Fix streaming ci failure (#12830)

streaming/java/streaming-runtime/src/main/java/io/ray/streaming/runtime/client/JobClientImpl.java

@@ -44,9 +44,12 @@ public class JobClientImpl implements JobClient {
 
       if (submitResult.get()) {
         LOG.info("Finish submitting job: {}.", jobGraph.getJobName());
+      } else {
+        throw new RuntimeException("submitting job failed");
       }
     } catch (Exception e) {
       LOG.error("Failed to submit job: {}.", jobGraph.getJobName(), e);
+      throw new RuntimeException("submitting job failed", e);
     }
   }
 }

streaming/java/streaming-runtime/src/main/java/io/ray/streaming/runtime/master/JobMaster.java

@@ -157,7 +157,8 @@ public class JobMaster {
       scheduler = new JobSchedulerImpl(this);
       scheduler.scheduleJob(graphManager.getExecutionGraph());
     } catch (Exception e) {
-      LOG.error("Failed to submit job.", e);
+      e.printStackTrace();
+      LOG.error("Failed to submit job {}.", e, e);
       return false;
     }
     return true;

streaming/java/streaming-runtime/src/main/java/io/ray/streaming/runtime/master/scheduler/JobSchedulerImpl.java

@@ -1,5 +1,6 @@
 package io.ray.streaming.runtime.master.scheduler;
 
+import com.google.common.base.Preconditions;
 import io.ray.api.ActorHandle;
 import io.ray.streaming.runtime.config.StreamingConfig;
 import io.ray.streaming.runtime.core.graph.executiongraph.ExecutionGraph;
@@ -82,7 +83,7 @@ public class JobSchedulerImpl implements JobScheduler {
     Map<ExecutionVertex, JobWorkerContext> vertexToContextMap = buildWorkersContext(executionGraph);
 
     // init workers
-    initWorkers(vertexToContextMap);
+    Preconditions.checkState(initWorkers(vertexToContextMap));
 
     // init master
     initMaster();
@@ -119,17 +120,13 @@ public class JobSchedulerImpl implements JobScheduler {
    * @param vertexToContextMap vertex - context map
    */
   protected boolean initWorkers(Map<ExecutionVertex, JobWorkerContext> vertexToContextMap) {
-    boolean result;
-    try {
-      result =
-          workerLifecycleController.initWorkers(
-              vertexToContextMap,
-              jobConfig.masterConfig.schedulerConfig.workerInitiationWaitTimeoutMs());
-    } catch (Exception e) {
-      LOG.error("Failed to initiate workers.", e);
-      return false;
+    boolean succeed;
+    int timeoutMs = jobConfig.masterConfig.schedulerConfig.workerInitiationWaitTimeoutMs();
+    succeed = workerLifecycleController.initWorkers(vertexToContextMap, timeoutMs);
+    if (!succeed) {
+      LOG.error("Failed to initiate workers in {} milliseconds", timeoutMs);
     }
-    return result;
+    return succeed;
   }
 
   /** Start JobWorkers according to the physical plan. */

streaming/java/streaming-runtime/src/main/java/io/ray/streaming/runtime/python/PythonGateway.java

@@ -228,7 +228,7 @@ public class PythonGateway {
   public byte[] newInstance(byte[] classNameBytes) {
     String className = (String) serializer.deserialize(classNameBytes);
     try {
-      Class<?> clz = Class.forName(className, true, this.getClass().getClassLoader());
+      Class<?> clz = Class.forName(className, true, Thread.currentThread().getContextClassLoader());
       Object instance = clz.newInstance();
       referenceMap.put(getReferenceId(instance), instance);
       return serializer.serialize(getReferenceId(instance));

streaming/java/streaming-runtime/src/main/java/io/ray/streaming/runtime/transfer/DataReader.java

@@ -246,8 +246,8 @@ public class DataReader {
 
     // kMessageBundleHeaderSize + kUniqueIDSize:
     // magicNum(4b) + bundleTs(8b) + lastMessageId(8b) + messageListSize(4b)
-    // + bundleType(4b) + rawBundleSize(4b) + channelID(20b)
-    static final int LENGTH = 4 + 8 + 8 + 4 + 4 + 4 + 20;
+    // + bundleType(4b) + rawBundleSize(4b) + channelID
+    static final int LENGTH = 4 + 8 + 8 + 4 + 4 + 4 + ChannelId.ID_LENGTH;
     private int magicNum;
     private long bundleTs;
     private long lastMessageId;

streaming/java/streaming-runtime/src/main/java/io/ray/streaming/runtime/transfer/channel/ChannelId.java

@@ -5,6 +5,7 @@ import com.google.common.base.FinalizableReferenceQueue;
 import com.google.common.base.Preconditions;
 import com.google.common.collect.Sets;
 import com.google.common.io.BaseEncoding;
+import io.ray.api.id.ObjectId;
 import java.lang.ref.Reference;
 import java.nio.ByteBuffer;
 import java.util.Random;
@@ -16,7 +17,7 @@ import sun.nio.ch.DirectBuffer;
  */
 public class ChannelId {
 
-  public static final int ID_LENGTH = 20;
+  public static final int ID_LENGTH = ObjectId.LENGTH;
   private static final FinalizableReferenceQueue REFERENCE_QUEUE = new FinalizableReferenceQueue();
   // This ensures that the FinalizablePhantomReference itself is not garbage-collected.
   private static final Set<Reference<?>> references = Sets.newConcurrentHashSet();
@@ -82,15 +83,15 @@ public class ChannelId {
   }
 
   /**
-   * Generate channel name, which will be 20 character
+   * Generate channel name, which will be {@link ChannelId#ID_LENGTH} character
    *
    * @param fromTaskId upstream task id
    * @param toTaskId downstream task id Returns channel name
    */
   public static String genIdStr(int fromTaskId, int toTaskId, long ts) {
     /*
-      |    Head    | Timestamp | Empty | From  |  To    |
-      | 8 bytes    |  4bytes   | 4bytes| 2bytes| 2bytes |
+      |    Head    | Timestamp | Empty | From  |  To    | padding |
+      | 8 bytes    |  4bytes   | 4bytes| 2bytes| 2bytes |         |
     */
     Preconditions.checkArgument(
         fromTaskId < Short.MAX_VALUE,
@@ -99,7 +100,7 @@ public class ChannelId {
         Short.MAX_VALUE);
     Preconditions.checkArgument(
         toTaskId < Short.MAX_VALUE, "toTaskId %s is larger than %s", fromTaskId, Short.MAX_VALUE);
-    byte[] channelName = new byte[20];
+    byte[] channelName = new byte[ID_LENGTH];
 
     for (int i = 11; i >= 8; i--) {
       channelName[i] = (byte) (ts & 0xff);