►
From YouTube: Ceph Orchestrator Meeting 2021-07-06
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
Is
it
is
it
okay
for
everyone,
if
we're
just
putting
those
into
the
orchestrator
weekly
other
pad,
or
should
we
write
design
documents?
I
don't
know
I'm
typically
kind
of
resistant
to
write
design
documents,
because
I
personally
often
think
they're
often
not
that
worth
writing
it
down,
because
for
me,
typically,
it's
okay
to
just.
B
I
I
mean,
for
example,
if
we
are
going,
for
example,
for
this
one
one
architecture
about
nfs
to
rule
them
all.
Maybe
it
makes
sense
to
have
it
published
somehow
somewhere,
where
all
members
and
people
and
even
downstream
people,
can
take
a
look
and
see
how?
What
is
that?
The
final
workflow
for
nfs
and
rtu
and
the
diagram
and
all
that
kind
of
things.
So
it
depends
on
the
complexity
right.
A
So,
for
example,
we
had
that
rook
orchestrator
weekly
rook
orchestrator
meeting
last
week,
where
we
had
a
bunch
of
of
minutes
in
the
orchestrator
weekly
as
I've
had,
but
nothing
definitive,
that's
a
bit
of
edge
case
because
there
was
just
no
no
solution
back.
Then
there
was
a
follow-up
meeting
in
the
miniature
rook
pv
pvs
as
ipad.
Where
we
came
then
was
a
proposal
that
that's
typically
how
we
work
right
in
upstream.
A
C
I
I
think
that
that
can
be
heavily
improved.
Okay,
what
we
have
in
the
other
part
is
just
a
brief
description
of
what
are
the
things
that
we
think
that
we
need
to
do.
Okay,
but
that
is
not
that
the
same.
That
is
not
an
architecture.
That
is
not
something
that
is
a
thinking
in
the
connection
between
different
pieces
of
the
system,
and
this
result
in
things
that.
C
All
the
elements
in
the
theft
cluster
are
not
perfectly
enlightened.
Okay,
and
this
is
because
we
do
not
have
a
white
vision.
Okay,
obviously,
we
need
the
people,
for
example,
like
sage
like
you
like,
more
people,
okay,
that
has
a
good
knowledge
of
all
the
ecosystem
in
order
to
do
this
kind
of
design.
C
C
A
Yeah
that
that
brings
me
to
redmine,
we
have
an
agile
board
in
the
orchestrator.
A
In
the
orchestrator
project,
but
we
are
not
really
using
it,
as
I
personally
find
it
a
bit
overwhelming
and
not
really
to
the
point.
A
But
there
are
certainly
ways
to
make
use
of
that
red,
mine,
bot
and
so
or
alfonso.
Do
you
make
use
of
that
red
mine
board?
Yes,
you
do
right.
B
A
Yeah
I
saw
that
someday
sometime
right.
I
think
it's
a
project,
isn't
it.
A
It's
a
project
where
you
are.
A
You
you
see,
pull
requests
right,
that's
you're!
Using
that
to
see
pull
requests.
B
B
I
have
to
rely
on
that
with
mine.
For
that.
A
But
I
see
your
point
yeah
the
redmine
atari
bot
is
not
really
great,
it's
certainly
usable,
but
it's
not
it's.
It's
not
great.
B
Well,
you
can
use
it
but
yeah.
If,
if
you
want
to
try
something
like
scrum,
instead
of
the
camera
approach,
you
will
need
to
cherry
pick
the
issues
that
you
want
to
work
during
a
sprint.
A
Yeah,
because
we
have-
I
don't
know-
maybe
200
issues
in
that
bot
and
that's
way
too
overwhelming,
especially
the
there
is
no
way
to
differentiate
between
the
backlog
and
issues
that
are
supposed
to
be
in
progress.
A
And
issues
that
also
in
in
issues
that
are
in
need
more
info
are
also
not
really
in
progress
because
of
the
way
to
to
further
clarify
if
they're
actually
going
to
be
tackled.
Sometime
soon,.
A
A
A
B
B
B
Yeah,
the
only
thing
is
that
you
have
to
take
care
of
link
the
the
vaxilla
with
the
tracker
issue
instead,
but
apart
from
this,
you
are
done.
That's
how
we
work.
Currently
I
mean
someone
creates
an
axilla
automatically
digital
car
is
created,
so
you
have
only
to
put
the
the
tracker
issue
link
or
the
pull
request,
even
in
the
maxilla
and
all
is
linked.
A
I
hear
you,
I
hear
you,
but
I
would
like
to
have
another
yeah.
A
B
A
No
baxilla
is
really
the
the
worst
tool
for
for
the
job.
In
my
opinion,
this
is
so
praczilla
is
so
bad.
B
A
For
for
it,
it
is
possible
for
for
redmine
to
have
private
issues,
but
private
issues
are
written,
are
still
upstream
issues.
B
A
What
we
have
certain
integration
between
the
tools?
We
have
no
proper
integration
between
between
github
pull
requests
and
and
redmine,
for
example,
everything
is
manual,
so
you
can't
close
tracker
redmine
issues
from
bachelor
police.
B
E
A
Which
board
theater
board?
No,
I
haven't
looked
at
it
before
why.
E
E
I
guess
yeah
like
just,
for
example,
like
the
new
category
like
then
we
like
prioritize
these
somehow
and
a
way
to
like
sort
them.
So
it's
what's
more
important
to
be
done,
then.
You
know
this
list
goes
on
for
three
four
hundred,
I'm
not
sure
which
one
would
be
the
you
know.
What
needs
to
be
done
what's
prior
to
the
priority
would
make
that
more
useful.
A
So
we
could
manually
mark
issues
like,
for
example,
tag.
We
recat
could
tag
issues
for
in
in
the
in
the
answer
board
and
then
either
remove
or
or
add
that
issue
at
that
tag
to
the
to
the
red
magnesium
and
we
could
then
maybe
filter
the
edge.
I
bought
only
for
things
that
are
attacked
with
their
child
or
I
don't
know.
A
E
So
we're
trying
to
use
it
to
kind
of
like
plan
our
weekly
and
keep
like
this
is
what
everybody's
working
on
like
some
people
are
gonna,
be
working
on
bugs
and
then
they're
just
not
accounted
for
they're.
Just
only
tracking
features.
D
I
mean
I
should
start
using
this.
I've
I've
been
using
working
off
of
the
the
trello
board,
but
I
will
pin
this
tab.
A
Okay,
maybe
it's
a
job
for
me
to
just
go
over
the
entire
ball.
What,
if
I
can
make
use
of
it.
D
D
I
mean
some
of
these,
like,
even
if
you
just
look
at
the
ones
at
the
top
they're
like
not
like
this
one
make
use
of
progress
items
I
already
merged
something
that
uses
progress.
Items
for
deploys.
The
only
thing
it
doesn't
do
is
the
drive
groups,
because
it's
very
awkward
with
the
drive
groups.
You
don't
know
how
many
they're
going
to
be
so
I
don't
even
know
if
it's
worth
implementing
it
feels
like
this
card
is
like
half
done
and
the
other
half
is
like.
A
Yeah
but
the
important
part
having
having
those
deployment
properly.
A
Visible
at
least
you
could
make
a
progress
over
the
hosts
say
if
you,
if
you
start
deploying
those
these
hosts
here,
you
have
to
look
at
and
then,
if
you're
finished
with
it,
then
you
can.
You
can
close
it
at
100,
so
that
could
work.
A
Over
200
issues
in
the
in
the
sub
project,
here
already
yep,
so
things
like
this
issue
here,
if
you're,
if,
if
your
host
is
by
chance
named
localhosted
local
domain-
and
we
are
deploying
two
monitors
on
the
localhost
with
two
different
names-
this
is
a
clearly
a
bug
and
I
would
like
to
to
get
it
solved,
but
it's
super
detailed
and
it's
in
in
that
case,
if
we
decide
on
keeping
those
edge
cases
properly
documented
then
yeah.
D
A
Things
like
safety
and
bootstrap,
just
docker
doesn't
actually
work.
A
A
Every
other
daemon
is
going
to
use
podman
because
it's
not
stored
persistently
in
the
in
a
manager
module.
So
it's
definitely
a
bug.
A
A
A
There
is
a
race
when
stopping
containers
that
sometimes
the
container
is
not
properly
stopped
and
you
have
to
restart
the
host
in
order
to
restart
a
container.
D
A
I
don't
know
it
it's
a
rare
event
where
we
cannot
properly
shut
down
containers.
It's
I
don't
know
one
one
in
ten
thousand
container
restarts
at
fair,
but
it's
often
enough
that
we
had
to
disable
center
seven
in
totality.
C
Yeah,
that's
me
yeah,
asking
things:
okay,
just
to
know
well
we're
going
to
use
different
pbs
providers,
because
this
is
something
what
we
have
in
the
book
records
now
is
something
that
is
based
in
the
in
the
local
storage
operator.
Okay,
which
is
installed
manually,
I
think
so
joseph
can
you
can
you
clarify
this
or.
F
Yeah,
so
I
think
you're
talking
about
how
the
the
whole
point
of
like
that
pr
is
to
say
like,
depending
on
like
what
storage
class,
the
user
specifies
we're
going
to
try
to
get
the
best
information
we
can
from
like
different
provisioners
or
whatever.
F
Basically,
the
idea
is
that,
like
whether
somebody
uses
like
lso
or
open
ebs
or
like
topo
lvm,
like
there's
going
to
be
like
the
orchestrator,
will
be
able
to
get
information
based
on
like
what
storage
class
the
user
specifies,
and
we
can
tell
like
what
provisioner
that
storage
class
is
from,
so
that
we
can
look
at
like
the
correct
resources.
F
C
Okay,
the
the
two
options
that
you
have
added
to
the
theft
list
of
options
is
in
order
to
to
know
what
is
the
storage
class
that
we
are
going
to
use.
Okay,
yeah,
okay,
okay
and,
and
so
the
the
installation
of
the
local
storage
operator
or
another
kind
of
operator
in
order
to
to
deal
with
pvs
is
something
that
is
outside
the
stator.
F
C
A
But
I
agree-
I
I
feel
it
a
bit
odd
that
we
are
that
that,
in
order
to
use
the
orchestrated
integration,
we
are
kind
of
forcing
users
to
install
another
operator
which
is
done
manually.
It's
I
don't
know.
F
Yeah,
I
think
it's
because
the
the
whole
like
direction
that
rook
is
going
in,
is
that
they
don't
want
to
use
like
raw
device
like
for
bare
metal
setups.
They
they
don't
want
to
use
like
just
raw
devices
directly.
They
want
to
like
abstract
it
through
pvs.
So
that's
that's
kind
of
where
this
comes
well.
C
A
G
Just
about
fetching
the
the
gather
effects
or
for
a
host
if
it's
in
the
main
transport,
but
I
see
currently
it's
returning
me
the.
G
It's
mostly
because
the
all
demons
are
down,
but
I'm
not
sure
like,
even
if
I
am
because
the
gather
effect
is
fetching
from
the
case
right
now,
so
I
I
was
assuming
that
maybe
it
should
return
the
host
fix,
but
it's
giving
me
the
empty
list
in
case.
If
the
host
is.
A
You
have
to
use
the
cast
results
that
there
is
no
way
you
can.
You
can
get
the
facts
directly
from
the
host.
It's
just
too
slow.
It's
just
not
going
to
work.
We
went.
G
Through
it,
the
apr2
for
this,
which
I
ended
for
introducing
the
yeah
which
got
merged
to
for
this
idiom,
it
does
the
same
thing.
It
gets
the
effects
from
the
cache
itself,
but
yeah.
My
point
here
is
like,
if
I
call
if
I
expose
the
endpoint,
which
yeah,
which
returns
this
getherfx,
it
returns
an
empty
list
for
post,
which
is
in
the
maintenance
mode
currently
so
yeah
for
other
hosts
yeah,
it's
working
but
yeah,
I
yeah.
So
I
want
to
know
the
reason.
Why
is
that
the
case,
if.
A
B
A
G
I
can
yeah
I
can
do
that
so
basically,
I
should
create
against
this
like,
like
even
the
fetching
from
the
cash
it
doesn't
return
the
okay
yeah
yeah.
I
will
yeah.
G
A
Yeah
in
the
orchestrator
project,
please,
because
you
are
probably
the
the
first
one
really
consuming
that
effect,
except
from
a
few
edge
cases
we're
using
get
a
gather
effect
to
to
to
to
to
change
how
we
are
deploying
the
ingress
service.
I
think
it's
using
other
effects
or
it
used
to
use
get
effect.
I
don't
know
the
very
very
details
here,
but
you
are
probably
the
the
first
real
users
of
get
effects
in
general.
A
A
Perfect
have
a
nice
week
and
see
you
next
week,
bye.