Launcher freezes when launching uxrst ctl command to relaunch Jobs
Article ID: 277428
Updated On:
Products
Issue/Introduction
In a context of a DUAS Node where there's a job that is auto-relaunched on the trailer uproc with a command "uxrst ctl", after some time the Launcher freezes and remains stuck with a wake up date in the past.
When checking the processes at system level, we can see a uxjobend process using a lot of cpu that provokes a deadlock situation with the Launcher due to a lock in the u_fmsb60 file.
In universe.log this can only be seen by this warning message during the online reorg:
| 2023-09-23 13:10:03 |WARN |X|IO |pid=p.t| o_lock_engines_during_pur | Lock timeout occurred (10 min.).
| 2023-09-23 13:10:03 |WARN |X|IO |pid=p.t| o_lock_engines_during_pur | Module Launcher still busy.
| 2023-09-23 13:10:03 |ERROR|X|IO |pid=p.t| o_io_on_purge | o_purge_job_logs not executed (lock timeout)
Environment
DUAS 6.10.x or 7.x
Cause
Defect
Resolution
Update to a fix version listed below or a newer version if available.
Fix version(s):
Component: Application Server (Node)
Dollar Universe 7.00.21 - Available
Will be fixed on Dollar Universe 6.10.121 - planned release March 2024
Additional Information
Defect ID: DU_AS-7441
Public Short Description: Launcher freeze may happen when using the autorelaunch feature.
Public Description: Issue was related to the communication between uxjobend and launcher engine, and only when using auto relaunch (or using command line to relaunch the same task/session/uproc)
Feedback
