Flink: Add retry limit for IcebergSource continuous split planning errors #7571
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
hililiwei
reviewed
May 10, 2023
flink/v1.17/flink/src/main/java/org/apache/iceberg/flink/source/IcebergSource.java
Outdated
Show resolved
Hide resolved
hililiwei
reviewed
May 10, 2023
flink/v1.17/flink/src/main/java/org/apache/iceberg/flink/source/IcebergSource.java
Outdated
Show resolved
Hide resolved
stevenzwu
reviewed
May 11, 2023
flink/v1.17/flink/src/main/java/org/apache/iceberg/flink/FlinkReadOptions.java
Outdated
Show resolved
Hide resolved
...nk/src/main/java/org/apache/iceberg/flink/source/enumerator/ContinuousIcebergEnumerator.java
Outdated
Show resolved
Hide resolved
914a2b5 to
4ac1f6d
Compare
stevenzwu
reviewed
May 11, 2023
...nk/src/main/java/org/apache/iceberg/flink/source/enumerator/ContinuousIcebergEnumerator.java
Outdated
Show resolved
Hide resolved
...nk/src/main/java/org/apache/iceberg/flink/source/enumerator/ContinuousIcebergEnumerator.java
Outdated
Show resolved
Hide resolved
stevenzwu
reviewed
May 13, 2023
...nk/src/main/java/org/apache/iceberg/flink/source/enumerator/ContinuousIcebergEnumerator.java
Outdated
Show resolved
Hide resolved
...nk/src/main/java/org/apache/iceberg/flink/source/enumerator/ContinuousIcebergEnumerator.java
Outdated
Show resolved
Hide resolved
...rc/test/java/org/apache/iceberg/flink/source/enumerator/TestContinuousIcebergEnumerator.java
Outdated
Show resolved
Hide resolved
...rc/test/java/org/apache/iceberg/flink/source/enumerator/TestContinuousIcebergEnumerator.java
Show resolved
Hide resolved
...rc/test/java/org/apache/iceberg/flink/source/enumerator/TestContinuousIcebergEnumerator.java
Outdated
Show resolved
Hide resolved
...rc/test/java/org/apache/iceberg/flink/source/enumerator/TestContinuousIcebergEnumerator.java
Outdated
Show resolved
Hide resolved
...rc/test/java/org/apache/iceberg/flink/source/enumerator/TestContinuousIcebergEnumerator.java
Outdated
Show resolved
Hide resolved
...rc/test/java/org/apache/iceberg/flink/source/enumerator/TestContinuousIcebergEnumerator.java
Show resolved
Hide resolved
...rc/test/java/org/apache/iceberg/flink/source/enumerator/TestContinuousIcebergEnumerator.java
Outdated
Show resolved
Hide resolved
...rc/test/java/org/apache/iceberg/flink/source/enumerator/TestContinuousIcebergEnumerator.java
Outdated
Show resolved
Hide resolved
stevenzwu
reviewed
May 15, 2023
Contributor
stevenzwu
left a comment
There was a problem hiding this comment.
this PR is very close now. left a few more nit comments.
| LOG.error("Failed to discover new splits", error); | ||
| consecutiveFailures++; | ||
| if (scanContext.maxAllowedPlanningFailures() < 0 | ||
| || consecutiveFailures < scanContext.maxAllowedPlanningFailures()) { |
Contributor
There was a problem hiding this comment.
nit: should be <=. if maxAllowedPlanningFailures is 3, we should allow 3 consecutive failures.
Contributor
Author
There was a problem hiding this comment.
Changed this and the corresponding tests
| enumeratorContext.triggerAllActions(); | ||
| Assert.assertEquals(0, enumerator.snapshotState(2).pendingSplits().size()); | ||
|
|
||
| // Trigger the planning again to recover from the failure, and we get the expected splits |
Contributor
There was a problem hiding this comment.
nit: avoid we in the comment. maybe like Second scan planning should succeed and discover the expected splits?
Contributor
Author
There was a problem hiding this comment.
Done.
Found another we and fixed it as well
| // Can not discover the new split with planning failures | ||
| for (int i = 0; i < expectedFailures; ++i) { | ||
| enumeratorContext.triggerAllActions(); | ||
| pendingSplits = enumerator.snapshotState(2).pendingSplits(); |
Contributor
There was a problem hiding this comment.
nit: change the checkpoint id to i.
Contributor
Author
There was a problem hiding this comment.
Checked the checkpoint ids and changed them - not really relevant for the tests, but still a valid point
| } | ||
|
|
||
| @Test | ||
| public void testOriginalRetry() throws Exception { |
Contributor
There was a problem hiding this comment.
nit: this method name is not informative. what does original refer to? also how is this test different with testTransientPlanningFailure?
Contributor
Author
There was a problem hiding this comment.
This is testing that we are ignoring failures (-1)
stevenzwu
approved these changes
May 17, 2023
Contributor
|
thanks @pvary for the contribution and @hililiwei for the review |
pvary
pushed a commit
to pvary/iceberg
that referenced
this pull request
May 17, 2023
…res in IcebergSource before failing the job (apache#7571) to 1.16 and 1.15
Contributor
Author
|
Thanks @hililiwei and @stevenzwu for the throughout review and the merge! |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
We found that if the current implementation of the IcebergSource faces some downstream error then it silently retries the planning again and again until the error persists. The only effect on the job is that there is no new record emitted - but no other alarms are raised.
This is similar how the Flink FileSource works but this is confusing for the users.
Also other sources might implement error handling differently, for example Kafka which fails immediately on an error. This is also not desirable for our user base because we expect more resiliency for our jobs.
Based on our discussion with @zhen-wu2 and @gyula-fora, I have created this PR which adds the possibility to retry the failed planning and to configure the number of retries in a few different ways:
IcebergSource.Builder.planRetryNum- if the source is created from java codeconnector.iceberg.plan-retry-numplan-retry-numThe default value is 3, which means that if the 4th planning is failed then the Flink job is failed.
If the original behaviour is needed, then the value
-1should be used.