►
From YouTube: 2019-09-24 :: 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
B
Debugging,
the
crimson
Oh
ladies
door
and
I
meet
a
problem
about
the
period
stage
and
wait
for
active.
It's
always
not
active,
so
I
just
walk
around
a
to
not
to
fix
it
and
then
currently
riddles
parents.
Writer
can
return,
has
the
response
keratin
and
but
the
riddles
pantry
do
still
has
problem
I'm,
not
sure
it
is
to
read
the
past
error.
All
the
right
now
to
write.
Well,
not
right,
work
right.
So
let's
do
debugging
on
the
reader.
With
those
page
reading.
B
B
C
C
They
say
the
one
argument
taking
two
variant
of
say
then,
basically
all
those
so
that
things
pointed
out
in
the
end
review,
but
re
very
plain
to
thanks
very
much
and
also
also
apply
it
also
of
DPR
also
applies
sorry.
The
second
version
of
the
branch
also
applies
the
a
writer
to
many
other
places.
Basically,
it
converts
now
the
iterator
ends.
Eg
do
OSD
ops.
E
Come
work
in
the
recovery
back
whole
thing.
The
first
thing
that
struck
me
is
that
we
really
don't
want
to
rewrite
it
so
I'm
extracting
I'm,
trying
to
create
a
cleaner
interface
between
hearing
state's
primary
log
PJs
various
for
cover
backs
well
recover
replicas
code
and
the,
as
it
turns
out
now,
rather
complicated
system
of
throttling
and
sleeping
off
of
the
OSC
that
puts
appropriate
but
I'd
like
to
generalize
a
little
bit
so
that
securing
state
interface
doesn't
have
to
like
expose.
All
of
that.
This
should
be
not
super
hard.
A
Damn
I
got
a
question
for
you.
How
do
you
imagine
we
should
consolidate
the
different
interface
between
between
the
classic
OST
and
the
cruisers
dear?
We,
for
example,
we
we
have
a
with
our
objective
class
in
like
like
peering
listener.
The
return
value
is
oh.
Is
that
not
enough
you
right,
even
yeah
yeah,
that's.
E
That's
okay
with
me,
and
the
reason
is
that
if
you
take
a
look
at
interface
closely,
while
there
are
a
few
callbacks
they're
like
I,
don't
the
peering
state
doesn't
do
multi-state
operations,
so
the
lack
of
future
is
Asian
is
a
problem
for
peering
state,
but
not
for
crimson.
E
I'm
not
going
to
attempt
to
future
eyes
intervals
of
classic
OST.
Well,
you
put
it
that
way.
What
I
will
do
is
make
it
so
that
the
interface
that
hooks
up
into
crimson
doesn't
require
much
interfacing
to
interact
with
us
and
then
making
a
few
tries
will
be
super
helpful
because
it's
going
to
be
wrapping
the
state
machine,
which
is
actually
an
even
better
way
of
representing
an
asynchronous
operations
in
a
futurism
in
this
case,
in
other
words,
I'm
not
going
to
future
eyes
it
just
to
have
things.
You
know
it's
not
helpful.
E
I
will
do
I
I
will
attempt
not
to
make
it
the
case
that
my
choice
of
interface
in
classical
ISTE
makes
Crimson
hard
to
read,
but
obviously
there
were
always
compromises
that
I
don't
see
that
one
as
being
likely
I
think
it'll
be
straightforward.
It'll.
A
E
A
E
Decals
gonna
be
the
same
way.
It'll
be
a
sequence
of
like
any
particular
PG
going
through.
Backfill
is
going
through.
The
state
machine
ask
what's
next
in
the
PG,
compare
that
to
what
I've
gotten
the
PG
send
out
some
operations,
each
of
which,
at
the
little
state
machine
kind
of
pulls,
get
pulls
back
right
down
local
leafs
entry
or
sent
out
right
to
the
other
replicas
get
the
responses
back
mark
the
object,
complete
release
the
lock.
Let
the
next
state
machine
up
proceed
forward
to
the
next
object
right.
Yes,.
A
E
That
would
be
awesome,
but
until
we
have
that
it's
not
worth
worrying
about
too
much
I
think
yes,
the
part
I'm
part
of
actually
worried
about
and
will
probably
require
us
think
about
a
little
bit
is
I.
Don't
I,
don't
think
we
should
skip
this
part.
We
should
do
it
now.
At
least
a
little
is
the
one
interest
that
were
at
the
same
time.
E
Something
but
I
don't
actually
want
it
to
be
the
case
that
we
just
dumped
a
bunch
of
continuations
on
the
reactor
and
assume
that
it'll
prioritize
the
client
ones,
because
it
probably
won't.
So
we
got
to
think
about
what
we
want
to
do
there.
It
might
be
changes
the
C
star
to
add
some
kind
of
priority,
but
I
don't
really
want
to
do
that,
because
you
can
do
a
lot
of
stuff
with
a
scheduler
and
it
doesn't
make
a
lot
of
sense
to
cram
that
into
the
interior
of
C
star.
F
A
E
Can't
be
customized
and
it's
not
clear
that
we
would
want
to
even
if
we
could.
The
question
is
more:
how
do
we
design?
How
do
we
integrate
em
clock
with
that,
so
that
we
get
to
decide
what
the
it
could
be
as
simple
as
we
simply
work
out?
How
many
currently
active
continuations
see
start
should
have
maybe
T
star
can
tell
us.
I
can
say
I've
been
idle
for
a
percent
of
the
time
and
then,
whenever
it's
to
idle,
we
just
we've
released
more
stuff
from
scheduler
for
it
to
work
on
on
in
priority
order.
D
D
E
Think
develop
our
testing
methodology.
Ideally,
we
would
cover
everything.
That's
led
I'm
using,
but
my
guess
is
at
the
moment
our
priority
should
be
developing
more
than.
D
E
D
A
Probably
we
can,
when
we
think
back,
we
probably
we
can
ask
him
to
do
some
performance
test.
If
he
can,
you
can
have
a
have
access
to
the
some
somebody
card,
which
is
supported
by
TBD
K
and
run
run
some
pasted
on
some
come
across
on
tests
using
crimson
and
crimson
messenger.
Do
you
have
the
hardware
which
is
supported
by
Tim
DeKay.
C
C
E
C
F
D
C
A
C
A
C
C
A
A
E
A
E
Any
time
for
the
Sunnydale
folk
to
get
up,
that's
true,
I,
don't
know
you
should
ask
me
how
that
she
probably
just
has
an
opinion
about
how
long
it
takes
to
go
for.