我正在为Cloudera的分布式日志聚合系统Flume编写一个自定义的装饰器插件。我的Java代码如下:

package multiplex;

import java.io.IOException;
import java.util.ArrayList;
import java.util.List;

import com.cloudera.flume.conf.Context;
import com.cloudera.flume.conf.SinkFactory.SinkDecoBuilder;
import com.cloudera.flume.core.Event;
import com.cloudera.flume.core.EventImpl;
import com.cloudera.flume.core.EventSink;
import com.cloudera.flume.core.EventSinkDecorator;
import com.cloudera.util.Pair;
import com.google.common.base.Preconditions;

public class JsonMultiplexDecorator<S extends EventSink> extends EventSinkDecorator<S> {
  private final String serverName;
  private final String logType;

  public JsonMultiplexDecorator(S s, String serverName, String logType) {
    super(s);

    this.serverName = serverName;
    this.logType = logType;
  }

  @Override
  public void append(Event e) throws IOException {
    String body = new String(e.getBody()).replaceAll("\"", "\\\"");

    String json = "{ \"server\": \"" + this.serverName + "\"," +
      "\"log_type\": \"" + this.logType + "\", " +
      "\"body\": \"" + body + "\" }";

    EventImpl e2 = new EventImpl(json.getBytes(),
        e.getTimestamp(), e.getPriority(), e.getNanos(), e.getHost(),
        e.getAttrs());

    super.append(e2);
  }

  public static SinkDecoBuilder builder() {
    return new SinkDecoBuilder() {
      @Override
      public EventSinkDecorator<EventSink> build(Context context,
          String... argv) {
        Preconditions.checkArgument(argv.length == 2,
            "usage: multiplexDecorator(serverName, logType)");

        return new JsonMultiplexDecorator<EventSink>(null, argv[0], argv[1]);
      }
    };
  }

  public static List<Pair<String, SinkDecoBuilder>> getDecoratorBuilders() {
    List<Pair<String, SinkDecoBuilder>> builders =
      new ArrayList<Pair<String, SinkDecoBuilder>>();

    builders.add(new Pair<String, SinkDecoBuilder>("jsonMultiplexDecorator", builder()));

    return builders;
  }
}

这可以使用ant很好地编译成JAR文件,我可以在运行时将其加载到Flume中并成功配置节点以使用它。但是,当实际上在加载了此插件的节点上发生事件时,我在日志中得到如下错误:
2010-10-19 21:03:15,176 [logicalNode xxxxx] ERROR connector.DirectDriver: Driving src/sink failed! LazyOpenSource | LazyOpenDecorator because null
java.lang.UnsupportedOperationException
    at java.util.Collections$UnmodifiableMap.put(Collections.java:1285)
    at com.cloudera.flume.core.EventBaseImpl.set(EventBaseImpl.java:65)
    at com.cloudera.flume.handlers.rolling.RollSink.append(RollSink.java:164)
    at com.cloudera.flume.agent.diskfailover.DiskFailoverDeco.append(DiskFailoverDeco.java:93)
    at com.cloudera.flume.core.BackOffFailOverSink.append(BackOffFailOverSink.java:144)
    at com.cloudera.flume.agent.AgentSink.append(AgentSink.java:109)
    at com.cloudera.flume.core.EventSinkDecorator.append(EventSinkDecorator.java:58)
    at multiplex.JsonMultiplexDecorator.append(JsonMultiplexDecorator.java:56)
    at com.cloudera.flume.core.EventSinkDecorator.append(EventSinkDecorator.java:58)
    at com.cloudera.flume.handlers.debug.LazyOpenDecorator.append(LazyOpenDecorator.java:69)
    at com.cloudera.flume.core.connector.DirectDriver$PumperThread.run(DirectDriver.java:92)

([logicalNode xxxxx]是EC2内部DNS名称的占位符)。我没有太多的Java经验,所以不确定在这里做错什么还是Flume错误。我应该提到的是,我使用Flume来源的HelloWorld插件示例编写了此代码,并且还使用了一些内置的Flume装饰器进行绘制。

最佳答案

构造EventImpl e2时,您要传递e.getAttrs(),该值不可修改。尝试将e.getAttrs()复制到您自己的地图中;使用new HashMap(e.getAttrs())的浅表副本就足够了。

参考:https://groups.google.com/a/cloudera.org/group/flume-user/browse_thread/thread/046b4a446877c8f9?pli=1

10-04 12:31
查看更多