前兩天折騰了mysqlsla,今天才發現hackmysql.com中的工具已淪爲歷史產物,建議使用Percona Toolkit代替。
1、下載安裝html
[root@VMUest tools]# wget https://www.percona.com/downloads/percona-toolkit/3.0.2/binary/tarball/percona-toolkit-3.0.2_i386.tar.gz [root@VMUest tools]# tar xf percona-toolkit-3.0.2_i386.tar.gz -C /usr/local/ [root@VMUest tools]# cd /usr/local/ [root@VMUest local]# mv /usr/local/percona-toolkit-3.0.2/ /usr/local/toolkit [root@VMUest local]# cd toolkit [root@VMUest toolkit]# cat README #配置編譯安裝 [root@VMUest toolkit]# perl Makefile.PL Checking if your kit is complete... Looks good Warning: prerequisite DBD::mysql 3 not found. Writing Makefile for percona-toolkit #根據需求安裝,以前mysqlsla已安裝部分依賴包 [root@VMUest toolkit]# yum install perl-DBD-MySQL [root@VMUest toolkit]# perl Makefile.PL [root@VMUest toolkit]# make && make test && make install
完成後相關工具自動添加到環境變量,個人機器上在/usr/local/bin/目錄下有對應的文件mysql
#查看安裝了哪些工具 [root@VMUest toolkit]# man percona-toolkit #查看具體工具如何使用 [root@VMUest toolkit]# man pt-query-digest [root@VMUest toolkit]# pt-query-digest --help
2、用法示例git
#1、直接分析慢查詢文件並將結果寫入118slow_report.log文件 [root@VMUest ~]# pt-query-digest /myshare/118mysqldumpslow-20161227.log > /myshare/118slow_report.log [root@VMUest ~]# pt-query-digest --group-by fingerprint --order-by Query_time:sum --limit 10 /myshare/118mysqldumpslow-20161227.log #2、查看指定fingerprint(參數化Query ID)的語句 [root@VMUest ~]# pt-query-digest /myshare/118mysqldumpslow-20161227.log --filter '$event->{fingerprint} && make_checksum($event->{fingerprint}) eq "6C0C2937AEF1D27C"' [root@VMUest ~]# pt-query-digest /myshare/118mysqldumpslow-20161227.log --no-report --output slowlog --filter '$event->{fingerprint} && make_checksum($event->{fingerprint}) eq "6C0C2937AEF1D27C"' #3、指定時間範圍內的查詢 [root@VMUest ~]# pt-query-digest /myshare/118mysqldumpslow-20161227.log --since '2016-12-27 07:51:20' --until '2016-12-27 10:10:34' #4、分析含有delete語句的慢查詢 [root@VMUest ~]# pt-query-digest --filter '$event->{fingerprint} =~ m/^delete/i' /myshare/118mysqldumpslow-20161227.log #5、把查詢保存到query_review表 [root@VMUest ~]# pt-query-digest --user=mydba --password=mysql5635 --review h=192.168.85.129,P=3306,D=test,t=query_review /myshare/118mysqldumpslow-20161227.log mysql> select * from test.query_review; #6、把查詢保存到query_history表 [root@VMUest ~]# pt-query-digest --user=mydba --password=mysql5635 --history h=192.168.85.129,P=3306,D=test,t=query_history /myshare/118mysqldumpslow-20161227.log mysql> select * from test.query_history; #7、分析binlog [root@VMUest ~]# mysqlbinlog /usr/local/mysql/log/mysql-bin.000001 > /myshare/mysql-bin000001.sql [root@VMUest ~]# pt-query-digest --type=binlog /myshare/mysql-bin000001.sql #8、分析general log mysql> set global general_log_file='/usr/local/mysql/log/mysql-general.log'; mysql> set global general_log =1; [root@VMUest ~]# pt-query-digest --type=genlog /usr/local/mysql/log/mysql-general.log
3、報告解釋github
[root@VMUest ~]# pt-query-digest --filter '$event->{fingerprint} =~ m/^delete/i' /myshare/118mysqldumpslow-20161227.log --limit 100% # 5.4s user time, 80ms system time, 13.06M rss, 17.15M vsz # Current date: Fri Mar 31 17:37:00 2017 # Hostname: VMUest # Files: /myshare/118mysqldumpslow-20161227.log # Overall: 8 total, 2 unique, 0.00 QPS, 0.03x concurrency ________________ # Time range: 2016-12-27 00:00:28 to 00:30:35 # Attribute total min max avg 95% stddev median # ============ ======= ======= ======= ======= ======= ======= ======= # Exec time 56s 256us 16s 7s 15s 5s 9s # Lock time 9s 20us 9s 1s 9s 3s 128us # Rows sent 0 0 0 0 0 0 0 # Rows examine 2.52M 0 657.18k 322.91k 650.62k 220.00k 410.39k # Query size 504 56 84 63 80.10 11.21 54.21 # Profile # Rank Query ID Response time Calls R/Call V/M Item # ==== ================== ============== ===== ====== ===== ============== # 1 0x6C0C2937AEF1D27C 56.1986 100.0% 6 9.3664 1.08 DELETE Ranking # 2 0xE0735FE2ADE92491 0.0008 0.0% 2 0.0004 0.00 DELETE Rank # Query 1: 0.00 QPS, 0.03x concurrency, ID 0x6C0C2937AEF1D27C at byte 5162025 # This item is included in the report because it matches --limit. # Scores: V/M = 1.08 # Time range: 2016-12-27 00:00:28 to 00:30:35 # Attribute pct total min max avg 95% stddev median # ============ === ======= ======= ======= ======= ======= ======= ======= # Count 75 6 # Exec time 99 56s 5s 16s 9s 15s 3s 10s # Lock time 99 9s 20us 9s 2s 9s 3s 37us # Rows sent 0 0 0 0 0 0 0 0 # Rows examine 100 2.52M 241.92k 657.18k 430.55k 650.62k 142.45k 487.82k # Query size 66 336 56 56 56 56 0 56 # String: # Databases Zipai # Hosts 192.168.85.129 # Users root # Query_time distribution # 1us # 10us # 100us # 1ms # 10ms # 100ms # 1s ################################################################ # 10s+ ################################ # Tables # SHOW TABLE STATUS FROM `Zipai` LIKE 'Ranking'\G # SHOW CREATE TABLE `Zipai`.`Ranking`\G delete from Ranking where eventTypeID=2 and rankTypeID=7\G # Converted for EXPLAIN # EXPLAIN /*!50100 PARTITIONS*/ select * from Ranking where eventTypeID=2 and rankTypeID=7\G # Query 2: 0 QPS, 0x concurrency, ID 0xE0735FE2ADE92491 at byte 5183064 __ # This item is included in the report because it matches --limit. # Scores: V/M = 0.00 # Time range: all events occurred at 2016-12-27 00:02:00 # Attribute pct total min max avg 95% stddev median # ============ === ======= ======= ======= ======= ======= ======= ======= # Count 25 2 # Exec time 0 827us 256us 571us 413us 571us 222us 413us # Lock time 0 730us 205us 525us 365us 525us 226us 365us # Rows sent 0 0 0 0 0 0 0 0 # Rows examine 0 0 0 0 0 0 0 0 # Query size 33 168 84 84 84 84 0 84 # String: # Databases Zipai (1/50%), test_db (1/50%) # Hosts % # Users root # Query_time distribution # 1us # 10us # 100us ################################################################ # 1ms # 10ms # 100ms # 1s # 10s+ # Tables # SHOW TABLE STATUS FROM `Zipai` LIKE 'Rank'\G # SHOW CREATE TABLE `Zipai`.`Rank`\G delete from Rank where typeID=16 or typeID=17 or typeID=13 or typeID=14 or typeID=15\G # Converted for EXPLAIN # EXPLAIN /*!50100 PARTITIONS*/ select * from Rank where typeID=16 or typeID=17 or typeID=13 or typeID=14 or typeID=15\G
3.一、整體統計結果
第一行:此工具執行日誌分析時的所用時間、內存資源(rss物理內存佔用大小,vsz虛擬內存佔用大小)
第二行:分析時的系統時間
第三行:進行分析的主機名,非記錄日誌的數據庫服務器
第四行:分析的日誌文件名稱
第五行:文件中總共的語句數量(Overall 8),惟一的語句數量(對語句進行了格式化,unique 2),QPS,併發數
Time range:查詢執行的時間範圍
total:總計; min:最小; max:最大; avg:平均;95%:把全部值從小到大排列,位置位於95%的那個數,這個數通常最具備參考價值;median:中位數,把全部值從小到大排列,位置位於中間那個數
Exec time:語句執行時間;Lock time:鎖佔有時間;Rows sent:發送到客戶端的行數;Row examine:掃描的行數;Query size:查詢語句的字符數
3.二、查詢分組統計結果
Response time:總的響應時間以及該查詢在本次分析中總的時間佔比
Calls:執行次數,即本次分析總共有多少條這種類型的查詢語句
R/Call:平均每次執行的響應時間
Item:查詢對象
3.三、每一種查詢的詳細統計結果
最上面的表格列出了百分比、總計、最大、最小、平均、95%等各項目的統計
Databases:庫名
Users:用戶
Query_time distribution:查詢時間分佈, 長短體現區間佔比,本例中1s-10s之間查詢數量是10s以上的兩倍
Tables:查詢中涉及到的表
Explain:示例,非select語句會轉換爲select語句
14:43 2017/6/20 補充sql
# 參數說明 --limit 95%:20 --outliers Query_time:1:10
limit默認返回前20條或者總響應時間佔比(從大到小)累計達到95%(取最早知足的)最差的查詢;outliers默認返回95%欄位的Query_time>=1秒,且執行次數>=10次。最終結果是取limit和outliers的並集。這也就是爲何顯示指定limit參數,返回的結果並不全知足limit參數,還會返回匹配outliers的結果。
若是使用了--limit或(和)--outliers參數,在返回結果的第二部分"查詢分組統計結果",rank的最末行是MISC,表明那些沒有計算入內的查詢統計結果。詳細內容參考官方文檔pt-query-digest.數據庫