►
From YouTube: Ceph Orchestrator Meeting 2022-07-12
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
there's
a
couple
that
I
kind
of
wanted
to
talk
about.
One
was
the
way
we
do.
Our
fb
filtering,
f,
idiom,
so
kind
of
what
we
have
is.
We
have
replaced
some
on
demon.
We
check
it
against
the
public
networks
and
then,
if
there
is
no
network
on
that
host
that
matches
the
public
network,
we
don't
allow
it,
but
somebody
was
asking
last
week
or
something
whether
we
could
be
a
bit
more
lenient
with
it.
B
A
And
so
it
has
like
an
exact
sort
of.
There
was
some
it's
a
slash,
24
network,
but
if
you
give
like
a
less
strict
one
like
if
you
do
like
the
slash
eight
that
technically
it
should
match,
it
will
still
deny
putting
on
this
host
because
it
wants
an
exact
match
for
the
public
network,
and
so
it
means
you
basically
have
to
put
in
like
a
set
of
different
networks
for
each
host.
A
A
I
don't
know
because
I
don't
know
when
this
has
been
around
for
a
while.
I
don't
know
if
it
was
originally
added,
then
if
there
was
a
reason
they
did
like
that
or
if
it
was
just
easier
to
implement
that
way.
At
the
time,
but.
C
Yeah,
I
definitely
think
there's
probably
okay
you're
pointing
out
like
if,
if
this
is
a
strict
subset,
that
kind
of
doesn't
make
sense
not
to
allow
it
but
yeah,
there's.
Also
the
general
question
like
I
was,
I
didn't
know
it
did
this,
and
it
makes
me
wonder
like.
Does
this
apply
to
everything
like,
for
example,
like
the
manager?
Does
the
manager
even
need
to
be
on
a
public
network
if
it
can
talk
to
other
things
like
stupid
questions
like
that
pop
into
my
head?
So
it's
only.
C
C
Yeah,
it
seems
reasonable
to
me,
but
I'm
the
noob,
so
there
could
be
other
subtle
issues,
but
I
think
it
would
be
worth
looking
into.
A
A
Yeah
yeah,
I
think
it's
worth
a
try.
We
had
to
sort
of
implement
something
that
can
check
if
a
subnet
these
falls
within
a
broader
subnet
or
not.
I
don't
think
we
have
anything
that
does
that
right
now,
details
are
anything
but.
C
A
C
A
C
And
worst
case
scenario
like,
like,
I
said,
if
you,
if
you
do
put
it
behind
like
a
specific
policy
that
someone
needs
to
opt
into,
it,
won't
affect
anybody
else
except
the
people
who
are
like.
Oh,
I
want
this.
Oh
okay,
you
have
this
thing
and
then,
if
it
turns
out
to
be
a
bad
idea,
you
can
just
drop
that
policy
or
deprecate
it.
That's.
A
So
the
other
one
I
have
in
here
is
something
a
bit
more
difficult.
Probably
so
somebody
else
was
asking
about
the
log
level
for
the
binary.
So,
as
you.
B
A
Know
already
we
deploy
the
binary
onto
the
host
and
then
we
run
the
commands
using
the
one
we
deploy.
Somebody
was
asking,
and
then
I
guess
all
those
logs
from
that
binary
end
up
in
this,
this
big
sort
of
setfiem.log
we
have
on
the
host,
I
guess
to
get
rotated
every
like
day
or
so.
A
Somebody
was
asking
about
the
log
level
in
there
because
we
already
allow
first
up
in
the
manager
module
you
can
set
the
log
level
and
everything
and
it'll
only
log
at
that
level,
but
for
anything
that
runs
on
the
with
the
binary
it
gets
put
in
sfdm.log,
it's
always
on
debug,
because
right
now
we
have
a
sort
of
hard-coded.
A
Of
or
like
I
don't
know,
logging,
config
and
so
they're
asking
if
there's
a
way
to
change
it,
which
I
don't
mean
right
now.
I
don't
think
that
there
is.
I
was
even
thinking
like
what
would
be
the
best
way
to
even
go
about
doing
that,
because
it
started
hardcoded
into
this
file
that
we're
deploying
on
the
host.
C
Yeah,
it's
it's
a
reasonable
request,
although
she
seemed
to
indicate
it
was.
The
concern
was
a
bit
more
about
like
the
log
filling
up,
but
I'm
like
the
log
rotation
is
already
there.
So
I
think
it's
more
of
a.
I
was
annoyed
by
the
debug
logging
and
then
turned
into
the
other
thing
to
make
it
sound.
You
know
more
have
a
better,
whatever
technical
reason.
Anyway.
I
think
it's
a
somewhat
reasonable
request
to
ask
for
configuration
of
belonging.
The
only
problem
is
it
I
almost
we're
like.
C
If
you
break
it,
you
own
it
it's
like.
I
don't
know
if
we
want
to
be
in
the
business
of
giving
lots
of
options,
but
I
don't
know
yeah
my
third,
my
first
thought
was
like:
oh
doesn't:
doesn't
the
logging
module
have
a
con
like
an
ini
config?
You
can
just
feed
it
if
you
want
full
control,
so
you
could
do
something
like
that.
If
the
config
file
exists,
you
can
use
it.
But
then
again
it's
like!
Oh
once
you
do
that
you
can
let
them
muck
it
up.
A
I
didn't
want
to
give
too
much
yeah,
it's
kind
of
like
the
log
that
one
level
field
yeah
and
I
wanted
to
be
able
to
set
that.
But
it's
just
hard
because
it's
in
the
file
that
we
then
deploy
in
the
host
and
then
so
we'd
have
to
like,
I
don't
know,
deploy
it
with
it,
saying
like
inspo
or
saying
debug
or
something
or
we'd
have
to
have
it
as
some
sort
of
command,
but
they'd
have
to
pass
that
flag
setting
then
or
yeah
it's
like
a
fire
system.
C
C
Environment
variable
excuse
me,
so
if
we
made
an
environment
variable
people
could
stick
the
environment
variable,
you
know
in
their
bash
rc
or
whatever,
and
that
would
be
an
option
and
that
way
it
wouldn't
have
to.
I
mean
I'd.
Do
both
honestly
if
I
was
going
to
do
an
environment
variable
and
also
do
a
cli
option
but
yeah
the
other
question.
C
A
A
D
And
that
one
gets
a
little
bit
more
obnoxious
because
I
think
a
lot
of
the
commands
don't
run
within
the
context
of
the
show.
So
the
environment
solution
wouldn't
work,
and
then
we
would
also
quickly
run
into
the
request
where
someone
will
say
well.
I
want
debug
logging
on
this
specific
host,
but
not
on
all
the
rest
and
then
the
complexity
of
configuration
and
the
flagging
gross.
A
D
It's
feasible,
I
guess
I
guess
the
other
question
too
is
there's
two
halves
to
this.
So
traditionally,
like
this
configuration
is
anything
that's
info
level
or
greater
is
to
the
console
and
anything
lesser
than
that
like
debug,
is
to
this
log
file.
A
D
A
D
A
lot
of
confusion,
yeah
and
there's
also
the
verbose
option,
but
also
they
put
steva
glugging
out
on
the
console.
So.
D
A
I
think
it
would
have
to
just
be
the
the
log
file
one.
We
need
the
console.
We
need
to
sort
of
manage
the
console
output
because
we
care
about
that,
for
our
comments
like
a
verbose
flag
is
fine,
because
if
you
manually
run
something
you
can
provide
that
and
it's
good
for
you,
but
we
need
when
we're
running
after
the
manager
module,
it
has
to
be
an
info
level.
Otherwise,.
B
A
I
really
wouldn't
want
to
let
them
mess
with
the
console
one
outside
of
if
they're
manually
running
something
in
case.
We
could,
you
know,
there's
a
flag,
they
could
maybe
do
for
that.
I
think
you
said
you
already
that
verbose
fog
does
that
I
just.
I
would
like
to
be
able
to
cut
them
to
have
like
a
global
customization.
Maybe
for
the
log
level,
one
log
file,
one.
A
That
right
now,
they
can't
control
at
all,
and
it's
not
actually
irrelevant
to
us
what
gets
put
in
that
log
file
like
we
don't
actually
care
as
part
of
our
operation.
It's
really
just
whatever
the
user
wants
to
look
at.
C
B
D
Do
we
have
any
idea
of
how
much
more
verbose?
This
is,
because
the
one
that
I
find
the
most
useful
is
where
we
log
the
command
that
we're
executing,
because
the
problem
is,
is
if
we
start
omitting
those
from
the
log
file,
then
the
context
isn't
terribly
helpful,
like
the
log
file
just
has
almost
random
messages.
We
don't
know
what
the
context
was.
D
B
C
A
C
Yeah
I
mean
the
other
thing.
Is
we
can
go
back
and
ask
so
in
particular?
Is
there
a
particular
message
because
I
think
of
the
bc
stream
z
is
something
like
this
particularly
spammy
message
has
already
been
removed,
but
the
original
request,
which
was
to
make
the
vlogging
configurable,
was
never
finished
or
never
done,
which
is
true.
C
C
A
I've
I
was
looking
at
one
actually.
I
was
looking
at
multiple
things
related
to
this
somebody
who
had
some
problem.
I
haven't
been
able
to
recreate
it
yet
with
that
that
file
rotating
something
with
the
permissions.
Let
me
see
if
I
can
find
an
error
there,
but
I
don't
know
if
I
just
think
I
was
on
an
older
version
or
not.
C
A
A
This
would
be
something
with
our
log
rotate
config
for
that,
according
to
what
I
was
looking
at
with
this,
some
people
were
getting
problems
with
the
permissions
on
that
directory
filezilla,
and
I
was
just
trying
to
bootstrap
with
like
a
most
recent
version
and
also
some
like
it
was
like
1627.
I
think
I
tried
and
I
didn't
have
any
issues
with
it.
I
was
able
to
just
force
the
log
rotates
and
it
just
worked.
Okay.
A
C
A
Yeah,
that's
why
I
was
picking
one
in
the
middle,
because
what
we
kind
of
want
is
we
almost
want
like
info,
plus
these
special
messages
to
be
in
the
logs
or
in
the
in
the
log
files.
We
don't
want
them
in
the
console,
the
console's
at
info
level,
so
they.
A
Okay,
so
if
they
don't
want
debug
logs,
but
they
still
want
to
have
useful
log
files,
you
need
something
that's
sort
of
in
the
middle
of
that
yeah.
C
B
C
A
A
A
A
D
Unless
something
happens
that
tends
to
block
this,
so
the
one
of
the
really
severe
nasty
bugs
that
I've
run
into
that
I
don't
like
is
occasionally
set
volume
will
hang
in
an
unadoptable
state
like
trying
to
do
an
inventory
and
when
that
happens
then
we'll
end
up
with
another
sephadm
instance
continually
spooling
on
this
lock
and
that's
the
easiest
way
to
find
that.
A
Yeah
we've,
I
know
guillaume,
has
been
looking
at
that
sort
of
stuff
for
a
while.
You
keep
trying
to
figure
out
why,
each
time
it
happens,
why
it's
happening,
but
usually
we
find
it
with
just
being
that
there's
a
it's.
A
volume
process
in
in
these
states
always
seems
to
have
enough
happening.
A
When
they
were
sort
of
thinking,
though,
is
that
like
so
what
people
could
do
is
they
could
just
leave
them
on
like
some
info
level
or
whatever?
Then,
if
this
something
is
going
wrong,
they
could
you
know,
set
it
to
debug
and
then
do
a
manager
failover
and
have
it
you
know,
try
again
do
what
happens.
A
A
At
least
you
could
still
find
it
if
just
change
the
log
level
back.
C
I
think
we're
getting
away
from
the
topic
of
strictly
of
the
logging,
but
my
next
question
is:
I
haven't
looked
at
the
lock
myself,
but
if
there's
a
timeout
on
the
lock,
you
should
log
that
at
a
like
a
warning
level
or
something
which
would
be
you
know,
still
still
indicate
the
fat
problem,
but
not
necessarily
spammy.
A
There
is
a
timeout
option:
oh
it
even
logs
if
you
go
over
the
timeout
at
warning
level,
but
I
don't
think
you
have
to
explicitly
pass
a
to
use
that
which
we
don't
pass
in
from
the
manager
module,
which
we
showed,
though,
there's
actually
another
tracker
open
for
doing
that.
A
D
So,
just
given
the
complexity
of
like
configuring
all
these
log
levels,
I
almost
learned
towards
what
john
had
suggested,
try
and
find
that
really
noisy
line
and
omit
it
like
maybe
we're
outputting
something
from
standard
out
on
some
command.
That's
excessively
verbose.
A
I
think
I
did
also
see
that
it
was
like
the
odd
man
run
command.
We
end
up
generating
for
like
demons.
I
think
we're
getting
put
in
there
at
debug
level,
I'm
getting
I'm
seeing
some
stuff
about
here.
I
happen
to
have
one
you
just.
D
Because
what
we
might
do
is
like
change
the
debug
level
for
this
f8m
command.
That
kind
of
indicates
the
context
of
this
command
that
raise
that
to
a
different
log
level
and
then
suppress
everything.
That's
debug
that
might
be.
A
A
C
C
I
have
so
many
in
the
mail,
so
I'm
trying
to
find
the
what's
happened
to
my.
B
A
So
I
just
printed
out
right
under
the
topic
that
that's
just
I
looked
at
my
blog
and
it's
just
that
over
and
over
and
over
again.
A
A
D
C
A
That
one's
almost
as
bad
as
the
the
other
facts,
one.
D
A
As
much
these
are
that
I
even
look
at
that
tracker.
That
seems
like
it's.
What
a
sebastian
tried
to
do
on
the
original
time.
This
was
okay
right.
You
just
remove
some
stuff
rather
than
trying
to
the
logging
level.
C
Yeah,
I
think
that's
actually
a
reasonable
approach.
So
again,
we'll
probably
want
to
explain
things
in
the
tracker
and
the
and
the
pc,
but
be
like
yeah.
Sorry,
we're
not
going
to
give
you
that,
but
what
we
will
give
you
instead
yeah
and
then
close
the
upstream
tracker.
So
people
don't
think
we're
going
to.
A
This
is
kind
of
unusable
at
this
point
right,
because
these,
like
so
stuff
like
inventories,
every
like
10
minutes,
gathered
back
every
one
minute
and
they're,
just
printing
out
a
ton
of
output
yeah.
Maybe
that
silent
thing
you
were
talking
about
is
actually
what
we
want,
like
only
print
this
out,
if
we
say
verbose
exclusively,
otherwise,
don't
log
this.
C
Right
we
we
did.
I
did
something
similar
in
ketty,
where
it
was
logging,
some
lvm
commands
that
had
a
lot
of
output
and,
oh
I'm
sorry
it
was
cluster
status
or
something
like
that
long
story
short.
It
doesn't
matter
what
command
it
was.
The
the
important
thing
is,
the
logs
are
just
full
of
this
stuff,
and
so
we
added
a
flag
to
the
function.
To
basically
say
you
know
what
this
one
don't
bother:
logging,
the
output.
C
It's
it's
rarely
interesting,
and
if
there
is
an
error,
then
we
can
log
the
output,
which
I
think
this
looks
like
based
on
this
call
verbosity
flag.
There
is
some
support
for
this
stuff,
but
maybe
it
wasn't
plumbed
all
the
way
through
to
all
the
different
all
sides.
I'm
also
seeing
that
there's
like
a
nested,
async
defs
inside
of
a
cult,
I'm
like
a
little
bit
much
for
my
puny
brain
in
the
middle
of
the
meeting,
but
it
looks
like
the
intent
is
there,
but
maybe
it
wasn't
thoroughly
implemented.
A
I
don't
remember
what
you're
talking
about,
I
think
there's
something
kefu
added
in,
and
I
remember
looking
at
it
when
we
had
some
other
bug
a
long
time
ago,
with
the
with
the
lock
stuff
it
looked
like
something
was
stuck
and
it
ended
up
being
like
weird
about
the
way
we
do.
Our
sub
processes
was
hanging,
but
I
never
really
figured
out
how
that
stuff
works
or
why
it
was
broken.
A
Yeah,
I
guess
maybe
that's
where
we
want
to
start.
Is
we
just
want
to
clean
up
the
debug
logs?
You
need
to
move
all
this
spammy
stuff
to
silent
level
and
then
maybe
once
they're
actually
usable,
you
can
actually
do
something
yeah,
even
like
the
ls
as
well,
the
one
that
looks
like
the
demons.
This
one
also
is
spamming.
The
logs
every
single
regular
command.
We
have
spams
the
logs.
C
Yeah,
so
it
would
be
like
worth
locking
all
the
output
like
say
on
a
parse
failure
like
if
we
were
parsing
this
output
and
it
wasn't
incur
it
wasn't.
You
know
we
couldn't
parse
it.
Then
it's
totally
legit
to
like
hey.
Well,
how
did
the
parser
go
wrong?
I
want
to
write
a
unit
test
to
fix
this
or
whatever
but
yeah.
If
it's
it's
spammy,
you
know
probably
not
log
it
if
the
command,
no
errors
and
everything's
hunky-dory.
B
A
A
All
right
that
was
it
for
topics
that
I
had
and
didn't
have
anything
else.
You
want
to
talk
at
all
about
the
binary
refactoring
or
any
of
that.
C
I
can
give
a
quick
status,
I
don't
have
a
ton
to
say
I'm
I'm
happy
to
say
that
I
do
think
the
recent
changes
are
improving
the
situation.
I
have
a
lot
of
tabs
there
with
me.
C
Yes,
I'm
doing
builds
right
now,
because
the
failures
in
this
group-
so
I
believe
this-
has
solved
some
of
the
issues
when
we're
intentionally
deploying
an
older
version
and
then
doing
upgrade
tests
so
especially
since
mike's
here
I
don't
know
if
you're
aware,
yet
what
I
ended
up
doing
is
taking
the
work
you
did
to
change
the
methodology,
tasks,
stuff
adm
stuff.
C
C
A
C
B
C
Once
that's
done
I'll,
probably
start
pestering.
You
folks
tell
me
if
there
are
other
tweets
that
should
be
run
in
addition
to
the
regular
cepheum
orchestration
ones,
just
to
kind
of
suss
out
any
weirdness
that
might
be
lingering
in
other.
C
C
A
A
So
it's
always
been
the
fs
said
the
radio
suite
that
we
break
when
we
mess
with
stuff,
probably
those
two
all
right.
Maybe
the
up
there
is
an
upgrade
suite
itself
and
it's
like
not
safety,
yamaha.
Definitely.
A
C
Those
assuming
that
all
goes
well
I'll,
probably
just
quickly
check
on
the
whether
we
like
squash
these
patches
together
and
then
hopefully
we
can
be
done
with
this
here
and
then
we'll
move
on
to
other
things
like
documenting.
It.
B
C
Yeah,
so
he
he
made
a
comment
on
a
log
line
there's,
but
there
aren't
any.
There
are
no
changes
to
the
documentation,
so
I
think
we
probably
want
to
mention
things
like.
This
is
how
you
get
stuff
adm
now
versus
like
I
haven't,
looked
for
it
yet
so
that's
probably
on
that's
on
me
to
actually
go
through
the
documentation
and
find
out.
C
A
C
A
Okay,
yeah
sounds
good,
I
don't
have
any
other
topics.
Does
anyone
have
anything
else
they
want
to
talk
about
in
here.
A
All
right,
I
also
take
a
look
at
the
log
stuff,
I
guess
later
other
than
that
it
seems
like
oh
yeah
and
then
the
ip.
So
I
kind
of
probably
even
talked
about
that,
but
it
sounds
like
we're
good
to
sort
of
start
moving
forward
enough.
A
Case
you
can
end
here
and
I
think,
given
who's
in
here,
I'd
probably
see
all
of
you
on
thursday
in
the
stand
up.