[
https://issues.apache.org/jira/browse/HIVE-29037?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
chiranjeevi updated HIVE-29037:
-------------------------------
Description:
The HiveSplitGenerator class incorrectly calculates elapsed time for split
serialization operations, resulting in negative time values in the logs. This
makes debugging and performance analysis difficult.
The calculation fileWriteStarted - Time.monotonicNow() is reversed. Since
Time.monotonicNow() returns the current time, which is always greater than
fileWriteStarted (the time when the operation started), this calculation always
produces a negative number.
Example log
2025-06-21 16:43:28,637 [DEBUG] [HiveSplitGenerator.SplitSerializer Thread -
#2|#2] |tez.HiveSplitGenerator|: Split #41 event to output path:
hdfs://clustername/tmp/hive/staging/.tez/application_1749793988888_0095/events/hive_20250621164323_c39588f6-0ae9-48ad-8db5-5e43cf4a1b1d/0_catalog_sales_InputDataInformationEvent_41
written in *-320 ms*
was:
The HiveSplitGenerator class incorrectly calculates elapsed time for split
serialization operations, resulting in negative time values in the logs. This
makes debugging and performance analysis difficult.
The calculation fileWriteStarted - Time.monotonicNow() is reversed. Since
Time.monotonicNow() returns the current time, which is always greater than
fileWriteStarted (the time when the operation started), this calculation always
produces a negative number.
```javascript
2025-06-21 16:43:28,637 [DEBUG] [HiveSplitGenerator.SplitSerializer Thread -
#2] |tez.HiveSplitGenerator|: Split #41 event to output path:
hdfs://clustername/tmp/hive/staging/.tez/application_1749793988888_0095/events/hive_20250621164323_c39588f6-0ae9-48ad-8db5-5e43cf4a1b1d/0_catalog_sales_InputDataInformationEvent_41
written in *-320 ms*
```
###
> Incorrect time calculation in HiveSplitGenerator causing negative elapsed
> time in logs
> --------------------------------------------------------------------------------------
>
> Key: HIVE-29037
> URL: https://issues.apache.org/jira/browse/HIVE-29037
> Project: Hive
> Issue Type: Bug
> Components: Hive
> Reporter: chiranjeevi
> Priority: Minor
>
> The HiveSplitGenerator class incorrectly calculates elapsed time for split
> serialization operations, resulting in negative time values in the logs. This
> makes debugging and performance analysis difficult.
> The calculation fileWriteStarted - Time.monotonicNow() is reversed. Since
> Time.monotonicNow() returns the current time, which is always greater than
> fileWriteStarted (the time when the operation started), this calculation
> always produces a negative number.
> Example log
> 2025-06-21 16:43:28,637 [DEBUG] [HiveSplitGenerator.SplitSerializer Thread -
> #2|#2] |tez.HiveSplitGenerator|: Split #41 event to output path:
> hdfs://clustername/tmp/hive/staging/.tez/application_1749793988888_0095/events/hive_20250621164323_c39588f6-0ae9-48ad-8db5-5e43cf4a1b1d/0_catalog_sales_InputDataInformationEvent_41
> written in *-320 ms*
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
