►
From YouTube: Ceph Crimson/SeaStore 2021-03-10
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
A
I'm
sure
has
worked
to
improve
the
alien
store
performance
by
starting
the
the
task
queue
and
the
threading
by
by
pjid
and
currently
still
reviews
them
part,
two
pr
to
wire
up
the
object,
store
and
extend
a
story,
the
systole
and
the
object,
omap
domain
manager
and
extend
map
manager.
That's
me.
A
C
B
Npd's
segmentation
report
and
the
founder,
it
is
caused
by
my
garbage
collection
and
that
mess
up
the
rb
tree
and
I
found
the
sum
has
already
fixed
some
bugs
in
a
garbage
collection.
So
that
should
also
solve
the
problem
in
the
in
the
in
my
past.
So
I
will
wi-fi
it's
a
return.
A
I
didn't
quite
follow.
You
so
seems
like
the
change.
B
Has
has
fixed
some
issues
in
the
pc,
so
he
said
that
should
fix
the
segmentation
fault.
C
C
I
spent
some
time
developing
a
way
to
reproduce
each
behavior
chutney
saw
mostly
it
was
a
pro.
It
was
a
minor
bug
in
the
lba
manager's
lower
bound
or
lookup
implementation.
It
was
just
doing
flatly
the
wrong
thing,
so
I
have
a
fix
for
that.
I
also
found
a
bug
with
how
the
garbage
collector
is
using
the
journal
cursor.
So
I'm
going
to
have
to
fix
that
as
well.
C
D
E
I
I'm
fixing
issues
from
pakistan
in
the
honor
3
and
I
think
the
issue
is
now
fixed
and
others
I
have
some.
I
have
some
reviews
and
discussions
to
support
multiple
devices
in
this
store.
C
C
B
E
And
okay,
so,
okay,
so
insert
all
node
and
the
mutate
or
node
right.
C
E
E
Okay,
so
so
do
you
think
if
I
add
more
transactions
in
in
the
be.
B
C
E
C
E
C
We
could
talk
about
it
more
if
you
want
it's
just
that
the
ethyl
tree
code
is
really
complicated.
We
need
better
unit
testing,
I'm
open
to
the
notion
that
your
existing
unit
testing
is
enough.
E
B
C
C
Okay,
I
believe
you
it's
just.
I
find
that
good
to
be
difficult
to
interpret
especially
the
unit
tests.
Well,
all
of
it
really
okay!
Well,
if
that's
true,
then
that's
good
that
addresses
my
concern,
so
the
other
part
would
just
be
filling
in
reasonably
comprehensive
tests
at
the
ethel
tree,
node
manager,
level.
E
A
A
By
the
way,
I
also
spot
the
the
timeout,
the
restoration,
the
the
failures
caused
by
chris
messenger
unit
test
every
every
node
and
then
even
we
after
the
timeout.
A
E
So,
while
that
should
be
stopped
should
be
killed
by
the
120
seconds.
A
No
not
likely,
actually,
I
tried
to
debug
the
the
existing
process
which
failed
to
exceed,
but
the
back
trace
doesn't
show
anything
it
all
is
it
showed
is,
is
question
marks
which
doesn't
have.
I
guess
it
was
the
it
was
a
mango
or
something
it's
so
crappy
or
something
I
I'm
not
sure
some
of
them
have
been
running
for
like
weeks
or
so
since
the
last
month.
A
A
A
A
Let
me
find
the
tracker
ticket
some
of
the
some
of
the
some
of
our
crimson
unit
tests
never
ends.
A
C
B
A
C
C
C
Okay,
so
what
I
mean
is
if
the
unit
test
corrupts
the
address
space
or
something
it
might
fail
to
crash,
it
might
just
run
forever.
In
other
words,
it
might
fail
to
trigger
the
c
test
timer,
so
yeah
timeouts
embedded
within
the
process
that
runs
the
unit
test,
are
not
as
reliable
as
timeouts
elsewhere.
So
embedding
one,
for
instance,
in
the
script
that
jenkins
runs
will
be
more
reliable.
A
C
A
B
Yes,
I
have
finished
the
test.
C
C
B
B
West
is
yeah
device
is
too
big
for
me
now.
It
took
me
a
long
time
to
meet
that,
to
make
it
a
dressful.
C
Well,
I
think
it'll
just
never
hit
it.
The
way
you
have
fio
set
up.
It's
only
ever
gonna
write
about
half
the
total
size
of
the
device.
The
working
set,
that
is,
will
be
half
the
total
working
set
size
of
the
device.
So
if
you
set
it
to
one
terabyte
of
disk,
I
believe
fio
is
only
set
up
to
use
512
gigs
of
it,
so
that
c
store
will
always
have
half
the
disc
available
to
it.
This
would
happen
if
you
set
fio
to
use
like
900
gigs
or
something
this
isn't
really
important.