[GitHub] [samza] mynameborat commented on a change in pull request #1385: SAMZA-2550: Move side input processing to use RunLoop

[GitBox <gi...@apache.org>]

mynameborat commented on a change in pull request #1385:
URL: https://github.com/apache/samza/pull/1385#discussion_r464518666



##########
File path: samza-core/src/main/scala/org/apache/samza/storage/ContainerStorageManager.java
##########
@@ -758,41 +745,50 @@ public void run() {
       sideInputSystemConsumers.register(ssp, startingOffset);
       taskInstanceMetrics.get(this.sspSideInputHandlers.get(ssp).getTaskName()).addOffsetGauge(
           ssp, ScalaJavaUtil.toScalaFunction(() -> this.sspSideInputHandlers.get(ssp).getLastProcessedOffset(ssp)));
+      sideInputTaskMetrics.get(this.sspSideInputHandlers.get(ssp).getTaskName()).addOffsetGauge(
+          ssp, ScalaJavaUtil.toScalaFunction(() -> this.sspSideInputHandlers.get(ssp).getLastProcessedOffset(ssp)));
+    }
 
-      SystemStreamMetadata systemStreamMetadata = streamMetadataCache.getSystemStreamMetadata(ssp.getSystemStream(), false);
-      SystemStreamMetadata.SystemStreamPartitionMetadata sspMetadata =
-          (systemStreamMetadata == null) ? null : systemStreamMetadata.getSystemStreamPartitionMetadata().get(ssp.getPartition());
+    Map<TaskName, TaskSideInputHandler> taskSideInputHandlers = this.sspSideInputHandlers.values().stream()
+        .distinct()
+        .collect(Collectors.toMap(TaskSideInputHandler::getTaskName, Function.identity()));
 
-      // record a copy of the sspMetadata, to later check if its caught up
-      initialSideInputSSPMetadata.put(ssp, sspMetadata);
+    Map<TaskName, RunLoopTask> sideInputTasks = new HashMap<>();
+    this.taskSideInputStoreSSPs.forEach((taskName, storesToSSPs) -> {
+        Set<SystemStreamPartition> taskSSPs = this.taskSideInputStoreSSPs.get(taskName).values().stream()
+            .flatMap(Set::stream)
+            .collect(Collectors.toSet());
 
-      // check if the ssp is caught to upcoming, even at start
-      checkSideInputCaughtUp(ssp, startingOffset, SystemStreamMetadata.OffsetType.UPCOMING, false);
-    }
+        RunLoopTask sideInputTask = new SideInputTask(taskName, taskSSPs, taskSideInputHandlers.get(taskName), sideInputTaskMetrics.get(taskName));
+
+        sideInputTasks.put(taskName, sideInputTask);
+      });
 
     // start the systemConsumers for consuming input
     this.sideInputSystemConsumers.start();
 
+    TaskConfig taskConfig = new TaskConfig(this.config);
+    SamzaContainerMetrics sideInputContainerMetrics =
+        new SamzaContainerMetrics(SIDEINPUTS_METRICS_PREFIX + this.samzaContainerMetrics.source(),
+            this.samzaContainerMetrics.registry());
+
+    this.sideInputRunLoop = new RunLoop(sideInputTasks,
+        null, // all operations are executed in the main runloop thread
+        this.sideInputSystemConsumers,
+        1, // single message in flight per task
+        -1, // no windowing
+        taskConfig.getCommitMs(),
+        taskConfig.getCallbackTimeoutMs(),
+        this.config.getLong("container.disk.quota.delay.max.ms", TimeUnit.SECONDS.toMillis(1)),

Review comment:
       I am fine if we can create a follow up ticket and knock them out all at once. Considered resolved once you have a follow up TODO created.




----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

For queries about this service, please contact Infrastructure at:
users@infra.apache.org