►
From YouTube: Ceph RGW Refactoring Meeting 2023-03-01
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/contribute
What is Ceph: https://ceph.io/en/discover/
A
Okay,
well
I'll
start
with
an
announcement
from
the
Upstream
stuff
leadership
call
earlier
today
the
decision
was
made
to
Fork
The
Reef
release
branch
that
should
be
happening
today,
so
anything
that
still
needs
to
get
to
Reef
will
have
to
merge
to
Main
and
go
through
backboards.
B
Hi
Katie,
yes,
so
I
have
a
request,
so
I've
tested
most
of
the
cases
with
the
bucket
sink
disabled,
for
which
using
sync
policy,
so
a
request
review,
also
I
updated
in
the
pr
that
I
see
an
issue
with
the
latest
master
that
out
of
three
zones.
If
only
two
zones
are
configured
to
get
synced
still
third
zone
is
pulling
the
data.
This
looks
like
a
regression
a
test
in
the
downstream,
which
is
equal
into
pin
series.
We
don't
see
this
issue.
B
A
Yeah
I'll
add
that
there
seems
to
be
a
related
issue
with
bi
log
trimming
and
archive
zones,
specifically
where
sync
policy
is
telling
it
to
query
the
other
zones,
even
though
they're
not
syncing
from
it.
So
that's
preventing
logs
from
being
trimmed
and
I,
think
the
archive
Zone
shouldn't
be
recording
any
logs
anyway.
A
Maybe
that
would
make
sense,
but
the
the
sync
module
interface
has
a
a
virtual
function,
saying
whether
the
module
should
export
data
and
archive
Zone
says
false
there.
So
I
don't
think
it's
intended
to.
A
B
When
I
encountered
another
issue
with
three
zones,
I
filed
a
Tracker
upstairs
okay.
B
Thanks
and
one
more
thing,
I
have
said,
is
wave
suppose
out
of
three
zones.
Some
zones
are
not
part
of
the
sink.
If
we
configure
policy
such
a
way,
do
we
still
need
to
I
mean
the
sync
status
still
shows
those
zones
which
are
out
of
sync
as
a
source
I
mean
it
looks
incorrect.
Do
you
think
we
should
correct
this
thing?
B
Status
that
is
bucket
sync
status
is
showing
appropriately
like
it
shows
only
the
source
zones
of
those
zones
which
are
part
of
the
thing,
but
sync
status
shows
all
the
so
eating
towards
all
the
zones
in
its
own
group,
even
though
they
are
not
thinking
from
those
zones.
B
I
I've
pasted
the
tracker
issue
I
raised
for
it.
Upstream
tracker
in
the
Theta
pad
I
just
want
to
check
if
it's
feasible
to
correct
the
timing.
I
haven't
looked
into
housing
status.
Works
in
this
sync
policy
case.
B
It's
supposed
to
own
three
suppose
not
part
of
the
sync
policy.
I
mean
it's
sync,
only
Zone
one
and
so
two
are
allowed
to
sync
among
each
other.
Still,
if
I
run
sync
status
command,
it
shows
to
Zone
3
as
the
source
for
both
Zone
one
and
so.
B
D
Order
to
solve
what
I'm
saying
but
like
there
can
be
the
obvious
case
where
it's
disabled
in
the
policies
and
yeah.
D
Well,
I,
don't
remember
the
the
specifics.
It
could
be
that
the
the
flow
is
permitted
but
not
enabled
in,
in
which
case
it
can
be
at
the
Bucket
Level.
B
B
D
C
B
B
D
B
B
That's
all
I
have
on
this.
Thank
you.
A
Thanks
Samia
thanks
Yehuda
I
added
rgwhttp
client
as
a
topic
just
kind
of
following
up
on
the
email
that
I
sent
to
the
list.
Yesterday.
A
But
I
think
it
could
also
be
useful
in
multi-site.
A
B
E
Casey
are
there
reasons
to
use
this
new
client
inside
the
attribute
beside
hpsp
in
quick,
just
control,
the
existing
HTTP
client
that
we're
using.
A
A
We
were
talking
about
using
Beast
earlier
in
the
in
earlier
discussions,
but
I
think
this.
This
new
curl
Library
could
could
work
really
well
there
and
it
has
the
the
templated
SEO
interface
to
to
work
with
any
kind
of
core
routine
or
async.
E
B
E
A
Yeah
but
Beast
is
missing
a
lot
of
features
that
we
discussed
earlier
in
that
thread
stuff,
like
connection
pooling
and
and
all
of
the
timeouts
and
retries,
and
all
of
HTTP
too.
A
Yeah
I
I'm,
finding
that
I
I
like
lib
curl,
more
and
more
as
the
years
go
on.
The
the
only
downside
really
is
that
it's
the
only,
and
so
you
can't
do
stuff
like
custom
allocators.
E
Yeah
but
I
guess
for
the
for
multi-site
any
any
advanced
HP
features
are
actually
very
useful
because,
like
it's,
a
server
to
server
communication
right,
so
HTTP
would
even
be
be
good
or
could
be
good
there
because
of
the
the
features
that
it
has,
which
are
better
for
serving
to
server
communication.
A
Yeah,
it
definitely
could
see
HTTP
3
being
useful
for
multi-site
on
the
client
side,
but
we
we
would
have
to
overcome
the
hurdle
of
openssl
versus
boring,
SSL
linkage.
A
A
Which
is
doable
but
the
Prototype
that
I
have
is
kind
of
just
relying
on
the
application
to
do
all
of
the
lib
curl
stuff
and
then
just
process.
It
asynchronously.
E
A
The
boring,
SSL
header
is
essentially
compatible
with
openssl's
headers,
so
it
was
kind
of
tractable
to
get
Seth
at
least
some
components
to
build
against
boring
SSL.
A
A
So
the
the
openssl
dependency
is
part
of
Seth
messenger,
which
is
part
of
Seth
common,
which
includes
you
know,
self-context
and
logging,
and
all
of
the
other
stuff
that
rgw
needs.
So
we
would
essentially
have
to
kind
of
detach
all
of
those
things
from
Seth,
common
and
I
guess
build
a
separate
messenger
stack
specifically
for
rgw.
A
So
I
don't
know
it's
worth
discussing.
Someday
openssl
is
going
to
support
quick,
I,
hope,
Maybe
and
once
that
happens,
I
think
this
would
simplify
life
for
for
everything
here,
but
that
may
be
years
down
the
road.
So
I
don't
know
I,
don't
think
there
are
easy
answers
here
for
now.
I
think
I
think
loading
the
front
end
as
a
plug-in
is
kind
of
the
the
easiest
thing
in
the
short
term.