►
From YouTube: 2020-04-07 :: Ceph Crimson 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
C
D
I'm
here
to
just
listen
in
and
just
follow
up
on
the
shallow
that
we
created
recently
I
know
some
of
us
were
not
in
the
planning
meeting
last
week,
just
wanna
make
sure
everybody
is
on
the
same
page
and
if
there
are
some
items
that
we
can
identify
as
low-hanging
fruits
or
things
that
nobody
is
working
on
in
general,
that
any
new
contributors
can
just
pick
up
and
try
to
walk
on
I.
Guess,
that's
yeah!
That's
that's
something
that
we
want
to
identify,
but
doesn't
it'll
happen
today.
D
B
E
Hello
last
week,
I
sent
a
smoke.
We
are
bringing
support
for
burn
for
Turing,
written
output,
buffers
and
return
valve
values
inside
PT,
lock.
Also,
the
link
right
now,
I'm
working
on
the
backfill
I
I,
implement
I've
fixed
and
some
kind
of
dead
loop
in
handing
curve,
and
always
be
alive
message
and
now
I'm
I'm.
Why
I
wire
it
up?
Also
the
PD
temp,
orchestration
and
I'm
investigating
an
issue
actually
I'm
working
with
I
want
the
transition
of
of
primary
and
to
doing
Bakula.
E
I
have
a
half
environment
where
I
have
a
scenario
among
in
a
scenario
when
new,
when
I
primary
an
empty
primary
joins
the
cluster
and
it
needs
to
get
back
field.
However,
it
seems
that
I
stuck
busy
in
white
in
weight,
acting
change
state
because
should
restart
feeling
and
particularly
past
identify
the
new
OSD
map
have
as
being
as
having
a
new
interval.
E
F
So
since
that
that
should
give
us
a
rough
outline
for
the
last
piece,
so
I'm
doubling
back
and
writing
unit
tests
for
that
for
the
transaction
manager
for
the
journal
and
for
the
cash
and
as
they
do,
that
I'll
fill
in
missing
methods
that
I
haven't
got
around
to
the
woman
I
get.
This
will
probably
take
a
couple
of
weeks
after
that,
I
should
have
a
PR
ready
for
the
LP,
a
lair.
That's
what.
B
B
B
F
B
F
F
Has
the
code
for
the
LB
a
tree
is
well,
I
mean
module
of
bug.
Fixes
and
improvements
is
the
same
as
it
will
be
when
there's
a
implementation
of
the
interface
that
writes
just
that
parts
identical
that
is
it's
using
the
layer
immediately
below
the
transaction
manager
to
do
it
to
do
what
it
needs
to
do
excusing
the
cast
extent
roughs,
and
it's
doing
all
of
all
of
that
stuff.
B
B
F
G
G
C
H
Week
I
had
identified
the
song
missing
features
in
the
current
heartbeat
implementation
and
I
have
implemented,
but
often
in
in
private.
In
my
current
private
branch
and
I
will
send
them
out
after
the
refactoring.
The
the
one
connection
P
is
is
merged,
I
think,
and
another
amazing
feature
from
heartbeat
is
is
the
is
the
consolidation
of
the
peers
update.
It
is
controlled
by
the
heartbeat
need
update
in
the
classic
OSD
and
in
the
second
part,
I
am.
H
H
H
B
D
F
F
F
F
A
D
F
It'll
be
quicker,
I
do
intend
to
include
the
relevant
bits
in
c-store,
but
given
a
choice
between
debugging
snapshots,
which
take,
which
will
take
a
really
long
time.
It's
just
hard
to
win
like
snap
trimming
alone
is
a
fair
amount
of
work.
So
there's
a
lot
of
administrative
stuff
in
the
OSD
that
doesn't
really
directly
contribute
to
the
mission
of
crimson
that
may.
F
As
making
sure
crimson
basically
works
and
is
fast,
and
that
is
the
debt,
the
value
I
think
we
have
to
demonstrate
by
Pacific
it
it's
that
for
a
straight-line
workload
we
need
to
correctly
and
properly
do
recovery
and
IO
and
snapshots
are
like
it's
a
valid
thing
to
work
on
and
by
all
means.
If
we
have
the
bandwidth
I
think
we
should
try
to
attack
it,
but
I
don't
know
that
I'd
feel
bad.
If
we
didn't
have
them.
D
F
D
F
Specifically,
because
I
don't
I,
don't
want
the
goal
of
Pacific
to
be
a
laundry
list
of
features
for
crimson
I,
want
it
to
be
speed
and
quality,
and
for
that
I
think
there
are
some
core
things
that
are
clearly
qualitatively
different
from
features
like
RVD
as
a
workload
has
to
be
demonstrated
to
be
faster
and
with
lower
CPU
cost
than
clásico
is
OSD
and
I.
Don't
think
snapshots
are
required
to
demonstrate
that
that
value.
F
Then
it's
just
a
lot
of
work
like
trim,
isn't
an
inherently
interesting
thing.
It's
just
a
thing:
we
have
to
pour
it
over
from
classic
it's
very
similar
to
the
work
that's
being
done
currently
for
scrub.
It's
just
a
matter
of
porting.
The
existing
algorithm
and
I
would
say
we
should
leave
off
scrub,
except
that
without
scrub.
We
can't
demonstrate
that
wrapping
that
recovery
works.
That's
the
risk
of
what
scrub
over
yeah.
A
D
All
right
I
mean
that
makes
sense
and
B
that's
pretty
much
flexible,
I
guess
and
you
guys
are
the
best
people
to
do
to
mine
what
goes
and
what
doesn't.
But
in
general,
are
there
any
other
items
that
are
not
here
and
should
be
here,
any
other
brainstorming
that
we
didn't
end
up
doing
last
week
and
should
be
done
now?
What.
A
B
B
F
So
that's
kind
of
what
I'm
saying
I:
don't
it's
not
that
we
can't
work
on
things
that
aren't
that
set
it's
that
those
are
the
things
that
I
think
are
important
for
Pacific
and
I.
Think
what
they
have
in
common
is
that
there
are
the
hard
things
in
the
OSD
I.
Don't
think
that's
an
accident!
It
turns
out
that
the
core
things
that
we
wouldn't
have
bothered
to
do
the
hard
things
if
they
weren't
in
some
way
and
important
right.
D
B
Yeah
is
to
probe
is
not
a
very
high
priority
or
importance
benefit
is
the
missing
piece
in
in
crimson
and
the
messenger,
for
we
can't
get
your
hands
dirty
for
the
purpose
of
getting
familiar
with
of
crimson.
That's
it
to
edit.
It
configuration
support
to
the
messenger,
because
currently
messenger
is
using
a
poor-man's
solution.
It
all
of
the
other
settings.
The
hardwired
I
had
editing
in
the
infra
or
clean
up
to
make
sure
it's
where
defined.
How
about
that
category.
F
F
No,
what
I'm
saying
is
chef
ADM
could
automatically
and
chef
ADM
could
automatically
take
a
one,
terabyte
disk
and
divide
it
into
ten
100
gigabyte
disks
I'm
saying
the
two-lane
could
just
automatically
divvy
up
a
disk
into
a
bunch
of
OSU's
I.
Don't
think
it's
a
good
answer
forever,
but
I
think
it's
adequate
for
for
now.
B
F
Teaching
an
existing
tool
to
just
do
it.
Chef
ADM
already
has
the
ability
to
deal
with
deploying
ghosties
and
stuff,
so
it's
just
a
matter
of
teaching
it
to
do
this
part
to
usability
yeah,
don't
knock
usability,
we'll,
never
be
able
to
teach
people
to
set
these
these
things
up
without
those
those
tools,
someone
taking
notes.
F
A
F
A
F
It
all
of
our
problems,
if
you
can't
get
it
to
work,
you
need
to
work
out.
Why
and
file
a
bug.
A
B
F
D
No
I
think
what
the
priorities
that
you
mentioned
earlier
are
right.
We
need
to
have
bearing
recovery,
backfill
sorted
out.
First
I
think
this.
The
snapshot
support
just
came
in
because
probably
Josh
was
thinking
in
the
lines
that,
if
we
have
this,
then
we
can
just
say
that
our
body
and
its
own
in
the
replicated
scenario
would
be
feature
complete,
but
I
understand
where
you
are
coming
from
so
I
think
this
is
more
discussion
with
product
and
all
the
downstream
stuff
that
we
don't
want
to
get
involved
right
now.
F
D
Think
this
is
a
really
good
discussion
that
we've
started
and
I
would
really
encourage
everybody
on
this
group
to
identify
those
kind
of
areas.
If
there
are
things
that
have
not
been
identified,
and
there
are
other
people
who
can
help
with
that
or
we
should
render
fight
them
and
like
that
them
in
the
Trello
in
some
form
and
as
chief
who
said,
we
can
follow
up
in
a
separate
meeting
as
to
how
we
can
make
that
process
of
introducing
other
people
to
crimson
easier
and
do
some
sort
of
you
know.
D
A
B
D
Guess
I
guess
the
idea
of
Trello
is
that
as
long
as
anything
is
not
much,
it
doesn't
go
to
the
specific
merged
column,
which
is
to
the
extreme
left.
You
see
the
first
column,
it
says
crimson
v1
Pacific
as
long
as
you're
working
on
something
and
it's
ongoing
it.
It
just
stays
wherever
it
is
and
once
it's
merged
you
can
move
that
particular
card
to
the
first
column.
So
it's
it's
tracked
as
complete
yeah.