►
From YouTube: Ceph Orchestrator 2023-01-24
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/contrib...
What is Ceph: https://ceph.io/en/discover/
A
Let
me
see
someone
does
have
comments
on
it
so
a
couple
weeks
ago,
or
maybe
it
was
just
last
week
there
was
an
issue
with
movie
and
also
some
detail,
an
issue
with
our
setup
of
systemd,
where,
if
the
service
timed
out
starting
up
and
the
last
time
out,
it
would
not
kill
the
container
process
because
we
have
kill
mode
on
none
for
other
reasons,
and
so
redo
actually
looked
into
that
later
last
week,
and
he
had
this
PR
here.
A
Basically,
we
found
out
is
that
stop
doesn't
run
when
at
times
are
like
that,
because
it
assumes
it's
already
stopped
because
it
failed
starting
up
and
so
Post
stop
actually
does
run
and
so
did
I
the
wrong
PR
for
that
one,
okay
and
second,
if
he
likes
the
rock
I
like
the
same
PR
in
both
of
those.
Let
me
find
the
actual
Post
stop.
A
Yeah
there
we
go
yeah
so
anyway.
The
idea
is
that
we
do
it
in
post-op,
basically,
the
same
thing
we
do
and
stop
right
now
and
that
way
the
container
will
always
be
taken
down.
If
the
timeout
happens,
we
can't
get
left
in
some
state
where
the
process
itself
is
running,
but
the
system
D
unit
considers
it
as
failed,
which
was.
C
A
So
basically,
I
only
brought
this
up
again
because
I
know
we
talked
about
it
last
week
and
I
think
we
had
said:
we've
already
discussed
it
a
bit
more
and
the
solution
seems
like
it.
It
works,
but
I
wrote,
I
was
testing
with
like
it
went
pretty
well,
so
I
just
wanted
to
give
people's
opinions
on
it
and
see
if
anyone
has
anything
to
get
sick.
Otherwise
it's
already
passatology
and
stuff,
so
I
was
am
I
going
to
disparge
it.
Otherwise.
C
A
A
All
right
and
the
other
one
was
the
problem
brought
up
two
days
ago
yesterday
in
the
stand-up,
the
host.containers
dot,
internal
Etsy,
host
entry,
causing
iSCSI
issues
and
we're
looking
for
potential
solutions
for
that,
and
so
I
have
that
pull
request
there
that
mounts
the
hosts
version
of
the
seos,
which
seems
to
at
least
work
and
we're
going
to
look
into
whether
there
were
some
alternative,
Solutions
I
didn't
really
have
much
time
to
look.
C
You
find
the
unmute
button
yeah,
so
part
of
the
fact
that
does
name
resolution
is
not
just
in
raise
this
enough
when
the
python
call
that
ultimately,
it's
socket
get
fqd.
C
You
can
see
I'm
really
prepared
for
this
meeting.
The
the
socket
call
that
we're
using
that
actually
gets.
The
name
of
the
host
goes
through
a
stack
component
in
G
LED
Z.
It's
doing
it
does
more
than
just
look
at
Etsy
hosts.
That
also
will
look
at
Betsy
nsswitch.com.
This
is
a
whole
Library,
sub
library
and
she'll
see
that
handles
this.
C
One
thing
that
I
was
looking
into
is:
there
is
a
my
hostname
plugin
in
the
NS
switch
file
and
the
order
it's
in
in
our
containers
today
is
not
the
recommended
order.
According
to
the
man
page.
A
C
A
All
right,
do
you
know
how
I
was
getting
set
right
now?
Is
it
just
brilliant
the
host
version
of
it,
or
is
it
writing.
C
Its
own
I
believe
it's
just
baked
into
the
Container
image.
I,
don't
think
it's
one
of
the
files
that
gets
tweaked
live
by
the
runtime,
not
like
Etsy
host,
which
gets
updated
dynamically
I'd
have
to
double
check
but
I'm
like
80
certain
that
whatever
your
at
CNS
switch
is
in
your
in
your
image
is
what
it
uses.
C
C
A
C
Yeah
it
was,
it
should
just
be
the
I
just
grabbed
the
main
page.
It
was
off
of
Arch,
but
I
think
that
it's
just
the
man
page.
C
Yes,
which
that's
the
only
line
we
care
about
enough
that
file
today,
most
most
other
things
the
lines,
the
other
lines,
don't
matter
in
containers
unless
you're
going
to
maybe
do
user
impersonation,
but
that's
more
like
maybe
an
SMB
or
NFS
which
I
won't
get
into
today.
A
C
A
Yeah
I'm
also
thinking
of
like
how
to
actually
do
it
because,
with
like
the
bouncing
seos
thing,
it's
just
this
amount.
A
C
Yes,
so
that's
one
way
to
do
it
to
do
it
in
the
docker
file
container
file
just
have
our
own
copy
and
and
put
it
in
there.
Alternatively,
we
could
do
Dynamic
stuff,
but
I
don't
know
of
how
necessary
that
is
so
yeah
I
think
it's
going
to
take
a
little
bit
more
discussion
before
we
settle
on
this
as
a
as
the
solution,
but
I
think
it's
at
least
worth
pursuing.
C
A
We
think
it's
worth
using
the
mounting
at
theosis
will
work
around.
In
the
meantime,
he
thinks
this
is
going
to
be
difficult,
so
that
right
now
I
think
is
because
he
is
actually
just
sort
of
broken
use.
Podman
4.1,
there's
just
an
unremovable
Gateway.
It
gets
place.
C
Yeah
I'm
not
objecting
to
that
as
a
workaround
I'm,
a
little
nervous
about
other
potential
things,
but
it's
probably
fine.
A
Okay,
maybe
we'll
just
do
that
for
now,
this
thing
loads
will
need
more
investigation
and
we
need
like
a
nice
way
to
do.
This,
would
have
to
be
better
if
we
could
just
get
this
into
the
the
build
process
thing.
If.
C
Yeah
I
mean
the
good
thing
about
the
workaround.
Is
it
doesn't
need
any
other
dealings
outside
of
the
scope
of
cyphenium,
so
yeah
I
just
would
like
to
double
check.
There
aren't
any
other
like
weird
side
effects.
You
know.
The
other
thing
is
like
I'm
running
a
cluster
locally
here
and
it's
doing
it's
putting
the
thing
into
Etsy
hosts
and
like
I'm,
not
seeing
any
breakage
but
then
again,
I'm
not
running
I,
suggesting
either
it's
interesting
that
most
of
the
staff
containers
don't
care.
C
A
C
A
Yeah,
so
we
definitely
try
that
and
see
if
it
works,
I
might
yeah.
I
might
just
go
forward,
try
to
get
the
mount
and
Yeti
host
in
the
meantime,
and
then
we
do
like
a
proper
investigation.
The.
C
C
A
C
A
C
A
Could
maybe
come
up
with
some
like
hacky
way
of
modifying
it
as
well
at
least
test
it
in
tautology?
You
see
if
it
breaks
anything
there,
and
then
you
try
to
do
like
a
proper
thing
after
Maybe.
A
No
did
that
to
be
sort
of
a
I'm
gonna
put
the
doc
what
the
line
is.
We
have
to.
C
C
A
C
After
well,
I
think
the
problem
with
Essie
hosts
is
I,
believe
it
it's
I
think
it
detects
it
as
the
host
name
using
we're
using
host
networking
on
these
containers
right
we're
not
using
user
networking.
So
I
think
what
happens.
Is
it
sees
that
the
IP
address
matches
its
own
IP
address
and
goes
well?
C
That
must
be
my
host
name,
even
though
it's
really
there
for
the
container
to
talk
to
the
host
when
you're
using
non-host
networking
like
I,
can't
even
remember
the
thing's
name
right,
but
the
a
little
sub
thing
for
user
mode
containers.
A
A
A
Yeah
anyway,
well,
let's
look
at
this
a
bit
more.
This
is
a
pretty
complex
thing.
A
Well,
the
original
point
is
I.
Guess
we'll
all
go
forward
with
mounting
the
seos
for
now
as
a
workaround
that
seems
to
work
and
physically.
We
want
to
sort
of
use
the
host
name
resolution
anyway
here
and
then
we
can
look
more
at
maybe
if
this
is
a
better
permanent
solution
and
is
more
resistant
to
any
modifications
to
seos
in
the
future.
This
kind
of
switch,
which
doesn't
seem
to
be
getting
touched
by
anything.
A
Add
since
Roto
I
see
you're
here
now,
we
did
I
did
briefly
bring
up
the
post-op
actions
PR,
but
it
seems
like
nobody
had.
Okay,
any
strong
opinions
on
it,
so
we'll
probably
go
forward
with
that
one
as
well.
A
A
Okay,
that's
all
I
had
for
topics
which
is
those
two
things.
Does
anyone
have
any
other
things
they
want
to
talk
about
in
here,
foreign.
C
C
Yeah
I'll
link
this
in
chat.
It's
just
kind
of
one
of
these.
What
I'm
wondering
is,
if
we
can
reuse
functions
from
async
SSH
inside
of
the
the
path
here
where
we
do
setting
the
keys,
because
right
now
you
can
basically
feed
it.
Garbage
and
it'll,
take
them
and
store
them,
and
then
you
try
and
use
them
later
and
things
will
fail.
So
it
might
be
nice.
You
know
we
could
even
optionally
hey.
B
Robert
I
remember
fixing
some
similar
issue
with
another
tracker,
basically
to
check
if
scch
is
working
properly,
but
it
was
when
cfdm
is
being
provided
by
by
user.
C
Yeah
I,
don't
know,
I
didn't
I,
didn't
see
a
duplicate,
but
I
also
don't
know
how
if
Kyle
really
searched
for
it.
You
know
a
duplicate
before
it's
a
long,
long
story
short.
What
I'm
wondering
is
I,
don't
know
how
much
validation
we
can
do
in
that
path.
But
if
we
can
do
validation,
it
might
be
nice
to
do
some
sanity
checking
because
otherwise
I
always
you
know
it's
always
difficult.
C
B
C
A
The
problem
is
just
that
I
think
he's
setting
it
with
directly
setting
the
config
keystore
yeah
I
think
we
have
a
function
for
actually
setting
those.
B
A
A
C
Takes
a
string,
you
know,
that's
a
really
good
point,
so
what
I
can
I
can
talk
to
you
over
chat?
Later
I
can
reply
to
him
and
say:
hey.
You
know
what
there's
actually
a
better
way
to
do
this
action.
He
might
be
doing
that
out
of
habit
for
whatever
reason
it.
C
Heck
we
could
even
print
you
can
even
print
a
warning
based
on
the
the
key,
the
not
the
key,
but
the
key,
the
the
dictionary
key
key
of
the
name
and
say:
hey,
there's
a
better
way
to
do
this.
C
Okay,
yeah
so
Adam,
if
you
want
to
you,
can
either
reply
directly
to
that
tracker
or
I
can
do
it
later,
but
I
have
to
drop
because
I
have
another
meeting
to
join.
C
Hey
there,
all
right,
I'm
glad
I
actually
brought
this
up,
because
it
means
that
there's
a
something
better
than
writing
code,
which
is
code
that
already
exists.
A
Yeah
as
you
leave
yeah,
so
we
have
the
documentation
generally
on
a
safe
configuration.
That
includes
the
the
proof
key
and
it
says.
C
C
A
B
B
Anyway,
this
will
set
the
key
directly
in
the
config
manager
store,
so.
A
Store
directly
documents
that
we
have
this
function:
that
that
already.
B
A
So
I'll
look
through
the
documentation
and
check
what
we
have
but
I
guess.
We
definitely
I'll
replying
the
tracker
saying
that
we
have
the
function
and
I'll
check
if
a
more
recent
documentation,
if
there's
anything
and
if
not
we'll-
have
to
add
that
to
the
sh
configuration
documentation,
section
yeah,
but
we
do
have
one
and
it
does
do
some
validation.
I,
don't
know
if
the
validation,
how
good
the
validation
is,
but
I
I
would
at
least
catch
garbage.
Can
you
give
it
all
to
test
that.
C
A
C
Right
well,
I
gotta
go!
Thank
you
for
discussing
this.
We'll
I'll
follow
up
on
it
later.
I
just
thought.
It
was
interesting
because
it
came
out
of
a
side
conversation
we
had
over
chat
and
I
was
kind
of
interested
in
the
topic.
A
It
looks
like
it
tries
an
actual
connection
to
a
host
by
the
way,
with
the
validation.
A
And
we
use
the
Aperture,
we
use
this
function,
this
stat
vdm
that
proof
key.
Let
me
check
that.
A
B
A
Yeah,
so
we
do
that
verification.
So
as
long
as
the
verification
Works
in
there
are,
we
verify
any
keypads
of
bootstrap
it's
just
if
you
modify
it
in
the
middle,
because
if
you
use
the
function,
it's
just
they'll.
Do
it
just
because
again
that
they
directly
didn't
do
the
config
key
store,
which
we
don't
control
yeah
so
and
against.
Mostly
the
documentation
or
awareness
sort
of
thing.
B
Probably
I
mean
if
we
do
provide
discipline
or
we
have
to
move
this
from
the
documentation.
A
B
Yeah
or
you
have
to
to
update
the
documentation
as
part
of
this
tracker.