►
From YouTube: 2020-05-19 Apache Cassandra Contributor Meeting
Description
Update on CEP-2. Cassandra Kubernetes Operator
Question and Answers on Cassandra build CI/CD
A
All
right,
everyone's
famous
now,
okay,
let's
get
this
going,
we
had
a
couple
of
things
on
the
agenda
item
I
just
wanted
to
I
put
my
my
agenda
item
on
there
and
it's
just
a
real
quick
update.
We
have
a
the
Casandra
kubernetes
connector
cig
meeting.
Thursday,
there's
been
quite
a
bit
of
work
done
on
this
EP
draft
document.
That's
sitting
in
Google.
Many
thanks
to
Mick
for
helping
me
out
with
a
lot
of
stuff
around
CEP
I.
Think
that's
gonna,
be
I'm.
Gonna!
Try
to
do
some
cleanup
on
the
dock!
A
There's
some
other
folks
that
want
to
do
that,
but
essentially
I
think
what
we're
gonna
do
is
is
try
to
get
this
thing
grounded
out
on
Thursday
like
okay,
are
we
had
a
place,
and
can
we
start
talking
about
moving
forward
here
and
the
like,
like
what
is
moving
forward,
mean
I?
Think
from
what
we've
talked
about
is,
since
there
are
so
many
operators
out
there
already
that
this
needs
to
be
kind
of
a
more
of
a
like
a
kubernetes
Kassandra
mini
community
discussion
about
how
how
that
even
even
is
gonna.
A
Look
before
we
start
putting
all
this
energy
right
into
the
Cassander
project.
I
think
it's
a
that'll
be
a
lot
of
noise
that
that's
good
I,
think
that
is
the
right
move.
Just
after
talking
to
a
lot
of
folks
about
this,
it's
like
well
kind
of
sort
out
what
this
actually
is
and
then
add
that
will
be
the
proposal
and
there's
not
only
a
specification,
but
some
idea
of
what
the
code
would
look
like
as
well
that
that's
where
we're
at
right
now,
but
any
discussion
around
that.
B
Yeah,
one
of
the
questions
that
I've
had
is
if
we
had
a
chance
to
talk
with
some
people
who
have
been
successful,
running
Cassandra
and
kubernetes,
to
figure
out
what
some
of
the
harsher
lessons
they
might
have
learned
about
it.
It's
it's
important
to
be
able
to
adapt
our
current
cyber
model
that
many
people
are
using
to
q,
but
I'm
also
curious
what
rough
spots
people
have
found
along
the
way.
A
It's
good
Scott,
actually,
that's
what's
been
encouraging
is
that's
pretty
much
what's
been
in
these
sig
meetings.
Is
we've
had
a
mostly
significant
amount
of
people
that
are
running
organizations
that
are
running
Cassandra
and
kubernetes.
Now
we've
only
had
a
couple
of
words
like
it
was
interesting
as
Bloomberg
and
New
Relic
both
showed
up
as
and
they're
interested
in
doing
it.
So
it
was
really.
There
was
some
interesting
interactions
between
the
people
who
are
running
a
kubernetes
operator
in
their
own
way
asking
well.
What
do
you
want
if
you're
interested
in
this?
A
C
Have
a
question
on
the
clock
process,
which
is
likely
slightly
basic
here.
It's
a
lot
of.
Firstly,
two
goals
that
we
are
talking
about
have
alignments
on
from
CP
one
as
well.
There's
a
lot
of
overlap
there.
So
did
we
put
a
thought
process
of
contributing
those
towards
CEP
1
so
that
kubernetes
operated,
not
2
minute
is
deployments?
Can
utilize
as
well.
C
So,
like
one
thing,
I
particularly
see
which
is
like
CP,
1
and
CP
2
talks
about
is
about
running
well
commands
and
running
your
desired
state
commands
like
notebook
maps
which
can
really
bend
which
can
really
benefit
non-punitive
operators
as
well,
and
that
one
of
the
goals
for
keeping
that
in
CP
one
was
there's
a
lot
of.
As
for
like,
for
example,
you
know
this
desired
way
to
run
a
note,
no
to
command.
C
So
I
would
like
to
attend
no
to
Teaneck,
mind
on
my
interfere
beside
requesters,
putting
their
desires
to
it
and
kubernetes
operated,
taking
care
of
it
and
getting
it
and
all
the
on
the
process
in
the
same
way
in
October.
It's
an
element,
you
know,
that's
something.
That's
really
useful
as
well.
So
I
was
wondering
if
we
can
contribute
all
of
that
to
CP
one
and
benefit
both
both
communities.
A
D
D
So
for
for
for
for
the
operator
to
to
execute
the
the
app
action
on
each
node
is
not
that
big
of
a
deal
but
I
get
the
point
of.
It
would
be
nice
to
have
that
kind
of
action
built
in
I
just
worry
about.
How
do
you
secure
it?
Well,
so
so
that
you
don't
like
accidentally,
do
the
wrong
thing,
because,
because
if
the
option
is
basically
call
this
one,
you
you
URL
for
just
one
node
and
a
different
one
for
for
all
nodes
or
like
just
just
one
rack.
C
I
mean
we
discussed
about
security
in
CEP.
One
side
card
is
well
within
I
mean
it's
always
subjective
well
between
the
security
boundaries
having
having
being
on
the
instance
along
the
Cassander
process
and
also
having
attended
Z's
to
be
laid
out.
So
if
I
mean
you
would
be
authenticated
and
authorized
against
calling
cycle
process
inside
commences
all.
A
A
A
C
A
C
D
A
F
You
know
one
thing
I
wonder
about,
and
this
actually
came
up
from
a
phrase
you
use
Scott
recently,
which
is
the
whole.
Let
a
thousand
flowers
bloom
approach
like
on
the
testing
side
is
like
if
we
have
seven
different
kubernetes
operators
out
there
and
each
of
them
have
a
different,
distinct
point
of
view
on
what
to
optimize
for
an
operator
totally
makes
sense.
F
I
would
wonder
if,
with
a
sidecar,
we're
in
that
same
boat,
like
whether
there's
one
clear
set
of
correct
compromises
for
a
sidecar,
that
we
should
have
a
a
sidecar
or
whether
or
not
it
makes
sense,
have
an
ecosystem
of
sidecars
I'm
not
putting
on
a
flag
in
either
camp.
But
as
with
anything
like
this
is
going
to
take
significant
effort
for
us
to
figure
out
the
right
compromise.
F
G
H
G
C
Yeah
so
I
by
no
means
I'm
saying
no
to
buildings
too
many
Tigers.
As
long
as
there
is
one
difference,
implementation
that
people
can
benefit
from,
but
not
putting
community
in
a
spot
way
well,
I
have
a
sidecar
which
tells
XYZ
and
another
sidecar
which
does
ABC
good,
should
I
have
more
like
what
is
right.
C
F
G
The
way
I
think
we
should
be
approaching
is
through
the
CIP,
because
there
are
lot
of
issues
which
have
happened
in
the
past,
specially
with
automation,
because
there
are
more
people
who
are
involved
in
all
the
bitching.
If
you
are
talking
about
a
very
specific
storage
engine
change,
you
will
only
have
five
people
who
will
commit
on
it.
But
if
you
are
talking
about
sidecar,
you
will
have
100
people
comment
on
it,
because
it's
just
more
people
understand
that
area
more
than
very
specific
area
and
code
base.
G
You
know
like
if
you
send
a
start,
a
thread
on,
so
you
will
get
three
replies.
If
you
start
a
thread
on
running
repair,
you
will
get
hundred
opinions
right,
which
is
good
right.
It's
not
I'm,
not
saying
it's
bad,
but
what
we
need
to
do
is
basically
have
that
process
so
that
things
move
forward
and
as
a
project
we
should
actually
make
sure
that
the
progress
is
being
made.
Then
actually
people,
you
know-
and
you
know,
meaning
we
need
people
to
help,
make
the
process
go
forward.
A
All
right
well,
I
mean.
Is
there
any
other
comments
on
this
particular
thread?
I
mean
this
is
I
think
these
are.
This
is
kind
of
like
we're
somewhere
yeah
there's
some
really
cool
energy
here
and
that's
I
see
what
what's
happening
with
the
Cooper
names
operator.
There's
a
lot
of
energy
and
it's
like:
where
do
we
fill
most?
You
know
it's
like
that.
You
could
go
eight
different
directions
or
one
different
direction
or
two
different
directions,
but
it's
not
always
apparent.
I
I
didn't
actually
want
to
talk
about
it,
but
you
guys
rocked
me
in
I
think
it
was
you
Patrick
and
someone
else
said
you've
got
to
go
talk
about
it.
There's
been
a
few
messages
to
the
dev
list
on
the
work
that
I've
done
with
Apache
Jenkins
and
because
everybody
loves
Jenkins,
so
much
I've
been
kind
of
left
alone
in
the
forest
to
deal
with
that
one.
I
So
I've
got
some
questions
to
everybody,
but
also
hoping
that
people
will
reach
out
and
have
some
questions
about
that.
Looking
at
who's
here,
I,
don't
think
I
need
to
run
through
or
give
an
introduction
to
what
we
have
now
with
CI
Cassandra
and
the
machines
which
have
been
donated
by
data
stacks
and
Amazon
and
instaclip
and
Highland
and
and
I
do
know
that
everyone
is
using
predominantly
circle
CI
on
a
day
to
day
basis,
but
one
of
the
one
of
the
ideas
and
motivations
behind
there's
getting
the
Cassandra
CI.
I
The
Apache
Jenkins
up
and
running
again
was
to
have
a
permanent
record
of
test
results
for
our
main
branches
and
to
encourage
more
a
more
repeatable,
transparent,
permanently
recorded
process
to
identify
bugs
and
fixing
them
in
the
community.
If
you
look
through
back
through
some
of
our
older
tickets,
it
can
be
a
bit
difficult
to
know,
look
at
progressions
and
that
type
of
stuff.
B
Think
there
is
some
additional
value
there
too,
largely
due
to
a
change.
The
circle
CI
you've
made
recently
the
inability
to
view
a
build
that
you
yourself,
hadn't,
triggered
in
your
own
personal
org,
has
made
it
really
difficult
to
share
circle.
Ci
results
which
I
think
is
just
even
more
reason
to
place.
Greater
emphasis
on
ASF,
see
I.
E
K
Been
talking
to
them
about
it
for
actually
a
back
and
forth
and
I
have
had
a
back
and
forth
with
many
as
their
support
people.
The
new
UI
wants
to
change
it
so
that
everyone
has
to
be
authenticated,
so
we
might
be
able
to
get
everything
miserable,
but
they
want
everyone
to
do
a
login
first.
So
that
is
the
one
thing
I'm
trying
to
push
them
to
change
is
to
make
it
so
we
don't
need
logins,
but
they're,
trying
to
push
back
saying
everyone
should
create
a
login
first,
so
they
can
track.
You.
I
I
think
it
this
is
free
for
everyone
to
jump
in
and
give
their
opinion.
My
view
on
this
is
I've
been
doing
it
whenever
I
review,
something
because
just
for
the
sake
of
double
accounting,
I
I'm,
always
comparing
the
circle,
see
I
and
the
Jenkins
results.
There
asks
a
different
set
of
tests.
There
are
things
that
circle
see,
I,
do
and
things
that
Jenkins
does.
I
And
like
for
that,
crowd
of
people
who
are
not
committees
and
not
working
in
one
of
the
big
companies
that
are
providing
the
chromium
circle,
CI
they're,
left
out
in
a
cold
I
think
the
only
the
only
way
we
can
look
after
those
people
is
when
a
reviewer
comes
along
on
that
pachu
is
a
committer
one
way
or
another.
They
can
help
them
out.
Otherwise,
those
people
just
have
to
use
the
free
circle
CI,
which
is
just
you
unit,
Ted's
tests,
I.
K
Can't
see
any
way
around
that
so
I've
reached
out
to
the
infra
about
this
and
I
keep
on
forgetting
to
poke
them.
So
what
I
have
noticed,
at
least
from
Hadoop's
background,
is
that
you
don't
need
to
be
a
commuter
to
run
CI
and
a
lot
of
CI
happens
even
pre-commit
rather
than
post
commit,
and
then
how
do
QA
comes
back
and
annotates
the
jura
itself.
So
when
I've
asked
him
why?
What
is
unique
about
Hadoop?
That
is
not
true
for
the
Cassandra,
they
came
back
and
said
they
will
look
into
it.
A
I
It's
a
that's
the
reason
you
don't
have
circle
see
I
own,
the
Apache
account
because
your
patch
account
just
would
get
blasted
with
400
projects
or
something
for
the
Jenkins
stuff.
It's
because
it's
their
infrastructure
and
they
can't
have
unauthenticated
unauthorized
people
just
random,
randomly
running
stuff
and
like
there
are
plenty
of
jobs
out
there
which
are
running
scripts.
Bash
scripts.
I
I
like
on
the
on
the
agent
which
can
build
your
and
deploy
your
website.
Then
you've
got
this.
That's
a
secured
machine
in
their
infrastructure
and
I've
heard
David
that
other
projects
are
doing
it
as
well.
I
think
beam
was
doing.
A
theme
seems
to
be
leading
the
way
and
a
lot
of
these
things,
but
I'm
cysts
I'm
wondering
if
what
they've
done
is
they're
whitelisted
forks
of
the
repository.
So
if
a
contributor
comes
along
and
then
the
committee's
can
go,
they're
trusted
will
will
pre
commit
their
branches,
and
that
gets
around
that.
K
I,
don't
know
about
be
my
no
spark.
It
doesn't
use
a
SF
infrastructure
for
its
builds,
so
they
can
do
whatever
they
feel
like.
But
I
know
what
like
a
Hadoop
itself.
How
I
understand
it
when
I
was
talking
to
a
couple
of
friends
over
there
is
that
they
have
a
bot,
polling,
JIRA
or
listen
to
JIRA,
and
whenever
GRC
is
an
attachment,
it
then
tries
to
trigger
the
the
bill.
K
So
it
looked
like
it's
at
least
JIRA
annotated,
and
what
I
was
hearing
was
that
the
primary
issue
was
that,
like
with
github
itself,
anyone
can
come
in
and
do
whatever
they
feel
like,
but
with
JIRA
you've
already
authenticated
with
Apache
itself,
so
it
seemed
like
that
was
like
a
better
area.
So
even
if
that
can
get
worked
out,
where
is
like
it's
annoying,
but
you
can
submit
a
patch
for
Yadira
and
then
see
I
could
work
there.
N
N
A
I
There's
fallout
this
Harry
this
the
driver
tests,
the
upgrade
tests
are
not
in
Jenkins,
yet
either
it's
a
good
question
and
if
I
link
it
back
to
the
previous
question,
is
there
like
the
value
of
pre-commit
building
on
Jenkins?
Is
that
something
that
we
want
to
push
for
anyway,
because
no
matter
what
it's
always
limited
to
how
much
it
can
test
while
circle
CI?
It's
it's
in
comparison,
unlimited
and
maybe
Jenkins
actually
has
more
value.
I
I
I
K
A
K
Always
say
the
same
thing
along
with
Jenkins
right,
so
for
me,
one
of
the
things
that
I'd
love
to
start
seeing
is
like
what
is
the
nice
thing
about
circle,
CI
and
one
of
the
nice
things
that
I
have
access
to
it's.
We
can
partition
the
work
over
all
the
different
tests
across
all
the
different
workers
itself,
which
drastically
lowers
the
execution
time
so
in
circles,
CI
with
a
higher
config
you
get
D
tests
or
anything
in
about
30
minutes
without
the
higher
resource
test.
K
I
think
if
I
remember
correctly
and
unit
tests
are
about
four
minutes,
but
I
would
love
to
see
in
Jenkins
is
if
we
can
get
similar
results
to
get
most
of
those
things
or
any
faster.
One
thing
that
I
see
itself
at
the
same
time
is
that
like,
since
the
work
is
also
skewed,
you're,
going
to
have
a
small
number
of
containers
that
take
longer
than
others?
K
So
even
when
you
have
multiple
different
builds
going
at
the
same
time
to
conflict,
they
end
up
balancing
each
other
out,
so
that
I
think
would
be
the
greatest
way
to
like
get
people
to
start
using
it.
A
lot
more
often
is
my
circle.
See
I,
like
your
done
within
an
hour,
so
I
think
that
is
the
biggest
hindrance
right
now.
I
I
So
Roberts
been
testing
using
Gradle,
Enterprise
and
distributed
testing
and
he's
got
the
D
tests
running
in
54
minutes
because
that's
the
longest
t-test
so
I
have
no
idea
how
feasible
that
is
on
on
various
fronts,
but
kind
of
makes
me
want
to
wait
and
see
what
comes
out
of
that
before
container
izing.
More
of
those
tests.
One
of
the
things
I
really
would
like
to
see
is
all
of
our
CI
scripts
made
as
agnostic
as
possible
so
that
the
circle
CI
configuration
and
the
Jenkins
configuration
it's
as
thin
as
possible.
A
Well,
on
that,
on
that
vein,
make
some
folks
have
been
talking
to
in
the
communities
community
have
asked
like
hey.
You
know:
we've
been
talking
about
CI
CD
as
a
topic
around
kubernetes
and
Cassandra,
and
there
was
a
really
good
idea
that
and
I
don't
know
how
feasible
it
is,
but
providing
a
hell
in
charge
of
the
community
that
would
build
basically
here's
your
complete
set
up
for
you
know
this
CI
CD
set
up
for
building
Cassandra
and
testing
it
and
then
using
that
to
help
grow
the
amount
of
resources.
A
Potentially
you
know
so
big
company
access
yeah.
We
would
love
to
provide
this
infrastructure.
We
here's
your
helmet
art
and
then
we,
you
know
it's
everything's
containerized,
it's
portable.
You
know
it
kind
of
goes
into
this
portability
sphere.
Again.
This
is
a
kubernetes
operator
topic
as
well,
but
it's
also
containers
we
really
haven't
grounded
out
on
the
container
discussion.
N
I
Yeah
I
think
that's
I,
think
there's
a
lot
to
do
there.
Patrick
I
like
being
aspirational,
yep
but
I.
Think,
certainly,
if
so
I
in
the
in
confluence,
I
would
started
a
wiki
page
to
try
and
kind
of
put
down
the
pros
and
cons
of
Jenkins
and
circle
CI,
and
hopefully
we
can
move
it
in
that
direction.
Patrick
once
we
kind
of
start
kind
of
creating
something
which
is
more
base
or
a
general
vision,
regardless
of
the
one
particular
CI
system.
O
I
So
it's
in
not
the
latest,
but
the
second
latest
CI
status,
email
on
the
dev
list.
Circle
CR
doesn't
do
the
C
Q
LSH
Lib
tests,
that's
yeah!
That's
correct,
David
and
there's
some
hesitation
about
adding
that,
because
some
people
want
to
remove
those
tests
or
together,
because
there's
duplicate
test
coverage
in
D
tests.
Is
that
correct?
So.
K
I,
remember
that
with
Dinesh
was
commenting
on
is
the
efforts
you've
run
it
in
circusy
I
might
be
more
effort
than
porting
it
over
to
a
Python
D
test,
so
I
think
his
argument
was
why
not
just
bite
the
bullet
and
go
Python
D
test,
but
then
that
ended
up
having
it.
So
neither
of
those
tracks
happened
so
running
whatever
the
test
is,
however,
it
is,
would
still
be
good
and
making
sure
that
circle,
CI
and
Jenkins
are
at
least
doing.
The
same
thing
would
be
great
yeah.
I
Someone
is
working
on
the
JDK
11
building
tests.
That's
in
jenkin,
that's
really
easy
because
you
just
there's
a
matrix,
plug-in
and
all
of
the
JDK
s
are
there
already
and
with
Alex's
recent
benchmarking
on
JDK
14.
That
raises
a
question
of
whether
whether
we
want
to
look
into
providing
support
or
start
testing
in
our
CI
Cherica
14.
B
In
principle,
no
objection
to
testing
with
later
J
decays
I
think
that
were
probably
stretched
a
little
thin
to
consider
it.
Something
that's
supported,
but
it's
decently
likely
that
doing
so
could
give
us
a
heads
up
on
something
we
might
need
to
fix
in
the
future
or
something
that
we
might
need
to
report
to
the
open
JDK
project.
K
Yeah,
how
I
kind
of
see
it
it's
after
8
and
11
are
good
and
we're
all
satisfied
with
what
that
means
for
us,
then
14
is
a
great
nice
to
have
if
they
don't
or
if
they
break.
That
might
not
be
our
fault,
which
is
or
does
something
for
the
future
to
deal
with.
So
it's
great
to
monitor,
but
from
a
resource
perspective.
I
want
to
make
sure
that
8
and
11
are
exactly
where
we
need
before.
A
A
F
Excited
about
ZD
see
that
I'm
willing
to
eat
a
little
pain,
not
that
I'd
be
the
one
eating
the
pain,
because
God
forbid
I
have
time
to
work
on
code
these
days,
but
yeah
I
mean
there's.
Oh,
do
we
have
a
precedent
for
like
some
of
these
tests
running
it
like
a
less
frequent
cadence
like
weekly
tests,
etc?.
I
No
so
when
I
jumped
in
and
did
the
cloudBees
set
up
and
the
pipeline
the
decorative
pipeline,
that's
entry,
the
idea
there
was
just
just
to
create
one
pipeline
not
and
not
try
and
have
all
this
area.
I
think
that
kind
of
comes
back
to
it.
Overlaps
to
what
Patrick
was
saying,
trying
to
make
it
just
as
simple
in
its
context,
free
as
possible.
Yeah.
F
F
Stuff
is
beefy
right,
like
those
things
take
some
time
to
run
so
saying
we're
going
to
run
those
for
every
single
commit
before
they
merge
like
at
some
point
it
becomes
a
you
know:
ecological,
bad
choice
for
the
planet,
much
less
for
the
project.
So,
like
maybe
there's
the
question
of,
do
we
run
some
things
nightly?
F
I
Absolutely
I
have
seen
engine
I
have
seen
in
Jenkins
some
of
the
longer
tests
they
can.
Actually
they
can
get
one
stage
actually
reused
in
different
pipelines.
It
seems
to
when
it
queues
up
a
test.
It
like
I,
think
there's
some
things
you
can
do
there
I'd
like
to
look
more
into
that,
both
combining
how
often
you
poll
your
repository,
but
also,
if
you
have
longer
running
tests
and
they're
queued,
for
example,
though
they
concurrency
is
limited
than
they.
F
F
Yeah
JDK
14
so
gonna
be
thinking
about
that
because
like
if
we
ran
that
weekly
I'm
sure
Monday
like
it
would
you
know
tickle
Roberts
to
to
wake
up
to
that
report
because,
for
whatever
reason,
that's
kind
of
thing
that
seems
to
make
him
happy
I.
That
he's
compelled
to
do
it
either
way,
but
that
would
at
least
give
us
like
a
non-invasive
way
to
kind
of
our
turning
the
crank
on
some
of
the
you
know
future
facing
things
without
them
being
part
of
the
day-to-day
operations
on
the
project
kind
of
separate
those
concerns.
I
M
F
72-Hour
smoke
test
is
wrong.
You
didn't
hear
the
wrong
thing.
I
said
the
wrong
thing.
We
got
like
those
those
multi-day
tests
and
the
drivers.
For
instance,
we
could
get
value
out
of
running
these
things
more
frequently
than
just
right
before
release,
but
obviously
it
wouldn't
make
any
sense
to
run
those
things
like
even
on
a
daily
Cadence's.
H
H
Are
there
plans
to
do
that
at
a
project
level,
or
is
that
more
from
a
company
to
company
contribution
level
of
testing
for
the
correctness
testing
in
particular,
because
sometimes
you
have
to
run
it
for
days
because
you
have
to
establish?
You
know
certain
patterns
that
that
may
or
may
not
be
present
in
a
single
day
test
or
even
a
week
test.
F
Theoretically,
like
the
entire
value
prop
of
fallout,
is
to
be
able
to
do
that
kind
of
generative
fuzz
testing,
where
you
write,
those
generators
have
them
run,
have
them
soak
and
then
start
pounding
weird
States
onto
the
cluster
and
see
what
happens
when
notes
go
down,
come
on
bootstrap,
etc
and
I'm
pretty
sure.
There's
people
right
now
working
on
writing
some
of
those,
because
we
had
to
gut
see
tool.
F
That's
like
the
internal
Cassandra
tool
that
was
bad
and
make
it
kubernetes
native
before
open
source
and
fallout
so
I
think
that's
something
that's
open
source,
but
that
mostly
contributors
employed
by
data
stacks
are
pushing
right.
Now,
it's
just
kind
of
a
like
a
fragmented
tapestry
right
now,
Jeremy
right,
like
different
different
companies,
have
different
resources,
they're
applying
in
different
ways.
H
F
The
any
generator
should
come
with
a
validator
that
confirms
that
the
data
that
comes
out
the
other
side
is
the
data
you
expect.
It's
not
purpose-built
for
the
same
kind
of
thing,
like
Cassandra,
did
for
use
kind
of
two
different
clusters
and
make
sure
that
the
data
that's
in
there
is
the
same.
It's
more
about.
You
have
a
single
cluster
and
you
do
bad
things
to
it,
make
sure
at
the
end
of
the
bad
things
it's
the
way.
You
expect
it
to
be
right.
H
Oh
Ted
an
important
thing
and
that's
the
subtle
difference.
What
one
is
more
like
lots
of
different
failure,
conditions
that
are
very
specific
and
the
other
is
more
specific
I
just
didn't
know
if
that
those
either
of
those
like
the
Cassandra
DIF
methodology
or
the
follow-up
methodology,
was
in
the
cards
for
putting
it
into
a
regular
cadence
as
we
get
closer
to
the
release
yeah
for
what
it's
worth
like
a
my
personal
to-do
list,.
F
H
Okay,
yeah
and
I
don't
mean
to
say
that
you
know
before
the
project
gets
the
4.0
out
of
the
door
or
whatever
it
has
to
be
a
project
level
thing
I
just
didn't
know
if
there
was
a
formal
plan,
for
you
know,
for
each
significant
milestone
having
something
done
from
disparate
company
resources
perspective.
So.
K
One
thing
that
I
think
would
actually
be
pretty.
Nice
is
a
lot
of
the
testing
that
we
want
to
do,
and
a
lot
of
testing
that
everyone
is
kind
of
working
on
I,
don't
see
without
additional
resources
of
getting
added
into
the
ASF
Jenkins
environment.
I,
don't
see
those
running
on
Jenkins
for
a
long
time.
K
We
would
need
to
beat
those
up
in
order
to
actually
run
them
there,
which
still
be
nice.
But
what
would
be
kind
of
nice
to
also
say
is
that
different
milestones
is
the
different
companies
that
are
working
on
different
things
contribute
these
are
we
ran
the
test
against
this
release?
So,
if
we
want
to
say,
alpha
is
done,
let's
say
that
diff
test
ran
fall
out,
ran
and
here's
the
test
report.
K
Here's
what
we're
finding
if
we
want
to
say
beta,
is
done
doing
the
same
thing,
so
the
different
companies
that
are
working
on
it
could
at
least
contribute
the
test
results.
That
would
I
think
be
a
very
beneficial
thing,
given
the
fact
that
we
don't
think
as
a
project
open
source
being
able
to
run
them
is
not
going
to
be
easy.
Yeah.
H
B
At
that
point,
I
think
one
of
the
things
that
we're
learning
is
that
we're
able
to
produce
the
the
highest
rate
of
new
issues
for
investigation
and
resolution
when
we're
putting
bb's
through
their
paces,
rather
than
just
in
a
simple
low
testing
environment,
but
finding
ways
to
without
potential
for
customer
or
user
impact
model
simulated
workloads,
either
through
capture
and
replay
through
diff
testing
that
type
of
stuff.
As
pre-flight,
that's
telling
us
a
whole
lot
more
than
independently
running
one
of
these
tools
against
a
static
builders.
O
B
A
As
much
as
we
can
help,
okay,
is
there
anything
else
we
have
here
we're
kind
of
at
the
top
of
the
hour
I,
don't
know
how
much
more
we
can
bang
on
anything
else,
yeah
all
right.
Well,
in
the
spirit
all
the
Apple
employees
will
love.
This
I
got
one
more
thing:
Kassandra
shirts,
since
you
got
to
the
end
of
this
particular
meeting,
I
got
a
whole
batch
of
these
things.
They
just
got
delivered
today,
so
they're,
just
Apache,
Cassandra
shirts,
my
yeah.
A
The
the
artist
I
have
his
name
is
Johnny.
He
is
awesome.
Comic
book
artist
in
Brazil,
I
love,
giving
him
money
to
do
cool
things
and
I
just
give
him
some
basic
ideas.
He
came
up
with
this
one,
which
is
really
cool.
If
you
would
like
one
send
me
an
email
with
your
your
mailing
address
and
your
size,
and
we
have,
we
have
women's
sizes
and
we
have
men's
sizes
and
just
send
me
an
email
Patrick
at
data
stacks.
A
A
So
you
know
like
what
we're
getting
t-shirts,
people
still
gotta
wear
clothes
at
home,
sometimes
right
at
least
we're
on
zoom'
meetings
so
be
happy
to
do
that
for
anybody,
and
if
anyone
watches
this
video
after
the
fact
that
offer
still
stands,
you
don't
have
to
be
here
for
that.
That's
for
anybody,
I.