转载

CPU 100%负载的性能优化分析

今天收到报警邮件,提示在短时间内DB time有了很大的抖动。报警邮件如下:
ZABBIX-监控系统:

------------------------------------
报警内容: DB time is too high
------------------------------------
报警级别: PROBLEM
------------------------------------
监控项目: DBtime543 %
------------------------------------
报警时间:2015.12.07-10:48:37
可以看到在蓝色的框中是问题发生时间段的DB time情况,其实后面直接飙到了5000%这个效果是很恐怖的。

CPU 100%负载的性能优化分析
而紧接着CPU负载也开始急剧飙升,直接的反应就是机器反应开始非常慢。根据top得到的进程,可以看到cpu资源已经被耗光了。
这个从sar的结果也可以看得很明显。
11:30:02 AM     CPU     %user     %nice   %system   %iowait    %steal     %idle
11:31:02 AM     all     95.86      0.00      0.36      0.00      0.00      3.78
11:32:02 AM     all     94.76      0.00      0.25      0.00      0.00      4.99
11:33:01 AM     all     94.53      0.00      0.17      0.00      0.00      5.30
11:34:01 AM     all     95.83      0.00      0.19      0.00      0.00      3.98
11:35:01 AM     all     95.18      0.00      0.17      0.00      0.00      4.66
11:36:01 AM     all     96.47      0.00      0.96      0.02      0.00      2.55
11:37:01 AM     all     97.50      0.00      0.15      0.08      0.00      2.28
11:38:01 AM     all     96.08      0.00      0.17      0.00      0.00      3.75
11:39:01 AM     all     95.74      0.00      0.17      0.01      0.00      4.08
11:40:01 AM     all     96.24      0.00      0.17      0.00      0.00      3.58
11:41:01 AM     all     94.68      0.00      0.34      0.00      0.00      4.98
11:42:01 AM     all     97.30      0.00      0.15      0.00      0.00      2.54
11:43:01 AM     all     96.51      0.00      0.18      0.00      0.00      3.31
11:44:01 AM     all     96.14      0.00      0.14      0.00      0.00      3.71
11:45:01 AM     all     95.52      0.00      0.20      0.01      0.00      4.26
11:46:01 AM     all     96.67      0.00      0.94      0.01      0.00      2.38
11:47:01 AM     all     96.23      0.00      0.16      0.01      0.00      3.61
11:48:01 AM     all     96.59      0.00      0.16      0.00      0.00      3.25
11:49:01 AM     all     96.40      0.00      0.18      0.00      0.00      3.41
11:50:02 AM     all     96.15      0.00      0.21      0.00      0.00      3.64
11:51:01 AM     all     96.72      0.00      0.59      0.04      0.00      2.64
11:52:01 AM     all     96.51      0.00      0.16      0.00      0.00      3.32
11:53:01 AM     all     94.57      0.00      0.20      0.01      0.00      5.22
11:54:01 AM     all     93.82      0.00      0.14      0.00      0.00      6.04
11:55:01 AM     all     94.04      0.00      0.14      0.01      0.00      5.81
11:56:01 AM     all     96.01      0.00      0.85      0.00      0.00      3.14
11:57:01 AM     all     95.66      0.00      0.15      0.01      0.00      4.17
11:58:01 AM     all     94.93      0.00      0.14      0.00      0.00      4.93
11:59:01 AM     all     97.15      0.00      0.16      0.01      0.00      2.68
12:00:01 PM     all     99.76      0.00      0.23      0.00      0.00      0.01
12:01:01 PM     all     99.45      0.00      0.55      0.00      0.00      0.01
12:02:01 PM     all     99.76      0.00      0.24      0.00      0.00      0.00
12:03:02 PM     all     99.60      0.00      0.40      0.00      0.00      0.00
12:04:02 PM     all     99.76      0.00      0.23      0.00      0.00      0.00
12:05:01 PM     all     99.77      0.00      0.23      0.00      0.00      0.00
12:06:05 PM     all     99.37      0.00      0.62      0.00      0.00      0.01
12:07:02 PM     all     99.59      0.00      0.40      0.00      0.00      0.00
12:08:03 PM     all     99.57      0.00      0.43      0.00      0.00      0.00
12:09:02 PM     all     99.66      0.00      0.34      0.00      0.00      0.00
12:10:01 PM     all     99.80      0.00      0.20      0.00      0.00      0.01
12:11:15 PM     all     99.70      0.00      0.30      0.00      0.00      0.00
12:12:07 PM     all     99.87      0.00      0.12      0.00      0.00      0.00
12:13:03 PM     all     99.74      0.00      0.26      0.00      0.00      0.00
12:14:02 PM     all     99.79      0.00      0.21      0.00      0.00      0.00
12:15:02 PM     all     99.85      0.00      0.14      0.00      0.00      0.01
12:16:04 PM     all     99.58      0.00      0.42      0.00      0.00      0.00
所以问题其实变得还是挺着急的。这种负载情况下,看似问题还在恶化,着实让人捏了一把汗。
top的情况如下:
top - 13:34:40 up 469 days, 20:09,  6 users,  load average: 44.01, 42.83, 43.48
Tasks: 515 total,  48 running, 453 sleeping,   0 stopped,  14 zombie
Cpu(s): 99.9%us,  0.1%sy,  0.0%ni,  0.0%id,  0.0%wa,  0.0%hi,  0.0%si,  0.0%st
Mem:  32946232k total, 32753852k used,   192380k free,   209652k buffers
Swap: 16771776k total,   186968k used, 16584808k free, 28973828k cached

  PID USER      PR  NI  VIRT  RES  SHR S %CPU %MEM    TIME+  COMMAND                                                                                                    
 8517 oracle    18   0 10.1g 305m 300m R 64.7  0.9 263:48.43 oracletestdb (LOCAL=NO)                                                                                   
17796 oracle    19   0 10.1g 174m 169m R 56.7  0.5  14:54.71 oracletestdb (LOCAL=NO)                                                                                   
18473 oracle    17   0 10.1g 171m 166m R 39.8  0.5  14:31.95 oracletestdb (LOCAL=NO)                                                                                   
18153 oracle    16   0 10.1g 175m 171m R 38.1  0.5  13:55.82 oracletestdb (LOCAL=NO)   
根据top的情况抓取了一个进程,看看这个进程到底在干嘛.比如进程号是16908.
可以看都原来是一个客户端触发的一个select语句导致。
$ sh showpid.sh 16908
*******************************************
Process has found, pid: 16908  ,  addr: 00000002DF6199B0    

####### Process Information from OS level as below ########
oracle   16908     1 17 12:00 ?        00:14:48 oracletestdb (LOCAL=NO)
##############################################
       SID    SERIAL# USERNAME        OSUSER          MACHINE              PROCESS         TERMINAL        TYPE       LOGIN_TIME
---------- ---------- --------------- --------------- -------------------- --------------- --------------- ---------- -------------------
       771      41269 TESTAT_NEWBG                    mbilocalhost         1234                            USER       2015-12-07 12:00:05

SQL_ID                         SQL_TEXT
------------------------------ ------------------------------------------------------------
7jywxtcmcmcgv                  select t0.stat_time stat_time,t2.consume_5cnt today,t1.consu
                               me_5cnt yestoday,t3.consume_5cnt lastweek  from   (select to
                               _char(trunc(sysdate)-level*5/60/24,'hh24:mi') as stat_time f
                               rom dual connect by level <=288  order by stat_time) t0  lef
                               t join   (select        to_char(a.stat_time,'hh24:mi') stat_
查看执行计划的情况如下,其实看计划都是走了索引扫描。
$ sh showplan2.sh 7jywxtcmcmcgv
PLAN_TABLE_OUTPUT
--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
SQL_ID  7jywxtcmcmcgv, child number 0
-------------------------------------
Plan hash value: 3396463774
------------------------------------------------------------------------------------------------------------
| Id  | Operation                         | Name                   | Rows  | Bytes | Cost (%CPU)| Time     |
------------------------------------------------------------------------------------------------------------
|   0 | SELECT STATEMENT                  |                        |       |       |   171 (100)|          |
|   1 |  SORT ORDER BY                    |                        |     3 |   300 |   171   (2)| 00:00:03 |
|   2 |   NESTED LOOPS OUTER              |                        |     3 |   300 |   170   (1)| 00:00:03 |
|   3 |    NESTED LOOPS OUTER             |                        |     1 |    68 |     8   (0)| 00:00:01 |
|   4 |     NESTED LOOPS OUTER            |                        |     1 |    36 |     5   (0)| 00:00:01 |
|   5 |      VIEW                         |                        |     1 |     4 |     2   (0)| 00:00:01 |
|   6 |       CONNECT BY WITHOUT FILTERING|                        |       |       |            |          |
|   7 |        FAST DUAL                  |                        |     1 |       |     2   (0)| 00:00:01 |
|*  8 |      TABLE ACCESS BY INDEX ROWID  | DY_USER_ANALYSIS_MIN   |     1 |    32 |     3   (0)| 00:00:01 |
|*  9 |       INDEX RANGE SCAN            | IND_USER_MIN_DATE_GAME |     2 |       |     2   (0)| 00:00:01 |
|* 10 |     TABLE ACCESS BY INDEX ROWID   | DY_USER_ANALYSIS_MIN   |     1 |    32 |     3   (0)| 00:00:01 |
|* 11 |      INDEX RANGE SCAN             | IND_USER_MIN_DATE_GAME |     2 |       |     2   (0)| 00:00:01 |
|* 12 |    TABLE ACCESS BY INDEX ROWID    | DY_USER_ANALYSIS_MIN   |     3 |    96 |   162   (1)| 00:00:02 |
|* 13 |     INDEX RANGE SCAN              | IND_USER_MIN_DATE_GAME |   481 |       |   155   (1)| 00:00:02 |
------------------------------------------------------------------------------------------------------------
一边分析一遍生成了一个调优报告。看看dbms_sqltune怎么建议。
$ sh gentunerpt.sh  7jywxtcmcmcgv
 TASK_68645
GENERAL INFORMATION SECTION
-------------------------------------------------------------------------------
Tuning Task Name                  : TASK_68645
Tuning Task Owner                 : TESTDBA
Scope                             : COMPREHENSIVE
Time Limit(seconds)               : 120
Completion Status                 : COMPLETED
Started at                        : 12/07/2015 13:28:55
Completed at                      : 12/07/2015 13:30:59
Number of Index Findings          : 1
Number of Errors                  : 1
-------------------------------------------------------------------------------
Schema Name: TESTAT_NEWBG
SQL ID     : 7jywxtcmcmcgv
完整的语句如下:
SQL Text   : select t0.stat_time stat_time,t2.consume_5cnt
             today,t1.consume_5cnt yestoday,t3.consume_5cnt lastweek
              from
              (select to_char(trunc(sysdate)-level*5/60/24,'hh24:mi') as
             stat_time from dual connect by level <=288  order by stat_time)
             t0
              left join
              (select
                    to_char(a.stat_time,'hh24:mi') stat_time,
                    nvl(a.consume_5cnt,0) consume_5cnt
                 from   DY_USER_ANALYSIS_MIN a
              where
                a.game_type = :1
                and a.stat_time >= to_date(:2, 'yyyy-mm-dd')-(INTERVAL '1'
             DAY)
                and a.stat_time < trunc(to_date(:3, 'yyyy-mm-dd'))
                and a.group_id = :4
                order by a.stat_time) t1 on t0.stat_time=t1.stat_time
                left join (select
                    to_char(a.stat_time,'hh24:mi') stat_time,
                    nvl(a.consume_5cnt,0) consume_5cnt
                 from   DY_USER_ANALYSIS_MIN a
              where
                a.game_type = :5
                and a.stat_time >= to_date(:6, 'yyyy-mm-dd')
                and a.stat_time < trunc(to_date(:7, 'yyyy-mm-dd'))+(INTERVAL
             '1' DAY)
                and a.group_id = :8
                order by a.stat_time) t2  on t0.stat_time=t2.stat_time
                left join (select
                    to_char(a.stat_time,'hh24:mi') stat_time,
                    nvl(a.consume_5cnt,0) consume_5cnt
                 from   DY_USER_ANALYSIS_MIN a
              where
                a.game_type = :9
                and a.stat_time >= to_date(:10, 'yyyy-mm-dd')-(INTERVAL '7'
             DAY)
                and a.stat_time < trunc(to_date(:11, 'yyyy-mm-dd'))-(INTERVAL
             '6' DAY)
                and a.group_id = :12
                order by a.stat_time) t3  on t3.stat_time=t0.stat_time
                   order by t0.stat_time
给出的建议是创建一个索引。
-------------------------------------------------------------------------------
FINDINGS SECTION (1 finding)
-------------------------------------------------------------------------------
1- Index Finding (see explain plans section below)
--------------------------------------------------
  The execution plan of this statement can be improved by creating one or more
  indices.
  Recommendation (estimated benefit: 92.38%)
  ------------------------------------------
  - Consider running the Access Advisor to improve the physical schema design
    or creating the recommended index.
    create index TESTAT_NEWBG.IDX$$_10C250001 on
    TESTAT_NEWBG.DY_USER_ANALYSIS_MIN("GROUP_ID","GAME_TYPE","STAT_TIME");
然后在执行计划的对比中,可以看到新的执行计划选择度要好很多。
- Using New Indices
--------------------
Plan hash value: 2029808490

----------------------------------------------------------------------------------------------------------
| Id  | Operation                         | Name                 | Rows  | Bytes | Cost (%CPU)| Time     |
----------------------------------------------------------------------------------------------------------
|   0 | SELECT STATEMENT                  |                      |     1 |   100 |    13   (8)| 00:00:01 |
|   1 |  SORT ORDER BY                    |                      |     1 |   100 |    13   (8)| 00:00:01 |
|   2 |   NESTED LOOPS OUTER              |                      |     1 |   100 |    12   (0)| 00:00:01 |
|   3 |    NESTED LOOPS OUTER             |                      |     1 |    68 |     8   (0)| 00:00:01 |
|   4 |     NESTED LOOPS OUTER            |                      |     1 |    36 |     5   (0)| 00:00:01 |
|   5 |      VIEW                         |                      |     1 |     4 |     2   (0)| 00:00:01 |
|*  6 |       CONNECT BY WITHOUT FILTERING|                      |       |       |            |          |
|   7 |        FAST DUAL                  |                      |     1 |       |     2   (0)| 00:00:01 |
|   8 |      TABLE ACCESS BY INDEX ROWID  | DY_USER_ANALYSIS_MIN |     1 |    32 |     3   (0)| 00:00:01 |
|*  9 |       INDEX RANGE SCAN            | IDX$$_10C250001      |     1 |       |     2   (0)| 00:00:01 |
|  10 |     TABLE ACCESS BY INDEX ROWID   | DY_USER_ANALYSIS_MIN |     1 |    32 |     3   (0)| 00:00:01 |
|* 11 |      INDEX RANGE SCAN             | IDX$$_10C250001      |     1 |       |     2   (0)| 00:00:01 |
|  12 |    TABLE ACCESS BY INDEX ROWID    | DY_USER_ANALYSIS_MIN |     1 |    32 |     4   (0)| 00:00:01 |
|* 13 |     INDEX RANGE SCAN              | IDX$$_10C250001      |     1 |       |     3   (0)| 00:00:01 |
----------------------------------------------------------------------------------------------------------
查看表的属性情况如下,可以看到当前存在两个索引。
$ sh showtab.sh  TESTAT_NEWBG DY_USER_ANALYSIS_MIN|less
*******************************************
OWNER                          TABLE_NAME
------------------------------ ------------------------------
TESTAT_NEWBG                   DY_USER_ANALYSIS_MIN    
*******************************************
********** TABLE GENERAL INFO *****************
TABLE_NAME                     PAR TABLESPACE STATUS  INI_TRANS   NUM_ROWS     BLOCKS EMPTY_BLOCKS LOG MON ROW_MOVE LAST_ANALYZED
------------------------------ --- ---------- ------ ---------- ---------- ---------- ------------ --- --- -------- -------------------
DY_USER_ANALYSIS_MIN           NO  TE_DATA    VALID           1     372113       3268            0 YES YES DISABLED 2015-12-05 06:00:19
********** TABLE STORAGE INFO *****************
  INITEXT    NXTEXT     MINEXT      MAXEXT  FREELISTS  AVG_SPACE  CHAIN_CNT AVG_ROW_LEN CACHE      T DEPENDEN COMPRES
--------- --------- ---------- ----------- ---------- ---------- ---------- ----------- ---------- - -------- ----------
    65536   1048576          1  2147483645                     0          0          61     N      N DISABLED DISABLED
********** TABLE columns INFO *****************
 COLUMN_ID COLUMN_NAME                    DATA_TYPE       DATA_LENGTH NULLABLE   DATA_DEFAULT
---------- ------------------------------ --------------- ----------- ---------- --------------------
         1 ID                             NUMBER(11,0)             22 N
         2 STAT_TIME                      DATE                      7 Y
         3 GAME_TYPE                      VARCHAR2(50)             50 Y
         4 ZONE_ID                        NUMBER                   22 Y          0
         5 GROUP_ID                       NUMBER                   22 Y          0
         6 ONLINE_5CNT                    NUMBER                   22 Y          0
         7 ONLINE_20CNT                   NUMBER                   22 Y          0
....
19 rows selected.
********** INDEX DETAILS INFO *****************
INDEX_NAME                     TABLESPACE INDEX_TYPE UNIQUENES PAR COLUMN_LIST                    TABLE_TYPE STATUS   NUM_ROWS LAST_ANALYZED       G
------------------------------ ---------- ---------- --------- --- ------------------------------ ---------- ------ ---------- ------------------- -
IND_USER_MIN_DATE_GAME         USERS      NORMAL     NONUNIQUE NO  STAT_TIME,GAME_TYPE            TABLE      VALID      371010 2015-12-05 06:00:21 N
SYS_C00106568                  TL_DATA    NORMAL     UNIQUE    NO  ID                             TABLE      VALID      371010 2015-12-05 06:00:20 Y
集群因子的情况如下:
TABLE_NAME                     INDEX_NAME                     CLUSTERING_FACTOR     BLOCKS   NUM_ROWS
------------------------------ ------------------------------ ----------------- ---------- ----------
DY_USER_ANALYSIS_MIN           IND_USER_MIN_DATE_GAME                      4700       3268     372113
DY_USER_ANALYSIS_MIN           SYS_C00106568                               4575       3268     372113
对于创建索引的如下建议。
create index TESTAT_NEWBG.IDX_DY_USER_ANA_IDTYPT on TESTAT_NEWBG.DY_USER_ANALYSIS_MIN("GROUP_ID","GAME_TYPE","STAT_TIME");
可以通过如下的方式来做个简单的分析。这个索引列的顺序也是蛮讲究,通过直方图的信息可以看到三个相关列的数据分布情况。
select column_name,num_distinct,high_value,low_value ,avg_col_len,histogram from dba_tab_col_statistics where table_name='DY_USER_ANALYSIS_MIN'
COLUMN_NAME    NUM_DISTINCT HIGH_VALUE      LOW_VALUE       AVG_COL_LEN HISTOGRAM
--------------------------- --------------- --------------------------- -----------
ID                   503808 C33E403E        C30C5164                  5 NONE
STAT_TIME              2992 78730C07170101  78730B1B0C3301            8 NONE
GAME_TYPE                 1 746C6262        746C6262                  5 FREQUENCY
ZONE_ID                   1 3E6466          3E6466                    4 FREQUENCY
GROUP_ID                136 C25C44          3E6466                    4 FREQUENCY
为什么选用了GROUP_ID作为首选列呢,其中一个原因就是范围查询和等值查询,在这个例子中范围查询就是stat_time相关的查询,等值查询就是group_id相关的。这种情况下是优先选择等值查询的。而game_type的数据分布很单一,所以这个列也不能作为首选列。
当然这些信息也是在做了简单的评估之后发现可行,所以马上部署了。可以看到部署之后负载立马降了下来。
CPU 100%负载的性能优化分析
可见这个变更也确实起到了立竿见影的效果。但是问题还不止于此,为什么这个语句造成了严重的性能问题,经过后续和开发同事的讨论,他们说这个新需求已经上线两周了。目前存在大量的left join的原因就是需要查询这周,上周,上上周。。某一天的数据情况,按照这种思路,其实我还是建议他们去掉这种冗余的left join,直接使用stat_time in (xxxx,xxxx,xxxx,xxx)其实这种更简明直接。当然问题解决了还不是结束,还需要后续跟进,保证这种问题从根本上杜绝。



正文到此结束
Loading...