►
From YouTube: Ceph RGW Refactoring Meeting 2022-11-30
Description
Join us every Wednesday for the Ceph RGW Refactoring meeting: https://ceph.io/en/community/meetups
Ceph website: https://ceph.io
Ceph blog: https://ceph.io/en/news/blog/
Contribute to Ceph: https://ceph.io/en/developers/contrib...
What is Ceph: https://ceph.io/en/discover/
B
Yeah
I
think
this
will
be
pretty
quick.
I
am
working
on
making
or
adding
tags
to
all
the
configuration
variables
in
the
docs
to
see
whether
or
not
they
can
be
set
or
changed
at
runtime
or
not.
B
Casey.
Thanks
for
the
pointers
a
few
days
ago
on,
what's
missing
in
the
docks,
I
was
able
to
open
up
a
PR
and
get
those
rendered
I'm
wondering
what
the
next
step
should
be
to
actually
go
through
and
audit.
All
the
config
variables
thing
Matt
suggested
I
put
together
a
Google
sheet
or
something
and
share
it
with
the
rgw
team.
B
B
C
C
B
I'm
gonna
guess,
if
we
Market
or
we
don't
Market
we're
kind
of
using
our
common
sense
and
saying
this
variable
or
the
the
result
of
this
config
should
take
place
upon
restart
or
it's
fine
to
just
throw
this
in.
During
run
time,
I
mean
even
in
the
case
where
we,
where
we
say
we
Market
with
runtime
or
with
restart
or
whatever
we're
still
using
our
best
knowledge
of
the
Fig
variable
as
a
team
right.
C
B
Okay,
yeah
I'll
go
ahead
and
get
started
on
that
I
mean
luckily,
the
the
docks
are
kind
of
broken
up
in
a
way
that
it's
somewhat
easier
for
me
to
how
uncertain
people
on
certain
subsystems.
D
B
Yeah
I'll
get
I'll
get
started
on
that
and
update
the
group's
progress
over
the
coming
weeks.
A
All
right,
let's
see
or
is
not
here,
but
I,
just
had
a
a
chat
with
him
before
this
call
about
his
sync
fairness,
progress
and
he's
stuck
on
an
issue
with
rados
in
general
about
how
to
persist
the
bids.
A
C
A
A
And
this
would
introduce
a
way
where
we
would
go
on
and
apply
that
right.
Even
if
we
couldn't
acquire
the
lock.
C
A
Well
it
it
does
need
to
be
the
OSD.
That's
that's
storing
all
of
these,
but
yeah
it.
C
C
C
So,
what's
so,
what's
what
would
change
in
your
in
your
approach
is
so
so
the
bidding
currently
goes
to
like
a
central
location
and
if
we're
all
and
all
operations
have
to
that
are
that
are
that
are
that
are
negotiated,
have
to
have
an
extra
step
where
we
do.
We
do
that,
and
then
we
do
whatever
else
we're
going
to
do.
C
Or
what
so?
What
changes?
If
we,
if
we
move
it
to
the
to
the
two
to
the
object,
and
you
know
what
you're,
what
you're
calling
out
here,
that
if
we,
if
we
declobalize
it
to
this
at
the
point,
was
we're
not
using
CLS
lock.
But
then,
if
we've
switched
you're
proposing
to
use
CLS
lock
in
order
to
get
rid
of
the
global
map.
A
The
the
global
map,
the
sync
fairness
PR-
adds
a
global
map
to
CLS
lack
to
implement
the
bid
stuff.
A
C
A
A
A
C
E
Could
we
possibly
redefine
the
op
as
as
persisting
this
data
and
then
as
a
side
effect,
taking
the
lock
instead
of
the
other
way
around,
and
then
it
would
return
success
but
unlocked
or
success
but
locked
and
persist
the
data.
C
C
C
Does
to
allow
us
to
persist
a
particular
piece
of
data
violating
its
normal
rules,
but
the
comp
correct
for
executing
the
compound
op
Tim
am
I
making
sense
there.
D
C
C
A
A
Yeah
part
of
part
of
this
is
tied
into
the
PG
log
and
how
it
does
recovery,
so
it
would
need
to
recover
apps
that
returned
this
error
and
apply
the
rights
again.
A
Yeah
welcome
to
the
call
we're
this
part
of
the
project
is
rgw,
which
is
the
S3
layer
on
top
of
stuff.
A
F
C
A
A
C
A
I
mean
in
the
pr
as
it
is,
they
are
Global
static
variables
in
CLS,
lock,
CC,
so
it
would
be
OSD
wide,
but
there
may
be
some
way
that
we
can
hook
this
into
a
object
context
so
that
it's
thread
safe
and
its
lifetime
is
attached
to
the
abject
context.
Instead,.
C
D
A
Yeah,
maybe
maybe
we
just
need
some
CLS
interface
to
to
get
at
it
might
not
be
too
bad.
Sounds.
C
A
Yeah
I
mean
I
I
still
feel
like.
It
would
be
nice
if
we
could
persist
it
and
so
recover
from
crashes
or
whatever,
but.
C
A
C
A
A
A
What
do
you
mean
by
that?
It's
it's
weird:
to
persist;
data
that
that.
C
C
If
it's,
if
it's
significantly
about
which,
let's
respect
it
backs
up
right,
I
mean
I,
guess
having
some
having
some
specific
day,
you
can
just
stick
in
other
countries,
not
not
to
blow,
it
should
be
harmful.
It
may
be
harmless,
although
it
doesn't
it's,
it
sounds
like
we'll
get
an
argument
about
it,
but
but
but
let's
say,
but
let's,
let's
I
guess:
let's
talk
about
the
later
question.
This
is
how
significant
how
helpful
to
how
helpful
the
performance
is
it
that
it
be
persisted?
Is
it
worth
paying
for
a
Reno's
transaction
for
it.
A
So
say
that
OSD
crashes
and
another
OSD
comes
up
as
the
primary,
so
lock
request
would
go
to
that
one.
Instead,
there
wouldn't
be
any
bids,
and
so
the
first
Locker
would
acquire
the
lock
and
then
bids
would
pile
up
after
that
and
subsequent
lock
requests
would
would.
A
G
The
shootology
test
finally
ran
and
will
be
both
before
and
after
the
rebase
looks
like
both
before
and
after
the
rebase
had
to
multi-site
failures
and
toothology.
G
But
I
don't
know,
I,
don't
know
how
to
tell
if
those
test
failures
are
meaningful
or
not.
So
anyway,.
G
C
F
G
G
Matt
I
ran
one
at
Dash
P10,
because
Matt
told
me
to
and
then
the
second,
the
later
one
I
ran
a
dash
P20,
okay,
I
guess
that
I
don't
know
how
unfriendly
that
is,
but.
A
So
sorry
Caleb,
can
you
clarify
how
Matt's
changes
are
part
of
this?
Did
we
just
merge
Matt's
stuff.
C
A
C
C
G
Know
how
I
would
have
I
don't
know
what
the
right
steps
would
have
been
to
merge
to
include
yours
and
keep
the
merge
message,
but.
G
C
G
Okay,
these
are
git
get
commands
above
my
pay
pay
grade.
So
I
don't
know.
A
C
Yeah,
if
you
want
to
do
that,
if
you're
gonna
do
that
Kayla,
that's
okay,
I
mean
it'll,
just
or
else
I
don't
know.
A
C
H
C
E
So
relevant
to
this
is
the
sort
of
driver
changes
which
I
have
completed,
but
not
yet
run
through
QA,
because
that's
going
to
be
at
least
as
invasive
as
the
commit
that
just
merged.
E
A
Yeah,
yes,
if
we
have
to
rebase
and
retest,
then
it
might
make
sense
to
let
down
do
his
changes
at
the
same
time
and
then
include
both.
But
that's.
E
E
So
the
the
store,
the
driver,
renames
The,
Base
Class-
store
to
driver.
So
it
touches
just
about
everything
in
rtw.
C
E
C
A
G
A
Right
yeah
I
mean
maybe
we
pile
all
three
together
if,
if
it's
going
to
be
a
a
messy
rebase,
no
matter
what
we
do.
E
E
Yeah
so
I
suspect,
you've
all
can
get
his
merged
before
mine
and
then
I'll
just
rebase,
both
mine
and
mats,
on
top
of
it.
A
All
right,
Caleb
anything
more
to
discuss
on
your
topic.
G
A
C
A
A
F
A
F
Yeah
I
need
that
to
move
ahead
with
the
multi-site
coating
work.
A
Okay,
yeah,
maybe
just
carry
those
six
commits
in
in
your
branch.
In
the
meantime,.
C
B
H
H
A
F
D
Hello,
hey,
hey,
I,
want
to
start
contributing
or
make
changes
in
your
GitHub
repo.
So
how
can
I
start
I
can
stock
here?
I
can't
understand
that
how
can
I.
D
F
D
Okay,
okay,
so
I
can
I
can
anytime
mail.
You
Cassie.
A
D
Yeah,
okay
and
I,
and
also
I
included
that
the
development
project
is
on
Docker
Docker
image
available.
D
Yeah
I
should
start
with
container
or
native
on
my
native
machine
I
can
set
up
my
development
project.
D
Okay,
okay,
so
can
you
share
the
link
of
how
can
I
set
up
a
development
environment
in
my
system.
A
Foreign
yeah
no
problem
might
also,
if
you're
looking
for
things
to
work
on
specifically
I
point
you
to
our
bug
tracker,
there's
a
tag
called
Blow
hanging
fruit,
which
should
be
fairly
easy
stuff
to
start
working
on.
D
Okay,
they,
these
are
low
hanging
fruit
tag
name,
is
easy
to
easy
to
solve.
Yeah.
A
Exactly
yeah
shouldn't
be
too
involved
to
to
reproduce
and
fix
those.