Applications Manager collects Oracle metrics in groups of related queries, each limited by the monitor's Query Execution Timeout and Socket Read Timeout. When a group exceeds either limit, its attributes are not updated for that poll and the monitor reports the timeout.
You will see one of these messages, where <group> is the query group that exceeded the limit, such as INSTANCESTATUS, TABLESPACESTATUS or OracleAlertLogList.
| Message | Underlying error |
|---|---|
Query execution timed out due to this query group(s): <group> | ORA-01013: user requested cancel of current operationApplications Manager cancelled the statement, not a user. |
Socket read timed out due to this query group(s): <group> | IO Error: Socket read timed outThe database sent nothing back within the socket read limit. |
The failing statement and the seconds it ran are recorded in <AppManager Home>/logs/OracleMonitor/oraclemonitorOut.txt. Search for DBSQLTIMEOUT or DBSOCKETREADTIMEOUT to identify the SQL statement that timed out and review its execution duration. Perform this analysis before modifying any timeout settings to determine whether the configured timeout is insufficient or the query itself is taking longer than expected.
Both values are set per monitor, from the Oracle monitor's details page:
Monitor Actions → Edit Monitor → Socket Read Timeout
Monitor Actions → Edit Monitor → Query Execution Timeout
Increase the seconds configured in the Timeout field named by the alert. The defaults are 60 seconds for Query Execution Timeout and 300 seconds for Socket Read Timeout; use the elapsed seconds from the log as your guide rather than an arbitrary large value.
If the alert returns after raising the timeout, the query itself is the problem. Continue with sections 3 and 4.
These settings are inside Applications Manager. They take effect without any change to your Oracle database and you can reverse them at any time, so try them before touching the database.
| Setting | Where | Effect |
|---|---|---|
| Oracle Alert Log TableSpace Statistics | Settings → Performance Polling → Optimize Data Collection → Oracle | Stops collecting the metric entirely. Use when the data is not needed; the corresponding tab stops updating. INSTANCESTATUS has no equivalent switch. |
am.oracle.alertlog.upperlimit | Settings → Server Settings | Caps the alert-log rows fetched per poll. Default 1000. |
am.oracle.ignorealerts | Settings → Server Settings | Semicolon-separated text to exclude, so noisy expected ORA errors are never fetched. Empty by default. |
The server settings am.oracle.querytimeout.alert.enable and am.oracle.socketreadtimeout.alert.enable, with their oraclepdb equivalents, hide the message without fixing the timeout. Use them only while you work through the steps below.
This group reads Oracle's performance and dictionary views. On a busy instance the queries over V$SQLAREA and DBA_INDEXES are usually the slow ones, and stale optimizer statistics are a common cause. Check with your Oracle DBA whether the statistics are current.
SQL> EXEC DBMS_STATS.GATHER_FIXED_OBJECTS_STATS;
SQL> EXEC DBMS_STATS.GATHER_DICTIONARY_STATS;
Gather fixed-object statistics while the database is under a representative workload. Both calls consume resources and affect dictionary query plans across the database, so agree the timing with your Oracle DBA. Compare the query's execution time before and after — this is not a guaranteed fix.
The recycle bin is not involved in this query group. Purging it will not change an INSTANCESTATUS timeout.
To report space held by dropped objects, the tablespace queries read DBA_RECYCLEBIN on every poll, so a large recycle bin slows this group directly. Check the count, and time the statement itself:
SQL> SELECT COUNT(*) FROM dba_recyclebin;
If the count is high, purge it, then gather dictionary statistics as above:
SQL> PURGE dba_recyclebin;
PURGE DBA_RECYCLEBIN privilege, and those objects can no longer be recovered with FLASHBACK TABLE ... TO BEFORE DROP. Check with your Oracle DBA that nothing in the recycle bin is still needed before you run it.In a multitenant database each container has its own recycle bin, so repeat for every monitored PDB:
SQL> ALTER SESSION SET CONTAINER = <pdb_name>;
SQL> PURGE dba_recyclebin;
This group reads the alert log through Oracle's ADR diagnostic views, which expose the XML alert files in the Automatic Diagnostic Repository (ADR). Collection is incremental — only records newer than the last one collected — but these views are not indexed, so Oracle still works through the accumulated history on every poll. The more alert files there are, the slower it gets.
Find the alert directory from the database, then count the files in it:
SQL> SELECT name, value FROM v$diag_info WHERE name IN ('ADR Base','ADR Home','Diag Alert');
Linux: find "<Diag Alert>" -maxdepth 1 -type f | wc -l
du -sh "<Diag Alert>"
Windows: dir "<Diag Alert>" /b /a-d | find /c /v ""Use the Diag Alert path exactly as the query returns it; it is the directory holding the XML alert files for that instance.
Thousands of fragmented XML files, common on databases running for years without regular purging, is the condition to look for. With your Oracle DBA, purge the old alert data and adjust the retention policy so the backlog does not simply rebuild:
adrci> show homes
adrci> set homepath diag/rdbms/<db_name>/<instance_name>
adrci> show control
adrci> purge -age 43200 -type ALERT
adrci> set control (LONGP_POLICY = 720)
purge -age takes minutes, so 43,200 is 30 days. LONGP_POLICY takes hours, so 30 days is 720. Do not use the same number for both.LONGP_POLICY sets how long ADR keeps alert data, and defaults to one year. Both commands act on the ADR home currently selected, so set the homepath first; each Oracle RAC instance has its own. Agree both values with your Oracle DBA, and skip the purge if you need the history for auditing.
Afterwards, run the alert-log statement from oraclemonitorOut.txt against the database with SET TIMING ON, and compare the elapsed time with the figure logged before the cleanup and with the monitor's Query Execution Timeout.
If the group is not one of the three above, or the query is still slow in SQL*Plus after the steps here, write to appmanager-support@manageengine.com with: