►
From YouTube: Ceph Crimson/Seastore Meeting 2022-06-07
Description
Join us weekly for the Ceph Crimson/Seastore meeting: https://ceph.io/en/community/meetups
Ceph website: https://ceph.io
Ceph blog: https://ceph.io/en/news/blog/
Contribute to Ceph: https://ceph.io/en/developers/contribute/
What is Ceph: https://ceph.io/en/discover/
A
All
right,
let's
get
started,
let's
see
this
week
for
me,
was
mostly
reviews
and
I'm
getting
back
to
work
in
multicore
aravind
how's
it
going.
B
Yeah
so
yeah
I've
been
looking
into
the
the
zns
part
of
things
last
week
and
I
have
identified
a
few
issues
right
now
and
I'm
working
on
them.
I
also
did
some,
you
know
clean
up
of
the
log
messages
which
were
a
little
bit.
You
know
it
was
logging
most
of
the
info
and
debug
messages
as
error
messages,
so
kind
of
changed.
B
All
of
those
things
I
fixed
the
the
capacity
reporting
that
one,
so
I
I'll
send
a
pr,
probably
today
from
the
jet
fork
that
I
have
so
I
request
any
of
you
to
please
review
that,
and-
and
I
I'm
still
you
know,
along
with
these
patches-
I'm
still
trying
to
understand
the
interface,
how
it
is
interfacing
with
the
higher
layers,
and
all
of
that
I
am
getting
a
good
idea,
but
still
not
saying
100
there
yet.
B
So
I
think
that
was
a
quick
update
from
my
end.
Okay,
how
are.
A
C
Yeah
last
week
I
went
over
the
cash
code
and
added
some
fixes
for
review
and
for
the
generation
no
gc,
I
replace
the
code,
and
I
find
out
that
fio,
zip,
zip,
f
mode
might
be
more
appropriate
to
add,
generate
workload
to
test
gc,
because
it
will
the
probability
to
to
hit
some
blocks
will
be
not
even
in
this
mode.
C
And
also
I
added
the
matrix,
I
did
some
improvement
to
the
matrix.
That's
all.
A
A
A
Yeah,
so
in
more
realistic
systems,
what
you
do
is,
you
would
add,
a
sampling
system
to
well,
in
this
case
c
store
example,
the
actual
lbas
that
are
being
being
written
to
and
then
use
that
to
tune
the
zip
f
parameters
to
do
synthetic
testing.
But
since
we
don't
have
that
kind
of
data,
and
even
if
we
did
it
would
be
highly
worklet
dependent.
I
would
pick
I
would
go
find
a
paper
that
evaluates
pretty
much
any
recent
block,
block
implementation
and
just
copy
whatever,
whatever
synthetic
workload
they're
using.
A
The
other
option
is,
you
could
test
several
parameters
and
then
present
the
results
as
a
function
of
how
much
of
the
lba
space
falls
into
the
80th
percentile
or
something
does
that
make
sense.
B
A
I
don't
have
a
good
sense
for
what
I
would
expect
to
see
in
our
workloads.
Actually
do
you
have
any
customer
like
do
you
have
any
internal
customers?
You
could
deploy
testing
on.
A
Okay,
that'll
be
something
to
look
into
later
on
once
we
have,
and
we
don't
have
to
do
it
with
cs3
they're
doing
with
blue
story
just
just
fine
or
even
doing
it
in
the
rbd
client.
A
Something
to
think
about
chad,
knight
how's
it
going.
D
Last
week
fixed
a
rollback
issue
and,
let's
just
small
pr,
it's
called
watching
and
just
I
I
got
the
comments
for
the
error
log
and
we
are
working
on
the
comments
to
modify
the
error.
Log
appear,
I
will
appear
and
redact.
I
have
sent
you
an
email
about
the
historic
yeah.
E
So
I
responded,
I
think,
I'm
I
think
that
the
reason
behind
behind
doing
doing
the
creation
of
intrusive
pointer
without
adding
to
the
rif
count.
It's
is
sorry
for
for
preserving
for
guaranteeing
that
there
is
at
least
one
reference.
Even
if
the
operation,
if
the
client
request
is
out
is
is
handled,
is
finished
and.
E
It
and
only
the
history
registry
is
its
soul
user.
Well,
I'm
surprised
that
it.
It's
that
you
are
seeing
use
after
free
in
the
client
request,
part
we'll
take
a
look
on
that.
D
Okay,
thank
you
and
I'm
still
debugging
on
the
regressive
bug
and
the
before
it
passed
the
test,
but
currently
still
has
an
error,
I'm
not
sure
which
pr
introduced
in
the
the
back
in
the
c
raid
operation
and
still
debug
on
it.
But
that's
all.
E
Hello
not
much
for
crimson
last
week
mark
this
week,
I'm
working
on
bringing
back
the
buildability
of
crimson
on
jenkins
on
our
ci
cd
right
now.
I'm
experiencing
issues
with
system
d
libraries
with
udef
what's
interesting,
is
that
on
the
right,
after
only
after
upgrading
the
the
tool
set
to
gcc
11.
E
A
F
Last
week
I
made
several
pr
ready
to
merge
and
I
also
did
some
further
proof
tests
for
for
the
c
store.
The
result
shows
that
the
the
journal,
trimming
duration,
is
significant.
F
Okay,
I
mean
my
test.
My
prof
test
shows
that
the
journal
trimming
transactions,
duration.
F
Oh
sorry,
I
got
a
poor
connection
here
I
mean
the
difference
on
the
general
trimming.
Transactions
duration
is
not
caused
by
it's
not
caused
by
difference
in
the
systar's
kill
length.
I
will
do
more
to
try
to
see
why
there
is
such
difference.
A
All
right
so
I'm
out
next
week
I
may
see
email,
but
I'm
doing
some
personal
stuff.
So
I'm
not
precisely
sure
how
much
time
I'll
be
able
to
spend
on
work.
We
will
see,
but
I'll
be
back
the
week
after
that,
all
right
is
there
anything
else.
B
Yes,
some
one
question:
is
there
anything
like
any
requirement
before
sending
out
the
patches
like?
Is
there
anything
like
you
know
regarding
licenses
or
anything
like
that?
B
B
Patches.Rst,
it
did
say
about
signing
yours,
but
I
I
mean.
Obviously
I
signed
that,
but
anything
other
than
that.
A
No
try
to
match
your
style
to
existing
code.
All
the
usual
boring
stuff
for
committing
to
an
open
source
project
do
try
to
break
up
your
pr's
based
on
what
they
actually
do.
A
Smaller
prs
review
faster.
Basically,
yes,.
C
And
you
can
just.