dbi Blog

Attention.log is a feature, which was introduced in Oracle Database 21c, designed to capture high-level summaries of significant database events and activities. It differs from the alert.log in following points:

High-Level Summaries: The attention.log focuses on summarizing critical and significant events rather than logging every minor detail, including database startups and shutdowns, major configuration changes, errors or warnings that need immediate attention.

Consolidation of Critical Events: It provides a consolidated view of the most important events, making it easier for database administrators to quickly review and identify critical issues without rummaging through through detailed logs.

Accessibility: Designed to be easily readable and quickly accessible for a high-level overview of the database’s health and significant activities.

Complementary to Alert Log: While the attention.log highlights major events, it complements the alert.log rather than replacing it. Database administrators can use the attention.log for a quick overview and the alert.log for detailed diagnostics.

Location: Like the alert.log, the attention.log is also found in the DIAGNOSTIC_DEST directory, usually under $ORACLE_BASE/diag/rdbms/<db_name>/<instance_name>/log.

It can be very helpful for not so experienced database administrators or to get a quick overview in difficult or unexpected cases, as I had on a productive environment some time ago: an internal error had occurred, the database crashed and the alert.log was far too large to read without splitting it up which of course makes troubleshooting unnecessarily difficult (under time pressure).

How to get information about the attention.log:

The location of the attention log can be found by querying the V$DIAG_INFO view, it is in the same directory as the alert. log, (since Oracle 11g: $OH/diag/rdbms… )

select name, value
from   v$diag_info
where  name = 'Attention Log';

NAME                      VALUE
--------------------- -------------------------------------------------------------
Attention Log         /u01/app/oracle/diag/rdbms/cdb1/cdb1/trace/attention_cdb1.log

The oracle documentation proposes to query the V$DIAG_ALERT_EXT view to get relevant attention.log information, but it is a view over the XML-based alert.log (in the Automatic Diagnostic Repository for the current container), not the attention log! But nevertheless we can get very useful information out of it, divided into the same categories as in the attention.log:

--message_type 2=INCIDENT_ERROR, message_type 3=ERROR
SELECT message_type, message_level, message_text
FROM V$DIAG_ALERT_EXT 
WHERE message_type in (2, 3);

MESSAGE_TYPE MESSAGE_LEVEL MESSAGE_TEXT
------------ ------------- ---------------------------------------------------------
           3    4294967295 PMON (ospid: 3565): terminating the instance due to ORA error 471

Querying V$DIAG_ALERT_EXT the most important labels are:

MESSAGE_LEVEL:

1: CRITICAL: critical errors

2: SEVERE: severe errors

8: IMPORTANT: important message

16: NORMAL: normal message

MESSAGE_TYPE:

1: UNKNOWN: essentially the NULL type

2: INCIDENT_ERROR: the program has encountered an error for some internal or unexpected reason, and it must be reported to Oracle Support

3: ERROR: an error of some kind has occurred

4: WARNING: an action occurred or a condition was discovered that should be reviewed and may require action

5: NOTIFICATION: reports a normal action or event, this could be a user action such as “logon completed”

6: TRACE: output of a diagnostic trace

Opening the attention.log with vi

vi /u01/app/oracle/diag/rdbms/cdb1/cdb1/trace/attention_cdb1.log

will give you an output like this (JSON formatted), which is obviously pretty comfortable to read:

{
IMMEDIATE : "PMON (ospid: 3565): terminating the instance due to ORA error 471" 
CAUSE: "PMON detected fatal background process death"
ACTION: "Termination of fatal background is not recommended, Investigate cause of process termination"
CLASS : CDB-INSTANCE / CDB_ADMIN / ERROR / DBAL-35782660
TIME : 2024-07-10T14:15:16.159-07:00
INFO : "Some additional data on error PMON error"
}

It is possible to convert the output into plain text format:

jq -r '.tags[].name' input.json > output.txt

which gives us a formatted expression that might look like this:

2024-07-01T10:15:32.123456+00:00
[SEVERE] ORA-00600: internal error code, arguments: [1234], [], [], [], [], [], [], [], [], [], []
Action: Please contact Oracle Support Services.

2024-07-01T11:20:45.789012+00:00
[CRITICAL] ORA-01578: ORACLE data block corrupted (file # 23, block # 220734)
Action: This error signifies a corrupted data block. The data block has been marked as corrupt. Consider restoring from backup.

2024-07-02T08:42:27.654321+00:00
[ALERT] Database instance crashed due to unexpected termination.
Action: Investigate the cause of the instance termination. Review related logs and diagnostic information.

2024-07-03T12:34:56.987654+00:00
[INFO] System global area (SGA) resized. New size: 68GB.
Action: No immediate action required. Monitor performance and stability.

Key Points

Severity Levels: Entries are tagged with severity levels such as [SEVERE], [CRITICAL], [IMPORTANT] or [NORMAL] to highlight their importance.

Timestamp: Each entry begins with a timestamp in ISO 8601 format.

Messages and Actions: Each entry provides a brief description of the event and recommended actions.

Benefits for DBAs

Quick Identification: The attention.log helps DBAs quickly identify and respond to critical issues without sifting through the more detailed alert.log.

Conciseness: It captures only the most significant events, reducing noise and making it easier to focus on urgent matters.

Complementary to alert.log: It complements the alert.log by summarizing critical events, while the alert.log continues to provide detailed information for troubleshooting.

Overall, the attention.log is a useful addition for DBAs, enabling more efficient monitoring and quicker responses to significant database events.