►
From YouTube: Ceph Orchestrator Meeting 2021-08-10
Description
No description was provided for this meeting.
If this is YOUR meeting, an easy way to fix this is to add a description to your video, wherever mtngs.io found it (probably YouTube).
A
Oh
hi,
everyone
and
welcome
to
today's
orchestrator
meeting.
Looking
at
the
again
now
we
have
what
topics
the
first
one
is
comic.
I
wanted
to
to
show
a
bit
how
we
are
testing
the
manager
rook
module
in
in
the
rook
ci
and
then
another
topic
that
came
up
last
week
is
about
resetting
our
design,
setting
oc
flex,
and
then
we
have
another
topic
that
came
up
last
week:
what
about
storing
private
keys
and
yammer
specs,
and
then
we
have.
A
Zones
and
zone
group
creation
in
the
orchestrator
and
then
we
have
a
local
storage
discussion
discussion.
Okay,
let's
start
with
the
first
topic:
cool
meg,
oh
my
god!
Dude!
Do
you
just
want
to
briefly
describe
how
you
are
testing
the
rook
orchestrator
module
in
in
the
rocky
eye,
because
I
think
to
the
rest
of
the
folks.
That's
super
interesting
to
know,
especially
if
we
want,
if
we,
if
there
are
possibilities
to
to
avoid
applications
or
speed
up
the
the
coverage.
I
don't
know.
B
Yeah,
of
course,
okay.
Well,
I
have
prepared
a
small
presentation:
okay.
Well,
I
think
that
blaine
travis,
probably
will
will
know
more
detail
about
the
internal
of
the
drug
testing
framework,
okay,
because
what
what
I
have
done
is
just
basically
use
the
the
the
mechanism
that
are
implemented
in
the
project
in
order
to
create
tests.
Okay,
so
I
think
that
what
is
is
very
easy
to
follow
and
let
me
check
okay.
B
So,
let's
go
basically
what
what
I
have
done
is
just
to
to
create
a
new
test
for
the
the
safe
manager,
orchestrator
module,
and
I
I
have
used
the
the
same
strategy
that
is
used
in
other
tests
in
the
in
the
rook
environment
in
the
sci
environment.
B
If
you
take
a
look
to
this
direction,
this
url
this
is
the
test,
the
explanation
of
the
test
framework
and,
basically,
what
what
provides
is
that
what
you
need
to
to
set
up
fixed
uber
methods,
cluster
using
cube
adm
using
mini
cube
using
kcli
or
another
another
tool?
Okay,
and
once
you
have
a
quantus
running,
you
can
execute
the
test.
B
Yes,
basically
executing
the
test,
and
this
is
going
to
run
the
test
framework
that
is
going
to
create
a
theft
cluster
and
is
going
to
run
your
test
and
after
that
is
going
to
clean
the
environment.
Okay.
So
but
this
is
the
basic
working
behavior.
B
What
are
the
tests
how
the
test
is
executed?
Well,
there
are
a
special
folder
for
all
the
all
the
tests
in
the
hook
project
and
the
this.
The
test
for
the
focus
stator
is
place
it
in
this
folder.
B
B
This
is
important
because,
when
I
think
that
one
of
the
targets
of
the
or
or
the
the
most
important
things
of
the
integration
test
is
to
detect
as
soon
as
possible,
if
we
have
brick
something
one
modification
so
to
have
a
set
of
integration
tests
just
directly
inside
the
room
project
and
to
to
be
possible
to
to
detect
this
kind
of
of
problems
as
soon
as
possible
is
very
important.
Okay,.
C
D
B
Okay,
this
is
the
different
runs
that
we
have
of
the
test
theft
manager
suite.
If
I
I'm
not
wrong
okay,
so
all
these
brands
are
are
started
because
there
is
a
pull
request
against
master.
That
is,
is
his
launch.
Okay,.
B
B
B
B
B
This
is
what
what
we
have
discussed-
okay
about
the
it's
very
important
to
have
this,
this
kind
of
intelligent
test
after
well.
If
you
want
to
execute
it
manually,
you
directly
from
the
rook
folder,
you
can't
issue
this,
this
kind
of
command
that
is
common
to
every
kind
of
test
and
execute
the
test.
B
What
we
are
doing
is
just
to
to
be
sure
that
the
rook
orchestrator
module
is
enabled
and
is
set
as
default.
Okay,
and
a
part
of
that
is
just
to
work,
to
start
to
execute
the
different
commands
that
we
have
in
the
organizator
in
the
theft
mana
suite.
There
is
a
method
that
is
called
execute
that
is
going
to
to
be
very
useful
in
this
kind
of
of
test,
in
order
to
just
to
call
the
the
different
or
stator
commands,
and
after
calling
the
orchestrator
command
that
you
want
to
test.
B
I
think
that
we
we
should
keep
the
functionality
to
create
osds
using
the
discovery
diamond
and
directly
creating
the
the
the
osd
okay,
the
in
the
in
the
more
simple
form
and
part
of
that.
Well,
there
is
a
open
discussion
about
the
local
storage
operator
and
what
we
are
going
to
do
with
that.
Okay,
so
I
think
that,
basically,
this
is
all.
If
there
is
any
question,
I
will
be
happy
to
answer.
If
I
able
to
do
that.
A
Thank
you.
What
what
I'm
wondering
is?
Can
we
share
the
maintenance
burden
of
men
of
maintaining
the
tests
between,
but
between
self
titality
and
rook?
I
mean
because
we
want
to
test
the
very
same
thing
right.
We
want
to
test
if
device
ls
works,
we
want
to
test.
If
ls
works,
we
want
to
test
it
first
equation
works.
We
want
to
test.
If
I
don't
know,
and
we
are
end,
we
are
going
to
end
up
with
a
very
same
tests.
A
We
just
implemented
one
time
and
go
in
the
rooks
here
in
the
rook
project
and
another
time
in
python
and
enjoyed
in
the
typology
integration,
and
I
was
wondering
if,
if
we
can,
if
there
is
a
way
to
only
write
it
once
I
don't
know.
B
I
don't
know,
I
think
that
is
something
that
is
going
to
be
a
little
bit
difficult,
because
basically,
the
problem
is
that
we
are
talking
about
two
different
projects:
okay
and
well,
something
that
is
static
in
one
project
is
dynamic,
completely
dynamic
in
the
other
one.
B
Okay,
so
well,
I
think
that
just
to
to
launch
the
integration
test
from
the
continuous
integration
system
that
is
running
in
rook
is
not
going
to
be
possible,
basically
because
well,
if
we
are
not
able
to
run
technology
testing
automatic
wake
directory
from
the
third
project,
I
think
that
is
going
to
be
very
difficult
to
do
the
same,
because
I
we
have
done
a
modification
in
the
in
the
rook
in
the
operator
code.
B
C
B
D
D
Yeah,
probably
the
more
complicated
tests
are
going
to
end
up
on
the
left
side
for
the
manager
work,
module
itself.
That
works
with
your
commands,
since
it's
the
ones
changing
the
most
yeah
yeah,
that's
okay!
As
long
as
there's
something
basic
on
the
on
the
work
side,
someday,
hopefully,
it'll
be
possible
to
easily
launch
a
teethology
job
and
let
it
whatever,
but
that's
a
ways
off.
A
Yeah,
especially
if
we
are
in
rook,
we
are
not
developing
the
orchestra
module
right,
so
the
really
detailed
testing
needs
to
be
done
and
safe
anyway.
A
C
A
Okay,
if
there
are,
if
there
aren't
any
other
questions,
I
guess
we
can
move
to
the
second
point,
but
that
was
a
question
that
came
up
last
week.
So
when,
when,
when
cliff
administrator
is
restarting
osds,
we
are
not
setting
any
any
flex
fill.
We
are
not
setting
no
out
or
or
anything
you're,
just
restarting
the
ausd's.
A
C
A
A
A
We
are
storing
private
keys
on
the
yamaha
spectrum
for
rgw
ssl,
for
example.
Do
we
need
to
improve
things?
I
mean
we?
We
don't
have
a
proper
secret
store
like
kubernetes
in
in
the
monster.
A
I
also
think
that
you're
good
as
we
are,
I
mean
at
some
point
you
have
to
put
the
private
keys
and
then
secrets
onto
the
host
and
at
that
point
they're
there,
anyway
and
and
the
connection
to
the
most
to
the
money
should
be
secured.
So.
A
What
what
about
the
monitor
database
is
it
encrypted
by
default
and
software,
and
do
you
know
that
change.
D
I
mean
it's
just
rockstar,
so
if
you
went
and
you
looked
in
the
the
raw
files,
you
could
probably
pull
it
out.
So
no,
yes,
nothing
is
encrypted.
A
D
D
A
Okay,
the
next
stop
I
had
bit
pop-up
last
week,
was
that
self-answerable,
when
doing
the
migration
is
still
continuing
to
create
realms
and
zones.
D
A
Okay,
we
a
while
ago
we've
we
talked
about.
A
Okay,
a
while
ago,
we've
talked
about
introducing
some
kind
of
a
yellow
document
to
create
those
three
arms
and
zones
in
in
the
orchestrator
we
have
said,
or
that
we
have
set
those
yammer
documents
also
already
in
rook.
A
C
D
Yep
there's
already
a
ticket
for
it,
but
don't
think
it's
related
to
adoption.
C
A
C
Yeah,
so
we've
got
the
doc.
I
think
I
mentioned
it
last
week,
and
this
is
really
a
high
level
requirements,
doc
and
the
big
picture.
What
do
we
want
on
friday?
We
did
talk
about
it
with
the
lso
team
in
the
weekly
meeting
with
them
and
the
I
think
the
outcome
of
that
meeting
is
that,
okay,
let's
kind
of
get
to
the
next
level
of
requirements.
What
are
really
really
talking
about?
As
far
as
do
we
need
raw
volumes?
Do
we
need
partitions?
Do
we
need
lvs,
dynamic
provisioning?
C
I
think
I
mean
I
think
we
do
need
all
of
that,
maybe
not
partitions
as
much
yeah,
for
example.
I
think
I'll
be
using
before
right.
I
think,
or
my
head
is
right.
Now
is
okay
if
we
do
static
provisioning
for
whatever
you
want
to
set
up
in
advance,
maybe
even
raw
discs
we'll
just
use,
static,
provisioning
and
then
dynamic
provisioning
can
just
use
lv's,
not
just
just
like
or
or
like
top
oil
vm
does
it.
C
D
C
Like
esther,
I
think
the
difference
there
just
needs
to
be
a
different
storage
class.
So
if,
if
you
want
to,
you
could
have
a
storage
class
that
says,
give
me
a
raw
device
or
you
could
have
another
one
that
says
it's:
okay
to
have
a
partial
and
and
if
the
size
of
the
whole
disk,
maybe
you
just
give
back
the
whole
disc
too
yeah.
D
Potentially
separate
storage
classes,
it
seems
like
it
has
to
be
yeah.
It's
not
it's
not
quite
the
same
because
with
an
lv
you
can
resize
like
you
could
start
with
a
big
thing
and
then
resize
down
or
up
or
whatever
or
and
create
a
snapshot
where
there's
a
raw
disk.
You
can't
do
that,
but
does
it
have
to
be
a
separate
storage
class?
Is
there
a
way
to
have
a
something
on
the
on
the
pvc?
That
indicates
what
you
want?
What.
C
C
B
One
one
thought
about
that:
okay,
I
have
read
the
document.
Okay,
I
think
that
when
we
have
a
fixed
good
description
of
what
we
we
need.
Okay-
but
probably
we
are
thinking
in
more
things
that
only
theft,
okay
and
then
I'm
I'm
a
little
bit
worried
about
that.
Okay,
because
I
think
that
in
our
case,
I
think
that
the
problem
that
we
are
trying
to
solve
is
that
to
provide
to
the
final
user
that
is
going
to
install
what
storage
environmental
cluster.
B
B
That
could
be
provide
this
kind
of
resources,
pvs,
okay
to
other
applications,
because
but
maybe
what,
if
this
is
going
to
to
simplify
the
the
design
and
the
things
that
we
need
to
that
it
could
be
good
to
have
it's
my
my
view.
C
D
D
And
if
the
pvs
are
pre-created
like
you
can
do
that
you
can
have
a
pvc
that
just
names
the
pv
that
it
wants.
But
if
it's
dynamic
like
we
need
to
have
some
way
that
we
can
have
a
pvc
that
says
I
want,
isn't
I
thought
tv
pcs
had
a
way
to
like
put
extra
stuff
in
there.
That's
like
pass
through
the
provisioner,
but
maybe
I'm
wrong.
C
D
B
A
Without
further,
that's
can't
really
continue
here.
Unfortunately,.
A
Okay,
yeah-
I
guess
I
did
for
today.
If
we
don't
have
any
other
topics.
A
Okay,
then
yeah.
D
A
Okay,
all
right,
then,
let's
sing
up
next
week.
B
D
So
the
like
super
confused
because
suddenly
my
talks
tests
are
failing
and
they're
complaining
about,
like
the
manager.
Module.Pi
function
is
missing
a
type
annotation
that
code
was
merged
in
january,
but
it
also
only
activates
if
the
python
version
is
greater
than
or
equal
to
3.8
and
I
looked
and
my
machine
just
centos
8
has
python3.6.8,
but
it's
been
that
way
for
forever.
I
just
checked
on
like
a
random
2000
machine
and
it's
also
python
3.6.8.
D
A
D
C
A
A
That's
this
one
here.
B
A
D
D
D
A
A
D
A
D
A
D
D
For
roundstiff
adm
so
that
it
returns
different
output,
it
mocks
different
output
based
on
the
command
and
I'm
confused
about
how
to
do
that.
I'm
wondering
if
you
can.
A
A
D
D
D
D
Like
you
do
this
or
whatever
you
do,
the
the
function
name,
but
there's
something
weird
going
on
here,
because
here
it
mocks.
First,
it
mocks
that
it
wants
no
output
for
it
to
do
all
this
stuff
and
then
down
here.
It
changes
the
return
value
and
it
resets
it.
But
I
don't
know
how
to
what
I
need
is
like
run
step.
Adm
dot
thing
equals
my
function:
f,
volume.
A
A
Exactly
so,
you
have
to
go
on
an
outside
effect.
It
reset
mark
resets
the
mark
to
to
a
vanilla
mark
without
any
return
value
without
any
side
value.
It's
just
an
empty
and
again.
D
In
order
to
use
that
that
code.
A
A
Not
sure
this
is
working,
that's
the
problem
with
mock.
So
do
you
see
that
this
is
a
logging.info
with
a
cmd
nope.
A
A
D
Okay
and
yeah
for
that,
for
your
for
your
test,
I
think
you
need
python
368,
that's
the
reason
I
I.