SQLite doesn't support high concurrency. In case of a lot of concurrent access from multi-process or multi-thread, calling a SQLite r/w function is prone to hit the case that the database is locked by another process or thread. The default behavior is that the SQLite r/w function return SQLITE_BUSY immediately, and the caller should retry the call later.

Below is an example function. When the SQLite function returns SQLITE_BUSY, the caller sleep one second then retry. The sleep is required to optimize the CPU utilization, without the sleep, the CPU will be occupied by a lot of retries.

bool Execute(const string& sql) {
char* errmsg = NULL;
while (true) {
int status = sqlite3_exec(db_, sql.c_str(), NULL, NULL, &errmsg);
if (status == SQLITE_OK) {
return true;
} else if (status == SQLITE_BUSY) {
++total_busycount;
cerr << "Thread " << threadindex_ << ": failed to execute " << sql << " (" << errmsg << ")\n";
sqlite3_free(errmsg);
sleep(1);
} else {
cerr << "Thread " << threadindex_ << ": failed to execute " << sql << " (" << errmsg << ")\n";
sqlite3_free(errmsg);
return false;
}
}
return false;
}

In fact, the caller doesn't need to sleep explicitly, calling sqlite3_busy_timeout before SQLite r/w functions will make SQLite automaticlly insert sleep between two sequential calls of a r/w function. It's safe to set a big timeout value. SQLite splits the big timeout value into many small timeout values. With the optimization, if the lock is available while the caller is waiting, the caller won't need to wait to timeout. See below code,

1452  static int sqliteDefaultBusyCallback(
1453 void *ptr, /* Database connection */
1454 int count /* Number of times table has been busy */
1455 ){
1456 #if SQLITE_OS_WIN || HAVE_USLEEP
1457 static const u8 delays[] =
1458 { 1, 2, 5, 10, 15, 20, 25, 25, 25, 50, 50, 100 };
1459 static const u8 totals[] =
1460 { 0, 1, 3, 8, 18, 33, 53, 78, 103, 128, 178, 228 };
1461 # define NDELAY ArraySize(delays)
1462 sqlite3 *db = (sqlite3 *)ptr;
1463 int timeout = db->busyTimeout;
1464 int delay, prior;
1465
1466 assert( count>=0 );
1467 if( count < NDELAY ){
1468 delay = delays[count];
1469 prior = totals[count];
1470 }else{
1471 delay = delays[NDELAY-1];
1472 prior = totals[NDELAY-1] + delay*(count-(NDELAY-1));
1473 }
1474 if( prior + delay > timeout ){
1475 delay = timeout - prior;
1476 if( delay<=0 ) return 0;
1477 }
1478 sqlite3OsSleep(db->pVfs, delay*1000);
1479 return 1;
1480 #else
1481 sqlite3 *db = (sqlite3 *)ptr;
1482 int timeout = ((sqlite3 *)ptr)->busyTimeout;
1483 if( (count+1)*1000 > timeout ){
1484 return 0;
1485 }
1486 sqlite3OsSleep(db->pVfs, 1000000);
1487 return 1;
1488 #endif
1489 }

Another alternative is set a busy hander by calling sqlite3_busy_handler. If the busy callback returns 0, then no additional attempts are made to access the database and SQLITE_BUSY is returned to the application. If the callback returns non-zero, then another attempt is made to access the database and the cycle repeats.

SQLite busy handler的更多相关文章

  1. 【腾讯Bugly干货分享】微信iOS SQLite源码优化实践

    本文来自于腾讯bugly开发者社区,非经作者同意,请勿转载,原文地址:http://dev.qq.com/topic/57b58022433221be01499480 作者:张三华 前言 随着微信iO ...

  2. SQLite Design and Concepts

    API 分为两大类 core API. 基本的SQL操作 extension API. 创建自定义的SQL操作. 基本数据结构 需要了解的组成部分有连接.statments.B树.pager. 为了写 ...

  3. SQLite剖析之内核研究

    先从全局的角度把握SQLite内核各个模块的设计和功能.SQLite采用了层次化.模块化的设计,而这些使得它的可扩展性和可移植性非常强.而且SQLite的架构与通用DBMS的结构差别不是很大,所以它对 ...

  4. SQLite入门与分析(三)---内核概述(2)

    写在前面:本节是前一节内容的后续部分,这两节都是从全局的角度SQLite内核各个模块的设计和功能.只有从全局上把握SQLite,才会更容易的理解SQLite的实现.SQLite采用了层次化,模块化的设 ...

  5. sqlite遇到database is locked问题的完美解决

    这两天在项目中用大强度大频率的方法测试时遇到sqlite报database is locked的问题,分析下来原因是sqlite对数据库做修改操作时会做(文件)锁使得其它进程同一时间使用时会报该错误( ...

  6. 【原】FMDB源码阅读(一)

    [原]FMDB源码阅读(一) 本文转载请注明出处 —— polobymulberry-博客园 1. 前言 说实话,之前的SDWebImage和AFNetworking这两个组件我还是使用过的,但是对于 ...

  7. FMDB源码解析

    上一篇博客讲述SQLite的使用,本篇将讲述FMDB源码,后面也会讲解SQLite在使用与FMDB的区别.本篇读下来大约20-30分钟,建议大家先收藏一下. FMDB是以OC方式封装SQLite中C语 ...

  8. Register A Callback To Handle SQLITE_BUSY Errors(译)

    http://www.sqlite.org/c3ref/busy_handler.html留着自己看的. Register A Callback To Handle SQLITE_BUSY Error ...

  9. ArcGIS API For Android Errors汇总

    API客户端异常错误的发生通常是由于不正确的方法参数,对象状态依赖,或网络条件. //*******推荐使用方法,按下Ctrl+F搜索错误代码,再查看与代码对应的解释.********// ArcGI ...

随机推荐

  1. Django模板语言(一)

    1,Django模板语言 1.1>基础语法:1.1.1,变量相关:{{ 变量名 }},1.1.2,逻辑相关{% ... %} 在Django模板语言中按此语法使用:{{ 变量名 }},当模板引擎 ...

  2. java8--异常处理(java疯狂讲义3复习笔记)

    try,catch,finally,throw,throws java将异常分为两种,Checked异常和Runtime异常. IndexOutOfBoundsException NumberForm ...

  3. 编程题:1. var person = '{name:"Lily",sex:"famale",age:24,country:"US"}';将person转换成JSON对象并便利每个属性值。

    /// <summary> /// Json工具类 /// </summary> public class JsonUtility { private static JsonU ...

  4. C 项目案例实践(1)数据结构之链表(0)

    链表是通过一组任意的存储单元来存储线性表中的数据元素的,那么怎样表示出数据元素之间的线性关系呢?为建立数据元素之间的线性关系,对每个数据元素ai,除了存放数据元素的自身信息ai之外,还需要存放和ai一 ...

  5. Windows 异步IO操作

    Windows提供了4种不同的方法来接收I/O请求已经完成的通知:触发设备内核对象.触发事件内核对象.可提醒I/O和I/O完成端口. Windows的异步I/O     当线程向设备发起一个I/O异步 ...

  6. 建立自己的私有docker(ssl&login auth)

    建立私有docker需要先建立ssl证书,然后建立htpass的登陆证书 最后使用nginx配置docker-compose.yml 参考: https://www.digitalocean.com/ ...

  7. IJ:IJ来了

    ylbtech-IJ:IJ来了 1.返回顶部 1. 2. 2.2. 3. 4. 2.返回顶部 1. 2. 3. 3. Welcome to IntelliJ IDEA返回顶部 1. 2.2. 2.2. ...

  8. ava Double: 四舍五入并设置小数点位数

    public static void main(String[] args) { // 1. 先乘后四舍五入, 再除; double d = 62.31060027198647; double d2 ...

  9. wamp的手动安装

    Wamp的手动安装 (http://www.cnblogs.com/homezzm/archive/2012/08/01/2618062.html) 一.Apache2.4安装 1.修改\Apache ...

  10. 洛谷 P3357 最长k可重线段集问题【最大流】

    pre:http://www.cnblogs.com/lokiii/p/8435499.html 和最长k可重区间集问题差不多,也就是价值的计算方法不一样,但是注意这里可能会有x0==x1的情况也就是 ...