Incorrect handling of consecutive faulty items in chunk scanning

[fmbenhassine]

As of v5.0.1, the FaultTolerantChunkProcessor is unable to skip two consecutive faulty items when scanning chunks.

Here is a failing test (currently disabled in FaultTolerantChunkProcessorTests):

@Test
void testWriteRetryOnTwoExceptions() throws Exception {
	SimpleRetryPolicy retryPolicy = new SimpleRetryPolicy();
	retryPolicy.setMaxAttempts(2);
	batchRetryTemplate.setRetryPolicy(retryPolicy);
	processor.setWriteSkipPolicy(new AlwaysSkipItemSkipPolicy());
	processor.setItemWriter(new ItemWriter<String>() {
		@Override
		public void write(Chunk<? extends String> chunk) throws Exception {
			if (chunk.getItems().contains("fail")) {
				throw new IllegalArgumentException("Expected Exception!");
			}
		}
	});
	Chunk<String> inputs = new Chunk<>(Arrays.asList("3", "fail", "fail", "4"));
	Exception exception = assertThrows(RuntimeException.class, () -> processor.process(contribution, inputs));
	assertEquals("Expected Exception!", exception.getMessage());
	// first retry
	exception = assertThrows(RuntimeException.class, () -> processor.process(contribution, inputs));
	assertEquals("Expected Exception!", exception.getMessage());
	// retry exhausted, now scanning
	processor.process(contribution, inputs);
	// skip on this attempt
	exception = assertThrows(RuntimeException.class, () -> processor.process(contribution, inputs));
	assertEquals("Expected Exception!", exception.getMessage());
	// 2nd exception detected
	exception = assertThrows(RuntimeException.class, () -> processor.process(contribution, inputs));
	assertEquals("Expected Exception!", exception.getMessage());
	// still scanning
	processor.process(contribution, inputs);
	assertEquals(2, contribution.getSkipCount());
	assertEquals(2, contribution.getWriteCount());
	assertEquals(0, contribution.getFilterCount());
}

This test started failing after resolving #4314.

[mariusbreivik]

@fmbenhassine what's the status on this issue?

[saugion]

+1, I'm stuck with the same issue. Do you know any workaround? @fmbenhassine

[ozlemkaradeniz]

Hi, I am stuck with this problem, what is the current status of the issue? @fmbenhassine
Thanks

[fmbenhassine]

The status of this issue is in standby. It will be fixed in the new implementation of the chunk-oriented processing model, see #3950.

#3950 is currently released as experimental, and should land as beta in the core framework in v6.