alert日志中的一条ora警告信息的分析

今天照例检查数据库alert日志,发现一个错误。但是也没在意,想可能有大的操作导致的,马上会释放空间的,但是转眼一想,这是生产库,而且现在时早上,泰国的运营商还不算忙时,需要重视这个问题,看有没有什么潜在的问题,

从alert日志里面看到的

Fri Jul 12 09:08:23 ICT 2013

ORA-1652: unable to extend temp segment by 128 in tablespace                 TEMP

 

查询temp_usage,发现目前使用的只有goldengate的10多个session,占用的自用很少,查询现在的temp usage已经恢复正常了。

SQL> select TABLESPACE_NAME,TOTAL_BLOCKS,USED_BLOCKS,FREE_BLOCKS from v$sort_segment;


TABLESPACE_NAME                 TOTAL_BLOCKS USED_BLOCKS FREE_BLOCKS

------------------------------- ------------ ----------- -----------

TEMP                                 1023872        7936     1015936




导出awr报告,数据库整体负载很小。top sql里面看到的sql貌似都加了Hint,是被优化过的。


(awr报告时1小时一生成,可能有很多信息都不准确)

没办法,最后查ASH,精确到那一分钟,得到了以下的信息,


Service

Module

% Activity

Action

% Action

XXXX01

TOAD 9.6.1.1

83.08

UNNAMED

83.08

 

JDBC Thin Client

13.85

UNNAMED

13.85

 

并且发现下面的sql耗费了大量的资源,

Top SQL Statements

SQL ID

Planhash

% Activity

Event

% Event

SQL Text

7v8g1ffh5mwz7

3702571469

83.08

CPU + Wait for CPU

83.08

SELECT /*+ leading (ar1_charge...

d8x0ns0xjbrp9

1042878405

9.23

CPU + Wait for CPU

9.23

SELECT MT.SHORT_DESC, MO.ENTIT...

2979km1x69s3g

3257149028

1.54

CPU + Wait for CPU

1.54

SELECT AR_BALANCE FROM AR1_ACC..

 

猛一看,这个sql应用了大量的hint,细细一看,是一个很有问题的sql

关联了好几个大表,但是没有关联。

SQL details:

SQL Id

SQL Text

7v8g1ffh5mwz7

SELECT /*+ leading (xxxxx1 xxxx2  xxx3) use_nl (xxxxx1 xxxx2   xxx3) index (xxxxx1 xxxx2 _ix) index (xxxx2  xxxx2 _pk) */ xxxxx1 .CHARGE_ID, xxxxx2.debit_id, xxxx2.invoice_id, xxxx1.partition_id, xxxx1.period_key, ROW_NUMBER () OVER (ORDER BY xxxx2.DEBIT_ID DESC) RN FROM xxxx1, xxxx2, xxx3 WHERE xxxx1.ACCOUNT_ID = 10000027

 

最后马上和team里面确认了下,是有一个人执行的。

然后为了阻止隐患,为邮件给关联的team,对于sql的优化问题一点那个要优化转发到dba team。


看似一个很小的问题,可能包含着错误的操作。

 

来自 “ ITPUB博客 ” ,链接:http://blog.itpub.net/23718752/viewspace-766192/,如需转载,请注明出处,否则将追究法律责任。

转载于:http://blog.itpub.net/23718752/viewspace-766192/

  • 0
    点赞
  • 0
    收藏
    觉得还不错? 一键收藏
  • 0
    评论
Usage: ora [-u user] [-i instance#] [] General -u user/pass use USER/PASS to log in -i instance# append # to ORACLE_SID -sid set ORACLE_SID to sid -top # limit some large queries to on # rows - repeat Repeat an coomand time. Sleep between two calls Command are: - execute: cursors currently being executed - longops: run progression monitor - sessions: currently open sessions - stack get process stack using oradebug - cursors [all] : [all] parsed cursors - sharing : print why cursors are not shared - events [px]: events that someone is waiting for - events [read_by_other_session] events that someone is read by other session - ash [duration] [-f ] active session history for specified period e.g. 'ash 30' to display from [now - 30min] to [now] e.g. 'ash 30 10 -f foo.txt' to display a 10 minutes period from [now - 30min] and store the result in file foo.txt - ash_wait_graph [duration] [-f ] PQ event wait graph using ASH data Arguments are the same as for ash except that the output must be shown with the mxgraph tool - ash_sql Show all ash rows group by sampli_time and event for the specified sql_id - [-u ] degree degree of objects for a given user - [-u ] colstats stats for each table, column - [-u ] tabstats stats for each table - params []: view all parameters, even hidden ones - snap: view all snapshots status - bc: view contents of buffer cache - temp: view used space in temp tbs - asm: Show asm space/free space - space []: view used/free space in a given tbs - binds : display bind capture information for specified cursor - fulltext : display the entire SQL text of the specified statement - last_sql_hash []: hash value of the last styatement executed by the specified sid. If no sid speficied, return the last hash_value of user sessions - openv []: display optimizer env parameters for specified cursor - plan []: get explain plan of a particular cursor - pxplan : get explain plan of a particular cursor and all connected cursor slave SQL - wplan []: get explain plan with work area information - pxwplan : get explain plan with work area information of a particular cursor and all connected cursor slave SQL - eplan []: get explain plan with execution statistics - pxeplan : get explain plan with execution statistics of a particular cursor and all connected cursor slave SQL - gplan : get graphical explain plan of a particular cursor using dot specification - webplan get graphical explain plan of a particular [/] cursor using gdl specification []: optional: child_number, default is zero. optional: decorate to print further node information. default is 0, 1 => print further node information such as cost, filter_predicates etc. 2 => in addition to the above, print row vector information sample usage: # ora webplan 4019453623 print more information (decorate 1) # ora webplan 4019453623/1 1 more information, overload! (decorate 2) # ora webplan 4019453623/1 2 using sql_id along with child number instead of hash value # ora webplan aca4xvmz0rzup/3 1 - hash_to_sqlid : get the sql_id of the cursor given its hash value - sqlid_to_hash : get the hash value of the cursor given its (unquoted) sql_id - exptbs: generate export tablespace script - imptbs: generate import tablespace script - smm [limited]: SQL memory manager stats for active workareas - onepass: Run an ora wplan on all one-pass cursors - mpass: Run an ora wplan on all multi-pass cursors - pga: tell how much pga memory is used - pga_detail | -mem : Gives details on how PGA memory is consumed by a process (given its os PID) or by the set of precesses consuming more than MB of PGA memory (-mem option) - pgasnap [] Snapshot the pga advice stats - pgaadv [-s []] [-o graphfile] [-m min_size]: generate a graph from v and display it or store it in a file if the -o option is used. -s [] to diff with a previous snapshot (see pgasnap cmd) -o [graphfile] to store the result in a file instead of displaying it -m [min_size] only consider workareas with a minimum size - pgaadvhist [-f []] display the advice history for all factors or for factor between f_min and f_max - sga: tell how much sga memory is used - sga_stats: tell how sga is dynamically used - sort_usage: tell how temp tablespace is used in detail - sgasnap [] Snapshot the sga advice stats - sgaadv [-s []] [-o graphfile] generate a graph from v and v and store it in a file if the -o option is used. -s [] to diff with a previous snapshot (see sgasnap cmd) -o [graphfile] to store the result in a file instead of displaying it - process []: display process info with pga memory - version: display Oracle version number - cur_mem [ ] display the memory used for a given or all cursors - shared_mem [ ] detailed dump of cursor shared mem allocations - runtime_mem [ ] detailed dump of cursor runtime memory allocations - all_mem [ ] do all of the memory dumps - pstack |all [] run pstack on specified process (or all if 'all' specified) and store files in specified dir ( when not specified) - idxdesc [username] list all indexes for a given user or for a given user and table - segsize [username] list size of all objects(segments) for given user for a given user and object - tempu list temporary ts usage of all users or for a given user - sqlstats [ ] list sql execution stats (like buffer_gets, phy. reads etc) for a given sql_id/hash_value of statement - optstats [username] list optimizer stats for all tables stored in dictionary for a given user or for a given user and table - userVs list all user Views (user_tables, user_indexes etc) - fixedVs list all V$ Views - fixedXs list all X$ Views - px_processes list all px processes (QC and slaves) - cursor_summary summarize stats about (un)pinned cursors - rowcache summarizes row cache statistics - monitor_list lists all the statements that have been monitored - monitor [xml]: wraps dbms_sqltune.report_sql_monitor(). Directly passe the arguments to the PL/SQL procedure. Args are: sql_id, session_id, session_serial, sql_exec_start, sql_exec_id, inst_id, instance_id_filter, parallel_filter, report_level, type. Examples: - monitor xml shows XML report - monitor show last monitored stmt - monitor sql_id=>'8vz99cy9bydv8', session_id=>105 will show monitor info for sql_id 8vz99cy9bydv8 and session_id 105 Use simply ora monitor 8vz99cy9bydv8 to display monitoring information for sql_id 8vz99cy9bydv8. Syntax for parallel filters is: [qc][servers([,] [,] )] Use /*+ monitor */ to force monitoring. - monitor_old [ash_all] [] [qc| [ []]] Old version of SQL monitoring, use a SQL query versus the report_sql_monitor() package. Display monitoring info for the LAST execution of the specified cursor. Cursor response time needs to be at least 5s for monitoring to start (use the monitor hint to force monitoring). Without any parameter, will display monitoring info for the last cursor that was monitored - ash_all will aggregate ash data over all executions of the cursor (useful for short queries that are executed many times). If parallel: - qc to see only data for qc - slave_grp# to see only data for one parallelizer - slave_grp# + slave_set# to see only data for one slave set of one parallelizer, - slave_grp# + slave_set# + slave# to see data only for the specified slave - sql_task [progress | interrupt | history | report ] progress: progress monitoring for executing sql tasks interrupt: interrupt an executing sql task history: print a history of last n executions report: get a sql tune report - sql_use_temp_segment Find Who And What SQL Is Using Temp Segments. - sh Run a shell command. E.g. ora repeat 5 10 sh 'ps -edf | grep DESC' - awr_dbid Show AWR dbid - awr_dbtime [dbid] Show AWR dbtime - awr_dbtime [dbid] [inst] Show AWR dbtime - awr_dbtime_order [dbid] Show AWR dbtime order by desc - awr_sql_elaps_time [dbid] Show AWR SQL elapsed time - awr_sql_elaps_time [dbid] [inst] Show AWR SQL elapsed time - awr_sql_elaps_time_order [dbid] Show AWR SQL elapsed time order by desc - awr_logical_reads_order [dbid] - awr_logical_reads [dbid] Show AWR logical reads M Show AWR logical reads M order by desc - awr_physical_reads [dbid] Show AWR physical reads M - awr_physical_reads_order [dbid] Show AWR physical reads M order by desc - awr_db_cpu_per [dbid] [inst] Show AWR db_cpu_time cpu percent - awr_user_cpu_per [dbid] [inst] Show AWR oracle user_time cpu percent including backgroud process - awr_sql sql_id [dbid] Show AWR sql_id executions, per elapsed time. - awr_fulltext sql_id [dbid] Show AWR sql fulltext - awr_plan sql_id plan_hash [dbid] Show AWR sql plan, if plan_hash is null, show all plans. - awr_binds sql_id end_snap_id [dbid] Show AWR bind values in end_snap_id. - tab_frag owner [frag_percent] Show table fragment. - index_frag owner [frag_percent] Show index fragment. - rman_fullrestore_scripts dest_dbfile_dir Generate rman full database restore scripts - top_buffers_gets Top 10 by buffer gets > 10000 - top_physical_reads Top 10 by Physical Reads (disk_reads > 1000) - top_executions Top 10 by Executions > 100 - top_parse_calls Top 10 by Parse Calls > 1000 - top_sharable_memory Top 10 by Sharable Memory > 1M - top_version_count Top 10 by Version Count > 20 - top_cpu_usage Top 10 by CPU usage (cpu_time) - top_running_time Top 10 by Running Time (first_load_time desc) - create_tbs path size Create test database's tablespace script - create_tbs path size [dbid]Create dbid's test database's tablespace script - hold_txlock Show sessions holding a TX lock - wait_txlock Show sessions waiting a TX lock - rowid Display rowid's file_id, file_name, block info, object info, extent_id Memory: The detailed memory dumps need to have events set to work. The events bellow can be added to the init.ora file event="10277 trace name context forever, level 10" # mutable mem event="10235 trace name context forever, level 4" # shared mem NOTE ==== - Set environment variable ORA_USE_HASH to 1 to get SQL hash values instead of SQL ids - Set environment variable DBUSER to change default connect string which is "/ as sysdba" - Set environment variable ORA_TMP to the default temp directory (default if /tmp when not set)

“相关推荐”对你有帮助么?

  • 非常没帮助
  • 没帮助
  • 一般
  • 有帮助
  • 非常有帮助
提交
评论
添加红包

请填写红包祝福语或标题

红包个数最小为10个

红包金额最低5元

当前余额3.43前往充值 >
需支付:10.00
成就一亿技术人!
领取后你会自动成为博主和红包主的粉丝 规则
hope_wisdom
发出的红包
实付
使用余额支付
点击重新获取
扫码支付
钱包余额 0

抵扣说明:

1.余额是钱包充值的虚拟货币,按照1:1的比例进行支付金额的抵扣。
2.余额无法直接购买下载,可以购买VIP、付费专栏及课程。

余额充值