►
From YouTube: Ceph Developer Summit Quincy: Dashboard Follow-up
A
Exactly
I
still
like,
I
still
like
what
I
you
know
what
I
said
earlier.
So
if
I'm
gonna
probably
drop,
then
I've
got
another
conflict,
but
you
know
that'd
be
the
only
thing
I
would
add
you
know
ernesto
and
again
you
don't
like
it.
So:
okay,
okay,
okay,.
B
We,
we
may
add
different
views
and
discuss
that
yeah
yeah
yeah
that'll
be
good
yeah.
In
fact,
I
will
add
that
one.
So
how
do
we
call
it?
Toolbox
is
the
way
that
is
called.
A
Yeah
ui
toolbox,
you
know
and
again
the
description
would
be
just
the
ability
to
do
low
level.
Seth
commands
never
leaving
the
ui.
I
mean
it's
like.
I
think
that
would
be
a
very
beneficial
thing
personally,
but
I
mean
we
can
debate
it
with
product
management
and
others.
I
mean
as
we
move
forward
so
and
again
it's
you
know.
In
that
case,
I
think
the
developers
would
enjoy
it
as
well,
especially
if
it
added
value
you
know
to
make
the
you
know
it.
A
A
Had
all
the
you
know,
everything
internalized
and
built
into
the
you
know
the
ui,
so
there's
no
guessing
about
ip
addresses
where
you
go,
you
know
how
you
know
how
where
the
shell
is,
that
you
have
to
go
to
I
mean
all
that
kind
of
crap
I
mean
if
we
could
simplify
that
and
then
basically
have
better
error
handling,
and
you
know
of
of
those
things
as
well.
I
mean
it'll,
be
it
would
be
awesome,
you
know
and
then,
and
then
that
way
they
never
leave.
C
B
Yeah
yeah
yeah.
Well,
let's,
let's
see
so
if
we
may
start
discussing
that,
I'm
not
sure
if
you
want
to
add
anything
else
about
this
topic
or.
B
Now
the
this
case,
I
think
and
well,
I
think
we
may
officially
start
that
this
is
going
to
be
recorded
and
posted
in
the
youtube
channel.
So
welcome
everyone
to
the
follow
up
session
for
the
cds
last
week,
dashboard
team,
so
we're
talking
about
this
first
feature
I
just
was
proposing
it.
It's
basically
embedding
a
cli
kind
of
eli
in
the
dashboard,
the
ui,
so
the
users
can
basically
fall
back
to
the
this
uicli.
When
way
they
don't,
they
cannot
perform
some
specific
workflow
from
from
the
dashboard.
B
So
we
will
probably
have
to
discuss
this
with
other
core
teams
really
with
real
estate,
because
this
might
have
an
impact
on.
I
mean
security
impact
right
now.
The
manager
technically
has
the
same
access
as
a
kind
of
an
admin
access.
So
I'm
not
sure
if
we
have
a
fine
grain
control
here.
B
Just
to
you
know,
impersonate
every
customer,
every
user,
because
technically
every
user
should
act
as
a
with
their
own
permissions,
so
in
this
case,
probably
the
openshift
world
and
that
world
a
little
bit
different
because
the
users
aren't
exactly
matching
the
same.
The
same
roles
here
but
well
I
mean
something
to
discuss.
I
I
will
keep
that
for
a
later
discussion.
B
Okay,
well,
let's
go
I'm
I'm
not
going
to
share
both
windows,
the
backlog
and
the
the
third
part.
I
will
paste
the
other
part
url
in
the
chat
of
this
meeting.
B
It's
also
published
in
the
details
of
this
meeting
so
basically
well,
we
may
start
with
the
high
level
topics
the
bottle
ham
bottom,
half
of
that
ether
pad.
So
basically
we
may
start
with
the
end-to-end
workflows.
B
B
B
And
also
the
cfs
mirroring
as
well,
so
this
would
cover
both
multi-site
and
this
might
be
more.
An
exploratory
work
basically
check
the
ability
for
the
dashboard
to
manage
multiple
set
clusters.
So
technically
it
might
be
feasible.
The
dashboard
at
least
the
front
end
is
just
plugged
to
a
specific
backend,
so
we
may
try
the
switching
dragons
and
see
how
the
dashboard
reacts.
So
we
may
have
this
kind
of
exploratory.
B
B
I'm
not
prioritizing
things
so
what
we
may
discuss
later.
I
don't
think
if
everything
is
prioritized
in
other
a
group
that
at
least
are
listing
things
here.
B
B
The
current
cli
is
kind
of
decoupled
and
you
really
need
different
cli
tools
for
the
rgw
mixing
the
rados
cli
right
with
the
s3
so
having
this
unified
in
a
in
the
front
end.
I
think
it
will
be
very
valuable
apart
from
that
on
the
monitoring
area.
Well,
we
had
this
idea
of
customizing
graphing
dashboards,
and
what
is
most
important
is
persisting
them
right
because,
right
now,
if
you
perform
any
kind
of
modification
in
aggro
final
dashboard,
that
will
be
well.
B
I
think,
right
now,
with
cdm
that
will
be
remove
or
will
be
basically
lost
as
soon
as
the
container
is
restarted.
B
B
That
would
simplify
a
bit
the
deployment
and
probably
improve
the
user
experience,
because
right
now
there
are
some
issues
with
the
ssl
and
different
well
security
constraints
that
use
using
iframes
is
having
on
on
the
front-end,
so
that
will
probably
improve,
and
also
we
have
a
high
availability
for
monitoring
monitoring
stack
right
now,
it's
not
fully
high
available.
So
that's
something
that
it
might
be
easier
to
do
right
now,
with
the
work
on
going
on
safe
avm
with
the
new
ingress
service,
we
might
be
able
to
do
this
lunardan
later.
B
I
saw
today
in
the
depth
mail
list
mailing
list,
someone
asking
for
the
greylock
support
for
for
chef,
so
I
think,
there's
still
some
interest
on
on
that
one,
having
a
centralized
logs
and
being
able
to
process
and
search
through
those
logs
from
a
centralized
location.
So
I
think
it's.
This
is
worthy
at
least
to
explore
that
we
may
just
do
a
poc
just
for
checking,
but
with
an
existing
stack,
elastic
or
whatever
is
being
used
right
now.
B
So
what
else
do
we
have
here?
There
was
a
question
from
telemetry
about
how
to
increase
the
level
of
reporting
that
there
is
a
regarding
dashboard,
like,
for
example,
reporting
which
areas
were
using
from
the
ui
from
the
dashboard
and
kind
of
tracking
those.
So
I
think
it
might
be
interesting
to
such
things,
so
this
could
be
element.
The.
B
B
Apart
from
that
regarding
the
performance
and
scalability
biggest
challenge
that
we
have
there
it's,
this
is
a
shared
concern
between
the
dashboard
and
also
the
core,
rather
the
manager
component.
So
it's
about
the
what
one
facilities,
what
what
I'm
very
exploring
right
now,
which
is
basically
bringing
the
cache
into
caching
at
save
manager,.
B
Api
and
the
other
big
area
there-
and
this
is
only
affecting
the
dashboard-
it's
basically
adding
the
board
for
pagination
filtering
sorting
and
fill
selection
in
the
ui
right
now.
Most
of
the
tables,
basically
fetch
all
the
data
from
the
back
end
and
or
drive
through
that,
and
that
my
what
we
surely
know
that
this
is
not
going
to
scale
and
it's
not
scaling
under
certain
circumstances.
So
that's
probably
another
area
of
work.
B
And
then
there
is
it's
probably
going
to
be
quicker
than
I
expected.
There
is
a
big
hotspot,
which
is
called
the
lean
dashboard,
and
these
are
more
or
less
a
a
series
of
well
initiatives
just
to
streamline
the
development
process
so
part
of
the
idea
there.
It
could
be,
for
example,
generic
ui
components,
and
but
I
will
just
paste
the
high-level
topic
here
and
there
are
some
sub
triggers
that
we
may
individually
address,
but
this
could
be
more
as
the
high
level
idea.
B
So
and
then
another
initiative
here
in
the
land
dashboard
is
having
cross-component
interfaces.
B
So
basically,
for
example,
our
main
well
here
in
the
dashboard
world
right
now
is
fav
so,
for
example,
just
having
a
common
interface
there,
but
it
might
be
also
interesting
to
have
something
with
rgw
components
or
with
other
components,
so
this
might
mean
just
having
shared
schemas
or
classes.
So
we,
both
both
components,
share
the
same
definition
and
we
don't
need
to
implement
and
ensure
that
these
same
data
structures
are
in
sync.
B
And
there
are
other
areas
like
we
talk
about
adding
python
runtime
type
checks,
so
this
other
one.
B
We
talk
about
the
replacing
the
rafana
jsons
with
generators,
this
could
be
python
or
it
could
be
also.
B
Jsonnet
so
right
now
there
are
multiple
alternatives
there,
but
yeah
main
idea
here
is
just
to
avoid
the
you
know:
pain
of
reviewing
a
thousand
lines,
json
file
with
a
lot
of
boilerplate
and
defaults,
etc.
So
we
can
concentrate
on
just
previewing
and
perhaps
adding
even
unit
testing
to
graphana
dashboards.
B
B
And
if
you
are
reviewing
also
the
we
see
priority
center
path-
and
I
I
miss
anything-
please
feel
free
to
step
in
and
write
directly
at
any
anything
there.
If
I
missed
that
one
so
last
comments
there,
how
about
the
I
using
yeah
this?
This
was
a
plc
that
we
were
talking
about,
so
it
would
be
how
we
call
this
one
native
dependencies
right.
We
call
it
native
dependencies.
B
Instead
of
I
will
refine
this.
These
journal
cards
after
the
meeting,
so
they
are
aligned
with
the
current
syntax
instead
of
packaging.
B
B
B
And
last
one
here
is
implement
the
self
test
for
the
dashboard.
This
is
something
feature
that's
already
there
in
ncf,
but
the
dashboard
currently
doesn't
implement
any
kind
of
search
test.
So
the
idea
would
be
to
provide
this
self-test
built
in
in
the
dashboard.
B
So
an
operator
after
installing
the
dashboard
can
run
this
shelf
test,
for
example,
and
they
can
quickly
receive
feedback
on
the
status
of
the
all
the
different
components
from
the
cli,
and
they
don't
really
need
to
go
page
after
page
checking
that
everything
is
okay
on
whether
there
is
an
exception
in
the
rjw2
page
or
whatever.
So
that's
mostly
about
the
bottom
half
of
that
ether
pack
and
the
top
half
is,
I
think
I
filed
most
of
these
trackers
under
tracking
trackers,
like
most
of
the
usability
suggesting
some
feedback.
B
B
The
only
big
thing-
and
I
I
when
I
was-
was
watching
the
cbs
follow-up
session
from
the
wrestling
they
talked
about
this.
B
Us
m
clock
us,
so
I
think
this
might
be
a
nice
improvement.
It's
just
this
plane
or
well
configuring.
Don't
figure?
U
s
and
clock
from
dashboard.
This
could
be
more
or
less
as
a
follow-up
improvement
with
what
I
think
it
was
tatiana
did
for
the
recovery
profiles.
So
I
think
there
are
new
ways
of
configuring,
the
internal
schedulers
and
stuff,
and
that
my
and
I
they
were
talking
about
this
in
the
in
the
core
brothers
team,
so
this
might
be
interesting
to
expose.
C
B
And
what
else
we
have
here
from
the
orchestrator
area?
Yeah,
that's
also
filed
within
the
workflows.
Hopefully
we
will
have
everything
there
in
sync.
B
B
B
B
Okay
and
the
last
thing
to
provide
feedback
was
ffs,
we've
got
the
surface,
mirror,
that's
all.
What's
included
there,
yeah
integration
with
the
snapshot,
scheduler.
C
B
B
C
B
B
B
Okay,
so
I
think
that's
all
now
it's
time
to
pre-order
things
so
well,
what
we
may
start
doing
is
just
moving
to
the
bottom
of
the
column,
the
things
that
we
consider
less
relevant
so
alfonso
you
mentioned
yeah
the
this
one's
right.
It
was
one
of
these.
B
And
probably
top
things
are
already
there,
like
the
plastic
installation,
which
are
these
already
improvements,
support
for
hdw?
This
probably
would
be
more
or
less
aligned
with
the
rgw
too
big.
B
B
B
The
integration-
I
think
this
might
be
interesting,
also
based
on
your
expression,
experience
with
the
nfs
right,
at
least
for
queen
c2.
C
Replay,
but
it's
more
priority
that,
if
the
way
features
or
this
one,
I
think
both
are
more
or
less,
I
think
maybe
objective
ways
a
little
bit
more
priority
from
the
feedback
collected,
but
yeah.
B
Yeah,
I
think
I
mean
this
might
be
just
like
cleanup.
The
thing
is
that
part
of
the
issues
that
we
are
facing
right
now
with
the
nfl
support,
probably
might
you
know
get
away
with
with
this
right
if
we
replace
that
at
least
we
unify
where
the
management
of
the
nfs
is
it's
been
done
and
we'll
have
to
have
this
quality
that
we
currently
have
with.
C
For
example,
this
unified
are
the
rtw
cli
is
tightly
related
to
multi-site
topper,
because
without
this
there
is
no
way
to
do
variable.
Human
yeah,
yeah,
yeah.
D
D
I
think
we
should
just
move
that
up:
okay
and
just
as
soon
as
we.
If
we
try
to
do
that
soon,
then
that'll
unblock
thanks.
D
Just
be
clear,
I
think
the
sort
of
the
conservative
first
step
there
is
to
just
pull
in
the
the
parts
related
to
round
management,
not
the
entire.
The
radio
skate
we
have
in
cli
is
huge,
but
just
the
pieces
that
we
need
for
all
the
rail
and
zone
configuration.
I
think
that's
is
that
the
part
that
you
need
or
is
there
are
there
more
that
you
need,
which.
D
Sorry
this
one,
when
you
say
unified,
rgwc,
I'm
just
wondering
which
yeah,
which
you
know,
which
and
maybe
yeah
just
describing
a
ticket
like
which.
B
Yeah
yeah
yeah-
I
was-
I
was
thinking
just
on
doing
this
offline,
just
refining
the
the
yeah
every
ticket.
So
we
can
keep
this
short
yeah,
but
the
idea
would
be
just
at
least
expose
the
commands
that
we
currently
need
for
the
configuration
of
realms
structure,
trying
to
do
all
the
things
that
currently
go
in
the
rados
admin
cli
in
in
this
in
this
tool.
So
we
can
send
commands
to
the
monitor.
B
Okay-
and
you
were
talking
about
this
right-
I
I
was
watching
the
rather
meeting-
that's
what
you
were
talking
about
exposing
this
from
the
from
the
dashboard.
So
where
would
this
should
go
in
in
the
dashboard.
D
I
honestly,
I
don't
know
it's
pretty
new,
so
I
would
make
sure
that
we
have
the
like
the
conversation
about
how
exactly
we
can
simplify
this.
But
okay
yeah.
I
don't
know,
I
don't
know.
I
don't
know.
I
don't
know.
B
So
this
is
managing
all
the
diamonds,
so
it's
or
specifically
attached
to.
D
I
mean
there
are
multiple
parts
of
it.
The
simplest
sort
of
phase
of
it
is
just
controlling
how
we
prioritize
client
work
versus
background
work
and
I'm
not
sure
if
that
should
be
a
global
set
of
switches
or
knobs
or
whether
something
we
want
to
control
purple
or
something
like
that,
possibly
both.
D
But
the
second
part
of
it
is
like
client
by
client
qos,
and
I
have
a
I
imagine.
That
is
something
that
would
be
pretty
slick
if
it
was
integrated
with
like
the
top
view,
or
something
like
that,
so
that
you
can
there's
some
unified
view
where
you
both
sort
of
see
what
clients
are
currently
using
and
also
what
your
qos
metrics
are
at
the
same
time.
D
D
Things
the
second
part
they
don't
exist
yet
for
the
clients
yeah,
it
would
eventually
it'll
be
like
by
rbd.
Image
would
be
one
place,
I'm
not
sure
where
it'll
go
first
ffs
and
it
doesn't
really.
The
rgw
is
its
own
thing,
so
yeah
it
needs
a
lot
of
planning
before
we
get
that
far
okay.
D
B
D
I
would
I
would,
if
I
had
to
just
choose
something,
I
would
say
that
the
first
phase
of
it,
which
is
controlling
just
qos
for
client
versus
background
work.
That
should
be
definitely
in
scope,
although
a
little
bit
of
scoping
needs
to
be
done
to
figure
out
exactly
how
detailed,
how
much
control
we
actually
want
to
give
people
there,
but
but
the,
but
the
client
part
is
probably
going
to
be
pretty
bleeding
edge,
even
as
we
wrap
up
quincy.
D
B
Not
sure
states,
if
you
were
here
when
jeff
was
talking
about
this
toolbox
thing.
C
D
B
Yeah,
it
might
be
just
I
mean
some
placeholder
for
injecting
monitor
commands,
but
even
though
that's
well,
the
thing
is
that
we
currently
don't
have
fine-grained
mapping
between
you,
the
safe
dashboard
users
and
chef
hearings,
not
sure
if
we
should
do
something
like
that
link
both
things.
So
we
could
map
permissions
or
something
right
now.
I
guess
every
user
of
the
dashboard
technically
would
escalate
to
a
manager
giving
right,
so
they
could
do
as
much
as
the
manager
can
does
in
I'm
doing
in
a
safe.
So
that's
my
very
risky.
D
B
And
what
else
we
have
here,
just
installation
with
sir
yeah?
This
is
something
that
we
will
yeah
well
paul
costner
was
prepared.
D
B
I
left
out
everything
related
to
group
because
in
the
meeting
I
remember
that
we
talked
about
having
a
follow-up
meeting
with
the
rook
team
yeah
so
but
yeah.
If
we
already
identify
this
one,
we
may
just
start
it
here:
okay,
probably
yes,
the
first
thing
is
not
here
so
yeah.
We
should
have
this
meeting
just
for
reviewing
the
current
support
from
the
book
orchestrator.
B
B
We
may
just
move
that
to
the
c
plus
website,
depending
on
also
yeah
and
yeah,
and
also
we
have
to
check
because
one
possible
concern
with
this
is
the
possibility
of
having
shared
objects
in
python
and
how
the
different
modules
would
consume
the
same,
shared
objects
and
that
might
well
private,
savvy
unexplored
territory.
So
the
sub
interpreters
thinks
is
allowing
you
to
do
things
that
probably
are
not
much
tested
in
in
python.
So,
okay,
that's
mostly
it
regarding
the
customization.
B
B
And
actually
it's
probably
over
customization
from
the
user
and
right
if
we
provide
you,
know
sufficiently
customize
the
dashboards
and
I
think
right
now
they
are.
They
are
dashboards
per
global
dashboards
for
air
component
and
also
per
instance
like
ferrari
image.
B
So
probably,
if
there
are
specific
requests
for
that
by
the
way,
I
I'm
digesting
the
save
user
survey,
at
least
the
dashboard
comments
and
feedback.
So
from
that,
I
will
try
to
also
collect
some
ideas
for,
for
this.
B
B
I'm
currently
not
aware
of
whether
alert
manager
provides
an
interface
for
customizing
that,
because
in
the
past
I
think
everything
the
alerts
were
configured
mainly
from
a
jumble
file
or
a
new
file.
So
it
wasn't
very
dynamic.
If
you
want
to
modify
anything
there,
I'm
not
sure.
If
there's
a
way
to
we
might
explore
that.
B
The
high
availability
of
the
monitoring
stack,
that's
something
that
I
guess
with
the
latest
changes
in
cf.
It
might
be
closer
than
we
suspected.
So
perhaps
it's.
D
Yeah,
I
just
I
don't
know
how
important
it
is
yeah
I
think
I
would.
I
would
be
inclined
to
move
that
one
down.
Okay,
there's
also
the
like
the
customizing
grafana
dashboards.
I
don't
know
if
that's
something
that
people
have
been
asking.
B
C
B
The
averaging
variant-
probably
this
this
one,
would
be
an
easy
one,
because
the
most
of
the
mirroring
the
rv
mirror
is
already
there.
So
probably
it
would
be
just
reducing
code
and
objects,
just
in
fact,
I
remember
that
the
pulled
mirroring
was
going
to
be
deprecated
right.
B
So
we
we
may,
I
think,
jason
opened
a
a
trucker
just
or
the
dashboard
just
too
well,
I'm
not
sure
if
deprecating
or
at
least
prioritizing
the
snapshot
and
the
image
based
mirror
okay.
So
we
may
give
this
more
priority.
Then,
let's
move
that
to
the
multi-side
etc.
B
And
what
else
the
log
aggregation
yeah.
B
Yeah,
it
could
mean
just
basically
gathering
all
the
different
looks
from
the
different
demos-
I
guess
mostly
from
osd
ones,
and
exposing
that
in
a
kind
of
elastic,
search
or
friendly
or
kind
of
a
centralized
way,
so
not
sure
if
we
may
also
embed
that
via
iframes
in
the
same
way
as
we
did
for
grafana.
I
also
provide
a
placeholder
blocks
tab
in
the
in
the
dashboard,
where
you
can
quickly
search
across
logs
coming
from
all
the
demos
in
the
class
instead
of
relying
on.
I
guess
right
now.
B
This
is
journal
right,
journally,
not
familiar
with
the
way
that
the
journal
deluxe
are
gathered,
but.
D
B
Yeah,
I
think
sebastian
recently
mentioned
that
the
potman
journal,
the
integration,
was
a
bit
faulty
right.
B
Okay,
so
we
made
a
very
test
this
a.
B
This
is
not
implemented
right
now,
and
I
I'm
not
very
familiar
with
what
other
modules
do
here,
but
I
think
from
the
dashboard
perspective,
it
might
be
interesting
to
have
this
specially
like
a
post-install
step,
so
you
install
the
dashboard
just
run
this
and
ensure
that
all
different
components
have
connectivity
with
their
counterparts
and
everything
is
working.
Okay,.
D
B
C
D
B
Yeah,
there's
a
huge,
a
huge
tracker
issue:
it's
here:
it's
basically
a
meta
tracker
where
we
are
filing
everything
that
is
kind
of
you
usability,
so
it's
more
or
well
there's
a
hierarchy
with
the
different
components
like
host,
etc,
but
yeah.
I
I
also,
I
think
I
filed
your
suggestions,
for
example,
recently
about
the
file
systems
and
diamonds,
the
services.
So,
but
if
you
anything
that
you
I
mean
missed
there,
you
can
simply
track
that
in
that
one
and
we
are
trying
to
implement
this
progressive.
C
B
Perfect,
so
thank
you.
Everyone
have
a
nice
day,
see
you
soon,
bye.