►
From YouTube: Ceph Crimson/SeaStore 2021-10-20
Description
Find more Ceph meetings: https://ceph.io/en/community/meetups/
A
Let's
see
the
multi-device
stuff
merged
last
week,
what
else
working
againstology
to
work
with
c-store
chimney
is
also
working
in
that
and
working
on
the
cash?
How
are
you
joyhot?
I
just
I
think
I
found
the
bug
in
lower
bound.
So
that's
the
first
bug
I
think.
Okay,
what
did
you
do
to
reproduce
it?
If
I
might
ask.
B
I
just
run
it
and
yeah
it
crashes
after
eight
or
nine
seconds
makes
sense.
A
A
C
I
can
use
shanghai
teams
to
solid
test
environment
currently
and
figure
out
how
to
build
the
beam
package
for
crimson
ost,
and
I
filmed
the
technology
switch
logger
that
I
I
saw.
The
system
is
set
up
and
the
printout
and
so
to
solve.
The
report
error
is
that
the
there,
the
osd
metadata
is
not
written
to
the
wireless
fosd,
this
folder,
so
for
the
blue
stone
every
metadata
will
write
a
copy
to
the
to
this
folder
to
this
directory.
C
So
so
I
just
want
to
confirm
with
you
and
just
as
well
c
star.
We
we
have
we.
We
also
need
to
support
that
feature
right.
C
Yeah
yeah,
so
so
that
will
mount
to
this
directory
and
save
the
os
metadata,
and
I
think
the
deploy
will
check
that
folder
and
yeah.
C
C
Okay,
so
that's
all
just
doing
the
cody
ended
back.
A
A
A
A
Yeah
exactly
so,
we
might
as
well
do
the
same
thing.
It's
we're
going
to
have
to
do
all
the
same
stuff.
We're
still
going
to
need
tooling,
to
be
able
to
detect
and
mount
d
store
block
devices.
So
we
may
have
used
this.
We
may
as
well
use
the
same
tricks.
Blue
store
is
using
that'll.
A
D
Sorry
keep
forgetting
that
it
mutes
me
on
entrance
or
something
after
previewing
me.
It's
not
muted,
I'm
trying
to
build
enough
interfaces
that
I
can
actually
do
like
a
separate
pg
reactor
or
something
and
feed
into
like
two
of
them
from
one
messenger
and
then
into
a
single
object
store,
or
you
know,
however,
many
numbers
we
want,
so
that
I
can
do
more
interesting
things
than
just
that,
but
I
need
interfaces
first,
that
can
speak
the
right
kinds
of
primitives
so
trying
to
make
that
happen.
D
A
B
Last
week
I
was
still
trying
to
see
how
my
implementation
of
the
sprite
lfs
strategy
in
sistor
works
and
at
first
I
I
can
run
it
smoothly,
but
I
found
I
found
that
the
segment
then
the
segment
cleaner,
just
didn't
clean
quickly
enough
to
catch
up
with
the
actual
rights.
B
So
I
I
increased
the
amount
of
data
that
the
signal
cleaner
can
reclaim
during
one
gc
cycle
from
one
megabyte
to
128,
megabytes
and
ice
came
up
and
it
came
up
with
the
the
bug
we
we
discussed
yesterday.
B
Yeah-
and
I
also
there
isn't
also
another
issue
that
I
met
during
the
test-
that
is,
we
didn't
handle
right
v.
The
function
called
correctly
when
the
the
actual
extents
tri
we're
trying
to
we're
trying.
We
were
trying
to
write
yeah.
This
is
the
iov
max
threshold,
so
I
submitted
pr
for
it.
That
was
all
for
me.