►
From YouTube: 2020-03-09 :: 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
C
C
D
D
C
B
B
I
think
that
well,
for
me,
the
right
one
is
to
what
a
new
parameter
to
all
the
commands
in
order
to
work,
to
hide
the
information
that
parameters
in
in
that
in
that
coma.
Okay,
because,
basically,
is
what
we
can
avoid.
We
should
avoid
to
put
this
information
in
the
doc
files
if
we
are
using
the
command
line,
for
example
a
phenol,
you
said
using
the
command
line
or,
for
example,
if
we
are
using
a
parameter,
what
a
command
object
from
another
Orchestrator
I
think
that
in
any
case,
we
should
hide
this
information.
B
B
C
B
B
External
options
in
the
in
the
back
I
think
that
the
world
of
Nestor
has
what
right,
with
with
lot
of
the
tiles
they're
possible
the
different
possibilities,
and
when,
for
me,
the
right
one
is
the
common,
adding
a
new
flap,
but
I,
don't
know
what
do
you
see.
E
That's
the
easiest.
The
procurement
is
the
easiest
one
to
implement,
but
it
obscures
some
information
because
there
are
lots
of
commands
that
are
pretty
generic
like
config
key
set.
That
could
be
anything
that
you're
setting
and
it's
it's
nice
to
know
what
you're
setting
I
think
it
might
be
better
to
do
the
per
argument
little
bit
trickier
to
implement,
though,
because
we
need
to.
E
We
need
to
make
sure
that
that
flag
in
that
in
that
string
that
describes
the
daylight
command,
doesn't
get
passed
with
a
CLI
because
the
allied
chokes
when
it
gets
a
command
full
type
of
flag,
whatever
that
doesn't
understand
that
needs
to
be
filtered
out
on
the
server
side,
but
just
means
it
would
be
a
little
bit
careful
there.
But
it's
not
that's
just
like
a
couple
lines
to
do
it
properly:
deletion,
okay,
I
think
argument,
maybe
the
pipe
for
the
way
to
do
it.
Okay,.
E
C
The
next
thing
on
the
agenda
is
so
the
Python
client
for
a
rook.
There
was
a
separate
procreated
a
few
weeks
ago
and
Sebastian
put
all
those
files
there
so
I
this.
Just
this
morning,
I
created
the
repo
to
marry
that
into
rook
org,
so
that
is
created
now
and
I.
Everything
that
was
had
been
committed
as
of
this
morning
is
now
also
committed,
they're
getting
up
made
it
easy
to
just
basically
marry
it
over.
So
I'll
follow
up
with
Sebastian
making
sure
we
use
that
going
forward
and
I
think
we
should
be
some
good.
B
Yes,
I
have
I,
am
working
with
the
with
your
cluster
state.
Okay,
when
trying
to
solve
this.
Okay
I
am
entering
in
trying
to
the
back
what
is
happening
with
the
service
and
why
it's
impossible
to
communicate
with
the
service
from
the
operator.
So
no,
let's
see
if
I
can
find
something.
I
have
what
or
net
information
from
your
call
it's
working
with.
Well,
basically
with
those
OCP
is
not
that
it's
not
the
same
more
or
less
okay,
and
let's
see
if
I
can
find
something
between
today
and
tomorrow.
B
E
B
Well,
I
think
that
it
should
work
with
all
day
on
the
port
network
providers.
Okay,
I
have
used
calico
in
in
your
cluster.
Okay,
probably
that
is
you
you
can
you
have
more
experience
using
different
port
networks,
but
what
I
have
seen
and
I
have
tested
all
of
them.
Is
that
what
basically
it
works?
Okay,
no.
C
B
B
B
C
B
B
You
use
hot
networking
in
the
problem
that
you
have
is
that
in
the
in
the
monitor
there
is
an
error
showing
that
you
can
bind
to
the
address
that
is,
is
going
to
be
using
the
service
in
kubernetes
and
if
you
don't
use
host
network,
then
it
seems
that
everything
works.
But
in
the
operator
you
have
an
error
message,
because
the
operator
is
not
able
to
connect
with
any
monitor
so.
B
E
B
Within
the
window,
with
a
with
the
last
test
that
I
am
doing,
they
have
force
the
placement
of
the
monitor
and
the
operator
in
a
stood
in
the.
What
is
the
the
control
plane
node
take
in
order
to
want
to
avoid
any
kind
of
additional
problems
in
network
and
and
the
problem
is
happening,
then,
in
the
same
host.
The
same
node
in
this
moment
in
the
worker
and
the
control
plane.
Node
is,
is
running
the
monitor
and
is
running
the
operator,
and
we
have
the
problem.
B
E
C
E
B
B
E
E
E
C
E
Fashions
gonna
be
late
and
so
I
think
I
should
just
wait
for
him
to
talk
about
it.
That's
all
that's
ad
and
stuff
I'm
trying
to
think.
If
there's
anything
else
in
work,
I
think
my
main
goal
is
to
get
my
cubeb
and
cluster
set
up
on
CentOS
8
on
my
development
box
and
then
figure
out
cube
jacker,
so
I
can
actually
develop
work
on
the
orchestrator
if
the
record
shooter
module
without
having
to
do
it.
E
So
hopefully
it
can
start
that
today,
I
seem
to
get
weird.
I
was
trying
I'm
like
googling,
for
how
to
install
you
baby,
a
DM
with
cryo
and
pod
man
instead
of
talker
and
it
like.
Aren't
there?
Aren't
you
just
as
far
as
I
can
tell
for
sense.
Awesome,
maybe
I'll
just
use
docker
under
the
dead
body.
Since
that's
what
everybody
else
does
anyway,
testing
yeah
the
future
hasn't
arrived
yet.
E
Okay,
both
its
back
I,
would
say
if
he'd
be
here
on
the
half-hour
or
he
and
they
he
just
said
he
would
view
later
heard
late,
yeah,
it's
like
conflicting
meeting
so
but
I
think
well
us
so
I
guess.
The
main
question
here
is
so
in
rook.
If
you
delete
the
CRD,
then
all
the
pods,
obviously
just
get
cleaned
up.
Everything
goes
away.
That's
right!
I'm
at
the
and.
E
It's
just
a
little
bit
different
acetate,
a
and
because
there's
we
have
like
a
the
equivalent
of
the
Sierra
Diaz,
a
spec
and
it's
per
demon
type.
Others
respect
for
monitors
that
basically
describes
where
your
monitors
go.
This
one
for
each
MDS
cluster
or
each
file
system,
that
maps
directly
to
a
CRT.
E
E
Yeah
so
I
mean
this
sort
of
CLI
commands
are
like
Steph
Forge
apply
under
the
need.
You
like
count,
5
or
just
5,
or
whatever,
like
how
many
you
watch
and
support
supply
I
want
three
of
those
or
I
want
Amman
hosts.
You
know
most
one
has
to
something
like
that
or
you
know.
Seth
supply
RVD
near
I
want
to
be
on
those
gold.
E
E
That'll
solve
this,
one
will
go
directly
apps
to
us
already,
the
stuff
filesystem
Zod,
but
these
other
ones
are
gonna,
be
modifying
bits
and
pieces
inside
the
clusters
here
to
your
ancestors
here,
and
you
can
do
all
these
things.
You
can
specify
account
and
just
specific
host
names
and
also
label
that
right.
Those
are
basically
the
three
ways
you
can
play:
sing.
C
E
E
Yeah
I
mean
I,
guess
Sebastian's
fly
right,
I
think
in
basically
we
just
current
anand
up
until
like
last
week.
You
could
just
create
demons
and
destroy
demons
at
whatever
like
manually,
and
they
would
just
be
there,
but
now
set
fate.
A.M.
let's
you
like
specify
this
placement
description,
the
spec
for
the
demon
type
and
the
question
is:
what
happens
if
the
spec
doesn't
exist,
but
the
demon
does
but
I
think
that's
really
just
like
an
upgrade
case
for
people
running
earlier
sees.
E
E
E
Yeah
aphelion
isn't
controlling
that
part.
It's
just
there
actually
is
a
pull
request
that
needs
to
be
refreshed
but
would
basically
automatically
adjust,
though
itself
there's
a
setting
of
how
many
active
you
want.
That
would
basically
go
the
other
way.
I'll
tell
the
orchestrator
to
go
spin
up
that
many
demons,
if,
if
there's
always
a
standby
or
in
standby,
use,
or
something
like
that,
I
think
we
need
to
decide
which
direction
go
there.
There's.
C
E
Of
a
philosophical
problem
we
keep
coming
back
to,
it
depends
on
whether
you're
driving
stuff,
the
Akoo
burn,
Eddie's
or
the
other
way
around.
If
you're,
using
like
the
dashboard
or
the
stuff
CLI
to
manage
yourself,
cluster
and
work
is
just
there
to
keep
demons
running
then
it
makes
sense
to
drive
at
that
direction.
But
if
you
are
kubernetes
native
user,
you
just
want
to
say
everything
here:
D
and
just
have
it
work.
E
E
C
A
C
Operator
we
just
implement
making
them
exclusive.
If
you
do
this
yeah.
C
B
I,
wasn't
me
it's
just
to
comment
at
that
with
that
manager
in
Brooke?
Well,
what
we
are
twin
is
to
avoid
to
have
more
than
one
manager,
okay,
yeah,
because
they
program
of
the
of
the
ports
I,
think
so,
and
one
that
is
possible
to
to
use
the
placement
properties
of
the
CRT
to
put
the
manager
in
the
hospital.
The
node
that
you
want
when
one
question
is
that
I'm
not
sure
because
I'm
not
too
much
update
with
with
FFA
EDM
in
EDM,
there
is
any
kind
of
CR
define
or
something
like
that.
E
E
E
B
B
C
E
F
F
E
Yeah
well,
I
mean
I,
think,
there's
even
even
independent
of
whether
you
actually
want
to
end
up
that
way.
There's
sort
of
this
question
of
whether
it
should
be
asynchronous
like
should
you
just
be
able
to
delete
the
demon
and
have
it
happen
to
me
return
immediately
and
then,
in
the
background,
go
and
clean
them
up
the
same
way
we
create
them
like
that
seems
nicer
perspective.
E
E
Which
is
I,
guess
proposal,
but
then
I
guess
in
order
for
that
to
work,
then
you
basically
need
to.
If
we
find
a
stray
demon
that
doesn't
have
a
spec,
we
should
just
delete
it
always
well.
E
F
E
I
mean
there
are
only
two
clusters
that
I
know
about
that
are
running
it
they're,
both
mine,
but
like
I
parented
five
clusters,
so
I
can
just
create
the
specs
before
upgrading.
Also
I
think
that'd
be
fine.
It's
a
little
bit.
Weird
I
think
we
probably
add
some
safety
checks
because,
like
if
the
config
fee
record
disappears
four
monitors
say
it's
just
going
to
delete
all
your
monitors.
I
think.
E
F
E
F
F
E
E
Correct
it's
just
like
you
have
to
it's
two
steps
enough
to
wait.
Just
sort
of
yeah
I
could
have
that
behavior
just
for
well,
so
the
ones
that
really
are
scary.
It's
really
just
the
monitor
one,
that's
the
scary
one
and
we
should
just
not
allow
you
to
delete
it
at
all.
Right,
not
allow
you
just
kill
it
down
below
one
I
think
that.
E
E
E
In
in
general,
there's
nothing
there
like
the
OS
DS,
it's
just
metadata
about
the
UST.
We
don't
actually
wipe
the
device
or
anything
that
you
could
bring
a
back
and,
in
the
monitor
case,
monitor
case
is
the
only
one
really
where
you're
deleting
let's
monitor
policies.
Matheus
has
some
like
historical
stats.
You
Julia
Prometheus,
then
you're
deleting
actual
data,
but
it's
not
like
for
important,
probably
and
then
the
monitors
the
only
one,
but
we
wouldn't
allow
you
to
delete.
E
E
E
F
E
E
E
Missing
piece
is
the
thing
that
automatically
refreshes
configs:
it
needs
to
learn
what
I
think
I
was
going
to
make
it,
so
it
remembers
what
monitor
epoch
it
last
generated
the
set
count
for
so,
if
you
add
or
remove
the
monitor,
then
it'll
go
regenerate
all
the
step,
comps,
all
the
demons.
They
know.