《Oracle_AWR_报告分析实例讲解.docx》由会员分享，可在线阅读，更多相关《Oracle_AWR_报告分析实例讲解.docx（45页珍藏版）》请在三一办公上搜索。
1、WORKLOAD REPOSITORY report for DB NameDB IdInstanceInst numReleaseRACHostICCI1314098396ICCI1110.2.0.3.0YESHPGICCI1Snap IdSnap TimeSessionsCursors/SessionBegin Snap:267825-Dec-08 14:04:50241.5End Snap:268025-Dec-08 15:23:37261.5Elapsed:78.79 (mins)DB Time:11.05 (mins)DB Time不包括Oracle后台进程消耗的时间。如果DB Ti
2、me远远小于Elapsed时间，说明数据库比较空闲。在79分钟里（其间收集了3次快照数据），数据库耗时11分钟，RDA数据中显示系统有8个逻辑CPU（4个物理CPU），平均每个CPU耗时1.4分钟，CPU利用率只有大约2%（1.4/79）。说明系统压力非常小。可是对于批量系统，数据库的工作负载总是集中在一段时间内。如果快照周期不在这一段时间内，或者快照周期跨度太长而包含了大量的数据库空闲时间，所得出的分析结果是没有意义的。这也说明选择分析时间段很关键，要选择能够代表性能问题的时间段。Report SummaryCache Sizes BeginEndBuffer Cache:3,344M3,3
3、44MStd Block Size:8KShared Pool Size:704M704MLog Buffer:14,352K显示SGA中每个区域的大小（在AMM改变它们之后），可用来与初始参数值比较。shared pool主要包括library cache和dictionary cache。library cache用来存储最近解析（或编译）后SQL、PL/SQL和Java classes等。library cache用来存储最近引用的数据字典。发生在library cache或dictionary cache的cache miss代价要比发生在buffer cache的代价高得多。因此sh
4、ared pool的设置要确保最近使用的数据都能被cache。Load ProfilePer SecondPer TransactionRedo size:918,805.72775,912.72Logical reads:3,521.772,974.06Block changes:1,817.951,535.22Physical reads:68.2657.64Physical writes:362.59306.20User calls:326.69275.88Parses:38.6632.65Hard parses:0.030.03Sorts:0.610.51Logons:0.010.01
5、Executes:354.34299.23Transactions:1.18% Blocks changed per Read:51.62Recursive Call %:51.72Rollback per transaction %:85.49Rows per Sort:#显示数据库负载概况，将之与基线数据比较才具有更多的意义，如果每秒或每事务的负载变化不大，说明应用运行比较稳定。单个的报告数据只说明应用的负载情况，绝大多数据并没有一个所谓“正确”的值，然而Logons大于每秒12个、Hard parses大于每秒100、全部parses超过每秒300表明可能有争用问题。Redo size：
6、每秒/每事务产生的redo大小（单位字节），可标志数据库任务的繁重程序。Logical reads：每秒/每事务逻辑读的块数Block changes：每秒/每事务修改的块数Physical reads：每秒/每事务物理读的块数Physical writes：每秒/每事务物理写的块数User calls：每秒/每事务用户call次数Parses：SQL解析的次数Hard parses：其中硬解析的次数，硬解析太多，说明SQL重用率不高。Sorts：每秒/每事务的排序次数Logons：每秒/每事务登录的次数Executes：每秒/每事务SQL执行次数Transactions：每秒事务数Block
7、s changed per Read：表示逻辑读用于修改数据块的比例Recursive Call：递归调用占所有操作的比率Rollback per transaction：每事务的回滚率Rows per Sort：每次排序的行数注:Oracle的硬解析和软解析 提到软解析(soft parse)和硬解析(hard parse)，就不能不说一下Oracle对sql的处理过程。当你发出一条sql语句交付Oracle，在执行和获取结果前，Oracle对此sql将进行几个步骤的处理过程：1、语法检查(syntax check)检查此sql的拼写是否语法。2、语义检查(semantic check)诸如
8、检查sql语句中的访问对象是否存在及该用户是否具备相应的权限。3、对sql语句进行解析(parse)利用内部算法对sql进行解析，生成解析树(parse tree)及执行计划(execution plan)。4、执行sql，返回结果(execute and return)其中，软、硬解析就发生在第三个过程里。Oracle利用内部的hash算法来取得该sql的hash值，然后在library cache里查找是否存在该hash值；假设存在，则将此sql与cache中的进行比较；假设“相同”，就将利用已有的解析树与执行计划，而省略了优化器的相关工作。这也就是软解析的过程。诚然，如果上面的2个假设中
9、任有一个不成立，那么优化器都将进行创建解析树、生成执行计划的动作。这个过程就叫硬解析。创建解析树、生成执行计划对于sql的执行来说是开销昂贵的动作，所以，应当极力避免硬解析，尽量使用软解析。Instance Efficiency Percentages (Target 100%) Buffer Nowait %:100.00Redo NoWait %:100.00Buffer Hit %:98.72In-memory Sort %:99.86Library Hit %:99.97Soft Parse %:99.92Execute to Parse %:89.09Latch Hit %:99.9
10、9Parse CPU to Parse Elapsd %:7.99% Non-Parse CPU:99.95本节包含了Oracle关键指标的内存命中率及其它数据库实例操作的效率。其中Buffer Hit Ratio 也称Cache Hit Ratio，Library Hit ratio也称Library Cache Hit ratio。同Load Profile一节相同，这一节也没有所谓“正确”的值，而只能根据应用的特点判断是否合适。在一个使用直接读执行大型并行查询的DSS环境，20%的Buffer Hit Ratio是可以接受的，而这个值对于一个OLTP系统是完全不能接受的。根据Oracle
11、的经验，对于OLTPT系统，Buffer Hit Ratio理想应该在90%以上。Buffer Nowait表示在内存获得数据的未等待比例。buffer hit表示进程从内存中找到数据块的比率，监视这个值是否发生重大变化比这个值本身更重要。对于一般的OLTP系统，如果此值低于80%，应该给数据库分配更多的内存。Redo NoWait表示在LOG缓冲区获得BUFFER的未等待比例。如果太低（可参考90%阀值），考虑增加LOG BUFFER。library hit表示Oracle从Library Cache中检索到一个解析过的SQL或PL/SQL语句的比率，当应用程序调用SQL或存储过程时，Ora
12、cle检查Library Cache确定是否存在解析过的版本，如果存在，Oracle立即执行语句；如果不存在，Oracle解析此语句，并在Library Cache中为它分配共享SQL区。低的library hit ratio会导致过多的解析，增加CPU消耗，降低性能。如果library hit ratio低于90%，可能需要调大shared pool区。Latch Hit：Latch是一种保护内存结构的锁，可以认为是SERVER进程获取访问内存数据结构的许可。要确保Latch Hit99%，否则意味着Shared Pool latch争用，可能由于未共享的SQL，或者Library Cach
13、e太小，可使用绑定变更或调大Shared Pool解决。Parse CPU to Parse Elapsd：解析实际运行时间/(解析实际运行时间+解析中等待资源时间)，越高越好。Non-Parse CPU ：SQL实际运行时间/(SQL实际运行时间+SQL解析时间)，太低表示解析消耗时间过多。Execute to Parse：是语句执行与分析的比例，如果要SQL重用率高，则这个比例会很高。该值越高表示一次解析后被重复执行的次数越多。In-memory Sort：在内存中排序的比率，如果过低说明有大量的排序在临时表空间中进行。考虑调大PGA。Soft Parse：软解析的百分比（softs/so
14、fts+hards），近似当作sql在共享区的命中率，太低则需要调整应用使用绑定变量。Shared Pool Statistics BeginEndMemory Usage %:47.1947.50% SQL with executions1:88.4879.81% Memory for SQL w/exec1:79.9973.52Memory Usage %：对于一个已经运行一段时间的数据库来说，共享池内存使用率，应该稳定在75%-90%间，如果太小，说明Shared Pool有浪费，而如果高于90，说明共享池中有争用，内存不足。SQL with executions1：执行次数大于1的sq
15、l比率，如果此值太小，说明需要在应用中更多使用绑定变量，避免过多SQL解析。Memory for SQL w/exec1：执行次数大于1的SQL消耗内存的占比。Top 5 Timed Events EventWaitsTime(s)Avg Wait(ms)% Total Call TimeWait ClassCPU time51577.6SQL*Net more data from client27,3196429.7Networklog file parallel write5,4974797.1System I/Odb file sequential read7,9003545.3User
16、 I/Odb file parallel write4,8063475.1System I/O这是报告概要的最后一节，显示了系统中最严重的5个等待，按所占等待时间的比例倒序列示。当我们调优时，总希望观察到最显著的效果，因此应当从这里入手确定我们下一步做什么。例如如果buffer busy wait是较严重的等待事件，我们应当继续研究报告中Buffer Wait和File/Tablespace IO区的内容，识别哪些文件导致了问题。如果最严重的等待事件是I/O事件，我们应当研究按物理读排序的SQL语句区以识别哪些语句在执行大量I/O，并研究Tablespace和I/O区观察较慢响应时间的文件。如
17、果有较高的LATCH等待，就需要察看详细的LATCH统计识别哪些LATCH产生的问题。在这里，log file parallel write是相对比较多的等待，占用了7%的CPU时间。通常，在没有问题的数据库中，CPU time总是列在第一个。更多的等待事件，参见本报告 的Wait Events一节。RAC StatisticsBeginEndNumber of Instances:22Global Cache Load Profile Per SecondPer TransactionGlobal Cache blocks received:4.163.51Global Cache bloc
18、ks served:5.975.04GCS/GES messages received:408.47344.95GCS/GES messages sent:258.03217.90DBWR Fusion writes:0.050.05Estd Interconnect traffic (KB)211.16Global Cache Efficiency Percentages (Target local+remote 100%) Buffer access - local cache %:98.60Buffer access - remote cache %:0.12Buffer access
19、- disk %:1.28Global Cache and Enqueue Services - Workload Characteristics Avg global enqueue get time (ms):0.1Avg global cache cr block receive time (ms):1.1Avg global cache current block receive time (ms):0.8Avg global cache cr block build time (ms):0.0Avg global cache cr block send time (ms):0.0Gl
20、obal cache log flushes for cr blocks served %:3.5Avg global cache cr block flush time (ms):3.9Avg global cache current block pin time (ms):0.0Avg global cache current block send time (ms):0.0Global cache log flushes for current blocks served %:0.4Avg global cache current block flush time (ms):3.0Glo
21、bal Cache and Enqueue Services - Messaging Statistics Avg message sent queue time (ms):0.0Avg message sent queue time on ksxp (ms):0.3Avg message received queue time (ms):0.5Avg GCS message process time (ms):0.0Avg GES message process time (ms):0.0% of direct sent messages:14.40% of indirect sent me
22、ssages:77.04% of flow controlled messages:8.56Main Report Wait Events Statistics SQL Statistics Instance Activity Statistics IO Stats Buffer Pool Statistics Advisory Statistics Wait Statistics Undo Statistics Latch Statistics Segment Statistics Dictionary Cache Statistics Library Cache Statistics Me
23、mory Statistics Streams Statistics Resource Limit Statistics init.ora Parameters Wait Events Statistics Time Model Statistics Wait Class Wait Events Background Wait Events Operating System Statistics Service Statistics Service Wait Class Stats Back to TopTime Model Statistics Total time in database
24、user-calls (DB Time): 663s Statistics including the word background measure background process time, and so do not contribute to the DB time statistic Ordered by % or DB time desc, Statistic name Statistic NameTime (s)% of DB TimeDB CPU514.5077.61sql execute elapsed time482.2772.74parse time elapsed
25、3.760.57PL/SQL execution elapsed time0.500.08hard parse elapsed time0.340.05connection management call elapsed time0.080.01hard parse (sharing criteria) elapsed time0.000.00repeated bind elapsed time0.000.00PL/SQL compilation elapsed time0.000.00failed parse elapsed time0.000.00DB time662.97backgrou
26、nd elapsed time185.19background cpu time67.48此节显示了各种类型的数据库处理任务所占用的CPU时间。Back to Wait Events Statistics Back to TopWait Class s - second cs - centisecond - 100th of a second ms - millisecond - 1000th of a second us - microsecond - 1000000th of a second ordered by wait time desc, waits desc Wait Class
27、Waits%Time -outsTotal Wait Time (s)Avg wait (ms)Waits /txnUser I/O66,8370.00120211.94System I/O28,2950.009335.05Network1,571,4500.00660280.72Cluster210,5480.0029037.61Other81,78371.8228014.61Application333,1550.0016059.51Concurrency5,1820.04510.93Commit9190.00440.16Configuration25,42799.46104.54Back
28、 to Wait Events Statistics Back to TopWait Events s - second cs - centisecond - 100th of a second ms - millisecond - 1000th of a second us - microsecond - 1000000th of a second ordered by wait time desc, waits desc (idle events last) EventWaits%Time -outsTotal Wait Time (s)Avg wait (ms)Waits /txnSQL
29、*Net more data from client27,3190.006424.88log file parallel write5,4970.004790.98db file sequential read7,9000.003541.41db file parallel write4,8060.003470.86db file scattered read10,3100.003131.84direct path write42,7240.003017.63reliable message3552.8218490.06SQL*Net break/reset to client333,0840
30、.0016059.50db file parallel read3,7320.001340.67gc current multi block request175,7100.0010031.39control file sequential read15,9740.001012.85direct path read temp1,8730.00950.33gc cr multi block request20,8770.00803.73log file sync9190.00440.16gc cr block busy5260.00360.09enq: FB - contention10,384
31、0.00301.85DFS lock handle3,5170.00310.63control file parallel write1,9460.00310.35gc current block 2-way4,1650.00200.74library cache lock4320.00240.08name-service call wait220.002760.00row cache lock3,8940.00200.70gcs log flush sync1,25942.02210.22os thread startup185.562890.00gc cr block 2-way3,671
32、0.00200.66gc current block busy1130.001120.02SQL*Net message to client1,544,1150.0010275.83gc buffer busy156.671700.00gc cr disk read3,2720.00100.58direct path write temp1590.00150.03gc current grant busy8980.00110.16log file switch completion290.001170.01CGS wait for IPC msg48,73999.87008.71gc curr
33、ent grant 2-way1,1420.00000.20kjbdrmcvtq lmon drm quiesce: ping completion90.000190.00enq: US - contention5670.00000.10direct path read1380.00010.02enq: WF - contention140.00090.00ksxr poll remote instances13,29158.45002.37library cache pin2110.00010.04ges global resource directory to be frozen9100.
34、000100.00wait for scn ack5830.00000.10log file sequential read360.00020.01undo segment extension25,34299.79004.53rdbms ipc reply2790.00000.05ktfbtgex6100.000100.00enq: HW - contention440.00010.01gc cr grant 2-way1580.00000.03enq: TX - index contention10.000340.00enq: CF - contention640.00010.01PX De
35、q: Signal ACK3721.62010.01latch free30.000100.00buffer busy waits6250.16000.11KJC: Wait for msg sends to complete1540.00000.03log buffer space110.00020.00enq: PS - contention460.00010.01enq: TM - contention700.00000.01IPC send completion sync40100.00000.01PX Deq: reap credit1,54499.81000.28log file
36、single write360.00000.01enq: TT - contention460.00000.01enq: TD - KTF dump entries120.00010.00read by other session10.000120.00LGWR wait for redo copy5400.00000.10PX Deq Credit: send blkd175.88000.00enq: TA - contention140.00000.00latch: ges resource hash list440.00000.01enq: PI - contention80.00000
37、.00write complete waits10.00020.00enq: DR - contention30.00000.00enq: MW - contention30.00000.00enq: TS - contention30.00000.00PX qref latch150100.00000.03enq: MD - contention20.00000.00latch: KCL gc element parent latch110.00000.00enq: JS - job run lock - synchronize10.00010.00SQL*Net more data to
38、client160.00000.00latch: cache buffers lru chain10.00000.00enq: UL - contention10.00000.00gc current split10.00000.00enq: AF - task serialization10.00000.00latch: object queue header operation30.00000.00latch: cache buffers chains10.00000.00latch: enqueue hash chains20.00000.00SQL*Net message from c
39、lient1,544,1130.0012,6268275.83gcs remote message634,88498.649,20314113.41DIAG idle wait23,6280.004,6161954.22ges remote message149,59193.454,6123126.72Streams AQ: qmn slave idle wait1670.004,611276110.03Streams AQ: qmn coordinator idle wait35147.864,611131370.06Streams AQ: waiting for messages in t
40、he queue488100.004,60594360.09virtual circuit status157100.004,596292720.03PX Idle Wait1,07297.112,58124070.19jobq slave wait14597.9342028960.03Streams AQ: waiting for time management or cleanup tasks1100.002702697470.00PX Deq: Parse Reply4040.00030.01PX Deq: Execution Msg12126.45000.02PX Deq: Join
41、ACK3842.11010.01PX Deq: Execute Reply3432.35000.01PX Deq: Msg Fragment160.00000.00Streams AQ: RAC qmn coordinator idle wait351100.00000.06class slave wait20.00000.00db file scattered read等待事件是当SESSION等待multi-block I/O时发生的，通过是由于full table scans或 index fast full scans。发生过多读操作的Segments可以在“Segments by P
42、hysical Reads”和 “SQL ordered by Reads”节中识别（在其它版本的报告中，可能是别的名称）。如果在OLTP应用中，不应该有过多的全扫描操作，而应使用选择性好的索引操作。DB file sequential read等待意味着发生顺序I/O读等待（通常是单块读取到连续的内存区域中），如果这个等待非常严重，应该使用上一段的方法确定执行读操作的热点SEGMENT，然后通过对大表进行分区以减少I/O量，或者优化执行计划（通过使用存储大纲或执行数据分析）以避免单块读操作引起的sequential read等待。通过在批量应用中，DB file sequential read是很影响性能的事件，总是应当设法避免。Log File Parallel Write事件是在等待LGWR进程将REDO记录从LOG 缓冲区写到联机日志文件时发生的。虽然写操作可能是并发的，但LGWR需要
