Oracle Data block 的物理结构

  对data block物理结构的认识

    1.Data Block 物理结构图:

    2.一次对block的dump过程:

    PHP code:

SQL> create table t9 (a varchar(10));

Table created.

SQL> insert into t9 values('a');

1 row created.

SQL> commit;

Commit complete.

SQL> set serveroutput on
SQL> exec show_space('T9');
Free Blocks.............................1
Total Blocks............................8
Total Bytes.............................65536
Unused Blocks...........................6
Unused Bytes............................49152
Last Used Ext FileId....................3
Last Used Ext BlockId...................121
Last Used Block.........................2

PL/SQL procedure successfully completed.

SQL> alter system dump datafile 3 block 122;

System altered.

SQL> select * from v$tablespace;

       TS# NAME                           INC
---------- ------------------------------ ---
         0 SYSTEM                         YES
         1 UNDOTBS1                       YES
         8 USERS                          YES
        18 TEMP1

    Trace 文件:

*** 2004-07-25 15:48:01.000Start dump data blocks tsn: 8 file#: 3 minblk 122 maxblk 122buffer tsn: 8 rdba: 0x00c0007a (3/122)scn: 0x0000.0068d716 seq: 0x01 flg: 0x02 tail: 0xd7160601frmt: 0x02 chkval: 0x0000 type: 0x06=trans dataBlock header dump: 0x00c0007aObject id on Block? Yseg/obj: 0x806d csc: 0x00.68d714 itc: 2 flg: O typ: 1 - DATAfsl: 0 fnx: 0x0 ver: 0x01Itl Xid Uba Flag Lck Scn/Fsc0x01 0x0004.00b.00000fac 0x00801885.008c.56 --U- 1 fsc 0x0000.0068d7160x02 0x0000.000.00000000 0x00000000.0000.00 ---- 0 fsc 0x0000.00000000data_block_dump,data header at 0x552105c===============tsiz: 0x1fa0hsiz: 0x14pbl: 0x0552105cbdba: 0x00c0007a76543210flag=--------ntab=1nrow=1frre=-1fsbo=0x14fseo=0x1f9bavsp=0x1f83tosp=0x1f830xe:pti[0] nrow=1 offs=00x12:pri[0] offs=0x1f9bblock_row_dump:tab 0, row 0, @0x1f9btl: 5 fb: --H-FL-- lb: 0x1 cc: 1col 0: [ 1] 61end_of_block_dumpEnd dump data blocks tsn: 8 file#: 3 minblk 122 maxblk 122

 

 

3.先介绍数据块中包括的3种头信息:

    首先,数据块是通过data block buffer cache完成读和写操作的,所以它提供了20bytes的Cache Header和4bytes的 Tail给Cache,用来读取和管理 。

    第2部分是为Transaction层提供的Header信息。它一共占据了48bytes,包括24bytes的控制信息,和一系列的Interested Transaction Slot (ITS)。

    剩下的部分都叫Data Area,用来存储用户数据。Data Area也包括data header,和row data及剩余空间。但是Cluster blocks, table blocks, index block的data header,row data结构是不相同的,这里主要介绍table blocks.

    4. 结合trace文件中的信息,详细介绍:

    1) The Cache Header and Tail:

    buffer tsn: 8 rdba: 0x00c0007a (3/122)

    scn: 0x0000.0068d716 seq: 0x01 flg: 0x02 tail: 0xd7160601

    frmt: 0x02 chkval: 0x0000 type: 0x06=trans data

    Database block address: 占用4 bytes,表示Tablespace relative database block address(RDBA)。包括tns=8,即Tablespace number (User 表空间),file#=3, block_id=122

    SCN: 占用6bytes,表示最后变化的scn.包括2bytes的高位字节(SCN wrap),和4bytes的低位字节(SCN base)

    Sequence: 占用1byte,用途不明确,可能是辅助SCN的变化

    Flag: 占用1byte

    Format: 占用1byte,应该是用来区分版本。Oracle 8之前值为1,之后为2.

    Checksum: 占用2byte,跟db_block_checksum 参数有关系。

    引用oracle document 的解释:“DB_BLOCK_CHECKSUM determines whether DBWn and the direct loader will calculate a checksum (a number calculated from all the bytes stored in the block) and store it in the cache header of every data block when writing it to disk. Checksums are verified when a block is read-only if this parameter is true and the last write of the block stored a checksum. In addition, Oracle gives every log block a checksum before writing it to the current log.

    If this parameter is set to false, DBWn calculates checksums only for the SYSTEM tablespace, but not for user tablespaces.“

    Block type: 占用1byte,经常用到的有,1=undo segment header block; 2=undo data block; 5= data segment header block; 6=data block

    Unused: 保留的字节,占用4bytes,用来前后兼容。

    Tail 包括了SCN中SCN base的低位(low-order)2bytes,然后是block type,还有Sequence number.每当block 被读的时候,都要检查Tail与 block header 是否一致,保证了这个block不是损坏的(corrupted)。

    2) The Transaction Header:

    一共占据48bytes,包括24bytes的控制信息,和一系列的Interested Transaction Slot (ITS)。这些ITS组合在一起称为Interested Transaction List (ITL)。初始的ITL slot 数量由 INITRANS 决定(index branch block 只有1个slot)。如果有足够的剩余空间,oracle会根据需要动态的分配这些slot,直到受到空间限制或者达到了MAXTRANS.

Block header dump: 0x00c0007aObject id on Block? Yseg/obj: 0x806d csc: 0x00.68d714 itc: 2 flg: O typ: 1 - DATAfsl: 0 fnx: 0x0 ver: 0x01Itl Xid Uba Flag Lck Scn/Fsc0x01 0x0004.00b.00000fac 0x00801885.008c.56 --U- 1 fsc 0x0000.0068d7160x02 0x0000.000.00000000 0x00000000.0000.00 ---- 0 fsc 0x0000.00000000

    24bytes的控制信息包括:

    Object number(seg/obj): 占用4bytes,指在OBJ$中记录的segment 的 object number(0x806d=32877)

    Cleanout SCN(csc): 占用6bytes,最后一次 full cleanout 的scn

    ITL count(itc): 占用1byte,ITL 的slot数量。

    Flag: 占用2bytes.O表示这个block在freelist 上。否则flag为“-”

    Block type: 占用1byte.1=data; 2=index

    ITL freelist slot(fsl): 占用1byte.Index to the first slot on the ITL freelist

    Next freelist block(fnx): 占用4byte.Segment freelist中下一个block的RDBA

    Version: 1 byte

    Unused: 4bytes,用来前后兼容。

    每个ITL entry包括以下的内容:

    Transaction id(Xid): 8bytes.其中包括rollback segment number, transaction table中的slot number等。

    Undo block address(Uba): 8bytes.其中包括rollback segment block的DBA,sequence number等。

    Flags: 1nibble.—— = transaction is active, or committed pending cleanout C—— = transaction has been committed and locks cleaned out -B—— = this undo record contains the undo for this ITL entry——U- = transaction committed (maybe long ago); SCN is an upper bound——T = transaction was still active at block cleanout SCN

    Locks: 3nibbles. 也就是所谓的行级锁(row-level locks)

    SCN or free space credit: 6bytes. 如果这个事务已经clean out,这个值就是SCN;否则,前两个字节表示由这个事务释放的此block中的空间数。

    3)Data Area

    包括14bytes的data header,4bytes/table的table dictionary,2bytes/row的row dictionary.table dictionary主要用于cluster block中,只不过table block中的table dictionary只有一个table.

data_block_dump,data header at 0x552105c===============tsiz: 0x1fa0 ==> total data area sizehsiz: 0x14 ==> data header size (14+ntabs*4 + nrows*2)pbl: 0x0552105c ==> pointer to buffer holding the blockbdba: 0x00c0007a ==> block dba / rdba76543210flag=-------- ==> n=pctfree hit (clusters),f=don’t put on freelist, k=flushable cluster keysntab=1 ==> number of tables (>1 so this is a cluster)nrow=1 ==> number of rowsfrre=-1fsbo=0x14 ==> free space begin offsetfseo=0x1f9b ==> free space end offsetavsp=0x1f83 ==> available space in the blocktosp=0x1f83 ==> total available space when all transactions commit0xe:pti[0] nrow=1 offs=00x12:pri[0] offs=0x1f9bblock_row_dump:tab 0, row 0, @0x1f9b ==> 3bytes row header tl: 5 fb: --H-FL-- lb: 0x1 cc: 1col 0: [ 1] 61 ==> column length(1 byte if length<250; otherwise 3 bytes) and valuesend_of_block_dump

    5. 对于一个没有插入任何row的block来说,并且假设使用了默认的INITRANS,它的大小应该是:

db_block_size- (cache header+ transaction header +data header + Tail + table dictionary*ntab + row dictionary*row )=8192 - ( 20+48+14+4+4 +0)=8192 - 90 =8102 bytes

你可能感兴趣的:(oracle,header,table,byte,Dictionary,transactions)