►
From YouTube: 2019-12-09 :: 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).
A
A
A
B
So
Eve
only
noteworthy
change
is
that
there
are,
you
can
have
multiple
drive
groups
and
therefore
you
should
name
them
also
for
understanding
and
describing
what
the
drive
room
actually
doesn't.
Naming
is
probably
a
good
idea,
though
yeah
I
added
that
for
people
to
have
a
look
at
yourself,
because
the
initial
implementation
from
the
dashboard
people
does
not
respect
the
fact
that
you
can,
where
you
have
to
give
your
draft
of
a
name
and
that
draft
groups
can
also
like
appear
in
multiple
and
not
only
in
singles.
B
A
B
A
C
A
I
got
no
go
ahead.
That
directly
expects
the
orchestrator
because
they
did
affect
how
we
are
creating
almost
is
from
from
the
dashboard
that
get
effects,
how
we're
creating
OS
these
from
the
CLI,
and
it
affects
how
we
want
to
deal
with
it
from
within
the
data
structures.
We'll
see
ya,
orchestrated
this
yeah
causing
more
trouble
yep
and
any
idea
how
you
want
to
modify
the
UI
the
despot
UI.
My.
C
D
C
C
D
A
C
Separated
those
two
things:
cuz
they're,
two
different
problems,
at
least
the
way
every
the
situation
of
the
problem
Sebastian's
up
against
is
that
LVM
and
and
EVS
don't
cooperate,
and
so
why
not
using
LVM
at
all,
then
that's
fine,
but
it
we
can
see
doesn't
need
a
wrapper
if
we
don't
use
lbm
okay,
but
I
think
we
do
need
a
wrapper
for
tough
demon.
Yes,
whether
when
the
unit
starts
out
picking
and
set
up
the
DM
crypt
volume
and
then
start
the
OST
and
then
when
it
stops,
the
container
it'll
tear
it
all
down
again.
C
A
D
D
D
D
C
D
C
B
C
C
C
D
A
C
C
C
C
Okay,
yeah,
it's
the
beginning,
Oh.
What
else
have
a
couple
things?
Oh
I
was
starting
to
do
the
reconfig
and
then
I
realized
I
want
to
make
it
behave.
Well,
if
I
made
the
changes,
I
thought
I
was
going
to
make,
it
would
collide
with
Paul's
monitoring
thing.
So
I
think
we
should
just
let
Paul
murchison
first
get
that
working
and
then
I
can
we
can
clean
up,
in
fact
or
whatever,
to
make
that
reconfig
work
and
then.
C
C
Sure
how
far
away
that
is,
and
then
the
last
thing
was
my
removal,
asti's
I
had
that
like
or
line
patch.
That
was
basically
just
cutting
based
in
your
stupid
suggestions
fashion,
maybe
that
you
should
just
get
adopted
by
whoever's
doing
there's
another
one
that
fixes
it
to
use,
drive
groups
and
I.
Think
this
another
there's
another
Orchestrator
flow
request:
mixing
honesty's
a.
C
Or
if
you
want
to
take
it
over
since
you
wrote
the
code
actually
I,
just
don't
know
why
that
yeah
I
don't
really
understand
that
I
should
suffer
well.
Oh,
the
last
thing
was
and
I
was
trying
to
work
on
the
upgrade
stuff.
But
again
the
completion
stuff
just
made
me
confused
because
I'm
in
the
serv
thread,
and
so
I
do
some
stuff
that
triggers
some
asynchronous
work
but
I
don't
return.
C
A
C
Think
so
I
think
we
should
have
a
big
fat
work
to
you,
so
we
can
deploy
lots
of
those
T's
that'll
be
my
first.
Then
we
we
shouldn't
look
yeah.
This
is
this.
This
is
a
serve
yeah.
I
didn't
really
I.
Didn't
care,
I,
don't
care
if
it
blog
actually
originally
thought
it
wasn't
going
to
block
at
all
so
I,
just
like
bar
the
events
and
then
I,
looped
and
West,
but
nothing
would
be
good
completion
like
that.
Then
then
then
run
and
it's
in
the
upgrade
said
come
on
this.
C
D
Speaking
about
upgrades,
I
I
didn't
upgrade
today
from
another,
is
to
octopus
in
work
and
I've
noticed
a
different
message
when
it
comes
to
blue
storage,
these
with
all
map
usage,
stats,
I
thought
it
was
fixed.
Maybe
the
conversion
is
not
happening
yet.
Can
you
look
at
the
comment?
I
passed
its
first
this
stage
in
the
chat
I.
D
D
C
C
B
D
C
D
C
But
it's
kind
of
like
a
it's
a,
maybe
a
somewhat
disruptive
thing,
I
mean
I
could
do
it
as
part
of
the
upgrade
I
guess
that
might
make
sense.
I
mean
yes,
that
might
actually
make
sense.
Maybe
when
it
appraised
you
chose
D.
If
it
knows
it's
doing,
a
major
upgrade,
it
could
just
actually
run
FS
check
on
it.
A
C
D
C
D
It's
just
too
bad
that
we
cannot
like
have
this
hook
when
the
OSD
exits
really
because,
like
I
think
it's
only
an
issue
for
openshift
at
this
point,
maybe
because
the
well
not
not
now,
because
the
powers
are
not
running
or
running
privileged.
So
it's
it's
not
a
problem
and
upstream
I
think
we
can
solve
this
like.
If
you
like
I
know
you
drop
religious
but
somehow
I
don't
know
like.
D
D
B
C
Think
that
the
underlying
issue
is
that
it
just
it
isn't
safe
because
stay.
The
very,
very
first
thing
that
the
process
does
when
it
starts
up,
is
installed
a
signal
handler,
but
still
not
an
atomic
operation.
So
it's
possible
that
the
OSD
slow
to
start
the
signal
human
who
doesn't
get
installed
and
the
cleanup
operation
doesn't
run
by
the
time.
D
C
D
C
Don't
really
what
we
do
need
is
I
think
the
the
system
do
unit
not
system
unit,
but
the
container
when
it
shuts
down,
should
clean
up
and
it
needs
to.
If
we
needs
to
do
the
initialization,
which
it
doesn't
do
right
now
and
it
needs
to
do
the
cleanup
which
that
it
should
do
the
cleanup
I
think
just
to
be
cleaner
from
the
inside.
You
mean
okay,
yeah
I
mean
it
doesn't
necessarily
have
to
happen
inside
the
container
or
outside
the
container
out
know
if
it
really
matters,
but
for
I.
C
C
To
Berlin:
well,
it's
right
now
it
doesn't
work
it
all
with
the
in
crypt,
unless
you
don't
reboot.
D
C
C
Like
I
was
replacing,
which
the
drives
in
my
cluster
and
I
like
yanked,
to
drive
that
had
been
dead
for
a
million
years
and
I
put
a
new
one
in
it
that
long
deploy
was
failing,
and
it's
because
there
was
just
old
stuff
that
was
like
still
mounted
off
that
dead
drive
from
like
ages
ago.
I!
Guess,
if
you
stop
the
demon,
it
should
should
tear
everything,
tear
everything
down
anything
that
it
sets
that
stuff
on
should
clean
up,
so
that
you
don't
have
to
go
manually
clean
things
up
like
undoing
the
crypto
Jim
crippling.
D
C
C
D
D
C
D
C
A
D
It's
like,
can
you
just
think
think
about
this
a
bit
more,
because
the
signal
handling
thing
is
like
not
so
difficult
to
do,
but
it's
just
like
over-engineering
when
you
get
all
the
logic
from
from
system
D
like,
for
example,
in
the
root
case,
I
need
to
have
something
that
handles
the
signal,
because
I
don't
have
any
system
the
unit
and
I
can't
do
any
post
stop
actions.
But
since
you
have
system
leader,
then
you
can
already
do
it.
It's
probably
better
to
do
it
with
system
D.