►
From YouTube: 2019-07-01 :: Ceph Orchestration 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
A
A
C
A
A
B
B
B
A
C
C
A
B
B
B
A
C
A
You
can
also
just
write
a
tool
that
just
imports
ratos
great
the
Raiders
object,
passes
it
to
Mon
command
API,
and
then
you
have
them
on
command
API.
So
it's
you
can
use
any
Raiders
object.
Do
you
have
available
within
the
orchestrator
as
a
as
part
of
the
miniature
module
or
imported
by
yourself
or
any
other
thing
that
provides
you
a
Raiders
object.
B
B
A
A
B
B
C
In
theory,
us
the
orchestrator
import
that
head
of
function
and
all
it
and
then
both
that
then
talks
that
API,
which
is
fine,
I
guess,
but
then
it
goes.
So.
Your
fear
is
that
it
goes
back
to
another
implementation
of
Orchestrator
land
by
the
way,
the
name
of
service
instance
and
OSD
zap
by
that
you've
been
round
three
round
yep.
A
D
B
C
But
maybe
we
shouldn't
define
those
those
high-level
functions
like
remove,
replace
into
a
library,
but
maybe
we
should
just
offer
people
the
fine-grain
steps
like
was
the
outpost
empty
and
then
in
the
orchestrator.
You
can
just
compose
whatever
you
need,
because
every
single
implementation
is
slightly
different
and.
C
C
Sure,
but
but
it's
but
it's
fine,
but
they
still
need
the
stuff
out
and
self
urge
and
whatever
that's
what
they
still
need
and
when
they
do
it
in
the
orchestrator
they
can
use
set
out
and
they
can
use
to
surf
purge
and
then,
instead
of
doing
a
self
SFO
liam
albion,
zap
or
assistant
CDL,
stop
disabled
then
use
their
hook
into
their
goal.
Code,
staying
keel
container
or
whatever
my.
A
A
A
A
C
A
From
an
API
standpoint,
there
is
a
description
of
how
how
the
Sturridge
looks
like
in
in
the
Kuban.
It
is
api.
There
is
a
cuban
it
isaiah
api
objective
court,
def
cluster
and
which
is
the
llamo
file,
and
the
mo
file
contains
a
storage
element
and
in
that
storage
element
there
is
a
description
of
which
devices
are
used
by
rook.
A
A
It
has
a
bad
feeling
and
hedge
to
it,
because
if
you
want
to
remove
an
LSD,
then
you're
just
removing
the
device
from
that
big,
safe
cluster
customary
resource
object
in
the
Kuban
Etta's
api
and
then
rook
compares
the
previous
Jeff
Kloster
object
to
the
current
safe.
Just
erotic
just
covers
it.
Oh
that
OSD
is
missing
or
that
device
is
missing
from
the
object
and
then
it
starts
to
basically
remove
the
OSD
container.
A
It's
it's
very
generic.
It
does
not
contain
OS.
The
idea
is,
it
just
contains
list
of
drives.
Okay,.
A
E
C
C
Okay,
so
and
then,
and
then
so,
Kira
Nerys
computes
the
difference
between
those
between
the
old
file
and
the
new
file,
and
it's
he
is
okay.
There
is
one
file,
others
one
device
missing,
and
then
it
takes
it.
Computes
know
where
it
is
which
the
container,
what
is
a
container
that
was
attached
to
that
device
in
and
it
kills
the
OSD
yep.
C
A
Don't
know
if
they
actually
do
wait
until
the
data's
gone
I.
Don't
know
that's
okay,
but
that's
in
detail.
I
mean
let
the
detail
yeah,
but
in
general
removing
that
it
was.
The
specification
is
like
a
delete
request
due
to
the
Cuban
ettus
API.
That's
a
function
called
like
any
of
the
function.
Call
to
let
go
is
the
odd
or
I
was
the
purge,
but.
C
C
But
we're
trying
to
go
from
the
other
side
and
saying
Steffes
is
also
truth,
and
we
want
to
leave
that
OSD
and
suffer
figure
out
which
device
is
attached
to
that
Louis
D
and
then
triggering
the
action
on
the
host
side,
which
is
probably
the
better
idea,
and
that
could
also
be
done
by
rook,
because
they
because
staff
knows
that
a
certain
at
this
is
attached
to
an
OSD.
And
then
you
say
Allah
you
know
remove.
C
Then
you
do
the
self
things,
because
you
have
to
do
them
first
anyways
and
then
you
trigger
Orchestrator
five
things
that
they
rook
delete
this
line
in
that
file
and
then
we
trigger
whole
humanity's
build
a
delta
thing
and
remove
the
pod
contain
or
whatever
no
we're
going
from
the
south
side.
They're
going
from
the
Canaries
yeah.
A
A
Is
it
kubernetes
distill
this
source
of
truth,
but
we
have
dropped
groups
that
match
what
the
other
part
of
this
ecosystem
is
using,
even
though
that
might
mean
that
we
have
to
extend
the
drive
groups
with
tough,
especially
for
cuban
interpreted
problem.
Yeah,
it's
I,
don't
think
it's
a
problem.
It's
just
nothing
optional.
It
does
not
need
to
be
implemented.
C
I'm
really
into
player
because
handling
and
handling
files,
especially,
but
we
currently,
as
it
seems,
they're
only
standalone
or
steers,
and
but
when
you
have
dedicated
right
ahead
logs
and
databases
and
then
caching
devices
and
you
know
not
really
persistent
names.
And
then,
if
you
want
to
start
refacing
within
a
yellow
file.
Certain
chose
these,
but
not
the
pieces
of
the
right
head
law
from
other
those
teas.
Then
it
starts
to
get
really
messy
and
that's
why
we
actually
move
to
not
specifying.
C
C
E
A
A
Providing
a
good
example,
as
a
good
way
like
like
showing
that
deeps
is
using
drive,
groups
and
tipsy
is
successful
in
using
drive
groups.
That's
what.
A
But
it's.
A
B
They're
there
a
way
of
doing
SEF
volume
stuff
if
I
can
state
that
poorly
and
staff
volume
is
the
way
that's
F,
you
know,
creates
and
manages
OS
DS.
Then
like
it's,
it's
a
natural
thing
to
present
to
people
right.
You
know
if
you're,
if
you
want
to
deploy
surf,
if
you
wanted
to
play
our
SD,
is
you
have
to
use
self
oleum
underneath
and
drive?
Groups
are
a
thing
that
matches
as
well.
Then
you
know
just
use
this
specification.
Please
yeah.
A
A
Drive
groups
is
the
thing
that
can
also
be
integrated
into
the
dashboard.
So
if
we
have
that,
we
would
have
a
complete
chain
of
integrating
the
dashboard
into
the
orchestrator
into
passing
it
through
the
extra
actual
orchestrate
implementation
to
set
volume
on
the
house
itself,
which
would
be
really
awesome.
That.
C
C
Pretty
neatly
yeah
so
now
I
mean
it
definitely
makes
sense
to
put
it
somewhere
up
the
stack
if
it's
a
folium
or
if
it's
tough
itself
or
if
it's
you
know,
Orchestrator
I,
don't
really
care,
but
probably's
have
only
makes
sense,
because
that's
actually,
because
you
actually
could
generate
a
set
volume
commands
out
of
it.
The
only
thing
that
bugs
me
is
if
in
the
near
future
or
in
the
far
future,
somebody
decides
that
subvolume
is
not
something
for
them.
C
They
have
to
use
cell
volume
if
they
want
to
use
right
groups
or
they
have
to
re-implement
something.
So
that's
the
only
thing
that
you
know
with
future
in
mind.
That's
maybe
not
a
good
idea,
because
you
know
we
all
know
that
tools
get
replaced
with
you
look
at
self
disk
and
no
cell.
All
you
will
be
definitely
not
not
the
end-all
be-all.
C
B
If
my
my
guess,
based
on
just
you
know,
phase
of
the
moon
and
gut
feeling
and
that
sort
of
thing,
if
if
SEF
volume
goes
away,
it'll
be
at
least
two
or
three
years
from
now,
just
based
on
the
rate
of
change
of
you
know
other
other
pieces
we've
seen
so
it's
it's
not
a
I,
don't
think
we
have
any
immediate
doom.
There.
C
Yeah
but
I
mean,
if
you
just
implement
them
on
stick
above,
you
will
always
have
those
drive
groups
and
they
will
just
an
either
generate
a
set
of
several
elements
or
a
set
of
deaf
future
tool.
Whatever
commands,
you
know,
and
we
have
never
been
implemented,
no
I'm
fine
with
either
way
I
just
wanted
to
raise,
raise
the
point
from.