利用Python监控Oracle当前跑的TOP SESSION

Posted 2022-04-09 robinson1988

利用linux top命令抓出Oracle当前正在跑的进程，记录进程消耗的%CPU，%MEM，RES

根据PID查询当前跑的SQL，以及当前等待事件，监控程序每3秒执行一次

import cx_Oracle
import os
import time
os.environ['NLS_LANG']='SIMPLIFIED CHINESE_CHINA.UTF8'
command="top -bi -n 1 | grep oracle | grep -v top | grep -v _orcl | awk '"+'print $1 "," $9 "," $7"," $10'+"'"
#你需要将_orcl换成_sid
con=cx_Oracle.connect('scott', 'tiger', "192.168.56.10/orcl:pooled", cclass="PYTHON",
                            purity=cx_Oracle.ATTR_PURITY_SELF, encoding="UTF-8")
cur=con.cursor()
cur.arraysize = 100
sql='select decode(sql_id,null,a.program,sql_id),a.event from v$session a, v$process b where a.paddr=b.addr and b.spid='
while True:
 date = time.strftime("%Y-%m-%d %H:%M:%S", time.localtime())
 with os.popen(command, "r") as p:
   process = p.read()
 if process=='':
     print(str(date)+' System is idle')
 elif process!='':
  line=process.splitlines()
  for rows in line:
     row=rows.split(",")
     cur.execute(sql+row[0])
     for x in cur:
       sql_id=str(x[0])
       event=str(x[1])
       print(str(date)+','+row[0]+','+row[1]+'%CPU,'+ row[3]+'%MEM,'+str(round(int((row[2]))/1024))+'MB,'+sql_id+','+event)
 time.sleep(3)
cur.close()
con.close()

代码运行示例：

[oracle@server ~]$ python3 monitor.py
2020-05-18 13:41:07 System is idle
2020-05-18 13:41:11 System is idle
2020-05-18 13:41:14,3023,75.0%CPU,23.0%MEM,439MB,34149m3vuwxuh,direct path write temp
2020-05-18 13:41:17,3023,68.8%CPU,23.2%MEM,443MB,34149m3vuwxuh,db file scattered read
2020-05-18 13:41:20,3023,58.8%CPU,23.2%MEM,444MB,34149m3vuwxuh,db file scattered read
2020-05-18 13:41:24,3023,62.5%CPU,23.2%MEM,444MB,34149m3vuwxuh,direct path write temp

第一列是当前时间，第二列是PID，第三列是CPU使用率，第四列是内存使用率，

第五列是当前耗费内存，第六列SQL_ID，第七列等待事件

可以将上面python代码稍作改动，将查询的信息保存到CSV文件中，早上再导入数据库分析