Werk #19250: Oracle scheduler jobs: handle job class SCHED$_LOG_ON_ERRORS_CLASS
| Component | Checks & agents | ||||||||
| Title | Oracle scheduler jobs: handle job class SCHED$_LOG_ON_ERRORS_CLASS | ||||||||
| Date | Jun 23, 2026 | ||||||||
| Level | Trivial Change | ||||||||
| Class | Bug Fix | ||||||||
| Compatibility | Incompatible - Manual interaction might be required | ||||||||
| Checkmk versions & editions |
|
Previously, Oracle scheduler jobs belonging to the Oracle-supplied job class
SCHED$_LOG_ON_ERRORS_CLASS could stay CRIT forever once they had failed.
Jobs in this job class only write a log entry when a run fails — successful
runs are never logged, and there is no last run state for them. Because the
monitoring derives the last run status from the job log, the last logged
(failed) run remained the reported status indefinitely, even after the job had
run successfully many times since. The ORA <SID> Job service therefore
stayed CRIT and never recovered on its own.
The Linux/UNIX agent plugin mk_oracle now detects this case. For jobs in the
SCHED$_LOG_ON_ERRORS_CLASS job class, it compares the job's last start date
with the start date of the last logged run. If the job has started again after
its last logged failure, the stale error is no longer reported and the last run
status is left empty.
An empty last run status is handled by the State in case of Job has no log
information option of the Oracle scheduler job rule set (default: WARN).
You can set this to OK there if you do not want such jobs to be flagged at
all.
This werk is marked incompatible because the empty last run status still
alerts (default: WARN), and you now have the possibility to use the State in
case of Job has no log information option of the Oracle scheduler job rule
set to control the resulting state.
Note: this change affects the Linux/UNIX agent plugin only. The Windows plugin is unchanged.