►
From YouTube: 2018-07-17 Rook Community Meeting
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
All
right,
the
recording
is
started,
so
this
is
the
July
17th
2018
rook
community
meeting
and
the
first
topic
we
need
to
get
into
is
0.8.
The
leader
I
believe,
are
very
close
to
getting
this
release
out
and
ready
and
I
think.
The
last
image
we
have
in
the
mouse
down
here
is
about
the
arm
64
image
having
amd64
binaries.
So
is
there
anyone
that
can
provide
a
status
on
that.
C
I,
maybe
it's
caching
image.
Caching,
there
was
a
logic
backing
to
make
file,
not
that's
fixed.
Now.
C
I
think
the
work
to
use
daemon
base
when
network
went
in
quite
a
bit.
Quite
a
while
ago
the
make
files
were
following
the
amd64
base
image
instead
of
the
arm.
64
base
image,
and
that's
been
the
case
since
my
work
went
in
so
it's
it's
been
a
been
a
while
so
I
it
was
a
really
simple
bug,
but
it
because
we
weren't
building
properly
on
arm.
It
uncovered,
like
a
bunch
of
other
issues,
installing
young
packages
and
on
CentOS.
A
B
A
A
It
doesn't
even
work,
so
I
still
feel
that
this
should
be
in
the
mouse
done
myself
personally,
but
it
would
be
good
to
be
able
to
round
up
the
resources
to
finish
this,
since
this
is
the
last
issue
we
have
that
I
know
of
so
whatever
we
can
do
to
get
everybody
or
get
the
right
people
to
follow
up
on
this,
it's
gonna
be
really
good.
Now.
Does
this
speak
to
potentially
like
started
to
think
about
pipelines
for
other
architectures
in
our
testing,
because
this,
you
know,
didn't
get
caught
for
a
while
for.
D
A
Okay,
cool
all
right
so
Travis
are.
Can
we
consider
you
on
point
for
EM?
You
know,
following
up
and
with
in
getting
the
right
people
on
the
right
resources
to
track
this
down?
Can
you
be
the
go-to
person
here
for
that?
Okay,
Thank,
You,
Travis,
oh
yeah,
so
I
believe,
that's
the
last
that
that
is
the
last
issue,
and
you
know
these
these
issues
here,
we'll
talk
about
in
a
second,
but
so
the
you
know
the
big
architectural
changes.
A
That's
you
know
we
put
a
lot
of
effort
into
and
tried
to
do
a
lot
of
vetting
for
such
as
you
know,
refactoring
everything
to
support
multiple
storage
providers
beyond
SEF
and
that's
f,
OS
DS
getting
moved
to
one
pod
per
OSD.
You
know
those
are
very
big
changes
and
they're
all
in
master.
Now
and
we've.
You
know
supporting
them
through
upgrade
from
0.7,
and
you
know
there's
a
lot
of
effort
to
make
sure
they're
right
or
that
they
were
working
so
in
the
field.
D
D
D
B
A
I
think
that
was
the
regression
from
discovery
from
the
device
discovery
implementation
phase.
Then
it
was
fixed,
but
when
we
went
to
master
with
the
OSD
one
OST
per
pod
could
do
wonderful
extra
Travis.
Could
you
respond
to
this
and
follow
up
on
that
and
begin
note
that
the
version
here
is
probably
doesn't
have
the
fix
yeah?
If
you
want
to
you,
can
follow
up.
D
A
D
D
At
it,
because
I
was
simply
hoping,
we
get
it's
Richard
and
much
like
today.
If
nobody
has
any
changing
feedback,
because
so
we
can't
really
get
into
what
this
this
protocol
fixes.
Is
that
when
you
specify
to
operations
in
a
placement
in
the
cluster
see
of
you,
they
are
not
getting
applied
to
their
OSD
deployments
so
rewriting
and
it
wasn't
applied
for
deployments.
It
was
only
applied
for
the
job
I
think,
but
not
for
the
actual
deployment.
A
D
It
just
a
small
thing:
I
would
like
to
get
in
because,
at
least
in
a
few
last
few
days,
I
had
a
good
amount
of
people
who
are
like
yeah.
Hey
I
want
to
place
a
starch
on
something,
no,
no
it's
Federation's
and
stuff
on
it,
but
they
weren't
getting
applied,
and
during
our
tests
locally
I
at
least
saw
to
put
like
that.
That
determinations
were
added
to
the
O's
D
deployments.
To
knows
the
one
ID,
for
example,
yeah
that
there
wasn't
an
enumeration
edit,
even
though
I
have
specified
it
so
right.
D
D
A
D
I
had
two
rich
Reiko
I'm,
not
sure
if
it's
upon
and
thinking
about,
we
from
time
to
time
seem
to
have
some
issues
with
ABS
on
kuba
news,
1.7,
but
I'm
not
sure
right
now,
if
it's
ready,
mm-hm
and
multiple
times,
I
just
saw
it
on
a
few
PRS,
always
ABS
aw
years
with
companies.
One
does
something
failing,
but
I
didn't
follow.
The
fire
I
found
a
pudding
head
on.
You.
A
A
B
A
D
B
A
Yeah
this
sounds
like
from
the
scope
there
that
I
would
not
hold
the
the
major
8.
That's
8
is
that
0
release
for
it,
especially
if
we
were
able
to
get
the
you
know
they
md6
re.
The
arms
for
package
is
done
and
we've
done
all
our
testing
and
you
know
very
close
to
signing
off
on
it.
I
would
yeah
I
would
not
hold
the
0
date
for
this
and
I
think
the
example.
As
on
the
as
written
on
the
in
the
documents
it
uses
replication
for
for
the
block,
storage
for
right.
B
A
E
B
B
D
A
That's
not
my
I
hadn't,
even
thought
of
I'm
glad
that
some
other
smarter
people
thanks
Caitlin
and
then
this
last
issue
here
about
the
field
monitor
now
that
was
from
fixed
in
ACF
luminous,
dot,
release
right,
Travis
and
now
do
we
have
that
or
what's
the
status
that
I
saw
a
whole
bunch
of
talk
about
sex
release
builds
what's
going
on
there.
Yes,.
B
So
with
this
one,
the
dot
six
release
was
supposed
to
have
it.
Well,
then
dot
six
release,
actually
it
was
built
I,
don't
think
he
was
officially
released
because
I
found
another
issue
with
it,
so
they're
working
on
of
dot,
seven
now,
which
should
be
out
any
day,
so
this
one
shouldn't
be
hit
near
as
often
now,
because
we
have
a
flag
that
you
must
set
to
reuse
multiple
file
systems.
The
operator
has
that
flag.
Now,
so
without
that
flag
set,
you
can't
hit
this
issue,
but
I
would
suggest.
B
A
A
F
A
So
we've
added
that
to
the
release,
documentation
here
and
I'll
go
ahead
and
send
that
in
a
chat
to
so
that,
there's
a
link
to
it,
then.
So
this
is
the
criteria
here
that
will,
you
know
a
first
pass
at
at
least
that
we
will
use
to
make
sure
that
the
release
is
of
a
certain
quality.
And
you
know
all
these
important
items
have
not
been
forgotten
and
you
know
everyone
is
signed
off
on
the
release
that
needs
to
need
to
approve
it.
So
a
lot
of
these
items
have
been
done
here.
A
Then
the
other
thing
I
wanted
to
talk
about
real,
quick
too,
is
that
now
project
status.
So
before
we
had
been
declaring
a
single
project
status
for
the
entire
Brooke
project,
and
you
know
now
that
we
have
various
storage
writers
being
supported
like
SEF
and
in
addition
to
that
Mineo
and
cockroach
DB
so
far
in
in
row,
hands
working
on
NFS.
A
We
need
the
ability
to
be
able
to
version
them
independently,
and
you
know
it's
kind
of
specify
or
describe
the
maturity
level
of
each
of
those
different
storage
providers
and
the
support
for
them
on
an
independent
basis.
So
we
are
no
longer
using
a
project
status
for
rook
as
the
whole,
but
we
will
be
talking
about
the
support
for
each
storage
provider
individually
and
then
of
note
there
is
that
SEF,
specifically
with
this
milestone,
is
graduating
to
beta.
And
what
really?
A
What
that's
saying
is
we're
making
a
statement
that
the
types
the
SEF
types
you
know
the
API
that
resurface
force
F
is
going
to
be
at
a
certain
level
of
stability
going
forward
where
if
there
are
changes
to
it,
then
you
know
we
expect
to
support
those
with
migration
or
upgrade
through
upgrade
such
that
the
SEF
types
should
honor.
You
know
backwards
since
the
backwards
compatibility,
the
other
types
you
know,
cockroach
TV,
are
still
very
early
days
and
still
alpha
support
so
yeah,
so
there
so
yeah.
Basically,
overall
rook
itself
does
not
have
a
status.
C
A
A
Recapture
get
there,
and
so
that
came
up
in
I
think
one
of
Travis's
recent
pr's,
where
the
home
chart
we
want
at
first
staff.
We
want
the
helm
chart
to
also
be
you
know
in
the
beta
channel,
but
the
whole
process
of
promoting
to
beta
is
gonna.
Take
you
know
the
other
storage
shredders
along
for
the
ride,
so
what
Travis
and
I
decided
there?
You
know
it's
basically
just
removed
the
beta
image
tag
that
accidentally
goes
that
gets
published
for
cockroach
and
menu
I
I.
C
D
D
Just
meaning
it
it's
I
think
we
could
do
it
in
one
step
with
my
overhauling
the
whole
test
pipelines
with
that.
Looking
in
a
regard
of
parallelity
and
stuff
off
the
bullet
jumps
and
everything
to
just
take
a
look
at
this
that
we
eat,
I
can
even
version
stuff
separately
and
and
so
on,
without
triggering
I,
releasing
other
pulse.
D
E
A
trade-off
here
that,
when
you're
developing
something
that's
trying
to
catch
up,
not
being
forced
to
operate
at
the
same
level
is
actually
quite
advantageous.
Even
it's
going
to
slow
down
your
how
quickly
people
on
test
with
your
branch,
they
not
being
forced
to
be
prematurely
mature,
is
quite
valuable.
Yeah.
C
And
that's
the
circuit,
didn't
we
I
think
that's
achieved
by
marking
the
CRD
types,
the
resource
types
as
alpha?
So
if
you
were
to
add
next
Center
you'd
say
it
Martinez
also
and
I
think
that
that
totally
achieves
what
you're
saying
and
you
could
decide
how
long
and
you
want
to
go
before
it
goes
to
beta
we're.
Just
the
granularity
is
no
longer
at
the
Paul
repo.
It's
now
at
a
resource
level.
C
A
G
Right,
that's
good,
so
my
name
is
Janice
and
I
contacted
you
on
the
slack
channel
and
I
won't
explore
the
possibility
of
building
at
the
center
operator
on
top
of
rope.
So
let
me
give
a
little
context.
I'm
doing
this
project
for
my
diploma
thesis
and
I
think
a
sander
is
a
good
feed,
first
of
all,
because
it's
a
very
popular
database
and
second
because
there
is
very
little
automation
right
now
in
the
out-of-the-box
Cassandra
I,
don't
know
about
the
Enterprise
editions
the
various
vendors
may
offer,
but
we
the
default
installation
of
Cassandra.
G
It
needs
very
much
manual
intervention
in
order
to
work
correctly.
So
you
know
it's
a
rule,
and
so
we
had
a
very
robust
community
and
it
seemed
like
a
quickly
evolving
project.
So
I
thought.
Maybe
if
we,
if
I,
could
fit
this
operator
into
the
whole
project,
it
would
be
better,
maintain
better
testing
and
more
used.
G
G
G
C
G
C
G
Right,
of
course,
I
want
to
take
the
dome
and
the
solution
I
have
arrived
to
so
first,
let
me
just
improve
some
requirements,
so
my
operator,
what
I
wanted
to
do
is
to
automate
the
creation
of
new
nodes
scaling,
the
replacement
of
a
node,
the
repair
of
the
data
with
a
manual
operation
Cassandra
and
right
now
it
is
done
by
hand.
You
know
by
I,
should
say
setting
into
a
note,
then
issuing
the
operation
backup
and
provide
the
metrics
and
polling
to
integrate
with
Prometheus,
so
yeah.
A
G
I'm
glad
I'm,
aiming
for
a
very
friendly
UI
goes
that
right
now
is
something
Cassandra
doesn't
have
it's
the
same,
because
it's
a
very
good
software
solution,
so
yeah.
In
order
to
do
this,
however,
I
have
to
talk
to
Jay,
makes
interface
in
the
pods,
so
AJ
makes
interface.
If
you
know,
is
an
RPC
protocol
in
Java
and
in
order
to
do
that,
the
best
solution
I
have
arrived
to
is
to
use
is
to
run
a
sidecar
container
alongside
the
main
Cassandra
container.
This
is
a
solution
that
vendors
have
used
successful
in
the
past.
G
Netflix,
for
example,
runs
aside
carol
oates
a
cassandra
spotify
also
runs
a
sidecar
and
looks
at
chandra.
Those
projects
are
called
Priam
and
reaper.
If
you
want
to
look
them
up,
so
it's
more
or
less
a
tested
approach
and
I
think
it
would.
It
could
bring
very
good
results
and
I
have
tickled
I
got
taken
a
look
into
the
mooc
architecture
and
from
my
understanding
you
have
a
kind
of
sidecar,
but
at
the
node
level
the
equation
incorrect.
A
Yeah
that
the
rock
agent
you
know
has
fairly
specific
functionality
and
that
it
is
an
interface
to
the
kubernetes
storage
subsystem
to
service
kind
of
like
a
volume
plugin.
So
it's
you
know
it
does.
You
know,
take
commands
like
map
or
attach
or
detach
a
volume
when
they
know
the
cubelet
was
telling
it
to
do
so
and
then
performs
that
work,
but
it
hasn't
necessarily
I
mean
yeah,
which
is
in
a
turret
system.
They.
G
C
G
A
very
rough
design
document
right
now,
I
marked
on
document.
Do
you
have
a
way
you
wanted
to
provide?
Maybe
open
gigabit,
you
yeah.
C
G
C
G
Have
the
document
were
less
ready,
so
I
will
open
the
github
issue,
so
you
can
review
it
and
then
go
on
with
that
and
in
the
so
I
was
talking
about
the
root
page.
Yet
because
I
was
thinking
that
this
particular
approach,
the
CRV
plus
controller
plus
a
sidecar
to
run
alongside
the
database
and
maybe
I-
think
it's
a
partner
that
other
databases
could
use
as
well.
G
I
have
seen
other
operators
trying
to
do
this
particular
thing
for
post
press,
for
example,
so
I
was
one
of
my
hopes
for
this
undertaking
for
Cassandra
operator
is
by
completing
this
project.
I
should
be
able
to
provide
a
generic
implementation
of
a
sidecar
that
you
know.
Other
databases
can
extend
to
integrate
with
Luke
and
Fernandez
and.
A
I,
like
that
mindset
as
well,
because
you
know
with
the
you
know-
we
have
a
set
of
functionality
that
is
common
across
storage
providers.
You
know
in
the
in
the
rook
framework
that
you
know
can
help
other
storage
riders,
build
operators,
and
you
know
contain
some
of
that.
You
know
operational
logic
in
there
and
so,
if
any
further
steps
to
you
know
beef
that
up
or
add
functionality
in
it
that
can
further
help.
Other
storage
platforms
in
the
future
is
definitely
initiative
that
I
support
wholeheartedly.
A
B
Thought
to
consider
in
the
design
is
when
thinking
about
using
the
agent
or
how
you
can
share
code
there.
You
know
think
of
it
at
the
code
sharing
level,
because
each
of
the
operators
do
run
independently.
You
know
in
their
own
pods,
so
the
even.
If
you
share
code
with
the
rogue
agent,
you
know
it
would,
whatever
you
do
with
for
Cassandra
Cassandra
operator
I
would
anticipate
will
run
as
its
own
pod.
They
might
be
called
the
Cassandra
agent
or
or
maybe
it's
a
sidecar
with
the
database
pod
yeah.
G
G
A
You
know
the
in
terms
of
developer
Doc's
about
you
know,
what's
available
in
the
in
the
you
know,
Brooke
codebase
and
libraries
I.
Think
a
lot
of
our
documentation
has
been
focused
on
end
users
or
you
know,
operators
or
admins
about
how
to
use
Brooke
to
do.
You
know
useful
functional
work
and
their
clusters
and
I
don't
know
if
we
really
have
much
documentation
about
focus
that
developers.
A
So
that's
maybe
something
that
we
need
to
beef
up
over
time,
but
you
can
definitely
use
you
know
me
Jared
here
as
a
resource,
because
you
know
I
wrote
the
cockroach
TB
operator
implementation
recently
and
have
you
know
been
working
on
some
that
Rug
framework
stuff.
So
you
know
you
feel
free
to
reach
out
to
me
on
slack-
and
you
know
we
can
kind
of
you
know.
I
can
give
you
some
some
tips
or
maybe
walk
through
a
couple
of
things
with
you
and
then
maybe
we
can
identify.
A
C
B
B
A
G
A
let
me
do
one
last
question
so
do
that
for
the
operator
I
will
develop.
Do
you
target
a
specific
burn
in
this
version
and
after
or
no
that's
in
your
pay
rate,
or
we
can
target
one
thing,
for
example,
the
sub
resources
for
CRTs,
with.
A
A
G
G
G
C
A
Interesting
sounds
like
it
would
be
a
very
beneficial
thing
to
the
community
as
a
whole,
so
I
I
like
this
effort,
a
lot
so
I
think
so
we
already
talked
about
this
pull
request
here
and
it's
an
action
item
we
captured.
So
there
wasn't
anything
else
left
in
the
agenda.
Was
there
anything
on
someone
else's
mind
that
needs
to
be
brought
up.
H
Hello,
hello,
yeah
can
I
quickly
top
sure
I
just
want
to
introduce
myself
myself.
My
name
is
Indra
Ariana
I
am
the
person
who
submitted
the
feature
request
for
Luke,
so
I
just
quickly
want
to
talk
because
I
hear
the
previous
speaker
talking
also
but
I
think
a
Sandra
as
an
operator,
their
intro,
so
I
just
want
to
introduce
myself
quickly.
Here.
I
spoke
with
a
couple
of
you
with
regard
to
the
process
of
contributing
in
rope,
especially
I'm
interested
in
integrating
illusio
in
in
in
rope.
Yes,.
H
H
Maybe
it
is
not
very
deep
at
the
moment,
however,
I
probably
would
like
to
share
as
soon
as
possible,
so
that
I
can
get
feedback
and
comments
and
I
use
a
full
guide
as
to
how
to
develop
the
operator
and
also
the
agents
in
rock
yeah
I.
Think
that's
actually
today.
I
do
not
really
want
to
go
into
be
good
things,
but,
but
if
you
have
some
question
with
regard
to
my
motivation
or
the
purpose,
what
is
the
use
case
in
thing
I
then
I
hope
we
can
quickly.
Also
answer
about
that.
A
Yeah
I
think
from
the
introduction
that
we've
gotten
so
far
on
the
the
mailing
list
and
you're
going
to
be.
You
know,
opening
or
extreme
with
us
a
you
know,
one
page
or
design
write
up
soon.
Then
that
sounds
to
me
at
least,
like
that's
a
great
channel,
to
follow
up
on
on
all
that
stuff
on
I
mean
if
anyone
else
has
specific
questions
right
now,
you
know
if
it
feel
free
to
ask
but
I'm
content
with
with
following
up
on
the
one
pager.
H
A
H
H
D
We
support
the
combinations
that
are
supported
by
the
client
going
after
which
metro
and
matrix
are
we
going
because
they're
comfortable
compatibility
metric
their
matrix.
There
is
well
not
not
really
well
showing
which
it's
in
what
even
supports
a
convenient
five
as
far
as
I
understand
the
matrix
there,
so
yeah.
C
D
C
D
F
If
one
brief
mentioned
that
that
problem
with
the
OST
is
not
restarting
when
the
node
restarts
several
people
drop
the
hint
that
it
might
have
something
to
do
with
using
host
path
instead
of
devices,
so
I've
got
another
thing:
I
can
try
to
reproduce
with,
and
at
least
one
of
those
people
said
that
the
OSD
came
back
up
with
a
different
cluster
FS
ID,
which
would
certainly
cause
the
problem
yeah.
There
might
be
some
light
at
the
end
of
that
home.
Awesome.