Hi Sushma,
Thanks for your investigations! We already noticed the serializing
risk on GenericObjectMap/DBObjectMap. In order to improve performance
we add header cache to DBObjectMap.
As for KeyValueStore, a cache branch is on the reviewing, it can
greatly reduce lookup_header calls. Of course, r
Hi Haomai/Greg,
I tried to analyze this a bit more and it appears that the
GenericObjectMap::header_lock is serializing the READ requests in the following
path and hence the low performance numbers with KeyValueStore.
ReplicatedPG::do_op() -> ReplicatedPG::find_object_context() ->
ReplicatedPG:
OH, I didn’t know there was a Calamari mailing list.
Subscribing now, thanks.
On a semi-related note, iWeb mirrors a lot of Open source projects and Linux
distributions: http://mirror.iweb.com/
I’d love to add Ceph to that list - is there any particular process ?
If anything, I’d love to build a
Hmm. You could replace do a s/method/_method/ kind of thing on the
read methods and have method in ObjectStore ping the trace point and
then call _method. The writes are slightly trickier/simpler. I
suggest you add ObjectStore level _start_transaction(uint64_t seq,
Transaction*) and _end_transac
Hi,all.
I finally was succeded.
Maybe somebody will be intresting.
A script read the content from a fuse-rbd files (i wonder, what is actual use
case of fuse-rbd?) with "dd"
and, in a case of timeout (alarmed by a background process), killed entire fuse
daemon, remount fuse-rbd and resumed at
Hi Zack,
Thanks for the tip, I'll try it next time :-)
Cheers
On 30/06/2014 16:10, Zack Cerza wrote:
> Hi Loic,
>
> At this point I don't really have a way to look back in time to see
> what was going on, but in the future when jobs are blocked waiting for
> machines for unreasonable periods it
On Mon, 30 Jun 2014, David Moreau Simard wrote:
> Hi,
>
> Is there any short term plans for Inktank/Redhat to mirror packages for
> Calamari and Calamari-Clients under ceph.com or is that currently in the
> hands of the community ?
John just sent an update about this yesterday to ceph-calamari
-- All Branches --
Alfredo Deza
2013-09-27 10:33:52 -0400 wip-5900
Dan Mick
2013-07-16 23:00:06 -0700 wip-5634
David Zafman
2014-06-05 00:22:34 -0700 wip-8231
Greg Farnum
2013-02-13 14:46:38 -0800 wip-mds-snap-fix
2013-02-22 19:57:53 -0800 w
On Mon, Jun 30, 2014 at 5:39 PM, Alex Elder wrote:
> On 06/25/2014 12:16 PM, Ilya Dryomov wrote:
>> Introduce ceph_osdc_cancel_request() intended for canceling requests
>> from the higher layers (rbd and cephfs). Because higher layers are in
>> charge and are supposed to know what and when they a
On Mon, Jun 30, 2014 at 5:50 PM, Alex Elder wrote:
> On 06/25/2014 12:16 PM, Ilya Dryomov wrote:
>> Linger requests that have not yet been registered should not be
>> unregistered by __unregister_linger_request(). This messes up ref
>> count and leads to use-after-free.
>>
>> Signed-off-by: Ilya
Hi Loic,
At this point I don't really have a way to look back in time to see
what was going on, but in the future when jobs are blocked waiting for
machines for unreasonable periods it's useful to know what is holding
them:
teuthology-lock --brief -a --machine-type plana | sort -k +4
Thanks,
Zac
On Sat, Jun 28, 2014 at 9:47 AM, Yuri Weinstein
wrote:
> altho 'teuthology-kill' is not
> working for me today :(
Uhoh, why not?
--
To unsubscribe from this list: send the line "unsubscribe ceph-devel" in
the body of a message to majord...@vger.kernel.org
More majordomo info at http://vger.kerne
On 06/25/2014 12:16 PM, Ilya Dryomov wrote:
> Linger requests that have not yet been registered should not be
> unregistered by __unregister_linger_request(). This messes up ref
> count and leads to use-after-free.
>
> Signed-off-by: Ilya Dryomov
> ---
> net/ceph/osd_client.c | 15 +++
On 06/25/2014 12:16 PM, Ilya Dryomov wrote:
> Introduce ceph_osdc_cancel_request() intended for canceling requests
> from the higher layers (rbd and cephfs). Because higher layers are in
> charge and are supposed to know what and when they are canceling, the
> request is not completed, only unref'
Hi,
On 30/06/2014 14:59, Andreas Joachim Peters wrote:
> Hi Loic,
>
> I was reading through the code and I have got the impression that they have
> formulated something simple with a lot of mathematical formulas in this
> paper. Indeed all ci = 1 ... which makes it simple ...
It makes things
On 06/25/2014 12:16 PM, Ilya Dryomov wrote:
> We should check if request is on the linger request list of any of the
> OSDs, not whether request is registered or not.
>
> Signed-off-by: Ilya Dryomov
That was a difficult to spot bug. Very good.
Reviewed-by: Alex Elder
> ---
> net/ceph/osd_cl
On 06/25/2014 12:16 PM, Ilya Dryomov wrote:
> Linger requests that have not yet been registered should not be
> unregistered by __unregister_linger_request(). This messes up ref
> count and leads to use-after-free.
This makes sense. The problem can occur when updating the OSD
map. An OSD *clien
Hi Loic,
I was reading through the code and I have got the impression that they have
formulated something simple with a lot of mathematical formulas in this paper.
Indeed all ci = 1 ... which makes it simple ...
The code seems to work like this e..g 10 + 2 + 4 :
Data = (1,2,3,4,5,6,7,8,9,10
On 06/25/2014 12:16 PM, Ilya Dryomov wrote:
> Add some WARN_ONs to alert us when we try to destroy requests that are
> still registered.
>
> Signed-off-by: Ilya Dryomov
Good idea. Especially the RB_CLEAR_NODE() call.
Reviewed-by: Alex Elder
> ---
> net/ceph/osd_client.c |7 +++
> 1
On 06/25/2014 12:16 PM, Ilya Dryomov wrote:
> It is important that both regular and lingering requests lists are
> empty when the OSD is removed.
Looks good.
Reviewed-by: Alex Elder
>
> Signed-off-by: Ilya Dryomov
> ---
> net/ceph/osd_client.c |2 ++
> 1 file changed, 2 insertions(+)
>
On 06/25/2014 12:16 PM, Ilya Dryomov wrote:
> Add dout()s to ceph_osdc_request_{get,put}(). Also move them to .c and
> turn kref release callback into a static function.
You can pretty much take the identical comments from
what I said on [PATCH 03/14].
Reviewed-by: Alex Elder
> Signed-off-by:
On 06/25/2014 12:16 PM, Ilya Dryomov wrote:
> Add dout()s to ceph_msg_{get,put}(). Also move them to .c and turn
> kref release callback into a static function.
>
> Signed-off-by: Ilya Dryomov
This is all very good.
I have one suggestion though, below, but regardless:
Reviewed-by: Alex Elder
On 06/25/2014 12:16 PM, Ilya Dryomov wrote:
> So that:
>
> req->r_osd_item --> osd->o_requests list
> req->r_linger_osd_item --> osd->o_linger_requests list
>
> Signed-off-by: Ilya Dryomov
This looks good and I prefer it too.
Reviewed-by: Alex Elder
> ---
> include/linux/ceph/osd_client.h |
On 06/25/2014 12:16 PM, Ilya Dryomov wrote:
> Abstract out __move_osd_to_lru() logic from __unregister_request() and
> __unregister_linger_request().
>
> Signed-off-by: Ilya Dryomov
Looks good.
Reviewed-by: Alex Elder
> ---
> net/ceph/osd_client.c | 26 ++
> 1 file
On 06/11/2014 11:40 AM, Ilya Dryomov wrote:
> The following check in rbd_img_obj_request_submit()
>
> rbd_dev->parent_overlap <= obj_request->img_offset
>
> allows the fall through to the non-layered write case even if both
> parent_overlap and obj_request->img_offset belong to the same RADOS
Hi Andreas,
TL;DR: which part of the code chooses the coefficients to maximize the fault
tolerance of the code as suggested in the Xorbas paper ?
If I understand correctly, the locality is computed (i.e. how many local parity
chunks are created) with:
https://github.com/madiator/HadoopUSC/blob
image_id is leaked if the parent happens to have been recorded already.
Fix it.
Signed-off-by: Ilya Dryomov
---
drivers/block/rbd.c |2 ++
1 file changed, 2 insertions(+)
diff --git a/drivers/block/rbd.c b/drivers/block/rbd.c
index d99aa81774f8..adedb393b374 100644
--- a/drivers/block/rbd.c
Hi Loic,
i think the best is to read along the sources. It is very readable!
https://github.com/madiator/HadoopUSC/blob/developUSC/src/contrib/raid/src/java/org/apache/hadoop/raid/SimpleRegeneratingCode.java
If there is a high interest in this, you could port the code from Java to C++
and use
Hi koleosfuscus,
It clarifies it enough to raise a question : where can I read code (or an
algorithm if not code) that chose the coefficients desirable to implement what
is suggested in the Xorbas paper ?
Cheers
On 30/06/2014 10:18, Koleos Fuscus wrote:
> Hi Loic,
>
> I am happy to contribute
Hi Loic,
I am happy to contribute with some clarifications. In fact,
erasure/reliability concepts are not blocking my progress with the
reliability model at ceph. It is the ceph model itself that has some
parts not clear to me, and nobody had time yet to review the state
model diagram that I publi
30 matches
Mail list logo