►
From YouTube: 2017-11-06 Rook Community Meeting
Description
Rook Community Meeting
A
A
D
A
D
D
C
So
so
what
I
got
from
the
team
is
that
if
you're
not
older,
if
you're
new
kernel,
when
I
say
new
kernel,
corporal
or
app
you're
fine,
but
if
you're
an
older
kernel,
what's
gonna
happen?
If,
if
a
user's
request
an
attachment
for
file
system,
it
would
check
whether
you
have
two
or
more
list
basis,
meaning
two
more
faucets
or
servers
in
the
current
era.
I'll
say:
hey,
you
know
what
you
need
to
delete,
one
because
I
don't
know
which
I
do
want
to
attach
to.
C
C
C
F
D
C
Okay,
just
I
thought
that
behavior
being
undefined
was
a
big
thing
in
case
because
we're
dealing
with
file
system
and
data
here
so
I
wouldn't
want
to
trick
the
user
that
hey.
You
know.
You
are
actually
some
student,
but
the
user
meant
to
attack
from
a
star
system
or
number
two.
So
faucets
are
number
one.
For
example,
I
feel.
D
D
C
F
D
C
D
D
C
A
D
D
Yeah
III
wonder
if
the
right
solution
right
now
is
to
check
the
kernel
version
check
if
it's
safe
to
pass
the
option
and
then
just
document
that
multiple
file
systems
are
only
supported
if
you
on
nodes
that
are
running
for
point
seven
and
above
for
now,
set
fuse
we'll
fix
that
in
the
future.
But
for
now
for
0.6.
D
D
F
C
A
D
A
D
A
A
A
D
A
F
A
F
In
there
so
fair
for
my
own
reminder
here,
the
levels
of
testing
that
we
have
are
the
integration
tests
that
just
run
as
part
of
continuous
item
integration
on
every
pull
request.
Then
we
have
a
full
regression
pass,
which
is
longer
and
incorporates
more
things,
and
then
we
also
have
long-haul
that
runs
over
like
an
entire
weekend.
So
it's
kind
of
those
three
levels.
A
D
A
A
D
A
D
D
A
F
A
F
A
A
D
A
As
I
suddenly
have
an
appointment
in
a
few
minutes,
I
just
have
a
quick
update
from
my
side
about
a
cube.
Con
I
just
got
my
flight
booked
today,
so
I
will
definitely
better
a
PDR
awesome
yeah
by
the
way,
just
one
small
thing
from
my
side,
I
posted
about
I
think
it
was
1194
it's
about
the
resource
constraints.
A
F
D
B
D
A
You're
good,
who
was
the
report
again?
I
didn't
catch
the
name
ska.
B
B
D
B
For
the
most
part
is
just
to
make
kind
of
security
contacts
clear
on
what
each
part
is
supposed
to
do
if
he
doesn't
need
to
run
in
push
mode,
who
don't
do
that
and
I
believe
for
staff?
They
don't
need
to
run
this
root
and
the
root
filesystem
has
to
be
rewrite
by
lease.
If
you
make
that
explicit,
the
the
security
context
in
different
most
should
have
no
problem
handling.
That
right
now
is
all
implicit
and
yeah.
D
F
D
F
D
Sure
sure
I
think
we
can
take
this
as
part
of
0.7
there's.
Our
goal
is
actually
to
get
to
a
point
where
we
don't
have
any
problems
contain
at
all,
but
that's
gonna
require
you
know
some
work
in
kubernetes
to
support
local
storage,
and
maybe
the
local
storage
can
do
the
mounting
of
devices
and
all
of
that
I.
Don't
think
we're
there
yet,
but
it
would
be
great
if
Brook
can
run
without
any
privilege
containers
right.
B
D
F
Was
I
was
kind
of
curious
here,
I
wanted
to
have
a
bit
of
a
bigger
discussion
and
not
long,
but
just
to
get
a
better
idea
of
what
Shaw
your
use
case
is
Shawn
cuz,
it
seems,
like
you
know,
you're
one
of
the
first
people
that
we've
seen
run
into
you
know
this
sort
of
a
more
lockdown
environment
where
you
don't
want
to
use
any
local
devices
and
such
and
I
kind
of
want
to
get
a
better
idea
of
you
know.
Do
you
know?
If
that's
you
know
a
fairly
common
scenario
and
then
beyond
that?
F
My
concern
when
I'm
kind
of
worried
about
is
that
you
know
it
seems
like
you
want
to
go
in
a
direction
where
you
know
kind
of
like
Mineo
is,
is
what
you're
mentioning?
Is
that
you're
only
using
kind
of
temporary
aggregating
temporary?
You
know,
storage
together
to
build
a
greater
storage
network
and
I'm
wondering
if
you
know
I
haven't
seen
OS
T's
work
very
well
with
that,
where
you
know
they
tend
to
be.
You
know
wanting
to
have
an
entire
disk,
so.
B
I
think
yeah
I
think
a
lot
of
those
conversation
I
took
belong
to
two
separate
topic.
One
is
how
can
I
get
route
up
and
running
with
a
minimal
amount
of
efforts,
or
the
least
assumptions
and
and
part
of
that
is
dead
and
kind
of
in
our
environment?
That
assumption,
obviously
just
end
right
at
you
just
doesn't
work.
We
we
actually
would
allow
people
to
mount
local
storage.
B
The
problem
is:
there's
really
no
good
story
on
how
many
gene
calls
for
local
storage
resources,
yet
so
what
we
have
been
trying
to
do
is
to
build
support
into
our
cluster
so
that
it's,
for
example,
use
the
integer
extension
resource
to
manage
the
court
out
of
that.
But
then
there's
a
result,
we
don't
expose
the
role
devices
to
our
users
and
to
make
the
cluster
wide.
So
we
could
build
a
cluster.
Why
Seth
FSO
look
for
other
people
to
use,
but
I
can't
imagine.
B
D
D
So
sorry,
our
goal
is
to
get
to
a
point
where
rooted
one
and
on
top
of
local
storage
I
mean
really
any
persistent
Dom.
But
typically
we
think
the
common
scenario
will
be
local
storage,
but
you
can
put
it
on
top
of
EBS
or
Google
persistent
disk.
If
you
wanted
to,
and
then
you
know,
if,
if
the
mounting-
and
you
know
formatting
and
all
that
stuff
that
happens
with
these
devices
is
done
by
the
local
storage
provider,
then
we
should
be
able
to
run
without
privileged
and
with
the
leak
leave
little
security
context.
So.
B
Like
I,
don't
need
to
I,
because
then
I
can
modify
the
part,
which
is
month
the
flex
volume
into
the
expected
directory
yeah,
because
the
if
you
host
use
host
pass
then
I
have
manga
on
the
underlying
host.
I
cannot
just
have
the
part
monk
that
into
its
own
storage,
namespace
or
file
system
in
space,
I,
think
that
is
the
cleaner
solution,
so
I
prefer
to
have
the
part.
So,
as
I
was
talking
about
in
the
hitch
in
the
in
a
slack,
room
is
I
prefer
to
add
options.
B
D
So
I
wonder
if
there
so
it
sounds
like
there
are
two
things
here:
there
is
the:
how
do
we
link
our
pods
run
lease
privileged?
Well,
not
yeah.
We
we
should
just
fix
in
0.7
and
then
the
other
discussion
is
really
I,
think
it
it
dovetails
into
the
local
storage
design
that
travis
has
so
maybe
we
could
make
sure
that
your
scenarios
are
captured
as
part
of
that
design,
because
I
think
there's.
Actually
there
is
more
work
to
be
done
on
the
kubernetes
site
for
local
storage.
D
So
maybe
this
sounds
like
this
pull
request
should
be
scoped
to
just
the
least
privileged
piece,
that's
correct,
yep
and
then
the
larger
design
discussion
we
should
have
in
the
context
of
local
storage.
What
does
that
make
sense?
Yep
so
Travis
can
you
can
you
maybe
work
with
with
chai
on
on
on
this
I
make
sure
we
have
all
the
design
calculate
I.
D
A
D
C
C
D
F
Yet
dad
that
being
said,
I
think
one
of
the
ones
that
was
kind
of
concerning
me,
because
just
within
the
last
week,
I
think
I'd
heard
like
three
separate
people
mentioned
it
as
an
issue
is
that
the
rookie
iPod
seems
to
be
pegging
a
CPU
for
some
people.
You
know
I
hadn't
remembered
acting
that
way
before.
So
it's
not
a
feature
request,
but
it's
kind
of
a
concerning
issue
and.
F
A
D
E
D
A
E
D
D
D
D
D
F
A
F
F
A
A
F
I
didn't
strike
me
as
a
you
know:
high-level
milestone
theme,
but
you
know
there
is
an
issue
I
think
in
region.
Maybe
just
one
issue
about
being
able
to
define
you
know
further.
Do
you
define
granularity
for
on
the
component
level
like
I
want
the
operator
to
be
this
level?
I
want
the
you
know,
Mons
to
be
at
this
level
right.
A
F
F
A
A
D
F
So
yeah,
that's
one
part
of
it
at
least,
which
is
now
in
master
that
we
will
clean
up
our
media
devices
for
you
now
and
that's
on
the
main.
You
know
clean
up
path
as
well.
If
you
follow
the
documentation
that'll,
everything
will
happen
nicely
for
you,
but
I
think
that
there
is
a
greater
you
know
a
thing
to
deal
with
with
just
you
know:
rebooting
like
wind
chorus.
Let's
say
it
gets
new
update
and
manoda
gets
restarted.
Do
we
die?
F
D
D
F
F
C
D
Yeah
I
mean
this:
this
there's
a
this
is
there's
a
class
of
things
that
we
are
gonna
need,
make
changes
in
kubernetes
for
and
I
think
what
every
time
we
talk
about.
This
is
like
we
have
to
have
an
active
role
and
asking
for
these
changes,
so
we
shouldn't
wait
around
and
ask
for
it
to
be.
You
know
see
what
they
do
with
it.
We
should
go
ask
specifically
for
it
to
become
data
or
enabled
and
make
a
case
for.
E
Can
can
we
structure
this
list
into
a
zero,
seven
must-have
and
a
zero
seven
like
that,
Prometheus
f
thing:
we
would
it's
great
there's
a
pull
request
out,
but
we
would
still
ship
zero
seven
without
it.
I
think
is
there
way
we
construction
list
to
take
that
into
account,
because
then
I
think
there's
something
like
performance
testing
for
block
storage.
I,
don't
know.
Could
that
be
in
that
group.
A
A
F
E
F
F
D
A
E
D
D
A
F
F
E
E
D
E
C
A
D
A
F
D
I
think
what
it
means
is
that
we
decide
the
rook
agents
besides,
what
the
best
way
to
do
block
and
file
is
going
to
be.
For
that
note,
and
then
it
will.
There
could
be
hints
as
well
from
from
at
the
cluster
level
even
pool
level,
but
it
will
decide
whether
to
use
the
kernel
modules
or
it
will
start
one
of
Ceph
fuse
or
env.
Dr
BD
PCM.
You
runner
and
run
them
in
the
agent
as
a
child
process
in
the
agent.
D
So,
and
by
doing
so
it
doesn't,
you
know,
like
the
staff,
use
examples,
it's
great,
because
that
means
that
there's
no,
you
know
besides
fuse,
there's
no
other
formal
activity
and
we
get
to
run
with
multiple
namespaces
and
all
the
features
that
are
supported
in
luminous
series
trooper
in
b'dar.
We
need
TCM
year
runner.
C
C
D
D
D
C
D
C
A
A
F
Also
for
tracking
I
don't
know
if
you've
always
done
a
good
job
of
when
we're
focusing
on
a
design
that
we
also
having
a
issue
on
the
board
for
it
as
well,
but
if
we're
gonna
front
load
some
of
our
design
work
now,
for
you
know
at
the
Sakura
milestone
we're
about
to
embark
on,
then
I
would
love
to
everyone.
We
have
a
ticket
open
on
the
board,
so
we
can
track
that
work.
D
D
E
A
E
E
D
F
D
F
A
F
D
D
C
F
C
C
F
A
F
A
F
F
A
F
I
think
on
slack
barrier
talk
about
Alexander
and
him
I'm
talking
to
Travis
and
I
about
that,
so
we'll
meet
up
at
them
and
do
that
in
general.
I
think
having
a
single
owner
from
you
know
that
the
core
team
is
makes
a
lot
of
sense
for
each
area.
You
know
instead
of
kind
of
a
dissemination
of
responsibility
where
there's
no
real
accountability.
I
hate
it
I
hate
doing
that.
I
always
want
to
avoid
that
yeah.
C
F
I
mean
said
that
I
totally
agree
with
you
Stephen
that
that
first
step
is
to
get
you
know,
issues
open
for
the
things
and
put
names
on
them,
which
we
don't
actually
have
right
now.
So
once
we
do
get
a
name
on
it,
then
sure,
then
the
accountability
and
responsibility
is
there.
But
we
need
to
take
that
first
step
to
get
names
on
things.
A
more
formal
way.