►
From YouTube: Ceph Crimson/SeaStore 2021-06-02
Description
A
So,
let's
start
last
week,
I've
been
trying
to
reproduce
the
the
failure
reported
by
by
tarvis
who
one
who
was
trying
to
run
crimson
in
in
rook
and
turned
out
himself
to
in
two
steps
when
he
tried
to
make
advance
using
crimson
it
fails
and
ways
to
try
to
launch
crimson
using
the
device
prepared
by
classical
it
fails
as
well.
So
I
addressed
the
second
problem
last
week,
I'm
trying
to
reproduce
the
first
problem
this
week
and
that
the
first
step
is
to
to
to
have
a
better
error
report.
A
When
we
fail
when
we
are
trying
to
start
up
crimson.
So
I
come
up
with
a
fix
in
last
week
which
basically
use
the
same
methodology
used
by
cell
db,
which
uses
the
different
helpers
to
to
hook
up
the
the
cleanup
functions.
A
You
know
our
rei
way
and
also
I
ran
into
a
asan
issue
which,
which
reports
that
we
are
accessing
the
deck
out
of
bound.
But
it
turns
out
it's
a
bug
in
acen
and
I
created
the
pr
to
address
it
and
also
I
was
looking
at
some
materials
on
zfs
and
try
to
understand
how
it's
address
problem
in
good
collection
and
allocator.
B
Last
week
I
fixed
the
transaction
manager,
read
exchange,
assert
issue,
that's
because
the
segmented
state
tracker
mess
up,
because
because
the
last
tracker
right
and
not
finished
and
the
the
last
criminal
thing
stopped.
B
A
B
B
A
B
A
B
A
B
I
think
we
will
try
catch
about
the
cash
message
not
to
print
out
anything,
no
exception
message
turned
out,
so
I
just
come
out
osd
stop
in
the
with,
without
that
the
sister
before
and
I
found
that
some
error
happened
in
rba
ping
ping
said
t
destructor.
B
The
e
constructor
function
that
there
is
reserved
there
other
means
that
the
pins
should
be
empty,
but
it
seems
not
not
empty,
so
I'm
still
debugging
on
it.
It's
weird.
B
Yes,
in
the
this
day,
constructor
function.
B
B
A
C
B
Okay,
yeah,
so
I'm
still
depressed
why
the
other
happened.
C
Oh
hello,
joseph
hi,
sorry
yeah,
just
a
quick
introduction,
I'm
actually
an
intern
who
started
at
red
hat
three
weeks
ago,
I'm
in
emerging
technologies
and
well
I'm
gonna,
be
working
on
storage,
I'm
not
sure
if
it's
gonna
be
on
crimson
yet.
But
I
was
told
that
I
should
probably
join
this
meeting
just
to
get
a
feel
for
what
crimson
is
like
so
yeah
hi
sam
welcome
yep
thanks.
D
Hi
hi
everybody
I'm
working
on
scrubbing
as
usual,
I'm
I
have
an
outstanding
pr
which
feels
pretty
stable,
which
is
this
one
now
the
only
reason
that
I'm
not
really
pushing
for
having
it
merged-
and
I
would
like
your
suggestion
on
this-
this
is
this
pr
contains
three
commits.
D
Two
of
them
are
part
of
a
earlier
prs
that
are
not
yet
merged.
Now
this
those
pr's
were
fixed
a
bit,
and
now
I
have
new
commits
there.
So
what
I
have
now,
I
have
created
a
branch
with
the
new
base
commits
and
the
changes,
and
this
and
this
branch
is
tested
to
my
abilities
and
is
ready
check.
So
what
would
you
prefer
going
over
this
old
pr,
which
already
has
some
comments
or
just
me?
Creating
a
new
pr
with
the
new
updated
base
commits.
D
D
Okay,
so
I
do
that
and
then
appreciate,
if
you
have
time
to
take
a
look
at
it,
I'm
apart
from
that
the
next
pr
and
that
I'm
testing
now
is
extracting
all
the
scrubber
backend.
It
now
works,
apart
from
some
bookkeeping
regarding
number
of
whether
we
count
number
of
repairs
etc,
which
I'm
trying
to
understand,
and
some
like
I
hinted
yesterday.
I
have
some
questions
that
I'll
probably
send
an
email
to
you:
okay,
okay,
so
that's
where
I
am,
and
apart
from
that,
I'm
I've
been
reading
some
vrs
and
learned.
D
D
A
Yeah,
well,
I
like
patterns
3006.,
but
I
don't
know,
I
think
you
one
of
here
is
ready
to
to
be
merged
right.
Four,
zero,
six,
five
two
which
have
been
approved
by
josh
and
it
has
been
tested.
Can
I
can
I
merge
it?
A
A
C
Yep
I
pushed
apr
that
fixes
the
build
under
clang.
I
don't
totally
get
why
it
worked,
I'm
pretty
sure
it's
actually
a
bug
and
clang,
I
think,
with
sufficiently
deeply
nested
lambdas
you
get
to
a
thing
where
it
starts.
Failing
to
it
doesn't
matter
anyway
that
peer
fix
for
now.
Once
I
get
some
time
I'll
make
a
reproducer,
so
I
can
submit
it
to
client
upstream
but
problem,
I'm
continuing
to
add
things
to
the
interruptable
aerator,
so
it
can
actually
be
used
in
c
store
and
that's
it.
C
I'm
still
debugging
the
extend
placement
manager.
This
is
the
code
and
I
hope
I
can
send
me
the
pr
this
week.
That's
all
for
me.
E
I'm
working
on
extending
the
the
oh,
no
three
note
size
sizes
and
I
went
through
size
calculations
to
implement
the
bounded
key
value
sizes
and
the
there's
a
pr
to
extend
the
lba
lookup
interface
and
it
has
been
merged
last
week.
That's
all
for.