Oracle效能 - 常用查詢診斷及調整指令碼[不斷更新] (final)

一，現象：系統執行緩慢，負載較高

1. 實時查詢長時間操作的SQL(或者透過OEM圖形介面檢視)

使用檢視 V$SESSION_LONGOPS，查詢執行超過6秒鐘的操作，這些操作也包括很多備份
恢復功能，統計資訊收集，查詢操作，不同版本可能有更多的操作加入。

Oracle OEM中的長時間操作就是讀取這個檢視。要監控query execution progress，
必須滿足以下前提條件:
1) Oracle最佳化器使用 CBO;
2) 資料物件已經收集了統計資訊;
3) 初始化引數 TIMED_STATISTICS = true 或是 SQL_TRACE = true ;

如果是Oracle RAC, 可以使用以下指令碼：
SELECT     a.opname, a.SID, a.serial#,
           a.sql_id, a.start_time, a.time_remaining, a.elapsed_seconds,
           b.sql_fulltext, a.MESSAGE
    FROM   gv$session_longops a, gv$sql b
   WHERE       a.start_time > SYSDATE - 0.1
           AND a.time_remaining > 0 AND a.sql_id = b.sql_id
ORDER BY   a.start_time DESC;

如果是單機, 可以使用以下指令碼：
SELECT     a.opname,
           a.SID,
           a.serial#,
           a.sql_id,
           a.start_time,
           a.time_remaining,
           a.elapsed_seconds,
           b.sql_fulltext,
           a.MESSAGE,
           b.module,
           b.executions
   FROM   v$session_longops a, v$sql b
   WHERE a.start_time > SYSDATE - 0.1
           AND a.time_remaining > 0
           AND a.sql_id = b.sql_id
ORDER BY   a.start_time DESC;

2. 10g或以上版本透過檢視SYS.wrh$_active_session_history查詢歷史等待事件及SQL

查詢等待事件及TX鎖表

SELECT session_id, sql_id, program, blocking_session,
       blocking_session_serial#
FROM SYS.wrh$_active_session_history a, v$event_name b
WHERE sample_time BETWEEN TO_DATE ('2011/01/24 21:50', 'yyyy/mm/dd hh24:mi')
                       AND TO_DATE ('2011/01/24 22:35', 'yyyy/mm/dd hh24:mi')
   AND a.event_id = b.event_id;

SELECT session_id, sql_id, program, blocking_session,
       blocking_session_serial#
FROM SYS.wrh$_active_session_history a, v$event_name b
WHERE sample_time BETWEEN TO_DATE ('2011/01/24 21:50', 'yyyy/mm/dd hh24:mi')
                       AND TO_DATE ('2011/01/24 22:35', 'yyyy/mm/dd hh24:mi')
   AND a.event_id = b.event_id
   AND b.NAME = 'enq: TX - row lock contention';

SELECT session_id, sql_id, program, blocking_session,blocking_session_serial#
FROM   SYS.dba_hist_active_sess_history
WHERE sample_time BETWEEN TO_DATE ('2011/04/14 06:00', 'yyyy/mm/dd hh24:mi')
                       AND TO_DATE ('2011/04/14 08:30', 'yyyy/mm/dd hh24:mi')
                       AND session_type = 'FOREGROUND'
       -- AND event = '' ;

SELECT * FROM v$event_name

1.獲取最大的磁碟讀的SQL語句
SELECT *
FROM (SELECT PARSING_USER_ID EXECUTIONS,
               SORTS,
               COMMAND_TYPE,
               DISK_READS,
               sql_text
          FROM v$sqlarea
         ORDER BY disk_reads DESC)
WHERE ROWNUM < 20;

2. 透過Linux PID及SID查詢相關SQL及程式：

select a.sid,a.serial#, a.program,b.spid from v$session a,v$process b
where a.paddr=b.addr and b.spid in ('1245','2985','5884');

spid為top檢視到的os段的process id .

緊急處理時候可以透過kill session方式或直接kill os process來結束程式。
Alter system kill session 'SID,SERIAL#' ;
Kill -9 SPID (unix,linux)
orakill sid(oracle_instance_name) spid (windows)

select a.username,a.machine,a.program,b.spid,c.sql_text
from v$session a,v$process b,v$sqlarea c
where a.paddr=b.addr and c.hash_value=a.sql_hash_value
and c.address=a.sql_address and b.spid=12984 ;

知道有問題的SPID情況下檢視正在執行的SQL .

SELECT a.username,a.machine,a.program,a.sid,a.serial#,a.status,c.piece,c.sql_text
from v$session a,v$process b,v$sqltext c
WHERE b.spid='14150'
AND b.addr=a.paddr AND a.sql_address=c.address(+)order BY c.piece

3. 查詢Oracle庫中的Lock (或透過Toad中session browser或OEM參看lock情況)

檢視資料庫中的鎖(LOCK),找出程式及SQL

SELECT   se.inst_id, se.SID, se.serial#,lk.SID,
         se.username,se.OSUser,se.Machine,se.program,
         DECODE (lk.TYPE,
                 'TX', 'Transaction',
                 'TM', 'DML',
                 'UL', 'PL/SQL User Lock',
                 lk.TYPE)
            lock_type,
         DECODE (lk.lmode,
                 0, 'None',
                 1, 'Null',
                 2, 'Row-S (SS)',
                 3, 'Row-X (SX)',
                 4, 'Share',
                 5, 'S/Row-X (SSX)',
                 6, 'Exclusive',
                 TO_CHAR (lk.lmode))
            mode_held,
         DECODE (lk.request,
                 0, 'None',
                 1, 'Null',
                 2, 'Row-S (SS)',
                 3, 'Row-X (SX)',
                 4, 'Share',
                 5, 'S/Row-X (SSX)',
                 6, 'Exclusive',
                 TO_CHAR (lk.request))
            mode_requested,
         TO_CHAR (lk.id1) lock_id1,
         TO_CHAR (lk.id2) lock_id2,
         ob.owner,
         ob.object_type,
         ob.object_name,
         DECODE (lk.Block,
                 0,
                 'No',
                 1,
                 'Yes',
                 2,
                 'Global')
            block,
         se.lockwait,
         sq.sql_fulltext
FROM   GV$lock lk, dba_objects ob, GV$session se, GV$sql sq
WHERE       lk.TYPE IN ('TM', 'UL')
         AND lk.SID = se.SID
         AND lk.id1 = ob.object_id(+)
         AND (lk.inst_id = se.inst_id)
         AND sq.address=se.sql_address ;

軟硬解析數：

select a.value as total_Parse, b.value as hard_parse,
a.value - b.value as soft_parse
from v$mystat a , v$mystat b,v$statname aa , v$statname bb
where aa.name = 'parse count (total)'
and bb.name = 'parse count (hard)'
and aa.statistic# = a.statistic#
and bb.statistic# = b.statistic# ;

查詢壞塊：

SELECT segment_name,segment_type,extent_id,block_id, blocks
from dba_extents t where file_id = 10
AND 51896 between block_id and (block_id + blocks - 1)

查詢當前執行的程式：

select a.username,a.machine,a.program,b.spid,a.serial#,a.sid from v$session a,v$process b
where a.paddr=b.addr and b.spid in () ;

select a.username,a.machine,a.program,b.spid,a.serial#,a.sid from v$session a,v$process b
where a.paddr=b.addr and a.sid in () ;

select a.username,a.machine,a.program,b.spid,a.serial#,a.sid from v$session a,v$process b
where a.paddr=b.addr and a.machine in () ;

select a.serial#, a.sid,b.spid from v$session a,v$process b
where a.paddr=b.addr and b.spid in ('1245','2985','5884');

Alter system kill session 'SID,SERIAL#' ;

Kill -9 SPID (unix,linux)

orakill sid(oracle_instance_name) spid (windows)

4. 透過AWR查詢效能問題(預設有1小時延遲，不能查詢當前時間的session)

SQL> conn / AS SYSDBA
SQL> @/u01/product/oracle/rdbms/admin/awrrpt.sql
輸入 report_type 的值:
輸入 num_days 的值: 2      --- 現在到過去兩天時間內的snap id (可以檢視到).
輸入 begin_snap 的值: 2147    --- 輸入的開始及結束的snap id 對應您要查詢的出現問題的時間段。
輸入 end_snap 的值: 2182
輸入 report_name 的值:
Report written to awrrpt_1_2177_2182.html
SQL> exit

下載awrrpt_1_2177_2182.html並開啟檢視。

ASH儲存了系統最新的處於等待的會話記錄，可以用來診斷資料庫的當前狀態；
而AWR中的資訊最長可能有1小時的延遲，所以其取樣資訊並不能用於診斷資料
庫的當前狀態，但可以用來作為一段時期內資料庫效能調整的參考。

5. 查詢物理讀寫嚴重的SQL及查詢哪個SID最消耗資源

檢視佔I/O較大的正在執行的session

SELECT se.sid,se.serial#,pr.SPID,se.username,se.status,se.terminal,se.program,se.MODULE,
       se.sql_address,st.event,st.p1text,si.physical_reads,si.block_changes
FROM v$session se,v$session_wait st,v$sess_io si,v$process pr
WHERE st.sid=se.sid AND st.sid=si.sid AND se.PADDR=pr.ADDR AND se.sid>6 AND st.wait_time=0
       AND st.event NOT LIKE '%SQL%'
       ORDER BY physical_reads DESC ;

查詢物理讀寫嚴重的SQL

SELECT *
FROM (SELECT sql_text, module,
         disk_reads / DECODE (executions, 0, 1, executions) AS tt
        FROM v$sqlarea a
        ORDER BY tt DESC)
WHERE ROWNUM <= 20

查詢哪個SID最消耗資源

select s.sid,s.value "CPU Used"
from v$sesstat s,v$statname n
where s.statistic#=n.statistic# and n.name='CPU used by this session'
and s.value>0
order by 2 desc;

查詢前十條效能差的SQL

SELECT * FROM
(select PARSING_USER_ID,EXECUTIONS,SORTS,COMMAND_TYPE,DISK_READS,sql_text
FROM v$sqlarea order BY disk_reads DESC )
where ROWNUM<10 ;

6. 查詢物件統計資訊是否不是最新的。

select * from dba_tables
where wner='DFMS' and table_name='TEST' and LAST_ANALYZED >= sysdate-1
and status='VALID';

select * from dba_indexes where LAST_ANALYZED
where wner='DFMS' and index_name='IDX_TEST' and LAST_ANALYZED >= sysdate-1;

select * from dba_tab_columns where LAST_ANALYZED
where wner='DFMS' and table_name='TEST' and LAST_ANALYZED >= sysdate-1
and COLUMN_NAME='XXXXX' ;

7. 繫結變數窺視(Peeking)問題

9i, 10g 繫結變數窺視使得執行計劃出現變化。11g有改善。如果出現此類
問題導致的效能問題，需要升級到11g, 或者加入hint進行強制改變執行計劃。

8. 壞塊導致系統效能 (當然一般alert log中都有error, 查詢壞塊)

SELECT segment_name,segment_type,extent_id,block_id, blocks
from dba_extents t
where file_id = 10
AND 51896 between block_id and (block_id + blocks - 1)

8. Oracle9i, 10g buffer cache及Library Cache的命中率及其他

9i 命中率 :
// oracle9i data buffer hit ratio .

select a.value + b.value "logical_reads", c.value "phys_reads",
round(100 * ((a.value+b.value)-c.value) / (a.value+b.value)) "BUFFER HIT RATIO"
from v$sysstat a, v$sysstat b, v$sysstat c
where a.statistic# = 40 and b.statistic# = 41 and c.statistic# = 42;

10g 命中率 :
// oracle10g data buffer hit ratio .

select a.value + b.value "logical_reads", c.value "phys_reads",
round(100 * ((a.value+b.value)-c.value) / (a.value+b.value)) "BUFFER HIT RATIO"
from v$sysstat a, v$sysstat b, v$sysstat c
where a.statistic# = 47 and b.statistic# = 50 and c.statistic# = 54;

監控SGA Library Cache的命中率，應該小於1%

select sum(pins) "Total Pins", sum(reloads) "Total Reloads",
sum(reloads)/sum(pins) *100 libcache
from v$librarycache;

select sum(pinhits-reloads)/sum(pins) "hit radio",sum(reloads)/sum(pins) "reload percent"
from v$librarycache;

監控記憶體和硬碟的排序比率，最好使它小於 .10，增加PGA
SELECT name, value FROM v$sysstat WHERE name IN ('sorts (memory)', 'sorts (disk)');

9. redo log都處於active狀態。

select * from v$log ; 檢視redo log是否都出於active狀態，以及大小及組數
是否需要調大。

10. 碎片程度定期檢視

檢視碎片程度高的表

SELECT segment_name table_name , COUNT(*) extents
FROM dba_segments WHERE owner NOT IN ('SYS', 'SYSTEM') GROUP BY segment_name
HAVING COUNT(*) = (SELECT MAX( COUNT(*) ) FROM dba_segments GROUP BY segment_name);

分析index的碎片程度：

SQL> analyze index dfms.IDX1_WIP_D_WO_DETAIL_COMID validate structure ;

SQL> select name, del_lf_rows, lf_rows, round((del_lf_rows/(lf_rows+0.0000000001))*100)
2 "Frag Percent" from index_stats where name = 'IDX1_WIP_D_WO_DETAIL_COMID' ;

11. OS及網路檢視命令

top, sar, vmstat, iostat, ping , tnsping, route, traceroute,

Oracle效能 - 常用查詢診斷及調整指令碼[不斷更新] (final)

相關文章