►
From YouTube: 2020-04-14 :: Ceph Crimson Meeting
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
Okay,
let's
start
that's,
we
carry
with
thirst
you're
working
on
another
tree
implementation
and
we're
trying
to
debug
in
it
and
also
reviewed
couple
crispy.
Ours.
B
That's
what
I'm
doing
right
now
I
saw
it's
essentially
it
goes
in.
There
are
three
stages
or
three
places
that
you
have
to
look
at
when
you
want
to
implement
new
functions
for
to
support
Sheila's
plugins,
you
have
to
look
at
object
class.
You
have
to
look
at
objects
either,
specifically
it
actually,
it
always
be
off
which
is
sort
of
the
dispatcher
which
was
written
by
a
part
of
it
was
related
by
by
Riddick,
and
then
you
have
the
actual
back-end
functions
during
the
actual
operation.
A
B
C
D
I
had
a
bikini
I'm,
taking
a
look
under
self
at
Amin
enter
and
if
father
some
address
has
already
worked
on
each
and
then
I
worked
on
the
remove
the
Hat
coded,
a
configuration
for
the
messenger,
the
the
the
messengerĂs
code
has
being
changed
and
the
unit
testify.
Messenger
has
verified
it,
but
I've
found
that
the
Aileen
Aiko
test
test,
any
Aiko
transit
worker
India
master
branch,
so
I
still
need
to
fix
the
any
I
call
you
need
chatter,
then
the
PAPR
will
be
ready.
D
F
Now,
I'm
implementing
the
primary
lock
PG
recover
back
counterpart
in
crimson.
Basically,
all
we
do
in
the
back
feeling
state
I
had
some
issues
with
the
Crimson
reaching
the
back.
Winning
state
I
had
to
serve
a
bunch
of
issues
and
solve
one
well.
One
of
them
has
been
actually
working
at
it
and
I've
got
a
question
about
a
bucket
our
viewing
machine
from
time
to
time.
F
This
runs
handlers
feeling
crystal
like,
let's
say,
unbuckle
reserved
in
classical
OSD.
They
are
implemented
as
basically
schedulers
events,
the
one
huge
main
of
work
queue.
So
the
code
is
actually
meet
the
code
of
handler
a
scheduled
code.
It's
actually
mutually
exclude
exclusive
with
that
with
the
one
that
that
scheduled
it
that's
not
the
case
in
in
crimson,
because
we
have
when
it
comes
to
feeding
and
recovery.
There
are
two
pipelines.
Actually
one
is
two
operations,
the
peering
event,
local
entry
mode,
but
be
sure
the
the
pipeline,
and
we
got
also
ground
recovery.
F
C
G
F
G
G
The
purpose
of
those
structures
is
not
to
it's
not
to
be
the
only
way
of
doing
ordering.
It's
just
I,
just
created
them
as
an
initial
example.
Whatever
you
do
to
make
it
better
or
to
make
it
more
complicated,
just
make
sure
that
anything,
that's
waiting
on
whatever
it
is.
You
introduce,
still
gets
registered
up
with
the
relevant
wait
with
the
waiting
thing
or
it
gets
dumped
by
the
formatter.
That's
the
important
part,
that's
the
part,
I
really
care
that
the
OSD
can
get
that
we
can.
G
F
G
Ping
me:
when
you
would
you
have
a
version
of
it
and
I'll
I'll,
take
a
look
but
yeah
I
I,
don't
I,
don't
think,
there's
any
way
that
any
of
that
is
complete,
that'll
it'll,
all
pretty
much
have
to
be
improved,
then
more
complicated,
throw
off
any
more
dependencies.
We
don't
really
want
dependencies,
but
some
of
them
do
exist.
So
you
know
we
catch.
Oh.
F
C
F
G
H
H
H
H
G
H
C
H
A
H
G
C
C
H
G
C
A
G
I
have
a
little
doodad
that
lets
me
run
unit
tests
that
use
futures
and
generally
the
sea
star
reactor
in
G
test,
because
I
wanted
that
and
I
am
implementing.
You
know
test
for
sea
star
or
a
for
C
store.
Rather
this
this
this
week.
Finally,
I
took
some
of
last
week.
Office
I
didn't
get
a
lot
done,
but
I
do
have
the
unit
test
framework
and
I'm
working
on
it.
Now
it's
straightforward.
G
G
A
C
G
To
actually
just
use
G
test
because
you
can
use
the
filter
like
G
test,
has
all
those
little
like
convenience
program
options,
things
that
lets
you
filter,
which
tests
you
want
to
run
or
whatever
it's
just
nice.
Okay,
that's
innate
value
plus,
if
you
plus,
if
you
look
at
like
the
existing
unit
test
for
C
star,
they
have
several
dozen
lines
of
boilerplate
code.
That's
a
little
bit
different
between
the
different
test
files,
which
just
bugged
me
so
I
factored
all
that
up
to
its
own
little
library.
That's
all!
Okay!.
C
I
I
Exactly
oh,
go
ahead:
yeah
actually
I'll
cover
it.
My
first
time
to
attend
this
mean
it
just
because
I
I'm
trying
to
understand
what's
a
major
assignment
you
for
their
persons
who
are
working
on
the
crimson
project
and
since
you
know,
I'm
moving
very
fast
on
my
sister's
design,
the
implementation.
So
that's
why
you
know
I
would
record
it
and
you
can't
just
keep
your
pace
to
roil
meeting
and
if
I
have
any
questions.
I
well,
I
try
to
interact.
Thank
you.
J
Almost
week,
I
was
modifying
the
pigeon
log
based
recovery
code,
kohu
and
Sam
suggested
and
I
was
implementing
a
clean
shutdown
of
their
current
crimson
OST,
but
because
this
clean
shutdown
is
kind
of
important,
this
was
a
smooth
early
run.
Some
test
code
about
recovery
and
I
think
that's
all
for
me.
E
Last
week,
I
helped
you
Han
to
dip
to
debug
and
improve
the
OST
shutdown
process
and
for
the
heartbeat
single
connection,
I
wrote
a
reaching
now
both
of
my
change,
detailed
vision,
house
in
the
PR
and
and
guy
for
the
note
Onochie
last
we
are
talked
with
food
to
decide
what
to
do
next
about
the
prototype
and
I'm
trying
to
define
the
scope
of
of
our
synchronous
prototype.
The
first
part
is
to
define
the
interface
to
define
the
interfaces
of
the
potential
users
of
of
the
Onochie
I.
E
So
so
our
current
goal
is
to
implement
this.
The
synchronous
memory
based
prototypes
also
I
also
want
to
have
been
simplified
dummy
interfaces
from
the
transaction
manager
and
the
logical,
cashy
Kesh
the
extent
to
manage
our
dummy
logical
blocks.
So
that's
the
two
parts
and
I
was
working
on
that
last
week.
Just.