MongoDB 的Replication是通过一个日志来存储写操作的,这个日志就叫做oplog。 在默认情况下,oplog分配的是5%的空闲磁盘空间。通常而言,这是一种合理的设置。可以通过mongod --oplogSize来改变oplog的日志大小。 oplog是capped collection,因为oplog的特点(不能太多把磁盘填满了,固定大小)需要,MongoDB才发明了capped collection(the oplog is actually the reason capped collections were invented).
oplog的位置 oplog在local库: master/slave 架构下 local.oplog.$main; replica sets 架构下: local.oplog.rs sharding 架构下,mongos下不能查看oplog,可到每一片去看。
mongos> use local switched to db local mongos> show collections Thu Mar 28 11:37:11 uncaught exception: error: { "$err" : "can't use 'local' database through mongos", "code" : 13644 }
oplog的格式 MongoDB 2.0版本
PRIMARY> db.oplog.rs.findOne() { "ts" : { "t" : 1354919611000, "i" : 196 }, "h" : NumberLong("-8946637877024029255"), "op" : "i", "ns" : "msg.msgToSend", "o" : { "_id" : ObjectId("50c26ecae7d64ae0b5f36cfe"), ... } }
MongoDB 2.2版本
PRIMARY> db.oplog.rs.findOne() { "ts" : Timestamp(1364362801000, 8247), "h" : NumberLong("8229173295225699173"), "v" : 2, "op" : "i", "ns" : "goods.Simigoods", "fromMigrate" : true, "o" : { "_id" : ObjectId("50b534310eba2018b88ba3b2"), ... } }
可以看到有个字段"fromMigrate" : true,之前以为是从2.0升级过来的,后查看源码发现并发如此,fromMigrate指的是chunk是迁移过来的,分片里的块移动,详见src/mongo/s/d_migrate.cpp, v表示OPLOG_VERSION,oplog版本。 新搭建的结构形如:
PRIMARY> db.version() 2.2.2 PRIMARY> db.oplog.rs.findOne() { "ts" : Timestamp(1364186197000, 58), "h" : NumberLong("-7878220425718087654"), "v" : 2, "op" : "u", "ns" : "exaitem_gmsbatchtask.jdgmsbatchtask", "o2" : { "_id" : "83f09a98-6a41-497b-a988-99ba5399d296" }, "o" : { "_id" : "83f09a98-6a41-497b-a988-99ba5399d296", "status" : 2, "content" : "", "type" : 17, "business" : "832722", "optype" : 2, "addDate" : ISODate("2013-03-25T04:36:38.511Z"), "modifyDate" : ISODate("2013-03-25T04:36:39.131Z"), "source" : 5 } }
MongoDB 2.4版本
{ "ts" : { "t" : 1361948104000, "i" : 325 }, "h" : NumberLong("-8795977166222676062"), "v" : 2, "op" : "i", "ns" : "test.log", "o" : { "_id" : ObjectId("51031ca0c86617a8811be893"), ... } }
格式大同小异,2.4版本又改回去了。ts格式2.2版本中是Timestamp(1364186197000, 58)形式,MongoDB2.0版本及MongoDB2.4版本是{ "t" : 1361948104000, "i" : 325 }形式,另外若用MongoDB2.4版本的客户端(mongo)查看2.2版本的,看到的是MongoDB2.4版本的格式,这个只与mongo版本有关。 oplog相关字段含义 ts: the time this operation occurred. h: a unique ID for this operation. Each operation will have a different value in this field. op: the write operation that should be applied to the slave. n indicates a no-op, this is just an informational message. ns: the database and collection affected by this operation. Since this is a no-op, this field is left blank. o: the actual document representing the op. Since this is a no-op, this field is pretty useless. The o field now contains the document to insert or the criteria to update and remove. Notice that, for the update, there are two o fields (o and o2). o2 give the update criteria and o gives the modifications (equivalent to update()‘s second argument). ts:8字节的时间戳,由4字节unix timestamp + 4字节自增计数表示。 这个值很重要,在选举(如master宕机时)新primary时,会选择ts最大的那个secondary作为新primary。 op:1字节的操作类型,例如i表示insert,d表示delete。 ns:操作所在的namespace。 o:操作所对应的document,即当前操作的内容(比如更新操作时要更新的的字段和值) o2: 在执行更新操作时的条件,仅限于update时才有该属性。 其中op,可以是如下几种情形之一: "i": insert "u": update "d": delete "c": db cmd "db":声明当前数据库 (其中ns 被设置成为=>数据库名称+ '.') "n": no op,即空操作,其会定期执行以确保时效性 。 20130719更新:今天发现修改配置,会产生 "n" 操作
{ "ts" : Timestamp(1372320938000, 1), "h" : NumberLong("2050563086860406946"), "v" : 2, "op" : "n", "ns" : "", "o" : { "msg" : "Reconfig set", "version" : 6 } } { "ts" : Timestamp(1372319914000, 1), "h" : NumberLong("5828735007195954091"), "v" : 2, "op" : "n", "ns" : "", "o" : { "msg" : "Reconfig set", "version" : 5 } } { "ts" : Timestamp(1372318223000, 1), "h" : NumberLong("512600544405470974"), "v" : 2, "op" : "n", "ns" : "", "o" : { "msg" : "Reconfig set", "version" : 4 } }
除了以上这些,还有两个bool型的字段,一个是上面提到的fromMigrate,另一是字段b,仔细看oplog我们发现有"b":true的文档,是在delete和update操作时的bool值(update一个或多个)。 举例:
{ "ts" : { "t" : 1354923335000, "i" : 2 }, "h" : NumberLong("563747339476084113"), "op" : "u", "ns" : "msg.device", "o2" : { "_id" : ObjectId("509fa1207386d978864c7833") }, "o" : { "$set" : { "flag" : "1", "pin" : "5126d5b23c303", "device" : "ceb27de6b9dd8f045130f046a7662630", "modified" : ISODate("2012-12-07T23:36:24.628Z") } } }
了解了oplog的详细结构,我们就可以根据原理写个程序,来达到同步数据的目的,详见mongosync。