►
From YouTube: 2019-06-12 :: Ceph Testing Weekly
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).
B
Actually
have
no
dates.
A
C
A
I've
been
working
on
a.
A
C
No
I
think
just
didn't
have
time.
We
put
a
downgrade
suite
together
just
to
get
like
into
testing
cycle,
and
you
know
I
hit
a
problem
that
you
know
we
could
not
proceed
because
we
detected
dumb
grades,
Georgians
and
Josh
needed
to
put
a
PR
for
tested.
So
we
can
test
it
out,
but
he
he
is
out.
I.
Think
he's
back.
Maybe
today
actually,
but
that's
just
the
beginning,
and
we
do
it
for
point
to
point.
C
C
A
I
think
it
was
recorded
I
think
it
was
the
CDM.
Wasn't
it
yeah.
C
C
A
C
B
C
C
Quick
fYI
I,
don't
remember,
I,
think
you're.
You
were
asking
about
like
smoke,
sweet
yep,
I,
don't
remember
what
in
what
context.
But
finally,
what
we've
done?
We
cleaned
up
smoke,
so
it's
actually
passing
right
now
and
cleaning.
You
know
mainly
man,
that
I
removed
staff
deploy
related
to
Suites
to
two
jobs.
They've
been
failing
like
forever
and
you
know,
I
asked
Josh
and
he
said
get
rid
of
it.
C
We
do
not
develop
step,
deploy
like
actively
right
now,
so
it's
actually
green
and
if
you
look
at
papito
results,
I
added
master,
plus
Nautilus
and
plus
mimic
at
the
time,
and
they
all
actually
pretty
pretty
green,
which
is
good.
We
just
as
it's
supposed
to
be
so
my
next
step
actually
is
that
I
want
to
talk
to
a
core
team.
Maybe
tomorrow
and
see,
is
it
feasible
to
not
run
Raiders
on
regular
basis,
for
name
releases
for
mimic
and
for
Nautilus
and
I?
C
Don't
know
what
they
will
say,
but
if
they
say
no,
you
know
we
will
can
discuss
what
needs
to
be
done
in
order
to
add
into
smoke.
So
we
can
actually
reduce
the
amount
of
jobs
and
also
like
increase
amount
of
head
of
checking
like
based
on
smoke,
and
then
same
question
probably
will
be
asked.
You
know
from
other
areas
as
well
but
later
on,
but.
C
B
C
B
Just
because
we
can
manage
at
least
I,
don't
see
any
I
don't
see
any
idea
how
to
manage
the
database
there.
We,
of
course
we
do
not
need
to
take
the
machines.
I
mean
they're
metal
machines
from
the
pool
of
the
main
tutorial
server,
and-
and
this
is
why
we,
but
it's
for
testing
to
tautology
and
perils
and
papito
changes.
I
understand.
B
Run
tests
on
metal
machines
on
demand
or
the
doji
if
they're,
if
we
are
sure
that
there
are
some
changes
related
to
the
hardware
stack
like
fog
relating
related
to,
for
example,
we
are
currently
on
downstream.
Only
does
it
we're
just
working
on
some
additional
mechanism
of
provisioning
nerds,
not
which
is
not
using
the
park.
We
will
add
this
support
for
extra,
so
we
will
need
to
check
it
in
the
before
merging
course.
B
A
It's
just
that
users
want
to
be
able
to
downgrade
in
case
something
went,
is
wrong
and
the
new
packages,
so
it's
not
sad
thoughts
or
anything.
It's
just.
You
have
a
life
clustering,
it
start
upgrading
you
or
you
finish
up
pretty
go.
Oh!
No!
That
was
a
bad
idea
and
you
revert
to
the
old
software
yeah.
B
We
also
had
some
input
from
our
users
on
her
during
summer
conversation
of
our
level
to
support
guys,
and
they
were
saying
that
our
customers
also
interested
in
this
feature.
But
I'm
I'm
not
sure
understand
how
what
will
be
the
outcome,
like
you
say,
just
try
to
upgrade
and
then
downgrade
and
what
connected
with
its
felt,
if
it.
If
it's
not
working
up
to,
don't
worry
to.
C
My
guess
would
be
that
you
know
you
have
a
class
consistent
of
different
components.
You
know
you
can
like.
If
downgrade
is
actually
reliable,
you
can
try
upgrade
on
several
and
see
if
it
works,
then
you
go
ahead
and
do
like
you
know
full-blown
upgrade.
If
it
doesn't
work,
you
have
a
fallback
option
to
downgrade
you
mean.
B
A
It's
we're
just
gonna
have
to
make
sure
that
when
we
do
back
forward
sort
when
we
do
our
new
point,
release
that
doesn't
make
incompatible
data
changes
like
I'm
I'm,
really
scared
of
the
amount
of
work
and
not
nothing's
impossible.
It's
just
like
it's
gonna,
be
a
thing,
but
we'll
I
guess
we're
developing
some
pretty
comprehensive
tests
that
will
do
the
downgrades
and
make
sure
that
we
have
the
self
object
corpus
dumps
for
each
point,
release
and
and
test
that
back
and
forth
in
both
directions,
and
so
we'll.
C
A
Know
I,
don't
we're
not
it's
not
gonna
do
anything
with
snapshots,
that's
not
his!
Then
you
would
like
lose
rights
or
something
in
addition
to
a
whole
bunch
of
other
complex
issues.
So
it'll
just
just
be
a
normal
running
cluster
and
like
we
already
have
the
thing
where
every
encoded
object
has
a
compact
version
and
an
incan
patent,
and
we
might,
you
know,
bump
the
compat
version
and
add
some
data
on
the
end
and
in
a
point
release,
but
we'll
make
sure
that
we
don't
ever
have
to
increment
the
ink
of
that
version.