[
https://issues.apache.org/jira/browse/KAFKA-20821?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Viktor Somogyi-Vass updated KAFKA-20821:
----------------------------------------
Description:
DeleteRecordsCommand iterates over the per-partition results of
{{Admin.deleteRecords()}} and prints an error for any partition whose future
fails, but it does not propagate that failure. The command still returns exit
code 0, so operators and scripts that check the exit status are told the
operation succeeded even though one or more partitions were not deleted.
*Steps to reproduce:*
# Run kafka-delete-records.sh with an offset JSON file that includes a
non-existent topic-partition (or one that otherwise errors).
# Observe the per-partition error: line in stdout.
# Check {{$?}} (it should be 0)
*Expected:* The command should exit non-zero when any partition fails, so
failures are detectable programmatically.
*Fix:* Collect the failed partitions while printing their errors and, if any
occurred, throw AdminCommandFailedException (consistent with the tool's
existing error handling). mainNoExit already maps a thrown exception to exit
code 1.
was:
DeleteRecordsCommand iterates over the per-partition results of
`Admin.deleteRecords()` and prints an error for any partition whose future
fails, but it does not propagate that failure. The command still returns exit
code 0, so operators and scripts that check the exit status are told the
operation succeeded even though one or more partitions were not deleted.
*Steps to reproduce:*
# Run kafka-delete-records.sh with an offset JSON file that includes a
non-existent topic-partition (or one that otherwise errors).
# Observe the per-partition error: line in stdout.
# Check `$?` (it should be 0)
*Expected:* The command should exit non-zero when any partition fails, so
failures are detectable programmatically.
*Fix:* Collect the failed partitions while printing their errors and, if any
occurred, throw AdminCommandFailedException (consistent with the tool's
existing error handling). mainNoExit already maps a thrown exception to exit
code 1.
> kafka-delete-records.sh exits 0 even when some partitions fail to delete
> ------------------------------------------------------------------------
>
> Key: KAFKA-20821
> URL: https://issues.apache.org/jira/browse/KAFKA-20821
> Project: Kafka
> Issue Type: Bug
> Components: tools
> Affects Versions: 4.4.0
> Reporter: Viktor Somogyi-Vass
> Priority: Minor
>
> DeleteRecordsCommand iterates over the per-partition results of
> {{Admin.deleteRecords()}} and prints an error for any partition whose future
> fails, but it does not propagate that failure. The command still returns exit
> code 0, so operators and scripts that check the exit status are told the
> operation succeeded even though one or more partitions were not deleted.
> *Steps to reproduce:*
> # Run kafka-delete-records.sh with an offset JSON file that includes a
> non-existent topic-partition (or one that otherwise errors).
> # Observe the per-partition error: line in stdout.
> # Check {{$?}} (it should be 0)
> *Expected:* The command should exit non-zero when any partition fails, so
> failures are detectable programmatically.
> *Fix:* Collect the failed partitions while printing their errors and, if any
> occurred, throw AdminCommandFailedException (consistent with the tool's
> existing error handling). mainNoExit already maps a thrown exception to exit
> code 1.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
