[
https://issues.apache.org/jira/browse/FLINK-36090?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
ASF GitHub Bot updated FLINK-36090:
-----------------------------------
Labels: pull-request-available (was: )
> Bug with IngestDB restore operation for priority queue state in backend
> -----------------------------------------------------------------------
>
> Key: FLINK-36090
> URL: https://issues.apache.org/jira/browse/FLINK-36090
> Project: Flink
> Issue Type: Bug
> Components: Runtime / State Backends
> Affects Versions: 2.0.0
> Reporter: Maxim Vershinin
> Priority: Major
> Labels: pull-request-available
>
> *Summary:* Incorrect handling of priority queue states in IngestDB during
> restoring due to missing {{equals()}} and {{hashCode()}} methods in
> {{{}RegisteredPriorityQueueStateBackendMetaInfo{}}}.
> *Problem Description:*
> During the restoration of IngestDB in my Flink project, an issue was
> identified where the priority queue states are not managed correctly in the
> backend. The problem stems from the absence of {{equals()}} and
> {{hashCode()}} methods in the {{RegisteredPriorityQueueStateBackendMetaInfo}}
> class.
> In particular, within the {{exportColumnFamiliesWithSstDataInKeyGroupsRange}}
> method of the {{RocksDBIncrementalRestoreOperation}} class, if the state is a
> priority queue, identical states from different subtasks are erroneously
> treated as distinct states within the {{exportedColumnFamiliesOut}} map. This
> leads to inconsistent behavior and errors during the restoration process.
> *Proposed Solution:*
> To address this issue, add {{equals()}} and {{hashCode()}} methods to the
> {{RegisteredPriorityQueueStateBackendMetaInfo}} class. Implementing these
> methods will ensure that priority queue states are accurately recognized and
> handled across different subtasks, thereby preventing errors during IngestDB
> restoration.
>
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
