cloudxtreme / simple_distribute_filesystem Goto Github PK

data-structure <1> s_string -- ref count (ok) <2> s_array -- ref count (ok) <3> s_packet -- ref count (ok) <4> s_queue (ok) <5> s_list (ok) <6> s_hash(s_string/double/void * -> data block) (4 hour) (ok) debug record: <<<1>>> 在一个笔误上debug了近2个小时！ s_hash_set(...)返回的是指向val的指针 我一直当成了指向struct s_hash_node的指针 这该怎么破！

 						(total 4 hour)

schedule <1> server group (connect - auth - reconnect - keep alive - ping pong - roundup time) (xxx hour) (ok) <2> file descriptor (32-bit id + 32-bit version) (0.5 hour) <3> block descriptor (file descriptor + 16-bit block id + 16-bit version) (0.5 hour) <4> file meta-data (4 hour) <<1>> filename <<2>> file descriptor <<3>> file size <<4>> locate blocks <4.x> file meta-data in meta-server( file meta-meta-data ) (2 hour) <<1>> filename <<2>> file descriptor <<3>> locate file meta-data

<5> create_file_meta_data( udata, callback ) --> meta-data (4 hour) <<1>> udata --> { filename, size, void * p } <6> get_file_meta_data( udata, callback ) --> meta-data (4 hour) <<1>> udata --> { filename, block start, block end, void * p } <7> expand_file_size( udata, callback ) --> meta-data (3 hour) <<1>> udata --> { filename, nsize, void * p }

<8.x> get_block_metadata( udata ) --> block meta-data (4 hour)

<8> write_block( udata, callback ) --> none (4 hour) <<1>> udata --> { buf, len, void * p nblock, block descriptor array[] } <9> read_block( udata, callback ) --> none (4 hour) <<1>> udata --> { buf, len, void * p nblock, block descriptor array[] } <10> read_block_array( udata, callback ) --> none (2 hour) <<1>> udata --> { buf, len, void * p nblock, block descriptor array[] } <11> write_block_array( udata, callback ) --> none (2 hour) <<1>> udata --> { buf, len, void * p nblock, block descriptor array[] } <12> read_blocks( udata, callback ) --> none (2 hour) <<1>> udata --> { buf, len, void * p nblock, block descriptor start } <13> write_blocks( udata, callback ) --> none (2 hour) <<1>> udata --> { buf, len, void * p nblock, block descriptor start }

 											(total 34 hour)

create file

1. client --(param:create)--> meta-serv

2.  meta-serv (check auth) ...
   

   meta-serv --(can create ?)--> all other meta-serv meta-serv (set timeout -> failed)

3.          all other meta-serv (check auth) ...
   

4.  meta-serv <--(can or cannot)-- all other meta-serv
   

5.  meta-serv --(decide auth and main/slave meta-serv)--> all other meta-serv
   

6.          all meta-serv (create file-meta-meta-data)
   

7.          main meta-serv (create file-meta-data) ...
    main meta-serv --(file-meta-data)--> related data-serv
    main meta-serv <--(accept) -- related data-serv
   

8.  meta-serv <--(file-meta-data)-- main meta-serv
   

9. client <--(file-meta-data)-- meta-serv

module layer

1. net-layer raw epoll net layer, point-to-point connection based on tcp support : send/receive packets

2. server-group-layer based on net-layer all servers defined in config.conf will compose a big server-group support : each server will be connected with all other servers reconnect when connection is broken ping/pong keep-alive caculate roundup-time

3. core-layer(simple-distribute-filesystem layer) based on server-group-layer support : client mserver dserver