►
From YouTube: Ceph Orchestrator Meeting 2021-08-17
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
A
I
I
think
that
one
is
pretty
much
a
set
right.
If
we
are,
if
we
are
properly
fixing
it,
then
if
you
are
probably
doing
the
backboard,
then
you
can.
You
can
put
your
backpot
on
top
okay.
A
A
Okay,
I
would
say:
let's
start,
we
have
already
two
minutes
past.
I
click
on
highlight:
okay,
so
hi
everyone
and
welcome
to
today's
orchestrator
meeting.
We
have
two
topics
for
today.
We
have
first
topic
is
that
we
have
another
potman
back.
That's.
A
You
lost
me
okay,
so
we
have
another
partnering
issue
and
that
that's
making
about
two-thirds
of
all
jobs
fail
of
all
safe
adm
jobs
fail,
and
I'm
a
bit
wondering
why
this
wasn't
discovered
in
the
portman
ci.
A
What
what
can
we
do
about
it?
I
mean
I'm
a
bit
lost
here.
A
A
B
A
The
two
thirds
of
us
have
adm
test
cases
are
failing
because
of
that
pot
management
and
some
in
some
random
locations
within
the
within
the
test.
Right
because
there
is
a
race
in
potman
that
prevents
containers
from
being
started
properly
or
reliably,
and
in
the
end
it
means
that
about
two-thirds
of
our
above
of
all
tautology,
jobs
are
failing.
A
A
I
don't
think
we
are
going
to
get
a
great
solution
here.
Okay,
let
that
switch
over
to
the
into
the
second
topic.
We
have
a
topic
that
that
came
from
downstream.
A
A
And
everything
is
already
supported
by
the
civilian
manager
module,
so
we
can
now.
We
are
installing
packages,
but
we
are
updating
the
self.confile
and
we
are
distributing
keyrings.
That's
that's
already
implemented,
but
we
have
another
requirement
that
we
are
not
going
to
be
able
to
do
constant
root
that
we're
not
going
to
be
able
to
have
constant
root
access
to
all
client
house,
and
this
is
making
it
kind
of
complicated
to
use
fading
because
fadi
mandates
that
that's,
if
adm,
has
constant
root
access
to
all
the
client
hosts,
which
we
cannot
have.
A
A
It
would
be
really
great
to
be
able
to
on
demand
deploy
newsef.com
files
to
the
client
host
because,
as
soon
as
cfdm
deploys
a
new
monitor
or
removes
a
monitor
or
evacuate
the
host
where
a
monitor
was
located
or
anything
that
changes
the
monmap,
we
are
having
the
it's
going
to
be
necessary
to
deploy
a
new
theft.conf
across
all
the
client
hosts,
and
it
would
be
really
great
to
have
it
automated.
D
A
A
D
No,
what
I'm
saying
is
that
in
the
client,
only
with
a
url
pointing
an
agent
in
any
of
the
theft
host
to
get
this
information
to
have
an
expected
endpoint
in
the
agent
and
using
clients
just
to
connect
to
the
fdm
agent
and
to
retrieve
this
information
and
to
put
this
information
in
the
place
where
the
the
the
user
decide.
Okay.
A
I
mean
yeah
right
that
that's
amazing.
Basically,
what
part
request
is
doing
right
if
I'm
not
mistaken,
given
dimitri
what
what
does
paul?
What
what's?
What
is
paul
doing
in
this
pull
request,
he's
deploying
the
the
client
keyring
and
deploying
the
theft.com
file
right
yeah.
Basically,
it
just
deployed
the
the
sub.conf
and
the
plan
carrying.
A
If
we
want
to
make
use
of
the
agent
to
update
the
self.conf
files,
which
would
be
great
anyway
right,
at
least
eventually,
then
we
need
to
have
a
way
to
deploy
the
agent
and
that's
not
going
to
be
happen
using
safe
alien,
because
at
that
point
we
would
need
a
constant
root
access.
So
if
we
wanted
to
use
a
staff
idm
agent
to
d
to
update
the
sev.conf
file,
then
we
would
need
to
magically
install
this
f8m
agent
on
the
house,
which
is
also
not
going
to
be
happen
without.
D
D
And
we
are
talking
about
clients
where,
but
probably
we
are
not
going
to
have
even
possibility
to
install
things.
Okay,
so
is
the
user,
the
one
that
is
going
to
decide
what
packages
he
wants
to
install
and
what
the
software
to
use.
So
I
think
that
the
solution
is
not
thinking
about
how
to
install,
for
example,
the
theft
you
imagine
in
these
clients.
D
A
A
So
there
is.
I
don't
think
that
there
is
a
need
to
to
have
an
additional.
D
A
A
I
think
paul's
pulse
per
request
is
already
okay,
because
it
provides
the
ability
to
deploy
this
effort
confident
the
key
rings.
A
If,
if
we
would
have
the
requirement
that
we
can't
easily
get
the
self.conf
file
generated
by
ceph,
then
at
that
point
we
would
need
to
do
that,
but
at
this
at
least
paul's
current
pull
request
makes
it
makes
use
of
it.
So
that
should
actually
work
properly.
E
A
Yeah
then,
my
question,
then,
is:
if
the
manager
of
the
cfid
manager
module
can
is
supposed
to
deploy
the
self-confident,
keyring.
E
The
like,
like
the
old
arcgs,
three
or
whatever
or
four
when
they
had
staff
ansible
when
they
used
that
to
deploy
clients
when
steph
you
know,
stephanie
uses
some
ssh
key
to
get
into
those
clients
to
you
know,
bring
the
files
over.
E
Does
that
get
cleaned
up
after
or
wouldn't
whatever
the
machine
that
ran
steph
ansible
still
have
access
root
access
to
those
machines.
After
the
effect
I
mean
that
ansible
is
not
running
constantly
like
the
manager
module
is,
but
they
still
would
have
access
to
those
machines.
I
guess
unless
the
ssh
keys
were
erased
after.
A
We
we've
seen
the
very
same
problem
from
the
openstack
folks
that
having
the
cluster
itself
has
admin
access
to
to
other
hosts
really
changes
the
security
considerations,
because
previously
only
the
admin
host
had
admin,
access
to
or
root
access
to
all
the
hosts
and
now,
basically,
all
hosts
of
the
theft
cluster
have
root
access
to
all
the
other.
All
the
cluster
hosts.
E
Well,
I
don't
know
I'm
guessing,
that's
the
concern
that
then
there's
really
no
way
to
have
cepheid
am
automatically
do
this.
That's
always
going
to
be
the
case
if
stephanie
m
is
going
to
be
in
charge
of
distributing
the
keys
to
the
video,
every
host
of
the
cluster
has
access
to
the
f80
mssh
key,
which
it's
that
I'm
going
to
use
to
distribute
those
keys.
So.
E
A
That
would
be
kind
of
a
way
to
completely
eliminate
the
ssh
requirement
to
self-idea
right.
If
we
are
installing,
if
you're
properly
installing
the
fadm
agent
before
even
installing
the
before,
without
using
sefadm
to
install
the
agent,
then
we
would
completely
eliminate
the
requirement
to
have
ssh
access
completely.
A
But
only
if
the
agent
is
capable
of
doing
everything
right
so
right
now
it
only
provides
information
to
the
manager
module,
but
then
we
would
need
to
be
able
to
do
much
more
than
that.
We
would
need
to
be
able
to
deploy
demons
and
stuff
like
that.
A
But
using
the
self-adm
engine
installed
on
the
host
seems
to
be
something
that's
also
kind
of
on
top
of
what
paul
does
right,
but
completely,
but
a
bit
it.
The
self-adm
agent
would
then
be
able
to
update
this
effort,
confirm.
A
Okay
yeah:
let's
conclude
this
topic
for
today.
Yes,
there
is
no
perfect
answer.
D
B
D
Yeah
yeah
yeah,
but
I
think
that
is
another
another
cause.
Okay,
I
I
don't
know
because
it's
it's
new
okay,
but
the
thing
is
that
what
the
problem
with
the
with
the
list
of
devices
is
fixed
and
it's
because
the
the
new
code
that
joseph
has
implemented
in
order
to
use
asteroids
class
okay
in
order
to
detect
what
are
the
the.
B
D
Yeah
yeah
many
tested
in
that
okay,
so,
let's
see
there
is
an
open
document.
I
think
that
we
have
a
list
of
requirements
that
which
is
great
okay,
but
well,
let's
see
I'm
I'm
probably
I
don't
know
who
is
going
to
be
that
the
best
people
to
to
start
with
the
project?
Okay,
maybe
the
the
same
people
that
have
implemented
the
local
storage
operator,
or
maybe
we
can
start
the
trade
by
ourself
and
but
having
some
kind
of
support
from
from
these
people.
B
Right,
the
other-
and
I
guess
one
update-
we
did
have
a
meeting
on
thursday
with
some
people
on
the
ocs
team
sahina
and
the
knot
and
a
few
a
few
others
just
to
talk
about
what
the
requirements
are
and
there
the
requirements
felt
like
it's.
The
same
thing
we've
already
been
expecting.
We've
already
been
talking
about
these
they're
already
in
the
dock
for
the
most
part
and
yeah.
So
the
next
question,
I
think,
is
how
we
start
off
the
project
and
how
do
we
work
with
the
lso
team
to
make
sure
we're?
C
Yeah,
so
to
begin
with,
I
have
been
working
for
past
two
months
on
implementing
a
new
kind
of
a
test
case
for
integration
testing,
using
the
behave
test
where
we
are
able
to
provide
the
specific
kind
of
commands
and
provide
the
specific
kind
of
outputs.
We
require
that
we
can
specify
that
how
the
system
should
work
or
the
command
should
work
and
we
have
been.
I
have
been
using
the
behave
framework
to
implement
these
test
cases,
so
we
have.
C
I
have
implemented
three
kinds
of
test
cases
to
initially
start
with
as
a
proof
of
concept
that
we
can
test
the
safe
shell
commands
and
osd,
adding
osds
and,
as
I'm
currently
working
on,
also
to
include
rdb
comma
functions
as
well,
and
we
would
like
to
have
the
review
from
the
community
on
this.
I
think
I'll
share
the
port
request,
so.
D
Okay,
another
important
thing
that
I
think
has
in
this
work.
It
was
worth.
The
comment
is
that
for
the
next
week
we
can
use
part
of
the
time
of
this
meeting.
E
D
Order
to
present
to
the
community
a
demo
of
this,
this
kind
of
integration
test.
Okay,
we
have
the
the
basic
functionality,
but
I
think
that
for
the
target
for
the
google
summer
of
code
project,
it's
completely
active,
I'm
very
happy,
okay,
because
well,
it's
very
simple.
The
code,
if
you
take
a
look
is,
is,
is
ridiculous,
simple
and
I
think
that
we
have
something
that
is
enough
power
powerful
in
order
to
to
test
almost
every
kind
of
feature
in
the
in
the
theft.
D
Cluster
and
very
easy
to
use,
obviously
can
be
improved,
obviously
can
be
extended.
Okay,
but
is
I
think
that
is
a
good
proof
of
concept
about
how
to
execute
intelligent
tests
in
a
very
easy
way
and
to
have
the
possibility
to
secure
this
test
in
your
own
laptop
or
to
link
the
test
automatically
with
the
yankee
system.
A
That
would
be
great.
I
think
the
behavior
tests
are
especially
interesting
for
self
volume
as
well.
A
A
Perfect
see
you
next
week.