8000 bug: datagen produces a seq twice when applying split change mutation · Issue #8677 · risingwavelabs/risingwave · GitHub
[go: up one dir, main page]
More Web Proxy on the site http://driver.im/
Skip to content
bug: datagen produces a seq twice when applying split change mutation #8677
Closed
@tabVersion

Description

@tabVersion
          it is strange that partition `3-0` produces data `11, 14, 17, 20` twice. 
2023-03-21T07:40:32.439464Z  INFO test_split_change_mutation: risingwave_stream::executor::source::source_executor: start with state actor_id=0 state=Some([Datagen(DatagenSplit { split_index: 0, split_num: 3, start_offset: None })])
2023-03-21T07:40:32.443087Z  INFO test_split_change_mutation: risingwave_stream::executor::source::source_executor: apply split change actor_id=0 state=[Datagen(DatagenSplit { split_index: 0, split_num: 3, start_offset: Some(3) }), Datagen(DatagenSplit { split_index: 1, split_num: 3, start_offset: None }), Datagen(DatagenSplit { split_index: 2, split_num: 3, start_offset: None })]
2023-03-21T07:40:32.443139Z  INFO test_split_change_mutation: risingwave_stream::executor::source::source_executor: actor 0 apply source split change to [Datagen(DatagenSplit { split_index: 0, split_num: 3, start_offset: Some(3) }), Datagen(DatagenSplit { split_index: 1, split_num: 3, start_offset: None }), Datagen(DatagenSplit { split_index: 2, split_num: 3, start_offset: None })]
2023-03-21T07:40:32.443266Z DEBUG test_split_change_mutation: risingwave_stream::executor::source::source_executor: take snapshot actor_id=0 state=[Datagen(DatagenSplit { split_index: 0, split_num: 3, start_offset: Some(3) }), Datagen(DatagenSplit { split_index: 2, split_num: 3, start_offset: None }), Datagen(DatagenSplit { split_index: 1, split_num: 3, start_offset: None })]
2023-03-21T07:40:32.443556Z TRACE test_split_change_mutation: risingwave_stream::common::table::state_table: commit state table table_id=9011 epoch=1
2023-03-21T07:40:32.443911Z TRACE test_split_change_mutation: risingwave_stream::common::table::state_table: commit state table table_id=9011 epoch=1

logs show that 3-0's offset is correctly recorded and need to find out why it produces the seq twice.

Originally posted by @tabVersion in #8674 (comment)

Metadata

Metadata

Assignees

Labels

No labels
No labels

Type

No type

Projects

No projects

Milestone

Relationships

None yet

Development

No branches or pull requests

Issue actions

    0