用户反映19c生产环境数据库慢,运行简单语句uat只需要2s,但是prod需要40s,对此对数据库进行检查 首先检查数据库的awr报告,发现异常等待 acknowledge over PGA limit占用较大Top 10 Foreground Events by Total Wait Time
| Event | Waits | Total Wait Time (sec) | Avg Wait | % DB time | Wait Class |
|---|---|---|---|---|---|
| acknowledge over PGA limit | 5,255 | 1050.3 | 199.87ms | 90.2 | Scheduler |
| cursor: pin S wait on X | 6 | 72.6 | 12.10 s | 6.2 | Concurrency |
| DB CPU | 4.6 | .4 |
查询mos文档,发现此类问题是由于PGA大小达到了PGA_AGGREGATE_LIMIT的值,防止ORA-4036错误,后面进程分配pga时需要等待其他进程释放pga。 12c: 'acknowledge over PGA limit' Wait Event (Doc ID 2138882.1) CAUSE The "acknowlege over PGA limit" is a new wait event that was introduced with PGA_AGGREGATE_LIMIT in 12.1, and it will force a process that wants more PGA to wait a bit if the instance is getting close to hitting the limit. The hope is some other process will release memory and avoid the ORA-4036 error. SOLUTION 1. Set PGA_AGGREGATE_LIMIT=0 (as SYS user) to revert to 11g behavior of PGA memory management using PGA_AGGREGATE_TARGET parameter and the wait event will be alleviated. ALTER SYSTEM SET PGA_AGGREGATE_LIMIT=0 SID='*' SCOPE=BOTH; 2. Set PGA_AGGREGATE_LIMIT to higher than the default value (as SYS user), which was calculated based on the PGA_AGGREGATE_TARGET, PROCESSES or "_pga_limit_target_perc" setting and that will also help to reduce this wait event. ALTER SYSTEM SET PGA_AGGREGATE_LIMIT= &new_value SID='*' SCOPE=BOTH; oracle建议设置PGA_AGGREGATE_LIMIT=0或者增大这个参数的值解决 ALTER SYSTEM SET PGA_AGGREGATE_LIMIT=0 SCOPE=BOTH; 但是,这个方法只是暂时解决,究竟是什么导致的呢,需要继续检查pga使用情况首先判断该应用pga设置是否合理,即连接数数量和pga大小,可以通过查询process数量和pga使用,如果明显出现process较小而pga使用很大,则有可能是某些应用进程或者数据库进程异常导致,可以通过下面语句查询出是哪个进程异常导致 1.列出占用pga最大的进程 select pid,spid,substr(username,1,20) "USER" ,program,PGA_USED_MEM,PGA_ALLOC_MEM,PGA_FREEABLE_MEM,PGA_MAX_MEM from v$process where pga_alloc_mem=(select max(pga_alloc_mem) from v$process where program not like '%LGWR%'); 2.查看数据库会话所有sid占用pga的的详细情况 set linesize 120 set pagesize 120 column spid heading 'OSpid' format a8 column pid heading 'Orapid' format 999999 column sid heading 'Sess id' format 99999 column serial# heading 'Serial#' format 999999 column status heading 'Status' format a8 column pga_alloc_mem heading 'PGA alloc' format 99,999,999,999 column pga_used_mem heading 'PGA used' format 99,999,999,999 column username heading 'oracleuser' format a12 column osuser heading 'OS user' format a12 column program heading 'Program' format a20 SELECT p.spid, p.pid, s.sid, s.serial#, s.status, p.pga_alloc_mem, p.PGA_USED_MEM, s.username, s.osuser, s.program FROM v$process p, v$session s WHERE s.paddr ( + ) = p.addr and p.BACKGROUND is null /* Remove if need to monitor background processes */ Order by p.pga_alloc_mem desc;3.检查pga随着时间的增长select snap_id,round(value/1024/1024,0) from dba_hist_pgastat where name='total PGA allocated' order by snap_id4.检查应用进程分配的pga,查询出大于50m的select a.sid,a.status,a.sql_id,a.last_call_et,b.pga_alloc_mem/1024/1024,a.event,a.state from v$session a,v$process b where a.paddr=b.addr and a.status='INACTIVE' and b.pga_alloc_mem/1024/1024 > 50;5.查看当前pga使用情况select * from v$pgastat 通过1可以查询出是否有异常的数据库进程占用较大的pga,2可以查询出是否有异常会话占用较大pga,3可以查询出pga在最近一段时间的增长情况,可以判断出什么时候开始出现异常,4则是对应用进程使用Pga大于50m进行排查 如果是数据库进程异常,则需要检查该系统进程对应的trc文件进行分析。 如果是应用进程异常,则通过该sid和spid去分析, 对不释放的进程做heapdump,看看内存中导致存储的是什么内容,然后和应用一起解决
