►
From YouTube: 2018-01-03 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).
B
A
A
B
B
We
want
to
think
that
there's,
you
know,
there's
some
architectural
changes
that
are,
you
know
for
this
release.
So
I'm
wondering
you
know
with
the
scope
of
0.7,
you
know
should
be
with
you
know
some
of
those
architectural
changes
that
are
needed,
and
you
know
what
all
we
really
want
to
accomplish
him.
You
know
when
we
would
have
this.
A
C
A
D
A
You
know
I
also
I'm
I'm,
trying
to
think
of.
If
we
will,
if
we
try
to
optimize,
for
you
know,
I'm
a
release
with
mimic
right.
There
is,
you
guys,
have
requirements
to
get
there,
and
some
of
them
should
get
front.
Loaded
like
what
Jerry's
stuff
at
some
of
these
architectural
changes
are.
Are
would
be
nice
to
get
these
in
earlier
than
later
yeah.
A
A
F
Is
there
any
notable
feature
or
it
opens
every
because
I
don't
recall
having
being
any
apart
from
the
usual
bad
fixes
and
know
if
they're,
not
any
I
actually
would
rather
get
some
of
this
architecture.
Change
that
we
talked
about
like
I
know
is
one
part
pro
se,
because
that
will
that
will
allow
us
to
you
know
easily
attack
and
solve
this
all
these
issues
regarding
OSC,
not
restarting
and
then
crashing,
and
all
that
well.
A
There's
I
mean
just
looking
at
the
milestone
right
here.
We
talked
about
a
bunch
of
things,
they're
all
incremental
changes,
there's
nothing.
You
know
big
here,
but
the
this
whole
idea
of
using
local
storage
or
underlying
persistent
volumes
and
how
it
relates
to
permissions
and
how
it
relates
to
volumes
being
mounted.
There's
this
whole
category
of
things
that
I
think
is
still
still
speculative
at
this
point.
I,
don't
I,
don't
think
we
have
a
you
know.
A
We
have
a
good
solution
yet
and
now
adding
adding
the
open
shifts
and
some
of
the
red-hot
scenarios
just
makes
it
even
clearer
that
we
need
to
figure
out
how
you
know
we
use
hosts
bath
or
local
storage
and
permissions
around
them
and
everything
else
so
I.
You
know
I
I
feel
like
we
should
think
about
that
4:07.
E
D
E
E
A
Agree-
and
you
know
when
I
work
backwards.
It's
looking
like
you
know
if
you
say
that
May
is
when
we
went
to
if
we,
if
we
shoot
for
May
as
a
v1
for
rope
right,
let's
work
backwards
from
that,
and
it
would
seem
to
me
that
you
know
we
do
something.
Like
a
point.
Seven
point:
seven
release
a
point:
eight
release,
probably
in
March
and
then
v1
in
May
and
not
the
main.
The
main
gate
is
interesting
because
it's
you
know
it's
cubic
on
Europe
it
against
that
sebasi
mothers
can
fill
in
here.
A
C
B
And
you'd
have
one
question
on
that
topic
too:
Bassam
that
I
would
have
is
I
would
I
would
appreciate
a
bit
of
a
refresher
in
terms
of
what's
going
to
be
in
mimic.
Some
of
the
big
features
there,
one
of
them
I'm
kind
of
curious
about
is
that
I
know
there
had
been
talk
about.
You
know
managing
placement
groups
better
and
kind
of
being
able
to
abstract
that
away
from
any
users
concern.
You
know
the
not
expose
placement
groups
there
so
is.
A
A
E
A
C
Think
of
anything
at
the
moment
it's
not
really
safe
correlated,
but
we
really
have
to
start
using
cell
volume
as
of
mimic
yeah.
So
it's
not
a
corset,
but
it's
a
configuration
of
noisy's,
so
we
have
to
get
this.
This
is
this
is
part
of
the
sequence
of
having
one
OSD
per
pod
and
then
start
using
cell
volume
and.
A
A
A
A
F
Even
even
if
we
don't
get
the
whole
complete
solution,
we
local
storage
and
pretty
early
container,
all
done
even
I,
think
having
one
OSE
per
pod
is
still
gonna,
be
beneficial,
for
example,
in
quantum
we
have
a
lot
of
issues
where,
if,
if
one
way
C
just
fails
like
processes,
II
just
fail,
the
whole
cluster
doesn't
come
up.
So
issues
like
this,
we
better
manage
if
we
were
had
Oh
a
sleeper
pod,
because
in
this
case
we
can
just
restart
a
pod.
A
A
F
F
A
B
It
seems
like
there
needs
to
be
some
sort
of
you
know
overarching.
You
know
design
here,
that's
kind
of
it
more
of
a
high
level
of
how
all
of
these
you
know.
Smaller
components
are
going
to
be
fit
together
and,
what's
gonna,
be
our
approach,
for
you
know
a
lot
of
these
things
because
it
seems
like
individual
issues
open
for
all
of
them,
but
how
they
all
tie
in
together-
and
you
know,
what's
the
overall.
A
B
A
A
A
A
C
There
is
some
kind
of
a
kiss
kidding
effect
on
where
you,
when
you
do
one
OS
deeper
part,
then
you
have
to
come
up
with
a
strategy
on
how
to
detect
devices.
So
I
thought
we
agreed
on
what
almost
agreed
on
having
something
like
an
init
container
that
scan
all
the
devices
available.
And
then
you
have
the
second
step
word,
which
is
for
people
about
the
preparation
and
then
you,
the
ultimate
step,
is
to
convert
all
of
that
using
cell
volume.
F
A
Yeah
you
say:
I
go
back
to
so
the
desire
to
make
these
changes
first,
but
I
think
I
feel
like
we
have
to
have
some
kind
of
blueprint
design
for
how
we
you
know
what
we're
going
to
do,
and
ideally
we
can
sequence
them
set
so
that
we
can
take
smaller
bites
and
keep
master.
You
know
stable,
otherwise,
I
think.
A
Well,
if
we're
gonna
take
in
a
really
big
chunk,
then
we'd
have
to
branch
work
in
a
you
know:
a
bigger
branch
which
I
prefer
not
to
do
like,
for
example,
we
can
switch
to
using
apt-get
Deb
packages
inside
the
container.
If
we
can
figure
out
some
issues
around
it
and
then
we
could
layer
other
stuff
on
top
of
it,
but
it
would
be
great
if
we
can
have
you
know
a
overall
design
of
what
we
want
to
accomplish
in
what
order.
B
Into
some
of
a
quick
thought
on
that,
is
you
know
a
way
that
one
way
that
we've
done
some
of
our
bigger
designs
and
recent
history
here
for
rook
is
where
you
know
one
person
goes
off
and
as
a
full
and
complete
you
know
deep
dive
design
document.
That's
you
know
submitted
as
a
pull
request
to
the
repo,
and
you
know
that
that
approach
has
been.
You
know
it
works
well
for
things
that
are
designs
that
are
kind
of
more
isolated
bone.
You
have
designs
that
are
really
really
kind
of
overall
in
broad
in
scope.
B
Here.
I
think
that
a
potential
is
a
suggestion
for
an
approach
here
would
be
more
of
a
like
a
live
collaborative
sort
of
effort
like
on
a
shared
Google,
Doc
or
something
where
you
know
you
can
kind
of
keep
that
you
don't
have
one
person
working
off
in
isolation.
You
have
it
more
as
a
live
effort
by
you
know
more
than
one
person,
you
probably
need
an
owner
on
the
whole
effort,
but
yeah,
but
you
know
more
live
and
collaborative
effort
is
probably
gonna,
be
nice
here.
A
B
I
can
I
can
probably
own
it.
You
know
I,
but
what
I'm
trying
to
finish
off
right
now
is
you
know
being
able
to
update
the
cluster
CRD?
You
know
adding
and
removing
storage
to
a
live
cluster,
basically,
which
has
been
kind
of
a
pretty
large
feature
in
itself,
but
I'm
running
into
a
lot
of
those
issues
about
you
know.
Well,
how
would
things
be
done
with
you
know?
You
know
it's
these
down
to
their
own
pods,
and
things
like
that.
B
A
B
Okay,
a
full-time
Sebastian,
that's
pretty
awesome.
A
B
A
B
A
We've
got
to
pull
them
all
together
and
and
then
there's
some.
You
know,
there's
even
more
investigation
exploration
to
be
done
in
some
areas
so
and
there's
some
stuff
upstream
to
I'm,
not
sure
I
understand
like,
for
example,
will
there
be
armed,
hf
builds
published
by
upstream
stuff,
and
there
were
a
couple
more
issues
around
the
repository,
apt
repositories
and
others
some
of
those
go
upstream
and
we
should
probably
front
load
these
as
well,
because
they're
probably
longer
cycle
there.
C
A
A
A
E
A
F
Think
we
we
should
engage
with
CSI
in
v1.
They
recently
is
because
when
we
could,
when
we
create
volumes,
the
volume
made
our
data,
this
object
metadata
is
a
it
would
not.
It
will
have
like
clicked
on
you,
for
example,
you
would
not
have
CSI
and
then,
if
that
happens,
and
if,
when
we
do
an
app
upgrade
to
UCS
I,
would
turn
the
switch
uses
I.
It
would
be
a
difficult
migration
because
we
had
to
go
through
all
these
PVC
objects
or
people
objects
and
convert
them
from
using
flexball
you
underneath
to
CSI.
A
F
A
Trying
to
understand,
if
we're
committed
to
it
and
who's
gonna
do
the
work
around
CSI
and
then
there
is
also
there's
also
the
user
mode
support.
As
part
of
that
which
is
currently
you
know,
we
have
plans
to
put
down
in
a
rogue
agent,
but
but
I
wonder
if
we're
you
know
that
belongs
in
CSI
as
well.
Yeah.
F
There's
two
things
that
that
we
need
for
CSI
and
having
use
remote
volume.
Those
are
the
thing
those
are
CSI
and
propagation,
which
are
both
alpha
right
now,
Cooper,
so
coordinated
1.9.
But
as
talking
to
the
kubernetes
guys
this
the
sad
set,
he
sees
CSI
and
my
probation
to
be
a
beta
for
one
for
one
that
ten.
We
also,
we
need
rob
block
storage
as
well
to
be
beta.
F
E
A
A
E
F
F
F
So
this
there's
a
lot
of
pieces
that
we
need
to
be
beta
and
in
roblox
stories
and
it's
never
gonna
be
part
of
a
flex
volume.
I
asked
during
the
Q
Khan
whether
we
are
whether
flexibility
ever
gonna
support
roblox
storage,
because
that's
needed
for
local
volume
and
they
say
no,
it's
not
gonna,
be
supported
and
with
no
gonna
accept
it
because
we
won't
didn't
want
to
change
the
API.
So
you.
B
Steven
I'm
not
sure
you're
answering
but
Dan
is
getting
at
it.
So
it's
more
of
a
question
of
you
know:
win
win.
Do
should
be
really.
You
know,
get
off.
Our
asses
infinity
go,
get
something
ships
food
that
you
know
that
supports
this.
You
know,
CSI
I,
think
that's
really
the
more
the
question
you
know
my
my
response.
That
would
be,
at
least
after
you
know,
1.10
beta.
You
know
not
at
all
before
that
is
you
know
about.
B
A
Guess
I'm
coming
at
this
from
the
perspective
of
if
we're
gonna
keep
building,
you
know
into
our
flex
volume,
and
you
know
investing
in
the
flex
volume.
Is
that
a
is
that
a
rational
decision
or
do
we
should
we
just
start
investing
in
CSI,
that's
kind
of
I'm,
especially
thinking
about
when
we
start
doing
user
space
stuff
like
running
and
BD
RBE,
or
you
see
mu,
Runner
and
others
should
that
work
go
into
the
Flex
volume?
My.
E
A
A
E
E
E
A
E
F
A
F
E
Kind
of
even
though
there's
a
Cuban
SCSI
organization
and
currently
most
of
the
drivers
over
there
in
the
past,
as
Cassie
is
become
under,
it's
been
tested
as
well,
so
Daphna
Steinberg
can
fork
it
and
the
crew
eats
his
own
repo.
So.
F
Plugins
but
stuff,
like
general
generic,
like
CSI,
ice,
kasi
and
and
MBDA
know
that
there
will
be
a
special
repository
for
it,
but
from
from
what
I
heard
that
they
don't
want
them
in
teen
over
a
super
story
maintain
something
that
is
that
that
that
that
they
don't
understand
the
code.
For
so
that.
Does
it
think
that
at
that
that
I'm
not
sure
about.
A
A
A
F
B
A
A
D
A
G
Yeah
well
that's
kind
of
depending
on
how
Jared
sees
it
too,
because
right
now
the
status
is
that
it
would
work
with
one
manager
right
now,
but
I
think
I,
don't
know
his
name
right
now,
but
he
wrote
that
he's
working
on
a
patch
upstream
for
prometheus
metrics
module
and
that
this
would
fix
so
more.
We
can
revert
our
change
to
back
to
one
manager
back
to
2
and.
G
E
G
So
yeah
I
said
we
kinda
can
go
ahead
and
go
merge
it
right
now
when
I
have
when
a
fixed
chair
is
common,
and
then
we
have
many
metrics
on
one
manager
and
when
this
patch
is
released,
as
I
wrote,
we
can
just
with
patch
to
pull
request
to
revert
the
reward
of
I'll,
see
using
only
one
manager
back
to
two
managers.
We
can
just
kind
of
put
in
the
patched
version
of
the
map
for
ratios,
metrics,
module
or
diversion
suffers
and
I.
Don't.
G
When
a
node
fails
and
it
takes
when
a
node
fails,
it
takes
about,
depending
on
how
you
set
it
about
1
or
2
minutes
on
titles
and
noticed
even
set
is
not
ready
and
the
second
point
there
is
when
another
is
not
ready.
It
takes
another
I
think
it
was
about
five
minutes
till
the
on
that
note
are
you've
moved
to
another
note.
So
if.
B
Yeah
I
think
that
I
think
you're
right
Alex.
That
I
mean
go
ahead
forward
now
with
merging
the
work
you
have
now.
After
the
last
couple
comments
in
that
P
R
such
that
we
have,
you
know
the
manager
module,
you
know,
reporting
metrics
and
then
you
know
in
the
future.
When
you
can
have
multiple
instances
of
manager,
then
they'll
go
back
and
revisit
that,
but
yeah
I
think
we'll
be
good
to
go
ahead
as
soon
as
you've
finished.
The
last
comments
on
that
PR
yeah.
E
A
D
A
E
G
A
G
G
Yeah,
just
one
thing,
maybe
to
keep
in
mind
that
it
depends
on
which
level
the
users
want
to
do
the
backups,
because
they
kind
of
have
two
things
they
want
to
do
right
now.
They
want
to
bring
back
a
through
storage
providers
like
us,
Luke
with
an
API
and
backups
on
the
files
file
level,
first,
something
like
stash
or
rustic
yeah
for
us
or
information.
So
they
kind
of
want
to
do
both
depending
on
what
you
you
are
using.
As
a
user.
F
E
A
E
B
D
Yeah,
it's
just
you
know,
with
ODOT
7,
for
example,
I
know,
there's
at
least
one
change
I
had
where
you
know
the
the
upgrade,
maybe
not,
that
the
upgrade
steps
were
out
of
date.
But
if
you
didn't
follow
them,
then
you
run
into
issues
series
upgrade.
Maybe
that
was
it
but
yeah.
If
upgrades
a
complicated
problem,
so
I
was
you
know
not
having
started
on
that.
Yet,
if
we're
saying
we
want
it,
when
we
reach
beta
I
guess
it'd
be
after
beta
when
we
say
we'd
have
it
it.
B
Feels,
like
you
know,
we've
been
kind
of
limping
along
right
now,
with
you
know,
a
manual
user
guide
and
a
fully
fleshed
out
design
for
how
upgrade
would
work.
You
know,
we've
accomplished
those
tasks
but
yeah
the
the
the
fact
that
we
haven't
actually
started
writing
code
for
it
and
the
operator
is
you
know
it's
that's
gotten
it's
concerning,
but
I
don't
know.
If
I
would
you
know
place
it
ahead
of
any
other
things
we
have
to
do
right
now
to.
E
A
A
I'm
just
saying,
and
then
the
other
one
I'm
concerned
about
is
kind
of
refactoring,
the
CR
DS
to
support
multiple
backends
or
at
least
making
sure
that
we
have
the
ability
to
add.
You
know
at
least
separate
the
parts
that
are
self
specific
from
others.
I
know
Travis.
You
started
some
of
this
work
right.
G
Think
we
should
go
ahead
and
create
some
kind
of
design
talk
again
for
that
Smurf
like
a
I.
Think
such
a
blueprint
for
hey,
you
want
to
add
storage
back-end.
How?
Where
would
you
need
to
put
change
things
or
overall,
how
about
the
CR
D
log?
Because,
yes,
I
can
decide
what
the
thing
with
upgrade
Oh
topic
before,
with
zero
to
seven
figure
like
yeah,
it's
better
and
we
are
going
to
change
society.
Most
users
understand
that
there
are
change,
is
going
to
happen
in
alpha,
but
not
in
beta
or
possible.
A
Once
once
the
series
declared
beta,
then
it's
you
can
have
breaking
changes
to
them.
You
have
to
have
migrated
them.
So
that's
that's
exactly
why
I'm
bringing
this
up
right
now
at
minimum
at
our
cluster
and
pool
series
should
get
some
treatment
for
you
know
what
parts
were
so
specific
and
what
parts
are
not
so
that
when
we
declare
them
beta,
it's
not
a
it's,
not
a
big
pain.
I.
E
Danna,
it
feels
a
little
naive
to
jump
ahead
and
write
a
design
doc
about
what
it
would
take
to
do
all
this
work.
You
know,
given
that
you're
not
actually
doing
the
work.
I
mean
it.
It
just
sounds
unlikely
you're
going
to
get
that
correct.
That
I
think
I
sort
of
agree
about
the
Sun
said:
let's
do
the
obvious
thing
and
I
could
certainly
imagine
writing
up.