For compression, I prefer to implement it in ECpool, it is much easier
because objects in ECpool are already striped, which is what we have
already finished now(and in testing). And the only Append write
operation is allowed in EC, which is also lead us to implement it
conveniently. Moreover, as is
improve this?
Regards
Ning Yao
2015-10-31 21:18 GMT+08:00 Sage Weil :
> On Sat, 31 Oct 2015, ??? wrote:
>> hi, all:
>>
>> There are two ObjectStore::Transaction in
>> ReplicatedBackend::submit_transaction, one is op_t and the other one
>> is local_t. Is that s
> Good point. In my previous response I did "echo garbage >
> ./foo__head_7FC1F406__1" to corrupt a replica.
I think this may just happen when mixing O_DIRECT and buffer_io, which
may just happen in Newstore.
Or, inode content changes such as FileStore write "
./foo__head_7FC1F406
> BTW, latest code base is already separating out 2 transaction. No more append
> call.
Yeah, Got it. But still extra memory allocation in local_t like op_bl.
we may use the previous op_bl in op_t in most cases.
--
To unsubscribe from this list: send the line "unsubscribe ceph-devel" in
the body o
nable filestore_max_inline_xattr in the first test? If
not, it may be reasonable. In my previous test, I remember just about
20%~30% improvement.
And can you also provide cpu cost per Op on osd node?
Regards
Ning Yao
2015-10-30 10:04 GMT+08:00 Xue, Chendi :
> Hi, Sam
>
> Last week I introduced ab
, data_map (which can act as
the inode in filesystem), so that we can achieve the whole HDD iops as
real data without the interference of filesystem-journal and inode
get/set.
Regards
Ning Yao
2015-11-04 23:19 GMT+08:00 Chen, Xiaoxi :
> Hi Ning,
>
> Yes, we doesn’t save any IO, or may even need m
?
We may avoid frequently updating those information based on the pool types?
Regards
Ning Yao
--
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.kernel.org/majordomo-info.html
the things
like pg_info, attrs as they are always not reused in FileJournal
writing?
Regards
Ning Yao
2015-11-11 16:25 GMT+08:00 Evgeniy Firsov :
> Rb-tree construction, insertion, which needs memory allocation, mutex
> lock, unlock is more CPU expensive then streamlined crc calculation of
>
>>the code logic would touch crc cache is bufferlist::crc32c and invalidate_crc.
>Also for pg_log::_write_log(), but seems it is always miss and use at
>once, no need to cache crc actually?
Oh, no, it will be hit in FileJournal writing
Regards
Ning Yao
2015-11-11 18:03 GMT+08
er::ptr length diff with ::encode(crc, bl), right?
So the previous ebl.crc32c(0) calculation would be also no need to
cache.
Regards
Ning Yao
2015-11-11 18:05 GMT+08:00 Ning Yao :
>>>the code logic would touch crc cache is bufferlist::crc32c and
>>>invalidate_crc.
>>Also
2015-11-11 21:13 GMT+08:00 Sage Weil :
> On Wed, 11 Nov 2015, Ning Yao wrote:
>> >>>the code logic would touch crc cache is bufferlist::crc32c and
>> >>>invalidate_crc.
>> >>Also for pg_log::_write_log(), but seems it is always miss and use at
>>
osition &spos) {
dout(15) << __func__ << " " << cid << "/" << hoid << dendl;
Index index;
int r;
if(hoid.pgmeta())
goto out;
***
***
***
out:
r = object_map->set_keys(hoid, aset, &spos);
dout(20) << __func__
for all
backends.
Regards
Ning Yao
2015-11-18 21:12 GMT+08:00 Sage Weil :
> On Wed, 18 Nov 2015, Ning Yao wrote:
>> Hi, Sage
>>
>> pgmeta object is a meta-object (like __head___2) without
>> significant information. It is created when in PG::_init() when
>>
108.444us averagely (about 15% improvement), and reduce
the whole cpu time 0.5% ~ 1% globally.
Regards
Ning Yao
2015-12-08 21:37 GMT+08:00 Sage Weil :
>
> On Tue, 8 Dec 2015, Ning Yao wrote:
> > Umm, it seems that MemStore requires in memory meta object to keep the
> > attribut
this would
not be occurs.
We encounter this situation under two cases:
1) High load parallel requests comes from client
2) Or High miss rate for cache tier
Regards
Ning Yao
2015-12-09 1:56 GMT+08:00 Rongze Zhu :
>
> Hi guys,
>
> I found out a strange issue in a ceph cluste
Can we also consider to coalesce two OP_SETATTR transaction to a
single OP_SETATTRS transaction?
Regards
Ning Yao
2015-03-05 15:04 GMT+08:00 Haomai Wang :
> I think the performance improvement can be refer to
> https://github.com/ceph/ceph/pull/2972 which I did a simple benchmark
> c
2015-03-16 22:06 GMT+08:00 Haomai Wang :
> On Mon, Mar 16, 2015 at 10:04 PM, Xinze Chi wrote:
>> How to process the write request in primary?
>>
>> Thanks.
>>
>> 2015-03-16 22:01 GMT+08:00 Haomai Wang :
>>> AFAR Pipe and AsyncConnection both will mark self fault and shutdown
>>> socket and peer wi
2015-03-17 15:25 GMT+08:00 徐昕 :
> Hi Alexandre,
>
> I have tried this out. It can improve the performance of rbd rollback
> greatly when the difference between the image and the sanpshot is
> small.
>
If the clone does not happen in the rollback process, you may consider
it would properly happen wh
Thanks all guys. I got the ideas
Regards
Ning Yao
2015-03-17 21:58 GMT+08:00 Gregory Farnum :
> On Tue, Mar 17, 2015 at 6:46 AM, Sage Weil wrote:
>> On Tue, 17 Mar 2015, Ning Yao wrote:
>>> 2015-03-16 22:06 GMT+08:00 Haomai Wang :
>>> > On Mon, Mar 16, 2015
process if the content is owned by object data, or it may lead to osd
error, if the inconsistent data occurs in pg_log, pg_info or osdmap?)
Regards
Ning Yao
--
To unsubscribe from this list: send the line "unsubscribe ceph-devel" in
the body of a message to majord...@vger.kernel.org
More
or four
hours and tend to be stable after ten hours.
So does LRU or LRU2 makes sense here? or other strategies to make
training process converge faster?
Regards
Ning Yao
2015-03-18 2:18 GMT+08:00 Josh Durgin :
> On 03/17/2015 01:58 AM, Loic Dachary wrote:
>>
>>
>>
>>
, and some data in the rbd is
modified but some not, then after calling rollback. It seems lots of
objects (which is not modified before) also run clone() and generate
the new object. I do not investigate the reasons, you may try it.
Regards
Ning Yao
2015-03-17 17:13 GMT+08:00 徐昕 :
> Hi Ning
2015-03-20 10:22 GMT+08:00 Shu, Xinxin :
> I think rocksdb can support this configuration.
>
I do not find this option in rocksdb. If you know, can you provide
this option to redirect the WAL file?
Regards
Ning Yao
> Cheers,
> xinxin
>
> -Original Message-
>
bject when Calling can_skip_promote() and send a
CEPH_OSD_OP_DELETE op to cold pool from the Objecter interface, which
would be much better when deleting file occurs. Is that possible?
Regards
Ning Yao
--
To unsubscribe from this list: send the line "unsubscribe ceph-devel" in
the body
rollbacks so that we do not need to promote all
snap object if we just want the head (and actually most cases are)
Regards
Ning Yao
2015-03-30 15:05 GMT+08:00 Wang, Zhiqiang :
> How about handling the DELETE op in the cache tier like this:
> 1) If the object is in the cache tier, we delete it in
set the hit : miss rate for the test. Is there any way to
do this?
Regards
Ning Yao
--
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.kernel.org/majordomo-info.html
Currently, yes before we can improve the osd code efficiency further.
You can achieve better performance by using client writeback cache if
application allowed.
Regards
Ning Yao
2015-12-11 18:00 GMT+08:00 Zhi Zhang :
> Hi Guys,
>
> We have a small 4 nodes cluster. Here is the
also for a short-time
maintaining. Otherwise, it will trigger the backfill process. So wait
for Sage's opinion @sage
If you are interest on this, we may cooperate to do this.
Regards
Ning Yao
2015-12-25 14:23 GMT+08:00 Dong Wu :
> Thanks, from this pull request I learned that this issue is n
28 matches
Mail list logo