►
From YouTube: 2019-04-23:: 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
At
the
same
time,
any
time
I'm
was
working
on
integrating
sorry
in
consolidating
the
v1
as
an
occasion
with
v2
but
the
but
the
classic
mess
v2
v1
messenger
is
still
too
under
deal
working
in
progress
passage,
though
I
will
continue
working
on
it
and
probably
will
also
work
on
the
red
past.
We
discussed
last
last
last
and
up-
and
that's
me
me
how's
it
going.
B
Yes,
so
I
try
to
set
up
the
craft
server
test
environment
and
my
newly
set
up
the
self
cluster
if
I
killed
at
the
surface
T
in
the
restart
is
a
feisty
and
the
riddles
bank
cannot
connect
it
with
the
surface
T.
So
even
for
the
Southwest,
not
crimson,
ste
I
can
to
make
them
connection.
So
I
just
give
up
just
wait
for
you
to
finish
the
the
right
path.
Then
we
need
Intel
start
a
sec
wise
you.
B
First
we
can
use
crimson
as
the
tested
directory
so
give
up
the
cross,
the
road
test
environment
and
then
I
continue
on
the
demon,
supportive
for
the
configure
file
under
the
command
line
for
the
night
options
and
the
configure
file.
Sightings
and
I
can't
I
found
that
I
can't
use
our
configure
proxy
outside
the
sister
framework.
Before
the
sister
framework
start
up.
We
couldn't
use
the
the
configure
proxy.
We
decided
and
also
I
can't
use
to
originally
configure
proxy.
So
I,
just
separated
leti
find
that
the
configure
Y,
alleles
and.
B
Configure
process
so
I
just
used
to
configure
while
use
and
define
a
configure
I'm
the
configure
and
just
use
that
one
to
call
the
configure
Pascal
your
file
and
the
pass
pass
to
configure
value
as
the
parameter.
Then
I
transfer
the
configure
value
to
the
crimson
configure
proxy
as
initialize
initialize
the
values
so
I'm
just
evaluating
a.
B
C
B
B
A
B
A
C
C
B
Okay,
no
problem,
so
when
I
finish,
the
the
demon,
debugging,
yeah
and
I
can't
go
back
a
to
to
find
out
to
debug
it
or
what
happened
and
another
one
is
that
in
--see
and
the
chimp
on
sent
out
there
a
bottom
eyes,
the
test
result
so,
for
instance,
optimized
optimize
patch,
the
remote
cast
I
mean
to
Hofstra
tester
and
seems
that
the
criminality
is
in
his
test
case
is
better
than
I.
Think
the
sync
messenger
premise:
messenger
is
better
than
a
sick
messenger,
but
the
his
chat
is
simple.
B
Only
one
I
think
only
one
connection
and
there.
So
so,
if,
if
you
have
time
so,
please
tester
the
Crimson
OH,
oh,
we
couldn't
test
a
Crematory
across
the
server
to
hostage
so
that
yes
I
want
to
do
so.
It's
no
problem
and
there
so
I
will
ask
a
temple
to
test
a
more
complex
cases,
for
example
multiple
connections
to
verify
in
his
chest
a
result.
B
B
B
And
yeah
adjusted
to
want
to
find
out
the
why
a
singer
messenger
in
the
local
test
environment
is
worse
than
crimson
messenger.
So
they
are
the
chances,
some
configurations
increments
in
a
single
messenger
and
so
before
the
Cinco
messenger
is
worse
than
crimson
messenger.
But
after
the
configuration
it's
the
similar,
they
are
the
similar
performance.
B
D
D
The
plan
reused
future
state
of
promise
as
often
as
possible.
If
there
is
a
if
there
is
a
future
and
created
from
a
given
promise,
it
will
be
always
the
promise
will
always
use
the
space
in
inside
future.
The
well
is
this
quiz,
not
only
conditional
jump,
an
extra
load
that
my
pubs
did
but
also
even
kill
the
last
memory
in
directional,
but
that
was
there
it's.
D
In
short,
it
steams
I'm,
going
to
give
it
a
test,
but
it
is
be
a
background
job
on
core
of
sistar
anchor
of
crimson
I'm,
sorry
I
started
work,
I
started
working
on
the
very
basic
rock
apart,
includes
always
be
by
very
basic
I
mean
no
replication
means
all
mean
size.
Equal
one
note:
eg,
lock
recovery,
all
that
stuff
I.
A
D
On
basic
right
path,
inside
criminal,
okay,
basic
rights
by
basically
I
mean
no
replication,
no
PG,
lock,
etc.
I
guess
we
are
in
huge
debate
about
the
log
based
recovery
in
Barcelona.
Definitely
mark
will
have
some
opinion
about
pity,
based
recovery.
You
sure
about
that.
I'm
going
just
be
very
simple
case.
Just
or
just
right,
over
replicated
pool
with
minerals
with
mean
size,
equal
one
I
will
be
focused
just
on
ordering.
Basically.
E
Week,
hi,
yes,
hello,
hi,
sorry
for
being
late,
not
yet
I'm,
starting
working,
trying
to
and
trying
to
get
things
set
up
and
yesterday
ie
first
time
I
compiled
the
the
crimson
version
of
the
OSD,
but
I
got
from
my
bad
stuff.
I
got
a
list
of
commands
that
might
help
the
last
one
failed
by
the
way
which
why
today
again
to
see
why
I
can't
see
restore
the
system
with
the
new
with
Crimson
OSD.
C
Alright,
so
I'm
getting
test
runs
that
look
a
little
bit
pretty
much
like
everyone.
Else's
master
runs
on
mirrors,
which
is
to
say
a
staggering
number
of
failures,
but
mostly
due
to
very
silly
warns
in
the
logs.
So
pretty
okay,
so
it's
converging
I'm
getting
close
tomorrow,
I'm
gonna
I
have
a
bunch
of
fixes
that
I
haven't
pushed
to
the
branch.
That's
in
the
PR,
I'm
gonna
merge
those
down
into
their
respective
original
cadets
rebase
and
push
and
then
it'll
start
people
more.
Like
you
know,
you
ready
for
review,
hey
I,
guess.
C
In
the
meantime,
I've
been
working
sort
of
whenever
I,
don't
have
sort
of
test
runs
to
look
up
on
integrating
into
crimson
I've
gotten
a
little
further
I
needed
to
come
up
with
like
a
shim
layer,
so
that
this,
the
central
log
users
in
peering
state,
don't
need
to
talk
to
a
central
log
that
doesn't
exist
in
crimson
are
all
so
of
them,
creating
some
abstract
class
sort
of
interfaces
to
deal
with
that
their
stubs
in
crimson
right
now
they
don't
do
anything,
but
that's
okay.
We
can
figure
out
what
we
want
them
to
do.
A
C
Sorry
I'm
at
PG
log,
the
the
canto
K
so
that
the
like
basic
operating
principle
of
ratos
that
each
PG
operates
as
a
ordering
for
a
log
right.
That's
fundamentally,
what's
what's
going
on
the
fact
that
it's
used
for
recovery
is
sort
of
optional,
but
you
basically
always
want
to
do
that
with
that
sort
of
design
and
then
backfill
is
a
whole
separate
thing
that
doesn't
have
to
be
talked
out
here.
C
It's
like
that
that
thing
that
knows
what
the
ordering
requirements
are,
and
that
knows
like
how
to
deal
with
the
log
you
use
to
talk
to
other
PT's.
That
thing
makes
sense
as
a
thing
that
isn't
the
PG
back
end
and
then
several
you
have
a
PG
back-end.
That
knows
the
difference
between
replicated
PJs
and
ECGs.
That
to
me
makes
sense
and
that's
a
sound
thing
to
carry
over.
We
just
want
to
be
careful
as
we
populate
that
API
to
do
a
better
job.
That's
all!
C
Okay,
yeah,
so
I
think
that
old
concept
makes
sense
and
we
definitely
do
not
need
if
the
PG
primary
log
PG
split.
That
was
really
dumb
for
wait.
No
wait!
Ichi
yeah
primary
log,
P
G,
yeah,
okay,
original
classic
list.
He
has
those
two
files
for
no
good
reason.
C
That
is
PG
and
primary
log
PGCC.
Those
two
things
are
there,
because
once
upon
a
time
there
was
a
raid
for
PG
that
didn't
work
like
a
long
long
time
ago
anyway.
So
in
other
words,
no
I
think
the
broad
structure
is
okay,
but
you
know
pull
two
reviews
and
like
watch
it
as
it
develops
and
think
about
whether
it's
doing
what
we
need
it
to
do.
Okay,
and
also
we're
not
doing
a
racial
coding
up
so
other
than
kind
of
vaguely
thinking
about
what
it
will
mean.
C
I,
don't
know
that
there's
much
we
need
to
do
about
it.
Yet,
as
far
as
marks
thing
about
logging
or
about
the
marks
thing,
if
I
have
understood
his
argument
correctly,
is
that
we're
spending
a
staggering
number
of
CPUs
dealing
with
this,
like
rotating
bit
of
metadata
for
figuring
out
after
a
crash,
what
operations
happened
recently
and
yeah?