MongoDB 从 3.6 开始为开发者提供了 Change Streams 功能,利用 Change Streams 功能可以非常方便地监听指定 Collection 上的数据变化。例如在 mongo shell 中,我们可以通过如下方式监听 shopping 数据库 order 表上的变化:
watchCursor = db.getSiblingDB("shopping").order.watch()
while (!watchCursor.isExhausted()){
if (watchCursor.hasNext()){
printjson(watchCursor.next());
}
}
利用 Play Mongo 可以方便地实现数据监听功能,并且我们可以将 Change Stream 转换成 Akka Stream,然后以流的方式处理指定 Collection 上的数据变化,
mongo
.collection[Order]
.watch()
.fullDocument
.toSource
.groupedWithin(10, 1000.millis)
.throttle(elements = 1, per = 1.second, maximumBurst = 1, ThrottleMode.shaping)
.runForeach{ seq =>
// ...
}
上面的代码实现了以下几个功能:
上面的代码并没有考虑可用性,如果在监听过程中发生了网络错误,如何从错误中恢复呢? 上面的实现代码底层是基于官方的 mongo-java-driver 实现的,关于可用性官方文档有如下描述:
Change streams provide a way to watch changes to documents in a collection. To improve the usability of this new stage, the MongoCollection API includes a new watch method. The ChangeStreamIterable sets up the change stream and automatically attempts to resume if it encounters a potentially recoverable error.
文档中提及程序可以自动从可恢复的错误中恢复。经测试验证,如果网络中断在 30 秒以内均属于可恢复错误;但是如果大于 30 秒,则会报连接超时错误并且无法从错误中自动恢复:
com.mongodb.MongoTimeoutException: Timed out after 30000 ms while waiting for a server that matches ReadPreferenceServerSelector{readPreference=primary}. Client view of cluster state is {type=UNKNOWN, servers=[{address=127.0.0.1:27117, type=UNKNOWN, state=CONNECTING, exception={com.mongodb.MongoSocketOpenException: Exception opening socket}, caused by {java.net.ConnectException: Connection refused}}]
at com.mongodb.internal.connection.BaseCluster.createTimeoutException(BaseCluster.java:401)
at com.mongodb.internal.connection.BaseCluster.handleServerSelectionRequest(BaseCluster.java:309)
at com.mongodb.internal.connection.BaseCluster.access$800(BaseCluster.java:65)
at com.mongodb.internal.connection.BaseCluster$WaitQueueHandler.run(BaseCluster.java:482)
at java.lang.Thread.run(Thread.java:748)
幸运的是,Akka Stream 的 RestartSource 可以帮我们解决这种不可恢复错误,解决方式就是通过指数规避(exponential back-off)方式不断重试。下面是一个通用的创建 RestartSource 的方法实现:
def restartSource(colName: String): Source[ChangeStreamDocument[JsObject], _] = {
RestartSource.withBackoff(
minBackoff = 3.seconds,
maxBackoff = 10.seconds,
randomFactor = 0.2,
maxRestarts = 1000000
) { () ⇒
Logger.warn(s"Creating source for watching ${colName}.")
mongo.collection(colName).watch().fullDocument.toSource
}
}
通过 Backoff 参数可以指定重试策略:
当发生错误时,RestartSource 会尝试重新创建一个 Source:
Logger.warn(s"Creating source for watching ${colName}.")
mongo.collection(colName).watch().fullDocument.toSource
完整代码如下:
val colName = "common-user"
restartSource(colName)
.groupedWithin(10, 1000.millis)
.throttle(elements = 1, per = 1.second, maximumBurst = 1, ThrottleMode.shaping)
.runForeach{ seq =>
try {
Logger.info(seq.toString())
} catch { case t: Throwable =>
Logger.error(s"Watch change stream of ${colName} error: ${t.getMessage}", t)
}
}
需要注意的是 runForeach 中需要显式捕获异常并处理,否则会导致 Source 结束并退出。