Flink - Fix UPSERT delete file metadata [Backport to Flink 1.12]

[kbendick]

This backports #4364 to Flink 1.12.

See also #4417 for Flink 1.13.

Co-authored-by: liliwei hililiwei@gmail.com

[kbendick]

It seems that the new tests aren't passing.

I will scan tomorrow to see if I missed anything when backporting. And check if anything wasn't backported previously.

This also opens the wider discussion of how much longer we'll actively support Flink 1.12 once Flink 1.15 is released. There have been several major API changes to areas that directly impact our work. Once supporting Flink 1.12 becomes a high burden, we'll have to really consider how important it is for us and our users.

[rdblue]

@kbendick, should we replace this with a PR that disables UPSERT in 1.12?

[kbendick]

Opened a PR to simply throw UnsupportedOperationException in FlinkSink if it's determined that the program is running in upsert mode: #4519

[kbendick]

Closing in favor of #4519

[kbendick]

I updated this PR to use DATE instead of TO_DATE, as we discovered in #4532 that TO_DATE was causing tests to process rows out of order during INSERT VALUES.

However, the tests all still fail in 1.12.

I think that So I'm still not able to get 1.12 to correctly process upserts. I think we should disable upsert mode in Flink 1.12 in the next patch with a recommendation to upgrade to 1.13 as proposed here: #4519

> Task :iceberg-flink:iceberg-flink-1.12:test
org.apache.iceberg.flink.TestFlinkUpsert > testPrimaryKeyEqualToPartitionKey[catalogName=testhadoop, baseNamespace=default, format=PARQUET, isStreaming=true] FAILED
java.lang.AssertionError: expected:<[2,aaa, 3,bbb]> but was:<[1,aaa, 3,bbb]>
at org.junit.Assert.fail(Assert.java:89)
at org.junit.Assert.failNotEquals(Assert.java:835)
at org.junit.Assert.assertEquals(Assert.java:120)
at org.junit.Assert.assertEquals(Assert.java:146)
at org.apache.iceberg.flink.TestHelpers.assertRows(TestHelpers.java:140)
at org.apache.iceberg.flink.TestFlinkUpsert.testPrimaryKeyEqualToPartitionKey(TestFlinkUpsert.java:178)
org.apache.iceberg.flink.TestFlinkUpsert > testPrimaryKeyFieldsAtBeginningOfSchema[catalogName=testhadoop, baseNamespace=default, format=PARQUET, isStreaming=true] FAILED
java.lang.AssertionError: expected:<[aaa,2022-03-01,2, bbb,2022-03-01,3]> but was:<[aaa,2022-03-01,1, bbb,2022-03-01,3]>
at org.junit.Assert.fail(Assert.java:89)
at org.junit.Assert.failNotEquals(Assert.java:835)
at org.junit.Assert.assertEquals(Assert.java:120)
at org.junit.Assert.assertEquals(Assert.java:146)
at org.apache.iceberg.flink.TestHelpers.assertRows(TestHelpers.java:140)
at org.apache.iceberg.flink.TestFlinkUpsert.testPrimaryKeyFieldsAtBeginningOfSchema(TestFlinkUpsert.java:219)
org.apache.iceberg.flink.TestFlinkUpsert > testPrimaryKeyFieldsAtEndOfTableSchema[catalogName=testhadoop, baseNamespace=default, format=PARQUET, isStreaming=false] FAILED
java.lang.AssertionError: expected:<[2,aaa,2022-03-01, 3,bbb,2022-03-01]> but was:<[1,aaa,2022-03-01, 3,bbb,2022-03-01]>
at org.junit.Assert.fail(Assert.java:89)
at org.junit.Assert.failNotEquals(Assert.java:835)
at org.junit.Assert.assertEquals(Assert.java:120)
at org.junit.Assert.assertEquals(Assert.java:146)
at org.apache.iceberg.flink.TestHelpers.assertRows(TestHelpers.java:140)
at org.apache.iceberg.flink.TestFlinkUpsert.testPrimaryKeyFieldsAtEndOfTableSchema(TestFlinkUpsert.java:262)
org.apache.iceberg.flink.TestFlinkUpsert > testPrimaryKeyEqualToPartitionKey[catalogName=testhadoop, baseNamespace=default, format=PARQUET, isStreaming=false] FAILED
java.lang.AssertionError: expected:<[2,aaa, 3,bbb]> but was:<[1,aaa, 3,bbb]>
at org.junit.Assert.fail(Assert.java:89)
at org.junit.Assert.failNotEquals(Assert.java:835)
at org.junit.Assert.assertEquals(Assert.java:120)
at org.junit.Assert.assertEquals(Assert.java:146)
at org.apache.iceberg.flink.TestHelpers.assertRows(TestHelpers.java:140)
at org.apache.iceberg.flink.TestFlinkUpsert.testPrimaryKeyEqualToPartitionKey(TestFlinkUpsert.java:178)
org.apache.iceberg.flink.TestFlinkUpsert > testPrimaryKeyFieldsAtEndOfTableSchema[catalogName=testhadoop, baseNamespace=default, format=AVRO, isStreaming=true] FAILED
java.lang.AssertionError: expected:<[2,aaa,2022-03-01, 3,bbb,2022-03-01]> but was:<[1,aaa,2022-03-01, 3,bbb,2022-03-01]>
at org.junit.Assert.fail(Assert.java:89)
at org.junit.Assert.failNotEquals(Assert.java:835)
at org.junit.Assert.assertEquals(Assert.java:120)
at org.junit.Assert.assertEquals(Assert.java:146)
at org.apache.iceberg.flink.TestHelpers.assertRows(TestHelpers.java:140)
at org.apache.iceberg.flink.TestFlinkUpsert.testPrimaryKeyFieldsAtEndOfTableSchema(TestFlinkUpsert.java:262)
org.apache.iceberg.flink.TestFlinkUpsert > testPrimaryKeyEqualToPartitionKey[catalogName=testhadoop, baseNamespace=default, format=AVRO, isStreaming=true] FAILED
java.lang.AssertionError: expected:<[2,aaa, 3,bbb]> but was:<[1,aaa, 3,bbb]>
at org.junit.Assert.fail(Assert.java:89)
at org.junit.Assert.failNotEquals(Assert.java:835)
at org.junit.Assert.assertEquals(Assert.java:120)
at org.junit.Assert.assertEquals(Assert.java:146)
at org.apache.iceberg.flink.TestHelpers.assertRows(TestHelpers.java:140)
at org.apache.iceberg.flink.TestFlinkUpsert.testPrimaryKeyEqualToPartitionKey(TestFlinkUpsert.java:178)
org.apache.iceberg.flink.TestFlinkUpsert > testPrimaryKeyFieldsAtEndOfTableSchema[catalogName=testhadoop, baseNamespace=default, format=AVRO, isStreaming=false] FAILED
java.lang.AssertionError: expected:<[2,aaa,2022-03-01, 3,bbb,2022-03-01]> but was:<[1,aaa,2022-03-01, 3,bbb,2022-03-01]>
at org.junit.Assert.fail(Assert.java:89)
at org.junit.Assert.failNotEquals(Assert.java:835)
at org.junit.Assert.assertEquals(Assert.java:120)
at org.junit.Assert.assertEquals(Assert.java:146)
at org.apache.iceberg.flink.TestHelpers.assertRows(TestHelpers.java:140)
at org.apache.iceberg.flink.TestFlinkUpsert.testPrimaryKeyFieldsAtEndOfTableSchema(TestFlinkUpsert.java:262)
org.apache.iceberg.flink.TestFlinkUpsert > testPrimaryKeyEqualToPartitionKey[catalogName=testhadoop, baseNamespace=default, format=AVRO, isStreaming=false] FAILED
java.lang.AssertionError: expected:<[2,aaa, 3,bbb]> but was:<[1,aaa, 3,bbb]>
at org.junit.Assert.fail(Assert.java:89)
at org.junit.Assert.failNotEquals(Assert.java:835)
at org.junit.Assert.assertEquals(Assert.java:120)
at org.junit.Assert.assertEquals(Assert.java:146)
at org.apache.iceberg.flink.TestHelpers.assertRows(TestHelpers.java:140)
at org.apache.iceberg.flink.TestFlinkUpsert.testPrimaryKeyEqualToPartitionKey(TestFlinkUpsert.java:178)
org.apache.iceberg.flink.TestFlinkUpsert > testPrimaryKeyEqualToPartitionKey[catalogName=testhadoop, baseNamespace=default, format=ORC, isStreaming=false] FAILED
java.lang.AssertionError: expected:<[2,aaa, 3,bbb]> but was:<[1,aaa, 3,bbb]>
at org.junit.Assert.fail(Assert.java:89)
at org.junit.Assert.failNotEquals(Assert.java:835)
at org.junit.Assert.assertEquals(Assert.java:120)
at org.junit.Assert.assertEquals(Assert.java:146)
at org.apache.iceberg.flink.TestHelpers.assertRows(TestHelpers.java:140)
at org.apache.iceberg.flink.TestFlinkUpsert.testPrimaryKeyEqualToPartitionKey(TestFlinkUpsert.java:178)

[kbendick]

Also, Github is saying that all checks are passing as Flink 1.12 has been removed in master. Checking out this branch and running it locally shows that the issue is still present.

[openinx]

If this failure can be reproduced locally, then we can just open the intellij idea debugger to analysis the written columnar files to see what's wrong there.

[rdblue]

@openinx have you had a chance to look into the failure? It would be great to fix this instead of disabling in 1.12.

[github-actions]

This pull request has been marked as stale due to 30 days of inactivity. It will be closed in 1 week if no further activity occurs. If you think that’s incorrect or this pull request requires a review, please simply write any comment. If closed, you can revive the PR at any time and @mention a reviewer or discuss it on the dev@iceberg.apache.org list. Thank you for your contributions.

[github-actions]

This pull request has been closed due to lack of activity. This is not a judgement on the merit of the PR in any way. It is just a way of keeping the PR queue manageable. If you think that is incorrect, or the pull request requires review, you can revive the PR at any time.