依据开发反馈。近期每天早上7:30应用会报警。应用的日志显示数据库连接池满了。新的连接被拒绝。
首先。我做了ASH报告(报告区间:7:25 ~ 7:35),从ASH的等待事件发现enq: TX - row lock contention竟然高达76.54%。例如以下所看到的:
Top User Events
Event | Event Class | % Event | Avg Active Sessions |
---|---|---|---|
enq: TX - row lock contention | Application | 76.54 | 0.81 |
CPU + Wait for CPU | CPU | 12.76 | 0.14 |
db file sequential read | User I/O | 7.40 | 0.08 |
下一步就是找这个等待事件主要由哪些SQL引起的:
Top SQL with Top Events
SQL ID | Planhash | Sampled # of Executions | % Activity | Event | % Event | Top Row Source | % RwSrc | SQL Text |
---|---|---|---|---|---|---|---|---|
4rm17788qwxuy | 1272661853 | 54 | 69.45 | enq: TX - row lock contention | 69.45 | UPDATE | 69.45 | update shift_case set expertId... |
1cqbcdr0ufyk6 | 1272661853 | 10 | 5.20 | enq: TX - row lock contention | 5.20 | UPDATE | 5.20 | update shift_case set daySecti... |
1anu5c146v8d7 | 1272661853 | 4 | 1.89 | enq: TX - row lock contention | 1.89 | UPDATE | 1.89 | update shift_case set daySecti... |
gbw4zk8jv0n0u | 2588599834 | 10 | 1.57 | CPU + Wait for CPU | 0.79 | TABLE ACCESS - BY GLOBAL INDEX ROWID | 0.47 | select sc.scId, sc.estId, ct.c... |
dvmk92c1umc97 | 905317021 | 9 | 1.42 | CPU + Wait for CPU | 1.42 | CONNECT BY - NO FILTERING WITH START-WITH | 0.63 | select h.hospitaluuid id, h.pl... |
4rm17788qwxuy | update shift_case set expertId = :1 , shiftDate = :2 , daySection = :3 , rcLimit = :4 , orderingCount = :5 , shareRccount = :6 , clinicTypeUuid = :7 , fee = :8 , isTimeDivision = :9 , state = :10 , isopen=:11 , stateTime = :12 , updateTime = sysdate where scId =:13 |
好了,既然已经定位到问题就好办了,立即把应用开发者找来一问,真相大白:原来该应用须要从外部系统获取数据,为了让内部的数据库和外部的尽量保持一致,每次查询外部系统时,会在数据库里运行update语句。
解决的方法也简单:因为每次的Update都会把前一次的update覆盖(等于前面的update做的都是无用功),所以根本不是必需每次查询都update。仅仅要最后一次查询做update就能够了。