►
From YouTube: 2019-08-06 :: 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).
B
A
C
C
A
C
A
C
A
C
Testing
my
replica
right
here
using
engines
working
progress,
useless
connection,
know
the
connection
support
and
with
retakes
CBT
branch.
It
can.
You
can
imagine
last
night,
so
it's
worked
great
and
it's
not
introduced
performance
regressions.
So
I'd
like
to
get
it
in
merchants.
They
said
but
I'm
not
sure
if
I
should,
because
that
the
the
it
needs
the
support
up
from
its
need
to
dilute
the
connection
loose
lips
connection
support
in
crimson
messenger.
C
A
C
Not
really
it's
a
you
know,
replica
right,
yes,
that
submits
the
transaction
to
to
to
to
more
remotes.
So
it's
what
I
can
see
is
that
the
not
hurt
sir
performance
Ennis
significantly
I,
think
and
did
if,
if
I,
if
I
see
the
mean
size,
I
decide
to
one,
the
performances
are
almost
the
same
as
it
was
so
I
think
performance
wise.
If
the
PR
is
a
joke.
Okay,
that's
that's.
What
is
the
one
thing
I
would
all
I
was
concerned.
A
C
No
I
didn't
compared
it
with
with
a
class-d
compared
to
the
performance
of.
A
C
I
believe
the
criminal
before
the
change,
the
performance
of
a
chromosome
LT
also
perform
from
that
classically.
So
across
those
these
then
there's
our
eyes,
our
in
no
commutator,
so
I'm
not
talking
needed.
What
I'm
I'm
war
I
was
worried
is
that
at
the
world
the
performance
of
the
Crimson
II
will
talk
because
of
some
some
some
Pia.
C
That's
why
that's
that's
also
listing,
as
the
next
thing
I
would
talk
about,
because
after
that,
silicon
wafer
we've
been
working
on
integrating
the
performance
test
into
Jenkins
and
Ritter
had
rated
already
already
did
a
TD
site
and,
and
the
criminality
is
pretty
much
container
to
just
a
poor,
the
readers
bench.
So
what
is
what's
missing
is
let
me
poster
the
Kanban
you
are
you
chart
along
holistic
per
chips.
C
C
The
performance
taking,
incidentally,
and
other
things
too,
to
write
a
small
SMO
toothpaste
on
CBD
test
without
the
two
to
see
if
it's
a
much
worser
o-o-or
on
par
ways,
the
baseline
performance
and
then
we
need
to
work
on
who
comes
tip
safe
builder,
to
it
added
job
to
crews,
cool
them
up
to
two
shirts
to
shoulder
to
return
the
to
show
the
retirees
out
on
github
and
I
think
that
that'll
be
next
step
and
then
I
will
I
will
try
to
come
up
with
some
some
some
test
cases
and
and
the
wrong
wrong
with
some
of
our
recent
changes.
C
For
example,
inching
star
is
use
tests.
Oh
sorry,
innings
changes,
messenger
and
my
my
replica
right
to
verify
it
and
I
said
from
eight
I
may
need
it
to
not
manage
it,
but
that
I,
partially
port
rock-beetle
sister
and
the
tedford
story.
The
first
test
case
the
past,
so
I
think
that
the
approach
is
workable,
functional
functionality,
wise,
but
I
wish
you
have
a
like
Suri
support
problems
ahead,
wise
to
to
make
sure
that
it
is
performance
can
at
least
on
par
with
the
Pollack's
version.
C
A
second
is
to
make
sure
they
can
be
upstream
meter
by
facebook
as
him
and
I
I
post
a
post
on
Google
or
sorry.
A
face
book
discussion
group
and
one
of
the
maintainer
replied
to
me
that
it's
the
proaches
looks
great,
but
it
cannot
promise
that
the
change
will
be
reviewed
permanently
so
I
think
I,
probably
I
will
then
send
the
preliminary
PR
to
top
trim
to
get
it
reviewed
and
get
more
attention.
But
I
need
to
some
to
some
a
performance
test
against
it,
but
I
do
think
that
remain.
A
C
Like
condition
variable
and
the
Sun
is
not
it's
not.
We
have
not
a
rewrite.
We
write
them.
We
write
on
the
portal
using
sister
but
I
think
the
the
building
block
is,
you
know
the
building
block
of
blue,
so
I
just
rocks
TV.
So
once
we
we
have
function
walkable,
we
have
a
working
rocks
deep,
ESRI
the
rocks
DB,
and
we
can
we
once
we
ensure
that
its
performance
is
good
or
on.
How
is
that
with
with
food
extrusion
we
need
to
think
about.
C
C
C
C
A
So
why
why
there
are
difference
between
the,
for
example,
hi,
Peter
disk
and
there's
no
speakers.
What
what's
the
difference
for
our
coding.
C
E
Exactly
because
it's
because
the
huge
seek
time
of
of
spinors,
if
you
need
to
move
the
record,
the
head
somewhere
to
point
some
more
some
more
other
location
on
disk,
it
takes
a
lot
of
time.
That's
not
the
case
of
sequential
right
and
blue
star
and
to
a
current
or
SD
can
see
common
are
doing
a
lot
to
liner
to
linearize
those
rights
we
are,
we
are
burning.
E
A
A
D
A
C
C
A
C
A
A
C
A
D
E
I'm
working,
the
main
focus
is
implementation
of
safe
classes.
Object
classes
for
crimson
I
have
I,
have
a
branch,
it's
almost
ready
at
the
moment.
Well,
the
first
stage
is
almost
ready
at
the
at
the
moment.
I'm
passing
all
tests
from
cirrus
hello,
test
test
cases
except
CLS,
hello
filter,
which
is
about
filtering
during
PD
reason,
I'm
implementing
this
missing
parts
right
now
and
fortunately
we
have,
and
hopefully
we'll
have
a
PR,
maybe
today
or
tomorrow,
this
will
be
the
first
stage
and
the
second
one
will
be
just
moving
to
our
BD
related
test
cases.
C
E
At
the
beach
it's
postponed
I
plant
I
plan
to
talk
I
want
to
reiterate,
basically
to
avoid
effect,
having
proposed
things
that
will
affect
the
performance
of
the
PDK
based
messenger,
so
we
would
improve.
We
would
sacrifice
mule
trader
improvements
for
posix
stock
at
the
price
at
the
at
the
cost
of
making
the
PDK
performance
worse.
That's
because
advanced
not
only
to
have
a
unified
interface.
No
ifs
like
are
we
go,
is
DP
decay
that
then
a
otherwise
go
another
way,
everyone's
also
other
ones.
E
E
C
E
Exactly
but
I'm
afraid
we
need
to
segment
I,
don't
those
requirements
I
basically,
are
basically
contrary
and
we
we
need
to
drill,
sacrifice
something
or
perfect
performance
or
uniformity,
and
personally
I
bet
that
performance
is
much
much
much
more
important
persist
that
that
nobody
would
even
think
about
employing
ste
start
because
of
because
it's
I
don't
think
that
anybody
employed
sister,
because
it's
nice,
because
it's
not
it's
done
T
star-
is
only
because
of
performance.
That's.
E
D
E
C
E
Input
power
factories
at
the
moment
is
the
postponed
I
plan,
I
plan
to
get
back
to
it
and
reiterate
the
best
idea
will
be
to
meet
avi
face
to
face.
We
had
internal
in
front
important.
We
had
a.
We
had
a
lot
of
discussions
of
input,
buffer
and
I
I.
Think
I
really
have
I
really
have
reasons
to
think
that
Abby
still
misunderstands
the
the
concept.
E
B
Thank
you.
We
said
that
we
said
we
write
it
together
and,
let's
do
it
I
forgot
about
it.
Okay,
we
do
have
to
write
something
that
might
convince.
Are
we
I
have
to
leave
in
five
minutes
so
I?
Just
sorry,
if,
ok,
if
ok
I'll,
just
tell
you
what
what's
going
on
and
I
live
and
I
added
the
support
for
the
server-side
of
UNIX
domain
sockets
to
see
stuff
and
I
have
to
remind
you.
B
The
idea
was
to
support
the
admin
interface
that
goes
through
UNIX
domain
socket,
it's
a
love,
for
example,
to
change
the
logging
level,
which
is
what
I
needed
and
what
I
wanted
to
support.
So
the
first
thing
was
to
support
to
add
native
support
for
listening
and
acts
in
the
getting
data
of
a
UNIX
domain
sockets.
B
B
Now
what
I
have
now
is
that
it's
working
apart
from
a
few
slight
problems,
the
one
is
that
I
got
I,
probably
got
the
call
attachment
one
I
think
inking.
Is
they
probably
asked
using
help
with
this,
and
what
I
have
now
is
that
if
I
have
multiple
CPUs
and
only
if
by
chance,
I'm
connecting
today
I'm
you,
the
client
uses
the
UNIX
domain
socket
that
is
listened
by
the
specific
CPU
will
I
get
a
real
working
connection.
B
C
B
C
B
D
B
And
when
I
will
eradicate
and
we
I
was
trying
to
change
the
log,
the
log
level
during
comparation
and
I
wasn't
being
able
to
do
so.
So
this
is
a
separate
and
they
couldn't
self
contained.
The
issue
I
could
tackle
without
understanding
too
much
about
more
than
I
know
about
the
USD
just
by
introducing
it
to
system
now.
C
D
D
The
existing
async
test
intercept
both
sides
when,
when
doing
tests
with
with
part
with
different
policy,
combinations
and
different
for
cases,
but
we
have
two
new
combinations
like
to
combine
the
rings,
OH
and
I
think
test,
and
currently
we
we
can't
make
it
running
in
the
same
process
because
they
are
using
the
same
common
library,
so
I
designed
a
different
way
to
wrong.
If
it
is
a
sync
messenger,
I
run
it
in
a
separate
process
and
only
intercept
and
evaluate
one
side
and
which
is
always
the
Crimson
messenger.
D
Just
intercept
Quinzel
and
they
evaluate
if
crimson
messenger
works
properly
with
with
the
peer,
the
pure
can
be
a
crimson
messenger
or
async
messenger,
and
also
the
test
code
can
be
simplified
with
with
this
one
side:
implementation,
because
we
don't
need
to
implement
the
same
interception
and
evaluation
for
async
messenger
and
and
but
it
still
needs
some
implementation
to
notify
the
pierside
to
connect
the
test
side
as
a
client
and
set
different
policies
and
the
restart
to
as
a
different
as
a
clean
new
test
suit.
So
comment.
D
Come
on
and
I
think
it
will
be
simple
if
we
just
need
to
start
if
it
is
crimson
messenger
here
we
just
run
it
in
the
same
in
same
program
and
it
is
a
default
unit
test
and
if
we
want
to
test
with
a
Singh
messenger,
we
just
start
another
process.
Maybe
with
a
script
or
a
single
squared
to
start
it
and
it
will.
The
tests
will
understand
it
and
when
test
is
finished,
if
it
can
ask
the
pr2
to
shut
down
I
think
it's
just
a
simple
command
to
start.
Okay,.