高可用的MongoDB集群-实战篇
1.概述
最近有同学和网友私信我,问我MongoDB方面的问题;这里我整理一篇博客来赘述下MongoDB供大家学习参考,博客的目录内容如下:
- 基本操作
- CRUD
- MapReduce
本篇文章是基于MongoDB集群(Sharding+Replica Sets)上演示的,故操作的内容都是集群层面的,所以有些命令和单独的使用MongoDB库有异样。具体集群搭建可以参考我写的《高可用的MongoDB集群》。
2.基本操作
常用的 Shell 命令如下所示:
db.help() # 数据库帮助
db.collections.help() # 集合帮助
rs.help() # help on replica set
show dbs # 展示数据库名
show collections # 展示collections在当前库
use db_name # 选择数据库
查看集合基本信息,内容如下所示:
#查看帮助 db.yourColl.help(); #查询当前集合的数据条数 db.yourColl.count(); #查看数据空间大小 db.userInfo.dataSize(); #得到当前聚集集合所在的 db db.userInfo.getDB(); #得到当前聚集的状态 db.userInfo.stats(); #得到聚集集合总大小 db.userInfo.totalSize(); #聚集集合储存空间大小 db.userInfo.storageSize(); #Shard版本信息 db.userInfo.getShardVersion() #聚集集合重命名,将userInfo重命名为users db.userInfo.renameCollection("users"); #删除当前聚集集合 db.userInfo.drop();
3.CRUD
3.1创建
在集群中,我们增加一个 friends 库,命令如下所示:
db.runCommand({enablesharding:"friends"});
在库新建后,我们在该库下创建一个user分片,命令如下:
db.runCommand( { shardcollection : "friends. user"});
3.2新增
在MongoDB中,save和insert都能达到新增的效果。但是这两者是有区别的,在save函数中,如果原来的对象不存在,那他们都可以向collection里插入数据;如果已经存在,save会调用update更新里面的记录,而insert则会忽略操作。
另外,在insert中可以一次性插叙一个列表,而不用遍历,效率高,save则需要遍历列表,一个个插入,下面我们可以看下两个函数的原型,通过函数原型我们可以看出,对于远程调用来说,是一次性将整个列表post过来让MongoDB去处理,效率会高些。
Save函数原型如下所示:
Insert函数原型(部分代码)如下所示:
3.3查询
3.3.1查询所有记录
db. user.find();
默认每页显示20条记录,当显示不下的情况下,可以用it迭代命令查询下一页数据。注意:键入it命令不能带“;” 但是你可以设置每页显示数据的大小,用DBQuery.shellBatchSize= 50;这样每页就显示50条记录了。
3.3.2查询去掉后的当前聚集集合中的某列的重复数据
db. user.distinct("name"); #会过滤掉name中的相同数据 相当于: select distict name from user;
3.3.3查询等于条件数据
db.user.find({"age": 24});
#相当于: select * from user where age = 24;
3.3.4查询大于条件数据
db.user.find({age: {$gt: 24}}); # 相当于: select * from user where age >24;
3.3.5查询小于条件数据
db.user.find({age: {$lt: 24}}); #相当于: select * from user where age < 24;
3.3.6查询大于等于条件数据
db.user.find({age: {$gte: 24}}); #相当于: select * from user where age >= 24;
3.3.7查询小于等于条件数据
db.user.find({age: {$lte: 24}}); #相当于: select * from user where age <= 24;
3.3.8查询AND和OR条件数据
- AND
db.user.find({age: {$gte: 23, $lte: 26}}); #相当于 select * from user where age >=23 and age <= 26;
-
OR
db.user.find({$or: [{age: 22}, {age: 25}]}); #相当于: select * from user where age = 22 or age = 25;
3.3.9模糊查询
db.user.find({name: /mongo/}); #相当于%% select * from user where name like '%mongo%';
3.3.10开头匹配
db.user.find({name: /^mongo/});
# 与SQL中得like语法类似 select * from user where name like 'mongo%';
3.3.11指定列查询
db.user.find({}, {name: 1, age: 1}); #相当于: select name, age from user;
当然name也可以用true或false,当用ture的情况下和name:1效果一样,如果用false就是排除name,显示name以外的列信息。
3.3.12指定列查询+条件查询
db.user.find({age: {$gt: 25}}, {name: 1, age: 1}); #相当于: select name, age from user where age > 25;
db.user.find({name: 'zhangsan', age: 22});
#相当于:
select * from user where name = 'zhangsan' and age = 22;
3.3.13排序
#升序:
db.user.find().sort({age: 1});
#降序:
db.user.find().sort({age: -1});
3.3.14查询5条数据
db.user.find().limit(5); #相当于: select * from user limit 5;
3.3.15N条以后数据
db.user.find().skip(10); #相当于: select * from user where id not in ( select * from user limit 5 );
3.3.16在一定区域内查询记录
#查询在5~10之间的数据
db.user.find().limit(10).skip(5);
可用于分页,limit是pageSize,skip是第几页*pageSize。
3.3.17COUNT
db.user.find({age: {$gte: 25}}).count(); #相当于: select count(*) from user where age >= 20;
3.3.18安装结果集排序
db.userInfo.find({sex: {$exists: true}}).sort();
3.3.19不等于NULL
db.user.find({sex: {$ne: null}}) #相当于: select * from user where sex not null;
3.4索引
创建索引,并指定主键字段,命令内容如下所示:
db.epd_favorites_folder.ensureIndex({"id":1},{"unique":true,"dropDups":true}) db.epd_focus.ensureIndex({"id":1},{"unique":true,"dropDups":true})
3.5更新
update命令格式,如下所示:
db.collection.update(criteria,objNew,upsert,multi)
参数说明: criteria:
查询条件 objNew:update对象和一些更新操作符
upsert:如果不存在update的记录,是否插入objNew这个新的文档,true为插入,默认为false,不插入。
multi:默认是false,只更新找到的第一条记录。如果为true,把按条件查询出来的记录全部更新。
下面给出一个示例,更新id为 1 中 price 的值,内容如下所示:
db. user.update({id: 1},{$set:{price:2}}); #相当于: update user set price=2 where id=1;
3.6删除
3.6.1删除指定记录
db. user. remove( { id:1 } ); #相当于: delete from user where id=1;
3.6.2删除所有记录
db. user. remove( { } ); #相当于: delete from user;
3.6.3DROP
db. user. drop(); #相当于: drop table user;
4.MapReduce
MongoDB中的 MapReduce 是编写JavaScript脚本,然后由MongoDB去解析执行对应的脚本,下面给出 Java API 操作MR。代码如下所示:
MongdbManager类,用来初始化MongoDB:
package cn.mongo.util; import org.slf4j.Logger; import org.slf4j.LoggerFactory; import com.mongodb.DB; import com.mongodb.Mongo; import com.mongodb.MongoOptions; /** * @Date Mar 3, 2015 * * @author dengjie * * @Note mongodb manager */ public class MongdbManager { private static final Logger logger = LoggerFactory.getLogger(MongdbManager.class); private static Mongo mongo = null; private static String tag = SystemConfig.getProperty("dev.tag"); private MongdbManager() { } static { initClient(); } // get DB object public static DB getDB(String dbName) { return mongo.getDB(dbName); } // get DB object without param public static DB getDB() { String dbName = SystemConfig.getProperty(String.format("%s.mongodb.dbname", tag)); return mongo.getDB(dbName); } // init mongodb pool private static void initClient() { try { String[] hosts = SystemConfig.getProperty(String.format("%s.mongodb.host", tag)).split(","); for (int i = 0; i < hosts.length; i++) { try { String host = hosts[i].split(":")[0]; int port = Integer.parseInt(hosts[i].split(":")[1]); mongo = new Mongo(host, port); if (mongo.getDatabaseNames().size() > 0) { logger.info(String.format("connection success,host=[%s],port=[%d]", host, port)); break; } } catch (Exception ex) { ex.printStackTrace(); logger.error(String.format("create connection has error,msg is %s", ex.getMessage())); } } // 设置连接池的信息 MongoOptions opt = mongo.getMongoOptions(); opt.connectionsPerHost = SystemConfig.getIntProperty(String.format("%s.mongodb.poolsize", tag));// poolsize opt.threadsAllowedToBlockForConnectionMultiplier = SystemConfig.getIntProperty(String.format( "%s.mongodb.blocksize", tag));// blocksize opt.socketKeepAlive = true; opt.autoConnectRetry = true; } catch (Exception e) { e.printStackTrace(); } } }
MongoDBFactory类,用来封装操作业务代码,具体内容如下所示:
package cn.mongo.util; import java.util.ArrayList; import java.util.List; import org.slf4j.Logger; import org.slf4j.LoggerFactory; import cn.diexun.domain.MGDCustomerSchema; import com.mongodb.BasicDBList; import com.mongodb.DB; import com.mongodb.DBCollection; import com.mongodb.DBObject; import com.mongodb.util.JSON; /** * @Date Mar 3, 2015 * * @Author dengjie */ public class MongoDBFactory { private static Logger logger = LoggerFactory.getLogger(MongoDBFactory.class); // save data to mongodb public static void save(MGDCustomerSchema mgs, String collName) { DB db = null; try { db = MongdbManager.getDB(); DBCollection coll = db.getCollection(collName); DBObject dbo = (DBObject) JSON.parse(mgs.toString()); coll.insert(dbo); } catch (Exception ex) { ex.printStackTrace(); logger.error(String.format("save object to mongodb has error,msg is %s", ex.getMessage())); } finally { if (db != null) { db.requestDone(); db = null; } } } // batch insert public static void save(List<?> mgsList, String collName) { DB db = null; try { db = MongdbManager.getDB(); DBCollection coll = db.getCollection(collName); BasicDBList data = (BasicDBList) JSON.parse(mgsList.toString()); List<DBObject> list = new ArrayList<DBObject>(); int commitSize = SystemConfig.getIntProperty("mongo.commit.size"); int rowCount = 0; long start = System.currentTimeMillis(); for (Object dbo : data) { rowCount++; list.add((DBObject) dbo); if (rowCount % commitSize == 0) { try { coll.insert(list); list.clear(); logger.info(String.format("current commit rowCount = [%d],commit spent time = [%s]s", rowCount, (System.currentTimeMillis() - start) / 1000.0)); } catch (Exception ex) { ex.printStackTrace(); logger.error(String.format("batch commit data to mongodb has error,msg is %s", ex.getMessage())); } } } if (rowCount % commitSize != 0) { try { coll.insert(list); logger.info(String.format("insert data to mongo has spent total time = [%s]s", (System.currentTimeMillis() - start) / 1000.0)); } catch (Exception ex) { ex.printStackTrace(); logger.error(String.format("commit end has error,msg is %s", ex.getMessage())); } } } catch (Exception ex) { ex.printStackTrace(); logger.error(String.format("save object list to mongodb has error,msg is %s", ex.getMessage())); } finally { if (db != null) { db.requestDone(); db = null; } } } }
LoginerAmountMR类,这是一个统计登录用户数的MapReduce计算类,代码如下:
package cn.mongo.mapreduce; import java.sql.Timestamp; import java.util.ArrayList; import java.util.Date; import java.util.List; import org.bson.BSONObject; import org.slf4j.Logger; import org.slf4j.LoggerFactory; import cn.diexun.conf.ConfigureAPI.MR; import cn.diexun.conf.ConfigureAPI.PRECISION; import cn.diexun.domain.Kpi; import cn.diexun.util.CalendarUtil; import cn.diexun.util.MongdbManager; import cn.diexun.util.MysqlFactory; import com.mongodb.DB; import com.mongodb.DBCollection; import com.mongodb.DBCursor; import com.mongodb.DBObject; import com.mongodb.MapReduceOutput; import com.mongodb.ReadPreference; /** * @Date Mar 13, 2015 * * @Author dengjie * * @Note use mr jobs stats user login amount */ public class LoginerAmountMR { private static Logger logger = LoggerFactory.getLogger(LoginerAmountMR.class);
// map 函数JS字符串拼接 private static String map() { String map = "function(){"; map += "if(this.userName != \"\"){"; map += "emit({" + "kpi_code:'login_times',username:this.userName," + "district_id:this.districtId,product_style:this.product_style," + "customer_property:this.customer_property},{count:1});"; map += "}"; map += "}"; return map; }
private static String reduce() { String reduce = "function(key,values){"; reduce += "var total = 0;"; reduce += "for(var i=0;i<values.length;i++){"; reduce += "total += values[i].count;}"; reduce += "return {count:total};"; reduce += "}"; return reduce; }
// reduce 函数字符串拼接 public static void main(String[] args) { loginNumbers("t_login_20150312"); } /** * login user amount * * @param collName */ public static void loginNumbers(String collName) { DB db = null; try { db = MongdbManager.getDB(); db.setReadPreference(ReadPreference.secondaryPreferred()); DBCollection coll = db.getCollection(collName); String result = MR.COLLNAME_TMP; long start = System.currentTimeMillis(); MapReduceOutput mapRed = coll.mapReduce(map(), reduce(), result, null); logger.info(String.format("mr run spent time=%ss", (System.currentTimeMillis() - start) / 1000.0)); start = System.currentTimeMillis(); DBCursor cursor = mapRed.getOutputCollection().find(); List<Kpi> list = new ArrayList<Kpi>(); while (cursor.hasNext()) { DBObject obj = cursor.next(); BSONObject key = (BSONObject) obj.get("_id"); BSONObject value = (BSONObject) obj.get("value"); Object kpiValue = value.get("count"); Object userName = key.get("username"); Object districtId = key.get("district_id"); Object customerProperty = key.get("customer_property"); Object productStyle = key.get("product_style"); Kpi kpi = new Kpi(); try { kpi.setUserName(userName == null ? "" : userName.toString()); kpi.setKpiCode(key.get("kpi_code").toString()); kpi.setKpiValue(Math.round(Double.parseDouble(kpiValue.toString()))); kpi.setCustomerProperty(customerProperty == null ? "" : customerProperty.toString()); kpi.setDistrictId(districtId == "" ? 0 : Integer.parseInt(districtId.toString())); kpi.setProductStyle(productStyle == null ? "" : productStyle.toString()); kpi.setCreateDate(collName.split("_")[2]); kpi.setUpdateDate(Timestamp.valueOf(CalendarUtil.formatMap.get(PRECISION.HOUR).format(new Date()))); list.add(kpi); } catch (Exception exx) { exx.printStackTrace(); logger.error(String.format("parse type or get value has error,msg is %s", exx.getMessage())); } } MysqlFactory.insert(list); logger.info(String.format("store mysql spent time is %ss", (System.currentTimeMillis() - start) / 1000.0)); } catch (Exception ex) { ex.printStackTrace(); logger.error(String.format("run map-reduce jobs has error,msg is %s", ex.getMessage())); } finally { if (db != null) { db.requestDone(); db = null; } } } }
5.总结
在计算 MongoDB 的MapReduce计算的时候,拼接JavaScript字符串时需要谨慎小心,很容易出错,上面给出的代码只是一部分代码,供参考学习使用;另外,若是要做MapReduce任务计算,推荐使用Hadoop的MapReduce计算框架,MongoDB的MapReduce框架这里仅做介绍学习了解。
6.结束语
这篇博客就和大家分享到这里,若是大家在研究学习的过程当中有什么问题,可以加群进行讨论或发送邮件给我,我会尽我所能为您解答,与君共勉!
邮箱:smartloli.org@gmail.com
QQ群(Hive与AI实战【新群】):935396818
QQ群(Hadoop - 交流社区1):424769183
QQ群(Kafka并不难学):825943084
温馨提示:请大家加群的时候写上加群理由(姓名+公司/学校),方便管理员审核,谢谢!