►
From YouTube: 2019-09-03 :: Crimson SeaStor OSD Weekly Meeting
Description
Crimson SeaStor OSD Weekly Meeting
A
B
Thank
you:
hey
guys,
hurry
up
being
late,
it
starts,
and
last
week
I
was
working
on
via
the
come
back
to
us
to
talk
to
compare
the
performance,
the
result
against
the
gains
that,
from
the
baseline.
B
And
I
think
the
the
next
step
is
to
integrate
into
the
Jenkins
I'm
I'm,
currently
and
I'm,
really
waiting
different
approaches
to
integrate
the
screw
it
into.
You
know
github.
Well,
the
problem
is
to
link
the
the
test
allows
result
to
profile
to
to
to
show
to
show
that
the
difference
between
the
different
between
that
the
current
test,
without
with
the
peer
against
the
difference
between
the
test,
the
PRP
baseline
but
I'm,
I'm
I'm,
also
looking
at
the
checks
API
offered
by
github,
which
which
is
more
flexible
and
and
more,
is
more
most
proceed.
B
B
No
I
think
there
are
two
ways
to
put
represents
the
accessories
out.
Why
to
this
is
a
comment
as
well.
What
do
we
have
currently
when
it
could
a
metric
when,
when
they're?
Oh
here,
we
have
a
result
from
a
check
mixture
Iran?
We
have
a
new
comment
right,
which
shows
that,
if
there's
a
clear
path
to
make
shake
wrong
or
not,
but
that
github
also
offers
that
way
amid
you
might
want
to
check
check
out
a
PR
PR
web
web
page,
which
we,
which
have
another
tab
name
that
checks.
C
B
C
No
problem
for
me:
it's
currently
figure
out
how
to
how
to
make
this
writer
piece.
Rather
drawing
works
in
the
gdb
environment.
I
have
to
mine,
you
always
switch
the
ride
and
it
just
keep
the
single
thread
running
then
in
the
substrata
will
bill
will
stopped
ended
a
piece
writer.
Drawing
real
quick
will
will
come
back,
so
it's
not
very
nominal,
for
the
other
in
the
normal
environment
is
old.
When
you're
writing
GDP,
you
just
continue.
C
All
this
rider
will
works
operated
correctly,
but
in
in
my
gdb
environment
our
prettiest
ride
until
I
have
to
that
at
the
gdb
Runyon
in
singles,
ride
and
switch
to
the
to
the
singles
rider
substrate
to
run,
and
they
will
hate
that
stop
and
quit.
Then
the
peace
rider
drawing
will
come
back,
but
if
not
into
the
environment,
it
works
well.
So
it
I
think
it's
not
a
problem.
C
It's
just
in
my
gdb
environment,
so
I
can
continue
debug
on,
but
it's
not
very
convenience
because
I
have
to
manually
switch
the
engine
run
each
riders
separate
for
this
several
chance.
There
are
many
spiders.
There
were
attempts
you
needed
to
switch
and
random
itself.
I
can
eat
my
breakfast
to
do
that.
The
packet
work,
so
I'm
eternally
I'm
debugging
on
the
on
the
Opie
sequence.
Ob
sequencer
will
pass
in
the.
A
C
B
C
Yeah
in
the
environment,
it's
a
block
into
P
stretch
on
the
seams
this,
but
this
apps
right
it
not
not
to
trigger
too
well
so
I,
wouldn't
know
you're.
Treating
me
I
just
used
to
output
a
lot
and
as
I
saw
the
code
the
death
function
will
passed.
It,
though,
that
the
print
can
be
turned
off
turned
out,
though
it
another
clock
there,
but
only
he
might
you
two
be
environment,
and
only
for
this
Macedon
recruits
right
and
Eustace
and
it'll
code
running
in
that's
right.
C
Though
but
anyway,
so
it
not
interesting
environment
that
the
code
works.
It
capacitor
piece
for
Android,
so
I
have
the
output
to
the
standard
of
the
screen
and
it
works
so
I'm
still
working
on
the
other.
The
following
following
tip
at
work
and
dislike,
sometimes
some
code
that
needed
a
safe
context,
a
the
transfer,
the
we
define
if
all
deceased,
our
framework
yourself.
C
E
E
E
F
Yeah,
so
the
usual
model
for
this
is
you
do
something
like
we
maintain
a
particular
branch.
We
are
currently
working
that
he
fix.
We
have
to
make
to
see
star,
we
submit
upstream,
but
we
also
submit
it
to
our
own
branch
and
one
every
so
often
will
we
base
our
branch
on,
bears
and
remove
any
any
common
cadets
from
ours
right.
A
F
E
Okay,
so
I
will
send
a
pure
will
may
create
a
peer
in
our
system.
It's
a
different
way
of
first
creating
the
patch.
They
are
using
an
email
and
we
get
hub
and
I
will
try
to
do
both
of
them.
There
is
a
there
is
the
issue
of
they
might
refuse
the
specific
changes
that
I'm
making
that
might
create
a
problem.
They
might
refuse
the
general
idea
of
supporting
domain
sockets
the
way
I'm
doing
it,
but
we
have
I'll
deal
with
that
if
this
happens.
E
So
this
is
one
thing.
Another
question
is
a
technical
one.
Regarding
to
the
to
the
way
connections
are
accepted,
insist
on
if,
when
we
are
talking
about
the
POSIX
tack,
there
are
two
stacks.
One
is
the
real
stack
only
one
thread:
only
one
call
is
one
with
the
real
stack
which
interacts
with
the
operating
system
and
performs
the
accept
the
actual
incident.
The
actual
accept
and
all
the
other
calls
are
running
the
AP.
A
version
of
the
stack
which
is
not
does
not
interact
with
the
system
just
accepts
for
commonly
through
some
make
you
some.
E
Associative
containers,
they
are
accepting
the
connections
from
domain
from
Co
0
handing
them
now
the
problem,
the
problem
with
the
way
it
simply
it's
usually
implemented,
is
that
only
cos
0
is
distributing
connections,
but
if
cos,
0
is
currently
handing
a
connection
and
incoming
connection.
There
will
be
no
distribution
of
new
connections,
even
if
the
rest
of
the
COS
are
doing
nothing
afraid
to
handle
a
new
connection,
though
there
was
some
discussion
I
think
a
few
months
ago,
someone
raised
it
in
the
many
lists,
but
I
saw
no
real
discussion
of
possible
solutions.
E
So
what
I
did
in
this
version,
specifically
for
the
accept
of
the
listen
and
accept
for
UNIX
domain
sockets
is
I,
won't
allow
call
0
to
actually
handle
incoming
connection
each.
We
just
distribute
it
to
the
other,
a
civic
we
have
more
than
one
Co
Co
Co
will
be
left
free
to
accept
and
distribute
incoming
connections,
which
is
different
from
the
way
the
rest
of
the
system
works.
Any
comments
any.
G
E
The
accept
is
just
looking
at
the
shirt
container
and
there
are
two
containers
concur
and
the
sockets
country,
which
are
simply
who
is
the
list
of
things
of
connections
that
were
where
they
accepted
that
nobody
is
handling
them
and
the
other
one
here
is
the
a
list
of
connections
that
a
I
accepted,
but
I
cannot
handle
because
I'm,
not
cool
I
cannot
really
accept
it
because
I'm
not
called
zero,
so
we
always
need
if
I
understand
correctly,
you
always
need
call
zero
to
actually
perform
a
the
process
of
accepting
and
distributing
the
connection.
So
what's.
E
Okay,
is
that
cause
you,
okay,
continue
the
issues
if
Co
zero
is
busy
handing
in
a
specific
incoming
connection
that
might
take
it's
a
connection.
It
might
take
time
it
won't,
and
this
is
what
I'm
saying
in
the
test
it
won't
handle
the
next
incoming
connection,
even
though
there
are
many
other
pause
that
are
free
to
handle
the
new
incoming
connections
to
the
years.
Try
to
prevent
the
interchange.
G
A
B
B
The
causing
in-seat
are
not
supposed
to
be
brought
blocking,
so
it
is
a
force
problem.
I
would
say
the
good
city
should
not
block
to
knock.
Who
does
the
the
score
zero
for
a
long
time?
Otherwise,
do
it
to
consider
that
a
store,
the
core,
a
broken
core,
it's
a
it
will
be
tech
detect
by
reactor
and
another
reactor
will
be
there
a
warning
or
something
will
you
take
a
such
kind
of
a
stall.
B
E
B
A
B
A
E
E
B
G
Okay,
so
so
the
first
thing
is
is
P
is
on
the
review
of
the
fail.
They
are
over
test
and
not
so
common
for
you
and
the
problem
is
I
cannot
reproduce
the
test.
Failure
on
my
side,
I
think
it's
strange
I
have
repeated
like
a
hundred
times,
but
it
always
passing
so
so.
G
G
Also
I
think
I
still
need
to
add
more
tests
to
to
check
the
stored
connections
because
a
connection
may
be
stored
and
another
and
recovered
and
find
this
connection
is
stored
and
need
to
to
make
sure
it
is
test
passed.
Also
so
total
a
three
three
major
kinds
of
tests
and
and
I
think
after
the
these
three
seconds,
I
implemented,
most
of
the
cases
can
be
covered.
B
A
G
B
G
A
B
B
G
B
D
No
question,
but
just
a
quick
note
in
the
matter
of
of
hunting
accepting
sister.
Actually
there
is
actually
there
is
a
possibility
to
also
issue
except
directly
from
to
the
car,
to
the
operating
system
directly
from
all
cars.
It's
the
choice
is
made
by
zinc
and
POSIX
reuse
detect
of
sister
reactor.
However,
it's
turned
off
by
default
with
following
commands.
D
A
E
D
E
Oh
yeah
and
there,
if
I
think
there
are
even
worse
problems,
at
least
what
this
is.
What
I
understood
from
a
discussion
in
the
system,
meaning
list
about
it
of
months
ago,
I
won't
try
to
do
this.
I
prefer
to
do
the
way
they
they
run
most
of
the
corner,
which
is
accepting
actual
accepting
only
in
the
cozy.
Oh.