►
From YouTube: Ceph Orchestrator Meeting 2021-06-15
Description
A
So
here
are
the
weekly
meeting
minutes
selfie
idm
ntw
set
up
in
the
dashboard.
There
is
a
pull
request
for
that
open.
A
I've
painted,
I
can't
do
anything
more
iphones
or
okay.
I
think.
B
B
Can
you
share
my
screen?
Are
you
able
to
see
my
screen?
Yes,
yep?
Okay,
so
the
current
status
is
that
we
set
credentials.
So
we
can
connect
on
upper
gateway
basis,
so
you
can
select
the
object
gateway.
Currently,
if
you
go
to
demons,
I
have
in
this
mock
setup,
multi
multi
site
that
I
have
for
development,
two
demons
belonging
to
different
realms.
Okay.
B
B
B
B
So
my
question
here
is
that,
from
my
past
experience
when,
for
example,
selecting
or
creating
buckets
and
selecting
the
placement
target,
I
found
that
if
a
demon
was
running
in
zone
one
and
you
were
to
create
a
bucket
and
place
it
in
a
placement
target
that
was
a
belonging
to
a
different
zone,
you
could
not
do
it.
I
got
an
error.
B
Through
a
credentials
and
a
host
an
abort,
what
should
be
the
the
the
main,
the
main
concept,
to
connect
to
a
zone,
a
real
or
a
gateway
that
that
is
my
my
my
doubt
because
from
I
I
know
from
past
experience,
I
know
that
the
zone
is
important
to
select,
but
I
think
that
the
proposal
is
to
only
sell
everyone
and
just
pick
up
a
diamond
running
once
on
a
certain
realm,
and
I
I
I
have
doubts
about
it.
B
To
okay
and
currently
what
we
are
doing
here,
what
we
are
doing
is
right
now
this
is
the
current
status.
We
are
creating
the
url
to
connect
to
the
admin,
ops,
api,
using
the
host
and
the
port
and
the
access
key
and
the
secret
key
that
we
pick
up
from
the
credentials
and
my
question
is
then:
the
new
the
new
proposal
should
be
instead
of
selecting
gateways
just
offer
the
user,
the
zones
that
the
diamonds
are
running
on
and
then
pick
up
a
zone,
select
a
zone
right,
but
but
not
be
doing
real.
B
D
So
I
backing
up
a
step
like
I
don't
think
it
makes
sense
to
target
gateways
separately,
because
if
you
have
two
gateways
running
on
the
same
zone,
they're
just
going
to
see
the
same
data
for
everything.
D
Sense
to
have
zone
as
the
thing
here,
but
I
don't
know
if
it
makes
sense
to
mix
zones
that
are
in
different
realms,
because
different
realms
have
completely
different
sets
of
data
and
users
and
buckets
so
you
might
want
to
have
like
realm
selection
as
just
a
completely
separate
level.
So,
like
you,
click
on
realm
one,
and
then
you
see
what
zones
are
there
and
you
can
operate
on
it.
D
B
Other
so
my
understanding,
it
should
be
that
provides
somehow
like
two
levels
of
selection,
and
so
you
you
pick
up
the
real,
but
then
you
have
to
select
at
least
one
zone
or
other
or
the
before
zone
from
that
realm.
B
D
B
Okay,
so
at
some
point
the
user
must
select
the
zone
apart
from
the
rim,
so
maybe
it
should
be
a
double
or
where
we
can
find
out
a
better
way
to
well
the
ui
is
we
can
take
out
the
the
ui
design
here,
but
because
what
I
want
to
clarify
the
concept?
Okay,
so
what
I
see
here
then
yeah-
I
I
I
I
don't
know
if
it's
joined
or
not
in
order
to
clarify,
then
when
he
says
the
important
part
is
to
select
the
realm.
B
I
assume
that
we
had
a
two,
a
two
drop
down,
selection
or
well,
some
kind
of
enhancement,
but
it
should
be
enough
and
that
then
the
implications
here
are
well.
We
need
this.
This
pull
request,
because,
right
now
we
we
don't
have
this
info
until
this
is
merged
the
real
name
and
ram
id
from
the
service
map,
because
without
credential
we
can
only
retrieve
the
the
gateway
list
from
the
and
the
zones
from
the
service
map.
B
So
what
I
will
do
is
to
to
leave
here.
Following
that
conversation,
a
comment
regarding
this
about
the
zones
and
then
to
see
if
the
then
the
fvm
and
dashboard
can
be
consistent
on
set
up
the
credentials.
The
only
thing
is
that,
maybe,
as
we
are
now
setting
up
the
credentials
on
a
per
gateway
basis
like
here
maybe
doing
on
a
personal
basis
will
require
from
from
one
point
release
for,
for
if
we
are
backporting
this
to
pacific
from
one
point
release
to
pacifist
to
another,
this
will
imply
some
kind
of
upgrade.
E
Yeah
is
that
demon
one
id
is
that
the
key
for
that
map
is
that
used
for
something
specific
or
is
it
just
like?
Could
I
just
replace
that
demon
id
with
the
zone
id
or
does
it
need
to
be
a
valid
demon
id.
B
No
yeah!
Well,
we
are
talk.
We
were
talking
about
you
just
joined.
We
were
talking
about
that.
I
think
that
the
relevant,
the
well
the
relevant
thing
in
order
for
operate
correctly
with
packets
and
users
is
the
zone.
B
So
from
my
understanding
from
the
casey
feedback
is
that
we
should
set
up
credentials
based
on
zones
not
on
gateways,
because
you
could
have
the
the
case
that
you
are
having
more
than
one
diamond
running
on
the
same
zone.
So
it
should
not
matter
what
time
when
you
connect
on,
but
it
should
be
the
credentials
that
allows
you
to
connect
to
a
diamond
running
on
that
zone.
So
the
drop
down
should
be
based
to
have
two
levels.
B
E
D
D
E
D
E
B
B
B
So
we
should
be
careful
that
use
the
credentials
for
the
real,
but
taking
into
account
that
I
have
to
select
a
zone
because
otherwise
I
have
I
can
have
problems
when
selecting
when
creating
buckets
on
a
zone
that
it
should
not.
That
is
not
belonging
to
the
demon.
So
then,
what
I
understand
the
credentials
should
be
per
real
and
the
diamond
selection
in
the
ui
person
that
that
is
the
agreement,
then
that
we
have.
A
B
Yeah,
okay,
so
then
this
should
be
perry
on
the
settings.
The
credential
okay,
then
my
comment
says:
is
that
the
legacy
way
that
you
only
put
an
access
secret
key
for
a
single
site
is
enough
for
the
multi-site,
then
yeah.
We
need
this
dictionary
like
but
simplified
on
upper
regional
basis
sounds
good,
okay,
okay,
so
I
will
try
to
then
to
contribute
on
that
p
area.
So
we
can
have
an
atomic
feature
I'll
update
this
full
request.
Today,.
B
B
We
should
take
care
of
this.
This
upgrade
tool
on
the
on
the
cfd
and
on
the
orchestrator
side,
yeah
yeah.
I
can
do
it
on
the
other
side.
B
E
I
guess
the
last
thing
is
this:
the
way
this
flow
requests
work,
it
always
disables
get
verification.
Do
we
worry
about
that.
E
E
Yeah
I
I
went
through,
I
was
going
through
the
comments
again
and
you
had
suggested
nfs
export
create,
which
is
better,
but
I
think
the
way
that
it
needs
to
work
is
as
a
it
needs
to
be
item.
Potent
which
makes
me
wonder
if,
instead
of
create
it
should
be
apply.
A
That
that
would
mean
we
need
some
kind
of
a
unified
update
and
create
the
way
that
the
current
implementation
works
right
now
yeah
and
it
should
accept
on
the
things
on
the
cli
and
it
should
accept
a
json
document
or
I
don't
know
we
we
could
even
pass
in.
A
I
I
don't
really
like
the
json
document,
because
it's
some
kind
of
artificial,
adding
a
new
kind
of
representation
of
those
ganesha
configurations.
So
what
would
you
think
of
just
accepting
a
ganesha
config.
E
E
E
E
Yeah,
I
mean
that's
sort
of
the
the
cheap
easy
way
to
do.
It
is
if
it
exists,
then
just
do
nothing
and
don't
actually
compare,
but
the
way
it's
written
right
now
it
basically
sort
of
merges,
update
and
the
create
and
update
paths
are
basically
the
same,
except
that
if
you
use
the
update
command,
if
it
doesn't
exist,
it'll
error
out
but
like
otherwise
it's
basically
the
same.
A
Not
perfect,
you
know
there
is
this.
A
E
So
for
the
for
the
create
command
like
the
create
ffs,
like
the
cli
version,
that
version.
C
E
Isn't
always
okay,
so
that
so
that
should
definitely
be
item
potent.
If
it's
not
that's
a
that's
a
bug,
I
mean
if
we
don't
need
like
a
an
update
that
errors
out.
If
the
export
doesn't
exist,
then
everything
gets
simple
right,
there's
just
like
one
there's
one
command
that
takes
jason
and
so
I'd.
Actually,
I
would
actually
prefer
that
I
wasn't.
I
just
wasn't
sure
if
we
wanted.
E
A
A
F
I
I
do
have
an
update,
so
I
think
I
got
the
apply
rgw
with
the
the
realm
and
zone
flags
enabled
the
realm
flag
is
pretty
useless
from
what
I
can
tell,
but
the
zone
flag
works,
so
yeah,
that's
kind
of
done.
For
now.
I
just
have
to
make
a
pr
and
test
it
some
more
for
the
thing
that
the
rook
client
python
thing
sebastian.
I
don't
know
if
you
saw
my
messages
in
irc
but
yeah
I
sent
some
some
links
concerning.
F
Okay,
I
can
send
those
messages
with
the
links
again.
If
you
didn't
see.
C
A
C
All
right,
in
my
case,
yes,
the
what
trying
to
to
create
something
in
order
to
provide
a
fast
way
to
create
a
rook
cluster,
okay,
the
cluster
using
the
rock
operator
and
the
rook
orchestrator,
and
a
part
of
that.
Well,
I
have
take
a
look
to
the
pull
request
that
you
did
about
the
modifications
in
the
root
client
and
the
group
python,
client,
celeste,
and
what
I
want
to
do
is
just
to
to
generate
the
the
classes
and
and
to
use
the
classes.
Okay,
in
order
to
be
sure
that
everything
is
working.
C
In
my
view,
I
think
that
the
python
client
is
useful.
Okay,
because
it
basically
to
two
reasons:
okay,
one
is
that
is
going
to
to
be
more
easy
to
work
with
the
different
troops
here.
This
using
the
the
python
client
and
the
other
one
is
that
we
are
going
to
avoid
problems
and
mistakes
using
in
different
ways
the
the
series.
C
A
A
B
I
have
sorry
yeah.
I
just
asked
to
the
dashboard
fox-
and
they
mentioned
about
these
facts
thing,
so
I
haven't
been
following
that
very
closely,
but
the
I
think
the
discussion
was
versus
getting
the
data
from
the
cache
right
from
the
key
value,
the
store
right
or
providing
a
new
command
for
doing
that
right.
A
I
I
don't
know
if
it
really
makes
sense
to
him
to
to
already
do
a
cli
command
for
it,
because
that
would
mean
that
we
have
to
properly
clean
up
the
data
that
then
any
other
effects,
because
as
soon
as
we
are
exposing
it,
we
are
going
to
have
people
depending
on
the
concrete
structure
of
it.
But
it
has
to
make
sense.
A
B
A
It
will
always
be
there.
Okay,
what
what
you
can't
do
in
any
case
is
fetching
the
live
data
from
all
the
different
hosts,
because
it
would
to
take
too
long,
especially
if
you
have
servers
that
are
running
into
timeouts,
you
would
block
everything,
so
you
really
can't
do
any
network
calls
from
from
the
cli
handler.