You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@flink.apache.org by "spuru9 (via GitHub)" <gi...@apache.org> on 2026/09/02 10:14:05 UTC

Re: [PR] [FLINK-40519][network] Deliver priority barrier to a credited remote reader while the subpartition is blocked during recovery [flink]

spuru9 commented on code in PR #29056:
URL: https://github.com/apache/flink/pull/29056#discussion_r3913033200


##########
flink-runtime/src/test/java/org/apache/flink/runtime/io/network/partition/PipelinedSubpartitionTest.java:
##########
@@ -472,6 +473,35 @@ void testConcurrentTimeoutableCheckpointBarrier() throws Exception {
                 .isInstanceOf(ExecutionException.class);
     }
 
+    @TestTemplate
+    void testPriorityBarrierAvailableToCreditedReaderWhileBlocked() throws Exception {
+        PipelinedSubpartition subpartition = createSubpartition();
+        subpartition.setChannelStateWriter(ChannelStateWriter.NO_OP);
+
+        // Block the subpartition, mirroring the RECOVERY_COMPLETION event emitted during recovery.
+        subpartition.add(
+                EventSerializer.toBufferConsumer(EndOfOutputChannelStateEvent.INSTANCE, false));
+        pollBufferAndCheckType(subpartition, Buffer.DataType.RECOVERY_COMPLETION);
+
+        // While blocked and without any priority element, a credited reader sees no data.
+        assertThat(subpartition.getAvailabilityAndBacklog(true).isAvailable()).isFalse();
+        assertThat(subpartition.pollBuffer()).isNull();
+
+        // Enqueue an unaligned checkpoint barrier as a priority element.
+        CheckpointOptions options =
+                CheckpointOptions.unaligned(
+                        CheckpointType.CHECKPOINT, CheckpointStorageLocationReference.getDefault());
+        subpartition.add(
+                EventSerializer.toBufferConsumer(
+                        new CheckpointBarrier(1L, System.currentTimeMillis(), options), true));
+
+        // The credited-reader availability check must now report available so the remote reader is
+        // enqueued and the priority barrier is delivered even though the subpartition
+        // stays blocked.
+        assertThat(subpartition.getAvailabilityAndBacklog(true).isAvailable()).isTrue();
+        pollBufferAndCheckType(subpartition, Buffer.DataType.PRIORITIZED_EVENT_BUFFER);

Review Comment:
   (Optional) nit: We could also assert that after the priority buffer is polled, the subpartition immediately returns to the blocked state (no longer available, pollBuffer() returns null).
   ```suggestion
           pollBufferAndCheckType(subpartition, Buffer.DataType.PRIORITIZED_EVENT_BUFFER);
           
           // After the priority element is consumed, the subpartition should immediately re-block.
           assertThat(subpartition.getAvailabilityAndBacklog(true).isAvailable()).isFalse();
           assertThat(subpartition.pollBuffer()).isNull();
   ```



-- 
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.

To unsubscribe, e-mail: issues-unsubscribe@flink.apache.org

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