thanks, sage weil.

writing  fs is a serious matter,we should make it clear, includes coding style.
there are other places we should fix.
thanks









At 2014-09-29 12:10:52, "Sage Weil" <sw...@redhat.com> wrote:
>On Mon, 29 Sep 2014, yuelongguang wrote:
>> hi, sage will1.
>> you mean if i use cache tiering, client's objecter can know how to connect
>> to osd daemon's objecter?
>> if i can see it  throught osdmap?
>> 2.
>> i know if rbd use cache, it use objecter, so i thought objecter is a cache
>> for IO scatter/gather.
>> i do not know 'COPYFROM operation ', you mention.
>>  
>> 3.
>> i search radosclient.cc, i know objecter is  a client for osd .
>> but for osd daemon, the code tells me  osd's objecter is a listen socket.
>> ceph-osd.cc : main:
>> ms_objecter->bind(g_conf->public_addr)
>>  
>> OSD:OSD  :: objecter_messenger(ms_objecter)
>>  
>> objecter_messenger->add_dispatcher_head(&service.objecter_dispatcher) , this
>> tells me all messages received from pipes(which is accepted by
>> objecter_messenger) are handled by  Objecter::dispatch. right?
>>  
>> no code tells me  this objecter connect other osds actively.
>
>Yeah.  I mean that the bind() call for ms_objecter is a mistake.  It only 
>initiates client-side connections, so it doesn't need to bind/listen.  
>I'll fix it shortly.  (It's pretty harmless, though.. just confusing.)
>
>sage
>
>
>>  
>> maybe i   miss something.
>> thanks
>> 
>> 
>> 
>> At 2014-09-29 11:23:37, "Sage Weil" <sw...@redhat.com> wrote:
>> >On Mon, 29 Sep 2014, yuelongguang wrote:
>> >> hi,all
>> >> 1.
>> >>  
>> >> and who will connect it? as for osd, this ms_objecter is listen socket.
>> >> it  is not included in osdmap. so how to know ms_objecter's  listen addre
>> ss
>> >> and connect it.
>> >
>> >Ah, that's a mistake.  It is only used to connect to other OSDs as a 
>> >client for the COPYFROM operation and for cache tiering.
>> >
>> >sage
>> 
>> 
>> 
>> 
_______________________________________________
ceph-users mailing list
ceph-users@lists.ceph.com
http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com

Reply via email to