You are viewing a plain text version of this content. The canonical link for it is here.
Posted to commits@hudi.apache.org by "Vinoth Chandar (Jira)" <ji...@apache.org> on 2019/10/07 16:18:00 UTC
[jira] [Commented] (HUDI-292) Consume more entries from kafka than
specified sourceLimit.
[ https://issues.apache.org/jira/browse/HUDI-292?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16945994#comment-16945994 ]
Vinoth Chandar commented on HUDI-292:
-------------------------------------
is it because `long toOffset = Math.min(toOffsetMax, range.untilOffset() + eventsPerPartition);` does not honor the `remainingEvents`?
Good catch. but wondering how severe can the actually effect be since we bound by eventsPerPartition anywya?
> Consume more entries from kafka than specified sourceLimit.
> -----------------------------------------------------------
>
> Key: HUDI-292
> URL: https://issues.apache.org/jira/browse/HUDI-292
> Project: Apache Hudi (incubating)
> Issue Type: Improvement
> Components: Utilities
> Reporter: leesf
> Assignee: leesf
> Priority: Major
> Fix For: 0.5.1
>
>
> When _CheckpointUtils#computeOffsetRanges_ for consuming kafka messges.
> Given
> topic = "test",
> fromOffsets(partition -> offset pair) = (0 -> 0), (1 -> 0), (2 -> 0), (3 -> 0), (4 -> 0),
> toOffsets = (0, 100), (1, 1000), (2, 1000), (3, 1000), (4, 1000),
> numEvents = 1001.
> The output of _CheckpointUtils#computesOffsetRanges_ is
> OffsetRange(topic: 'test', partition: 0, range: [0 -> 100])
> OffsetRange(topic: 'test', partition: 1, range: [0 -> 226])
> OffsetRange(topic: 'test', partition: 2, range: [0 -> 226])
> OffsetRange(topic: 'test', partition: 3, range: [0 -> 226])
> OffsetRange(topic: 'test', partition: 4, range: [0 -> 226])
> Total count is 1004(100 + 266 * 4), more than 1001, and thus consume more entries from kafka than specified 1001.
> CC [~vinoth]
--
This message was sent by Atlassian Jira
(v8.3.4#803005)