►
From YouTube: Ceph Orchestrator Meeting 2022-09-27
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,
so
we
have
a
couple
of
topics
in
here
for
this
week:
first
something
I
added
in
a
while
ago.
This
came
up
again
during
the
upgrade
of
the
giveaway
cluster
to
1624.
There
was
a
host
that
was
offline
and
basically
the
ads
have
things
the
upgrade.
A
It
says
it
like
fails
with
an
unexpected
exception,
or
something
because
I
can't
reach
the
host,
obviously
isn't
good,
but
that
does
raise
the
question
of
what
we
should
do
in
those
cases,
because
if
you
just
try
to
upgrade
everything
else-
and
you
end
up
with
up
with
the
problem
level
up
in
these
out
of
order
or
you
just
end
up
with
some
demons
getting
left
over
on
upgraded
if
you
force
moving
the
host
I
mean
it's,
it
works
it's
consistent,
but
it's
a
little
Annoying
user.
If
we
could
do
something
better
so.
A
They
had
to
remove
the
host
from
the
cluster
there's
you
have
to
right
now,
because
right
now
again,
it
just
throws
an
exception
that
you
have
an
offline
host
in
there,
so
that
was
the
only
way
to
get
it
to
work.
A
I
said
like
for
there,
I
mean,
like
the
the
minimum
that
we
need
to
do
is
have
a
better
reception.
Like
I
said
you
can't
just
say,
like
unknown
exception
or
whatever
I
think
I
have
copies
there.
Yeah
they'll
do
to
an
unexpected
exception.
This
is
obviously
not
not
good
enough
for
an
error
message,
but
they
did
have
to
you
know
they
had
to
remove
those
basically
before
upgrade
would
go
on.
A
Enough,
there's
nothing
at
all
right
now.
It's
not
accounted
for
just
like
assumed
I,
guess
that
the
hosts
are
offline,
like
there's
other
places
in
the
code
where
we
have
some
stuff
for
like
making
sure
we
don't
do
things
off
my
house,
but
this
is
one
Jeffrey
doesn't
have
it
right
now,
even
like
I
know,
I
just
a
while
ago
had
a
pull
request
for
the
client
files
and
things.
So
we
were
trying
to
write
those
to
offline
hosts
still,
but
there's
a
few
places.
We're
missing.
A
I
think
this
is
one
of
the
big
ones.
That's
left.
It
doesn't
account
for
offline
house
at
all,
yeah.
C
Do
you
remember
some
VR
introduces
recently
it
was
like
to
exclude
offline
hosts
from
from
scheduler
will
placing
demons.
A
Yeah
I
think
that
yeah
there's
there's
that
in
the
scheduler
it's
been
around
for
I
think
some
time
had
some
Logic
for
it,
but
I
think
upgrade
doesn't
use
the
scheduler
or
not
I
I
believe
I
want
to
say
that
was
farther
back
though
I
don't
remember
exactly
when
that
got
added,
then
I
know
I
know.
Recently
there
was
a
PR
that
added
the
draining
hose
to
the
do
the
scheduler,
but
I
think
the
offline
hosts
were
included
under
these
called
like
unreachable
hosts.
A
There
I
know,
I
think
that
one's
been
around
for
some
time
there
so
like
what
it
does.
There
is
it'll
automatically
filter
out
any
host
and
the
marks
is
unreachable
and
if
those
unreachable,
it
won't
even
try
to
like
the
big
things,
won't
try
to
remove
a
demon
from
that
house,
even
though
we
don't
really
want
it
in
the
placement,
it'll
just
skip
the
remains
in
the
camp.
C
The
way
in
the
past
I
worked
on
some
software
upgrade
and
we
had
like
silver
nodes,
but
it
was
maybe
the
architecture
was
different.
C
So
whenever
some
node
joined
the
cluster
first,
it
asks
like
for
the
the
software
versions
to
to
some
Central
node,
plus
the
centralized
architecture
and
the
text
against
the
versions
it
has.
If
the
the
versions
the
node
is
running
are
below
the
the
version
that's
supposed
to
run.
It
just
then
starts
an
upgrade
process
and
starts
like
asking
for
new
images
or
all
the
stuff
that
it
has
to
run.
But
in
our
case
we
don't
have
such
like
Central
Intelligence
that
handles
the
version
of
software.
As
far
as
you
know,.
A
Yeah,
we
really
do
have
something
like
that
and
you'd
almost
have
to
like
we'd
upgrade
the
rest
of
the
host.
You
have
to
track
that
this
host,
like
was
offline
at
the
time
and
have
never
came
back
online.
You
could
automatically
do
something
with
it
a
bit
more
complex,
I,
guess
it
would
sort
of
do
we
want,
but
I
mean
I,
don't
know,
but
it's
hard
because
it
definitely
seems
like
having
the
demons
be.
The
same
version
is
kind
of
important,
so.
C
B
B
So
we
do.
We
have
the
special
case
that
too
there's
a
couple
different
scenarios:
one
is
legit
dead.
Host
one
is
offline
host.
It's
not
reporting
you
could.
You
could
write
down
a
few
of
the
scenarios
and
then
kind
of
say,
okay.
Well,
if
if
we
were
to
do
the
like
the
simple
and
dumb
approach
of
not
allowing
upgrade
when
the
host
is
done
down,
what
do
you
do
to
recover
from
this
afterward?
B
And
then
you
know
maybe
just
pick
the
simplest
thing
after
kind
of
going
through
these
these
various
scenarios,
we
had
some
very
similar
issues
in
the
kitty
and
cluster
where
it
was
just
kind
of
like
well,
what
do
you?
What
do
you
do
and
in
those
cases
we
always
ended
up
punting
and
kind
of
pushing
it
back
on
support
teams
or
people
just
to
kind
of
work
around
them
manually,
yeah,
so
I'm
inclined
to
say,
do
the
simplest
things
first,
but
I
think
just
writing
stuff
down
and
walking
through
the
various
scenarios.
B
First
is
helpful.
I
mean
Adam's
already
done
some
of
that
in
the
in
the
in
The
Ether
pad
there,
but
I
think
it
needs
to
be
kind
of
fleshed
out
a
little
bit.
A
Yeah
I
was
sort
of
leaning
towards
just
sort
of
doing
similar,
just
like
pushing
it
back
on
them,
because
you
can
always
remove
a
host.
We
have
a
force
flag
for
that
already
like.
If,
if
you
have
demons
on
it
or
whatever,
where
you
can,
let
you
enforce,
remove
and
just
stuff
him
just
forgets
about
it
entirely
and
that's
what
they
actually
did,
because
they
didn't
get
the
host
back
up
when
they
were
doing
this.
A
He
said
it's
not
the
ideal
scenario.
If
you
can
get
the
host
back
at
some
point
because
then
you
know,
it'll
come
back
and
it'll
still
be
on
the
old
versions
and
because
the
upright's
not
in
progress
anymore,
it
would
just
sit
like
that,
but
you
I
guess
I.
Don't
think
that's
the
question.
A
B
C
It
would
just
just
like
maybe
first
we
have
to
at
least
like
the
as
you
commented
before,
John,
to
least
liked
the
different
situations,
and
she
likes
the
basic
ones
to
see
if
our
addressing
them
or
not,
then
to
go
with
more
complicated
situations
like
first
the
the
basic
ones
like
it's
almost
crashes,
basically
during
an
upgrade
or
just
goes
offline,
because
some
Hardware
or
whatever
issue,
then
what
to
do
like?
If
you
will
stop
the
upgrade
or
we'll
just
mark
this
post
as
offline
and
continue
without
it.
C
A
B
B
Those
are
the
ones
that
you
can
say:
okay,
well,
you
know
which
approach
to
upgrading
helps
this
case
the
most
and
and
that's
where
you
could
actually
do
some
more
smart
things
like
if
you
had
an
agent
or
something
you
could
say.
Oh
am
I
am
I
actually
in
in
sync
with
the
rest
of
the
cluster
at
boot
or
something,
but
that's
something
we
could
talk
about
later
after
we've
kind
of
charted
out
the
different
paths
that
they
could
take.
A
Yeah
I
might
just
start
with
trying
to
fix
the
error
handling
and
then
maybe
go
put
a
deeper
and
maybe
in
the
future,
there's
something
where
you
can
sort
of
track
that
a
host
in
the
cluster
was
offline.
When
the
update
was
happening
and
then
you
can
later
right
upgrade
the
demons
on
there.
Yeah.
B
I
was
about
to
say
something:
sorry
go
ahead:
I'll,
I'll,
try
and
remember
what
I
was
about
to
say.
You.
C
Know
I
would
just
the
just.
Do
we
write
down
the
version
of
images
in
in
the
monitor,
store
somewhere
versions.
A
Because
we
track
those
as
part
of
the
refreshing,
the
demons
getting
that
metadata
and
they
shows
up
in,
like
the
HPS
I'll,
put
all
those
yeah.
C
So
self
EDM
is
able-like.
Once
we
are
up,
we
should
be
able
like
to
get
the
current
image
version
and
by
the.
A
C
Yeah
not
in
terms
of
upgrade
in
terms
like
if
you
are
running
and
you
are
running
in
host
and
then
you
the
host
goes
offline
for
whatever
reason
then
we
upgrade
the
cluster.
Then
we
switch
on
the
the
host
newly
you
join
the
cluster.
A
So,
as
far
as
actually
like
joining
the
cluster
and
like
the
purest
sense
for
Stuff,
Stephanie
is
actually
totally
control
that
stuff
like
if
there's
an
OSD
that
was
down
and
it
comes
back
up
and
it's
on
the
wrong
version
it'll
still
join.
We
don't
have
to
do
anything
for
that.
It
just
does
that
as
far
as
we
are
concerned
for
the
upgrade,
what
all
we
track
is.
A
Basically,
the
user
gives
us
some
image
or
something
or
a
version
in
which
we
then
convert
to
an
image,
and
then
we
figure
out
what
the
the
digest
is
for
that
image
and
then
for
checking.
If
the
image
is
upgraded
or
not,
are
the
demons
upgraded?
We
just
check
if
it
has
its
images
using
that
same
digest.
A
So
if
we
send
it
like
hey,
if
they
both
did
come
back
like
you,
re-added
it
to
the
cluster
and
it
already
had
those
demons,
we
would
then
detect
those
demons
and
we
would
be
able
to
see
what
image
they're
on.
So
we
would
know
immediately
that
they're
not
on
the
same
image
that
all
the
other
ones
are.
A
So
as
long
as
we
have
stored,
which
we
do
as
like
config
options,
what
the
image
is
supposed
to
be
for
each
of
the
demon
types,
you
could
typically
compare
them,
and
you
would
know
that
they're
on
an
older
one.
In
that
case,.
A
B
Mean
there's
other
things
to
consider
too
like
if
the
cluster
is
so
small
that
you
don't
have
Quorum
or
whatever
you're
like
well,
I,
probably
shouldn't
do
the
upgrade
anyway,
just
fix
my
cluster
before
moving
on,
but
you're
right
I,
like
a
huge
scale,
you
might
say:
oh
I,
don't
know
if
I
wanna
block
the
upgrade,
but
the
other
thing
I
was
going
to
say
is
if
you,
if
you
choose
to
block
the
upgrade
thing,
it's
very
simple
and
it
does
not
preclude
changing
that
in
the
future
to
do
something
smarter.
B
So
if
the
procedure
is
oh
block,
the
upgrade
oh
I
have
to
mark
these
hosts
as
gone
then
I
can
upgrade.
Then
I
can
ring
them
back
somehow,
if
the
hardware
or
the
OS
is
still
good
great,
we
can
come
up
with
procedures,
but
yeah
I
still
think
having
a
a
more
thorough
plotting
out
of
all
these
scenarios.
Yeah.
A
B
A
So
technically
I
just
felt
made
upgrade
so
I
guess
we
could
do
better
than
that.
Actually,
we
just
said
it
tends
to
fail
super
early,
because
every
single
host
is
a
crash
demon
on
it,
and
so
it
usually
finds
pretty
fast.
Oh
okay,
those
are
down
but
yeah
I
guess
it
probably
should
be
another
thing
that
we,
if
we
know
this
is
going
to
happen
that
we
put
if
we
used
to
have
any
hosts
in
the
offline
mode,
because
we
do
Dropbox,
so
we
should
just
we
should
insta
fail
as
well.
C
Yeah
yeah
so
like
before,
starting
the
upgrade
just
go
and
do
some
sanity
check
on
all
the
notes
that
are
supposed
to
be
up
and
running
and
there's
something
phase
we
can
just
block
this
node
is
not
responding,
so
you
can
take
it
off
the
cluster
or
do
whatever,
but
we
can
go
with
that
great.
A
Yeah
yeah,
so
I
I
think
that
that's
probably
that
might
be
a
good
starting
point,
is
and
have
some
quick
check.
I
don't
want
to
do
too
much
during
the
upgrade
start
command,
because
that's
all
like
synchronous
like
it's
there
they're
waiting
for
that
to
end,
but
since
we
do
typically
already
track
those
that
are
offline,
if
we
know
something's
offline
right,
then
maybe
just
tell
them
in
this
instant
Market
failed
and
then
are
they
in
the
middle
of
the
club.
A
Stickers
have
to
be
host
and
go
offline
in
the
middle
of
the
upgrade
as
well
right
now
that
would
just
fail
with
this
weird,
like
unexpected
exception
error,
we
should
probably
have
a
different
exception
for
like
couldn't
reach
host
or
something
like
that.
Yeah,
at
least
then
they
know
what's
going
on
and
we'll
mark
the
host
offline
as
well,
when
that
happens,
I
imagine
so
yeah
at
least
that
way.
They
know
why
it's
happening.
C
A
Yeah,
we
already
do
that
when
we
do
like
a
refresh
those
demons.
There's
like
a
decorator
that
okay
put
on
and
it
does
it
tries
to
do
it.
It
only
goes
up
to
10
at
once,
because
how
many
threads
we
have
in
our
worker
pool
I
think,
but
it
does
some
amount
of
parallelization
for.
A
Right,
I
guess
yeah,
that's
what
we'll
do
for
now
is
that
maybe
there's
something
more
complex
in
the
future,
we'll
probably
have
to
deal
with
a
drama
saying:
go
through
all
the
different
possible
scenarios
in,
but
for
now
I
guess
we
will.
A
Yeah
we'll
do
that
and
yeah
for
now,
we'll
just
do
the
basic
stuff,
better
error,
handling
and
we'll
we'll
stop
the
upgrade
immediately.
If
we
know
that
it's
going
to
fail
to
the
offline
Mouse
that
should
at
least
improve
the
user
experience
if
it
doesn't
necessarily
improve
the
outcome
right
now,.
A
All
right,
I
think
that's
probably
good,
so
we
have
another
one
here
about
the
RW
spec
format.
You
want
to
throw
this
one
right
up.
C
Yeah,
so
basically,
this
is
related
to
the
rtw
multi-site
stuff
I'm
working
on
so
first,
when
it
started
we're
gonna
I
didn't
take
into
account
much
the
things
that
has
to
do
with
the
demons,
but
then
I
think
it
was
after
a
comment
from
Adam
about
the
placement.
Then
I
started
thinking
about
it
and
since
that
we
are
introducing
this
new
support.
I
said
that
it
should
be
a
good
idea
to
provide
the
same
supports.
C
We
already
have
for
rgw
demons
in
cephyrium
in
this
new
model,
so
the
idea
is
to
be
able
to
introduce
or
to
use
the
same
parameters
we
already
use
for
safe
idiom
in
this
new
spec
that
we
use
to
deploy
rtw
multi-site.
C
So
this
way
we
can
use
the
same
parameters
for
placements
for
for
specification
and
also
for
SSL.
If
we
need
to,
because
probably
if
we
introduce
the
support,
then
users
will
start
asking.
How
can
I
use
the
same
thing
that
I'm
already
using
for
safety
in,
but
now
in
this
new
model,
since
we
can
access
at
least
these
specs
from
this
model,
because
they
are
in
in
Python,
common
I
was
thinking.
C
C
A
Yeah
I
mean
I.
Think
you
also
have
the
pros
and
cons
fairly
well.
I
mean
the
pros
are
basically
that
it's
it's
simpler
and
that
you
can
reuse
it
and
everything
there's
also
the
little
like
the
top
of
that
stuff.
You
have
on
there
the
nice
detail
that,
if
you
do
it
that
way,
that
means
when
you
eventually
have
to
tell
the
orchestrator
to
deploy
the
demons.
You
can
just
pass
that
spec
directly
there.
A
You
don't
have
to
worry
about,
maybe
modifying
it
or
something
you
could
just
pass
it
along
that
benefit
as
well,
and
then
the
cons
I
think
the
big
one
is
just,
and
what
you
have
in
here
is
that
people
might
get
confused
yeah.
B
Is
the
feature
set
possibly
a
like
a
super
set
of
one
of
basically
I'm
wondering
if
you
could
embed
one
or
one
inside
the
other,
so
that
for
the
all
the
common
stuff,
you
kind
of
have
an
embedded
rgw
spec
inside
of
a
rgw
multi-site
spec,
not
sure
if
it's
possible
or
makes
sense.
But
that's
the
first
thing
that
pop
and
it's
a
kind
of
you
get
kind
of
the
benefits
of
both.
C
In
fact,
when
you
use
the
spec
I
mean,
if
you
don't
like
to
create
a
Zone
to
start
stack,
Rising
till
then
you
only
have
one
real
with
one
zero
group
with
one
zone.
I
think
then
you
don't
have
like
multi-side
support
because
we
haven't
created
a
secondary
cluster.
C
B
Sort
of
I
I
was
getting
I
was
kind
of
trying
to
I'm
like
looking
at
your
two.
You
know
your
pros
and
cons
or
two
options
and
I
was
saying
I
wonder
if
you
can
kind
of
hybridize
them
and
kind
of
reuse
as
much
of
the
existing
rgw
spec,
maybe
maybe
not
even
the
parent
class.
But
if
you
could
just
somehow
reuse
as
much
as
possible
that
way,
people
don't
have
to
relearn
anything.
They
just
know
to
embed
the
rgws
back
inside
of
the
rgw
multi-size
bank,
again
I've
I've.
C
B
C
It
should
be
that
should
be
good
for
a
simple
setup.
Without
a
multi-site
and
using
the
additional
commands
you
can
go
and
create
your
secondary
cluster
and
create
another
Zone
that
syncs
with
the
real
you
have
just
created.
I
pushed
the
I
post,
the
the
pr
I
updated
a
little
bit
the
command
the
the
documentation.
So
maybe
it
gives
you
more
information
about
how
how
it
works
right
now.
C
So,
if
you
go
to
the
pr,
you
will
see
that
we
have
two
different
specs
like
the
first
one
is
to
create
durian
Zone
group,
blah
blah
blah.
Once
you
create
all
of
this,
then
you
have
another
command
to
get
to
get
the
tokens
and
with
the
tokens,
then
you
go
to
your
secondary
cluster
and
you
have
to
provide
another
spec
where
you
include
the
token
and
you
get
your
secondary
Zone
synchronizing
with
your
primary
cluster.
B
C
File
good
question
I
thought
on
that
also,
but
you
have
to
feel
like
a
specific
Fields
like
the
host,
and
that
would
depend
on
your
secondary
cluster.
You,
you
need
to
know
the
host
names,
the
ports.
You
will
be
deploying
on
yeah
well,
but
you
you
can
generate
like
a
skeleton
and
yeah
yeah.
That
could
be
possible,
but
it
will
not
be
very,
very
useful
because
you
have
to
put
this
specific
details
on
it.
A
I'll
just
go
out
and
say
that
I'm
kind
of
in
favor
of
just
having
it
be
the
same,
spec
I,
know,
there's
that
the
confusion
is
possible
that
people
apply
to
the
wrong
place,
but
that,
maybe
that's
that's
okay
and
that's
the
other.
The
two
options:
it's
the
better
one,
like
the
benefits
of
doing
it,
are
worth
it
yeah
be
careful
when
we're
documenting
it.
I
guess
maybe
like
a
warning
in
the
documentation.
C
And
I
was
trying
to
think
about
like
the
use
cases
in
some
notes
familiarize
with
the
rgw
so
now
that
that
we
have,
this
supports,
I,
wasn't
sure
but
like
in
which
case
you
will
go
and
use
the
fadm
for
the
deployment
directly.
Not
using
this
model.
A
I
mean
it
would
just
be
if
you'd
already
manually
set
up
the
like
the
Realms
and
zones
and
things
like
the
way
people
do
it
right
now.
Basically,
is
you
manually
set
up
your
Realms
and
zones
with
other
external
commands,
and
then
you
just
want
to
deploy
it?
Even
so,
you
just
have
to
begin
to
deploy
demons.
A
I
assume
some
people
will
still
want
to
do
that,
especially
if
this
is
a
new
feature,
we
might
not
immediately
you
go
for
it
or
whatever
so
I
think
there'll
still
be
a
reason
like
I.
Don't
think
we
can
get
rid
of
like
say
applying
this
spec
with
with
the
orchestrator,
because
people
are
also
going
to
want
to
use
that
use
case.
C
A
I
mean
the
way
it's
definitely
is
I.
Imagine
there's
some
people
just
don't
just
never
want
to
change
away
from
anything.
I,
say:
I
I
think
that's
letting
the
spec
be
applied
to
both
of
them
sort
of
makes
sense
again,
even
though
there
is
that
sort
of
risk.
A
A
If
you
apply
to
just
orchestrator
versus
the
RW
module
at
least
we
can
like
link
that
to
anyone-
and
we
can
say
you
know
there-
anyone
actually
had
read
through
it
or
something,
because
that's
probably
where
people
will
go
I
guess
if
they
they
think
it's
not
working,
is
they'll,
go
check
the
documentation
and
see
what's
going
on
and
we'll
go
into
the
users
list
or
something.
B
C
Okay,
so
I
will
just
introduce
this
warning
in
safety
and
dogs
to
see
that's.
That
is
this
new
model.
That's
gonna,
do
you
know
working
with
things
for
you?
If
you
want
to
to
use
it,
then
just
go
and
use
it.
Otherwise
you
know
that
otherwise
you
have
to
create
everything
manually
as
we
use
it
to
do.
A
Yeah
because
I
think
there's
a
lot
of
nice
benefits
to
being
able
to
use
the
same
one
and
I
don't
want
to
like
get
rid
of
those
for
something
that
I
think
is
is
sort
of
a
minor
issue.
C
A
All
right,
those
were
the
topics
we
had
in
there
today.
C
I
have
the
last
topic
introduce
it
okay,
this
is
related
with
the
the
new
we
will
package
Cafe
ADM
this.
Now
we
have
this
support
packaging
and
there
is
no
issue
with
the
not
having
a
version
or
for
safety
theme.
So
maybe
now
we
are
at
a
good
points
to
see
if
we
can
go,
introduce
some
some
version.
B
Yeah
the
I
think
compared
to
the
previous
two
topics,
this
one's
this
one's-
an
easy.
Yes,
let's
just
do
it.
When
do
you
want
it
tomorrow,
Thursday,
no
I'm
just
kidding
it
shouldn't
be
very
hard.
It
should
be
a
matter
of
teaching
that
build.py
script
to
generate
a
I.
Think
a
chat.
I
said
something
like
underscore
versions,
py
and
then
stick
it
in
there
and
initially
we,
it
might
just
just
be
there
doing
nothing,
and
then
we
could
add
a
function
to
Steph
ADM
to
additionally
attempt
to
import
it.
B
A
Out
of
curiosity,
guys,
I
haven't
actually
checked
this
part
when
we
deploy
the
stefidium
from
the
manager
module
now.
Are
we
deploying
just
an
already
compiled
one
or
are
we
piling
it
on
there.
B
B
The
only
the
only
place
where
you
can
kind
of
get
a
your
own
homegrown
one
right
now
is
that
whole
manual
process
of
of
you
know
checking
out
the
tree
and
running
the
build
script.
Yeah.
B
A
B
It
should
be
yes,
if,
if
you
find
that's
not
the
case,
let
me
know
I'm
like
98
sure
that
should
be
the
case,
because
one
of
the
things
I
did
is
check
that
the
RPM
builds
and
I
didn't
really
go
deeply
into
the
devs
I
just
assumed
if
the
RPMs
were
working.
B
The
The
Depths
were
too
that,
because
the
build
is
ultimately
run
by
cmake,
which
I'm
assuming
is
going
to
be
doing
the
same
thing,
it's
just
the
kind
of
post
build
step
that
actually
creates
the
packages
that
differ
so
yeah
cmake
is,
is
running
build.py
sources,
Fabian
build.py
as
part
of
the
chain
of
things,
and
and
that's
where
we
probably
need
to
get
some
of
our
version
numbers.
B
So
what
we
teach
build.py
how
to
do
stuff,
we
probably
need
to
plummet
through
cmake.
That's
the
only
probably
challenging
part
to
the
suggestion.
A
Okay,
that's
what
I
was
wondering
about
is
the
one
that
came
from
the
manager
module,
so
I
think
it
makes
it
seems
really
simple
for
the
one
on
the
host
who,
just
on
the
compile
thing
and
it'll
spit
out
the
version
thing
and
then
either
but
yeah,
the
one
in
the
management
was
expecting
some
things.
Sorry
to
be
there.
So
it's
a
bit!
Well,
it's
I!
Guess:
yeah
Twitter
sayings!
That's
the
build
process
a
little
bit.
B
You
know,
basically,
we
can
look
to
see
if
there's
any
prior
art
in
the
cmake
I.
Don't
know
if
the
like,
you
know,
oh
the
manager
binary.
The
mod
binary
actually
gets
like
the
get
hash
embedded
in
it.
I've
seen
some
of
this
stuff
returned
from
some
of
the
API
commands,
but
I'm
not
sure
how
they're
getting
them,
but
if
it's
anything
remotely
similar,
we
can
just
plummet
through
add
some
command
line:
option
something
and
and
and
kind
of
hook
it
in
there.
B
A
Okay,
I
guess.
The
only
question
left
then
is
more
of
when
you
want
to
put
this
in
I.
Don't
know
if
you
want
to
like
wait
closer
to
the
reef
release
or
or
when
we
want
to
do
this
and
whatever
you
want
to
start
with,
and
all
those
things.
B
B
C
B
Not
sure
again,
it
goes
back
to
that.
How
much
work
do
we
want
to
make
our
for
ourselves
with
like
regards
to
backboarding
into
into
the
current
cefadium
or
the
previous
I
guess,
I
should
say,
because
if
we
start
making
a
bunch
of
changes
that
are
specific
to
the
compiled
version
that
might
make
backboarding
a
little
harder
I
don't
know,
I
haven't
been
around
long
enough
to
know
how
challenging
these
kind
of
things
are,
that
backboard
process.
A
A
B
Yeah
yeah
I
think
the
issue
is
like
say:
we
start
making
changes.
Now
we
don't
even
break
it
up.
We
just
add
a
bunch
of
commands
that
are
specific
to
the
new
version.
Then
we
find
a
big
bug.
Oh,
we
need
to
fix
the
big
bug:
oh
fix
it
on
Main,
oh
back
Port.
Up
now
we
have
tons
of
merge
conflicts
because
all
the
lines
got
shifted
around
and
stuff
in.
A
B
Yeah
I
think
it
might
make
sense
to
defer
giving
stuff
ADM
the
ability
to
print
out
its
own
version
for
now,
and
then
you
know
come
back
to
that.
So
you've
got
the
tracker
already.
We
could
like
I
said:
I
could
do
it
in
phases.
We
could
add
the
embedding
of
the
version
numbers
like
this
week
or
next
week
and
then
come
back
in
a
month
or
two
and
actually
teach
it
how
to
interpret
that
file.
A
Yeah
I'm,
just
looking
now
back
at
the
refactoring
binary
I'll
show
you
guys
I
figured
this
would
just
go
into
one
of
these
phases.
Yeah.
B
B
A
Like
right
before,
we
split
things
up,
but
right
after
we've
done
all
the
testing
and
everything
yeah.
C
Is
there
any
reason
that
we
don't
just
use
the
version
like
from
get
described
instead.
B
B
Let
me
forget:
I
have
a
new
topic
or
follow
up
on
that
one,
because
I
keep
forgetting
so
you
you
know,
person,
X
goes
off,
grabs
gets
us
f80
on
binary
from
somewhere
pass
it
on
their
system
and
they
want
to
know
it's
provenance
or
whatever,
so
they
run
likes
f80
and
dash
dash
version,
and
it
says:
oh
you
know,
I
was
built
on
such
and
such
a
date
such
and
such
a
issue
get
hash
such
and
such
a
good
version.
C
A
A
B
A
On
right,
yeah,
so
I
guess
we'll
wait
on
this
until
we
get
through
all
the
testing
and
everything,
but
right
before
we
split
things
up,
we'll
try
to
figure
this
one
out
because
that'll
be
kind
of
a
good
idea
of
what
we
want
to
do
as
well.
We
just
have
to
figure
out
some
of
the
details
with
the
build
process
and
everything.
B
Yeah
so
well
I
kind
of
forgot
to
follow
up
on
last
week.
Was
this
whole?
What
do
we
do
about?
How
do
people
get
cephanium?
Now,
with
this
compiled
version,
the
whole
like
download,
stuff.com
I,
think
we
had
discussed
it
briefly
like
we
need
to
reach
out
to
someone,
probably
who
has
more
knowledge
about
this
whole
thing,
so
I'm,
assuming
we
just
you
just
mail,
devl
and
CeCe,
a
couple
specific
people
or
does
anyone
have
any
suggestions.
C
A
Still
around
I
think
what
else
that
I
don't
know
maybe
like
Denmark,
also
knows
something
about
this
I
think
you
know
a
lot
about
some
of
these
periphery.
Things
see,
if
knows
anything,
because
he
knows
about
the
staff,
build
repo,
at
least,
and
maybe
there's
something
related
to
there.
A
B
A
B
All
right,
so
what
I'll
do
probably
this
afternoon
and
I'll
I'll
send
a
draft
to
you,
Adam,
just
kind
of
double
check
and
then
I'll
try
and
send
that
off
today.
A
There's
not
much
of
it,
but
there
is
like
one
or
two
spots
where
it
tells
you
to
curl
directly
from
GitHub,
basically
which
still
works
on
the
Quincy
and
Pacific,
but
if
you're
going
to
do
it
on
Main.
At
this
point
you
we
wouldn't
work.
No.
C
B
So
if
you
look
at
that
part
of
it
is
just
changes
to
the
now
I
spelled
instructions
from
her
fix
that
later,
yeah
I
just
threw
a
whole
bunch
of
to
do
to
do
to
do
just
kind
of
to
make
it
obviously
wrong.
C
Are
we
publishing
as
far
as
I
know,
we
are
building
daily
images
of
safe
and
they
are
available
somewhere
in
the.
A
And
they
make
packages
every
day,
yeah
on
the
sfci
back
and
you
can
find
them
in
Shaman
I,
don't
think
they'll,
there's
any
reason:
they're
gonna
ever
publish
the
video
there
because
we
they
already
put
them
on
chakra,
I.
Think
as
part
of
that
build
process.
Now.
A
I,
don't
I
mean
maybe
there's
some
reason.
Somebody
would
want
to
go.
Look
for
anything
else.
You
can't
access.
You
can't
find
them
somewhere,
though.
It's
definitely
possible
to
find
like
the
these
built
ones
on
there,
but
this
is
just
for
basically
the
official
release
ones.
We
want
to
have
a
different
place.
C
But
for
the
daily
The
Daily
releases
of
the
images
with
bits
of
vein:
where
are
they
publish
it?
I.
A
Don't
know
exactly
I
know
we
pull
them
as
part
of
the
tautology
test
now,
because
that's
so
you
get
Stephanie
I,
don't
remember
where
it's.
C
So,
like
Shaman
actually
pushes
the
Earth
of
the
build
thing
out
on
chakra
and
then
it's
retained
for
a
little
bit
of
time
for
the
topology
testing.
So
that's
like
with
the
daily
builds
or
the
you
know.
A
A
A
A
C
Oh
and
actually
I
do
see
it
for
Quincy,
so
here
I'll
post
in
the
so
check
out
this
link
here
and
you'll
see
there's
a
stuff
ATM.
This
is
for
Quincy
yeah,
because
I
think
he
switched
over
the
build
process.
Yeah.
You
went
to
like
the
dev
new,
build
processor
or
something
like
this,
so
it
did
affect
Pacific
and
Quincy.
A
It's
like
they
were.
The
link
ends
up
being
it's
really
like
RPM
Dash
reefs
or
something.
A
A
A
All
right
is
there
anything
else
you
have
to
say
on
that
topic.
It
sounds
like
we
kind
of
know
we're
going
without.
A
Confirm
with
them,
what's
going
on,
I'm
actually
allowed
to
fear
of
the
signatures,
it
sounds
like
it's
actually
in
a
better
spot
than
I
thought.
It
was.
B
A
All
right
do
we
have
any
other
topics
or
yeah.
A
Okay
sounds
like
we're
done:
yes,
we'll
call
it
there
and
I'll
see
you
all
next
week.