►
From YouTube: 2019-07-16 :: Crimson SeaStor OSD Weekly 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
A
And
it's
seems
it's
we
still
need
before
we
move
further.
We
still
need
to
address
the.
A
We
still
need
to
support
the
ruthless
collection
connection
and
access
from
it.
I
attend
the
meeting
with
ritek
and
and
david
galloway,
and
also
mark
and
neha,
to
discuss
the
next
step
to
to
integrate
during
great
performance
testing
into
a
jenkins,
and
I
think
we
lay
out
the
person
accounts
of
different
approaches,
and
I
think
the
next
step
is
to
allocate
a
machine
dedicated
for
performance
tests
and
then
enable
cpc
to
run
with
jenkins.
A
I
will
I
would
I
will
coordinate
with
with
with
afraid
and
david
on
this
and
that's
me:
redeck
hold
going.
B
I'm
working
on
the
cbt
side,
which
is
related
to
the
meeting
here
kiev
has
mentioned
here.
I'm
planning
I'm
doing
two
things
in
a
little
bit
right
now,
I'm
integrating
the
money
performance
monitoring
with
the
emu
statistic
guttering
for
for
cbt
with
perfstad.
However,
in
order
to
do
that,
I
want
a
variant
of
cbt
that
is
able
to
run
locally
without
an
independent
without
putting
any
dependency
on
network
things
like
ssh,
like
pdfh.
B
B
A
Actually,
I
got
a
question
regarding
to
you
your
your
approach
to.
I
think
the
idea
to
enable
cp2
to
run
without
just
just
on
a
single
machine
right,
but
in
my
mental.
B
Do
we
want
to
allocate
more?
Do
we
want
to
use
a
multi-node
configuration
for
our
junk
instant?
Stick,
I'm
something
like
osd
on
onenote
clients
on
the
second
one.
A
But
I'm
not
sure
it's
it's
viable.
B
B
Well,
an
insert
on
on
those
machines
that
we
plan
to
put
in
strings.
I
believe
we
will
have.
We
will
have
a
plenty
of
cpu
cars.
It
should
be
quite
easy
to
to
just
pin
raiders
benches
to
one
set
of
cpus,
and
always
this
two
different
one.
A
B
But
well
the
changes
I'm
working
right
now
are
not
for
well.
The
entire
idea
is
just
to
strip
the
requirement
of
ssh.
Is
the
ability
to
run
cbt
completely
locally
without
any
extra
networking
involved?.
A
B
Mostly
for
yep,
it's
just
easier,
it's
just
far
easier
to
start
developing
that
you
don't
need
to
install
ssh
pds-h,
all
that
unnecessary
stuff
that
is
okay
required
when
you
have
a
memory
is
essentially
required.
If
you
have
a
multi-node
configuration
but
on
in
case
of
testing
on
one
node,
it's
just
an
overkill.
B
The
what
is
important
for
crimson
is
integration
with
with
perfstat
it's
gathering
pmu
statistics,
I'm
doing
I'm
eradicating
ssh
before
the
development.
It's
far
easier.
B
Second
topic
is
that
I'm
getting
I'm
getting
I'm
doing
some
context,
videos
I'm
getting
back
to
the
input,
factory,
buffer
changes,
reworking
and
further
discussion.
After
reading
the
kernel
connection
multiplexer,
that's
me.
C
Hi,
do
you
hear
me
by
the
way,
okay?
Well,
I
had
two
days
to
work
on
crimson
this
week.
What
I'm
doing
is
following
sam's
a
comment
from
last
week.
C
The
new
method,
if
it
makes
sense
to
push
it
into
that
code
and
I'm
hoping
I'm
getting
along
next
week,
starting
sunday
sunday
to
friday,
I
will
be
at
a
radical
office
working
with
using
him
as
a
help
tool,
a
direct
help
in
trying
to
get
this
up
as
quickly
as
possible
if
it
makes
sense.
That's
that's
what
I
managed
this
week.
D
A
E
Yeah
last
week,
I'm
working
on
this
pr.
I
will
paste
it
in
the
chat
window
and
I
think
this.
This
is
a
prerequisite
of
the
to
implement
the
loose
list
policy,
because
it
addresses
some
issues
to
have
better
debugging
and
to
have
better
debugging.
A
handshake
happens
and
also
addresses
some
issue
of
address
exchange
between
messengers.
E
E
A
A
Regarding
to
my
my
replica
right,
I'm
actually
I
see
is
I
I
would
like
to
discuss
you
guys
regarding
to
how
we.
A
A
A
We
can
take
a
look
at
the
comments
by
redec,
starting
from
starting
with
to
do
because
I.
D
A
So
we
don't
care
about
which
is
which
which
replica
oc
is
blocking
the.
D
A
D
I
don't
feel
strongly
about
the
way
I
did
the
blockers
originally.
My
truly
my
only
goal
is
for
us
to
have
some
basic
way
of
debugging
stuff.
That's
stuck
and
getting
like
some
ongoing
information
about
what
an
ostomy
yeah,
but
obviously,
if
you've
got
the
osd
slow,
then
that
would
be
counterproductive.
A
A
D
D
B
A
D
Oh,
I
see
you
want
to
you
want
to
throw
off
one
continuation
per
logical
read
so
that
the
underlying
store
maybe
gets
a
little
bit
of
a
a
little
bit
of
q
depth
to
the
read
yeah
yeah.
My
instinct
is
that
rbd
never
does
that.
A
B
A
D
Almost
always
it's
a
little
different
for
rights,
because
you
also
need
to
do
more
than
one
thing
atomically.
D
I
could
be
lying,
though,
if
if
there
are
ever
two
reads,
I'm
I'm
gonna
guess
that
it's
not
like
two
reads
to
the
data
payload.
It's
like
an
omap
read
and
a
regular
read,
or
it's
like
a.
I
know,
rgw
likes
to
send
metadata
atomic
checks
and
then
do
the
read,
but
it
doesn't
even
send
those
back.
So
you
you
couldn't
parallelize
that
anyway,
because
you
need
to
you
need
to
know
the
answer.
D
D
D
D
D
F
D
D
By
this
name,
please
load
it
in
cache,
even
if
we're
not
processing
like
before
it
even
gets
to
the
pgq
or
whatever
pg
pipeline
stage.
What
have
you,
but
you
know
my
guess-
is
we
don't
actually
want
to
fork
off?
One
read
for
every
author
and
osd
at
least
not
yet.
A
You
know
I
sent
it
in
the
mail
to
them,
given
a
survival
list
to
continue
the
discussion,
and
the
one
thing
we
can
be
sure
is
that
rbg
does
not
do
a
component
rate,
so
it
doesn't
make
sense
to
parallelize
the
rate
in
a
single
m
object:
osd
op.
Well,
we.
D
F
A
lit
a
little
league
four
minutes
late,
sorry,
so,
finally,
I
can
build
the
blue
style
into
the
crimson
os
with
the
crimson
and
I
have
to
add,
because
in
the
chromosome
coma
and
chromium
ost
it
will,
with
the
macro
definition
with
sister
and
in
the
crimson
os
in
the
blue
star.
F
I
have
to
build
the
original
self
code,
so
I
add
a
new
macro
definition
with
alien,
so
they
they
will
go
different
code
branch
and
for
some
common
files
in
chromosome,
common
and
increments
permission
os
they
have
the
same
common
file
and
but
they
go
different
macro
definition,
so
they
will
tell
them
tell
them
multiple
definition.
So
I
added
some
name
space
to
that.
F
The
code
existing
together
the
same
function
existing
together,
I'm
not
sure
if
you
have
the
better
idea
to
build
the
blue
style
into
the
commercial
s,
but
that
is
what
I
did
now.
So
I
pasted
the
and
my
tree
here
so
this
can
be
built
with
bluestar
into
the
chromosos.
But
I
didn't
the
alien
interface
it
is.
The
interface
is
empty
now,
so
next
step
I
will
implemented.
The
alien
interface
will
call
the
from
the
c-star
thread
to
the
blue
star
thread.
F
F
F
A
How
do
you
address
the
problem
of
the
single
confliction,
because
we
have
a
two
set
of
symbols?
One
is
one
is
what
the
the
common
the
common
expecting
another
is
a
crimson
common
expecting
and
and
now
we
have
a
the
third
one,
a
third
set
of
them,
which
is
expected
by
by
by
crimson
sorry,
the
alien,
alien
blue
star.
F
Do
for
the
crimson
comma,
I
add
the
for
the
conflicting
files.
I
add
the
name
space
self
common,
so
in
that
space
they
will
build
that
name
space
and
in
the
alien
it
will
appear
the
same
as
the
self
namespace.
F
So
before
we
use
some
macro
definition
with
c
star
to
identify
which
code
branch
going
to
the
chromosome,
which
could
go
into
the
self
original
step,
and
but
currently
we
want
to
run
the
blue
star
in
the
traditional
south
thread
and
the
sister
thread
will
talk
to
the
traditional
cells
right.
So
we
have
to
build
the
blue
star
into
the
chromosos,
but
it's
running
in
the
just
like
the
existing
staff,
though
the
the
code
should
be
built,
the
two
branches
should
be
in
the
same
binary.
F
So
that's
the
problem.
So
there
are
many
build
error
and
I
just
do
something
to
fix
that.
Build
error.
F
So
if
you
have
time,
please
take
a
look
at
the
my
stuff
tree
and
currently
that
tree
can
be
built,
and
so,
if
you
have
a
better
idea,
because
I
have
to
modify
many
files
to
fix
the
confliction
either
some
aft
file
or,
if
not
defined,
something
like
that.
A
F
Yeah,
when
I
I
did
the
I
had
the
namespace
how
to
modify
many
files
because
they
are
so
common,
so
many
common
files,
they
use
the
incrimination
and
the
style.