►
From YouTube: Ceph Orchestrator Meeting 2020-07-20
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
Welcome
to
today's
orchestrator
meeting,
we
have
a
few
things
on
the
gunner
today,
so,
let's
get
directly
in,
we
have
the
we.
We
have
an
undocumented
mode,
that's
called
packets,
mode
and
self-idea,
which
which
depends
on
on
a
safe
adm
binary
which
is
installed
on
all
hosts.
So
the
problem
with
this
mode
is
that
upgrading
is
complicated,
because
it
has
very
strong
demands
on
the
compatibility
of
that
cdm
binary
because
it
if
it's
not
part
of
of
the
containerized
infrastructure,
it's
basically
outside
of
it,
and
thus
it's.
A
It's
then
shipped
in
the
version
that
the
the
distribution
packages
and
which
make
it
makes
it
really
hard
to
properly
upgrade
and
update
devadm
when
it
comes
to,
especially
when
it
comes
to
upgrading
from
-2
version,
because
at
that
point
we
don't
really
know
when
users
are
going
to
update
self-adm
itself.
So
we
have
to
keep
compatibility
with
minus
two
major
versions
and
plus
two
minor
versions,
which
is
something
that
we
cannot
provide
right
now.
A
So
the
idea
is
to
just
get
rid
of
it
for
now,
especially
that
is
not
documented
and
care
about
it.
When,
when
we
or
if
the
demands
comes
back,
then
we
have
to
or
we
we
can
try
to
resurrect
it
again,
but
for
now
I
think
it's
that
and
I
think
it's
already
dead
any
additional
thoughts
on
on
removing
the
package
mode
of
staff.
Idiom.
B
B
A
And
if,
if
adm
is
going
to
be
the
way
to
deploy
ceph,
then
we
have
to
be
on
power
with
sev.
So
we
have
to
provide
upgrades
from
minus
two
major
versions.
B
Okay,
because
that
has
done
that
yeah,
I
guess
rook,
we
haven't
tried
this
support
plus
or
minus
two,
at
least
for
what
it's
worth,
but
yeah.
Okay,.
A
B
Right
since
with
rook
people
are
honestly
upgrading
rook
much
more
often
than
seth.
So
that's
not
as
I'm
since
they're
already
accustomed
to
upgrades
regularly
and
wrote.
Maybe
the
upgrade,
and
you
know
for
steph
you'd
upgrade
to
the
next
version.
Major
version
of
seth
anyway,
before
before
the
next
one
even
comes
out.
A
A
C
One
question
why
why
this
one
of
the
things
that
you
you
say
that
is
a
pain
is
that
we
have
to
stay
backward,
cli
compatibility,
two
minutes
two
and
four
percent.
A
C
A
A
That
that's
a
big
difference
and
the
the
difference
between
the
root
mode
and
the
package
mode
and
self-adm
is
that
in
the
package
mode
we
don't
we
don't.
Instead
of
using
the
cipher
adm
binary,
which
is
distributed
within
the
container.
We
are
using
the
binary,
which
is
distributed
by
the
distribution
of
the
of
the
host
system.
A
C
A
Now
we
we
neither
have
proper
testing,
we
don't
have
we,
we
don't
have
anything
when
it
comes
to
having
a
loose
coupling
between
the
fadi
and
binary
and
the
manager
module
right
now.
The
code
just
makes
assumptions
that
everything
is
one
version.
C
Okay,
but
I
I
think
that
it
could
be
possible,
for
example,
in
the
orchestrator
to
have
a
any
kind
of
check
that
the
binary
that
we
have
in
each
of
the
host
is
the
right
one.
A
C
A
Yeah
but
introducing
a
version
flag
needs
some
minor
rear,
factorization
of
specific
thing,
with
a
proper
package
between.
C
So,
instead
of
of
what
you
said
that
in
this
moment,
we
can
do
nothing
with
this
feature
of
package
mode,
because
this
problem,
so
instead
of
telling
that
now
we
haven't,
got
any
any
possibility
to
to
fix
this.
C
Why
we
don't
go
to
to
make
this
refactorization
to
add
the
version
in
the
package
mode
and
in
the
and
and
start
to
using,
to
use
this.
These
kind
of
things.
A
Yeah,
but
which
has
strong
demands
on
any
features
we
are
going
to
implement
because
at
right
now,
if,
if
we
add
a
flag
to
the
sev
adm
binary,
we
just
assume
it's
there
and
then,
if
we
have
a
loose
coupling,
we
no
longer
can
assume
that
the
black
is
there
in
the
in
a
in
the
binary.
A
So
it
has
some
some
strong
demands
on
how
we,
how
we
develop
fossilia
and
having
a
tight
coupling
between
the
the
the
binary
and
the
manager
module
really
beats
up
development.
C
Okay,
but
in
any
case
well,
I
think
that
what
we
need
from
3d
and
binary
just
to
deploy
in
containers-
okay,
so
probably
probably,
we
do
not
need
to
know
even
the
version
and
which
is
fatm.
C
Okay,
it
is
a
good
compromising
in
this
moment
is
something
that
we
need
in
this
moment,
because
probably
we
we
do
not
have
all
the
functionality
that
we
need
in
the
in
the
the
mb9,
but
in
any
case,
the
solution
for
this
could
be
maybe
to
include
diversion
in
the
fad
binary
and
to
include
in
the
check
of
the
orchestrator,
the
health
check
that
we
have
the
the
right
versions
in
the
in
each
of
the
host.
B
A
Now
you
get
a
proper
package
and
then
this
one
here
yeah,
that's
a
tracker
actually
for
adding
a
minus
minus.
A
A
A
A
A
A
A
Yes,
exactly
so
right
now,
we
we
depend
on
users,
reading
the
cef
users,
mailing
list,
filtering
all
release
announcements
and
then
manually
scheduling
the
update
to
the
latest
version,
which
is
a
major
regression
when,
when
comparing
it
to
rpm
packages,
because
on
with
apm
packages,
you
can
just
simply
run
a
pm
update
or
upgrade
and
and
then
you're
good,
then
you're
on
the
latest
version.
We
no
longer
have.
D
A
Capability
with
selfie
m.
D
As
a
user,
you
probably
know
which
major
version
you're
running
right,
so,
if
you're,
on
whatever
nautilus
octopus
and
the
only
thing
that
you're
interested
in,
is
to
get
the
latest
version
of
that
specific
major
release
right.
So,
I
think
adding
a
latest,
which
is
only
in
scope
of
the
currently
running
major
version,
make
sense.
D
Or
let's
call
it
latest
stable
or
whatever
yeah.
D
Yes-
and
I
mean
latest,
is
always
relative
to
to
your
actually
vendor.
A
Yeah,
I
I
don't
think
that
you
can
use
cotton
laters.
We
can't
use
the
cotton
latest
tag
because
docker
io,
slash
slash,
slash,
theft
does
not
have
the
latest
stack,
so
we
have
to
be
more
intelligent
and
and
figure
something
out
that
actually
works.
A
Oh,
that
might
also
not
work
because
the
it
needs
to
be
bound
to
the
minor
release.
The
existing
current
minor
release,
yeah
right
plus
minus
some
reg
exchange.
A
A
But
but
in
the
context
of
of
kubernetes,
that's
that's
how
users
are
used
to
it
to
to
always
specify
the
the
the
correct
container
image
tag.
B
Yeah,
the
there's
been
a
lot
of
debate
too
about
the
image
tag,
because
a
lot
of
projects
do
have
like
the
latest
tag
or
yes,
like
the
ceph
images,
also
have
like
a
b15
flag
for
the
latest
octopus
release
with
rook,
though
we
don't
recommending
recommend
using
those
tags
that
are
more
general
and
can
move
with
a
different
release,
because
it
can
cause
issues
with
upgrade
where
rook
doesn't
really
know
what
version
is
working
with.
So
we
we
just
use
very
and
rick.
B
We
don't
have
a
latest
tech,
because
we
want
to
know
exactly
what
version
we're
running
and
how
we
can
work
with
version
changes
where
we
need
to,
and
we
know
what
version
we're
running.
A
Exactly
so,
if
you're
using
a
moving
tag
yeah,
you
will
easily
end
up
with
a
mixed
cluster
and
you
really
didn't
want
that.
A
Nine,
the
problem
is
that
I'm
facing
is
that
I
either
have
to
properly
support
denture,
7
and
7
and
then
find
out
how
to
document
that
problem
with
potman
on
center
7,
or
we
no
longer
run.
A
To
the
next
topic,
which
is
actually
a
sub
part
of
a
sub
or
it's
blocking
four
four
ticket,
four,
four,
nine
two
five,
which
is
the
minus
minus
version
that
we
just
just
discussed
a
few
minutes
ago,
so
I
think
from
a
separate
there
endpoint
much
and
there
are
no
more
topics
for
now.
Do
you
have
any
topics
for
self-idm
anything
that
that's
in
your
mind,
anything
that
we
need
to
discuss
today.
E
Discuss
regarding
sephiroth-
I
don't
know
if
this
is
like
really
deserving
of
discussion,
but
by
thursday
of
this
week
so
wednesday,
for
you
guys,
I
hope
to
have
the
cephadium
install
guide
up
on
the
website
for
or
at
least
in
a
pull
request
or
comment,
and
so
anyone
who
has
anything
to
say
about
it
is
warmly
encouraged
to
rip
it
apart
and
make
it
as
strong
as
possible,
so
that'll
be
in
72
hours
or
so.
A
E
Or,
should
we
just
wait
for
the
pull
request?
I
can
I
can
send
you
a
pdf
if
you
like,
but
it's
like
it's
just
a
libre
office
exported
to
pdf
pdf
right
now,
so
I'd
say:
wait
for
the
pull
request.
It'll
come
soon.
E
Sure
can
you
give
me,
I
don't
know
if
you're
on
irc
or
not,
but
if
you're
not
you
can
put
in
the
chat
here.
Give
me
your
email
address
and
I
will
send
it
to
you.
Okay,
I'm
putting
it
in
the
chat.
Now
all
right.
C
Well,
in
any
case
suck,
I
think
that
is
something
that
could
be
interesting
for
everybody.
So
if
you
can
put
something
like
a
pull
request,
it
could
be
nice
okay,
because
it
will
act,
help
everybody
to
to
to
see
the
changes.
What
you
are
proposing
and
probably
a
lot
of
people,
the
in
the
same,
will
provide
a
lot
of
interesting
feedback.
E
All
right,
I'll
I'll
I'll
spend
all
day
tomorrow
doing
it.
Maybe
I
can
get
it
done
within
the
day.
A
B
Yeah,
just
a
couple
of
general
things
coming
up,
no
real
discussion
items
that
I
have
today,
but
just
fyi
1.4
is
planned
around
the
end
of
the
month.
The
main
thing
we're
waiting
on
at
this
point
is
a
csi
3.0
release
that
we
want
to
go
out
with
so
the
anyway.
This
by
the
end
of
this
week,
we'll
plan
on
the
feature,
freeze
and
then
yeah
around
end
of
month,
then
we'll
get
the
release
out
and
part
of
that
drive
groups
is
emerged.
B
A
A
C
A
Then
enjoy
the
week
and
let's
wake
up
either
tomorrow
in
the
stand
up
or
next
week
in
the
next
orchestra
meeting.