
half
and.
B
B
E
B
E
Of
its
reused
for
the
tools
right
so
like
no
no
tool
and
Nancy
qsh
at
least
no
tool
but
I,
think
oh
and
all
the
SS
table,
yeah,
manipulating
things
use
those
files
to
so
I
understand
the
purpose
of
yeah.
Why
it's
chopped
up
so
much
but
I,
don't
know!
I
I
came
in
to
working
at
the
aesthetics
completely
new
to
Cassandra
and
I'm.
Just
like
you
know,
I
tried
it
I
think
I
got
to
this
with
fresh
eyes
and
I'm
like
this
is
this
is
challenging.
E
B
A
B
There's
I
mean
there's
other
areas
of
the
other
areas
of
the
you
know.
As
far
as
the
spec
you
go
through,
but
I
mean
we've
talked
about,
topology
talked
about
storage
or
at
least
enough.
You
know
there's
enough
that
we
just
looked
at
that.
If
we
can
or
I
think,
is
probably
worth
iterating
on
just
those
areas
to
save
and
then
can
we
move
it
further
along
on
those
areas
and
say
well
start
talking,
some.
A
E
I
think
I
think
you
know
it
keeps
being
questioned
each
week,
but
I
think
the
volume
stuff
has
been
the
best
example
of
why
we
need
to
try
to
combine
our
resources
and
not
just
say:
oh,
let's
just
start
with
somebody's
code
that
we
already
have,
because
that
I
know
the
parts
I
think
from
cast
operator.
You're
great
and
I
told
John
I'm
like
well
volumes
is
not
our
strong
point.
I
know
why
we
have
it
so
simplified.
But
right
you
know,
yeah.
A
E
What
John
has
here
is
like
a
way
more
fully
specified
thing.
That's
sort
of
like
it'll,
be
tedious
to
program
up,
but
it
will
pay
off
that.
You
know
everybody's
use
case.
That's
pretty
valid
can
be
hit
by
this
versus,
like
and
I'll
say
like
we
used
to
have
some
different
stuff
for
volumes,
and
it
was
very
confusing
and
it's
why
we
simplified
it
and
that's
why
I
was
also
like.
A
B
E
F
F
No,
no
sorry
like
I
didn't
talk
doing
the
great
stuff,
I
have
to
say
that,
like
I'm
not
able
to
join
or
4:30
on
the
Thursday
sync,
so
today,
I
got
the
opportunity
and
yeah
like
I
would
say.
Yeah
like
these
are
like
information,
details
and
I
would
really
like
to.
But
one
thing
that
I
think
we
can
already
start
is
like
I
would
like
to
add
to
digitize
some
base
for
the
some
sort
of
testing
or
conformance
testing
right,
so
where
we
say
like
hey
yeah.
F
This
is
the
operator,
and
we
say
this
is
like
we
ready
for
like
some
scoreboard
that
the
only
to
it
we
have
set
up
for
the
customer
operators
and
it
has
to
pass,
and
we
can
you
like
work
on
some
standardizing
that
one
but
yeah
like
on
other
parts
they're
like
I'm,
also
like
it's
specially
like
today.
What
you
present
is
like
really
great
stuff
and
later
I
will
comment
on
that.
Like
or
I
have
some
question
on
the
storage,
and/or
logging,
but
yeah
like
I.
Don't
have
still
all
the
background
area,
yeah.
E
A
E
A
A
It's
not
in
there
so
I
need
to
go.
Make
sure
that
happens.
It's
not
in
the
calendar,
so
I'm
gonna
go
out.
That's
what
that
needs
to
happen.
Yeah.
It
should
be
I'll
figure
out
what
time
it
should
be.
Anyhow,
I
will
I
will
get
back
together,
I
think
we're
gonna
be
doing
it.
10:00
a.m.
yeah
10:00
a.m.
this
time,
Pacific
time
so
it'll
be
at
the
same
time
next
week
on
Thursday
this
time.