►
From YouTube: 2022-11-15 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
Well,
first
of
all
for
the
1.9
release.
We
did
have
a
patch
release
199.13
almost
two
weeks
ago
now,
and
it
did
include
a
a
couple.
Little
security
fixes
as
far
as
the
crash
collector,
pod,
running,
less
privileged
and
then
I
think
also
getting
the
latest
a
binary
for
the
S5
command
tool
yeah.
So
your
release
notes
to
look
yeah
running
the
crash,
collector
zap
user
and
updates
S5
command,
which
I
don't
know
of
any
issues
with
that
spoke
command
other
than
just
some
security
scan
was
complaining
about
an
older
version.
A
Yep
so
for
1.9
we
don't
currently
have
another
release
planned
for
it.
It's
certainly
possible,
depending
on
what
comes
up
in
the
future,
but
I
think
we're
looking
pretty
stable
for
1.9
any
comments
on
1.9.
A
All
right,
then,
for
1.10
we
also
at
the
same
time
had
a
1.10.5
release
with
those
same
security
releases
and
a
number
of
other
fixes.
Let's
see
if
I
just
glance
at
the
release,
notes
real
quick
yep,
a
number
number
of
things
backboarded,
let's
see
anything
worth
mentioning
here,
for
the
the
recording.
A
Let's
see,
yeah
updated
the
latest
separately,
1725
doc
updates
some
other
minor
things,
but
I
guess
nothing
stands
out
for
me
to
mention
so
much
for
1.10.6,
we'll
plan
on
that.
This
Thursday
for
bi-weekly
Cadence
we'll
keep
that
and
get
a
few
fixes
out
there
for
yeah,
probably
just
on
that
Cadence
I.
Don't
think,
there's
really
anything
of
interest
to
look
at
in
the
project
board.
Honestly
I
missed
looking
at
it
this
week
to
make
sure
we're
up
to
date.
A
All
right
anybody
else
have
comments
on
1.10
nope,
all
right.
Moving
on
then
1.11,
no
real
updates
on
my
side,
still
planning
on
around
mid-January,
probably
or
whatever
how's
the
project
board.
Looking
I've
got
a
few
things
added
here:
I
guess
as
we're
working
on
new
features
that
we'll
stay
in
master.
For
now,
please
just
make
sure
they're
tagged
for
the
1.11
board.
A
All
right
and
that's
all
I
have
for
land
releases
anything
else.
Anybody
all
right
go
ahead
for
the
topic
section,
kubecon
retrospective,
so
Alex
I
think
you
were
the
only
one
in
the
group
here
that
that
made
it
so
curious
for
your
feedback
on
how
the
booth
went
or
or
other
things
there
related.
B
Yeah,
which
was
nice
meeting
anyone
there,
anyone
that
was
there.
B
Well,
there
isn't
too
much
to
say,
like
the
we
had
like
a
small
kiosk,
which
was
only
part-time,
we
were
able
to
get
to
give
out
quite
a
lot
of
well
they're
a
bit
aged
now,
but
five
year
anniversary,
Rook
t-shirts.
So
there's
that.
A
B
Quite
some
with
size
s,
small
left,
but
okay,
well,
yeah.
A
B
How
did
that
work?
It
was
then
changed
to
I
think
yeah,
okay,
which
I
think
to
put
it
more
or
less
into
that
perspective,
that
we
were
basically
two
people.
B
Well,
Mike
Paris,
mainly
yeah
Annette
I,
think
like
or
like
some
time
as
well.
Mainly
it
was
Mike
and
me
basically,
so
it
was
good
that
it
was
only
half
time.
B
Yeah
I,
don't
really
have
anything
to
say
to
the
virtual
talk.
I
was
at
the
booth
at
the
time
right.
A
D
B
B
B
Or
April
I
think:
okay,
let
me
I'll
Google
in
a
Google
it
in
a
second
yeah,
oh
April,
17th
to
21st
April,
okay
by
21
April.
B
You're
right,
a
one
that
didn't
join
Us
in
Detroit.
B
A
A
All
right,
that's
all
for
that.
A
couple
small
items
on
the
agenda
to
to
just
bring
up
real
quick,
so
this
Snick
security
scan
is
failing
due
to
something
a
CV,
that's
already
fixed
in
Rook
I,
just
mentioned
that
in
huddle
a
few
minutes
ago,
but
just
to
reiterate
here,
there's
a
link
to
it
there
and
like
to
need
to
figure
out
how
to
resolve
that.
So
our
scan
turns
green
again,
but
that's
the
tool
issue.
Since
we
know
it's
already
fixed.
A
Now
the
next
topic
I
put
in
the
agenda
there
is
just
thinking
about
the
community
and
questions
on
slack
and
GitHub
and
yeah
I.
Just
want
to
you
know:
ask
people's
thoughts
on
you
know.
How
can
we
spread
the
load?
I
I
really
like
I
enjoy
spending
time
there
answering
questions
and
so
I
intend
to
maybe
take
on
too
much
there,
but
I
think
it.
You
know
just
to
get
time
zone
coverage
or
make
sure
people
can
see.
A
Multiple
people
answering
questions,
I
think
it's
just
healthier
for
the
community.
If,
if
I'm,
not
the
main
one
answering
questions
all
the
time,
so
like
people
will
tag
me
and
to
answer
the
question.
Well,
there's
other
people
who
can
answer
the
questions
too.
So
I
think
if
we
can
spread
somehow
I've
never
felt
like
there's
a
good
way
to
spread
the
load
without
making
it
an
assignment,
and
then
it's
just
a
pain
to
track
who's
assigned
or
whatever
I.
A
B
I'm
currently
hoping
that
that
I
can
get
garof
to
help
on
the
slack
more
from
our
side.
Basically,
did
you
put
it
like
that?
Yeah.
C
B
I
get
maybe
just
as
a
I
think
we
kind
of
had
like
I,
think
I
initially
tried
talking
about
it
some
time
ago
already,
but
it's
like
it
still
feels,
like
slack,
isn't
really
the
right
tool
to
help
people
because,
like
it's,
if
we
like,
we
we're
using
threadsden
and
like
at
least
if
you
help
three
people
at
the
same
time,
the
same
time
and
it's
just
or
like
it's
a
mix
between
hey
yeah.
Can
you
run
this
no
response
for
like
a
week
or
even
a
month
or
something
someone
comes
quickly?
Oh
yeah!
B
B
At
other
times,
where,
like
you
immediately
getting
a
response,
or
it's
just
that's
like
I,
don't
know
it's
a
bit
like
it's
itself
is
a
slack
even
with
like
the
real-time
communication.
There
isn't
really
the
right
thing
there,
but
I
wouldn't
really
know
an
alternative,
and
that
would
involve
that
would
not
involve
us
switching
to
like
a
different
communication
tool
that
is
not
Slack.
A
B
It's
like
it's
getting,
not
unorganized,
but
it's
like
getting
all
over
the
place,
the
more
people
you
try
to
help
at
the
same
time,
because
again,
like
you'll,
respond
and
then
either
five
minutes
later
or
a
week
later,
you're
getting
a
response
like
what
is
going
on,
basically,
depending
on
when
the
response
is.
A
Yeah
but
I
feel
like
as
far
as
the
tool
I
mean.
There
are
other
tools
like
that
at
Red
Hat
we
had
to
have
Gchat
and
I
find
it
way
more
annoying
than
Slack,
so
I
don't
know
that
it
would
help
even
to
try
and
find
another
chat
tool
if
we
just
assume,
let's
keep
slack
and
yes,
it's
annoying,
how
it's
organized
but
I,
think
any
sort
of
real-time
chat
tool
is
going
to
be
it's
going
to
have
that
feel
of
not
so
much
organization.
A
B
C
C
Yeah,
just
on
my
opinion,
like
I,
usually
go
to
GitHub
issues
and
see
what
I
can
do,
but
for
slack
majorly
I
find
some
difficulties
in
saying
that,
like
there
are
lots
of
messages
at
the
end,
I
I
go
and
open
on
channels,
but
most
of
the
times
it's
helpful
like
whenever
someone
or
you
like,
add
me
or
tag
me
there,
then
I
go
and
try
to
spend
more
time,
so
maybe
at
the
starting.
C
Maybe
you
can
try
like
that
attacking
different
people
like
us
and
others
so
I
guess
it
would
be
more
better.
A
And
that's
very
and
I
think
it's
totally
fine.
If
some
people
plan
responding
on
GitHub
easier
or
some
people,
like
slack
YouTuber
yeah,
it's
where
it's
still,
it
is
ultimately
an
upstream
Community
where
we
don't
have
a
commitment
to
go
respond
in
a
certain
timeline
or
anything.
It's
just
I
just
want
to
make
sure
we're
finding
ways,
especially
as
maintainers,
to
kind
of
figure
out
how
to
keep
the
community
going
as
part
of
that
maintainer
role.
A
So
maybe
yeah,
let's
think
about
it
as
just
something
on
my
mind:
I
don't
have
a
formal
solution:
I,
don't
I
hope
we
don't
need
a
formal
solution.
D
B
But
the
problem,
at
least
for
example,
for
us,
is
that
at
least
we,
for
example,
use
slack,
connect
and
stuff
on
which
their
snow,
like
slack,
connect
to
solid,
basically,
which
is
like
I
expected,
but
yeah
I.
Don't
think,
there's
I,
don't
think.
There's
really
a
way
to
do
that
I've.
Maybe
one
idea
would
be
to
eat
a
I,
don't
necessarily
want
to
say
Force,
more
moderation,
but
at
least
I
guess
kind
of
like
there's
still
a
lot
of
like
as
like
Rook
is
focusing
on
purely
safe
at
the
moment.
B
It's
still
that
does
quite
some
questions
of
like
General
discussions
and
a
general
journal
as
an
example
where
it's
like,
hey
I,
have
to
sufficient.
Like
can
you
please
use
the
surf,
Channel,
basically
kind
of
like
this
yeah
I,
don't
know
like
it's
good
might
make
sense
to
I
guess,
boss
people
into
the
correct
channels,
depending
on
what
the
issue
is.
A
B
A
A
B
A
A
B
Yeah,
let's
check
it
up
as
a
as
a
top
action
item.
Could
you
edit
yeah.
A
B
Hope
they
have
it
open,
but
I,
don't
where
is.
A
B
Because
it
is
for
me
to
I
guess,
that's
I
think
the
most
I
get
infuriating
partner.
It's
just
really
attractive
like
this.
Hey
I
need
help
and
in
other
chances
like
so
you
know
it
helped
with
one
person
or
at
least
throw
them
the
question
and
then
five
seconds
later,
it's
like
wait,
didn't
I
just
already.
A
B
Them
help
you
and
yeah
well,.
B
Not
really
because
maybe
look
into
like
using
some
apps
or
something
that
I
guess
force
people
to
use
like
Fred's
or
like
autocrate
threats,
if
that's
even
possible
but
I,
don't
know
if
something
like
apps
for
that
exist.
So.
A
Okay,
well
with
that,
let's
move
on
and
so
Alex
said
the
bulk
of
I
think
our
meeting
will
be
your
topic.
So
let
me
just
skip
real
quick
to
the
one
Jerry,
not
Jared
added
and
we'll
come
back
to
you,
so
he
he
forwarded
an
email
that
someone
sent
him
directly.
I
guess
something
about
some
issue:
they're
having
so
I'll
just
respond
to
that
or,
if
needed,
ask
them
to
open
an
issue
or
or
whatever,
but
I
haven't
looked
at
that
yet
so
anyway,
that's
all
for
that!
A
B
Yeah,
so
this
is
still
pretty
rough
at
certain
aspects.
Yeah,
it's
like
the
idea
is
to
well
improve
Cloud
nativeness,
more
more
or
less
it's
more
like
a
I
guess.
The
computer
kubernetes
integration
there
of
self
component
slash
Rook
into
this
whole
ecosystem.
B
Next
slide.
Please.
B
Great,
you
know
how
it
goes.
It
just
works,
but
sometimes
it
doesn't
yeah
so
kind
of
the
idea
would
be
to
like,
instead
of
I
guess
allowing
local
disks
or
like
like,
instead
of
directly,
for
example,
accessing
like
a
Dev
SDA.
B
Sure,
okay,
then
yeah
second
point
is
basically
about
like
adding
like
something
else
like
a
safe
OSD
resources.
I
think
one
of
the
most
prominent
ones
to
rook
in
some
way,
so
that,
if
you
want
to
OST
that
is
running
on
a
certain
disk
or
something
or
even
that,
it's
like
a
mirror,
more
or
less
of
what
is
in
the
surf
cluster.
B
That
depends
on
like
what
way
we
would
want
to
go
down
is
that
you
could
technically
use
just
Cube
CTL
to
do
like
Cube
CTL
create
manifest
with
like
a
safe
ostn,
which
targets
a
certain
persistent
volume
or
even
an
existing
PVC,
and
that
will
be
then
created
that
we
basically
have
like
custom
resource
for
that.
B
We
can
do
like
at
least
something
like
Cube
CTO
get
safe
OSD,
and
we
would
at
least
have
like
a
mirror
of
what
is
going
on
in
the
surf
cluster,
like
oh
osd0123
and
so
on,
listed
naturally
in
kubernetes
thread
point
it's
mainly
about
with
kind
of
this,
these
resources,
or
these
custom
resource,
for
example,
of
like
surfersd.
The
idea
would
be
that
we
could
move
certain
customizabilities
to
these
objects
so
that,
if
some
customer
needs
some
special
labels
on
this
one,
specific
OSD
or
even
some
other,
more
specific
things.
B
It's
on
like
a
cephosd
object,
instead
of,
unlike
the,
for
example,
with
the
ceft
cluster
I,
guess
been
quite
long
depending
on
how
many
options
you
or
parameters
you
want,
especially
even
if
how
detailed
you
want
to
go
per
node,
and
this
might
be
like
a
I
guess.
I
was
a
bit
of
like
splitting
it
up
in
regards
to
the
resources
we
offer
to
the
users
and
also
kind
of
has
like
a
splitting
up
certain
parts
or
like
certain
Logics
of
the
operator,
to
be
more
independent
of
each
other
yeah.
B
Okay.
So
for
this
kind
of
like
with
the
PVCs
kind
of
I,
guess,
doesn't
this
disclaimer
on
like
the
next
slide
as
well?
There
are
already
projects
that
do
that.
B
The
main
idea
is
that
simply
we
for
the
first
point
on
the
plan,
we
would
need
something
which
would
see
disks
and
well
represent
them
in
some
way
in
kubernetes
is
like
an
object,
or
even
immediately
as
a
persistent
volume
for
Acclaim
than
to
be
made
on
this
persistent
volume
to
be
used,
then,
as
in,
for
example,
for
example,
foreign
OST,
the
stress
kind
of
at
least
so
I
thought
about
it
was
like
that's,
you
could
have
discs
or
in
partitions,
and
you
see
the
class
of
time
and
so
on.
B
B
I
think
this
is
more
or
less
like
a
if
we
want
to
really
go
for,
like
everything,
PVC
or
everything,
persistent
volume,
basically
what's
project
to
prefer
to
use
here,
or
at
least
recommend
to
the
users
if
they're
on
their
metal.
But
in
the
end
we
shouldn't
like
reinvent
the
wheel.
Like
that's
a
good
point
from
you.
There
Travis,
like
we
shouldn't
reinvent
the
wheel.
Basically,
it's
like
it
just
I
wasn't
aware
of
top
lvm
and
local
storage
operator
there
when
I
created
the
initial
slide.
B
A
Yeah
and
I
mean
throughout
the
past
couple
of
years,
since
we
we've
added
the
storage
class
device
sets,
which
is
all
about
managing
with
PBS
and
PVCs
under
osds
right,
so
the
I
mean
we've
had
discussions
about
well,
should
we
get
away
completely
from
the
raw
devices
and
try
and
convert
everybody
to
the
PVCs
and
where
I
felt
like
in
my
mind,
ended
up
is
that?
Well,
since
we
don't
want
to
reinvest
the
wheel,
we
would
be
pulling
into
a
dependent,
it
required
dependency
of
topology
lso
or
something
similar.
A
So
if
we
pulled
in
any
a
dependency
and
then
that
it
would
just
come
with
a
lot
of
challenges
like
converting
the
Clusters
or
pulling
in
this
dependency,
where
we
don't
have
the
dependency
at
all
today
yeah,
so
there
was
just
those
are
just
some
of
the
challenges
but
I.
Think
Blaine
yeah
have
some
thoughts
on
that
too,
or
since
we've
talked
about
it,
but
I,
don't
know
playing
anything
to
add.
D
Nothing
particular
I
I
think,
like
a
general
ideas,
I'm
sound
good
I
mean
I.
Also
know
it's
going
to
be
a
lot
of
churn
to
make
the
kind
of
those
kind
of
changes.
A
B
B
D
D
Conceptually,
it
is
pretty
pretty
simple:
I
I
think
it
is
just
a
lot
of
Parts
with
like
complex
interactions
between
them
all
and
then
and
then,
as
you
mentioned,
also
the
migration
which
I
mean
we
have
done
OSD
migrations
before
and
I
think
we
have
like
when
we
did
file
store
to
Blue
Store.
We
required
users
to
like
delete
osds
one
by
one,
if
I
remember
right,
so
that
might
end
up
kind
of
being
a
decent
option.
D
D
Do
feel
like
some
specialized
tooling,
that
might
be
important,
like
Topo
lvm
can
very
easily
carve
up
disks
and
do
stuff,
but
it
may
be.
It
may
be
useful
to
have
something
that
is
akin
to
like
the
what's
it
called
the
discovery
demon.
D
That
is
also
aware
of
like
how
to
check
discs
for
whether
or
not
they
have
like
a
ceph
system
on
them
like
they
know
how
to
check
for
the
Blue
Store
header
effectively,
as
well
as
using
standard
Linux
tools
for
determining
whether
the
disk
is
in
use
or
things
like
that,
which
could
help
with
migration
or
potentially
could
be
used
to
like
feed.
The
idea
of
use
all
disks
that
are
available.
B
D
D
Yeah
as
I
like
as
a
technical
detail
on
the
like
side
of
this
I
I
I
talked
to
Travis
about
this.
A
bit
ago,
when
we
were
talking
about
the
Cozy
driver,
I.
D
D
It
becomes
really
challenging
like
I
know,
manipulating
a
yaml
isn't
much
different,
but
there
are
a
lot
of
good
tools
that
help
people
like
autofill
yaml
files
when
crds
are
defined,
that
I
I
think
if
we're
like
changing
stuff
around
this
like,
if
we're
like.
Oh,
we
can
like
sort
of
create
or
like
reuse
parts
of
the
discovery.
D
Demon,
I
I
would
also
suggest
that
maybe
we
create
acrd
for
that,
and
even
if
we
set
up
the
system
such
that
it
is
a
oh
I
forget
what
the
programming
terminology
is
for
this,
but
we
basically
just
set
it
up
so
that
Rook
says
if
there
are
two
of
them.
D
We
just
give
an
error
like
there
should
be
one,
and
there
should
only
be
only
ever
be
exactly
one
in
the
system
to
help
configure
that
demon,
just
because
putting
configurations
on
that
operator
starts
to
seem
to
just
kind
of
blow
up
after
dealing
with
everything
like
annotations
and
resources
and
Etc.
B
Maybe
to
add
photos,
I
guess.
Is
it
like
other
approach,
except
on
it
in
like
such
a
way?
The
solando
postgres
operator,
for
example,
has
two
ways
with
either
reads
the
configuration
from
a
conflict
map
like
for
itself
like
for
the
operator.
It
reads
it
from
a
config
map
or
it
reads
it
from
like
a
crd
which
is
well
to
some
degree,
just
everything
you
they
ask
if,
in
for
the
in
the
conflict
map,
they
just
put
it
into
the
CID,
but.
D
A
Yeah
and
in
the
chat
I
just
put
a
link
to
an
lvm
operator
which
it
works
with
topple
lvm.
The
top
lvm
is
basically
the
CSI
driver
right
for
that.
You'd
create
the
osds
with
on
the
PVS,
but
the
and
this
lvm
operator,
it's
something
we
use.
Downstream
I
mean
it's
open
source,
but
we've
only
we
haven't
pushed
a
community
for
it,
and
so
it's
really
just
a
downstream
thing.
But
really
this
is
exactly
what
it
does.
B
Regarding
like
Sears
or
like
well
crds,
and
all
that
it's
like
yeah
I,
don't
know
it's
it's,
it
would
just
feel
cool
in
the
end,
if
you
could
do
like
Cube
City
against
fosd-
and
you
know
like
you-
would
get
us
to
see
all
the
osds
and
even
if
you're
like
oh,
what
disc
is
obviously
using,
you
could
just
do
like
a
like.
B
You
can
kind
of
do
it
already
with
like
what
Seth
is
able
to
talk,
or
so,
for
example,
this
might
be
even
something
to
implement
on
like
the
crew
plug-in
side
like
this,
you
know
with
the
custom
status
columns
the
swelling
kubernetes
as
we
already
use
them
for
surf
cluster
I.
Don't
know
you
know
it's
just
something
different,
it's
so
cool
to
be
able
to
run
cubes.
Did
you
get
suffocate?
B
All
of
them,
don't
know
they're
on
which
disc
or
what
device
class,
and,
if
you,
if
I,
want
to
replace
it
or
so
I
would
I,
don't
know
edit
it
and
set
like
a
you
will
be
replaced
and,
if
I
just
like,
if
the
disk
is
removed,
the
system
is
automatically
cleaning
that
up
and
waiting
for.
Like
a
new
disk
and
if
a
new
disk
is
even
added,
that's
like
on
the
idea
Sheet
later
on
as
well,
it's
not
here.
It
would
just
automatically
oh
new
disk
and
magically
heal
itself.
Basically
again,.
B
A
Well,
I
think,
like
you
said:
if,
if
it's
just
like
print
the
status
like
show
me
the
osds
and
what
their
device
classes
are
and
stuff
like,
the
group
plugin
could
easily
go,
get
that
status
and
print
it
out
for
sure
or
there's
theft
commands
that
can
show
all
sorts
of
details.
A
There's
and
the
crew
plugin
could
leverage
that
and
I
mean
the
operator
already
does
a
number
of
things
to
let
you
deploy
these
automatically
today,
based
on
the
Discovery
demon
and
not
with
PVCs,
though,
because
right
now,
today,
the
approach
with
storage
class
device
sets.
Is
you
tell
us
well
how
many
osds
do
you
want?
A
So,
if
you're
adding
a
disk,
that
number
has
to
be
updated
currently
anyway,
so
yeah,
there's,
certainly
areas
for
improvement,
but
I
think
without
completely
redesigning
how
we
deal
with
PBS
I
think
we
can
get
a
long
ways,
or
maybe
everything
we
need
to
do
can
be
done
with
existing
patterns.
So.
B
Just
like
on
at
least,
we
currently
Envision
it
with
like
the
core
search
distribution
there,
but
the
idea
is
that,
instead
of
like
having
I,
guess
something
that's
just
the
root
stuff
operator,
something
as
small
as
just
sitting
on
top
and
being
like
Oh
there's
a
new
disc
OST
created
and
like
so
that
it's
small
as
like
just
sitting
on
top
and
creating
what
is
desired
by
the
user
that,
like
oh,
you
use
every
disk.
Oh
so
every
new
disk
is
also
just
their
own
new
disc
stefo
is
the
object
created
and
so
on.
B
So
yeah
I
would
maybe
propose,
if
that
we
move
on
with
the
slides
at
this
moment,
just
to
I
guess
get
through
with
them
and
not
all
of
them
are
I
said
not
everything
there's
fully
fought
out
yet,
but
it's
just
like
to
I
guess
that
we
get
Freedom
more
or
less,
and
then
we
can
just
also
talk
more
about
discussions
there
and
see
if
we
might
want
to
take
some
discussions,
also
I,
guess
offline
from
this
meeting.
B
At
the
moment
you
brought
up
the
point
three
percent
in
regards
to
like
crds
versus
group
plugin.
B
As
you
like
said
with
it's,
even
for
like
a
get
OST,
we
could
have
something
like
that
in
the
crew
plug-in
easily.
Let's
put
for
example,
then
just
call
out
to
like
surf
osd3
command,
for
example,
or
ask
the
list
of
deployments
the
OSD
deployments.
We
have
it's,
it's
just
I,
guess
a
different
way
than
to
like
I
guess
show
this
information.
Potentially
when
you
like
to
cube
City,
get
service,
OSD
versus
you
have
to
crew
plug
and
install
it,
and
at
least
how
I
see
this
with
cids
versus
group
plugins.
B
It's
mainly
about
this.
How
can
you
deploy
it
with
crds?
It's
technically
you
just
throw
it
into
your
like
get
UPS
or
whatever
tool,
and
it's
taken
care
of
and
for
like
running
a
map
like
a
crew
plug-in,
create
OSD
command
or
something
that
would
need
to
be
done
like
well
manually
by
an
admin
normally
and
like
you
could
automate
this
like
ansible
and
well.
Bash
scripts
are
out
there,
but
for
most
I
think
it's
might
be
a
bit
of
like
for
like
creating
something.
B
So
like
this,
as
you
said
it
like
this,
the
science
state-
it
might
be
a
bit
problematic
to
have
it
like
just
full
Automation
Way,
but
where
the
crew
plugin
is
awesome,
it's
kind
of
as
I
put
it
as
like
in
in
Brackets.
Here
it's
like
technically
with
the
crew
plugin,
you
could
have
like
a
new
OSD
wizard
like
with,
if
anyone
still
installs
programs
I,
would
like
the
nice
installation
bristles
which
go
like
hey.
Where
do
you
want
to
software
to
be
installed?
Which
features
do
you
want
to
install?
B
You
know
this
is
kind
of
how
I
see
with
the
crew
plug-in,
like
the
group
link,
is
more
of
this
simply
interactive
tool
to
like
get
input
from
the
user.
I
guess
right.
D
Yeah
I
think
all
that
makes
sense
to
me
also
sort
of
extending
this
to
what
we
were
talking
about
a
little
bit
later
or
a
bit
a
little
bit
earlier.
Rather.
D
We
we
could
consider
something
like
a
wizard
for
doing
maybe
the
sort
of
like
we've
talked
about
kind
of
like
the
the
play
install
of
Rook,
where
it's
just
like
use
all
devices
where
we
don't
necessarily
recommend
people
do
that
in
their
actual
environments.
But
we
like
it
is
an
option
in
the
crd.
If
we
had
a
like
crew
plug-in
command.
That
was
just
like
you
know,
quick
start
that
would
just
install
stuff
really
quickly
and
also
run
like
a
a
job
or
like
a
demon
set
on
all
of
your
nodes.
D
That
would
just
look
at
the
disks
and
populate
it
can
a
config
map
of
stuff
that
could
be
used,
then
that
plugin
could
also
automatically
fill
in
some
details,
and
then
it
wouldn't
necessarily
be
on
us
to
make
a
like
constantly
running
sort
of
system.
To
do
the
same
thing.
It
would
be
fine
for
it
to
be
like
kind
of
logic
that
just
exists
within
the
group
plugin,
just
as
a
sort
of
kind
of
idea
and
CNF.
B
The
way
it's
not
about
like
removing
the
crew
Plugin
or
something
like
making
it
obsolete
as
well,
it's
just
about
like
yeah,
I,
guess
finding
ways
to
like
make
it
I,
guess
a
bit
more
native
for
your
kubernetes
and
then
just
being
able
to
do
like
even
like
a
Cube,
CTL
Rooks
F
create
OSD
and
like
then
it
as
we,
as
you
said
like
then
there's
like
the
wizard
a
lot
of
stuff
popping
up
and
all
it
would
do.
Would
then
just
like
create
a
suffered.
It
was
the
like.
B
A
Yeah
and
and
the
way
I
see
that
working
too
with
the
crew
plugin
is
the
crew
plug-in
it
would
take
input.
It
would
figure
out
what
you
want
to
do.
It
would
see
what
your
existing
CRS
are,
especially
across
your
CR.
If
you're
adding
OSD,
then
it
would
say:
oh
now,
I'm
going
to
update
or
create
the
cluster
CR
so
that
it
will
create
the
osds
for
you
and
the
crew
plugin
itself
wouldn't
go
like
create
osds
or
anything.
D
Yeah
I
yeah
I
mean
depending
on
which
route
we
want
to
go.
It
could
potentially
be
multi-parted
where
it
sends
some
requests
to
lvmo
and
then
once
the
disk
is
like
created
in
lvmo,
then
it
puts
that
into
the
storage
class
device.
Sets.
D
And
then
we
potentially
start
getting
users
more
toward
the
PPC,
only
kind
of
future
that
we
envision.
C
D
B
Well,
I
guess:
how
I
see
this
a
bit
with
more
like
LCD
perspective,
on
with
asset
like
this,
someone
else
could
sit
on
top
and
be
like?
Oh
and
you
always
see
a
new
OSD
or
something
like
that,
or
even
Rook,
adding
mechanisms
to
be
to
get
smart
about
like
replacing
disks
as
an
example,
that's
like
yeah
yeah,.
D
Yeah,
the
this
replacement
is
one
of
the
like
I,
don't
know
it's
kind
of
one
of
the
more
challenging
aspects
of
of
kind
of
dealing
with
the
stuff
cluster
like
having
some
functionality
around
OSD
management
and
the
crew
plugin
would
be
pretty
pretty
nice.
A
A
B
Sure,
but
this
might
be
done
things
like
that,
we
would
be
able
to
work
on
to
some
degree
also
as
kind
of
like
I
guess
you
could
call
it
extensions
and
like
having
something
that,
like
watches
for
the
new
disk
on
the
same
node,
and
it's
done
like
I
said
like
oh
disk-
has
been
removed,
so
we
need
to
clean
up
and
oh
a
new
disk
has
been
added,
and
now
we
need
to
replace
this
OC.
Basically
well.
A
Yeah
and
and
I
guess
to
be
clear
in
terminology
too,
for
replacing
an
OSD
and
stuff
I
think
means
something
specific
where
they
try
and
have
the
same
LSD
ID
and
all
that
so
usually
in
Rook,
we
just
say:
remove
the
old
one,
create
a
new
one
and
there's
no
or
it's
just
not
considered
a
technical
replacement
from
the
Seth's
perspective.
A
B
B
But
like
you
know,
even
if
it
would
create
like
a
new
one
but
yeah
well,.
A
B
B
Yeah
well,
I
just
tried
to
I
guess
scribble
up
a
little
bit
of
like
I,
didn't
really
have
too
much
Infinity
on
Hall,
like
I
said
first,
he
could
look
like
it's
like
the
yaml
structure.
At
the
moment,
yeah
I,
don't
know
this
would
be
something
up
for
debate
for
later
on
I
think,
but
maybe
like,
let's,
let's
just
skip
to
the
next
slide.
I
think
this
is
one
of
those
points
where,
like
yes,
you
said
with
like
the
if
it
would
be
like
a
read-only
access
of
like
again
I.
B
Also
like
this
question
of,
should
it
should
like
a
crd
exist
for
like
every
safe
demon,
then,
should
it
be
like
a
maybe
then
just
read
only
for
sethmon
and
self-manager,
representing
the
demon
with
like
the
current
status
of
itself,
but
for
the
osds
it
would
make
sense
of
like
a
you
can
see.
What's
what
its
status
is,
but
also
create
ones
by
just
creating
a
sfos.
The
object
with
the
right
option
specified
and
I.
A
B
B
A
D
Yeah
I'm
I'm,
yeah
I'm,
still
I
think
trying
to
to
parse
this
one
like
we.
We
can
use
PVCs
for
the
mon
like
monmap
data
and
stuff
like
that.
Okay,
are
you?
Are
you
talking
about
something
yeah?
It
was.
It
was
lady.
She
had
some
stuff
to
say.
D
Know
well,
no
she's
left
so
I
finally
get
breakfast
but
yeah
I
guess
is
this.
D
A
B
Yeah,
it's
just
a
few
like
that's
kind
of
the
read
access
there.
It's
just
like
a
pure
status
kind
of
view
into
those
components,
for
example,
if
like
a
mon,
would
be
unavailable
for
whatever
reason
the
operator
could
like
set
the
status
of
this
object,
and
it
appears
in
like
a
column
when
you
do
Cube
Studio
against
someone
of
like.
D
D
I
see
I
see
so
it
effectively
would
be
a
crd
that
contains
nothing
except
a
status
element.
Yeah.
B
D
C
D
You
know
all
over
the
ends
of
the
Earth
to
get
information
about
their
cluster
that
we
do
have,
or
at
least
that
we
continue
to
have
some
aggregate
information
as
well.
Yeah.
B
Yeah,
okay,
yeah
next.
B
The
slide
is
like
some
way
to
see
the
device
devices
on
the
servers
having
like
wrapper
objects
or
something
kind
of
like
Lexus
on
the
last
slide,
and
that's
like
the
different
we
could
adjust
it
like
if
I
guess
the
status
only
or
like
a
you
can
create
a
receipts
or
even
that
like
it,
would
then
reflect
the
status
of
like
oste,
basically
kind
of
like
replicating
like
this
self-adm
orchestrator
or
surf
ADM,
but
self
manager
at
least
right.
D
Yeah
I
I
might
be
looking
ahead
here.
I
was
looking
at
the
osds
already
and
there's
like
the
question
of.
Is
this
a
mirror?
What's
an
SF
cluster
like?
Is
this
just
kind
of
a
status
only
thing
like
demons,
or
is
this
a
way
that
we
can
both
create
and
manage
osds
I
I
think
we
could
also
consider
I.
Don't
want
to
call
this
a
hybrid
approach.
We
could
consider
that
there
could
be
both
Greenfield
and
Brownfield
cases,
something
that.
D
Something
that
the
cozy
like
design
has
done
is
generally
the
way
that
users
interact
with
creating
a
bucket.
Is
they
create
a
bucket
request
and
then
cozy
eventually
creates
a
bucket
resource
that
represents
that
and
users?
Don't
normally
you
know,
users
should
never
actually
modify
the
the
bucket
objects.
D
We
can
I'm
kind
of
Imagining
the
you
know,
step
OSD
to
be
the
bucket
object,
where,
generally
speaking,
you
never
actually
interact
directly
with
the
OSD.
You
usually
interact
with,
like
I.
Have
this
definition
of
kind
of
what
I
want,
and
it
does
some
things
automatically,
but
I
ca
like
the
administrator
can,
if
they
so
desire,
create
a
manual
self
OSD,
which
could
be
a
case
that
we
want
to
help.
Allow
users
enable
a
Brownfield
case
where
they're
like
I,
have
this
pre-existing
OSD.
D
It
has
a
metadata
device,
I'm
gonna
like
put
stuff
in
the
right
places,
so
that
it
picks
that
up
that
could
allow
us
to
have
a
nicer
migration
path
for
users
like
if
we
had
a
tool
that,
like
looked
at
whatever
an
existing
cluster,
was
and
I
created
the
the
Brownfield
OSD
and
then
obviously
like
a
user.
If
they,
if
they
were
really
really
interested,
they
could
create
OSD
objects
manually
themselves.
D
I
just
feel
like
that's,
probably
a
level
of
granularity.
Most
most
users
don't
want,
but
I
think
certainly
like
the
a
an
operator
that
sits
above
the
Rook
operator.
Like
red
hat,
has
the
OCS
operator.
Poor
may
also
have
some
controlling
functionality.
D
B
It
that
would
be
the
idea,
basically
with
it
yeah
that,
like
you,
either
have
there's
some
something
on
top
that
you
can
just
like,
for
example,
with
the
safe
cluster
is
like
a
kind
of
a
Brownfield.
Then
I
see
this
kind
of
already.
As
like
the
Brownfield
situation,
it
should
say
these
are
the
disks
I
would
like
to
use,
or
even
these
are
the
PVCs
or
volume
climb,
teams
or
device
sets.
B
I
want
to
have
created
and
all
the
operator
would
then
just
do
is
like
create
a
correct,
cfosds,
I
guess,
underneath
to
some
degree
yeah.
A
B
So
if
we
just
go
to
what
there
were
two
more
slides
like
or
like
one
more
slide,
real,
quick
here
and
I
think
the
game
plan
is,
we
can
ignore
that
idea.
I
said
like
maybe
some
way
to
have
like
disks
being
Auto
replaced,
then
that
is
kind
of
with
the
hot
plot
detection,
but
then,
like
with
you,
know,
then,
with
the
cephos,
the
object
being
cool
and
all,
and
at
least
so,
I
kind
of
saw
it
with
the
surf
manager.
Having
like
this
orchestration
interface
that
that
we
raise.
B
B
D
I
really
kind
of
wish
that
the
surf
dashboard
existed
independently
of
Seth
in
a
way
right
that
it.
You
know,
you
could
say,
I
want
to
deploy
the
dashboard
and
that's
the
first
thing
that
you
can
do
versus
having
it
have
to
exist
within
the
manager.
I
think
that
would
make
yeah
it
would
make
a
lot
of
things.
D
Manager,
please
yes,
yeah,
because
then
we
wouldn't
have
I
I,
don't
I
I,
just
think
it
would
be
a
much
better
design
like
for
for
Seth
and
for
this
use
case.
Where
then
it
becomes
a
lot
easier
for
the
admin
to
like
decide
what
their
kind
of
security
risk
tolerance
is.
They
can
be
like
I'm
going
to
do
some
operations
spin
up
the
dashboard,
use
the
dashboard
and
then
like
bring
the
dashboard
application
back
down
and
because
the
dashboard
application
could,
in
that
scenario,
be
not
like
running
24
7..
D
B
A
B
B
A
B
That
and
then
like
adding
some,
not
new
objects
there,
at
least
of
like
a
read-only
perspective
first
and
then
look
into
how
we
can
maybe
create
them
and
maybe
then
look
into
like
even
splitting
up
like
Logics
in
the
operator
like
making
the
OSD
creation
independent
of
not
necessarily
depend
independent
of
a
cluster
creation,
but
that
a
cluster
creation
technically
is
just
about
getting
one
running,
get
the
manager
running
done
and
then
the
server
OC
would
be
like.
C
B
B
A
A
great
discussion,
lots
of
great
design
discussions
to
be
had
I
think
if
we
yeah
we'll
just
continue
those
offline
or
yeah
yeah
sounds
good
well.
That
brought
us
to
the
end
of
the
agenda
as
well.
Let's
see
where
it'd
be
good
to
go.
I
stopped,
sharing
my
screen
anyway,
but
I
think
that's
all
we
had
so
thanks.
Everyone
for
the
all
the
discussion
and
I
think
we'll
call
it
good
for
now.