►
From YouTube: Ceph Orchestrator Meeting 2022-09-20
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,
let's
start.
B
There's
only
really
one
topic
for
today,
which
is
the
compiles
of
video
stuff,
got
merch
yesterday,
so
especially
I
guess
in
the
source
tree,
and
you
can
do
stuff
with
it.
B
This
is
why
today,
I
wanted
to
go
back
over
this
other
path.
We
made
a
while
ago
make
sure
that
we
still
agree
with
sort
of
at
least
the
next
steps.
A
B
B
I
think
the
idea
what
we're
discussing
was.
We
are
going
to
check
the
coverage
of
the
video
dot
PI
right
now.
This
is
the
one
file
and
we're
going
to
try
to
increase
fabric
up
to
maybe
like
80
percent
I
guess
it
depends
we'll
have
to
see
how
it
looks.
I
guess
as
we
get
higher
percentages
and
then
once
we're
there,
then
maybe
we
start
breaking
things
up
once
we're
confident
that
we
can
test
to
make
sure
that
breaking
up
hasn't
ruined
anything.
C
So
so
on
this
one
and
I'm
willing
to
do
this,
maybe
today
I
was
thinking.
Maybe
we
had
pie
test
coverage
to
that
and
that'll
give
us
like
a
coverage
report
where
we'll
know
you
know
what
items
are
maybe
possibly
missing
and
then,
when
we
have
that
list,
perhaps
we
can
go
through
that
list
and
identify
which
things
are
critical,
that
we
really
want
tests
around
and
then
try
and
meet
that
80,
Milestone
or
whatever?
That
might
be.
That
sounds.
D
Yeah
I
think
we're
on
the
same
wavelength
I.
If
you
look
at
the
pr
I
shared
it's
a
a
non-enabled
test
environment
for
Cox
that'll
generate
the
coverage
report
feel
free
to
suggest
anything
on
the
pr,
if
you
think
it
needs
a
tweak
or
two,
but
I
was
thinking.
We
could
kind
of
merge
that
I'd
shared
this
one-on-one
with
Adam
yesterday,
just
to
make
sure
that
it
mostly
worked.
D
C
D
C
Do
we
have
a
list
of
which
what's
missing,
though,
like
which
routines
or
CLI
man.
B
I
haven't
gone
through
it
formally,
we
sort
of
informally
looked
over
it
once
yesterday
on
the
coverage
report
showed
and
a
lot
of
it
is
like
the
classes
for
the
services
like
say,
if
there's
like
an
NFS
class
in
the
binary
where
all
the
stuff
for
there,
where
it's
just
like
you,
know
the
files
and
does
all
that
stuff,
a
lot
of
those
aren't
tested
at
all
our
list,
demons
isn't
tested
at
all.
B
Removing
the
cluster
isn't
testing
at
all
like
arm
cluster,
like
that,
so
I,
don't
even
compile
the
full
list
yet
I
didn't
like
formally
do
it,
but
we
could
see
I,
guess
some
of
the
the
big
things
that
we
testing.
That
is
a
lot
of
the
service,
specific
stuff
that
like
config
stuff
for
those
services
and
then
like
some
of
the
operations
that
are
sort
of
hard
to
test
in,
like
an
RM
cluster
of
those
demons.
D
So
the
idea
that
just
popped
to
my
head
is
that
maybe
what
we
could
do
is
add
to
a
list
right,
maybe
in
The
Ether
pad
like
create
a
new
subsection
of
like
top
level
like
the
class
name
or
a
top
level
function,
name
that
is
lacking
coverage,
and
then
we
can
come
back
either
in
the
stand
up
or
the
next
weekly
and
kind
of
they
divvy
it
up
amongst
like
three
or
four
people
or
two
people
or
whatever.
But
you
know
what
I'm
getting
at
that
way.
C
And
I
think
the
approach
I
had
kind
of
followed
with
the
unit
testing,
because
it's
kind
of
you
know
which
unit
did
we
break
it
down
to
I?
Think
I
was
kind
of
mostly
walking
down
all
of
the
CLI
functions
from
the
usage
and
so
I
think
I
got
part
way
through
that
I
did
like
a
bunch
of
bootstrap
tests
and
a
few
others.
So.
C
A
C
B
B
B
But
like
when
we
get
down
to
here,
like
this
class,
is
like
very
lightly
festive
like
pretty
much.
This
whole
thing
is
red,
so
that's
an
example
and
I
remember
right
when
I
think
it
was
RM.
Cluster
was
the
one
of
the
big
ones
that.
B
Arm
demon
as
well,
but,
like
all
this
stuff
is
untested,
RM
demon
arm
cluster
is
the
same
way.
All
the
adoption
stuff
is
not
tested
like
this
whole
area
down
here
is
missing.
B
A
lot
of
that
and
I
remember
with
I
think
this
is
listings
here.
This
is
another
one
of
the
big
ones.
That's
really
important,
but
it's
not
really
tested
at
all
outside
of
this
very
initial
part.
Basically,
once
you
get
down
to
the
part
where
it
actually
starts
looking
into
the
directories,
there's
nothing
any
of
that
either.
B
A
A
B
So,
like
the
agent
list,
demons
is
a
big
one.
All
the
adoption
stuff
RM
cluster
any
of
these
like
if
any
of
these
whole
trendline
functions
that
aren't
tested
at
all,
I,
really
need
to
get
something.
D
B
Yeah,
that's
like
the
way,
the
or
like
the
argument.
Parsing
at
the
bottom
is
set
up
where
they
all.
D
C
Yeah
that
was
kind
of
the
path
I
was
following
prior
for
this,
so
so
pretty
much
anything
that
has
like
a
you
know
manned
under
bar
Foo.
You
know
that
maybe
those
top
level
functions
so
yeah
and
then
I
think
I.
C
Oh
sorry,
I
I
think
I
selectively,
because
some
of
them
are
a
little
bit
different.
So
I
selectively
choose
different
routines
as
they
made
sense,
but
we're
outside
of
the
commands,
but
generally
the
yeah.
B
B
Yeah
I
guess
we
sort
of
have
to
have
a
list
instead
of
sort
of
Fuel
tickets.
I
don't
know
if
I
don't
want
to
wait
until
next
week
to
start
to
start
doing
any
of
this
or
be
maybe
just
like
put
it.
Maybe
what
you
want
to
work
on
one
of
them
we
can
just
go
put
it
into
this
other
pad
or
into
the
Deputy
I'm
refactoring
other
pad
just
like
write
down
the
highway.
D
B
B
Is
there
anything
else
to
say
about
that
testing
stuff?
Do
we
need
to
go
over
more.
D
D
Weird
stuff
that
might
be
like
really
hard
to
write
a
good
test
case
for
and
the
other
issue
is
maybe
you
could,
but
it
would
take
like
so
many
mocks
that
it
wouldn't
be
worth
it.
You're,
not
testing
anything
anymore,
so
yeah,
some
somewhere
between
80
and
90
I
think
would
be
ideal,
but
I
don't
want
to
set
that
as
like
a
hard
goal.
Yet.
B
B
I'm
not
super
tied
to
a
percentage
really
either
that's
fair
or
the
functionality
stuff.
I
don't
have
to
worry
about
individual
lines.
If
you
guys
watching
the
environment
plus
series
test
as
well,
though,.
A
All
right,
yeah
I,
jumped
a
little
bit
late,
because
I
didn't
see
that
that
time
is
this
presented,
something
that
we
have
to
achieve
before
a
start
refactoring.
Is
it
so.
B
What
we
decided
is
because
we're
worried
about
breaking
things
when
we
move
all
the
files
around
yeah
is
you
want
to
increase
the
test
coverage
generally,
and
we
were
just
talking
about
what
the
percentage
maybe
should
be,
that
we
should
aim
for
I.
Think
we
were
sort
of
saying.
Maybe
the
percentage
itself
shouldn't
be
the
goal.
B
It
should
just
be
like
list
out
the
big
items
that
are
not
tested
yet
and
make
sure
all
of
them
have
some
amount
of
testing
and
then,
if
like
say,
there's
like
random
individual
lines
all
over
the
place
and
maybe
that's
10
of
it-
isn't
tested.
Because
of
that,
then
that's
fine.
As
far
as
the
big
things
you
want
to
have
unit
tests.
B
For
so
what
we're
going
to
do
is
in
this
setting
refactoring
pad
I'll,
just
put
it
in
the
Google
Chat
in
case
you
don't
have
it
open
here
inside
the
phase,
one
section
I
started
a
subsection
things
that
need
testing,
and
the
idea
is
basically,
if
you
just
like,
say
like
arm
clusters
I
command
arm
clusters
on
there.
That's
one
of
the
things
that,
in
the
coverage
report
is
almost
entirely
red.
We
don't
have
any
tests
for
that
at
all.
B
So
that's
one
of
the
big
items
we
want
to
have
some
testing
for
before
we
move
forward,
and
so
the
idea
which
we're
just
going
to
go
through
the
coverage
report
and
find
things
that
are
sort
of
like
Big,
Red
Blocks
or
that
they're
like
high.
Like
top
level
functions,
you
can
call
the
CLI
that
don't
have
any
testing
for
like
I
know
like
adoption.
Stuff
is
another
big
one.
B
I
think
there's
nothing
for
and
we're
going
to
see
if
we
can
write
some
unit
tests
around
those
and
then
once
we've
essentially
checked
off
all
the
items
whatever.
That
percentage
is
so
that
it
should
in
theory,
if
we
do
all
the
items
be
something
in
the
at
least
in
the
80s,
but
once
we
have
all
the
items
done
anyway,
then
we're
good
to
start
breaking
things
up,
because
we
can
be
a
bit
more
confident
that
we
can
find
if
things
are
broken.
Yeah.
B
Yeah
and
so
the
the
idea
behind
these
other
pad
was
that
doing
it.
This
way
allows
us
to
do
a
bit
more
asynchronously
I,
don't
have
to
meet
up
and
say,
like
I'm,
going
to
work
on
this
one
right
now,
I
can
just
put
my
name
next
to
whatever
thing
like
say:
I'm
going
to
test
I
probably
will
end
up
doing
the
agent
stuff.
I
probably
should
edit
that.
B
D
Think
the
healing
issue
with
the
tracker
is
it's
hard
to
keep
track
of,
for
lack
of
a
better,
somewhat
unintentional
punny
thing
it's
like
if
we
were
to
start
filing
trackers.
You
have
to
be
like
on
top
of
the
tracker,
whereas
this
is
like
kind
of
like
at
a
glance
you
can
see
like
oh
John
has
decided
to
take
the
NFS,
stuff
or
or
Adam
is
doing
agent.
A
Last
or
or
maybe
we
can
use
some
tech
and
special
attack
for
all
the
tickets
related
to
the
refactoring.
B
And
obvious
otherwise
we
could
do
it.
I
would
enjoy
this
way.
This
is
just
really
like
lightweight
super
easy
just
to
put
it
in
here
as
long
as
you
can
do
that,
link
to
the
pad
right
and
also
we
don't
one
thing,
because
we're
not
backboarding
any
of
this.
B
We
don't
need
to
worry
about
tracking
that
sort
of
stuff,
which
I
think
is
one
of
the
things
that
the
tracker
is
really
good
for
I
mean
I
would
be
okay
with
the
idea
trackers,
but
I
just
think
this
might
be
a
bit
easier
on
people.
As
long
as
you
know
where
this
link
is
it's
just
all
the
things
are
listed
right
here.
It's
just
a
very
like
concise
way
of
looking
at
it.
D
Okay
and
I
think
the
other
thing
that's
easy
about.
This
is
It's.
You
know,
as
it's
immediate
like
I
can
see,
you
know
say
right
after
lunch,
Mike
adds
a
line
and
he
takes
a
thing.
I
can
say:
oh
wow,
okay,
he's
done
it
already
like
I.
Don't
have
to
worry
about
like
refreshing,
my
web
browser,
or
you
know
what
an
email
has
not
come
into
my
inbox
and
I
start
the
work
on
the
same
one.
That
kind
of
thing
it's
it's
ether,
pad's,
very
okay,
so
immediate!
B
I,
don't
know
you
even
technically
make
sub
trackers
that
could
all
be
linked
from
there,
but
again
I'm.
Just
not
sure
it's
worth
the
effort
to
do
all
that
yeah
and
given
a
group
of
people
like
everyone's
here
sort
of
on
the
same
page
as
what
we're
doing-
and
we
all
know
where
to
look
here
and
it
could
be
easier
to
do
it
this
way.
B
A
B
Yeah,
we
could
have
one
oh
tracker,
that
could
all
work
that
they
fixed
that
one,
but
we
can
just
leave
it
as
sort
of
in
progress,
States,
I,
guess
until
everyone
until
we're
happy
with
the
coverage,
then
we.
A
A
B
A
B
I
guess
we
yeah.
That
sounds
like
a
good
idea.
Let's
we'll
do
that
too.
I'll
open
one
big
tracker,
I'll
paste
it
in
that
same
spot
with
the
other
stuff
in
the
things.
I
need
testing
parts,
and
then
you
guys
can
just
everyone
can
just
link
that
one
track
for
the
fixers,
then
we'll
just
make
sure
to
in
the
tracker
like
comment
that
you
know,
that's
that
there
are
as
relevant,
maybe
just
say
like
this
is
a
PR,
and
this
is
exactly.
B
A
B
There's
yeah
another
in
the
main
other
pad
on
December
20th.
There's
a
one
of
the
pull
requests
link
there.
Okay
is
a
thing
from
John
to
update
the
talks
file
to
make
the
coverage
report
yeah.
D
And
so
because
I
don't
really
want
to
deal
with
infrastructure,
it's
not
going
to
be
a
thing
that
the
CI
does
for
you
and
hosts
somewhere
in
a
web
server.
It's
just
going
to
be
a
talk
cnv
that
is
not
in
the
EnV
list,
so
it
won't
get
run
automatically
yeah,
as
the
comment
says
that
so
you
just
have
to
remember
to
run
like
talks,
Dash
e
coverage,
and
then
it
will
plonk
down
a
directory
full
of
HTML
files
and
some
JavaScript
in
your
in
your
local
in.
A
B
B
A
B
D
A
B
D
Oh
I
was
gonna,
say,
I.
Think
it's
a
good
idea.
I
think
what
we
want
to
do
first
is
to
break
things
up
at
a
file
level
before
we
actually
like.
Unless
we
find
like
a
major
bug
I'm,
my
suggestion
would
be
to
not
change
the
code
itself
like
if,
if
as
we're,
writing
coverage,
we're
like
oh,
this
is
like
bad.
We
should
fix
it
right
away.
We
can
fix
it,
but
what
I
would
suggest
if
it's,
if
it's
minor,
like
oh,
a
spelling,
mistake
in
the
exception
or
a
small
logic
error?
D
But
it's
not
going
to
lead
to
much.
We
should
leave
that
mostly,
as
is
do
the
breaking
up
of
typhadium.py
first
and
then
come
back
in
and
say:
oh
okay,
now
that
we've
broken
up
say
the
agent
into
its
own
file,
so
the
agent
class.
D
D
You
should
do
that
after
we've,
like
separated
it
from
the
main
code
base,
because
if
we
make
too
many
changes
at
once,
it
becomes
hard
to
understand
the
history
of
the
code.
Yeah.
D
Future
yeah
I
would
love
more
automated
quality
checks.
I
I
know
in
some
of
the
go
code,
bases
I
work
in
you
know
we
have
some
like
function.
Length
checks,
that's
basically
like
yeah.
This
function
is
looks
too
complicated,
yeah
and
oftentimes.
If
you
spend
a
few
minutes
looking
at
it,
you
can
see
places
where
it's
easy
to
say
break
it
up.
When
Adam
was
sharing
a
screen
and
scrolling
down
one
of
these
functions,
the
level
of
indent
was
very
deep.
D
So,
that's
always
to
me
a
candidate
yelling,
hey,
hey
I
deserve
to
be
refactored
a
bit
yeah.
B
And
one
other
benefit
is:
if
you
just
do
something
for
now:
that's
just
testing,
then
we
don't
have
to
run
through
pathology
in
a
test
which
we
save
a
lot
of
time
as
well,
and
then
you
know
at
the
end,
when
we
do
the
refactoring,
we
can
start
doing
things
through
yeah.
D
D
B
D
I
didn't
really
capture
in
that
document,
which
is
like
G2
application
stuff
today,
that's
like
redundant
in
sefadium
or
the
manager
module,
probably
going
to
python
com
and
things
like
the
the
various
types
and
the
the
yaml
conversions
yeah,
because
I
I
know
right
now,
like
I
think
I
saw
it
was
like
a
couple
weeks
ago,
dashboard
team,
adding
you
know
something
and
it's
like
oh
I've,
got
to
add
it
in
two
places.
Eventually
we
wanted
to
fix
that,
but
I
think
that
should
be
done.
B
And
then
we
may
also
say
about
the
coverage
or
the
immediate
plans.
Are
we
all
good
with
this?
What
we
had
here,
where
we're
gonna
just
go
through
and
start
adding
us.
B
Yeah
I
was
going
to
briefly
bring
it
up,
so
this
is
infinity
with
also
the
fact
that
there's
really
a
Seth
developer,
the
emea
one
is
on
October
I
think
it
was
so
it's
a
Wednesday
yeah
and
so
we're
going
to
do
one
of
those
small
presentation
there
or
just
bring
it
up.
I
guess
to
say
that
this
is
what's
going
on
in
case.
Anyone
ever
cares.
B
B
It's
October
5th.
For
me
it
is
11
A.M,
which
would
be
5
P.M.
A
It's
Wednesday
yeah
called.
B
Yes,
we
wanted
to
do
Ernesto
basically
recommended
yesterday
that
we
do
a
small
sort
of
thing
on
that
there
just
again
to
keep
bringing
up
to
a
wider
audience.
We
already
did
the
users
list
and
the
CLC
call
before
before
we
merged
it
and
now
maybe
bring
it
up
there.
Now
that
and
it's
actually
there
and
people
can
play
with
it
a
bit
more
and
then
part
of
that
is.
We
wanted
to
have
some
amounts
of
documentation
on
this.
Nothing,
nothing
big!
B
Yes,
but
it's
just
like
a
basic
thing
like
this
is
how
you
do
the
compiling
and
then
like,
linking
from
the
spot,
where
we
tell
people
right
now
to
just
pull
from
GitHub
and
saying
like
this
is
how
you
can
compile
it
yourself
and
then
in
the
future.
Maybe
we'll
have
like
an
actual
link
to
download.com,
whatever.
D
So
I
shared
a
link
in
the
in
the
today's
section
for
the
weekly
either
pad
it's.
It's
got
the
part
where
it
tells
you
how
to
compile
it.
Hopefully
the
instructions
are
simple
enough,
they're,
a
one
very
short
paragraph.
D
Hopefully
the
code
is
bulletproof
enough
that
they
don't
need
a
lot
of
help
hand-holding
to
run
the
script
and
the
rest
of
it
is
basically
I
mean
it's
in
draft
and
to
do
partly
because
I
haven't
cleaned
it
up
yet,
but
even
when
I
do
we
don't
actually
have
a
download.zeff.com
link.
So
that's
something
that
I
did
want
to
discuss
today,
which
is
who
is
the
person
we
should
reach
out
to
at
least
first
to
kind
of
get
the
ball
rolling
on
having
a
download.sef.com
and
the
whole
signing
business.
A
D
D
Packages
are
signed,
I
think
it
was
Patrick
mentioning
that
in
the
CLT
call
a
couple
weeks
ago,
but
I
don't
know
more
than
that.
So
that
is
one
person
we
could
potentially
start
with
and
then
could
redirect
us
to
someone
else.
D
B
I'm,
not
entirely
sure
is
the
right
person,
especially
now
that's
David,
Galloway's
gone
I
would
have
had
him
on
the
list
of
people.
Maybe
ask,
maybe
we
could
ask
Dan
Nick
again.
He
please,
the
chakra
stuff
I
think
is
somewhat
connected
to
this
and
he
might
know
something
yeah,
maybe
Patrick
might
know
something
as
well.
Really
anyone
that's
super
like
high
up
like
like
Josh,
Sturgeon
or
Ernie
light,
might
know
something
yeah
yeah
with
David
Dunn
I'm.
B
Not
he
was
again
he
would
be
better
than
what
person
I
would
have
gone
to
before,
but
one
of
those
other
people,
I
mentioned
I,
think
could
know
something.
B
Yeah
I'm
sure
there's
some
form
you
can
ask
I
could
even
ask
him
like
the
CLT
call,
maybe
or
something
like
that
again,
because
you
think
all
the
people
who
are
in
charge
of
big
things
would
be.
B
Yeah,
so
there's
definitely
a
few
forms.
We
could
go
to
to
figure
that
out
and
luckily
there's
no
huge
rush
for
it,
because
we
don't
necessarily
really
need
this
until
there's
a
release
right
now
again,
this
is
a
the
main
branch.
It's
not
meant
to
be
stable.
It's
people
want
to
use
it.
They
can
compile
themselves.
B
It's
not
going
to
be
that
hard
right,
I'm
more
concerned
actually
with
having
some
documentation
for
how
to
do
the
compiling,
even
if
it
is
very
basic
and
then
the
download
just
have
to
come,
link
can
come
when
it's
available
sure
So
Adam.
D
If,
if
you
want,
I
can
certainly
start
breaking
that
PR
up
into
multiple
Parts
I
just
started
with
the
existing
documentation,
which
says
you
know,
pull
it
from
the
ceft
tree,
which
is
obviously
incorrect.
Now,
so
I
could
preserve
the
part
about
the
documentation
and
just
add
up
thinks
like
warning
above
saying.
B
Of
like
how
messy
can
we
have
the
main
branch
documentation
like
in
theory,
if
it's
just,
maybe
we
could
even
just
say
like
yeah.
This
is
a
there's.
There's
no
way
to
do
this
right
now.
You
have
to
use
this
compiler
thing
or
whatever
and
just
link
to
something
doesn't
have
to
be
like
an
official
nice
way,
yeah.
B
That
would
be
fine,
even
if
even
there's
a
link
I
was
looking
briefly
at
your
documentation
that
has
like
the
to
do
Lincoln
even
that
being
in
main
branch.
Just
as
a
way
of
saying
like
you
can't
do,
this
anymore
is
I,
don't
even
think
would
be
a
huge
deal
but
yeah
as
long
as
there's
something
that
signals
like
at
least
in
Maine.
You
can't
do
this
anymore,
this
curl
and
then
use
a
link
to
the
compile
how
to
do
that.
I
think.
That's
the
important
part.
B
There's
a
few
things
to
do
it,
but
I
think
if
you're
talking
about
the
warning
makes
sense
as
well
and
really
anything
that
gets
the
idea
across.
D
B
B
B
Right
that
was
the
only
topic
I
had
in
the
pad.
I.
Don't
have
any
other
topics
they
want
to
bring
up
in
here.
A
B
All
right,
in
that
case,
you
can
end
here
and
I'll,
see
you
all
right.
Sometime
later
this
week,.