►
From YouTube: Ceph Orchestrator Meeting 2020-09-14
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
Okay,
let's
start,
I
think
we
are
not
totally
in
full
force,
but
I
think
we're
good
okay,
hi
everyone
and
welcome
to
today's
orchestrator
meeting.
As
always,
let's
have
a
look
at
the
again
for
today,
so
we
have
a
new
pull
request
by
paul,
because
now
that's
one
three,
oh
one,
three,
zero!
That's
going
to
be
huge
because
it
introduces
some
kind
of
persistence
of
adm
daemon
running
on
all
holes.
B
One
thing
is:
is
I
think
that
the
first
of
all,
I
need
to
to
clarify
something
here:
okay,
because
well,
this
could
be
a
fifth
step
in
order
to
have
this
diamond
mode
with
a
lot
of
functionality.
B
But
now
the
scope
is
just
together
to
gather
the
information
of
things
that
we
have
in
the
in
the
host
about
hardware
and
diamonds.
Okay,
so
this
is.
This
is
the
the
current
scope?
Okay,
maybe
it
could
be
the
seat
of
something
bigger,
okay,
as
the
fadm
diamond,
with
all
the
functionality
that
that
we
need
okay,
but
this
this
is
not
our
target
in
this
moment.
B
In
this
moment,
the
target
is
just
to
solve
the
problem
about
the
time
that
we
need
to
retrieve
information
about
what
we
have
in
the
host
okay-
and
this
is
something
that
is
going
to
be
good
for
the
organizator
and
also
for
for
the
dashboard.
So
this
is
the
the
thing
that
I
I
want
to
transmit.
Okay,
because
this
morning
I
have
talked
with
paul
and
well.
B
We
had
another
conversation
on
friday
in
the
organizer
meeting,
and
do
you
remember
that
I
think
that
joshua
one
expects
express
some.
Some
worry
worries
about
the
the
design,
the
architecture
of
this
diamond,
okay
and
well.
This
is
not
a
our
current
use
case.
What
we
want
to
have
is
just
to
solve
the
problem
around
the
scalability,
because
we
are
taking
a
lot
of
time
in
order
to
to
get
all
the
information
that
we
need
from
the
host
okay.
B
B
A
B
A
B
A
A
Okay,
but
I
I
mean
we
now:
we,
we
have
the
opportunity
to
have
a
small
discussion
about
the
performance
and
the
scalability
of
sephardium
itself,
not
just
gathering
information
about
hosts
and
demons
and
in
the
inventory,
but
having
a
discussion
about
a
bigger
picture
right.
The
this
pure
request
from
paul
really
provides
a
way
forward
if
we
on
on
the
way
to
on
on
the
way
forward,
want
to
improve
or
have
an
idea
how
to
improve
the
performance
even
further.
C
A
So
like,
for
example,
we
have
a
self
loop
yeah
that
that
does
four
steps
and
then
it
does
nothing
and
then
it
it
does
four
steps
and
the
first
step
is
it
refreshes,
the
host
and
the
demons
to
basically
display
the
information
to
the
dashboard.
Then
the
the
second
step
is
to.
A
Restart
demands
or
reconfigure
demands
if
they,
if
they
need
to
get
reconfigured
and
refresh
host
and
demon.
That's
the
one
that
paul
is
really
looking
into
right.
Now,
that's
blue,
because
it
has
to
go
to
it
host
and
that's
going
to
execute
some
expensive
steps
like
calling
their
volume
inventory
or
getting
only
host
information
or
demon
information,
and
I
mean
that's
important
for
for
the
dashboard,
but.
A
There
are
also
things
that
we
might
need
to
consider
like
the
the
general
thing
that
connections
to
all
house
are
expensive
if
we
are
calling
each
host
from
from
the
manager
daemon,
because
the
manager
is
already
overloaded,
especially
from
from
the
promises,
module
and
the
dashboard
module,
then
the
second
problem
we
have
is
that
config
key
the
config
key.
A
I
don't
know
how
how
much
it
scales.
Actually,
it
might
be
okay
for
us,
or
it
might
become
a
problem,
so
we
always
have
to
be
careful
about
putting
too
much
stuff
into
configure
updating
country
keto
off.
A
The
next
problem
is
that
ssh
connections
are
slow,
making
ssh
connections
to
all
hosts
is
going
to
be
slow
and
expensive,
especially
if
we
want
to
keep
thing,
keep
ssh
connections
open,
persistently
and
stuff
like
that,
and
the
the
last
big
problem
that
I
see
is
imod.
If
we
end
up
with
servers
that
are
offline
or
we
have
network
issues
and
we
are
running
into
timeouts
and
everything
is
basically
stalled.
A
B
Well,
I
I
would
like
to
to
have
really
an
https
interface,
a
full
appeal,
rest
interface,
in
order
to
operate
with
the
fidm
diamond
okay.
But
this
is
the
then
design,
okay,
no,
it's
not
what
we
have
in
this
moment.
I
think
that
we
should
be
centered
basically
in
in
trying
to
solve
the
the
the
urgent
problems
that
we
have
in
this
moment.
Okay,.
C
B
A
Yeah,
we
are,
we
are
still
creating
lots
of
http
and
this
is
edge
connections
to
all
hosts
I
mean
we
are.
We
are
going
to
improve
the
display,
the
the
refresh
rate
of
information,
but
we
are
still
creating
lots
of
ssh
connections.
B
A
A
big
potential
for
having
future
scalability
improvement.
If
we,
if
the
time
comes.
B
B
We
have
put
a
new
document,
a
design
document
about
high-level
workflows
about
the
device,
ost
management
and
storage
device
selection.
Okay,
I
think
that
that
is
a
high
level
view
of
the
problem,
the
things
that
we
need
to
solve:
okay
and
even
a
proposal
of
interface.
Okay,
that
can
be
changed,
okay,
but
I
think
that
it
illustrates
perfectly
the
kind
of
things
that
I
think
we
we
need
to
to
have
in
the
in
the
user
interface.
B
I
think
that
we
should
do
exactly
the
same
for
the
vdm.
We
now
well
perfectly.
B
B
So
maybe
we
should
start
to
do
the
thing
in
the
big
picture:
okay,
not
not
for
pacific
for
the
next
version,
probably
okay,
about
how
to
provide
a
well-defined,
the
design?
Okay,
because
I
think
that
what
is
something
that
I
I
am
saying
in
this
moment:
okay
in
the
project
that
what
we
have
been
working
here
for
one
more
one
year
more
one
than
one
year
in
this
month,
more
or
less-
and
there
are
a
lot
of
effort
here-
okay,
but
we
do
not
achieve
to
to
have
something
really
robust.
B
B
I
think
that
is
a
big,
a
very
big
effort.
Okay,
it's
incredible!
It's
just
a
component
of
dev
and
we
have
800
pull
requests.
Okay
and
the
other
thing
that
we
have
is
that
basically,
we
are
with
exactly
with
the
same
number
of
packs
in
the
last
six
months.
Okay,
we
do
not
it's
not
possible
to
to
reduce
the
number
of
packs
or
to
improve
heavily
the
the
robustness
of
the
of
the
of
the
fedm.
B
So
I
think
that
this
illustrates
we.
What
what
is
our
current
situation?
Okay,
but,
okay!
We
have
something
something
that
I
think
that
is
very
good
to
to
show
what
what
is
the
way
to
follow?
Okay,
but
we
need
to
to
think
twice
about
how
we
are
doing
the
things
in
order
to
improve
that.
We
have
started
to
do
that.
B
Okay,
because,
for
example,
with
the
with
the
atom
that
we
are
having
in
this
moment
about
splitting
the
code
in
different
parts
in
order
to
restrict
to
the
code
and
to
make
more
easy
to
to
deal
with
all
the
gold
okay.
So
we
are
in
the
right
way,
but
I
think
that
what
this
is
something
that
we
need
to
do
and
I
I
would
like
to
start
in
the
same
way
that
I
have
put,
for
example,
the
information
about
the
the
workflow
in
the
in
the
about
the
osds
and
stress
device
management.
B
I
think
that
we
should
start
with
a
clear
definition
of
what
we
want:
okay
and
then
we
should
analyze
what
different
possibilities
we
have.
Okay.
In
order
to
implement
that
high
level
then
structure
a
right
architecture
with
blocks
with
how
each
block
is
interacting
with
the
other
blocks.
What
are
the
inputs?
What
are
the
outputs?
B
What
is
going
to
be
our
interface
and
after
that
start
to
implement,
because
in
this
moment
is
my
is
my
view
that
I
think
that
what
we
have
in
a
in
a
situation
where
we
have
something
okay,
this
is
going
to
work
okay,
but
it's
going
to
be
very
difficult
to
maintain.
We
are
going
to
have
problems
about
that's
difficult,
very
difficult
to
to
solve.
Okay,
because
basically,
we
have
started
with
something.
B
That
is
that
it
was
just
a
well
a
proof
of
content,
okay,
that
it
was
real
okay,
because
I
think
that
it
shows
exactly
what
was
the
the
sage
there
about
how
to
how
to
manage
this
problem
about
our
status.
C
So
I
just
recently
saw
your
your
design
workflow.
I
didn't
read
it
until
the
end.
Actually,
I
just
came
like
10
to
20
through
before
the
meeting,
but
I
think
that's
a
that's
a
very
good
start
and
we
should
definitely
start
the
discussion
when
we
want
to
change
something
very
early
and
should
also
come
to
a
to
a
consensus
before
we
do
bigger
changes.
C
B
Yes,
this
is
exactly
what
I'm
saying:
okay,
that
we
have
something
now
okay,
but
if
we
want
to
have
something
robust
and
we
have-
and
we
want
to
avoid
to
have
problems
in
the
future,
probably
we
should
do
this
kind
of
thing,
okay,
but
I
think
that
we
are.
We
are
early.
A
It's
a
bit
too
too
much
text
in
order
to
get
the
and
get
the
high
level
overview
out
of
it.
B
Yes,
yes,
yes,
sure,
okay,
but
the
the
aim
of
this
pull
request
is
just
to
start
a
discussion
about
how
we
want
to
to
to
manage
osds
and
stretch
the
bases
in
a
theft,
cluster,
okay,
and
what
is
the
best
way
to
do
that
and
what
is
the
the
the
more
easy
way
for
the
final
user
thinking
always
in
the
final
use?
Okay?
So
it's
just
my
vision.
We
have
commented
this
vision.
Okay,
I
have
talked
with
ernesto.
B
A
B
A
A
Yeah
and
then
yeah,
let's
have
a
look
at
the
purchase
and
discuss
things
in
more
detail,
but
yeah
a
big
thank
you
for
for
the
effort
you
put
into
this
pure
request.
B
I
think
that
we
are,
everybody
is
setting
the
same,
that
is
to
to
achieve
the
better
product
possible,
okay
and
something
that
would
be
easy
for
everybody
to
maintain
and
to
to
go
forward.
A
Indeed,
and
yeah,
with
with
the
current
implementation,
we
have
at
least
some
starting
point,
and
and
from
from
here
we
can
continue
to
improve,
and
I
I
think,
with
the
architecture
of
cfadm
and
with
the
dashboard,
there
is
a
huge
opportunity
to
keep
improving
the
usability
of
of
self-idm,
not
just
in
this
release,
but
also
in
the
release
afterwards
and
for
the
following.
Yes,
so
I
I
see
this
as
a.
A
A
A
A
That
doesn't
work.
Okay,
yeah
have
a
look
at
the
the
lines
I
just
pasted
into
the
miniature.
A
Module
and
the
the
problem
with
keeping
ssh
connections
open
for
for
all
hoes
is
that
it's
going
to
be
a
huge
burden
on
the
resources
of
the
manager.
So
we
have
to
somehow
limit
the
number
of
connections
we
keep
open,
but
in
general,
yes,
that's
already
done.
A
Zach,
I
think
that's
from
you.
We
have
two
different
documentation.
D
D
D
Docs.Com
has
a
has
a
substandard
search
that
allows
you
only
to
search
for
the
first
string
before
white
space.
So
if
you
search
for
sef
ls,
for
instance,
then
you're
going
to
get
staff,
which
is
obviously
a
crummy
search-
that's
not
great!
D
So
I
bought
kefu
about
this
back
in
like
march
and
april,
and
then
he
came
up
with
this
solution
with
this
read
the
doc
site,
which
has
a
great
search,
but
we
don't
get
the
benefits
of
the
sphinx
auto
documentation
for
the
api
on
the
read
the
docs
site.
So
I
don't
know
how
to
solve
this
yet,
but
that's
the
dilemma
that
we're
in
you
can
have
one
site
that
gives
you
good
api,
docs
and
no
search
or
another
site
that
gives
you
search
with
no
api
docs.
D
Eventually,
we
want
to
have
a
site
that
has
both
and
that's
that
issue
I
just
want
to
like
put
that
on
the
stack
so
that
people
are
thinking
about
it.
I
don't
expect
any
real,
like
useful
discussion
out
of
this
right
now.
That's.
C
A
I
mean
one:
one
idea
would
be
to
just
commit
the
the
source
code
of
the
api
documentation
into
into
the
source
tree
and
then
let
read
the
docs
properly
generate
the
the
documentation
of
everything.
A
It's
a
bit
bad
because
you
would
commit
things
that
are
that
can
be
generated.
Yeah.
D
D
D
I
don't
know
I
want
to
call
it
expensive,
but
it's
not
such
an
extreme
extreme
solution,
but
yeah.
That's
that's
a
thing,
that's
a
discussion
that
will
have
to
be
had
over
the
next
few
weeks
before
we
come
up
with
something
like
this
says.
He
doesn't
want
to
maintain
the
read
the
doc
site.
B
D
People
aren't
really
using
it
that
much
and
he
doesn't
want
to
maintain
it
if
people
aren't
using
it
for
the
next
two
months,
but
like
the
search
is
so
much
better
and
the
very
first
thing
that
I
confronted
when
I
came
into
the
stuff
project.
The
the
first
big
issue
was
that
when
I
used
the
search,
I
just
couldn't
find
anything
at
all.
So
I
I
can
only
imagine
that
other
people
who
are
coming
to
a
cold
are
frustrated
with
the
lack
of
good
search.
A
Is
there
a
better
way
to
have
a
good?
Is
there
a
way
to
have
a
better
search
when,
when
generating
the
the
docs
with
swings.
D
D
The
defaults
are,
as
far
as
I
can
see,
it's
the
default
search
and
that's
it.
I
just
put
this
in
here
because
it's
like
one
of
the
things
on
my
mind
for
for
this
week
and
for
this
this
upcoming
month.
D
I
have
notes
on
this
from
last
march,
and
yet
the
sphinx
search
is
just
crummy
and
people
have
raised
issues
about
it
and
it's
just
like
it's
not
a
priority
for
that
project.
D
A
A
Okay
doesn't
seem
to
be
the
case,
then
thank
you,
everyone
and,
let's
sync
up
then
next
time,
bye.
Folks.