Agent in mode Batch generates core files for ucxjlx6m with "Segmentation fault"

book

Article ID: 223920

calendar_today

Updated On:

Products

CA Automic One Automation

Issue/Introduction

In some Unix Agents, after having changed the start_Type from fork to batch, some Jobs ( only those launched with a system account other than root) generate core files for the Job Messenger (ucxjlx6m) process.

Despite the "crash" of the Job Messenger, the Jobs are correctly executed and complete with Ended_OK.

Example of the error in the Report when launching a Job that is launched as Oracle that Starts the Instance Oracle:

***************************************************************************
**  ucxjlx6m     version 12.2.3+build.1558154951870 changelist 1558082178  **
**  JOB 0200008373 (ProcID:0000025888) START AT   12.08.2021 / 10:36:15  **
**                                     UTC TIME   12.08.2021 / 08:36:15  **
**  TEXT="        Job started             "                              **
***************************************************************************
Segmentation fault (core dumped) 
#############################################
JOB LOG
---------------------------------------------------
[OK] JU.COM.ORA.START.INSTANCE - Startup Instance ORACLE FINISHED
---------------------------------------------------
***************************************************************************
**  ucxjlx6m     version 12.2.3+build.1558154951870 changelist 1558082178  **
**  JOB 0200008373 (ProcID:0000025888) ENDED AT   12.08.2021 / 10:36:31  **
**                                     UTC TIME   12.08.2021 / 08:36:31  **
**  TEXT="        Job ended               "       RETCODE=00             **
***************************************************************************
Segmentation fault (core dumped) 

 

If we go to the Oracle folder, we can find the core files:

[[email protected]~]# ls -lrt /opt/oracle/core*
-rw------- 1 oracle dba 667648 Aug 12 10:35 /opt/oracle/core.25562
-rw------- 1 oracle dba 663552 Aug 12 10:36 /opt/oracle/core.25889
-rw------- 1 oracle dba 667648 Aug 12 10:43 /opt/oracle/core.29771
-rw------- 1 oracle dba 667648 Aug 12 10:44 /opt/oracle/core.30153
-rw------- 1 oracle dba 667648 Aug 12 10:53 /opt/oracle/core.2021
-rw------- 1 oracle dba 667648 Aug 12 11:24 /opt/oracle/core.10003

 

When checking the core, we can see it was generated by the ucxjlx6m process:

[[email protected] oracle]# strings core.10003
CORE
CORE
ucxjlx6m
/opt/oneautomation/current/agent/bin/ucxjlx6m IPA=X.X.X.X PNR=12050 MNR=19
IGISCORE
CORE
ELIFCORE

 

Cause

System configuration issue, the ulimit -s (Stack size) parameter had been set to a huge value of 4GBytes  (4194304)  for the impacted System Accounts instead of the default 8192 Kbytes

Environment

Release : 12.x

Component : AUTOMATION ENGINE

Sub-Component: Agent Unix/Linux

Resolution

Reduce ulimit -s for all users from the huge value 4194304  to 39936  (default is 8192 for Linux).

This can be done either in /etc/security/limits.conf or the .profile of the users.

Additional Information

We do not recommend using the batch mode for Agents as it's by far more limited than the fork method.