►
From YouTube: Ceph Orchestrator Meeting 2022-05-17
Description
Join us weekly for the Ceph Orchestrator meeting: https://ceph.io/en/community/meetups
Ceph website: https://ceph.io
Ceph blog: https://ceph.io/en/news/blog/
Contribute to Ceph: https://ceph.io/en/developers/contribute/
What is Ceph: https://ceph.io/en/discover/
A
All
right,
if
we
can
start.
A
The
first
topic
in
here
like
it's
from
blaine,
something
about
rook
with
nfs.
I
want
to
talk
about
that.
B
Yes,
yeah,
I
guess
I
just
wanted
to
okay.
Let
me
find
the
the
meeting
notes
here.
B
Yeah,
I
think,
there's
a
bug
linked
there
in
sep
upstream
but
effectively.
I
I'm
pretty
sure
I
was
able
to
verify
it
as
a
a
regression
from
pacific
where,
if,
if
the
rook
orchestrator
module
is
enabled,
then
nfs
operations
will
fail
because
of
the
work
that
john
mulligan
did
to
to
allow
the
nfs
cli
to
work
without
the
work.
Orchestrator
module
enabled
that
that
pathway
still
is
like
still
is
operational,
and
that
is
the
current
workaround
to
get
the
nfs
cli
to
work
in
rick.
A
B
Correct,
I
I
think
in
in
pacific,
in
the
in
the
version
I
tested,
which
was
like
a
pre
a
version
like
shortly
before.
I
think
they
tagged
16.2.8.
B
It
was
working
both
with
it
enabled
and
with
it
disabled
in
quincy.
It
only
works
with
it
disabled
I
I
don't
particularly
have
a.
B
B
I'm
I'm
totally
fine.
If
the
only
code
path
that
exists
you
know
under
the
hood
is
the
one
where
the
orchestrator
is
not
enabled,
but
I
I
think
you
know,
I
think
that
decision
is
is
up
to
you
all
it's
just.
If
a
user
has
enabled
the
orchestrator
module,
the
commands
should
still
work
is
just
kind
of
the
bottom
line
for
me
or
a
further
project.
B
Yeah,
it
also
occurred
to
me
that,
because
it's
still
working
in
both
cases
in
pacific,
I
wondered
if
there
might
be
some
sort
of
pr
that
got
back
ported
to
pacific
and
missed
quincy.
C
My
question
is:
what
were
you
testing
the
same
version
of
rook
with
these
two
versions
of
stuff?
I
just
want
to
make
sure
that
change
wasn't
on
the
rook
side.
Okay,
great.
A
So
we
were
pretty
slow,
backboarding,
some
of
the
quincy
stuff.
I
think
around
this
time,
because
I
think
it
was
sort
of
close
to
the
release
and
we
were
trying
not
to
just
backboard
a
lot
of
things,
whereas
if
it
was
a
bit
farther
away,
we
still
were
so.
It's
definitely
possible
something
didn't
quite
get
into
the
1720
that
is
in
1628.
A
Trying
to
check,
because
we
had
pretty
much
all
of
our
quincy
backboards
we've
done
in
one
big
batch,
trying
to
look
there.
If
there's.
A
D
Yeah,
I
think
one
of
john's
manager,
nfspr,
might
not
have
been
backwarded
to
quincy,
but
I
mean
we
backboarded
everything
to
pacific,
so
that
might
be
it.
A
B
Yeah,
that's
all
I
really
had
on
that
on
that
topic.
So.
A
Okay,
yeah
we'll
we'll
go
through
and
we'll
try
to
figure
out.
If
there's
something
missing
it
gets
backboarded.
I
think
1721
is
actually
going
to
get
released
sort
of
soon,
because
there
was
a
couple
of
bugs
in
1720
that
make
them
want
to
release
a
new
one.
A
I
don't
know
exactly
when,
but
I
I
think
it
would
probably
be
like
I
guess
honestly
or
seventeen
two
one.
Okay,
so.
A
Yeah,
I
sort
of
want
to
get
it
so
it's
sort
of
almost
out
of
the
way
like
I
don't
want
to
keep
and
remind
myself
like.
Oh,
we
need
to
make
sure
we
backboard
this
yeah
manager,
nfs
thing
so
I'll.
Try
to
give.
C
A
A
look
later
today,
maybe
verify
with
john,
which
things
maybe
could
have
been
related
to
this
yeah.
C
A
Okay,
yeah,
I'm
not
sure
I
don't
think
too
much
has
happened.
I
managed
to
work.
I
have
to
check
with
with
the
I
guess
I
think
he's
the
one
who's
been
looking
at
that
that's
probably.
C
B
Yeah
the
well,
I
can
say
the
I
think,
john
pointed
out
to
me,
like
the
place
in
the
code
where
the
exception
occurs
is,
is
in
the
rook
module
and
it
is
related
to
some
changes
that
I
I
feel
like.
There's
a
weird
chicken
and
egg
problem
here.
There
were
changes
in
the
nfs
to
use
the
default
like
dot
nfs
tool,
instead
of
taking
pool
as
a
configuration
from
rook,
and
so
rook
removes
the
configuration
from
its
own
crs.
B
But
then
there
appears
to
still
be
this
part
of
code.
That's
looking
at
the
rook
crs
for
for
some
sort
of
sanity
check,
and
I
I
haven't
like
dug
into
the
code.
You
know
how
that
block
of
code
differs
between
pacific
and
quincy,
but
it
seems
likely
that
that
code
block
was
potentially
removed
for
pacific
or
that
it
was
changed
to.
B
Be
okay
with
that,
like
field
being
this
and
missing
from
the
resource
right
from
the
rook
resource.
C
A
B
Yeah
and
yeah
yeah
yeah,
I
mean,
if
you
have
any
questions
for
the
rook
side
of
things
you
can
always
reach
out
and
I'm
super
happy
to
I'll
try
to
clarify.
A
All
right,
it
seems
like
we
kind
of
know
what
we
need
to
do
with
this.
We
just
have
to
go
look
through
make
sure
I'm
just
going
to
manage
your
nfs
backports
and
check
if
there's
any
differences
in
the
rook
module.
A
All
right
move
on
then
I
see
there's
a
feature
request.
I
guess
for
the
nfs
stuff
we
talked
about
last
week,
yeah.
Okay,
I
see.
I
think
this
is
a
copy
paste
of
the
stuff
that
sage
had
in
his
mail
post
a
long
time
ago.
D
D
Yeah
I
went
ahead
and
created
this
feature
request
so
that
keep
track
of
it
and
whenever
the
firearm
team
has
time
to
work
on
it,
the
fadium
team
has
time
to
work
on
it.
They
can.
We
can
do
it
yeah,
so.
C
D
Discussed
last
last
week,
we
we
want
sephardium
to
come
up
with
a
solution,
a
non-ha
proxy
solution
to
provide
stable
virtual
ips
to
the
ganesha
demons.
So
that's
that
we
need
that,
for
the
openstack
use
case.
A
D
I
think
I'm
not
sure
exactly,
but
I
I
can.
I
can
figure
that
out,
but
there
is
a
timeline
for
the
upstream
open
stack
release.
So
that's
I
don't
know
the
release
of
open
openstack
release
cycle.
Well,
I
work
on
self.
So
I
I
I.
C
D
Get
you
the
right
number
release
number
and
the
approximate
release
date
so.
A
Okay,
thank
you.
Like
I
said
I
haven't
had
a
chance,
so
let's
look
at
it
yet.
So
I'm
not
sure
I
don't
have
too
much
to
say
about
it
right
now,.
D
Yeah,
I
was
also
wondering,
if
I
mean
if
rook
would
face
a
similar
issue
because
they
would
be
using.
I
suppose
the
english
controller
to
kubernetes
own
ingress
controller
to
you
know
provide
a
stable
virtual
ip
to
ganesha.
Is
that
right,
plane.
B
D
Yeah
and
in
the
refuse
case,
would
you
be
looking
at
a
similar
issue
where
you
expect
the
back
end
ganesha
server
to
see
the
client
id
is
that.
B
I
I
really
don't
know
I
I
don't
think
we've
talked
about
that
case
specifically
if
that
is
something
we
should
be
considering.
I'm
definitely
interested
in
kind
of
hearing
a
little
bit
more
about
that
and
why
that
might
be
something
that
we
need
to
need
to
plan
for.
D
Yeah,
okay,
I'll
just
briefly
want
to
go
through
this,
so
that
you
know
your.
I
mean
that
look
team
is
available
if,
if
the
back-end
server,
that
is
the
nfs
server
or
needs
to,
is
expected
to
enforce
restrictions
to
to
its
exports
based
on
client
id,
it
cannot
do
that.
D
It
cannot
do
that
if
there's
a
proxy
in
front
of
it,
because
the
back-end
server
just
sees
a
proxy's
ip
and
not
the
client
id.
So
I
mean
that's,
that's
the
basic
issue
and
there
are
ways
to
there
are
other
ways
to
get
around
it
like
setting
up
a
cha
proxy
in
a
transparent
mode
which
is
not
feasible
in
the
openstack
environment
and
the
other
approaches
have
ganesha
server.
The
backend
server
itself
support
what
they
call
a
proxy
protocol.
D
D
Do
in
their
their
own
protocol
called
live
in
dir
pc,
so
we
are
looking
into
that
too.
But
if
you
have
such
a
requirement
in
the
rook
use
case
where
the
client
ips
are
restricted
to
certain
exports
study
doctor,
the
exports
are
restricted
to
certain
client,
ips
and
then
you'd
pro.
A
That's
how
they
had
okay
good.
Well,
that's
a
I
guess,
just
in
you
or
make
sure
to
look
into
that
at
some
point,
and
you
know
what
we're
going
to
do
there.
A
A
Oh
I,
in
that
case
we
can
end,
I
guess
early
and
yeah.
I
guess
I'll
see
you
all
next
week.