Question

使用Akka Streams时似乎永远无法正确处理错误。

所以这是我的代码

var db = Database.forConfig("oracle")
var mysqlDb = Database.forConfig("mysql_read")
var mysqlDbWrite = Database.forConfig("mysql_write")

implicit val actorSystem = ActorSystem()
val decider : Supervision.Decider = {
  case _: Exception =>
      println("got an exception restarting connections")
     // let us restart our connections
     db.close()
     mysqlDb.close()
     mysqlDbWrite.close()
     db = Database.forConfig("oracle")
     mysqlDb = Database.forConfig("mysql_read")
     mysqlDbWrite = Database.forConfig("mysql_write")
     Supervision.Restart
}
implicit val materializer = ActorMaterializer(ActorMaterializerSettings(actorSystem).withSupervisionStrategy(decider))

我有这样的流程

val alreadyExistsFilter : Flow[Foo, Foo, NotUsed] = Flow[Foo].mapAsync(10){ foo =>
  try {
     val existsQuery = sql"""SELECT id FROM foo WHERE id = ${foo.id}""".as[Long]
     mysqlDbWrite.run(existsQuery).map(v => (foo, v))
  } catch {
     case e: Throwable =>
        println(s"Lookup failed for ${foo}")
        throw e // will restart the stream
  }
}.collect {case (f, v) if v.isEmpty => f}

所以基本上如果foo已经存在于MySQL中，则该流不应再对该记录进行处理。

我对这段代码的希望是，如果mysql查找失败（mysql机器非常糟糕且超时很常见），记录将被打印并丢弃，并且流将继续保留其余记录，由监督提供

当我运行此代码时。我看到像

这样的错误

[error] (mysql_write network timeout executor) java.lang.RuntimeException: java.sql.SQLException: Invalid socket timeout value or state
java.lang.RuntimeException: java.sql.SQLException: Invalid socket timeout value or state
    at com.mysql.jdbc.ConnectionImpl$12.run(ConnectionImpl.java:5576)
    at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1142)
    at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:617)
    at java.lang.Thread.run(Thread.java:745)
Caused by: java.sql.SQLException: Invalid socket timeout value or state
    at com.mysql.jdbc.SQLError.createSQLException(SQLError.java:998)
    at com.mysql.jdbc.SQLError.createSQLException(SQLError.java:937)
    at com.mysql.jdbc.SQLError.createSQLException(SQLError.java:926)
    at com.mysql.jdbc.SQLError.createSQLException(SQLError.java:872)
    at com.mysql.jdbc.MysqlIO.setSocketTimeout(MysqlIO.java:4852)
    at com.mysql.jdbc.ConnectionImpl$12.run(ConnectionImpl.java:5574)
    at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1142)
    at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:617)
    at java.lang.Thread.run(Thread.java:745)
Caused by: java.net.SocketException: Socket is closed
    at java.net.Socket.setSoTimeout(Socket.java:1137)
    at com.mysql.jdbc.MysqlIO.setSocketTimeout(MysqlIO.java:4850)
    at com.mysql.jdbc.ConnectionImpl$12.run(ConnectionImpl.java:5574)
    at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1142)
    at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:617)
    at java.lang.Thread.run(Thread.java:745)

和

[error] (mysql_write network timeout executor) java.lang.NullPointerException
java.lang.NullPointerException
    at com.mysql.jdbc.MysqlIO.setSocketTimeout(MysqlIO.java:4850)
    at com.mysql.jdbc.ConnectionImpl$12.run(ConnectionImpl.java:5574)
    at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1142)
    at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:617)
    at java.lang.Thread.run(Thread.java:745)

让我感到惊讶的一件事是，这些例外并非来自我的捕获区。因为我没有看到我的catch块的println语句。堆栈跟踪并没有显示它来自哪里......但是因为它说mysql_write我可以假设它是上面的Flow，因为只有这个Flow使用mysql_write。

最后整个流崩溃并出现错误

[trace] Stack trace suppressed: run last compile:runMain for the full output.
flow has failed with error Shutting down because of violation of the Reactive Streams specification.
14:51:06,973 |-INFO in ch.qos.logback.classic.AsyncAppender[asyncKafkaAppender] - Worker thread will flush remaining events before exiting.
[success] Total time: 3480 s, completed Sep 26, 2017 2:51:07 PM
14:51:07,603 |-INFO in ch.qos.logback.core.hook.DelayingShutdownHook@2320545b - Sleeping for 1 seconds

我不知道我做了什么来违反反应流规范!!

Answer 1

获得更可预测的解决方案的第一步是删除阻止行为（{ "userId": "1", "questions": { "questionId": "111", "type": "optional" } }）并使用Await.result。重写mapAsync流程可以是：

alreadyExistsFilter

有关Akka阻止的更多信息，请参阅docs。

Answer 2

斯特凡诺给出的答案是正确的。由于阻止了流程中的代码，错误确实会发生。

虽然，我的初始程序是针对scala 2.11运行的，甚至在切换到mapAsync之后，问题仍然存在。

由于这是一个命令行工具，我很容易切换到scala 2.12并再试一次。

当我尝试使用Scala 2.12时，它运行得非常好。

对我有很大帮助的一件事是在依赖项中有"ch.qos.logback" % "logback-classic" % "1.2.3",。这将向您显示正在执行的每个SQL语句，并轻松查看是否出现问题。

由于违反Reactive Streams规范而导致错误关闭，因此流失败

2 个答案: