►
From YouTube: 2020-10-21 :: 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
Yeah
I
start
that's
week.
I
was
reviewing
horns,
part-2
editor.
Let
me
soak
it
and
you
miss
PR
to
to
edit
alien
night
the
gusto
it's
her.
The
Aleppo
story,
pretty
close,
it's
just
I,
said
some
from
neat:
it's
ready
for
merge,
I,
guess
and
also
reviewing
retakes
change
to
bring
the
word
was
not
apply.
A
Support
to
to
crimson
also
I
was
reading
the
the
RFS
source
code,
inconnue
5.5
and
related
text
like
that
paper
on
PTR
vests,
also
I'm
I'm
I'm
working
on
the
memory
designing
for
for
the
for
Commission
I'm,
sorry
folders
or
the
Petrie
support
in
crimson,
I'm
posting
the
URM,
your
chat
window
assembly.
If
you
get
a
chance
you
can
break,
you
could
take
a
brick
quick,
look
at
it
to
see
if
there
is
something
terribly
wrong.
Whatever
that
word
from
your
Georgie
line,.
A
B
C
A
D
A
D
A
E
C
A
E
Last
we,
the
most
important
thing
from
last
week,
was
our
discussion
with
avi
and
serve
death
in
the
matter
of
our
rights
performance.
I
think
we
explained
why
we
really
need
to
differentiate
the
behavior
regarding
to
native
versus
plastic
stacks
of
sISTAR
and,
however,
I'm
still
waiting
for
response
from
Abby
accent.
Last
my
my
I
sent
last
email
on
Thursday
and
a
human
hasn't
responded.
Yet,
however,
I
found
that
Scylla's
I
found
that
Scylla
DB
has
slack
and
registered
their
Easter
day.
E
E
E
C
E
A
D
E
Don't
ruin
things
some
fixes
for,
but
related
to
last
week
in
handling
its
alcoholics
already
mad,
because
the
pencil
back
there,
however,
we
decided
to
resurrect
the
altar
old
idea
to
kill
last
be
entirely
and
I'm
working
on
it.
Right
now,
which
must
take
too
much
I
will
be
looking
for
a
new,
bigger
task
in
crimson
twice
what
we
have
only,
for
instance,
its
recovery
of
a
bullet,
a
Bible.
F
F
So
the
thing
about
recovery
is
you
actually
need
to
know
absolutely
everything
about
how
staff
does
recovery
so
step?
One
of
that
is
I.
Guess
learning
that
I've
been
working
through
this
with
is
do
mister
housing
but,
and
we've
been
working
at
a
document.
So
I
will
add
you
to
this
document
and
you
can
help
participate,
sort
of
filling
it
in
and
answering
questions
and
well
understanding
how
it
works
is
suppose.
Anyone
else
wanted.
D
A
F
F
Be
but
the
first
step
honestly
is
to
be
experts
on
how
recovery
works.
It
took
me
a
few
months
Lorna.
So
like
no
Joe
there's
no
way
you
can
we're
not
it's
it's
basic.
There
is
no
part
of
Rados
that
this
doesn't
touch
boy
fit
it
that
way.
I.
A
F
No
more
more
importantly,
the
exact
details
of
how
recovery
work
are
the
flip
side
of
why
the
right
guarantees
are
what
they
are.
So
if
you
changed
our
recovery
works,
you
also
have
to
change
our
rights
work
and
Reed's
work.
So
those
three
things
are
sort
of
a
triangle:
do
you
need
to
understand
how
they
all
fit
together
to
understand
why
the
rules
are
what
they
are,
some
parts
that
are
written
down.
F
Don't
get
me
wrong,
don't
be
like
too
scared.
It's
totally
learn
about
it's
just
you
know
it's
not
just
a
write
down
a
design
document
and
write
the
code
thing
the
exact
way
it
were
in
good
works
in
classic
as
a
matter
of
historical
happenstance
and
if
at
all,
humanly
possible,
we
don't
want
to
copy
the
same
structure
over
because
it's
much
harder
to
understand
than
it
needs
to
be,
but
to
do
it
in
fashion,
you
have
to
know
why
it
works
the
way
it
does
I.
F
E
A
A
F
D
F
Gonna
merge
for
now
I
suggest
that
we
just
do
that.
So
what
we
can
do
is,
would
we
get
a
little
closer
to
production?
We
can
just
say
that
any
cluster
that
has
crimson
O's
DS
in
it
you're
not
allowed
to
change
the
P
genome.
That
can
be
a
an
alpha
sort
of
scenario.
We
can
do
a
great
deal
of
testing
without
changing
the
P
genome.
It's
important
for
real
clusters,
but
it's
not
so
important
for
proving
that
the
core
our
system
works.
F
F
For
what
it's
worth
in
c-store
I
am
considering,
split
and
and
merge
like
there
will
be
structures
that
are
efficient
or
splitting
and
merging
Peachy
structures,
but
I'm
not
going
to
bother
implementing
it.
I'm
just
going
to
ensure
that
the
tree
structures
are
will
make
that
tractable.
When
we
decide
to
do
it.
C
F
F
F
F
E
F
We're
blood
on
recovery,
I
think
that's
the
next
thing.
You
can
only
put
it
off:
okay,
one
and
now
that
there's
a
blue
store
back
end.
It
makes
a
great
deal
more
sense
because
you
can
actually
turn
those
DS
off
and
turn
that
back
on
and
have
the
data
stuff.
You
didn't
work
so
in
testing
recovery,
yep.
F
D
D
C
Sock
admin
socket
implementation
is
back
for
review.
If
you
take
a
look
and
hopefully
approve
it
and
I
hope
there
aren't
any
issues
this
one
I
would
there
are
two
questions
I
had
regarding
this
one.
If
you
noticed
I
have
a
perf
pains,
perfect
automatic
proof
test
fails
with
strangely
compiler
errors,
which
I
couldn't
understand.
The
reason
for
I
am
compiling
game
both
with
clang
and
GCC
and
testing
they,
of
course,
before
before,
I
deliver
and
I'm,
not
sure
I
understand
the
someone
can
take
a
look
and
they
helped
me
understand.
C
Okay
moving
it
and
send
it
and
I'll
find
it
when
I
don't
want
to
hold
you
in
this,
but
I
found
it
in
a
second
note
that,
after
the
delivery,
we
have
a
problem
with
orderly
shutdown,
because
there
is
a
bug
in
Prague
that
I
talked
about
in
systole
in
a
boat
except
in
sister.
Now
my
pet
was
accepted
and
merged
in
system,
so
it
is
fixed.
Now,
sister
is
not
fixed
and
I
think
we
should
consider
updating
our
sister
version.
There
is
one.
C
The
reason
was
an
important
delivery
system
by
year,
inching
also
regarding
the
appropriate
that
we
encounter
in
in
the
tests
and-
and
there
were
a
few
other
things
that
emerged
regarding
clan,
your
work
in
clunk
people
and
the
others,
so
I
think
that
we
have
a
good
reason
to
take
version
of
sister.
At
this
point.
D
C
And
another
question
I'm
encountering
with
one
of
the
things
that
slowing
me
down
is
the
term
encountering
internal
errors
in
assisting
the
plus
plus
compilation
which
are
parodic.
They
are
not
repeatable,
but
they
when
compiling
a
full
staff
or
a
full
creation.
If
an
account
link
almost
every
time,
I
have
to
repeat
today
build
a
few
times
until
I
get
it
done.
Is
it
only
me?
C
E
D
D
F
C
C
C
I'm
on
fado,
31,
okay
and
another
way,
another
issue
is
the
sister
with
the
group
I'm
running
Fedora
31,
and
this
town
has
a
problem
with
cgroups
a
fat
one,
it's
use,
Dixie
could
be
one,
and
the
Fedora
is
now
so
good
to
be.
I
will
add.
I
will
try
to
add
minimal
support
for,
say
goofy
to
the
sister,
if
I
Heather,
if
I
find
the
time
to
do
it,
I'm
not
sure
I,
understand
why
it
seems
that
sister
only
looks
at
the
top
when
talking
about
C
groups.
C
G
A
F
Hey
sorry,
okay,
so
I
am
currently
making
progress
on
implementing
the
actual
writing
to
disk
portion
of
the
lowest
level
c-store
from
the
document
I
sent
out.
This
would
be
the
segment
manager
journal
and
transaction
manager.
It's
essentially
responsible
for
taking
a
set
of
blocks
and
deltas
and
writing
them
to
a
segment
sequentially
with
appropriate
check.
Something
what's
up.
Next
is
writing.
F
The
lot
is
finishing
that
and
then
writing
the
logic
for
finding
the
beginning
of
the
current
journal
sequence
of
Delta,
so
they
can
be
replayed
the
step
after
that
will
be
writing
a
dirty
block
cache
because
there's
no
way
to
do
this
without
one
and
then
the
actual
LBA
tree
implementation.
Key
foo
I'll
have
a
look
at
yours,
but
it's
missing
a
lot.
F
Good,
it's
just
that
the
reason
I
haven't
written
the
tree
part
yet.
Is
that
there's
a
lot
that
I
don't
know
about
the
way
the
next
like
to
interface
layers
are
gonna
work,
I'm
gonna,
send
you
an
email
with
more
what's
sort
of
what
might
branch
and
where
I
am,
but
the
short
answer
is
I.
Think
something
like
this
could
work.
It's
just
I,
don't
know.
F
Yeah
I
know
for
sure
that
I
don't
want
to
decode
Watts
I
would
like
the
B
tree
implementation
to
work
directly
in
terms
of
the
in
memory
state
I.
Don't
want
to
do
a
memory
copy
back
into
an
array
right.
That's
what.
F
I,
don't
want
to
decipher
the
devil,
no
I
wouldn't
on
this
layout.
That
is
tolerable
for
treating
as
an
in
memory
layout,
so
that
I
can
just
pull
the
block
directly
into
memory
and
then
interpret
it
directly,
which
will
be
a
little
bit
tricky
but
I'm
not
sure
to
what
extent
that's
gonna
work
yep.
So
sorry.
A
A
F
F
A
F
No
I
got
that,
but
I
want
to
cache
at
the
block
level.
I
want
to
catch
the
blocks
directly,
so
there
might
be
a
header
for
each
block
that
has
some
additional
information
like
pointers,
but
the
block
itself
I
want
to
be
uninterpreted.
Does
it
ask
for
four
kilobytes
of
keys
and
values
and
I
don't
want
to
decode
those
into
an
array?
The
keys
and
values
aren't
even
necessarily
adjacent
right.
A
F
A
D
D
F
Yeah,
but
the
larger
thing
is
just
I:
I
have
to
implement
two
more
layers
of
interface
before
I
know
exactly
what
I'm
gonna
watch
here.
If
that
makes
sense
now
it's
a
lot
more
likely.
This
will
work
for
the
Oh
node
tree
and
the
beach
and
the
oh
man
Oba
no
well.
The
LBA
tree
is
super
weird
because
it
has
physical
addresses
and
it's
write
out.
Rules
are
gonna,
be
really
strict.
F
I
mean
the
elbe.
A
tree
is
special,
it
doesn't.
It
behaves
more
like
a
clutter,
FS
tree,
but
the
other
tree's,
the
Ono
tree
and
the
OMAP
tree.
They
use
logical
interruptions,
so
you
can
rewrite
one
of
these
blocks.
Without
changing
the
parents,
you
have
to
change
the
I'll,
be
a
tree
parents,
but
that's
hidden
one
layer
down.
F
A
G
Or
I,
just
muted,
myself,
yeah
last
we
I
finished
the
figs
and
the
simplification
of
messenger
interfaces
and
there
are
impact
to
the
OSD
how
OSD
uses
messenger.
So
please
help
review
this
part
I
think
that's
the
most
important
change
and
also
sent
the
sistar
extension
to
to
the
Google
Group.
There's
no
reply
it
and
that's
not
for
me.
A
A
G
A
A
B
C
C
The
crystal
it
is
the
crimson
code,
I
copied
them
as
part
of
the
PR
and
listing
here
now
they
appear
I'm.
Talking
about
take
a
look.
This
is
what
I
used
to
format
the
files
that
I
added
in
this
specific
PR
I,
think
I
mean
almost
match
what
I
saw
in
our
code.
In
now,
which
files
are
you
look
up
just
at
they
had
take
a
look
in
the
Big
D
na
PR,
a
comment
at
the
PR
header.
C
It's
really
a
message,
I
think,
okay,
and
only
did
the
commit
message
to
it
appear
in
the
PR
messaging.
The
commitment
both
you
have
the
same
email
Ines,
take
a
look,
try
it
and
that's
agree
on
I.
Don't
care
exactly
what
they
what
we
agree
on,
but
let's
agree
on
a
set
of
fake
log
format,
tools
and
use
them.
It
simplifies
the
day's
work
and
saves
a
lot
of
time.
For
me
at
least
I.
Don't.
C
F
F
D
C
Okay,
once
that
two
different
issues-
let's
talk
about
it,
but
first
a
the
clock.
Former
doesn't
change
the
comment
format.
So
that's
unrelated
issue
as
far
as
I
know,
okay,
and-
and
so
this
is
one
issue
that
I'd
love-
that
you
take
a
look
at
it
and
the
second
I
did
try
after
your
codes,
I
did
try
to
copy
B
but
did
I.
What
did
I
do
wrong
in
the
comment
for
once
no.
C
F
C
F
If,
if
it
all
humanly
possible,
I
would
like
to
select
a
style.
That's
the
same
as
one
of
the
styles
that
already
exists
in
itself.
F
Is
it
possible
to
get
the
comment
format
as
part
of
that
thing?
You
don't
have
to
do
it.
Mr.
Nedim
just
curious,
I.
D
C
F
C
F
D
F
I
can
I
can
send
the
same
email
to
you
guys
too,
but
the
truth
is
I've
rewritten
that
code
three
times
in
the
last
three
days.
So
unless
you
want
the
local
icon
I'm
giving
it
to
keep
you
because
it
because
he's
working
on
the
Petri
code,
you
know
sent
him
an
email.
If
anyone
else
wants
to
see
in
the
branch
it's
this
one
but
seriously,
it
doesn't
always
even
make
sense
here.
Here's
here's
the
branch
I'm
making,
no
attempt
whatsoever
to
keep
the
commits
clean
and
it's
a
complete
mess.
F
G
A
B
A
Need
to
look
into
it,
but
the
the
problem
is
that
we
need
to
have
have
at
least
one
I
Laurie
I
need
I
really
need
to
look
into
sister
to
assist
our
to
understand
why
we
have
such
a
restriction.
I
think
we
we
added
is
this
restriction
couple
months
ago
to
to
address
the
test.
Failure
with
the
alien
I
think
it's
we
made
that
change
for
some
reason.